1 00:00:07,200 --> 00:00:13,520 >> I WOULD LIKE TO WELCOME 2 00:00:13,520 --> 00:00:16,040 EVERYONE FOR TODAY'S 3 00:00:16,040 --> 00:00:18,640 MORBIDITY WINNER'S WEBINAR, 4 00:00:18,640 --> 00:00:30,120 ALSO A MOUTHFUL NICHD--ENHANCE THE LIVES 5 00:00:30,120 --> 00:00:32,760 OF CHILDREN AND ADOLESCENCE AND OPTIMIZE 6 00:00:32,760 --> 00:00:35,040 ABILITIES FOR ALL. 7 00:00:35,040 --> 00:00:36,120 MATERNAL HEALTH AND A MAJOR RESEARCH 8 00:00:36,120 --> 00:00:38,840 FOLK ISOTOPE FOR OUR INSTITUTE WITH 9 00:00:38,840 --> 00:00:41,840 PARTICULAR EFFORTS FOCUSED ON REDUCING 10 00:00:41,840 --> 00:00:47,120 MATERNAL MORBIDITY AND MORTALITY. 11 00:00:47,120 --> 00:00:49,240 NEARLY A DECADE AGO, NICHD FUNDED THE 12 00:00:49,240 --> 00:00:52,440 NEW MOM TO BE PROJECT. 13 00:00:52,440 --> 00:00:56,640 NEW MOM TO BE STANDED FOR THE 14 00:00:56,640 --> 00:01:02,560 [INDISCERNIBLE] PREGNANCY OUTCOMES STUDY 15 00:01:02,560 --> 00:01:03,640 MONITORING MOTHERS TO BE. 16 00:01:03,640 --> 00:01:08,640 THIS WAS A RACIALLY ETHNICALLY AND 17 00:01:08,640 --> 00:01:10,160 GEOGRAPHICALLY DIVERSE SAMPLE OF 10,000 18 00:01:10,160 --> 00:01:12,800 PEOPLE WHO WERE PREGNANT FOR FIRST TIME. 19 00:01:12,800 --> 00:01:16,240 THIS COHORT WAS VERY RICHLY AND DEEPLY 20 00:01:16,240 --> 00:01:19,480 PHENOTYPES AND IT FOLLOWED THE WOMEN 21 00:01:19,480 --> 00:01:22,680 FROM THEIR SIXTH WEEK OF PREGNANCY 22 00:01:22,680 --> 00:01:24,000 THROUGH THEIR DELIVERIES. 23 00:01:24,000 --> 00:01:25,840 THE PRIMARY OUTCOME WAS TO DETERMINE THE 24 00:01:25,840 --> 00:01:30,480 RISK FOR PRETERM DELIVERY. 25 00:01:30,480 --> 00:01:32,040 BUT WE WONDERED WOULD IT BE POSSIBLE TO 26 00:01:32,040 --> 00:01:34,760 RELIEVE US FROM THE DATA FROM THE STUDY 27 00:01:34,760 --> 00:01:39,040 TO REANALYZE IT FOR RISK FACTORS FOR 28 00:01:39,040 --> 00:01:42,280 SEVERE MATERNAL MORBIDITY AND MORTALITY. 29 00:01:42,280 --> 00:01:46,880 WE DECIDED TO OFFER THIS AS NICHD'S 30 00:01:46,880 --> 00:01:49,600 FIRST EVER DATA CHALLENGE WITH THE GOAL 31 00:01:49,600 --> 00:01:52,120 OF DEVELOPING INNOVATIVE COMPUTATIONAL 32 00:01:52,120 --> 00:01:57,840 METHODS TO ANALYZE THE NEW MOM TO BE 33 00:01:57,840 --> 00:02:00,280 DATA TO DETERMINE RISK FACTORS FOR MA 34 00:02:00,280 --> 00:02:01,080 TERBAL MORBIDITY. 35 00:02:01,080 --> 00:02:03,240 WE RECEIVED A ROBUST RESPONSE WITH 36 00:02:03,240 --> 00:02:07,840 THREIVE SUBMISSIONS AND WE WERE THRILLED 37 00:02:07,840 --> 00:02:10,320 BY THE RANGE OF APPROACHED USED BY THE 38 00:02:10,320 --> 00:02:12,840 WINNING TEAMS, I'M LOOKING FORWARD TO 39 00:02:12,840 --> 00:02:14,480 HEARING ABOUT THEM TODAY. 40 00:02:14,480 --> 00:02:16,800 SOME FOCUSED ON UNIQUE APPROACHES AND 41 00:02:16,800 --> 00:02:18,240 METHODOLOGIES WHILE OTHERS ANALYZE THE 42 00:02:18,240 --> 00:02:19,840 DATA COLLECTION METHODS OF THE NEW MOM 43 00:02:19,840 --> 00:02:22,440 TO BE DATA SET. 44 00:02:22,440 --> 00:02:25,160 OTHER GROUPS USED INNOVATIVE STATISTICAL 45 00:02:25,160 --> 00:02:33,720 EQUATIONS TO FOSTER NEW PREDICTION 46 00:02:33,720 --> 00:02:36,840 MODELINGS FOR CO-MORBIDITIES, SPECIFIC 47 00:02:36,840 --> 00:02:39,960 TO MATERNAL MORBIDITY. 48 00:02:39,960 --> 00:02:44,080 ULTIMATELY WE AWARDED PRIZES 7 PRIZES 49 00:02:44,080 --> 00:02:49,320 FOR FOR INNOVATION AND 5 PRIZES FOR 50 00:02:49,320 --> 00:02:50,120 ADDRESSING HEALTH DISPARITIES. 51 00:02:50,120 --> 00:02:52,480 I WOULD LIKE TO TAKE THIS OPPORTUNITY TO 52 00:02:52,480 --> 00:02:54,880 THANK OUR STAFF AT NICHD AND THE STAFF 53 00:02:54,880 --> 00:02:57,560 AT THE NIH, OFFICE OF DIRECTOR AND THE 54 00:02:57,560 --> 00:03:03,440 NASA CENTERS FOR EXCELLENCE FOR 55 00:03:03,440 --> 00:03:05,360 COLLABORATIVE INNOVATION WHO HELPED 56 00:03:05,360 --> 00:03:08,000 ORGANIZE THIS CHALLENGE, I WOULD ALSO 57 00:03:08,000 --> 00:03:11,120 LIKE TO THANK FREELANCERS OUR CONTRACT 58 00:03:11,120 --> 00:03:13,320 PARTNER WHO HELPED US IMPLEMENT THIS 59 00:03:13,320 --> 00:03:13,880 CHALLENGE. 60 00:03:13,880 --> 00:03:16,200 I WOULD LIKE TO EXPRESS OUR APPRECIATION 61 00:03:16,200 --> 00:03:19,640 TO MANY JUDGES INCLUDING THOSE WHO SERVE 62 00:03:19,640 --> 00:03:21,520 NIH, THE MATERNAL CHILD HEALTH BUREAU OF 63 00:03:21,520 --> 00:03:25,120 THE HEALTH RESOURCES AND SERVICES 64 00:03:25,120 --> 00:03:27,040 ADMINISTRATION AND THE REPRODUCTIVE 65 00:03:27,040 --> 00:03:28,840 HEALTH BRANCH OF THE CENTERS FOR DISEASE 66 00:03:28,840 --> 00:03:33,120 CONTROL AND PREVENTION WHO HELPED US TO 67 00:03:33,120 --> 00:03:34,000 EVALUATE THE APPLICATIONS. 68 00:03:34,000 --> 00:03:35,200 THIS AFTERNOON WE ARE LOOKING FORWARD TO 69 00:03:35,200 --> 00:03:40,040 HEARING FROM OUR WINNING TEAMS TO LEARN 70 00:03:40,040 --> 00:03:41,440 MORE DETAILS ABOUT THEIR DESIGN AND 71 00:03:41,440 --> 00:03:42,920 METHODS, WHAT A LEARNED, AND HOW WE 72 00:03:42,920 --> 00:03:45,960 MIGHT BE ABLE TO USE THESE APPROACHES 73 00:03:45,960 --> 00:03:58,240 FOR FUTURE STUDIES AND,AINAL SIS. 74 00:03:58,240 --> 00:04:00,600 >> I WOULD NOW LIKE TO TURN IT OVER TO 75 00:04:00,600 --> 00:04:02,160 DR. DAVIS TO INTRODUCE OUR PRESENTERS. 76 00:04:02,160 --> 00:04:09,000 >> JUST A FOOT NOTE TO DR. BIANCHI'S 77 00:04:09,000 --> 00:04:13,120 COMMENTS, 1 OF OUR TEAM HAD A CONFLICT 78 00:04:13,120 --> 00:04:14,440 WITH THEIR PRESENTATION SO WE WILL HAVE 79 00:04:14,440 --> 00:04:15,120 6 TODAY. 80 00:04:15,120 --> 00:04:18,240 THE FIRST TEAM THAT'S GOING TO PRESENT 81 00:04:18,240 --> 00:04:22,320 THIS MORNING, THIS AFTERNOON WILL BE THE 82 00:04:22,320 --> 00:04:26,960 TEAM FROM COLUMBIA UNIVERSITY, 83 00:04:26,960 --> 00:04:27,760 DR. [INDISCERNIBLE] WILL REPRESENT THAT 84 00:04:27,760 --> 00:04:30,080 TEAM AND THE TITLE OF HER PRESENTATION 85 00:04:30,080 --> 00:04:33,480 IS PREDICTING AND UNDERSTANDING 86 00:04:33,480 --> 00:04:46,200 PREELAMPSIA, A MACHINE LEARNING 87 00:04:46,200 --> 00:04:46,440 APPROACH. 88 00:04:46,440 --> 00:04:49,440 >> GOOD AFTERNOON, THANK YOU FOR THE 89 00:04:49,440 --> 00:04:51,160 PRESENTATION,IME HERE ON BEHALF OF MY 90 00:04:51,160 --> 00:04:58,840 TEAM AND IT INCLUDES COLLABORATORS FROM 91 00:04:58,840 --> 00:05:03,280 COMPUTER SCIENCE [INDISCERNIBLE] AND 92 00:05:03,280 --> 00:05:08,280 DR. CHEN AND A DOCTOR HO IS AN OBGYN AND 93 00:05:08,280 --> 00:05:12,400 WE ALSO HAVE JOINT WORK WITH ANITA RAJA, 94 00:05:12,400 --> 00:05:16,400 AND HER STUDENTS, ADAM, DANIEL, AND 95 00:05:16,400 --> 00:05:17,040 ALISA. 96 00:05:17,040 --> 00:05:17,960 NEXT SLIDE, PLEASE. 97 00:05:17,960 --> 00:05:21,360 SO IN THIS PRESENTATION, I WILL 98 00:05:21,360 --> 00:05:23,840 INTRODUCE PREECLAMPSIA AND DESCRIBE THE 99 00:05:23,840 --> 00:05:25,480 NICHD CHALLENGE DATA AND THE STUDY 100 00:05:25,480 --> 00:05:27,640 COHORT, I WILL GO OVER OUR PIPELINE TO 101 00:05:27,640 --> 00:05:31,080 BUILD MODELS AND OUR RESULTS IN TERMS OF 102 00:05:31,080 --> 00:05:32,280 PREDICTION, ABILITY AND TOP FEATURES 103 00:05:32,280 --> 00:05:35,240 MPLET THEN WHY TALK ABOUT FAIRNESS CHECK 104 00:05:35,240 --> 00:05:37,520 IN AND MITIGATION IN OUR MODELS. 105 00:05:37,520 --> 00:05:39,600 AND CONCLUDE WITH OUR ONGOING RESEARCH 106 00:05:39,600 --> 00:05:43,680 AND NEXT STEPS. 107 00:05:43,680 --> 00:05:45,280 NEXT SLIDE, PLEASE. 108 00:05:45,280 --> 00:05:48,280 SO PREECLAMPSIA OR PE FOR SHORT IS A 109 00:05:48,280 --> 00:05:50,560 HYPER TENSIVE DISORDER AND 1 OF THE MOST 110 00:05:50,560 --> 00:05:54,960 SERIOUS AND LIFE THREATENING PREGNANCY 111 00:05:54,960 --> 00:05:57,440 OUTCOMES, IT IS CHARACTERIZED BY POOR 112 00:05:57,440 --> 00:06:00,920 PROFUSION IN TISSUE DURING PREGNANCY, PE 113 00:06:00,920 --> 00:06:04,760 RARELY DEVELOP A DANGEROUS COMPLICATION 114 00:06:04,760 --> 00:06:07,000 CALLED ELAMPSIA, A NEW ONSET OF SEIZURES 115 00:06:07,000 --> 00:06:09,680 NOT RELATE TOTD ANY OTHER MEDICAL 116 00:06:09,680 --> 00:06:14,680 REASON, THE TERM ECLAMPSIA ORIGINATES 117 00:06:14,680 --> 00:06:20,160 FROM THE GREEK WORD EKLAMPSIA, FOR 118 00:06:20,160 --> 00:06:22,440 LIGHTNING OR LIGHT BURST. 119 00:06:22,440 --> 00:06:25,600 DESPITE BEING KNOWN FOR CENTURIES 120 00:06:25,600 --> 00:06:27,120 PREECLAMPSIA REMAINS A CHALLENGE FOR 121 00:06:27,120 --> 00:06:28,440 CLINICIANS AND PREGNANT WOMEN. 122 00:06:28,440 --> 00:06:31,040 IN PARTICULAR WE STILL DON'T KNOW THE 123 00:06:31,040 --> 00:06:35,720 ORIGINS OF THE DISEASE OR OF THE AND IS 124 00:06:35,720 --> 00:06:38,320 UNDERLYING TYPES NOW WE KNOW HOW TO 125 00:06:38,320 --> 00:06:41,840 PREDICT, PREVEPT AND TREAT IT. 126 00:06:41,840 --> 00:06:43,080 NEXT, PLEASE. 127 00:06:43,080 --> 00:06:48,040 THE ONSET TEND TO OCCUR AROUND 20 WEEKS 128 00:06:48,040 --> 00:06:51,880 OF GUESTATION TODAY IT IS KNOWN AS A 129 00:06:51,880 --> 00:06:54,720 OBSTETRICAL SYNDROME RATHER THAN SCEEZ 130 00:06:54,720 --> 00:06:59,920 BECAUSE OF ITS COMPLICATIONS, FIRST IS 131 00:06:59,920 --> 00:07:03,640 THOSE RANGING FROM THE NEUROLOGICAL, 132 00:07:03,640 --> 00:07:05,360 RENAL, HEPATIC AND CAUSING SERIOUS 133 00:07:05,360 --> 00:07:11,840 DAMAGE TO BLOOD VESSELS BESIDES MATERNAL 134 00:07:11,840 --> 00:07:14,600 MORBIDITY AND PRENATAL MORBIDITY, THE 135 00:07:14,600 --> 00:07:17,240 RISK ARE ASSOCIATED WITH PREECLAMPSIA. 136 00:07:17,240 --> 00:07:21,840 NOTE THAT MOST DEATHS DUE TO PE OCCUR IN 137 00:07:21,840 --> 00:07:23,840 LOW INCOME COUNTRIES WHERE WOMEN DON'T 138 00:07:23,840 --> 00:07:25,840 HAVE ACCESS TO PROPER HEALTHCARE DURING 139 00:07:25,840 --> 00:07:29,440 PREGNANCY BUT EVEN IN DEVELOPED 140 00:07:29,440 --> 00:07:30,720 COUNTRIES MATERNAL MORTALITY IS VERY LOW 141 00:07:30,720 --> 00:07:34,640 BUT DUE TO PE IT'S VERY LOW. 142 00:07:34,640 --> 00:07:38,640 THERE IS A LARGE BODY OF LITERATURE ON 143 00:07:38,640 --> 00:07:41,320 PE WHICH INCLUDES PRIOR PE, CHRONIC 144 00:07:41,320 --> 00:07:43,400 HYPER TENSION, DIABETES, OBESITY AND 145 00:07:43,400 --> 00:07:52,360 MULTIPLE GUESTATION GUESTATION WHICH ALSO SUPPORTS 146 00:07:52,360 --> 00:07:57,400 DIFFERENT TYPE PREECLAMPSIA. 147 00:07:57,400 --> 00:07:59,640 THE DECODING CHALLENGE WAS DERIVED FROM 148 00:07:59,640 --> 00:08:03,760 THE NULLIP A ROUS PREGNANCY OUTCOMES 149 00:08:03,760 --> 00:08:07,480 STUDY, THIS WAS DONE BETWEEN 150 00:08:07,480 --> 00:08:09,040 OCTOBER 2010 AND MAY 2014. 151 00:08:09,040 --> 00:08:11,000 VARIOUS DATA WAS COLLECTED BY VARIOUS 152 00:08:11,000 --> 00:08:13,360 WOMEN ACROSS 4 TIME POINTS, 3 VISITS 153 00:08:13,360 --> 00:08:15,640 CORRESPONDING ROUGHLY TO THE FIRST, 154 00:08:15,640 --> 00:08:17,320 SECOND AND THIRD TRIMESTER THAT'S VISIT 155 00:08:17,320 --> 00:08:21,920 1, 2, 3, AS WELL AS DELIVERY VISIT. 156 00:08:21,920 --> 00:08:25,640 WE SELECTED AN INCREMENTAL SET OF 157 00:08:25,640 --> 00:08:28,040 FEATURES OR VARIABLES TO REACH 183 158 00:08:28,040 --> 00:08:31,560 FEATURES, AND MANY IN ACCORDANCE WITH 159 00:08:31,560 --> 00:08:33,720 RISK FACTORS, IDENTIFIED IN THE MEDICAL 160 00:08:33,720 --> 00:08:37,400 LITERATURE, ABOUT FOOD NUTRITION, 161 00:08:37,400 --> 00:08:40,680 PHYSIOLOGY, ULTRA SOUND, MEDICAL 162 00:08:40,680 --> 00:08:43,040 HISTORY, SLEEP CONDITION, DEMOGRAPHICS 163 00:08:43,040 --> 00:08:47,720 AND FAMILY MEDICAL HISTORY. 164 00:08:47,720 --> 00:08:48,160 NEXT SLIDE, PLEASE. 165 00:08:48,160 --> 00:08:52,880 SO WE WANT TO BUILD A PREDICTION MODELS, 166 00:08:52,880 --> 00:08:57,560 AND DIFFERENT VISITS FOR ELAMPSIA AND 167 00:08:57,560 --> 00:09:03,080 PREECLAMPSIA, AND DIG INTO OUR MODELS 168 00:09:03,080 --> 00:09:04,800 FOR DRIVING PREDICTION AND THE BEST 169 00:09:04,800 --> 00:09:11,400 THRESHOLD FOR THE RISK OF PE, SO WE 170 00:09:11,400 --> 00:09:20,760 FIRST POSED ABOUT PREVENTIVE DATA, SO 171 00:09:20,760 --> 00:09:22,520 2243 PATIENTS WERE SELECTED TO 172 00:09:22,520 --> 00:09:24,840 PARTICIPATE IN THE STUDY. 173 00:09:24,840 --> 00:09:30,000 SUBJECT WITH CHRONIC HYPER TENSION, MILD 174 00:09:30,000 --> 00:09:31,640 PREELAMPSIA, SUPER IMPOSED PREECLAMPSIA 175 00:09:31,640 --> 00:09:35,800 AND NEW ONSET HYPER TENSION WERE 176 00:09:35,800 --> 00:09:36,040 EXCLUDED. 177 00:09:36,040 --> 00:09:36,400 NEXT, PLEASE. 178 00:09:36,400 --> 00:09:39,440 SO WE WOULD A MACHINE LEARNING PIPELINE 179 00:09:39,440 --> 00:09:42,640 TO CREATE COMPOSITE PRECLICTIVE MODEL 180 00:09:42,640 --> 00:09:44,640 FOR ECLAMPSIA AT STUDY VISITS. 181 00:09:44,640 --> 00:09:47,640 WE CONDUCTED 100 TRIALS, EACH TAKING A 182 00:09:47,640 --> 00:09:50,280 SAMPLE OF TRAINING AND TESTING, 80% FOR 183 00:09:50,280 --> 00:09:53,280 TRAINING AND CROSS VALIDATION AND 20% 184 00:09:53,280 --> 00:09:55,560 FOR TESTING, GIVEN THE IMBALANCE IN THE 185 00:09:55,560 --> 00:09:57,720 DATA SET, WE DON'T SAMPLE THE MAJORITY 186 00:09:57,720 --> 00:10:00,160 CLASS TO BALANCE THE DATA AND ACHIEVE 187 00:10:00,160 --> 00:10:09,560 THE SAME PROPORTION OF CASES TO 188 00:10:09,560 --> 00:10:09,840 CONTROLS. 189 00:10:09,840 --> 00:10:12,760 WE RAN A SEARCH TO BUILD THE BEST 190 00:10:12,760 --> 00:10:13,120 MODELS. 191 00:10:13,120 --> 00:10:15,640 THEY WERE THEN OBTAINED TO THE TEST SET. 192 00:10:15,640 --> 00:10:18,840 WE AVERAGED A RESULT OF THE PREDICTION 193 00:10:18,840 --> 00:10:23,600 SCORES OVER THE TRIALS WE MADE. 194 00:10:23,600 --> 00:10:24,080 NEXT PLEASE. 195 00:10:24,080 --> 00:10:29,920 SO WE FOUND THAT IN SIMPLE METHODS, LIKE 196 00:10:29,920 --> 00:10:32,440 IN BOOSTING, XGBOOST AND RANDOM MODELS 197 00:10:32,440 --> 00:10:34,120 PERFORMED THE BEST, THIS RESULT 198 00:10:34,120 --> 00:10:37,880 INDICATED OUR MODELS ARE GOOD PREDICTERS 199 00:10:37,880 --> 00:10:42,040 ON PREECLAMPSIA RISK AS EARLY AS VISIT 1 200 00:10:42,040 --> 00:10:43,640 AND PREDICTION CAPABILITIES INCREASE 201 00:10:43,640 --> 00:10:44,720 WITH GUESTATIONAL AGE. 202 00:10:44,720 --> 00:10:48,640 WE OBTAINED PRETTY GOOD AUCs AND 203 00:10:48,640 --> 00:10:50,000 SCORES ACROSS THE VISITS. 204 00:10:50,000 --> 00:10:52,960 WHAT MAKES THIS COMPELLING FOR 205 00:10:52,960 --> 00:10:54,240 PREECLAMPSIA AND PROBABLY EXPLAINS THEIR 206 00:10:54,240 --> 00:10:55,720 PERFORMANCE IS THAT BUILDING THE VARIETY 207 00:10:55,720 --> 00:11:00,120 OF MODELS FROM THE SAME COHORT, GIVE A 208 00:11:00,120 --> 00:11:02,960 CHANCE FOR DIFFERENT TYPES TO EMERGE AND 209 00:11:02,960 --> 00:11:05,920 CAPTURE DIFFERENT SUBTYPES OF 210 00:11:05,920 --> 00:11:06,240 PREECLAMPSIA. 211 00:11:06,240 --> 00:11:08,240 NEXT PLEASE. 212 00:11:08,240 --> 00:11:10,320 WE ALSO EXTRACTED THE FEATURED 213 00:11:10,320 --> 00:11:12,120 IMPORTANCE OF LEADING TOP MANUFACTURES 214 00:11:12,120 --> 00:11:13,920 FOR THE LEADING MODELS, THE FEATURES 215 00:11:13,920 --> 00:11:17,400 FOUND TO BE MOST INFORMATIVE FOR VISIT 216 00:11:17,400 --> 00:11:19,120 3, INCLUDE, WEIGHT, RESTING BLOOD 217 00:11:19,120 --> 00:11:23,200 PRESSURE, DIET AND SERUMS BIOMARKERS, IN 218 00:11:23,200 --> 00:11:25,040 ADDITION TO THESE, [INDISCERNIBLE] INDEX 219 00:11:25,040 --> 00:11:26,840 WAS FOUND TO BE IMPORTANT. 220 00:11:26,840 --> 00:11:33,240 THE BIOMARKERS FOUND TO BE INFORMATIVE 221 00:11:33,240 --> 00:11:35,760 ARE [INDISCERNIBLE], VGF, PHG AND ITEM 222 00:11:35,760 --> 00:11:37,400 12 AND [INDISCERNIBLE] WHICH IS 223 00:11:37,400 --> 00:11:40,120 CONNIVITIENT WITH THE LITERATURE. 224 00:11:40,120 --> 00:11:44,840 WE ALSO FOUND THAT COMSUMPTION OF FATTY 225 00:11:44,840 --> 00:11:46,400 ACIDS, SATURATED FAT AND CHOLESTEROL TO 226 00:11:46,400 --> 00:11:49,880 BE PREDICTIVE AS WELL. 227 00:11:49,880 --> 00:11:50,360 NEXT, PLEASE. 228 00:11:50,360 --> 00:11:56,080 WE FURTHER EXPLORED THE PLOTS OR PDPs 229 00:11:56,080 --> 00:11:58,520 TO DETERMINE THE MARGINAL EFFECT OF ANY 230 00:11:58,520 --> 00:12:00,760 GIVEN FEATURE OF THE AVERAGE PREDICTIVE 231 00:12:00,760 --> 00:12:01,880 RESULT ON THE MODEL. 232 00:12:01,880 --> 00:12:04,400 FOR INSTANCE FOR THE RISK FACTOR OF THE 233 00:12:04,400 --> 00:12:10,760 PREGNANCY BMI AND WE SEE INCREASED RISK 234 00:12:10,760 --> 00:12:12,640 OF PE, OF 23.38, ON THE LEFT CURVE AND 235 00:12:12,640 --> 00:12:21,160 110 FOR THE RESTING RESTING RESTING SYST OLIC PRESSURE ON 236 00:12:21,160 --> 00:12:22,480 THE RIGHT. 237 00:12:22,480 --> 00:12:24,840 WE FOUND THIS COULD BE NOR CONSERVATIVE 238 00:12:24,840 --> 00:12:26,720 THAN CURRENT PRACTICE. 239 00:12:26,720 --> 00:12:27,440 NEXT SLIDE, PLEASE. 240 00:12:27,440 --> 00:12:29,760 >> OUR MODELS FOR RACIAL AS WELL FOR 241 00:12:29,760 --> 00:12:34,440 FAIRNESS AND OBSERVED A BIAS MAINLY 242 00:12:34,440 --> 00:12:35,960 AGAINST NONHISPANIC BLACK WOMEN AS SHOWN 243 00:12:35,960 --> 00:12:37,440 HERE IN THE RED BAR. 244 00:12:37,440 --> 00:12:39,840 USING THE WHITE RACE AS A REFERENCE 245 00:12:39,840 --> 00:12:43,640 RACE, WE IDENTIFIED THAT THE PREDICTIVE 246 00:12:43,640 --> 00:12:46,720 REACH OF A BLACK WOMEN WAS 1.314 WHICH 247 00:12:46,720 --> 00:12:48,880 IS HIGH ACCORDING TO THE FIFTH FIGHT 248 00:12:48,880 --> 00:12:49,760 RIEWFUL. 249 00:12:49,760 --> 00:12:52,480 IN OTHER WORDS OUR MODELS FALSE-POSITIVE 250 00:12:52,480 --> 00:12:55,040 RATE WAS HIGHER IN THE BLACK POPULATION 251 00:12:55,040 --> 00:12:56,560 WHICH IS PROBLEMATIC SINCE PATIENTS WITH 252 00:12:56,560 --> 00:12:58,440 NO RISK OF PE IN THE SUBPOPULATION WOULD 253 00:12:58,440 --> 00:13:01,920 BE SUBJECTED AND THIS IS VERY--THIS 254 00:13:01,920 --> 00:13:03,400 TESTS TREATMENT AND POTENTIAL TERM 255 00:13:03,400 --> 00:13:07,800 NATION OF PREGNANCY IF THIS MODEL WERE 256 00:13:07,800 --> 00:13:09,520 TO BE USED. 257 00:13:09,520 --> 00:13:10,560 NEXT, PLEASE. 258 00:13:10,560 --> 00:13:12,360 ONE WAY TO CORRECT FOR THE BIAS IS TO 259 00:13:12,360 --> 00:13:15,040 DETERMINE THE BEST PREDICTION THRESHOLD 260 00:13:15,040 --> 00:13:16,840 FOR THE AFRICAN AMERICAN POPULATION TO 261 00:13:16,840 --> 00:13:22,720 ACHIEVE THIS WE PLOTTED SUCH A PLOT OF 262 00:13:22,720 --> 00:13:26,240 THE SUBPOPULATION TO DETERMINE THE 263 00:13:26,240 --> 00:13:29,440 SELECTION CRITERIA. 264 00:13:29,440 --> 00:13:29,760 NEXT PLEASE. 265 00:13:29,760 --> 00:13:31,240 BY SLIGHTING INCREASING THE THRESHOLD WE 266 00:13:31,240 --> 00:13:35,040 WERE ABLE TO CHOOSE THE PREDICTIVE 267 00:13:35,040 --> 00:13:37,880 QUALITY RATIO, FOR AFRICAN AMERICAN 268 00:13:37,880 --> 00:13:40,840 MEANT 1.114 AS SHOWN IN THE FIGURE WITH 269 00:13:40,840 --> 00:13:41,600 THE [INDISCERNIBLE]. 270 00:13:41,600 --> 00:13:44,000 TO SUMMARIZE OUR WORK, AND OUR 271 00:13:44,000 --> 00:13:46,320 EXPERIMENTS, OUR WORK SHOWS THAT IT IS 272 00:13:46,320 --> 00:13:48,560 POSSIBLE TO CREATE A COMPOSITE SCREENING 273 00:13:48,560 --> 00:13:54,040 MODEL TO PREDICT THE WOMAN AT RISK OF 274 00:13:54,040 --> 00:13:56,040 DEVELOPING SEVERE PREELAMPSIA AND 275 00:13:56,040 --> 00:13:58,840 PREELAMPSIA AT 13 WEEKS GUESTATION, THE 276 00:13:58,840 --> 00:14:01,120 LINES ARE IN LINE WITH THE LITERATURE. 277 00:14:01,120 --> 00:14:06,560 THEY STRESS THE IMPORTANCE OF USING 278 00:14:06,560 --> 00:14:09,240 SEVERAL TESTS AND USING THE INDEX, WE 279 00:14:09,240 --> 00:14:12,680 INCLUDED IT TO INCLUDE ANY SOCIOECONOMIC 280 00:14:12,680 --> 00:14:14,320 OR RACIAL DISPARITY IN THE MODEL AND 281 00:14:14,320 --> 00:14:15,880 PROCESS TO METIGATE THEM. 282 00:14:15,880 --> 00:14:17,680 FINALLY THE PROPOSED WORK COULD BE 283 00:14:17,680 --> 00:14:21,280 EASILY ADAPTED TO ADDRESS ALL THE 284 00:14:21,280 --> 00:14:25,280 MATERNAL MORBIDITYS AND APORKSs, NEXT 285 00:14:25,280 --> 00:14:25,640 SLIDE, PLEASE. 286 00:14:25,640 --> 00:14:30,680 OUR TEAM HAS BEEN WORKING ON THE 287 00:14:30,680 --> 00:14:32,680 CHALLENGE AND THIS OPENED A NEW RESEARCH 288 00:14:32,680 --> 00:14:36,360 OPPORTUNITY THAT WE WOULD LIKE TO 289 00:14:36,360 --> 00:14:36,640 PURSUE. 290 00:14:36,640 --> 00:14:38,720 CURRENTLY, WE ARE PREPARING PUBLICATION 291 00:14:38,720 --> 00:14:42,400 TO SHOW RESULTS, WE ARE ALSO BUILDING 292 00:14:42,400 --> 00:14:45,680 MODELS TO COMPARE EARLY, FOR BEFORE 34 293 00:14:45,680 --> 00:14:52,080 WEEKS AND LATE ONSET OF PE, AFTER 34 294 00:14:52,080 --> 00:14:54,440 WEEKS AND WE ALSO WILL RUN OTHER TESTS 295 00:14:54,440 --> 00:14:56,600 FOR THE THRESHOLD MODELS, HOWEVER, 296 00:14:56,600 --> 00:14:59,040 FURTHER RESEARCH IS USED TO DEFINE TO 297 00:14:59,040 --> 00:15:03,720 FIND CAUSES OF THE BIAS AND ALSO 298 00:15:03,720 --> 00:15:05,680 IDENTIFYING PREECLAMPSIA BITES WITH THE 299 00:15:05,680 --> 00:15:07,280 MACHINE LEARNING AND INTERPRETABLE 300 00:15:07,280 --> 00:15:09,640 MODELS, I WOULD LIKE TO THANK 301 00:15:09,640 --> 00:15:11,800 DR. MAURICE AND THE CHALLENGE TEAM FOR 302 00:15:11,800 --> 00:15:14,640 THIS EXCITING CHALLENGE AND THANK ALL OF 303 00:15:14,640 --> 00:15:17,480 OUR COLLABORATORS FROM COLUMBIA AND ALL 304 00:15:17,480 --> 00:15:22,000 OF OUR STUDENTS. 305 00:15:22,000 --> 00:15:24,040 THANK YOU. 306 00:15:24,040 --> 00:15:26,360 >> THANK YOU DR. SALEB. 307 00:15:26,360 --> 00:15:27,840 WE DO HAVE TIME FOR QUESTIONS SO WE'RE 308 00:15:27,840 --> 00:15:31,960 GOING TO TAKE QUESTIONS AFTER EACH 309 00:15:31,960 --> 00:15:32,240 RESENTATION. 310 00:15:32,240 --> 00:15:33,840 SO WE DO HAVE A FEW MINUTES. 311 00:15:33,840 --> 00:15:37,600 SO AND I WANT TO REMIND ALL PARTICIPANTS 312 00:15:37,600 --> 00:15:40,120 THAT IF YOU WOULD LIKE TO ASK A 313 00:15:40,120 --> 00:15:45,680 QUESTION, PLEASE USE THE QUESTION AND 314 00:15:45,680 --> 00:15:50,200 ANSWER BUTTON AT THE BOTTOM OF YOUR 315 00:15:50,200 --> 00:15:50,440 SCREEN. 316 00:15:50,440 --> 00:15:51,640 THERE IS 1 QUESTION THAT I WOULD LIKE TO 317 00:15:51,640 --> 00:15:55,920 ASK AND THAT IS HOW IS THE 4-FIFTH RULE 318 00:15:55,920 --> 00:16:01,000 APPLIED IN THE CONTEXT OF FAIRNESS CHECK 319 00:16:01,000 --> 00:16:01,640 NOTHING YOUR APPLICATION. 320 00:16:01,640 --> 00:16:02,040 >> THANK YOU. 321 00:16:02,040 --> 00:16:06,040 THIS IS A GREAT QUESTION. 322 00:16:06,040 --> 00:16:08,840 SO FROM THE 4-FIFTHS RULE, IT'S APPLIED 323 00:16:08,840 --> 00:16:12,040 IN THE PROCESS AND THINGS LIKE THAT, IN 324 00:16:12,040 --> 00:16:13,640 OUR CONTEXT FOR THE UNDERPRIVILEGED 325 00:16:13,640 --> 00:16:15,960 GROUP WHICH IS THE AFRICAN WOMEN IN THE 326 00:16:15,960 --> 00:16:18,640 DATA WHEN THEY RECEIVE THE FIRST 327 00:16:18,640 --> 00:16:19,840 POSITIVE RATE OUTCOME, THAT'S BECAUSE 328 00:16:19,840 --> 00:16:23,360 THE GROUP IS MORE THAN 20%, SO THAT'S 329 00:16:23,360 --> 00:16:26,160 PARTICULARLY FOR THE GROUP TO REPRESENT 330 00:16:26,160 --> 00:16:28,200 IF THE--IF THE RATE OF THE 331 00:16:28,200 --> 00:16:33,040 FALSE-POSITIVE RATE IS BIGGER BY 20% AND 332 00:16:33,040 --> 00:16:35,920 THE RATE OF THE WHITE CAUCASIAN FEMALE 333 00:16:35,920 --> 00:16:37,960 IN THE DATA THEN WE DISCOVER THE 334 00:16:37,960 --> 00:16:39,040 DISPARITY AND THAT'S WHAT WE WANT TO 335 00:16:39,040 --> 00:16:43,480 BRING IT DOWN TO LESS THAN BASICALLY 336 00:16:43,480 --> 00:16:43,960 20%. 337 00:16:43,960 --> 00:16:47,840 AFTER THE MITIGATION, WE REDUCED THE 338 00:16:47,840 --> 00:16:51,760 DISPARITY FROM 1.31 TO 1.2 WHICH IS MORE 339 00:16:51,760 --> 00:16:55,040 SUSCEPTIBLE AND FILLS THE RULE OF THE 340 00:16:55,040 --> 00:16:58,520 4-FIFTHS RULE. 341 00:16:58,520 --> 00:16:59,640 THANK YOU. 342 00:16:59,640 --> 00:17:00,480 >> GREAT, THANK YOU. 343 00:17:00,480 --> 00:17:01,840 THERE IS 1 MORE QUESTION THAT I JUST 344 00:17:01,840 --> 00:17:05,720 RECEIVED FROM A PANELISTS, I WOULD LIKE 345 00:17:05,720 --> 00:17:07,880 TO ASK YOU DID YOU COMPARE THE 346 00:17:07,880 --> 00:17:11,560 PERFORMANCE OF DOWN SAMPLING VERSUS 347 00:17:11,560 --> 00:17:16,600 UPSAMPLING FOR SYNTHETIC DATA SUCH AS 348 00:17:16,600 --> 00:17:16,840 SMOOT? 349 00:17:16,840 --> 00:17:19,840 >> THIS IS A GOOD QUESTION. 350 00:17:19,840 --> 00:17:22,200 WE TRIED SMOOT TO SAMPLE THE MINORITY 351 00:17:22,200 --> 00:17:27,000 CLASS AND WE ALSO TRIED ENDOSAMPLING, SO 352 00:17:27,000 --> 00:17:30,040 WE HAD SLIGHTY BETTER PERPERFORMANCE 353 00:17:30,040 --> 00:17:31,440 WITH UNDERSAMPLING AND IT WAS FASTER TO 354 00:17:31,440 --> 00:17:33,640 TRAIN THE MODELS THAN USING 355 00:17:33,640 --> 00:17:33,960 OVERSAMPLING. 356 00:17:33,960 --> 00:17:35,280 SOPHISTICATEDY WE FOUND BETTER AREY 357 00:17:35,280 --> 00:17:39,840 SULTS WITH THE SMALL MARGIN VERY RESULTS 358 00:17:39,840 --> 00:17:43,040 WITH UNDERSAMPLING, SO YES, THIS WOULD 359 00:17:43,040 --> 00:17:46,920 BE WAY OR THE OTHERS TO ACCOMPLISH THE 360 00:17:46,920 --> 00:17:50,240 BALANCING OF THE DATA SET. 361 00:17:50,240 --> 00:17:55,640 >> AND I THINK WE HAVE TIME FOR 1 MORE 362 00:17:55,640 --> 00:17:57,440 QUESTION, DID YOU TRY OTHER BAGGING 363 00:17:57,440 --> 00:18:00,640 METHODS SUCH AS RANDOM FORCE, 364 00:18:00,640 --> 00:18:02,440 CLASSIFIERS VERSUS BOOSTING METHODS? 365 00:18:02,440 --> 00:18:05,040 >> THIS IS A GREAT QUESTION. 366 00:18:05,040 --> 00:18:07,800 QUESTION YES WE TRIED BOTH, EXTRA BOOST 367 00:18:07,800 --> 00:18:10,920 WHICH IS A BOOSTING METHOD, SO THERE ARE 368 00:18:10,920 --> 00:18:14,240 BOTH SAMPLES, THEY OFFER A BIT 369 00:18:14,240 --> 00:18:16,920 DIFFERENTLY SO FOR THE STUDY FOR US IT 370 00:18:16,920 --> 00:18:18,640 DEMONSTRATED THAT YOU MAY VOTE FOR THE 371 00:18:18,640 --> 00:18:21,720 MODEL AND IMPROVE IT BY DOING SOME 372 00:18:21,720 --> 00:18:22,080 [INDISCERNIBLE]. 373 00:18:22,080 --> 00:18:25,040 AND WE TRIED BOTH AND WE SAW CLOSE 374 00:18:25,040 --> 00:18:27,560 PERFORMANCE BETWEEN RANDOM FORCE AND 375 00:18:27,560 --> 00:18:29,520 EXTRA BOOSTING, GREAT, QUESTION, THANK 376 00:18:29,520 --> 00:18:31,040 YOU. 377 00:18:31,040 --> 00:18:32,720 >> THANK YOU DR. SALLEB, IF WE HAVE TIME 378 00:18:32,720 --> 00:18:35,320 AT THE END, I MAY ASK SOME ADDITIONAL 379 00:18:35,320 --> 00:18:37,320 QUESTIONS BUT FOR RIGHT NOW, SINCE WE 380 00:18:37,320 --> 00:18:39,160 HAVE MORE PRESENTATIONS I THINK WE NEED 381 00:18:39,160 --> 00:18:40,640 TO MOVE FORWARD JUST A LITTLE BIT. 382 00:18:40,640 --> 00:18:41,960 BUT THANK YOU VERY MUCH FOR YOUR 383 00:18:41,960 --> 00:18:53,080 PRESENTATION AND FOR YOUR WORK. 384 00:18:53,080 --> 00:18:55,720 >> OUR NEXT PRESENTER IS MISS BRITTANY 385 00:18:55,720 --> 00:18:57,520 JOHNSTON FROM THE JOHN TON COMPANY OUT 386 00:18:57,520 --> 00:19:01,920 OF SALT LAKE CITY, UTAH AND THE TITLE OF 387 00:19:01,920 --> 00:19:05,640 HER PRESENTATION IS: THE RELATIONSHIP 388 00:19:05,640 --> 00:19:08,240 BETWEEN MARGINALIZING BEHAVIORS AND 389 00:19:08,240 --> 00:19:11,320 POSTPARTUM COMPLICATIONS FOR NULLIP A 390 00:19:11,320 --> 00:19:16,240 ROUS WOMEN RECEIVING AN UNDESIRED 391 00:19:16,240 --> 00:19:16,920 C-SECTION. 392 00:19:16,920 --> 00:19:17,440 BRITNEE? 393 00:19:17,440 --> 00:19:18,080 >> HI, THANK YOU. 394 00:19:18,080 --> 00:19:19,720 >> YOU'RE ON, THANK YOU. 395 00:19:19,720 --> 00:19:22,560 >> HI, EVERYONE I'M BRITNEE JOHNSTON. 396 00:19:22,560 --> 00:19:24,040 THANK YOU FOR HAVING ME TODAY, I'M 397 00:19:24,040 --> 00:19:26,640 EXCITED TO SHARE A SUMMARY OF MY DATA 398 00:19:26,640 --> 00:19:28,240 PROPOSAL WITH YOU ALL. 399 00:19:28,240 --> 00:19:29,600 LET'S GO TO THE NEXT SLIDE. 400 00:19:29,600 --> 00:19:34,040 SO IN MY PROPOSAL MY HYPOTHESIS WAS THAT 401 00:19:34,040 --> 00:19:36,040 A CORCALCULATIONS EXISTS BETWEEN 402 00:19:36,040 --> 00:19:38,680 MARGINALIZING BEHAVIORS AND POSTPARTUM 403 00:19:38,680 --> 00:19:52,600 COMPLICATIONS, SPECIFICALLY WITH AN 404 00:19:52,600 --> 00:19:55,880 ENULLIPPOR OUS DATA SET. 405 00:19:55,880 --> 00:19:58,360 --IF THEY SPEAK UP IF THEY'RE TREATED 406 00:19:58,360 --> 00:20:06,760 UNFAIRLY AND IF THEY HAD LOW 407 00:20:06,760 --> 00:20:07,320 SOCIOECONOMIC STAT US. 408 00:20:07,320 --> 00:20:07,640 NEXT SLIDE. 409 00:20:07,640 --> 00:20:11,480 SO I WANT TO PROVIDE DATA HOW I CAME UP 410 00:20:11,480 --> 00:20:11,960 WITH THE IDEA. 411 00:20:11,960 --> 00:20:15,640 SO IN THE NICHD, THERE ARE THOUSANDS OF 412 00:20:15,640 --> 00:20:20,120 ROWS OF POSSIBLE VARIABLES DATA THAT WE 413 00:20:20,120 --> 00:20:22,800 COULD EXPLORE AND 1 THAT STOOD OUT TO ME 414 00:20:22,800 --> 00:20:24,080 IF SOMEONE EVER EXPERIENCED DISCROSS 415 00:20:24,080 --> 00:20:25,440 EXAMINE NATION IN A HEALTHCARE SETTING, 416 00:20:25,440 --> 00:20:26,640 I THOUGHT IT WAS INTERESTING. 417 00:20:26,640 --> 00:20:30,360 SO I LOOKEDDA THE PAST RESEARCH THAT 418 00:20:30,360 --> 00:20:32,360 NICHD HAD CONDUCTEDUTESSING THE NEW MOM 419 00:20:32,360 --> 00:20:34,560 TO BE DATA SET AND I SAW THAT 420 00:20:34,560 --> 00:20:35,640 DISCRIMINATION HAD NOT BEEN EXPLORED 421 00:20:35,640 --> 00:20:37,600 BEFORE SO I THOUGHT THIS MIGHT BE AN 422 00:20:37,600 --> 00:20:39,760 INNOVATIVE WAY TO LOOK AT THE DATA. 423 00:20:39,760 --> 00:20:41,840 I TOOK IT 1 STEP FURTHER BY NARROWING 424 00:20:41,840 --> 00:20:44,760 THE GROUP I WOULD LIKE AT TO WOMEN WHO 425 00:20:44,760 --> 00:20:46,640 RECEIVED C-SECTIONS AND SO IN MY BRIEF 426 00:20:46,640 --> 00:20:48,520 LITERATURE REVIEW, I SAW THAT C-SECTIONS 427 00:20:48,520 --> 00:20:51,080 WERE KNOWN TO BE AN UNNECESSARY RISK IF 428 00:20:51,080 --> 00:20:52,480 IT WAS NOT MEDICALLY NEEDED AND THEY 429 00:20:52,480 --> 00:20:55,720 COULD BE MORE RISKY THAN VAGINAL 430 00:20:55,720 --> 00:20:56,200 DELIVERIES. 431 00:20:56,200 --> 00:20:58,360 YET THE RATES OF C-SECTIONS HAVE 432 00:20:58,360 --> 00:21:02,040 INCREASED FROM 7% IN 1990 TO 21% IN 433 00:21:02,040 --> 00:21:02,240 2021. 434 00:21:02,240 --> 00:21:04,480 SO IT IS BECOMING A MORE COMMON 435 00:21:04,480 --> 00:21:04,840 PROCEDURE. 436 00:21:04,840 --> 00:21:09,640 I ALSO FOUND THAT C-SECTIONS THEIR RATES 437 00:21:09,640 --> 00:21:14,400 WERE HIGHER FOR BLACK WOMEN N. A 438 00:21:14,400 --> 00:21:15,960 CALIFORNIA HEALTH REPORT, THEY WERE 439 00:21:15,960 --> 00:21:17,320 QUESTIONING WHETHER BLACK WOMEN WERE 440 00:21:17,320 --> 00:21:19,080 BEING LISTENED TO AND INCLUDED IN THE 441 00:21:19,080 --> 00:21:20,840 TEAM FOR MAKING DECISIONS FOR THEIR 442 00:21:20,840 --> 00:21:21,040 HEALTH. 443 00:21:21,040 --> 00:21:23,840 SO THAT MADE ME QUESTION IF THERE WAS A 444 00:21:23,840 --> 00:21:25,440 LINK BETWEEN BEING DISCRIMINATED AGAINST 445 00:21:25,440 --> 00:21:27,760 AND NOT FAIRLY LISTENED TO AND MAKING A 446 00:21:27,760 --> 00:21:29,120 DECISION IN THE DELIVERY ROOM AND 447 00:21:29,120 --> 00:21:30,800 LEADING TO A C-SECTION AND TO A 448 00:21:30,800 --> 00:21:32,000 COMPLICATION AFTER THE FACT. 449 00:21:32,000 --> 00:21:33,560 SO THAT'S WHERE MY INITIAL THINKING CAME 450 00:21:33,560 --> 00:21:37,880 FROM AND HOW I CAME UP WITH THIS 451 00:21:37,880 --> 00:21:38,640 RESEARCH PROPOSAL. 452 00:21:38,640 --> 00:21:39,040 NEXT SLIDE. 453 00:21:39,040 --> 00:21:41,040 SO FOR THE DATA AND SAMPLE, I USE THE 454 00:21:41,040 --> 00:21:43,160 NEW MOM TO BE DATA SET THAT WAS PROVIDED 455 00:21:43,160 --> 00:21:45,040 TO US IN DATA CHALLENGE. 456 00:21:45,040 --> 00:21:48,040 I CREATED A SAMPLE GROUP OF JUST OVER 457 00:21:48,040 --> 00:21:51,640 1100 WOMEN WHO-FIC WANTED A VAGINAL 458 00:21:51,640 --> 00:21:53,840 DELIVERY BUT RECEIVED A C-SECTION 459 00:21:53,840 --> 00:21:55,560 INSTEAD AND COMPARISON WITH SUMMARY 460 00:21:55,560 --> 00:21:59,200 STATISTICS WITH JUST OVER 3600 WOMEN WHO 461 00:21:59,200 --> 00:22:00,960 RECEIVED THEIR DESIRED VAGINAL DELIVERY 462 00:22:00,960 --> 00:22:03,800 SO I DIDN'T USE JUST A 1 OUTCOME RESULT 463 00:22:03,800 --> 00:22:05,280 VARIABLE THAT THEY RECEIVED A C-SECTION, 464 00:22:05,280 --> 00:22:07,240 I WANT TO MAKE SURE IT REFLECTED THAT A 465 00:22:07,240 --> 00:22:09,000 DECISION HAD TO HAVE BEEN MADE TO CHANGE 466 00:22:09,000 --> 00:22:11,240 THEIR MIND TO END UP WITH A C-SECTION 467 00:22:11,240 --> 00:22:17,320 INSTEAD AND THAT IT WASN'T AN ELECTIVE 468 00:22:17,320 --> 00:22:18,360 DECISION OR PREPLANNED. 469 00:22:18,360 --> 00:22:18,800 NEXT SLIDE. 470 00:22:18,800 --> 00:22:24,600 SO FOR MY METHOD, I USED A REGRESSION 471 00:22:24,600 --> 00:22:27,240 MODEL USING R, MY DICHOT MOUSE DEPENDENT 472 00:22:27,240 --> 00:22:30,440 VARIABLE WAS AN INSTANCE OF POSTPARTUM 473 00:22:30,440 --> 00:22:33,000 COMPLICATION, THAT INCLUDED HEMORRHAGE, 474 00:22:33,000 --> 00:22:34,080 RETAINED PLACENTA, WOUND INFECTION AND 475 00:22:34,080 --> 00:22:34,760 SO ON. 476 00:22:34,760 --> 00:22:38,400 FOR MY INDEPENDENT VARIABLES, THOSE 477 00:22:38,400 --> 00:22:40,080 INCLUDED AGE, RACE AND ETHNICITY, 478 00:22:40,080 --> 00:22:48,440 WHETHER THEY WERE MARRIED, THEIR 479 00:22:48,440 --> 00:22:57,440 EDUCATION LEVEL, IF THEY--THEY WERE 200% 480 00:22:57,440 --> 00:23:00,040 BELOW THE POVERTY LEVEL, THEY INCLUDED 481 00:23:00,040 --> 00:23:01,160 HEALTH RELATED VARIABLES INCLUDING IF 482 00:23:01,160 --> 00:23:02,760 THEY WERE BORN BY VAGINAL DELIVERY 483 00:23:02,760 --> 00:23:05,800 THEMSELVES WHEN THEY WERE A BABY AND 484 00:23:05,800 --> 00:23:10,440 ALSO MEDICAL REASONS FOR THEIR 485 00:23:10,440 --> 00:23:10,720 C-SECTION. 486 00:23:10,720 --> 00:23:13,800 THAT INCLUDED IF THE BABY WAS TOO BIG OR 487 00:23:13,800 --> 00:23:16,280 CERVIX WAS NOT DILATING AND ALSO 488 00:23:16,280 --> 00:23:18,240 INCLUDED IF THE WOMAN DIDN'T KNOW HER 489 00:23:18,240 --> 00:23:21,280 REASON FOR HER C-SECTION. 490 00:23:21,280 --> 00:23:22,040 NEXT SLIDE. 491 00:23:22,040 --> 00:23:24,320 SO CONTINUING ON MY METHOD, I CHECKED 492 00:23:24,320 --> 00:23:25,720 FOR CLASSIFICATION BIAS AND I FOUND THAT 493 00:23:25,720 --> 00:23:28,520 THE PROPORTION OF INSTANCES AND 494 00:23:28,520 --> 00:23:29,680 NONINSTANCES OF POSTPARTUM COMPLICATIONS 495 00:23:29,680 --> 00:23:32,800 WAS NOT THE SAME. 496 00:23:32,800 --> 00:23:36,800 SO I TESTED SAMPLING METHODS TO MODIFY 497 00:23:36,800 --> 00:23:38,440 THE IMBALANCED DATA TO A BALANCED 498 00:23:38,440 --> 00:23:41,040 DISTRIBUTION. 499 00:23:41,040 --> 00:23:46,680 SO I TESTED UNDER SAMPLING, AND 500 00:23:46,680 --> 00:23:48,920 OVERCHAMPLING AND BOTH UNDER AND OVER 501 00:23:48,920 --> 00:23:53,440 SAMPLING, AND DATA GENERATION, SO OF 502 00:23:53,440 --> 00:23:55,880 THOSE TESTS, OVERSAMPLING WAS THE BEST 503 00:23:55,880 --> 00:23:59,240 MODEL FOR ACCURACY WITH AN AUCOF 0.74. 504 00:23:59,240 --> 00:24:02,920 AND THEN I TOOK AN ADDITIONAL STEP TO 505 00:24:02,920 --> 00:24:05,600 GROUP THOSE TOGETHER BY RACE ASK 506 00:24:05,600 --> 00:24:07,000 ETHNICITY FOR STATISTICS TO SEE IF THERE 507 00:24:07,000 --> 00:24:08,600 WERE TREPPEDS THERE AS WELL. 508 00:24:08,600 --> 00:24:09,040 NEXT SLIDE. 509 00:24:09,040 --> 00:24:11,560 OKAY, SO ON TO THE FINDINGS, SO I WILL 510 00:24:11,560 --> 00:24:13,880 SHARE 4 OF THE MAIN FINDINGS FROM MY 511 00:24:13,880 --> 00:24:15,920 PROPOSAL, THE FIRST IS THAT C-SECTIONS 512 00:24:15,920 --> 00:24:19,960 HAD A HIGHER RATE OF POSTPARTUM 513 00:24:19,960 --> 00:24:22,280 COMPLICATIONS AT 9% COMPARED TO THOSE 514 00:24:22,280 --> 00:24:24,960 WHO RECEIVED VAGINAL DELIVERY AT 4%. 515 00:24:24,960 --> 00:24:27,040 BY RACE AND ETHNICITY, BLACK WOMEN WITH 516 00:24:27,040 --> 00:24:32,480 C-SECTIONS HAD THE HIGHEST RATEST 517 00:24:32,480 --> 00:24:33,920 POSTPARTUM COMPLICATIONS AT 15% FOLLOWED 518 00:24:33,920 --> 00:24:37,040 BY ASIAN WOMEN AT 14%. 519 00:24:37,040 --> 00:24:39,440 SO THIS MATCHED WHAT I HAD SEEN IN MY 520 00:24:39,440 --> 00:24:43,120 LITERATURE REVIEW. 521 00:24:43,120 --> 00:24:43,480 NEXT SLIDE. 522 00:24:43,480 --> 00:24:46,080 DISCRIMINATION HAD A HIGHER LIKELIHOOD 523 00:24:46,080 --> 00:24:47,480 OF POSTPARTUM COMPLICATIONS SO THE ODDS 524 00:24:47,480 --> 00:24:50,000 OF HAVING A POSTPARTUM COMPLICATION WAS 525 00:24:50,000 --> 00:24:52,680 2.7 TIMES HIGHER IF THE WOMAN 526 00:24:52,680 --> 00:24:53,920 EXPERIENCED DISCRIMINATION IN HEALTHCARE 527 00:24:53,920 --> 00:24:55,560 SETTING COMPARED TO THOSE WHO DID NOT. 528 00:24:55,560 --> 00:24:59,320 AS FAR RACE AND ETHNICITY, BLACK WOMEN 529 00:24:59,320 --> 00:25:02,120 HAD THE HIGHEST RATES OF EXPERIENCING 530 00:25:02,120 --> 00:25:04,440 DISCRIMINATION AT 6%, AND KEEPING TO 531 00:25:04,440 --> 00:25:11,320 THEMSELVES IF TREATED UNFAIRLY AT 15%. 532 00:25:11,320 --> 00:25:12,400 NEXT SLIDE. 533 00:25:12,400 --> 00:25:13,520 SOCIOECONOMIC STATUS PLAYED A PART AS 534 00:25:13,520 --> 00:25:17,720 WELL, SO IN MY MODEL THE ODDS OF HAVING 535 00:25:17,720 --> 00:25:19,320 A POSTPARTUM COMPLICATION WERE 69% 536 00:25:19,320 --> 00:25:21,440 HIGHER IF THE WOMAN WAS LOW INCOME. 537 00:25:21,440 --> 00:25:25,600 IT WAS 49% LOWER IF THE WOMEN WAS 538 00:25:25,600 --> 00:25:28,120 MARRIED AND 33% LOWER IF THE WOMEN HAD A 539 00:25:28,120 --> 00:25:29,560 HIGH SCHOOL EDUCATION OR LESS. 540 00:25:29,560 --> 00:25:32,160 AND THEN AGAIN LOOKING AT DESCRIPTIVE 541 00:25:32,160 --> 00:25:34,560 STATISTICS, BLACK WOMEN HAD THE HIGHEST 542 00:25:34,560 --> 00:25:38,200 RATES OUT OF RACIAL/ETHNIC GROUPS FOR 543 00:25:38,200 --> 00:25:39,960 BEING LOW INCOME, BEING UNMARRIED AT 80% 544 00:25:39,960 --> 00:25:43,640 AND HAVING A HIGH SCHOOL EDUCATION OR 545 00:25:43,640 --> 00:25:44,160 LESS. 546 00:25:44,160 --> 00:25:44,560 NEXT SLIDE. 547 00:25:44,560 --> 00:25:47,120 I ALSO LOOKED AT THE MEDICAL REASONS FOR 548 00:25:47,120 --> 00:25:49,320 C-SECTION AND THEY HAD A HIGHER 549 00:25:49,320 --> 00:25:50,840 LIKELIHOOD OF THE POSTPARTUM 550 00:25:50,840 --> 00:25:54,160 COMPLICATION, SO OUT OF THE 8 VARIABLES 551 00:25:54,160 --> 00:25:57,480 RELATED TO MEDICAL REASONS, 3 OF THEM 552 00:25:57,480 --> 00:25:59,680 WERE STATISTICALLY SIGNIFICANT, SO THE 553 00:25:59,680 --> 00:26:02,640 BABY WAS TOO BIG, IF THE C-SECTION 554 00:26:02,640 --> 00:26:08,400 REASON WAS UNKNOWN TO THE MOTHER AND IF 555 00:26:08,400 --> 00:26:13,840 THEY WERE EXPERIENCING CHORIOAMNIONITIS. 556 00:26:13,840 --> 00:26:21,800 HISPANICS HAD THE HIGHEST HIGHEST RATES O F 557 00:26:21,800 --> 00:26:25,040 BABIES BEING TOO BIG, AND THEN ALSO 558 00:26:25,040 --> 00:26:27,960 ASIANS HAD THE HIGHEST RATES OF NOT 559 00:26:27,960 --> 00:26:29,320 KNOWING THEIR C-SECTION REASON. 560 00:26:29,320 --> 00:26:31,440 OVER ALL THE MEDICAL REASONS FOR 561 00:26:31,440 --> 00:26:33,320 C-SECTIONS WERE NOT HIGH FOR BLACK WOMEN 562 00:26:33,320 --> 00:26:36,840 BUT THE RATES WERE HIGH IN ALL OTHERS WE 563 00:26:36,840 --> 00:26:40,600 WERE TESTING, THIS FURTHER SUGGESTS THAT 564 00:26:40,600 --> 00:26:41,560 BLACK WOMEN'S POSTPARTUM COMPLICATIONS 565 00:26:41,560 --> 00:26:44,480 MAY BE DRIVEN MORE BY MARGINALIZING 566 00:26:44,480 --> 00:26:48,440 BEHAVIORS AND SOCIOECONOMIC STATUS 567 00:26:48,440 --> 00:26:50,400 RATHER THAT PHYSICAL HEALTH ISSUES. 568 00:26:50,400 --> 00:26:53,400 NEXT SLIDE, MY PROPOSAL SHARED A FEW 569 00:26:53,400 --> 00:26:55,840 FUTURE RESEARCH RECOMMENDATIONS 570 00:26:55,840 --> 00:26:57,800 INCLUDING ADDRESSING DISCRIMINATION IN 571 00:26:57,800 --> 00:26:58,640 HEALTHCARE. 572 00:26:58,640 --> 00:26:59,760 FACILITIES THAT PROVIDE DISCRIMINATION 573 00:26:59,760 --> 00:27:01,800 TRAINING TO THEIR MEDICAL STAFF, MIGHT 574 00:27:01,800 --> 00:27:04,960 HELP MITIGATE THESE OUTCOMES. 575 00:27:04,960 --> 00:27:07,600 SO THE FUTURE RESEARCH COULD LOOK AT 576 00:27:07,600 --> 00:27:09,720 FACILITIES THAT HAVE IMP LIAISON 577 00:27:09,720 --> 00:27:10,640 ELEMENTED DISCRIMINATION TRAINING AND 578 00:27:10,640 --> 00:27:12,480 COMPARED TO THOSE THAT HAVEN'T AND SEE 579 00:27:12,480 --> 00:27:17,040 IF THERE'S A DIFFERENCE IN POSTPARTUM 580 00:27:17,040 --> 00:27:18,360 OUTCOMES, NEXT SLIDE. 581 00:27:18,360 --> 00:27:20,040 FUTURE RESEARCH COULD ALSO FOCUS ON HOW 582 00:27:20,040 --> 00:27:23,600 SINGLE WOMEN ARE SUPPORTED DURING THE 583 00:27:23,600 --> 00:27:26,400 DELIVERY, SO LIKE I MENTIONED THE ODDS 584 00:27:26,400 --> 00:27:28,360 OF A POSTPARTUM COMPLICATION ARE 49% 585 00:27:28,360 --> 00:27:30,360 LOWER PER THOSE WHO ARE MARRIED HOWEVER 586 00:27:30,360 --> 00:27:32,880 ONLY 20% OF BLACK WOMEN WERE MARRIED. 587 00:27:32,880 --> 00:27:35,640 SO FUTURE RESEARCH COULD LOOK AT 588 00:27:35,640 --> 00:27:37,960 PROGRAMS THAT PROVIDE SUPPORT ARE TO 589 00:27:37,960 --> 00:27:39,160 SINGLE WOMEN DURING DELIVERY TO 590 00:27:39,160 --> 00:27:44,400 DETERMINE IF THERE ARE ANY DIFFERENCES 591 00:27:44,400 --> 00:27:45,200 IN THOSE OUTCOMES. 592 00:27:45,200 --> 00:27:45,520 NEXT SLIDE. 593 00:27:45,520 --> 00:27:48,560 AND THEN JUST A FEW OTHER SUGGESTIONS 594 00:27:48,560 --> 00:27:51,440 INCLUDED THAT FUTURE EFFORTS COULD FOCUS 595 00:27:51,440 --> 00:27:53,400 ON COLLECTING MORE DATA ON AMERICAN 596 00:27:53,400 --> 00:27:55,960 INDIAN WOMEN AND THEIR C-SECTION 597 00:27:55,960 --> 00:27:56,240 OUTCOMES. 598 00:27:56,240 --> 00:27:58,040 I DIDN'T HAVE ENOUGH DATAOT SUBGROUP TO 599 00:27:58,040 --> 00:28:02,120 BE ABLE TO REPORT ON THEM ON THE 600 00:28:02,120 --> 00:28:02,520 PROPOSAL. 601 00:28:02,520 --> 00:28:04,320 ALSO HEALTHCARE FACILITIES COULD WORK ON 602 00:28:04,320 --> 00:28:05,840 IMPROVING COMMUNICATION BETWEEN THE 603 00:28:05,840 --> 00:28:08,320 STAFF AND PREGNANT WOMEN, THIS MAY BE 604 00:28:08,320 --> 00:28:10,000 RELATED TO A LACK OF COMMUNICATION IN 605 00:28:10,000 --> 00:28:11,880 GENERAL FOR THOSE WHO DO NOT KNOW THE 606 00:28:11,880 --> 00:28:14,040 REASON OF THEIR C-SECTION AND MAYBE 607 00:28:14,040 --> 00:28:15,640 THEY'RE ALSO NOT BEING COMMUNICATED TO 608 00:28:15,640 --> 00:28:18,360 ABOUT HOW TO TAKE GOOD CARE OF THEIR 609 00:28:18,360 --> 00:28:21,320 C-SECTION LEARNED TO AVOID INFECTION 610 00:28:21,320 --> 00:28:23,240 AFTER THE FACT. 611 00:28:23,240 --> 00:28:24,800 LASTLY MEDICAL PROFESSIONALS THAT HELP 612 00:28:24,800 --> 00:28:27,840 PATIENTS FOCUS ON HEALTHY HABITS DURING 613 00:28:27,840 --> 00:28:28,440 THE PREGNANCY. 614 00:28:28,440 --> 00:28:30,720 WOULD BE MOST HELPFUL FOR HISPANIC WOMEN 615 00:28:30,720 --> 00:28:32,480 SINCE THEY HAD THE HIGHEST RATES OF 616 00:28:32,480 --> 00:28:36,960 MEDICAL REASONS FOR A C-SECTION. 617 00:28:36,960 --> 00:28:37,480 >> ONE MINUTE. 618 00:28:37,480 --> 00:28:38,640 >> NEXT SLIDE. 619 00:28:38,640 --> 00:28:41,280 SO IN CONCLUSION, AS MORE RESEARCH 620 00:28:41,280 --> 00:28:43,560 FOCUSES ON MARGINALIZING BEHAVIORS AND 621 00:28:43,560 --> 00:28:46,840 REVEALING THOSE TRENDS, IT WOULD PROVIDE 622 00:28:46,840 --> 00:28:49,280 MORE MOTIVATION TO HEALTHCARE FACILITIES 623 00:28:49,280 --> 00:28:53,080 TO IMPROVE PRACTICES AND TRAINING TO 624 00:28:53,080 --> 00:28:55,880 HELP REDUCE DISCRIMINATION AND BIAS AND 625 00:28:55,880 --> 00:29:00,480 POSSIBLY LEAD TO REDUCED POSTPARTUM 626 00:29:00,480 --> 00:29:02,120 COMPLICATIONS. 627 00:29:02,120 --> 00:29:02,880 NEXT SLIDE. 628 00:29:02,880 --> 00:29:05,400 SO FEEL FREE CONTACT ME IF HAVE YOU ANY 629 00:29:05,400 --> 00:29:06,960 QUESTIONS ABOUT MY DATA PROPOSAL. 630 00:29:06,960 --> 00:29:08,120 MY INFORMATION IS HERE AND I WANT TO 631 00:29:08,120 --> 00:29:10,840 WRAP UP BY SAYING THANK TO YOU NICHD FOR 632 00:29:10,840 --> 00:29:12,840 THE OPPORTUNITY TO PARTICIPATE IN THIS 633 00:29:12,840 --> 00:29:13,240 DATA CHALLENGE. 634 00:29:13,240 --> 00:29:14,720 IT'S BEEN A GREAT EXPERIENCE AND THANK 635 00:29:14,720 --> 00:29:16,840 TO YOU EVERYONE FOR YOUR TIME TODAY TO 636 00:29:16,840 --> 00:29:23,760 KIND OF LEARN MORE ABOUT OUR PROPOSALS. 637 00:29:23,760 --> 00:29:24,080 SO THANK YOU. 638 00:29:24,080 --> 00:29:26,360 >> THANK YOU VERY MUCH FOR A VERY 639 00:29:26,360 --> 00:29:26,880 INFORMATIVE PRESENTATION. 640 00:29:26,880 --> 00:29:29,920 I DO HAVE A COUPLE OF QUESTIONS IN THE 641 00:29:29,920 --> 00:29:30,240 CHAT FOR YOU. 642 00:29:30,240 --> 00:29:32,200 ONE QUESTION IN PARTICULAR, SAID THANK 643 00:29:32,200 --> 00:29:34,040 YOU FOR THIS INTERESTING TALK. 644 00:29:34,040 --> 00:29:38,680 WERE YOU ABLE TO EVALUATE POTENTIAL 645 00:29:38,680 --> 00:29:41,040 EFFECTS OF ACTIVITY SUCH AS IMMIGRATION 646 00:29:41,040 --> 00:29:44,240 STATUS AND PRIMARY SPOKEN LANGUAGE? 647 00:29:44,240 --> 00:29:46,880 >> YEAH, THAT'S A GOOD QUESTION AND A 648 00:29:46,880 --> 00:29:47,680 GOOD IDEA. 649 00:29:47,680 --> 00:29:50,600 THOSE WERE INCLUDED IN MY MODEL. 650 00:29:50,600 --> 00:29:55,360 WHOOPS--SORRY, THAT MIGHT BE ME. 651 00:29:55,360 --> 00:29:56,720 [BEEPING ], THOSE WERE NOT IN MY MODEL 652 00:29:56,720 --> 00:29:57,880 BUT WOULD HAVE BEEN INTERESTING TO 653 00:29:57,880 --> 00:30:00,280 INCLUDE AND I DON'T REMEMBER OFF THE TOP 654 00:30:00,280 --> 00:30:03,920 OF MY HEAD IF IMMIGRANT STATUS AND THE 655 00:30:03,920 --> 00:30:04,960 PRIMARY LANGUAGE SPOKEN WAS INCLUDE 656 00:30:04,960 --> 00:30:07,280 INDEED THAT DATA SET WE'RE GIVEN. 657 00:30:07,280 --> 00:30:12,160 IT WOULD BE INTERESTING TO LOOK AT. 658 00:30:12,160 --> 00:30:12,640 >> GREAT. 659 00:30:12,640 --> 00:30:14,720 LET'S SEE I THINK THERE--OKAY, I THINK 660 00:30:14,720 --> 00:30:17,280 THAT'S IT FOR THE QUESTIONS. 661 00:30:17,280 --> 00:30:19,480 THANK YOU VERY MUCH AGAIN FOR THAT 662 00:30:19,480 --> 00:30:21,560 PRESENTATION AND LIKE I SAID EARLIER IF 663 00:30:21,560 --> 00:30:23,280 WE HAVE TIME AND ADDITIONAL QUESTIONS 664 00:30:23,280 --> 00:30:25,160 COME UP AT THE END, WE WILL REACH OUT 665 00:30:25,160 --> 00:30:26,560 AND ASK YOU THOSE QUESTIONS THEN. 666 00:30:26,560 --> 00:30:32,200 THANK YOU VERY MUCH FOR THAT GREAT TALK. 667 00:30:32,200 --> 00:30:33,240 >> THANK YOU. 668 00:30:33,240 --> 00:30:38,920 >> OKAY, OUR NEXT PRESENTATION IS FROM 669 00:30:38,920 --> 00:30:48,880 THE IBM DATA SCIENCE AND AI ELITE TEAM. 670 00:30:48,880 --> 00:30:50,440 MR. ANISH [INDISCERNIBLE] WILL BE THE 671 00:30:50,440 --> 00:30:54,160 PRESENTER FOR THIS TEAM AND THE TITLE OF 672 00:30:54,160 --> 00:30:57,480 HIS PRESENTATION IS TRACKING CHANGES IN 673 00:30:57,480 --> 00:31:01,040 HEALTH METRICS BETWEEN VISITS TO MODEL 674 00:31:01,040 --> 00:31:05,440 ADVERSE PREGNANCY OUTCOME AMONG NULLIP A 675 00:31:05,440 --> 00:31:06,040 ROUS WOMEN. 676 00:31:06,040 --> 00:31:08,840 I'M SORRY, I MESSED YOUR NAME UP. 677 00:31:08,840 --> 00:31:11,280 >> DON'T WORRY ABOUT IT. 678 00:31:11,280 --> 00:31:11,560 [LAUGHTER] 679 00:31:11,560 --> 00:31:11,920 >> THANK YOU. 680 00:31:11,920 --> 00:31:15,680 >> I CAN TAKE IT FROM HERE, HELLO 681 00:31:15,680 --> 00:31:18,880 EVERYONE I'M AINESH P A NDEY, AND I 682 00:31:18,880 --> 00:31:23,040 SERVE AS THE IBM LEAD DATA TECHNICIAN 683 00:31:23,040 --> 00:31:27,480 FOR THIS CHALLENGE, I WROTE THE SOLUTION 684 00:31:27,480 --> 00:31:29,080 FOR GABRIEL [INDISCERNIBLE] AND 685 00:31:29,080 --> 00:31:33,240 [INDISCERNIBLE] WERE ALSO ON THE PROGRAM 686 00:31:33,240 --> 00:31:39,480 AS WELL. 687 00:31:39,480 --> 00:31:40,920 WE DECIDED TO--ESSENTIALLY WE BELIEVE 688 00:31:40,920 --> 00:31:42,040 THAT RELATIVE CHANGES IN MOZ 689 00:31:42,040 --> 00:31:44,000 MEASUREMENTS AND NOT THE RAW 690 00:31:44,000 --> 00:31:45,200 MEASUREMENTS THEMSELVES COULD PREDICT 691 00:31:45,200 --> 00:31:47,960 THE DEVELOPMENT OF APOsALATENER 692 00:31:47,960 --> 00:31:48,400 PREGNANCY. 693 00:31:48,400 --> 00:31:51,080 BECAUSE OF THE INPUTS TO OUR APPROACH 694 00:31:51,080 --> 00:31:52,240 WERE PRIMARILY ENGINEERING FEATURES WE 695 00:31:52,240 --> 00:31:54,400 HAD TO BE CAREFUL IN OUR INTERPRETATION, 696 00:31:54,400 --> 00:31:58,520 WE DECIDED TO CALCULATE DELTAS BETWEEN 697 00:31:58,520 --> 00:32:01,040 Z-SCORED MEASUREMENTS SO THE DELTA 698 00:32:01,040 --> 00:32:02,120 REPRESENT CHANGE BISE SPLIEWKS 699 00:32:02,120 --> 00:32:04,040 POPULATION FOR EXAMPLE, AN INCREASE IN 700 00:32:04,040 --> 00:32:06,080 WEIGHT OF 20-POUNDS DURING PREGNANCY, 701 00:32:06,080 --> 00:32:08,600 MEANS MUCH MORE FOR SOMEONE STARTING AT 702 00:32:08,600 --> 00:32:09,840 A LOWER BASE WEIGHT THAN SOMEONE 703 00:32:09,840 --> 00:32:11,840 STARTING AT A HIGHER BASE WEIGHT. 704 00:32:11,840 --> 00:32:14,400 THE DATA SET TOOK THESE CLINICAL 705 00:32:14,400 --> 00:32:16,360 MEASUREMENTS AT 4 STANDARD DOCTORS 706 00:32:16,360 --> 00:32:18,640 VISITS AS OUTLINED ON THE SLIDE. 707 00:32:18,640 --> 00:32:22,200 WE TOOK DELTAS BETWEEN VISITS 1 AND 2, 2 708 00:32:22,200 --> 00:32:24,960 AND 3, AND 1 AND 3. 709 00:32:24,960 --> 00:32:27,240 BECAUSE WE'RE INTERESTED IN IDENTIFYING 710 00:32:27,240 --> 00:32:28,640 APOs EARLY EARLY IN THE PREGNANCY AND 711 00:32:28,640 --> 00:32:31,120 NOT AT TERM, WE DID NOT GENERATE DALTON 712 00:32:31,120 --> 00:32:34,680 WITHERSPOONAS WITH THE FINAL VISIT. 713 00:32:34,680 --> 00:32:35,600 NEXT SLIDE. 714 00:32:35,600 --> 00:32:37,800 THERE ARE 3 CATEGORIES OF DATA WE CARED 715 00:32:37,800 --> 00:32:38,880 ABOUT FOR OUR ANALYSIS. 716 00:32:38,880 --> 00:32:41,640 I ALREADY TALKED ABOUT THE DELTAS WHICH 717 00:32:41,640 --> 00:32:44,040 INCLUDED CLINICAL MEASUREMENTS, SLEEP 718 00:32:44,040 --> 00:32:45,800 MONITORING DATA, INFORMATION ABOUT FETAL 719 00:32:45,800 --> 00:32:47,520 DEVELOPMENT, ET CETERA, THE SECOND 720 00:32:47,520 --> 00:32:51,120 CATEGORY THAT WE NEED TO INCLUDE IN OUR 721 00:32:51,120 --> 00:32:53,840 ANALYSIS WAS CO VARIANTS AND STATISTICAL 722 00:32:53,840 --> 00:32:55,440 MODELS LIKE THE 1S WE CREATE IT IS 723 00:32:55,440 --> 00:32:56,960 IMPORTANT TO CORRECT FOR CERTAIN 724 00:32:56,960 --> 00:32:57,240 VARIABLES. 725 00:32:57,240 --> 00:32:58,760 THE CO VALID AND RELIABLEIATES WITHIN 726 00:32:58,760 --> 00:33:02,440 OUR ANALYSIS WERE THE EGA, SINCE EACH 727 00:33:02,440 --> 00:33:07,520 VISIT FELL IN A RANGE OF LENGTH OF 728 00:33:07,520 --> 00:33:07,800 PREGNANCY. 729 00:33:07,800 --> 00:33:08,720 DEMOGRAPHIC DATA, HEALTH RELATED DAILY 730 00:33:08,720 --> 00:33:10,840 BASIS THEA AND SOCIOECONOMIC DATA. 731 00:33:10,840 --> 00:33:13,640 THE LAST CATEGORY OF COURSE WAS TARGET 732 00:33:13,640 --> 00:33:14,200 VARIABLES FOR ANALYSIS. 733 00:33:14,200 --> 00:33:16,880 WE WANTED TO SEE HOW ACCURATELY, WE 734 00:33:16,880 --> 00:33:19,000 COULD PREDICT PREGNANCY OUTCOMES AND THE 735 00:33:19,000 --> 00:33:21,200 DEVELOPMENT OF POSTPARTUM COMPLICATIONS 736 00:33:21,200 --> 00:33:24,160 OR MENTAL HEALTH CONDITIONS. 737 00:33:24,160 --> 00:33:27,760 FOR THIS ANALYSIS WE SELECTED DIFFERENT 738 00:33:27,760 --> 00:33:30,080 APOs AS TARGETS. 739 00:33:30,080 --> 00:33:30,480 NEXT SLIDE. 740 00:33:30,480 --> 00:33:33,640 NOW THAT WE HAVE THE MODELING DATA SET 741 00:33:33,640 --> 00:33:34,600 DEFINED, WHAT WITH ALL THE DAILY BASIS 742 00:33:34,600 --> 00:33:37,040 THEA CLEANING AND FEATURED ENGINEERING 743 00:33:37,040 --> 00:33:39,040 TO ASCERTAIN DATA QUALITY, WE CAN 744 00:33:39,040 --> 00:33:41,240 DISCUSS THE MODELING METHODOLOGIES THAT 745 00:33:41,240 --> 00:33:56,600 WE INPUT. 746 00:33:56,600 --> 00:33:56,880 NEXT SLIDE. 747 00:33:56,880 --> 00:34:01,320 --THROUGH ALL OF THAT I LEARNED THE 748 00:34:01,320 --> 00:34:04,600 MEDICAL RESEARCH COMMUNITY RELIES 749 00:34:04,600 --> 00:34:05,840 HEAVILY ON PROGRESSION TECH NEEGS 750 00:34:05,840 --> 00:34:10,360 BECAUSE THEY'RE EASY TO TRAIN, SIMPLE 751 00:34:10,360 --> 00:34:11,200 AND INTERPRETABLE, HOWEVER OUR 752 00:34:11,200 --> 00:34:12,440 CONTENTION IS THAT THERE'S MUCH MORE 753 00:34:12,440 --> 00:34:13,920 VALUE IN THE WORLD OF MACHINE LEARNING 754 00:34:13,920 --> 00:34:17,040 FELT OTHER METHODS WHICH ARE ONSEMBLING 755 00:34:17,040 --> 00:34:20,720 OR BOOSTING METHODS ARE BETTER SUITED TO 756 00:34:20,720 --> 00:34:25,120 MODEL COMPLEX RELATIONSHIPS AND OFTEN 757 00:34:25,120 --> 00:34:25,760 POWER BETTER. 758 00:34:25,760 --> 00:34:33,640 NEXT SLIDE. 759 00:34:33,640 --> 00:34:36,600 ENSEMBLE METHODS TAKE BOTH MODELS, A 760 00:34:36,600 --> 00:34:38,480 BOOT STRAP SAMPLE OF THE DATA FOR EACH 761 00:34:38,480 --> 00:34:40,000 MODEL AND ALLOWING IT TO SPECIALIZE 762 00:34:40,000 --> 00:34:42,080 DIFFERENT PARTS OF THE FEATURE SPACE, IN 763 00:34:42,080 --> 00:34:47,680 FACTIS WE SEE THAT ENSEMBLE MODELS 764 00:34:47,680 --> 00:34:51,360 GENERALLY OUTPERFORM SINGLE MODELS. 765 00:34:51,360 --> 00:34:52,040 NEXT SLIDE. 766 00:34:52,040 --> 00:34:56,600 BOOSTING TAKES A SPECIALIZATION THAT 767 00:34:56,600 --> 00:34:58,400 ENSEMBLE BRINGS FOR A WHOLE NEW MODEL 768 00:34:58,400 --> 00:35:03,240 AND IT RELYS ON LEARNERS FOR LARGE SWATH 769 00:35:03,240 --> 00:35:06,400 OF THE INPUT BASE, EACH LEARNER CAN BE 770 00:35:06,400 --> 00:35:08,680 ON THE FEATURE SPACE AT THE PREVIOUS 771 00:35:08,680 --> 00:35:10,600 LEARNER THAT WE PERFORM POORLY ON. 772 00:35:10,600 --> 00:35:13,920 FOR EACH OF OUR 18 TARGETS WE TRAINED TO 773 00:35:13,920 --> 00:35:16,880 THE LOGISTIC PROGRESSION MODEL, RANDOM 774 00:35:16,880 --> 00:35:19,440 MODEL AND LIGHT GBM MODEL TO SEE WHICH 775 00:35:19,440 --> 00:35:21,440 PERFORMED THE BEST. 776 00:35:21,440 --> 00:35:24,800 WE ELIMINATED 14 OF THE 18 APOs FOR 777 00:35:24,800 --> 00:35:30,840 LOW OR NEGATIVE CASES OR UNDER 778 00:35:30,840 --> 00:35:32,960 PERFORMING MODELS AND THE LIGHT GBM 779 00:35:32,960 --> 00:35:34,640 MODELS SHOWED THE BEST RESULTS. 780 00:35:34,640 --> 00:35:34,960 NEXT SLIDE. 781 00:35:34,960 --> 00:35:37,680 NOW WE WILL MOVE ON TO THE RESULTS OF 782 00:35:37,680 --> 00:35:38,480 OUR ANALYSIS. 783 00:35:38,480 --> 00:35:40,560 THE MAIN CHARTER OF THIS CHALLENGE WAS 784 00:35:40,560 --> 00:35:42,680 TO IDENTIFY AREAS OF RESEARCH INTO 785 00:35:42,680 --> 00:35:45,640 APOs WHILE ADDRESSING RACIAL 786 00:35:45,640 --> 00:35:45,920 DISPARITIES. 787 00:35:45,920 --> 00:35:47,680 HOWEVER, WE THINK THAT OUR ANALYSIS ALSO 788 00:35:47,680 --> 00:35:50,520 PAVES THE WAY FOR MORE INDEED SOLUTION 789 00:35:50,520 --> 00:35:53,800 THAT CAN IMPROVE THE LIVES OF PREGNANT 790 00:35:53,800 --> 00:35:55,600 WOMEN TODAY. 791 00:35:55,600 --> 00:35:56,800 NEXT SLIDE. 792 00:35:56,800 --> 00:35:59,800 TO GUIDE FUTURE RESEARCH INTO MATERNAL 793 00:35:59,800 --> 00:36:00,280 MORBIDITY. 794 00:36:00,280 --> 00:36:01,760 THE MOST IMPORTANT RESULT OF OUR 795 00:36:01,760 --> 00:36:03,240 ANALYSIS IS THAT FEATURED IMPORTANCES IS 796 00:36:03,240 --> 00:36:06,240 ASSOCIATED WITH EACH OF OUR MODELS, EACH 797 00:36:06,240 --> 00:36:07,720 HIGHLY IMPACTFUL FEATURE POTENTIALLY 798 00:36:07,720 --> 00:36:10,080 REPRESENTS A PATH FOR FUTURE MEDICAL 799 00:36:10,080 --> 00:36:10,360 RESEARCH. 800 00:36:10,360 --> 00:36:12,560 FOR EXAMPLE, WE FIND THAT THE CHANGE CAN 801 00:36:12,560 --> 00:36:16,360 WAIT BETWEEN VISIT IT IS 1, 2, AND 3, IS 802 00:36:16,360 --> 00:36:18,800 HIGHLY PREDICTIVE OF BOTH POART PART 803 00:36:18,800 --> 00:36:20,040 UMKC DEPRESSION AND ANXIOUS. 804 00:36:20,040 --> 00:36:22,840 THIS INDICATES THAT A BLINDED STUDY 805 00:36:22,840 --> 00:36:24,480 RELATING WEIGHT GAIN THROUGH 806 00:36:24,480 --> 00:36:27,360 PROGRESSINANCE SETHE DEVELOPMENT OF 807 00:36:27,360 --> 00:36:29,840 THESE POSTPARTUM MENTAL ILLNESSES IS 808 00:36:29,840 --> 00:36:31,120 LIKELY PRODUCE A STATISTICALLY 809 00:36:31,120 --> 00:36:32,840 SIGNIFICANT RELATIONSHIP BETWEEN THE 2. 810 00:36:32,840 --> 00:36:34,480 SIMILARLY, THE RESULTS OF OUR ANALYSIS 811 00:36:34,480 --> 00:36:38,000 SHOULD PROVIDE THE BASIS OF SEVERAL 812 00:36:38,000 --> 00:36:42,960 OTHERS EXPLAINED MEDICAL STUDIES. 813 00:36:42,960 --> 00:36:43,240 NEXT SLIDE. 814 00:36:43,240 --> 00:36:44,960 ANOTHER CHARTER OF THIS CHALLENGE WAS TO 815 00:36:44,960 --> 00:36:49,720 ADDRESS HOW IT SPECIFICALLY IMPACTS 816 00:36:49,720 --> 00:36:51,120 FACTIONS OF PREGNANT POPULATION, 817 00:36:51,120 --> 00:36:53,000 THEREFORE FOR EACH IMPACTFUL FEATURE 818 00:36:53,000 --> 00:36:54,800 IDENTIFIED BY OUR FEATURE IMPORTANCES WE 819 00:36:54,800 --> 00:36:59,080 WANTED TO SEE HOW THE DISTRIBUTION CROSS 820 00:36:59,080 --> 00:37:01,960 RACE, THERE'S A CHART MISSING FROM THIS 821 00:37:01,960 --> 00:37:03,840 SLIDE BUT IN THE CHART WE WOULD SHOW 822 00:37:03,840 --> 00:37:05,720 THAT THE RESTING DIASTOOL DNAIC BLOOD 823 00:37:05,720 --> 00:37:07,880 PRESSURE BETWEEN VISIT 2 AND 3, IS 824 00:37:07,880 --> 00:37:09,560 PREDICTIVE OF DEVELOPMENT OF CHRONIC 825 00:37:09,560 --> 00:37:10,840 HYPER TENSION. 826 00:37:10,840 --> 00:37:13,040 THE MAJORITY CLASS, WHICH IS WHITE 827 00:37:13,040 --> 00:37:15,920 FEMALES, SHOWS BEHAVIOR CLOSER TO ME, 828 00:37:15,920 --> 00:37:18,160 HOWEVER SOME RACIAL MINORITIES DISPLAY 829 00:37:18,160 --> 00:37:20,640 GENERAL BEHAVIOR, VERY DIFFERENT FROM 830 00:37:20,640 --> 00:37:21,640 THE NORM. 831 00:37:21,640 --> 00:37:23,760 IN THIS CASE, AMERICAN INDIAN AND ALASKA 832 00:37:23,760 --> 00:37:26,320 NATIVE SHOW A GENERAL--SHOWED A GENERAL 833 00:37:26,320 --> 00:37:28,640 DOWNWARD TREND IN RESTING DIASTOOL DNAIC 834 00:37:28,640 --> 00:37:30,160 BLOOD PRESSURE THROUGHOUT THEIR 835 00:37:30,160 --> 00:37:32,240 PREGNANCY WHILE NATIVE HAWAIIAN AND 836 00:37:32,240 --> 00:37:35,000 OTHER PACIFIC ISLANDERS SHOWED THE 837 00:37:35,000 --> 00:37:35,440 OPPOSITE. 838 00:37:35,440 --> 00:37:38,480 THESE DIFFERENCES IN RACIAL BEHAVIOR CAN 839 00:37:38,480 --> 00:37:42,240 PROVIDE A LENS IN THOSE BLINDED STUDIES 840 00:37:42,240 --> 00:37:43,640 MOVING FORWARD, ALLOWINGITOUS UNDERSTAND 841 00:37:43,640 --> 00:37:46,120 NOT ONLY HOW CERTAIN PREGNANCIES PREDICT 842 00:37:46,120 --> 00:37:48,320 APOs BUT HOW DIFFERENT FACTIONS OF THE 843 00:37:48,320 --> 00:37:51,040 FEMALE POPULATION MAY BE MORE OR LESS 844 00:37:51,040 --> 00:37:51,960 ADVERSELY AFFECTED. 845 00:37:51,960 --> 00:37:54,280 THIS LIST REPRESENTS THE MOST CRITICAL 846 00:37:54,280 --> 00:37:57,200 FINDINGS FROM OUR ANALYSIS, WHO 847 00:37:57,200 --> 00:37:58,600 IDENTIFIED DELTA FEATURES, WHICH 848 00:37:58,600 --> 00:38:01,680 MORBIDITYS THEY IMPACT AND HOW THE 849 00:38:01,680 --> 00:38:06,000 MAJORITY BEHAVES AND WHICH ARE ADVERSELY 850 00:38:06,000 --> 00:38:07,120 AFFECTED. 851 00:38:07,120 --> 00:38:09,920 EACH ROW REPRESENTS A POTENTIAL BLINDED 852 00:38:09,920 --> 00:38:11,360 STUDY TO ASCERTAIN OUR FINDINGS. 853 00:38:11,360 --> 00:38:13,560 NEXT SLIDE. 854 00:38:13,560 --> 00:38:16,280 THE MOST EXCITING ASPECT OF OUR ANALYSIS 855 00:38:16,280 --> 00:38:17,440 HOWEVER, GOES BEYOND THE CHARTER OF THIS 856 00:38:17,440 --> 00:38:18,760 CHALLENGE, WE BELIEVE THAT OUR WORK 857 00:38:18,760 --> 00:38:22,000 OPENS THE PATH TO A MORE IMMEDIATE 858 00:38:22,000 --> 00:38:25,400 SOLUTION THAT CAN START IMPROVING THE 859 00:38:25,400 --> 00:38:28,720 LIVES OF PREGNANT WOMEN NOW INSTUD OF 860 00:38:28,720 --> 00:38:29,880 THROUGH FUTURE RESEARCH. 861 00:38:29,880 --> 00:38:31,680 THE HIGH PERFORMANCE OF OUR MODELS 862 00:38:31,680 --> 00:38:33,360 INDICATES WE CAN DEVELOP REALTIME TOOL 863 00:38:33,360 --> 00:38:36,040 FOR DOCTORS TO USE IN THE EXAM NATION 864 00:38:36,040 --> 00:38:37,680 ROOM TO PROACTIVELY IDENTIFY MORBIDITY 865 00:38:37,680 --> 00:38:38,040 RISK. 866 00:38:38,040 --> 00:38:40,000 ONCE THE SCREENING DATA IS COLLECTED OUR 867 00:38:40,000 --> 00:38:42,680 TOOL CAN PRODUCE A RISK PROFILE WITHIN 868 00:38:42,680 --> 00:38:45,080 SECONDS WARNING TD DOCTOR OF ANY APO 869 00:38:45,080 --> 00:38:48,960 THAT THE PATIENT MAY SHOW SIGNS OF 870 00:38:48,960 --> 00:38:49,240 DEVELOPING. 871 00:38:49,240 --> 00:38:51,280 IN REALTIME, DOCTORS WILL BE ABLE TO 872 00:38:51,280 --> 00:38:53,880 VERIFY USING THEIR OWN EXPERTISE AND 873 00:38:53,880 --> 00:38:54,840 RECOMMEND ALLEVIATING MEASURES BEFORE 874 00:38:54,840 --> 00:38:57,040 THE PATIENT EVEN LEAVES THE HOSPITAL. 875 00:38:57,040 --> 00:38:58,800 FOR EXAMPLE, IF THE RISK PROFILE 876 00:38:58,800 --> 00:39:02,040 INDICATES A HIGH CHANCE OF DEVELOPING 877 00:39:02,040 --> 00:39:03,800 POSTPARTUM DEPRESSION, THE DOCTOR MAY 878 00:39:03,800 --> 00:39:05,720 FOCUS ON PARTS OF THE EXAM NATION 879 00:39:05,720 --> 00:39:07,600 RELATED TO MENTAL HEALTH AND DETERMINE 880 00:39:07,600 --> 00:39:10,120 THE VALIDITY OF THE RISK, IF THE DOCTOR 881 00:39:10,120 --> 00:39:11,920 CONCURRINGS, THEY CAN ADDRESS THIS 882 00:39:11,920 --> 00:39:14,040 CONCERN BEFORE IT ACTUALLY DEVELOPS. 883 00:39:14,040 --> 00:39:15,720 SUCH PREVENTATIVE CARE MAY GO A LONG WAY 884 00:39:15,720 --> 00:39:18,800 TO IMPROVE THE HEALTH OF PERSPECTIVE 885 00:39:18,800 --> 00:39:21,720 MOTHERS AND THEIR CHILDREN. 886 00:39:21,720 --> 00:39:22,760 NEXT SLIDE. 887 00:39:22,760 --> 00:39:24,960 WE ARE ALL VERY EXCITED ABOUT THE WERE 888 00:39:24,960 --> 00:39:31,000 AND REAL WORLD APPLICATIONS OF OUR 889 00:39:31,000 --> 00:39:31,240 ANALYSIS. 890 00:39:31,240 --> 00:39:39,120 YOU CAN FOLLOW THE LINKS ON THE SLIDE TO 891 00:39:39,120 --> 00:39:43,200 FOLLOW ALONG, AND I'VE ALSO PUBLISHED A 892 00:39:43,200 --> 00:39:46,080 BLOG, AND YOU CAN FOLLOW THAT LAST LINK 893 00:39:46,080 --> 00:39:48,120 AND MESSAGE ME DIRECTLY IF YOU WOULD 894 00:39:48,120 --> 00:39:49,240 LIKE TO WORK ON THE TEAM. 895 00:39:49,240 --> 00:39:53,440 THANK YOU? 896 00:39:53,440 --> 00:39:54,280 >> THANK YOU. 897 00:39:54,280 --> 00:39:56,600 REALLY APPRECIATE THIS INFORMATIVE 898 00:39:56,600 --> 00:39:57,080 PRESENTATION. 899 00:39:57,080 --> 00:40:03,640 HAVE 2 QUESTIONS FOR YOU FROM THE CHAT. 900 00:40:03,640 --> 00:40:06,040 THE FIRST QUESTION SAYS YOU MENTION THE 901 00:40:06,040 --> 00:40:07,320 POTENTIAL RELATIONSHIP BETWEEN WEIGHT 902 00:40:07,320 --> 00:40:11,840 CHANGE OR CHANGE IN BMI AND RISK OF 903 00:40:11,840 --> 00:40:13,920 POSTPARTUM DEPRESSION, HOW DO YOU 904 00:40:13,920 --> 00:40:16,600 RECONCILE THIS WITH THE FACT THAT WEIGHT 905 00:40:16,600 --> 00:40:18,960 GAIN IS NORMAL, UNAVOIDABLE IN 906 00:40:18,960 --> 00:40:20,360 PREGNANCY? 907 00:40:20,360 --> 00:40:20,920 >> RIGHT. 908 00:40:20,920 --> 00:40:22,600 THAT IS UNAVOIDABLE, IT IS A FACT OF 909 00:40:22,600 --> 00:40:25,640 PREGNANCY OF COURSE, THE PURPOSE OF OUR 910 00:40:25,640 --> 00:40:28,480 ANALYSIS IS TO TURN A BLIND EYE TOWARD 911 00:40:28,480 --> 00:40:30,040 WHAT HUMAN BEINGS BELIEVE AND TO LET THE 912 00:40:30,040 --> 00:40:34,840 DATA SPEAK FOR ITSELF, NOW WEIGHT GAIN 913 00:40:34,840 --> 00:40:37,120 IS NATURAL, AND AN EXPECTED PART OF 914 00:40:37,120 --> 00:40:39,800 PREGNANCY BUT THE FACT THAT IT IS ACTING 915 00:40:39,800 --> 00:40:41,440 AS YOUR MENTAL HEALTH AFTERWARD SYSTEM 916 00:40:41,440 --> 00:40:43,280 MAYBE A BIT OF AN OBVIOUS UNDERSTANDING 917 00:40:43,280 --> 00:40:44,360 BUT THERE ARE SEVERAL OTHER 918 00:40:44,360 --> 00:40:47,840 RELATIONSHIPS IN THE ANALYSIS AS WELL, 919 00:40:47,840 --> 00:40:52,720 BETWEEN SOME DATA ABOUT THE HEALTH OF 920 00:40:52,720 --> 00:40:54,440 THE FETAL--FETUS AND OTHER DEVELOPMENTS 921 00:40:54,440 --> 00:40:56,560 AS WELL, LATER ON AND WE THINK MIGHT BE 922 00:40:56,560 --> 00:40:57,080 MORE INDICATIVE. 923 00:40:57,080 --> 00:40:59,880 I JUST THOUGHT THAT THE RELATIONSHIP 924 00:40:59,880 --> 00:41:01,760 BETWEEN WEIGHT GAIN AND DEPRESSION WOULD 925 00:41:01,760 --> 00:41:03,840 BE MORE OBVIOUS AND EASY TO EXPLAIN BUT 926 00:41:03,840 --> 00:41:06,760 THERE ARE SEVERAL OTHERS THAT MIGHT BE 927 00:41:06,760 --> 00:41:10,360 MORE INTERESTING THAN THAT BASIC 1. 928 00:41:10,360 --> 00:41:13,720 >> OKAY. 929 00:41:13,720 --> 00:41:14,640 THANK YOU. 930 00:41:14,640 --> 00:41:16,560 EXCUSE ME, THERE'S 1 OTHER--ACTUALLY 2 931 00:41:16,560 --> 00:41:16,760 MORE. 932 00:41:16,760 --> 00:41:18,880 I THINK WE HAVE TIME FOR THESE. 933 00:41:18,880 --> 00:41:20,640 SECOND IS PRIOR AWARENESS OF MORBIDITY 934 00:41:20,640 --> 00:41:25,840 RISK IS GOOD BUT MAY LEAD TO STIGMA FOR 935 00:41:25,840 --> 00:41:28,360 PREGNANT WOMEN ESPECIALLY IF SEEING 936 00:41:28,360 --> 00:41:31,440 PROVIDERS WHO ARE NOT SUPPORTIVE OR WITH 937 00:41:31,440 --> 00:41:32,160 ACCESS TO LIMITED CARE. 938 00:41:32,160 --> 00:41:33,880 THIS IS A COMMENT IS BASICALLY ASKING IF 939 00:41:33,880 --> 00:41:36,480 HAVE YOU ANY THOUGHTS ABOUT THIS? 940 00:41:36,480 --> 00:41:39,040 >> YEAH THIS, IS A SYSTEMIC PROBLEM THAT 941 00:41:39,040 --> 00:41:40,760 EXISTS WHICH I DO UNDERSTAND THE REALITY 942 00:41:40,760 --> 00:41:43,520 OF WHAT THE PROBLEM WE'RE TRYING TO 943 00:41:43,520 --> 00:41:43,760 ADDRESS. 944 00:41:43,760 --> 00:41:48,760 OUR GOAL IS TO TRY TO IDENTIFY THE WAY 945 00:41:48,760 --> 00:41:51,240 WE CAN TRY TO CATCH THIS EARLY ON, USING 946 00:41:51,240 --> 00:41:55,200 DATA AND MAYBE PROVE THAT WITH SOME OF 947 00:41:55,200 --> 00:41:55,520 THESE STUDIES. 948 00:41:55,520 --> 00:41:56,880 THERE OF COURSE NEEDS TO BE WORK DONE TO 949 00:41:56,880 --> 00:41:59,520 FIX THE SYSTEM TO BE ABLE TO TAKE AND 950 00:41:59,520 --> 00:42:01,640 MAKE USE OF THESE FINDINGS BETTER AND IN 951 00:42:01,640 --> 00:42:07,080 A MORE EFFECTIVE WAY PRIMARILY FOR 952 00:42:07,080 --> 00:42:08,560 UNDERREPRESENTED WOMEN AND THAT'S GOING 953 00:42:08,560 --> 00:42:09,800 TO HAVE TO JUST BE SOMETHING THAT'S 954 00:42:09,800 --> 00:42:13,800 GOING TO HAVE TO HAPPEN. 955 00:42:13,800 --> 00:42:14,160 >> YEAH. 956 00:42:14,160 --> 00:42:20,840 OKAY, AND THE FINAL QUESTION I HAD FOR 957 00:42:20,840 --> 00:42:25,560 YOU, IT SAYS THANK YOU FOR A WONDERFUL 958 00:42:25,560 --> 00:42:27,240 PRESENTATION, AS DELTA IS RECALCULATED 959 00:42:27,240 --> 00:42:29,440 DID YOU LOOK INTO OTHER MODELS WHICH ARE 960 00:42:29,440 --> 00:42:31,920 BETTER WITH TIME SERIES DATA, FIRST 961 00:42:31,920 --> 00:42:38,200 VISIT, SECOND VISIT OR THIRD VISIT LIKE 962 00:42:38,200 --> 00:42:39,280 LSTM-NEUTRAL NETWORKS. 963 00:42:39,280 --> 00:42:40,720 >> NEURAL NETWORKS, YES, WE WERE 964 00:42:40,720 --> 00:42:41,240 CONSIDERING THAT. 965 00:42:41,240 --> 00:42:43,440 UNLIKE SOME OF THE TEAMS THAT CAME INTO 966 00:42:43,440 --> 00:42:46,760 THIS PROJECT WE ACTUALLY FOUND THIS 967 00:42:46,760 --> 00:42:48,840 PROJECT ON FREELANCER SO WE HAD THE 968 00:42:48,840 --> 00:42:50,760 LIMITED 2-3 MONTHS TO WORK WITH IT GIVEN 969 00:42:50,760 --> 00:42:52,720 THE TIME CONSTRAINTS WE DECIDED TO GO 970 00:42:52,720 --> 00:42:55,040 WITH THIS APPROACH BUT NEURAL NETWORKS 971 00:42:55,040 --> 00:42:56,240 WAS A CONSIDERATION FOR US. 972 00:42:56,240 --> 00:42:59,280 THE REASON WE MARKED IT OFF OR DECIDED 973 00:42:59,280 --> 00:43:01,640 NOT TO PURSUE THAT AVENUE IS THROUGH THE 974 00:43:01,640 --> 00:43:03,240 SHEER AMOUNT OF DATA AVAILABLE TO US 975 00:43:03,240 --> 00:43:03,640 WHICH IS LOW. 976 00:43:03,640 --> 00:43:09,640 WE ARE TRYING TO MODEL ON APOs AND A 977 00:43:09,640 --> 00:43:12,440 VERY GOOD ASPECT OF THE DATA AND VERY 978 00:43:12,440 --> 00:43:13,760 BAD MODELS PERSPECTIVES IS THAT THERE IS 979 00:43:13,760 --> 00:43:17,440 A LOW NUMBER OF APOs COMPARED TO OTHER 980 00:43:17,440 --> 00:43:19,960 FREG NANCYS THAT DID NOT DEAL WITH 981 00:43:19,960 --> 00:43:20,280 APOs, RIGHT? 982 00:43:20,280 --> 00:43:23,160 SO FOR THAT REASON, NEURAL NETS PERFORM 983 00:43:23,160 --> 00:43:25,240 AT BEST WHEN GIVEN A HIGH AMOUNT OF 984 00:43:25,240 --> 00:43:27,960 DATA, THERE'S A WELL KNOWN CHART IN THE 985 00:43:27,960 --> 00:43:29,720 MACHINE LEARNING WORLD WHERE OVERTIME, 986 00:43:29,720 --> 00:43:33,040 OVER THE AMOUNT OF DATA AVAILABLE TO US 987 00:43:33,040 --> 00:43:34,760 STANDARD MACHINE LEARNING MODELS 988 00:43:34,760 --> 00:43:36,240 OUTPERFORM UNTIL THE AMOUNT OF DATA 989 00:43:36,240 --> 00:43:38,800 INCREASES TO A SIZE WHERE NEURAL NETS 990 00:43:38,800 --> 00:43:40,040 BECOME MORE EFFECTIVE. 991 00:43:40,040 --> 00:43:42,120 THE AMOUNT OF DATA WE HAVE RIGHT NOW WAS 992 00:43:42,120 --> 00:43:52,080 NOT CONDUCIVE TO A GOOD NEURAL NET. 993 00:43:52,080 --> 00:43:53,320 >> THANK YOU AINESH, WE REALLY 994 00:43:53,320 --> 00:43:54,320 APPRECIATE THAT. 995 00:43:54,320 --> 00:43:54,760 GREAT PRESENTATION. 996 00:43:54,760 --> 00:43:56,880 ALL RIGHT, WE WILL KEEP MOVING AND OUR 997 00:43:56,880 --> 00:44:01,720 NEXT PRESENTER IS FROM THE DELFINI GROUP 998 00:44:01,720 --> 00:44:10,560 AND IT'S THE CHIEF TECHNOLOGY OFFICER 999 00:44:10,560 --> 00:44:11,640 DR. ALI EBRAHIM. 1000 00:44:11,640 --> 00:44:13,560 >> THANKS MAURICE FOR THE INTRODUCTION, 1001 00:44:13,560 --> 00:44:20,160 HI, I'M ALI AND I'M SUPER EXCITED TO BE 1002 00:44:20,160 --> 00:44:23,640 HERE WILL ITING ABOUT THE WORK WE DID AT 1003 00:44:23,640 --> 00:44:27,480 DELFINNA FOR THIS PROJECT CHALLENGE OURS 1004 00:44:27,480 --> 00:44:29,160 IS TITLED RANDOM FORESTS FOR ACCURATE 1005 00:44:29,160 --> 00:44:32,840 PREDICTION OF THE RISK OF HYPER TENSIVE 1006 00:44:32,840 --> 00:44:33,640 DISORDERS OF PREGNANCY AT TERM. 1007 00:44:33,640 --> 00:44:38,160 I AM SEE HAPPENING A LOT OF SIMILARITIES 1008 00:44:38,160 --> 00:44:41,600 WITH RANDOM FORESTS AND HYPER TENSION. 1009 00:44:41,600 --> 00:44:42,040 NEXT SLIDE PLEASE. 1010 00:44:42,040 --> 00:44:44,440 I THINK A BIG REASON HYPER TENSION WAS A 1011 00:44:44,440 --> 00:44:47,680 BIG FOCUS FOR ALL OF US IS THAT HYPER 1012 00:44:47,680 --> 00:44:49,200 TENSIVE DISORDERS AND PREGNANCY ARE 1013 00:44:49,200 --> 00:44:50,200 EXTREMELY COMMON HERE IN THE UNITED 1014 00:44:50,200 --> 00:44:52,640 STATES, THERE ARE STUDIES THAT SHOW THAT 1015 00:44:52,640 --> 00:44:57,120 IT COMPLICATES UP TO 10% OF PREGNANCIES. 1016 00:44:57,120 --> 00:45:02,120 HYPER TENSIVE DISORDERS CAN INCLUDE 1017 00:45:02,120 --> 00:45:03,920 GUESTATIONAL HYPER TENSION, CHRONIC 1018 00:45:03,920 --> 00:45:06,240 HYPER TENSION, PREELAMPSIA AND 1019 00:45:06,240 --> 00:45:10,560 ECLAMPSIA, AND THEY'RE ALSO ASSOCIATE 1020 00:45:10,560 --> 00:45:12,080 WIDE CARDIOVASCULAR DISORDERS AND HYPER 1021 00:45:12,080 --> 00:45:15,080 TENSIVE DISORDERS AND RESULT IN A LARGE 1022 00:45:15,080 --> 00:45:17,000 RISK FACTOR OF MATERNAL MORBIDITY. 1023 00:45:17,000 --> 00:45:17,440 NEXT SLIDE, PLEASE. 1024 00:45:17,440 --> 00:45:19,160 SO THIS IS SOMETHING WE REALLY WANT TO 1025 00:45:19,160 --> 00:45:21,480 FOCUS ON NOT ONLY BECAUSE IT'S SUCH A 1026 00:45:21,480 --> 00:45:23,800 LARGE PROBLEM IN THE UNITED STATES, 1027 00:45:23,800 --> 00:45:26,480 RELATIVE TO OTHER COUNTRIES AFFAIRS TEAM 1028 00:45:26,480 --> 00:45:28,040 LEADER A SIMILAR SOCIOECONOMIC DLEFUL 1029 00:45:28,040 --> 00:45:33,080 LEVEL OF DEVELOPMENT BUT ALSO THAT THIS 1030 00:45:33,080 --> 00:45:37,240 RISK IS SIGNIFICANTLY HIGHER FOR 1031 00:45:37,240 --> 00:45:39,120 INDIGENOUS AND ALSO BLACK PREGNANT 1032 00:45:39,120 --> 00:45:39,360 PEOPLE. 1033 00:45:39,360 --> 00:45:44,240 THIS IS A COMMONLY KNOWN PATTERN FOR 1034 00:45:44,240 --> 00:45:46,960 MANY RISKS OF MATERNAL MORBIDITY AND 1035 00:45:46,960 --> 00:45:49,080 HYPER TENSION ALSO SHOWS THE SAME 1036 00:45:49,080 --> 00:45:49,320 PATTERN. 1037 00:45:49,320 --> 00:45:51,280 SO THIS IS SOMETHING WE REALLY WANT TO 1038 00:45:51,280 --> 00:45:52,560 FOCUS ON. 1039 00:45:52,560 --> 00:45:55,640 THE SOCIAL DETERMINANTS OF HEALTH CAN 1040 00:45:55,640 --> 00:45:58,680 REALLY NOT BE IGNORED WHEN ADDRESSING 1041 00:45:58,680 --> 00:46:01,920 THESE DISPARITIES AND ANALYZING THEM. 1042 00:46:01,920 --> 00:46:08,240 AND WE BELIEVE THAT ANY ANALYSIS, OF 1043 00:46:08,240 --> 00:46:09,880 CONFOUNDING VARIABLES SUCH AS SYSTEMIC 1044 00:46:09,880 --> 00:46:13,240 RACISM THAT ARE PRESENT IN DELIVERY OF 1045 00:46:13,240 --> 00:46:15,760 HEALTHCARE AND ALSO OTHER FACTORS THAT 1046 00:46:15,760 --> 00:46:20,960 ACCOUNT AFFECT HEALTH OUTCOMES. 1047 00:46:20,960 --> 00:46:21,400 NEXT SLIDE, PLEASE. 1048 00:46:21,400 --> 00:46:24,600 SO THIS S&P SOMETHING THAT WE'RE REALLY 1049 00:46:24,600 --> 00:46:27,320 INTERESTED IN IS PREDICTION OF HYPER 1050 00:46:27,320 --> 00:46:28,240 TENSIVE RISK. 1051 00:46:28,240 --> 00:46:29,640 SO THE CURRENT STRATEGY THAT MOST 1052 00:46:29,640 --> 00:46:32,320 DOCTORS, A LOT OF THE DOCTORS THAT WE 1053 00:46:32,320 --> 00:46:34,320 TALK TO IS THEY USE THEIR CLINICAL 1054 00:46:34,320 --> 00:46:36,320 JUDGMENT TO ASSESS MATERNAL HISTORY AND 1055 00:46:36,320 --> 00:46:40,520 RISK FACTORS AND THEN FROM THAT EACH 1056 00:46:40,520 --> 00:46:42,680 DOCTOR WILL BASICALLY DETERMINE THE RISK 1057 00:46:42,680 --> 00:46:45,520 OF HIGH RISK PATIENTS, THEY TIKE A LOT 1058 00:46:45,520 --> 00:46:47,680 OF FACTORS INTO THEIR ACCOUNT AS THEY 1059 00:46:47,680 --> 00:46:48,120 ASSESS THIS. 1060 00:46:48,120 --> 00:46:51,440 THERE ARE GUIDELINES THAT ARE PUBLISHED, 1061 00:46:51,440 --> 00:47:00,320 AND ALSO THE DOCTORS ARE AWARE OF THE 1062 00:47:00,320 --> 00:47:01,600 INDIVIDUAL POPULATION AND WHERE THEY 1063 00:47:01,600 --> 00:47:03,520 SERVE, SO THERE ARE SOME STATES WHERE 1064 00:47:03,520 --> 00:47:06,240 HYPER TENSIVE RISKS CAN BE UP TO 40% OF 1065 00:47:06,240 --> 00:47:07,960 ALL PREGNANCIES MUCH HIGHER THAN 10% 1066 00:47:07,960 --> 00:47:08,920 ACROSS THE COUNTRY. 1067 00:47:08,920 --> 00:47:11,120 SO THERE IS SOME ROOM THAT DOCTOR SAYS 1068 00:47:11,120 --> 00:47:13,440 HAVE TO ADJUST TO HANDLE THE SPECIFIC 1069 00:47:13,440 --> 00:47:17,360 HEALTH NEEDS OF THEIR POPULATION. 1070 00:47:17,360 --> 00:47:20,640 WE BELIEVE THAT IMPROVED PREDICTION 1071 00:47:20,640 --> 00:47:22,440 STRATEGY CAN POTENTIALLY AID DOCTORS AS 1072 00:47:22,440 --> 00:47:24,160 THEY MAKE THESE RISK ASSESSMENTS AND 1073 00:47:24,160 --> 00:47:26,240 PROVIDE CARE TO THEIR PATIENTS AND IN 1074 00:47:26,240 --> 00:47:28,720 THIS WAYS WAY, WE THINK IT WILL ADD TO 1075 00:47:28,720 --> 00:47:31,680 THE CARE, SO A WEALTH OF INFORMATION IN 1076 00:47:31,680 --> 00:47:32,880 EHR, INCLUDES VITALS, LABS AND 1077 00:47:32,880 --> 00:47:36,000 MEDICATIONS THAT CAN DO A MORE 1078 00:47:36,000 --> 00:47:37,320 STATISTICALLY SOUND APPROACH TO AID THE 1079 00:47:37,320 --> 00:47:41,320 DOCTOR AS THEY THEN FURTHER ASSESS RISK. 1080 00:47:41,320 --> 00:47:44,920 WE BELIEVE THAT THIS WILL ALLOW 1081 00:47:44,920 --> 00:47:46,240 INCREASED SPECIFICITY AND SENSITIVITY OF 1082 00:47:46,240 --> 00:47:47,880 A CLINICIAN USING A MODEL LIKE THIS AND 1083 00:47:47,880 --> 00:47:50,040 WE BELIEVE IT CAN ALSO IPT GREATER 1084 00:47:50,040 --> 00:47:51,320 IMPROVE CLINICAL MANAGEMENT BY 1085 00:47:51,320 --> 00:47:53,400 PERSONALIZING CARE AND BEGINNING 1086 00:47:53,400 --> 00:47:57,200 MITIGATIONS EARLY, SO, FOR EXAMPLE, A 1087 00:47:57,200 --> 00:47:59,440 PREGNANT PERSON WITH A RISK OF HYPER 1088 00:47:59,440 --> 00:48:01,240 TENSIVE DISORDERS CAN BEGIN MONITORING 1089 00:48:01,240 --> 00:48:03,240 BLOOD PRESSURE EARLIER SO IF THEY DO FET 1090 00:48:03,240 --> 00:48:05,280 REALLY HIGH BLBS THEY GET MUCH MORE 1091 00:48:05,280 --> 00:48:07,960 QUICKLY, END UP HAVING THOSE ADDRESSED 1092 00:48:07,960 --> 00:48:10,040 BY THEIR PHYSICIAN, AND COME UP WITH A 1093 00:48:10,040 --> 00:48:12,480 BETTER STRATEGY AS OPPOSE TO IT COMING 1094 00:48:12,480 --> 00:48:14,600 OUT MAYBE MONTHS AFTERWARDS WHEN THEY 1095 00:48:14,600 --> 00:48:17,640 HAVE A HIGH BLOOD PRESSURE MEASURED AT 1096 00:48:17,640 --> 00:48:18,040 THE DOCTOR'S OFFICE. 1097 00:48:18,040 --> 00:48:20,680 ALSO SOMEONE WITH A HIGH RISK OF HYPER 1098 00:48:20,680 --> 00:48:22,200 TENSIVE DISORDERS CAN START DOG 1099 00:48:22,200 --> 00:48:23,720 SOMETHING LIKE TAKING ASPIRIN WHICH IS 1100 00:48:23,720 --> 00:48:27,040 GENERALLY LOW RISK BUT ALSO COULD 1101 00:48:27,040 --> 00:48:30,560 MITIGATE SOME STUDIES THEN CAN MITIGATE 1102 00:48:30,560 --> 00:48:33,920 THE DISORDERS AND THE DOCTOR CAN AT 1103 00:48:33,920 --> 00:48:36,120 THEIR DISCRETION DECIDE IF THIS IS A 1104 00:48:36,120 --> 00:48:39,360 NEED OR NOT BASED ON THE PATIENT'S RISK. 1105 00:48:39,360 --> 00:48:39,880 SO NEXT SLIDE. 1106 00:48:39,880 --> 00:48:42,040 THIS IS A LOT OF OTHER PRESENTERS TALKED 1107 00:48:42,040 --> 00:48:43,840 ABOUT HYPER TENSION AND GIIVE A LOT OF 1108 00:48:43,840 --> 00:48:45,240 GREAT INFORMATION SO I WANT TO JUMP 1109 00:48:45,240 --> 00:48:47,240 RIGHT INTO WHAT WE'VE DONE SO FAR. 1110 00:48:47,240 --> 00:48:55,280 SO WE USE THIS NULLIP A ROUS DATA SET, 1111 00:48:55,280 --> 00:48:57,840 IT WAS SUCH A WONDERFUL DATA SET, 1112 00:48:57,840 --> 00:48:59,960 PLEASURE TO WORK WITH, THE DATA QUALITY 1113 00:48:59,960 --> 00:49:02,160 WAS HIGH AND THERE WAS A LARGE NUMBER OF 1114 00:49:02,160 --> 00:49:03,680 VARIABLES COLLECTED AND IT MADE IT 1115 00:49:03,680 --> 00:49:04,880 REALLY INTERESTING FOR OUR TEAM AND 1116 00:49:04,880 --> 00:49:07,680 EVERYBODY WANTED A PIECE OF THIS DATA 1117 00:49:07,680 --> 00:49:09,920 ANALYSIS PIE BUT IT WAS ONLY THE LIMITED 1118 00:49:09,920 --> 00:49:13,640 NUMBER OF PEOPLE ON OUR TEAM WERE ABLE 1119 00:49:13,640 --> 00:49:14,040 TO DO IT. 1120 00:49:14,040 --> 00:49:16,840 WHAT WE DID IS WE CREATED MACHINE 1121 00:49:16,840 --> 00:49:19,000 LEARNING MODELS TO ASSESS THE RISK OF 1122 00:49:19,000 --> 00:49:21,200 THE DISORDERS AND WE BUILT FULL MODELS 1123 00:49:21,200 --> 00:49:22,680 AND MINIMIZE MODELS WHICH I WILL GET 1124 00:49:22,680 --> 00:49:24,520 INTO IN A LITTLE BIT AND WE MEASURE THE 1125 00:49:24,520 --> 00:49:26,600 PERFORMANCE OF ALL OF OUR MODELS AND WE 1126 00:49:26,600 --> 00:49:30,680 DEPLOYED OUR MOODLE TO WEB APPLICATION 1127 00:49:30,680 --> 00:49:32,080 DEMONSTRATION THAT YOU COULD TYPE DATA 1128 00:49:32,080 --> 00:49:34,800 INTO AND GET A SCORE OUT SO NEXT SLIDE, 1129 00:49:34,800 --> 00:49:35,040 PLEASE. 1130 00:49:35,040 --> 00:49:36,800 >> THE FIRST STEP WAS TO IDENTIFY THE 1131 00:49:36,800 --> 00:49:40,240 VARIABLES OF INTEREST, SO OUR INPUT 1132 00:49:40,240 --> 00:49:41,560 VARIABLES WERE CLIPICAL KNOWLEDGE FROM 1 1133 00:49:41,560 --> 00:49:47,800 OF OUR TEAM MEMBERS, DR. TIMATHY WEN WHO 1134 00:49:47,800 --> 00:49:50,680 IS A FETAL AND MATERNAL MEDICINE 1135 00:49:50,680 --> 00:49:52,520 SPECIALIST, HE TOOK THE LIST OF 9000 1136 00:49:52,520 --> 00:49:56,800 VARIABLE AND IT IS REDUCED IT TO US TO 1137 00:49:56,800 --> 00:49:58,440 750 VARIABLES KNOWN DO BE ASSOCIATE WIDE 1138 00:49:58,440 --> 00:50:00,200 HYPER TENSION THAT ARE USE INDEED 1139 00:50:00,200 --> 00:50:02,360 CLINICAL PRACTICE AND STUDY VS SHOWN 1140 00:50:02,360 --> 00:50:03,440 HAVE CORRELATION. 1141 00:50:03,440 --> 00:50:06,640 THIS IS INCLUDE LAB RESULTS AND VITALS, 1142 00:50:06,640 --> 00:50:08,480 MEDICAL HISTORY, FAMILY HISTORY AND 1143 00:50:08,480 --> 00:50:10,640 SOCIAL DETERMINANTS OF HEALTH, 1 INPUT 1144 00:50:10,640 --> 00:50:11,840 VARIABLE WHICH WE EXCLUDED FROM TRAINING 1145 00:50:11,840 --> 00:50:15,520 BUT KEPT FOR FUTURE ANALYSIS WAS RACE 1146 00:50:15,520 --> 00:50:19,240 AND EGHT INISSITY WHICH WE WILL DISCUSS 1147 00:50:19,240 --> 00:50:19,440 LATER. 1148 00:50:19,440 --> 00:50:21,000 SO FOR OUTCOME VARIABLES WE WANTED TO 1149 00:50:21,000 --> 00:50:24,040 SEE IF THE PERSON HAD HYPER TENSIVE 1150 00:50:24,040 --> 00:50:26,280 DISORDER FOR PREGNANCY AND THIS WAS THE 1151 00:50:26,280 --> 00:50:29,280 PRESENCE OF A DIAGNOOF THETIC COAT, 1152 00:50:29,280 --> 00:50:33,040 PREELAMPSIA EITHER MILD OR SEVERE, 1153 00:50:33,040 --> 00:50:34,440 ELAMPSIA AND HELLP SYNDROME. 1154 00:50:34,440 --> 00:50:34,840 NEXT SLIDE. 1155 00:50:34,840 --> 00:50:37,240 ONCE WE KIND OF HAD OUR DATA SET 1156 00:50:37,240 --> 00:50:39,840 ASSEMBLES WE CLEANED UP A DATA A BIT TO 1157 00:50:39,840 --> 00:50:43,480 PREPARE IT FOR OUR MACHINE LEARNING. 1158 00:50:43,480 --> 00:50:46,600 WE USED PYTHON AND COLAB AND PANDAS 1159 00:50:46,600 --> 00:50:48,400 STACK FOR OUR ANALYSIS, THERE WERE A FEW 1160 00:50:48,400 --> 00:50:49,960 CODING PROBLEMS WHICH WE FIXED AND THEN 1161 00:50:49,960 --> 00:50:55,280 1 THING WE DID WAS A 1-HOT ENCODING OF 1162 00:50:55,280 --> 00:50:57,240 REPEATED CATEGORICAL FIELDS TO HELP THE 1163 00:50:57,240 --> 00:50:57,760 MODELS LEARN BETTER. 1164 00:50:57,760 --> 00:51:00,760 A GOOD EXAMPLE IS IF YOU HAVE PEOPLE AND 1165 00:51:00,760 --> 00:51:03,120 WHAT THEIR FAVORITE COLORS THEY ARE, 1166 00:51:03,120 --> 00:51:06,360 HAVE YOU A FAVORITE COLOR 1 AND 2. 1167 00:51:06,360 --> 00:51:12,040 THIS IS NOT THAT CONDUCIVE TO MACHINE 1168 00:51:12,040 --> 00:51:14,520 LEARNING ALGORITHM THAT WANTS A CALMER 1169 00:51:14,520 --> 00:51:17,360 DATA FORMAT SO WE BROKE THIS DOWN, THIS 1170 00:51:17,360 --> 00:51:19,440 IS CALLED 1-HOT ENCODING SO WE WOULD SAY 1171 00:51:19,440 --> 00:51:21,640 DO THEY LIKE RED, YELLOW OR BLUE AND 1172 00:51:21,640 --> 00:51:25,320 THIS FIRST PATIENT FOR EXAMPLE, LIKES 1173 00:51:25,320 --> 00:51:30,320 RED AND BLUE, SO THEY ARE TRUE FOR HAS 1174 00:51:30,320 --> 00:51:36,440 RED AND TRUE FOR HAS BLUE, SO THIS 1175 00:51:36,440 --> 00:51:39,960 METHOD WAS REALLY VALUABLE FOR A LOT OF 1176 00:51:39,960 --> 00:51:43,600 COLUMNS WHICH HAD REPEATED FIELDS LIKE, 1177 00:51:43,600 --> 00:51:45,800 YOU HAD ANY FAMILY MEMBER HAD MEDICAL 1178 00:51:45,800 --> 00:51:46,640 HISTORY, ET CETERA. 1179 00:51:46,640 --> 00:51:48,800 THERE WAS ALSO A MISSING DATA PROBLEM. 1180 00:51:48,800 --> 00:51:53,400 WE WANTED TO BE MEASURING RISK AND NOT 1181 00:51:53,400 --> 00:51:54,440 WHETHER DATA VARIABLES IS COLLECTED IN 1182 00:51:54,440 --> 00:51:57,040 THE DATA SET TO ACTUALLY HAVE THAT BE 1183 00:51:57,040 --> 00:51:59,160 THE PREDICTIVE VARIABLE SO ANY VARIABLE 1184 00:51:59,160 --> 00:52:01,240 MORE THAN 90% MISSING DATA WE DROPPED 1185 00:52:01,240 --> 00:52:04,360 AND THIS LEFT US WITH 518 CO VALID AND 1186 00:52:04,360 --> 00:52:08,160 RELIABLEIATES TO PREDICT FROM. 1187 00:52:08,160 --> 00:52:08,880 NEXT SLIDE, PLEASE. 1188 00:52:08,880 --> 00:52:10,920 THE NEXT THING WE DID AFTER WE HAD A 1189 00:52:10,920 --> 00:52:13,840 CLEAN DATA SET WAS TO TRY MACHINE 1190 00:52:13,840 --> 00:52:19,000 LEARNING ALGORITHMS SO WE USED A 80/20 1191 00:52:19,000 --> 00:52:20,440 TRAINING EVALUATION SPLIT. 1192 00:52:20,440 --> 00:52:23,720 WE USED 80% OF THE DATA TO TRAIN AND 20% 1193 00:52:23,720 --> 00:52:25,720 WE USED TO EVALUATE. 1194 00:52:25,720 --> 00:52:30,200 WE EVALUATED THE AUROC WHICH IS THE AREA 1195 00:52:30,200 --> 00:52:32,600 UNDER THE RECEIVER OPERATOR 1196 00:52:32,600 --> 00:52:34,280 CHARACTERISTIC CURVE AND WE LOOKED AT 1197 00:52:34,280 --> 00:52:39,040 MODEL TRAINING TIMES SO WE USED AUTOML 1198 00:52:39,040 --> 00:52:40,760 APPROACHES WHICH IS APPROACHES WHERE YOU 1199 00:52:40,760 --> 00:52:42,240 DON'T THINK ABOUT MODEL ARCHITECTURE OR 1200 00:52:42,240 --> 00:52:45,480 HOW YOU SEGMENT YOUR DAT AYOU LET THE 1201 00:52:45,480 --> 00:52:47,040 MACHINE LEARNING MODEL AT IT AND YOU CAN 1202 00:52:47,040 --> 00:52:49,040 SEE WHAT HAPPENS, THESE CAN TAKE MANY, 1203 00:52:49,040 --> 00:52:52,880 MANY HOURS TO RUN. 1204 00:52:52,880 --> 00:52:55,200 WE ALSO USED SOME HAND TUNE NEURAL 1205 00:52:55,200 --> 00:53:00,720 NETWORKS, AND ALSO WE USED SCIKIT-LEARN 1206 00:53:00,720 --> 00:53:04,040 RANDOM FORESTS WE USED THE GRID SEARCH 1207 00:53:04,040 --> 00:53:05,240 FOR OPTIMAL PARAMETERS TO MAKE THIS 1208 00:53:05,240 --> 00:53:05,600 WORK. 1209 00:53:05,600 --> 00:53:07,840 WE FOUND THAT SIMILAR TO WHAT OTHER 1210 00:53:07,840 --> 00:53:09,840 PANELISTS HAVE DESCRIBED THAT RANDOM 1211 00:53:09,840 --> 00:53:12,920 FORESTS WORKED REALLY WELL AND WE FOUND 1212 00:53:12,920 --> 00:53:14,840 NO--IT OFTEN OUTPERFORMED THE MORE 1213 00:53:14,840 --> 00:53:16,440 COMPLEX MODELS AND THEY TRAIN INDEED A 1214 00:53:16,440 --> 00:53:18,240 MATTER OF MINUTES SO WE DEFINITELY USE 1215 00:53:18,240 --> 00:53:20,600 THESE MORE SO WE COULD HAVE A QUICKER 1216 00:53:20,600 --> 00:53:23,400 TURN AROUND AS WE BUILT THESE MODELS, 1217 00:53:23,400 --> 00:53:24,240 NEXT SLIDE, PLEASE. 1218 00:53:24,240 --> 00:53:27,280 THE NEXT STEP WAS REDUCING THE NUMBER OF 1219 00:53:27,280 --> 00:53:29,840 VARIABLES IN THIS A MINIMIZED MODEL, 1220 00:53:29,840 --> 00:53:31,920 THIS MODEL WE HAVE WITH THE CO VALID AND 1221 00:53:31,920 --> 00:53:33,400 RELIABLEIATES IS GREAT IF YOU CAN 1222 00:53:33,400 --> 00:53:35,200 INTEGRATE WITH AN EHR AND PULL THE DATA 1223 00:53:35,200 --> 00:53:38,040 IN WHICH IS STEPS WE'RE BUILDING AT 1224 00:53:38,040 --> 00:53:41,400 DELFINNA BUT THAT'S NOT GREAT FOR IF WE 1225 00:53:41,400 --> 00:53:43,280 ARE BUILDING SOMETHING FOR AN INDIVIDUAL 1226 00:53:43,280 --> 00:53:44,360 TO TRY TO BUILD AND USE. 1227 00:53:44,360 --> 00:53:49,360 SO WE USED A FEATURE CALLED RECURSIVE 1228 00:53:49,360 --> 00:53:53,840 FEATURE ELIM NATION AND FINDS THE NEXT 1 1229 00:53:53,840 --> 00:53:56,480 UNTIL YOU GET DOWN TO 20 VARIABLES, THIS 1230 00:53:56,480 --> 00:53:59,040 IS A AGREEDY ALGORITHM, MEANS YOU DON'T 1231 00:53:59,040 --> 00:54:01,840 ALL GETD THE BEST RESULT BUT IT'S RUN 1232 00:54:01,840 --> 00:54:03,560 EFFICIENTLY AND GREAT FOR 1233 00:54:03,560 --> 00:54:03,920 EXPERIMENTATION. 1234 00:54:03,920 --> 00:54:05,160 NEXT SLIDE. 1235 00:54:05,160 --> 00:54:07,320 THE FINAL STEP FOR WAS US TO REDUCE 1236 00:54:07,320 --> 00:54:09,440 RACIAL BIAS IN THE MODELS, SO THIS IS 1237 00:54:09,440 --> 00:54:13,520 SOMETHING I REFERRED BACK IN STEP 1, WE 1238 00:54:13,520 --> 00:54:15,440 ACTUALLY EXPLICITLY EXCLUDED RACE AND 1239 00:54:15,440 --> 00:54:16,680 ETHNICITY FOR THE MODEL BECAUSE WE 1240 00:54:16,680 --> 00:54:19,360 DIDN'T WANT THE MODEL TO REENFORCE 1241 00:54:19,360 --> 00:54:21,680 EXISTING BIAS IN HEALTH KOIR OUTCOMES 1242 00:54:21,680 --> 00:54:25,520 HOWEVER WE FOUND LIKE OTHERS FOUND THAT 1243 00:54:25,520 --> 00:54:26,840 THE MODEL SIGNIFICANTLY UNDERPERFORMED 1244 00:54:26,840 --> 00:54:27,880 FROM THESE AS WELL AS OTHER GROUPS SO 1245 00:54:27,880 --> 00:54:32,040 LAWN MOWER WE DID IS WE USE A TECHNIQUE 1246 00:54:32,040 --> 00:54:33,800 CALLED SMOKED WHERE WE WANTED TO BOOST 1247 00:54:33,800 --> 00:54:39,000 THE NUMBER OF ROWS IN THE DATA SET FOR 1248 00:54:39,000 --> 00:54:40,880 BLACK NONHISPANIC PATIENT IN THE 1249 00:54:40,880 --> 00:54:41,840 TRAINING SAMPLE. 1250 00:54:41,840 --> 00:54:42,240 >> ONE MINUTE. 1251 00:54:42,240 --> 00:54:44,080 >> THANK YOU, THIS FORCES THE OBJECTIVE 1252 00:54:44,080 --> 00:54:46,800 FUNCTION THAT THE MODEL IS TRYING TO 1253 00:54:46,800 --> 00:54:49,280 TRAIN TO FURTHER REDUCE THE ERROR RATE 1254 00:54:49,280 --> 00:54:50,960 FOR THESE PATIENTS RELATIVE TO THE OTHER 1255 00:54:50,960 --> 00:54:55,240 GROUPS TO TRY TO BRING THE RESULTS MORE 1256 00:54:55,240 --> 00:54:55,640 IN LINE. 1257 00:54:55,640 --> 00:54:57,640 AND WE FOUND THIS HELPED US REDUCE THAT 1258 00:54:57,640 --> 00:54:58,840 GAP. 1259 00:54:58,840 --> 00:55:00,320 NEXT SLIDE, PLEASE. 1260 00:55:00,320 --> 00:55:02,440 >> SO, HOW ARE MODELS PERFORMED, WE HAVE 1261 00:55:02,440 --> 00:55:06,760 THE FULL MODEL AND THE MINIMIZED MODEL 1262 00:55:06,760 --> 00:55:08,360 CONFUSION MATRIXES, THEY PERFORMED 1263 00:55:08,360 --> 00:55:08,880 SIMILARLY. 1264 00:55:08,880 --> 00:55:10,800 NEXT SLIDE, BETTER WAY TO VIECIALIZE 1265 00:55:10,800 --> 00:55:14,200 THIS IS THE AUROC, SO CAN YOU SEE THAT 1266 00:55:14,200 --> 00:55:16,080 THE MINIMIZED MODEL IS ALMOST AS GOOD AS 1267 00:55:16,080 --> 00:55:18,720 OUR FULL MODEL AND HAS VARIABLES, ON THE 1268 00:55:18,720 --> 00:55:19,840 RIGHT YOU SLEEP APNEA AND OBESITYY IT BY 1269 00:55:19,840 --> 00:55:23,600 RACE, SO YOU WILL SEE OUR MODEL STILL 1270 00:55:23,600 --> 00:55:24,280 UNDERPERFORMS FOR NONHISPANIC BLACK 1271 00:55:24,280 --> 00:55:26,640 PATIENTS BUT THE GAP IS MUCH LESS THAN 1272 00:55:26,640 --> 00:55:32,680 IT WAS PRIOR TO OUR [INDISCERNIBLE]. 1273 00:55:32,680 --> 00:55:35,640 SO THE FINAL THING WE DID WAS APPLY IT 1274 00:55:35,640 --> 00:55:37,240 TO A WEB APP. 1275 00:55:37,240 --> 00:55:44,520 CAN YOU ACCESS THIS AT HYPER 1276 00:55:44,520 --> 00:55:45,000 TENSION.DELFINNA.COM. 1277 00:55:45,000 --> 00:55:48,000 WE DEPLOY TODAY TO GOOGLE CLOUD FUNCTION 1278 00:55:48,000 --> 00:55:49,400 AND DOES A REPROTEIN COMPLEX PROCEDURE 1279 00:55:49,400 --> 00:55:51,480 TO THAT FUNCTION AND DISPLAYS THE RESULT 1280 00:55:51,480 --> 00:55:51,680 THERE. 1281 00:55:51,680 --> 00:55:54,440 I SHOULD NOTE THIS IS ONLY FOR 1282 00:55:54,440 --> 00:55:55,480 DEMONSTRATIVE APPROXIMATE UPPERS ONS AND 1283 00:55:55,480 --> 00:55:57,560 RESEARCH IT IS NOT MEANT TO BE USED FOR 1284 00:55:57,560 --> 00:55:58,800 CLINICAL USE. 1285 00:55:58,800 --> 00:56:01,400 AND THAT IS OUR SUBMISSION FROM 1286 00:56:01,400 --> 00:56:03,000 DELFINNA, I WANT TO THANK THE TEAM FOR 1287 00:56:03,000 --> 00:56:06,800 THE DATA SET AND DASH FOR GIVING US 1288 00:56:06,800 --> 00:56:08,840 ACCESS AND ALL THE TEAM MEMBERS FROM 1289 00:56:08,840 --> 00:56:10,040 DELFIN WHO WORK WIDE ME ON THIS. 1290 00:56:10,040 --> 00:56:12,160 THAT'S IT FOR US. 1291 00:56:12,160 --> 00:56:16,680 I AM HAPPY TO TAKE ANY QUESTIONS. 1292 00:56:16,680 --> 00:56:21,840 >> THANK YOU VERY MUCH, ALI. 1293 00:56:21,840 --> 00:56:22,440 VERY INFORMATIVE PRESENTATION. 1294 00:56:22,440 --> 00:56:24,960 THERE IS A COMMENT THAT SAYS, THIS 1295 00:56:24,960 --> 00:56:27,400 IS--THAT YOUR PRESENTATION IS VERY 1296 00:56:27,400 --> 00:56:30,040 INFORMATIVE AND TO THANK DELFINNA FOR 1297 00:56:30,040 --> 00:56:35,200 THEIR COMMITMENT TO HEALTH EQUITY. 1298 00:56:35,200 --> 00:56:37,000 >> THANK YOU ANONYMOUS ATTENDEE. 1299 00:56:37,000 --> 00:56:40,480 >> YEAH, I'M ALSO GLAD THAT YOU ENJOYED 1300 00:56:40,480 --> 00:56:43,080 ANALYZING THE NEW MOMS TO BE DATA SET. 1301 00:56:43,080 --> 00:56:47,520 JUST 1 QUESTION, ARE THERE PLANS TO 1302 00:56:47,520 --> 00:56:50,720 COMMERCIALIZE YOUR TOOL HERE? 1303 00:56:50,720 --> 00:56:51,880 >> THAT'S ACTUALLY SOMETHING WE'RE VERY 1304 00:56:51,880 --> 00:56:54,160 INTERESTED IN DOING, WE'RE CURRENTLY 1305 00:56:54,160 --> 00:56:56,600 TALKING TO NICHD ABOUT LICENSING DATA 1306 00:56:56,600 --> 00:56:58,560 SET FORIE COMMERCIAL APPLICATION, BUT 1307 00:56:58,560 --> 00:57:00,560 SEPARATELY WE WANT TO DO THE SAME 1308 00:57:00,560 --> 00:57:02,120 TECHNIQUES WE USED AND ACTUALLY WE ARE 1309 00:57:02,120 --> 00:57:05,240 WORKING WITH A CLINIC IN SOUTHERN 1310 00:57:05,240 --> 00:57:06,400 CALIFORNIA AND WE'RE INTERESTED IN 1311 00:57:06,400 --> 00:57:09,840 TRYING TO BUILD MODELS FOR HYPER TENSION 1312 00:57:09,840 --> 00:57:11,720 AND HOW THAT--THE DOCTOR IN THAT CLINIC 1313 00:57:11,720 --> 00:57:13,760 ACTUALLY USE THEM AND SEE IF IT IMPROVES 1314 00:57:13,760 --> 00:57:16,320 CARE, SO THAT'S SOMETHING THAT WE'RE 1315 00:57:16,320 --> 00:57:18,080 LASER FOCUSED ON, THANK YOU IF ARE THAT 1316 00:57:18,080 --> 00:57:18,960 QUESTION. 1317 00:57:18,960 --> 00:57:20,680 >> GREAT. 1318 00:57:20,680 --> 00:57:21,000 GREAT DEAL. 1319 00:57:21,000 --> 00:57:22,720 THANK YOU. 1320 00:57:22,720 --> 00:57:27,960 >> OKAY, OUR NEXT PRESENTER WILL BE 1321 00:57:27,960 --> 00:57:30,480 DR. MONICA KEITH WHO IS A DATA SCIENCE 1322 00:57:30,480 --> 00:57:40,440 FELLOW AT THE UNIVERSITY OF WASHINGTON 1323 00:57:40,440 --> 00:57:41,240 IN SEATTLE. 1324 00:57:41,240 --> 00:57:43,760 THERE WE GO. 1325 00:57:43,760 --> 00:57:44,280 DR. KEITH? 1326 00:57:44,280 --> 00:57:46,080 >> THANK YOU MAURICE, HELLO EVERYONE. 1327 00:57:46,080 --> 00:57:49,240 THANK YOU ALL FOR HAVING ME AND FOR 1328 00:57:49,240 --> 00:57:51,080 TUNING IN TODAY. 1329 00:57:51,080 --> 00:57:53,040 I WORKED WITH MELANIE MARTIN TO DEVELOP 1330 00:57:53,040 --> 00:57:56,240 A SOLUTION FOR THIS DATA CHALLENGE. 1331 00:57:56,240 --> 00:57:59,720 WE ARE BOTH BIOCULTURAL ANTHROPOLOGISTS 1332 00:57:59,720 --> 00:58:02,880 AT THE UNIVERSITY OF WASHINGTON. 1333 00:58:02,880 --> 00:58:05,480 AND TODAY I AM SHARING THE STRUCTURAL 1334 00:58:05,480 --> 00:58:09,440 EQUATION MODEL THAT WE DEVELOPED, OUR 1335 00:58:09,440 --> 00:58:12,440 MODEL IS DESIGNED TO IDENTIFY PATHWAYS 1336 00:58:12,440 --> 00:58:13,680 BETWEEN SOCIAL DETERMINANTS OF HEALTH 1337 00:58:13,680 --> 00:58:17,080 AND DOWN STREAM MORBIDITYS. 1338 00:58:17,080 --> 00:58:19,760 WE FOCUSED ON DISPARITIES AND HYPER 1339 00:58:19,760 --> 00:58:21,480 TENSIVE DISORDERS OF PREGNANCY FOR THIS 1340 00:58:21,480 --> 00:58:23,400 DATA CHALLENGE AND OUR MODEL IS ALSO 1341 00:58:23,400 --> 00:58:29,120 BASED ON THE CONCEPT EVER ALOSTATTIC 1342 00:58:29,120 --> 00:58:29,320 LOAD. 1343 00:58:29,320 --> 00:58:31,040 NEXT SLIDE. 1344 00:58:31,040 --> 00:58:31,280 PLEASE. 1345 00:58:31,280 --> 00:58:33,720 SO STARTING WITH THE FRAMEWORK IN SOCIAL 1346 00:58:33,720 --> 00:58:35,760 DETERMINANTINGS OF HEALTH, THERE HAS 1347 00:58:35,760 --> 00:58:38,400 BEEN MORE AND MORE DISCUSSION RECENTLY 1348 00:58:38,400 --> 00:58:40,400 REGARDING SOCIAL DETERMINANTS OF HEALTH. 1349 00:58:40,400 --> 00:58:48,080 AND THESE ARE OFTEN BROKEN DOWN AS SUCH 1350 00:58:48,080 --> 00:58:51,960 LIKE ECONOMIC AND SOCIAL SUPPORT 1351 00:58:51,960 --> 00:58:55,480 METRICS, DIFFERENT HEALTH VARIABLES, 1352 00:58:55,480 --> 00:58:57,800 DIFFERENT ACCESS TO AND HELP WITH 1353 00:58:57,800 --> 00:58:59,840 HEALTHCARE AND OF COURSE PHYSICAL 1354 00:58:59,840 --> 00:59:01,040 ENVIRONMENT AND BUILT ENVIRONMENT AS 1355 00:59:01,040 --> 00:59:01,240 WELL. 1356 00:59:01,240 --> 00:59:04,240 SO BASED ON THE DATA THAT WERE AVAILABLE 1357 00:59:04,240 --> 00:59:07,000 TO US, IN NEW MOM TO BE, WE WERE ABLE TO 1358 00:59:07,000 --> 00:59:11,880 MODEL A VARIETIES OF HEALTH BEHAVIORS 1359 00:59:11,880 --> 00:59:14,680 AND ALSO SOCIOECONOMIC FACTORS IN OUR 1360 00:59:14,680 --> 00:59:18,440 MODEL SOLUTION. 1361 00:59:18,440 --> 00:59:19,040 NEXT SLIDE, PLEASE. 1362 00:59:19,040 --> 00:59:21,440 SO AS I MENTIONED BEFORE, OUR MODEL IS 1363 00:59:21,440 --> 00:59:24,800 ALSO BASED ON THE CONCEPT OF ALOE STATIC 1364 00:59:24,800 --> 00:59:25,480 LOAD. 1365 00:59:25,480 --> 00:59:29,440 SO ALOE STATIC LOAD IS DEFINED AS THE 1366 00:59:29,440 --> 00:59:32,880 CUMULATIVE BURDEN OF PSYCHOLOGICAL AND 1367 00:59:32,880 --> 00:59:34,880 PHYSIOLOGICAL STRESS, EXPERIENCED ACROSS 1368 00:59:34,880 --> 00:59:38,440 THE LIFE COURSE, SO A MEASURE OF CHRONIC 1369 00:59:38,440 --> 00:59:41,600 STRESS, THAT REALLY CONCEPTUALLY LINKS 1370 00:59:41,600 --> 00:59:44,840 SOCIAL DETERMINANTS OF HEALTH WITH 1371 00:59:44,840 --> 00:59:45,240 CLINICAL BIOMARKERS. 1372 00:59:45,240 --> 00:59:47,120 AND CAN YOU SEE ON THE FIGURE TO THE 1373 00:59:47,120 --> 00:59:50,400 RIGHT, ON THIS SLIDE, HERE, THAT 1374 00:59:50,400 --> 00:59:53,400 BIOMARKERS OF ALOE STATIC LOAD ARE KNOWN 1375 00:59:53,400 --> 00:59:56,800 TO ASSOCIATE WITH MANY MATERNAL 1376 00:59:56,800 --> 00:59:58,160 MORBIDITYS INCLUDING PREELAMPSIA, WHICH 1377 00:59:58,160 --> 01:00:00,200 IS 1 OF THE MORBIDITIES THAT WE FOCUSED 1378 01:00:00,200 --> 01:00:04,960 ON IN OUR STRUCTURAL EQUATION MODEL FOR 1379 01:00:04,960 --> 01:00:10,760 THIS CHALLENGE, NEXT, SLIDE, PLEASE. 1380 01:00:10,760 --> 01:00:13,760 SO WE HYPOTHESIZED HERE THAT THE 1381 01:00:13,760 --> 01:00:16,640 CUMULATIVE EFFECTS OF SOCIOLOGICAL 1382 01:00:16,640 --> 01:00:18,680 ECOLOGICAL STRESSORS ACT DIRECTLY AND 1383 01:00:18,680 --> 01:00:21,920 INDIRECTLY VIA HEALTH BEHAVIORS TO 1384 01:00:21,920 --> 01:00:24,400 INCREASE ALOE STATIC LOAD ACROSS 1385 01:00:24,400 --> 01:00:26,120 GUESTATION WHICH THEN IN TURN INCREASING 1386 01:00:26,120 --> 01:00:29,600 THE RISK OF DEVELOPING MATERNAL 1387 01:00:29,600 --> 01:00:33,000 MORBIDITIES AND MORE SPECIFICALLY AN 1388 01:00:33,000 --> 01:00:35,120 INCREASED RISK OF HYPER TENSIVE 1389 01:00:35,120 --> 01:00:36,680 DISORDERS OF PREGNANCY. 1390 01:00:36,680 --> 01:00:40,160 AND THESE INCLUDE DIFFERENT SEVERITYS OF 1391 01:00:40,160 --> 01:00:44,080 PREECLAMPSIA AND NEW ONSET HYPER TENSION 1392 01:00:44,080 --> 01:00:45,640 DURING OR RESULTING FROM PREGNANCY. 1393 01:00:45,640 --> 01:00:48,560 AND AS YOU WILL SEE HERE, THERE ARE ALSO 1394 01:00:48,560 --> 01:00:50,440 SIGNIFICANT RACIAL AND ETHNIC 1395 01:00:50,440 --> 01:00:52,800 DISPARITIES IN HYPER TENSIVE DISORDERS 1396 01:00:52,800 --> 01:00:55,240 OF PREGNANCY WHICH HAS ALSO BEEN 1397 01:00:55,240 --> 01:00:59,440 HIGHLIGHTED IN SEVERAL OF THE PREVIOUS 1398 01:00:59,440 --> 01:01:03,760 TALKS TODAY AND WE PAUSE IT HERE THAT 1399 01:01:03,760 --> 01:01:06,560 THESE DISPARITIES RESULT FROM 1400 01:01:06,560 --> 01:01:08,760 SOCIOLOGICAL ECOLOGICAL INEQUITIES AND 1401 01:01:08,760 --> 01:01:11,480 RACIALIZED PROCESSES BY WHICH STRESSORS 1402 01:01:11,480 --> 01:01:13,280 BECOME ENBODIED AS HEALTH RISKS. 1403 01:01:13,280 --> 01:01:16,280 SOPHISTICATEDY WE DECIDES A STRUCTURAL 1404 01:01:16,280 --> 01:01:19,000 EQUATION MODEL TO IDENTIFY CAUSAL 1405 01:01:19,000 --> 01:01:24,240 PATHWAYS BETWEEN SOCIAL DETERMINANTS, 1406 01:01:24,240 --> 01:01:28,360 MEASURES OF ALLOSTATTIC LOAD AND HYPER 1407 01:01:28,360 --> 01:01:31,760 TENSIVE DISORD EROZANS OF PREGNANCY AND 1408 01:01:31,760 --> 01:01:33,880 WE HYPOTHESIZED THAT THESE FACTORS MAY 1409 01:01:33,880 --> 01:01:37,960 VARY IN MAGNITUDE AND SIGNIFICANCE AMONG 1410 01:01:37,960 --> 01:01:40,000 RACIAL AND ETHNIC GROUPS SO WE RAN OUR 1411 01:01:40,000 --> 01:01:42,080 STRUCTURAL EQUATION MODEL IN A 1412 01:01:42,080 --> 01:01:44,840 MULTIGROUP FRAMEWORK TO GET GROUP 1413 01:01:44,840 --> 01:01:49,240 SPECIFIC PATH CO-EFFICIENTS. 1414 01:01:49,240 --> 01:01:49,720 NEXT SLIDE, PLEASE. 1415 01:01:49,720 --> 01:01:51,720 SO I'LL TAKE A MOMENT HERE TO MENTION 1416 01:01:51,720 --> 01:01:55,560 THE KEY FEATURES OF STRUCTURAL EQUATION 1417 01:01:55,560 --> 01:01:57,600 MODELS, THIS SAY MULTIVALID AND 1418 01:01:57,600 --> 01:01:59,200 RELIABLEIATE ANALYSIS THAT MEASURES 1419 01:01:59,200 --> 01:02:00,960 RELATIONSHIPS BETWEEN LATENT AND 1420 01:02:00,960 --> 01:02:03,680 OBSERVED VARIABLES AND A CAUSAL PATHWAY 1421 01:02:03,680 --> 01:02:03,960 FRAMEWORK. 1422 01:02:03,960 --> 01:02:08,040 SO THESE MODEL LINEAR RELATIONSHIPS AND 1423 01:02:08,040 --> 01:02:10,840 PRODUCE PAST CO EFFICIENTS THAT COUNT 1424 01:02:10,840 --> 01:02:14,080 FOR THE VARIANCES AND CO VARIANCES AMONG 1425 01:02:14,080 --> 01:02:16,920 THE DATA IN THE MODEL AND STRUCTURAL 1426 01:02:16,920 --> 01:02:18,880 EQUATION MODELS HAVE 2 MAIN COMPONENTS. 1427 01:02:18,880 --> 01:02:22,720 SO THERE'S THE MEASUREMENTS MODEL OF 1428 01:02:22,720 --> 01:02:24,840 RELATIONSHIPS BETWEEN LATENT FACTORS AND 1429 01:02:24,840 --> 01:02:27,440 THEIR OBSERVED INDICATOR VARIABLES. 1430 01:02:27,440 --> 01:02:29,440 SO IN THE SOCIAL DETERMINANTS FRAMEWORK, 1431 01:02:29,440 --> 01:02:33,240 WE WILL HAVE A LATENT FACTOR FOR THE 1432 01:02:33,240 --> 01:02:35,000 SOCIAL ENVIRONMENT, OR SOCIOECONOMIC 1433 01:02:35,000 --> 01:02:35,560 POSITION. 1434 01:02:35,560 --> 01:02:37,560 AND WE HAVE INDICATOR VARIABLES THAT 1435 01:02:37,560 --> 01:02:42,240 LOAD ON TO IT, LIKE INCOMMAND 1436 01:02:42,240 --> 01:02:45,000 EDUCATIONAL ATTAINMENT BUT THESE ARE 1437 01:02:45,000 --> 01:02:47,400 REALLY INCOMPLETE PROXIES OF THE SOCIAL 1438 01:02:47,400 --> 01:02:48,400 ENVIRONMENT. 1439 01:02:48,400 --> 01:02:52,040 SO THIS LATENT VARIABLE FRAMEWORK IS 1440 01:02:52,040 --> 01:02:54,280 PARTICULARLY APPROPRIATE FOR MODELING 1441 01:02:54,280 --> 01:02:55,920 SOCIAL DETERMINANTS IN THIS WAY AND THEN 1442 01:02:55,920 --> 01:02:58,360 THE STRUCTURAL PART OF THE MODEL HAS ALL 1443 01:02:58,360 --> 01:03:02,040 OF THE CONNECTING PATHWAYS, BETWEEN 1444 01:03:02,040 --> 01:03:04,600 VARIABLES. 1445 01:03:04,600 --> 01:03:05,840 NEXT SLIDE, PLEASE. 1446 01:03:05,840 --> 01:03:10,960 SO OUR MODEL HAS 4 LATENT SOCIAL 1447 01:03:10,960 --> 01:03:13,240 DETERMINANT FACTORS THAT WE WERE ABLE TO 1448 01:03:13,240 --> 01:03:17,680 CHARACTERIZE REALLY ABOUT WITH THIS DATA 1449 01:03:17,680 --> 01:03:17,840 SET. 1450 01:03:17,840 --> 01:03:18,080 NEXT. 1451 01:03:18,080 --> 01:03:22,640 AND THEN WE HAD 3 ALLOSTATTIC LOAD 1452 01:03:22,640 --> 01:03:24,560 MEASURES THAT WERE MEASURED AT 4 1453 01:03:24,560 --> 01:03:25,640 DIFFERENT TIME POINTS THROUGHOUT 1454 01:03:25,640 --> 01:03:29,640 PREGNANCY AS WELL SO THESE 3 MARKERS AS 1455 01:03:29,640 --> 01:03:31,520 WELL, THESE ARE DIASTOOL DNAIC BLOOD 1456 01:03:31,520 --> 01:03:35,000 PRESSURE AND WEIGHT WERE MEASURED AT 4 1457 01:03:35,000 --> 01:03:41,360 DIFFERENT TIMES ACROSS GUESTATION AND WE 1458 01:03:41,360 --> 01:03:42,640 MODELED INDIVIDUAL ALLOSTATTIC 1459 01:03:42,640 --> 01:03:44,240 TRAJECTORIES AND INTERCEPTED THEIR 1460 01:03:44,240 --> 01:03:45,800 TRAJECTORIES AND SLOPES BECAUSE YOU CAN 1461 01:03:45,800 --> 01:03:48,040 IMAGINE THAT THE STARTING LEVEL OF BLOOD 1462 01:03:48,040 --> 01:03:50,920 PRESSURE AT THE INTERCEPT MAY IMPACT 1463 01:03:50,920 --> 01:03:53,120 DOWN STREAM HYPER TENSE OF RISK 1464 01:03:53,120 --> 01:03:54,760 DIFFERENTLY THAN THE SCOPE OF BLOOD 1465 01:03:54,760 --> 01:03:57,360 PRESSURE CHANGING ACROSS GUESTATION. 1466 01:03:57,360 --> 01:03:57,720 NEXT. 1467 01:03:57,720 --> 01:04:02,760 AND THEN WE MODELED 5 LEVELS OF HYPER 1468 01:04:02,760 --> 01:04:06,520 TENSIVE PREGNANCY OUTCOMES. 1469 01:04:06,520 --> 01:04:07,120 NEXT. 1470 01:04:07,120 --> 01:04:08,640 AND WE INCLUDED THESE VARIOUS CONTROLS 1471 01:04:08,640 --> 01:04:11,280 THAT YOU SEE HERE TO ACCOUNT FOR ASPECTS 1472 01:04:11,280 --> 01:04:14,560 OF BASE LINE MATERNAL PHYSIOLOGY, AND 1473 01:04:14,560 --> 01:04:20,240 RELEVANT CONFOUNDERS AS WELL. 1474 01:04:20,240 --> 01:04:20,840 NEXT; NEXT. 1475 01:04:20,840 --> 01:04:24,760 OKAY, THERE WE GO. 1476 01:04:24,760 --> 01:04:27,840 YEAH AND THEN WHEN WE FILL IN THE 1477 01:04:27,840 --> 01:04:29,880 STRUCTURAL DIAGRAM HERE, WE'RE MODELING 1478 01:04:29,880 --> 01:04:33,480 PATHWAYS FROM SOCIAL DETERMINANTS TO 1479 01:04:33,480 --> 01:04:35,960 THOSE ALLOSTATTIC LOAD MARKERS AND FROM 1480 01:04:35,960 --> 01:04:37,840 THOSE MARKER INTERCEPTS AND SLOPES TO 1481 01:04:37,840 --> 01:04:39,640 HYPER TENSIVE OUTCOMES AND WE RAN THIS 1482 01:04:39,640 --> 01:04:44,120 MODEL IN R WITH THE LA VON PACKAGE. 1483 01:04:44,120 --> 01:04:44,560 NEXT, SLIDE, PLEASE. 1484 01:04:44,560 --> 01:04:47,600 SO THIS IS WHAT THE SOCIAL DETERMINANTS 1485 01:04:47,600 --> 01:04:49,280 LATENT VARIABLE PORTION OF THE MODEL 1486 01:04:49,280 --> 01:04:50,760 LOOKS LIKE. 1487 01:04:50,760 --> 01:04:54,640 SO YOU CAN SEE THE OBSERVED INDICATOR 1488 01:04:54,640 --> 01:04:56,520 VARIABLES HERE THAT LOAD ON TO THESE 1489 01:04:56,520 --> 01:04:59,120 FACTORS AND SINCE WE MODELED THIS IN A 1490 01:04:59,120 --> 01:05:02,440 MULTIGROUP MANNER, WE GET GROUP SPECIFIC 1491 01:05:02,440 --> 01:05:04,560 COENTIOUS FICIENTS FOR EACH LOADING AND 1492 01:05:04,560 --> 01:05:07,120 THERE IS A LOT OF SOCIAL SCIENCE 1493 01:05:07,120 --> 01:05:10,440 RESEARCH TO SUPPORT MODELING RACIAL AND 1494 01:05:10,440 --> 01:05:13,040 ETHNIC AFFILIATIONS IN IN MANNER, WE 1495 01:05:13,040 --> 01:05:15,920 KNOW FOR EXAMPLE, THAT OFTEN TIMES, 1496 01:05:15,920 --> 01:05:18,640 BLACK AMERICANS, TAKE ON A LOT MORE DEBT 1497 01:05:18,640 --> 01:05:20,640 THAN WHITE AMERICANS TO ACHIEVE THE 1498 01:05:20,640 --> 01:05:25,440 EXACT SAME LEVEL OF EDUCATIONAL 1499 01:05:25,440 --> 01:05:27,200 ATTAINMENT FOR EXAMPLE. 1500 01:05:27,200 --> 01:05:29,640 THERE ARE OFTEN SIGNIFICANT 1501 01:05:29,640 --> 01:05:31,240 DIFFERENCINGS OVER WEALTH AND PROPERTY 1502 01:05:31,240 --> 01:05:32,880 ENROLLINGSHIP, THINGS THAT ARE OFTEN NOT 1503 01:05:32,880 --> 01:05:34,880 DIRECTLY CAPTURED VERY WELL IN OUR DATA 1504 01:05:34,880 --> 01:05:37,480 SO WHEN WE'RE RELYING ON METRICS LIKE 1505 01:05:37,480 --> 01:05:40,240 INCOME AND EDUCATION LEVEL TO SIGNAL THE 1506 01:05:40,240 --> 01:05:41,920 SOCIAL ENVIRONMENT, IT'S IMPORTANT THAT 1507 01:05:41,920 --> 01:05:45,200 WE RECOGNIZE THAT THESE METRICS MIGHT 1508 01:05:45,200 --> 01:05:49,240 NOT BE INDICATING COMPARABLE 1509 01:05:49,240 --> 01:05:49,840 CIRCUMSTANCES, ACROSS GROUPS. 1510 01:05:49,840 --> 01:05:52,520 ONE MORE THINK THIS I WILL ADD HERE IS 1511 01:05:52,520 --> 01:05:55,960 THAT THESE MODELS ARE VALIDATED BY THEIR 1512 01:05:55,960 --> 01:05:59,160 FIT STATISTICS AND I WILL JUST POINT 1513 01:05:59,160 --> 01:06:00,960 OUT, YOU CAN SEE IN THE BOX IN THE 1514 01:06:00,960 --> 01:06:02,320 BOTTOM RIGHT CORNER THERE THAT WE DO 1515 01:06:02,320 --> 01:06:04,320 HAVE STRONG SUPPORT FOR THIS PORTION OF 1516 01:06:04,320 --> 01:06:06,280 THE MODEL HERE, FITTING THE DAILY BASIS 1517 01:06:06,280 --> 01:06:10,360 THEA REALLY WELL AND WE SEE THE SAME 1518 01:06:10,360 --> 01:06:13,680 STRONG SUPPORT AND MODELS FIT FOR THE 1519 01:06:13,680 --> 01:06:15,320 FULL STRUCTURAL EQUATION MODEL ON THE 1520 01:06:15,320 --> 01:06:21,480 NEXT SLIDE AS WELL. 1521 01:06:21,480 --> 01:06:22,800 NEXT SLIDE, PLEASE. 1522 01:06:22,800 --> 01:06:25,440 AND HERE ARE OUR FULL MODEL RESULTS, SO 1523 01:06:25,440 --> 01:06:29,680 THIS DIAGRAM HERE HAS STATISTICALLY 1524 01:06:29,680 --> 01:06:32,640 SIGNIFICANT PAST CO EFFICIENTS ONLILY 1525 01:06:32,640 --> 01:06:35,800 SHOWN ON IT AND EXPECTEDLY THESE 1526 01:06:35,800 --> 01:06:40,240 CLINICAL MARKERS OF ALLOSTATTIC LOAD ARE 1527 01:06:40,240 --> 01:06:41,600 INFORMATIVE OF HYPER TENSIVE RISK. 1528 01:06:41,600 --> 01:06:43,840 NOW IF YOU LOOK OVER AT THE MORBIDITY 1529 01:06:43,840 --> 01:06:47,120 OUTCOMES ON THE RIGHT HAND SIDE OF THIS 1530 01:06:47,120 --> 01:06:49,760 FIGURE HERE, YOU WILL NOTICE THAT BLACK 1531 01:06:49,760 --> 01:06:52,000 MOTHER VS WORST HYPER TENSIVE OUTCOMES 1532 01:06:52,000 --> 01:06:56,080 AND MORE SEVERE HYPER TENSIVE OUTCOMES 1533 01:06:56,080 --> 01:06:59,720 PROPORTIONAL TO THEIR SAMPLE SIZE IN 1534 01:06:59,720 --> 01:07:00,680 THESE DATA. 1535 01:07:00,680 --> 01:07:02,240 AND THIS MODEL IDENTIFIES SEVERAL 1536 01:07:02,240 --> 01:07:06,040 PATHWAYS THAT ARE SPECIFIC TO THIS MOST 1537 01:07:06,040 --> 01:07:08,120 ADVERSELY AFFECTED GROUP SO YOU WILL 1538 01:07:08,120 --> 01:07:11,240 NOTICE, ALSO THAT NOT ALL OF THE 1539 01:07:11,240 --> 01:07:13,720 SIGNIFICANT PATHWAYS ON THIS DIAGRAM ARE 1540 01:07:13,720 --> 01:07:20,160 SHARED ACROSS ALL OF THE RACIAL AND 1541 01:07:20,160 --> 01:07:20,880 ETHNIC GROUPS. 1542 01:07:20,880 --> 01:07:22,720 SO I'LL DRAW YOUR ATTENTION TO THE 1543 01:07:22,720 --> 01:07:24,720 LARGEST CO EFFICIENT ON THIS DIAGRAM 1544 01:07:24,720 --> 01:07:27,560 WHICH IS THE 1.26 WITH IN BLACK TEXT 1545 01:07:27,560 --> 01:07:30,560 THAT'S ON THAT GREEN LINE, GOING FROM 1546 01:07:30,560 --> 01:07:33,160 THE SOCIAL ENVIRONMENT TO THE WEIGHT 1547 01:07:33,160 --> 01:07:37,680 INTERCEPT AMONG BLACK MOTHERS, SO THAT'S 1548 01:07:37,680 --> 01:07:39,840 THE LARGEST PATH CO EFFICIENT IN THIS 1549 01:07:39,840 --> 01:07:42,720 ENTIRE MODEL AND THIS ASSOCIATION 1550 01:07:42,720 --> 01:07:44,920 BETWEEN THE SOCIAL ENVIRONMENT AND 1551 01:07:44,920 --> 01:07:48,160 WEIGHT IS INDEPENDENT OF DIETARY OR 1552 01:07:48,160 --> 01:07:50,880 EXERCISE BEHAVIORS WHICH WE HAVE ALSO 1553 01:07:50,880 --> 01:07:53,760 ACCOUNTED FOR IN THIS MODEL AND WE 1554 01:07:53,760 --> 01:07:57,680 EMPHASIZE THIS HERE AS SOCIAL 1555 01:07:57,680 --> 01:07:59,320 SCIENTISTS, SINCE SO MUCH OF THE 1556 01:07:59,320 --> 01:08:01,440 DISCUSSION AROUND WEIGHT IN PARTICULAR 1557 01:08:01,440 --> 01:08:03,760 IS STILL FOCUSED SO HEAVILY ON 1558 01:08:03,760 --> 01:08:06,040 INDIVIDUAL BEHAVIORS AND WHAT ARE 1559 01:08:06,040 --> 01:08:10,240 USUALLY DESCRIBED AS HEALTH CHOICES 1560 01:08:10,240 --> 01:08:11,640 WHEREAS THESE STANDARDIZED PAST 1561 01:08:11,640 --> 01:08:13,840 COENTIOUS FICIENTS HERE ARE HIGHLIGHTING 1562 01:08:13,840 --> 01:08:15,640 THAT SOCIAL INIQUITIES ARE REALLY 1563 01:08:15,640 --> 01:08:21,560 THE--AT THE ROOT OF THIS PATHWAY. 1564 01:08:21,560 --> 01:08:24,720 SO THEN THIS MODEL IDENTIFIED PATHWAYS 1565 01:08:24,720 --> 01:08:27,160 AND QUANTIFIED RELATIVE RISK FACTORS FOR 1566 01:08:27,160 --> 01:08:30,440 HYPER TENSIVE MORBIDITIES THAT WARRANT 1567 01:08:30,440 --> 01:08:32,120 FURTHER RESEARCH, POLICY AND ATTENTION 1568 01:08:32,120 --> 01:08:35,320 IN CLINICAL SETTINGS TO ADDRESS HEALTH 1569 01:08:35,320 --> 01:08:38,760 DISPARITIES AND WE DO HAVE PLANS TO 1570 01:08:38,760 --> 01:08:40,520 EXPAND THIS MODELING FRAMEWORK MOVING 1571 01:08:40,520 --> 01:08:43,240 FORWARD TO OTHER HEALTH AND PREGNANCY 1572 01:08:43,240 --> 01:08:47,640 OUTCOMES AS WELL. 1573 01:08:47,640 --> 01:08:48,720 NEXT SLIDE, PLEASE. 1574 01:08:48,720 --> 01:08:50,400 WE WOULD LIKE TO ACKNOWLEDGE THE SUPPORT 1575 01:08:50,400 --> 01:08:54,120 AND HELPFUL FEEDBACK ON THIS MODEL FROM 1576 01:08:54,120 --> 01:08:57,640 THESE FOLKS HERE. 1577 01:08:57,640 --> 01:08:58,760 THANK YOU ALL FOR YOUR TAME AND 1578 01:08:58,760 --> 01:09:00,040 ATTENTION TODAY AND I AM HAPPY TO TAKE 1579 01:09:00,040 --> 01:09:07,320 QUESTION FIST WE HAVE A FEW MINUTES 1580 01:09:07,320 --> 01:09:07,520 HERE. 1581 01:09:07,520 --> 01:09:12,240 >> THANK YOU VERY MUCH FOR YOUR VERY 1582 01:09:12,240 --> 01:09:13,040 INFORMATIVE PRESENTATION MONICA. 1583 01:09:13,040 --> 01:09:15,840 AND HOPE WE DO HAVE SOME TIME. 1584 01:09:15,840 --> 01:09:18,880 SO IF THERE ARE BURNING QUESTIONS, 1585 01:09:18,880 --> 01:09:22,600 PLEASE TYPE THEM IN THE CHAT AND WE WILL 1586 01:09:22,600 --> 01:09:27,520 ASK THOSE QUESTIONS OF OUR PRESENTERS 1587 01:09:27,520 --> 01:09:27,840 GOING FORWARD. 1588 01:09:27,840 --> 01:09:30,520 AS OF RIGHT NOW, THERE ARE NO QUESTIONS 1589 01:09:30,520 --> 01:09:34,240 FOR YOU, BUT THANK YOU VERY MUCH MONICA 1590 01:09:34,240 --> 01:09:36,440 FOR YOUR PRESENTATION. 1591 01:09:36,440 --> 01:09:38,320 >> THANK YOU. 1592 01:09:38,320 --> 01:09:45,600 >> OUR FINAL PRESENTATION IS FROM 1593 01:09:45,600 --> 01:09:51,000 DR. NICOLE CARSON FROM THE SCHOOL OF 1594 01:09:51,000 --> 01:09:53,560 NURSING ASSISTANT SCHOOL OF NURSING AT 1595 01:09:53,560 --> 01:09:57,040 EMORY UNIVERSITY IN ATLANTA. 1596 01:09:57,040 --> 01:10:00,440 HER TITLE IS MATERNAL MORBIDITY AND 1597 01:10:00,440 --> 01:10:02,640 INTERSECTIONAL SOCIAL DETERMINANTS OF 1598 01:10:02,640 --> 01:10:04,560 HEALTH PHENOTYPE. 1599 01:10:04,560 --> 01:10:05,160 DR. CARLSON, WELCOME. 1600 01:10:05,160 --> 01:10:06,240 >> THANK YOU SO MUCH IT'S A PLEASURE TO 1601 01:10:06,240 --> 01:10:08,040 BE HERE WITH YOU ALL, I WILL PRESENT ON 1602 01:10:08,040 --> 01:10:13,560 A PROJECT THAT I CONDUCTED WITH 1603 01:10:13,560 --> 01:10:16,840 DR. ELISE ERICKSON WHICH WE ARE BOTH 1604 01:10:16,840 --> 01:10:18,200 ACADEMIC MIDWIFE SCIENTISTS ABOUT 1605 01:10:18,200 --> 01:10:21,240 MATERNAL MORBIDITY AND HOW IT RELATES TO 1606 01:10:21,240 --> 01:10:22,880 THE INTERSECTIONAL DETERMINANTS OF 1607 01:10:22,880 --> 01:10:24,800 PHENOTYPE IN THE NEW MOMS TO BE DATA 1608 01:10:24,800 --> 01:10:25,040 SET. 1609 01:10:25,040 --> 01:10:28,560 NEXT SLIDE, PLEASE MPLET I REALLY 1610 01:10:28,560 --> 01:10:31,160 APPRECIATE HOW ALL OF THESE PANELISTS 1611 01:10:31,160 --> 01:10:32,440 AND PRESENTATIONS HAVE REALLY HELPED 1 1612 01:10:32,440 --> 01:10:35,800 ANOTHER TO KIND OF BUILD, SO WE'RE GOING 1613 01:10:35,800 --> 01:10:38,320 TO DO A BIT OF BACKGROUND FOR THE WORK, 1614 01:10:38,320 --> 01:10:42,000 I WILL INTRODUCE YOU TO OUR THEORETICAL 1615 01:10:42,000 --> 01:10:43,640 MODEL AND GET THROUGH KEY METHODS AND 1616 01:10:43,640 --> 01:10:45,040 RESULTS AND TAKE AWAYS. 1617 01:10:45,040 --> 01:10:45,440 NEXT SLIDE. 1618 01:10:45,440 --> 01:10:47,440 SO I WANT TO ACKNOWLEDGE THE FIRST THE 1619 01:10:47,440 --> 01:10:50,080 NEW MOMS TO BE STUDY AND DASH FOR 1620 01:10:50,080 --> 01:10:51,720 PROVIDING THAT INFORMATION, ALSO BOTH 1621 01:10:51,720 --> 01:10:58,040 MYSELF AND DR. ERICSON ARE SUPPORTED BY 1622 01:10:58,040 --> 01:11:00,520 THE NIH AND NINR IN OUR WORK. 1623 01:11:00,520 --> 01:11:04,640 SO PART 1 IS THE BACKGROUND FOR THE 1624 01:11:04,640 --> 01:11:04,840 WORK. 1625 01:11:04,840 --> 01:11:05,440 NEXT SLIDE. 1626 01:11:05,440 --> 01:11:08,520 SO AS HAS BEEN MENTIONED HERE BY OTHER 1627 01:11:08,520 --> 01:11:09,400 PANELISTS, SOCIAL DETERMINANTS OF SHELGT 1628 01:11:09,400 --> 01:11:13,080 SOMETHING WE HAVEN'T KNOW SEEN A LOT OF 1629 01:11:13,080 --> 01:11:15,240 WORK, KIND OF TRADITIONALLY DONE IN 1630 01:11:15,240 --> 01:11:17,040 LOOKING AT MATERNAL MORBIDITY AND SO IN 1631 01:11:17,040 --> 01:11:22,680 FACT WHAT WE USUALLY SEE IS MATERNAL 1632 01:11:22,680 --> 01:11:24,040 MORBIDITY IS CLINICIAN CLINICALS LIKE 1633 01:11:24,040 --> 01:11:26,720 IT'S MORE ABOUT A FUNCTION OF 1634 01:11:26,720 --> 01:11:29,200 PREPREGNANCY HEALTH, AGE AND OTHER RISK 1635 01:11:29,200 --> 01:11:31,640 FACTORS, RACE AND ETHNICITY IS USED TO 1636 01:11:31,640 --> 01:11:32,560 IDENTIFY HIGH RISK PEOPLE WITHOUT 1637 01:11:32,560 --> 01:11:36,840 ATTENTION TO THE MECHANISMS OF HOW 1638 01:11:36,840 --> 01:11:38,640 MARGINALIZED SOCIAL EXPERIENCE CAN 1639 01:11:38,640 --> 01:11:40,640 UNDERPIN THE RELATIONSHIP BETWEEN RACE 1640 01:11:40,640 --> 01:11:43,000 AND MORBIDITY AND MORPALTALLITY. 1641 01:11:43,000 --> 01:11:44,320 AND SOCIAL FACTORS IF THEY'RE INCLUDED 1642 01:11:44,320 --> 01:11:47,240 AT ALL IN STUDIES ARE OFTEN INCLUDED 1643 01:11:47,240 --> 01:11:48,480 ONLIA AS CO VALID AND RELIABLEIATES 1644 01:11:48,480 --> 01:11:50,520 MAKING IT DIFFICULT TO TRACE THEIR 1645 01:11:50,520 --> 01:11:53,680 INTERPLAY. 1646 01:11:53,680 --> 01:11:54,120 NEXT SLIDE, PLEASE. 1647 01:11:54,120 --> 01:11:56,400 SO YOU'VE ALREADY SEEN INFORMATION ABOUT 1648 01:11:56,400 --> 01:11:58,560 THE SOCIAL DETERMINANTS OF HEALTH, THIS 1649 01:11:58,560 --> 01:12:00,320 IS AS IT'S DEFINED BY THE INSTITUTE OF 1650 01:12:00,320 --> 01:12:03,560 MEDICINE AND SO WHAT DR. ERICSON IS I 1651 01:12:03,560 --> 01:12:05,600 WERE INTERESTED IN DOING WAS TO TAKE 1652 01:12:05,600 --> 01:12:07,360 ADVANTAGE OF THE NEW MOMS TO BE DATA SET 1653 01:12:07,360 --> 01:12:09,720 TO LOOK AT THE SOCIAL DETERMINE NABTS OF 1654 01:12:09,720 --> 01:12:13,960 HEALTH AND SEE HOW THEY WOULD HELP 1655 01:12:13,960 --> 01:12:15,440 PREDICT MATERNAL MORBIDITY OUTCOMES. 1656 01:12:15,440 --> 01:12:16,200 NEXT SLIDE. 1657 01:12:16,200 --> 01:12:16,440 PLEASE. 1658 01:12:16,440 --> 01:12:18,200 WHAT DREW OUR ATTENTION TO THE NEW MOMS 1659 01:12:18,200 --> 01:12:20,680 TO BE DATA SET IS THAT IT IS VERY RICH 1660 01:12:20,680 --> 01:12:24,520 IN THE SOCIAL DETERMINANTS OF HEALTH 1661 01:12:24,520 --> 01:12:26,040 VARIABLES UNLIKE MANY OTHER CLINICAL 1662 01:12:26,040 --> 01:12:29,080 DATA SETS, IT ALSO INCLUDED THE RACE OF 1663 01:12:29,080 --> 01:12:30,040 PARTICIPANTS FROM DIFFERENT 1664 01:12:30,040 --> 01:12:30,840 PERSPECTIVES. 1665 01:12:30,840 --> 01:12:32,920 SO IN THIS AT THAT TIME DATA SET THERE 1666 01:12:32,920 --> 01:12:35,760 ARE PEOPLE WHO WOULD IDENTIFY ANCESTRY, 1667 01:12:35,760 --> 01:12:38,040 IDENTIFY HOW THEY SELF-IDENTIFIED WITH 1668 01:12:38,040 --> 01:12:40,240 RACE AND ETHNICITY AND FINALLY, THEY 1669 01:12:40,240 --> 01:12:42,680 IDENTIFIED HOW OTHER PEOPLE IN THE 1670 01:12:42,680 --> 01:12:43,640 UNITED STATES PERCEIVED THEIR RACE. 1671 01:12:43,640 --> 01:12:45,560 AND IN FACT IT WAS THAT LAST VARIABLE 1672 01:12:45,560 --> 01:12:49,960 THAT WE USE INDEED OUR ANALYSIS BECAUSE 1673 01:12:49,960 --> 01:12:51,360 THERE'S GOOD EVIDENCE THAT IT IS MORE 1674 01:12:51,360 --> 01:12:54,240 HOW PEOPLE ARE TREATED BASED UPON HOW 1675 01:12:54,240 --> 01:12:56,200 OTHER PEOPLE PERCEIVE THEIR RACE, THAT 1676 01:12:56,200 --> 01:12:59,320 HAS TO DO WITH THINGS LIKE HEALTH 1677 01:12:59,320 --> 01:13:01,280 BEHAVIORS FOR EXAMPLE, PEOPLE ARE LESS 1678 01:13:01,280 --> 01:13:05,040 LIKELY TO START PRENATAL CARE EARLY IN 1679 01:13:05,040 --> 01:13:07,640 PREGNANCY WHEN THEIR SKIN IS DARKER, IT 1680 01:13:07,640 --> 01:13:09,560 HAS TO GO BACK TO STRUCTURAL RACISM AND 1681 01:13:09,560 --> 01:13:11,840 WE WERE REALLY THRILLED TO HAVE THAT 1682 01:13:11,840 --> 01:13:13,760 KIND OF VARIABLE AVAILABLE TO US. 1683 01:13:13,760 --> 01:13:19,520 SO WE ALSO INCLUDED PERCEIVED 1684 01:13:19,520 --> 01:13:22,640 DISCRIMINATION AS DESCRIBED BY OTHER 1685 01:13:22,640 --> 01:13:25,080 PANELISTS AND LABOR PROCESSES AND 1686 01:13:25,080 --> 01:13:27,680 OUTCOMES, EMILY 8ITIONS FOR SDOH, NEW 1687 01:13:27,680 --> 01:13:31,440 MOM DATA SET FIRST THERE WERE NO 1688 01:13:31,440 --> 01:13:34,640 NEIGHBORHOOD LEVEL INFORMATION, SECOND 1689 01:13:34,640 --> 01:13:36,080 WE'RE NURSE-MIDWIFE RESEARCHERS SO WE'RE 1690 01:13:36,080 --> 01:13:38,040 INTERESTED IN THE CARE ENVIRONMENT AND 1691 01:13:38,040 --> 01:13:40,600 THE SPECIFICALLY THE IMPACT OF MIDWIFE 1692 01:13:40,600 --> 01:13:43,440 ASHES CYSTED CARE AND THAT WAS NOT 1693 01:13:43,440 --> 01:13:43,680 COLLECTED. 1694 01:13:43,680 --> 01:13:45,320 NEXT SLIDE. 1695 01:13:45,320 --> 01:13:46,120 NEXT SLIDE. 1696 01:13:46,120 --> 01:13:49,040 SO LET'S GO INTO OUR THEORETICAL MODEL, 1697 01:13:49,040 --> 01:13:49,400 NEXT SLIDE. 1698 01:13:49,400 --> 01:13:51,520 YOU CAN SEE HERE THAT IN BLUE, ARE THE 1699 01:13:51,520 --> 01:13:53,200 SOCIAL DETERMINANTS OF HEALTH, VARIABLES 1700 01:13:53,200 --> 01:13:54,960 FOR NEW MOMS TO BE. 1701 01:13:54,960 --> 01:13:57,120 WE LITRIALY WENT FROM THE IOM MODEL AND 1702 01:13:57,120 --> 01:13:59,640 TRIED TO FIND WHERE WE COULD SEE THOSE 1703 01:13:59,640 --> 01:14:02,320 SDOH VARIABLES IN THE DATA SET. 1704 01:14:02,320 --> 01:14:04,480 AND IDEALLY WHAT WE'RE THINKING HERE IS 1705 01:14:04,480 --> 01:14:07,040 THAT THE SOCIAL DETERMINANTS OF HEALTH 1706 01:14:07,040 --> 01:14:09,000 REALLY FLOWED THROUGH PREGNANCY 1707 01:14:09,000 --> 01:14:12,640 BEHAVIORS, CONDITIONS ASK LABOR AND 1708 01:14:12,640 --> 01:14:18,600 BIRTH CHARACTERISTICS TO CREATE TD 1709 01:14:18,600 --> 01:14:20,520 CONDITIONS AND MATERNAL MORBIDITY. 1710 01:14:20,520 --> 01:14:23,000 WE CREATE THAD FOR THE MORBIDITY OUTCOME 1711 01:14:23,000 --> 01:14:24,800 AND YOU CAN SEE IN THE ORANGE BOX WHAT 1712 01:14:24,800 --> 01:14:27,160 THE VARIABLES WERE WHICH WERE INCLUDED 1713 01:14:27,160 --> 01:14:27,440 IN THAT. 1714 01:14:27,440 --> 01:14:33,400 OKAY, NEXT SLIDE, WE WILL TALK ABOUT 1715 01:14:33,400 --> 01:14:34,560 METHODS. 1716 01:14:34,560 --> 01:14:36,720 WE WERE SPECIFICALLY INTERESTED IN THE 1717 01:14:36,720 --> 01:14:38,600 HEALTHY INDIVIDUALS IN THE NEW MOM DATA 1718 01:14:38,600 --> 01:14:40,640 SET, THIS IS PEOPLE WHO DIDN'T HAVE 1719 01:14:40,640 --> 01:14:41,640 CO-MORBIDITIES COMING INTO THEIR 1720 01:14:41,640 --> 01:14:44,400 PREGNANCY, IT WAS OKAY, IF THEY 1721 01:14:44,400 --> 01:14:45,160 DEVELOPED CO-MORBIDITIES, IN FACT THAT 1722 01:14:45,160 --> 01:14:47,120 WAS SOMETHING THEY WERE INTERESTED IN 1723 01:14:47,120 --> 01:14:48,640 BUT WE SELECTED FOR PEOPLE WHO WERE 1724 01:14:48,640 --> 01:14:52,040 HEALTHY IN THE OUTSET. 1725 01:14:52,040 --> 01:14:57,000 WE WANTED PEOPLE WITH TRIAL LABOR, WITH 1726 01:14:57,000 --> 01:14:58,840 NONANATIONAL LIBRARY OF MEDICINE LOWS 1727 01:14:58,840 --> 01:15:01,600 FETUS AND THAT MEANS OUR SAMPLE WAS 1728 01:15:01,600 --> 01:15:01,880 ALMOST 6000. 1729 01:15:01,880 --> 01:15:08,480 IN ACCIDENT WE USED LATENT MIX TOUR 1730 01:15:08,480 --> 01:15:10,440 MODELING TO UNCOVER LATENT CLASSES TO 1731 01:15:10,440 --> 01:15:12,720 DETERMINE VARIABLES WE USED IN THE DATA 1732 01:15:12,720 --> 01:15:13,800 SET AS INDICATORS. 1733 01:15:13,800 --> 01:15:15,920 WE CREATED OUR COMPOSITE MEASURE AS I 1734 01:15:15,920 --> 01:15:17,640 MENTIONED WHICH FOCUSED ON POSTPARTUM 1735 01:15:17,640 --> 01:15:20,000 MORBIDITIES AND WE LOOKED AT BI VALID 1736 01:15:20,000 --> 01:15:21,080 AND RELIABLEIATE MODERATION AND 1737 01:15:21,080 --> 01:15:23,800 MEDIATION TESTING AND WE DID SEM AND 1738 01:15:23,800 --> 01:15:25,640 VISUALIZATIONS TO SEE HOW THESE 1739 01:15:25,640 --> 01:15:26,800 INTERACTED AND NEXT SLIDE. 1740 01:15:26,800 --> 01:15:29,720 SO FOR OUR RESULTS, NEXT SLIDE, YOU CAN 1741 01:15:29,720 --> 01:15:31,680 SLEEP APNEA AND OBESITYY HERE'S OUR 1742 01:15:31,680 --> 01:15:33,000 PARTICIPANT SELECTION AND GROUPING. 1743 01:15:33,000 --> 01:15:34,400 THIS IS VERY SMALL SO I DON'T EXPECT YOU 1744 01:15:34,400 --> 01:15:36,640 TO SEE IT BUT IT GOES THROUGH HOW WE GOT 1745 01:15:36,640 --> 01:15:39,720 TO OUR FINAL SAMPLES, AND THE BOTTOM YOU 1746 01:15:39,720 --> 01:15:43,040 CAN SEE THAT THE RESULT AFTER LATENT 1747 01:15:43,040 --> 01:15:44,520 CLASS MODELING WHERE WE HAD 6 CLASSES 1748 01:15:44,520 --> 01:15:46,000 WHICH WERE DESCRIBED AND I WILL TELL YOU 1749 01:15:46,000 --> 01:15:48,840 MORE ABOUT THEM IN A MINUTE. 1750 01:15:48,840 --> 01:15:49,160 NEXT SLIDE. 1751 01:15:49,160 --> 01:15:51,920 SO HERE YOU CAN SEE A VISUALIZATION OF 1752 01:15:51,920 --> 01:15:55,160 HOW THESE DIFFERENT SOCIAL DETERMINANTS 1753 01:15:55,160 --> 01:15:58,040 OF HEALTH PHENOTYPES DISTRIBUTED ON 1754 01:15:58,040 --> 01:15:59,440 THEIR SOCIAL DETERMINANTS OF HEALTH. 1755 01:15:59,440 --> 01:16:01,680 SO I WILL DRAW YOUR ATTENTION TO A FEW 1756 01:16:01,680 --> 01:16:03,640 IMPORTANT DIFFERENCES FOR EXAMPLE, IN 1757 01:16:03,640 --> 01:16:07,480 THE MIDDLE IS CLASS 4, AND YOU NOTICED 1758 01:16:07,480 --> 01:16:09,160 THAT THE LIGHT BLUE KIND OF TINT ABOVE 1759 01:16:09,160 --> 01:16:12,520 THAT, THAT HAS TO DO WITH THE PERCENT OF 1760 01:16:12,520 --> 01:16:15,520 FEDERAL POVERTY LEVEL, CLASS 4 HAD 1761 01:16:15,520 --> 01:16:16,440 HIGHEST INCOME. 1762 01:16:16,440 --> 01:16:18,640 ALSO DRAW YOUR ATTENTION TO CLASS 4, YOU 1763 01:16:18,640 --> 01:16:27,800 WILL SEE THEY HAD THE LOWEST LEVELS OF 1764 01:16:27,800 --> 01:16:30,240 PROBLEMS WITH SOCIAL INSECURITY AND 1765 01:16:30,240 --> 01:16:32,440 LOWEST LEVELS OF EICATION, AND CLASS 6 I 1766 01:16:32,440 --> 01:16:36,000 WILL DRAW YOUR TENSION TO THE FAR RIGHT 1767 01:16:36,000 --> 01:16:37,720 OF THE SCREEN, THIS WAS THE GROUP OF 1768 01:16:37,720 --> 01:16:40,080 PEOPLE WITH THE LOWEST INCOME, LOWEST 1769 01:16:40,080 --> 01:16:41,720 EDUCATIONAL ATTAINMENT, MOST WERE USING 1770 01:16:41,720 --> 01:16:44,720 GOVERNMENT HEALTH INSURANCE AND HAD THE 1771 01:16:44,720 --> 01:16:46,240 HIGHEST LEVELS OF STRESS OF ALL, AND 1772 01:16:46,240 --> 01:16:48,920 THEN I WILL DRAW YOUR ATTENTION TO CLASS 1773 01:16:48,920 --> 01:16:51,480 2, THIS WAS THE PEOPLE WHO WERE MOST 1774 01:16:51,480 --> 01:16:53,800 RECENTLY IMMIGRATED TO THE U.S. 1775 01:16:53,800 --> 01:16:57,000 OVER HALF OF THEM DID NOT SPEAK ENGLISH 1776 01:16:57,000 --> 01:17:00,240 WELL, AND THEY DID HAVE 1 OF THE HIGH 1777 01:17:00,240 --> 01:17:01,800 LEVELS OF SOCIAL SUPPORT. 1778 01:17:01,800 --> 01:17:02,760 NEXT SLIDE, PLEASE. 1779 01:17:02,760 --> 01:17:05,320 SO WE ALSO SPLIT THESE GROUPS OUT TO 1780 01:17:05,320 --> 01:17:07,800 SHOW PRESENTING RACE AND EGHT INISSITY 1781 01:17:07,800 --> 01:17:10,360 BY SELF-REPORT AND CAN YOU SEE THAT 1782 01:17:10,360 --> 01:17:13,040 CLASS 4 WHICH WAS THE CLASS OF PEOPLE 1783 01:17:13,040 --> 01:17:18,120 THAT WERE THE MOST SOCIALLY SECURE, 94% 1784 01:17:18,120 --> 01:17:19,800 WERE WHITE PRESENTING BY CONTRAST CLASS 1785 01:17:19,800 --> 01:17:21,560 6 WHICH YOU WILL REMEMBER PEOPLE WHO 1786 01:17:21,560 --> 01:17:23,160 WERE THE YOUNGEST, THEY ALSO WERE THE 1787 01:17:23,160 --> 01:17:25,640 PEOPLE WITH THE LOWEST EDUCATIONAL 1788 01:17:25,640 --> 01:17:29,960 ATAIBMENT AND THE LOWEST INCOME, WAS 1789 01:17:29,960 --> 01:17:35,240 SPLIT SOMEWHAT EQUALLY BETWEEN BLACK 1790 01:17:35,240 --> 01:17:36,960 PRESENTING HISPANIC PRESENTING AND AND 1791 01:17:36,960 --> 01:17:37,400 WHITE PRESENTING. 1792 01:17:37,400 --> 01:17:39,040 CLASS 2 WHICH I DRAW YOUR ATTENTION TO 1793 01:17:39,040 --> 01:17:41,040 EARLIER WHICH WAS THE CLASS OF PEOPLE 1794 01:17:41,040 --> 01:17:42,080 THAT WERE MOST RECENTLY IMMIGRATED TO 1795 01:17:42,080 --> 01:17:46,000 THE UNITED STATES, YOU CAN SEE WAS 1796 01:17:46,000 --> 01:17:48,800 PREDOMINANTLY WHAT LITTLE OVER HALF OR 1797 01:17:48,800 --> 01:17:49,360 HISPANIC PRESENTING. 1798 01:17:49,360 --> 01:17:51,680 WE ALSO HAD 24% WHICH WERE WHITE 1799 01:17:51,680 --> 01:17:53,920 PRESENTING AND 19% WHICH WERE ASIAN 1800 01:17:53,920 --> 01:17:57,400 PRESENTING. 1801 01:17:57,400 --> 01:17:57,880 NEXT SLIDE, PLEASE. 1802 01:17:57,880 --> 01:18:00,600 OKAY, SO THIS GIVES YOU 1 OF THE FINAL 1803 01:18:00,600 --> 01:18:02,040 SLIDES FOR OUR OUR FINDINGS SO ON THE 1804 01:18:02,040 --> 01:18:05,800 LEFT, CAN YOU SEE IN THE PIE CHART HOW 1805 01:18:05,800 --> 01:18:09,800 THESE CLASSES DISTRIBUTED OVER ENTIRE 1806 01:18:09,800 --> 01:18:11,600 SAMPLE OF HEALTHY NULLIPOROUS PEOPLE, 1807 01:18:11,600 --> 01:18:14,320 AND YOU CAN SEE AT A GLANCE ALMOST 50% 1808 01:18:14,320 --> 01:18:17,320 OF THE WHOLE SAMPLE WAS CLASS 4, THAT 1809 01:18:17,320 --> 01:18:20,600 GROUP OF VERY SOCIALLY SECURE, MOSTLY 1810 01:18:20,600 --> 01:18:23,120 WHITE PRESENTING PEOPLE, 30% OF OUR 1811 01:18:23,120 --> 01:18:26,320 SAMPLE WAS CLASS 6, THAT WAS THE YOUNGER 1812 01:18:26,320 --> 01:18:27,800 LESS EDUCATED PEOPLE AND YOU COULD SEE 1813 01:18:27,800 --> 01:18:30,000 THE OTHER CLASSES ARE SMALLER 1814 01:18:30,000 --> 01:18:30,360 PROPORTIONS. 1815 01:18:30,360 --> 01:18:31,760 AND THEN, IT'S A LITTLE BIT HARD TO SEE, 1816 01:18:31,760 --> 01:18:34,800 BUT ON THE RIGHT, CAN YOU SEE OUR 1817 01:18:34,800 --> 01:18:37,760 PRIMARY MORBIDITY OUTCOMES BY CLASS AND 1818 01:18:37,760 --> 01:18:43,360 SO, THE FIRST SET OF BOXES CAN YOU SEE 1819 01:18:43,360 --> 01:18:46,280 HERE THAT IS THE COMPOSITE MATERNAL 1820 01:18:46,280 --> 01:18:46,760 MORBIDITY OUTCOME. 1821 01:18:46,760 --> 01:18:49,000 SO AT A GLANCE, CAN YOU SEE THE DARK 1822 01:18:49,000 --> 01:18:52,240 PURPLE IS THE CLASS 4, THE MOST SOCIALLY 1823 01:18:52,240 --> 01:18:53,840 PRIVILEGED CLASS AND YOU CAN SEE THAT 1824 01:18:53,840 --> 01:18:57,440 THEY HAVE THE LOWEST LEVEL OF COMPOSITE 1825 01:18:57,440 --> 01:18:59,960 MATERNAL MORBIDITY OUTCOME, IF YOU LOOK 1826 01:18:59,960 --> 01:19:03,360 ACROSS HERE ON THE X-AXIS YOU CAN SEE 1827 01:19:03,360 --> 01:19:06,560 FOR SEVERE PREECLAMPSIA, INFECTION, 1828 01:19:06,560 --> 01:19:09,040 HEMORRHAGE AND UNPLANNED CAESAREAN THEY 1829 01:19:09,040 --> 01:19:10,760 HAD LOWER LEVELS THERE. 1830 01:19:10,760 --> 01:19:12,640 BY CONTRAST, CAN YOU SEE THAT THE 2 1831 01:19:12,640 --> 01:19:16,320 CLASSES THAT HAD THE HIGHEST LEVEL OF 1832 01:19:16,320 --> 01:19:18,240 COMPOSITE MATERNAL MORBIDITY OUTCOMES 1833 01:19:18,240 --> 01:19:19,800 CLASS 6 WHICH IS THE ORANGE GROUP AND 1834 01:19:19,800 --> 01:19:24,360 CLASS 2 WHICH IS THE RED GROUP. 1835 01:19:24,360 --> 01:19:25,040 >> ONE MINUTE. 1836 01:19:25,040 --> 01:19:28,200 >> SO KEY TAKE AWAYS, WE DID FIND THE 1837 01:19:28,200 --> 01:19:31,360 IRPT SECTIONAL SOCIAL DETERMINANTS OF 1838 01:19:31,360 --> 01:19:32,640 HEALTH PREDICTED POSTPARTUM MORBIDITY 1839 01:19:32,640 --> 01:19:35,880 INIAL OF OUR ANALYSIS, THESE 2 DISTINCT 1840 01:19:35,880 --> 01:19:37,800 GROUPS OF PEOPLE WERE MORE LIKELY TO 1841 01:19:37,800 --> 01:19:39,640 HAVE A MORBIDITY EVENT THAT I DESCRIBED 1842 01:19:39,640 --> 01:19:40,000 BEFORE. 1843 01:19:40,000 --> 01:19:43,240 IN OUR MODERATION ANALYSIS WE FOUND MORE 1844 01:19:43,240 --> 01:19:44,480 ADVANCED SERVICEICAL DILATION AT THE 1845 01:19:44,480 --> 01:19:47,600 TIME OF HOSPITAL ADMISSION WAS 1846 01:19:47,600 --> 01:19:50,600 ASSOCIATED WITH LOWER MATERNAL MORBIDITY 1847 01:19:50,600 --> 01:19:53,480 IN CLOTHE CLASS 2 EXPW 6, AND FINALLY IN 1848 01:19:53,480 --> 01:19:56,520 OUR FINAL MODEL WE FOUND THAT ONLY 1849 01:19:56,520 --> 01:20:00,360 SOCIAL DETERMINANTS OF HEALTH AND 1850 01:20:00,360 --> 01:20:01,320 UNPLANNED CAESAREAN BIRTH ARE ASSOCIATE 1851 01:20:01,320 --> 01:20:03,440 WIDE THIS PART OF MORBIDITY. 1852 01:20:03,440 --> 01:20:05,280 NEXT STEPS WE WILL DO MORE EXPLORATION 1853 01:20:05,280 --> 01:20:07,200 OF THE NEW MODELS TO BE DATA SET, 1854 01:20:07,200 --> 01:20:10,680 LOOKING AT OTHER VARIABLES WE'RE ALSO 1855 01:20:10,680 --> 01:20:13,360 LOOKING TO USE LATENT CLASS ANALYSIS IN 1856 01:20:13,360 --> 01:20:15,960 THE FUTURE TO CREATE INTERSKSAL SOCIAL 1857 01:20:15,960 --> 01:20:17,760 DETERMINANTS OF HEALTH PHENOTYPES FOR 1858 01:20:17,760 --> 01:20:20,480 ANALYSIS EVER PREGNANCY OUTCOMES. 1859 01:20:20,480 --> 01:20:24,040 OUR INFORMATION THAT I SHARED HERE HAS 1860 01:20:24,040 --> 01:20:32,080 BEEN ACCEPTED FOR PUBLICATION AND SO, IF 1861 01:20:32,080 --> 01:20:34,040 YOU'RE INTERESTED IN GETTING IN TOUCH 1862 01:20:34,040 --> 01:20:36,480 WITH US AND RESOURCES ARE DOWN AT THE 1863 01:20:36,480 --> 01:20:38,800 BOTTOM FOR REFERENCE. 1864 01:20:38,800 --> 01:20:42,160 THAT'S IT FOR ME, THANK YOU. 1865 01:20:42,160 --> 01:20:44,600 >> THANK YOU DR. CARLSON AND 1866 01:20:44,600 --> 01:20:45,880 CONGRATULATIONS TO THE PUBLICATION 1867 01:20:45,880 --> 01:20:47,440 THAT'S FOURTH COMING SO DEFINITELY KEEP 1868 01:20:47,440 --> 01:20:54,080 US POSTED ON THAT AS WELL. 1869 01:20:54,080 --> 01:20:57,000 AND LOOKING AT FUTURE EXPLORATION OF THE 1870 01:20:57,000 --> 01:20:58,320 NEW DATA SET, SO. 1871 01:20:58,320 --> 01:21:00,280 >> --WE GOT IT CLEANED AT THIS POINT, WE 1872 01:21:00,280 --> 01:21:03,040 HAVE TO DO SOMETHING. 1873 01:21:03,040 --> 01:21:03,560 >> [LAUGHTER] 1874 01:21:03,560 --> 01:21:04,120 >> VERY GOOD, VERY GOOD. 1875 01:21:04,120 --> 01:21:05,000 THANK YOU. 1876 01:21:05,000 --> 01:21:07,000 WE ARE A LITTLE AHEAD OF SCHEDULE SO IF 1877 01:21:07,000 --> 01:21:13,360 ANYONE HAS ANY BURNING QUESTIONS FOR ANY 1878 01:21:13,360 --> 01:21:14,560 OF OUR PRESENTERS PLEASE POST YOUR 1879 01:21:14,560 --> 01:21:19,480 QUESTIONS AND WE WILL ASK THEM OF ALL OF 1880 01:21:19,480 --> 01:21:20,280 OUR PRESENTERS. 1881 01:21:20,280 --> 01:21:22,400 I WANT TO THANK EVERYONE AND AT THIS 1882 01:21:22,400 --> 01:21:33,240 TIME, I WANT TO MOVE FORWARD AND HAVE 1883 01:21:33,240 --> 01:21:35,760 CLOSING REMARKS FROM DR. ALLISON 1884 01:21:35,760 --> 01:21:38,960 CERNICH, I WAS JUST LOOKING TO BE 1885 01:21:38,960 --> 01:21:40,040 SURE--THERE SHE IS. 1886 01:21:40,040 --> 01:21:40,600 HI. 1887 01:21:40,600 --> 01:21:40,880 >> GREAT. 1888 01:21:40,880 --> 01:21:42,880 MAURICE, I WILL TAKE 1 MORE LITTLE PAUSE 1889 01:21:42,880 --> 01:21:47,080 BEFORE I START TALKING I APOLOGIZE MY 1890 01:21:47,080 --> 01:21:49,000 SET UP AT WORK IS NOT THE SAME SET UP 1891 01:21:49,000 --> 01:21:51,760 THAT I HAVE AT HOME SO I'M ALL OVER THE 1892 01:21:51,760 --> 01:21:55,160 PLACE TODAY, TRYING TO FIGURE OUT ALL OF 1893 01:21:55,160 --> 01:21:55,600 THE TECHNOLOGY. 1894 01:21:55,600 --> 01:22:00,320 FIRST I JUST WANT TO SAY, THANK YOU. 1895 01:22:00,320 --> 01:22:03,680 THIS WAS REALLY FUN. 1896 01:22:03,680 --> 01:22:07,800 NOT ALL THE TIME DO WE GET TO SAY, THIS 1897 01:22:07,800 --> 01:22:08,800 IS REALLY FUN. 1898 01:22:08,800 --> 01:22:11,800 FUNNY STORY FOR ALL OF YOU, WHEN WE TOOK 1899 01:22:11,800 --> 01:22:13,000 THIS ON IN 2020, I THOUGHT THIS WAS 1900 01:22:13,000 --> 01:22:14,920 GOING TO BE THE MOST COMPLICATED THING 1901 01:22:14,920 --> 01:22:19,760 THAT I DEALT WITH THAT YEAR. 1902 01:22:19,760 --> 01:22:22,040 AND I WAS WRONG BUT I THINK WHAT WAS 1903 01:22:22,040 --> 01:22:25,040 REALLY GREAT ABOUT THIS CHALLENGE WAS, 1904 01:22:25,040 --> 01:22:29,640 NUMBER 1, I GOT TO WORK WITH AN AMAZING 1905 01:22:29,640 --> 01:22:32,000 TEAM AT NICHD, SO I WANT TO FIRST SHOUT 1906 01:22:32,000 --> 01:22:34,800 OUT TO ALL OF THE FOLKS AT NICHD WHO 1907 01:22:34,800 --> 01:22:38,240 MADE THIS HAPPEN AND TO REALLY DID 1908 01:22:38,240 --> 01:22:41,800 ENGAGE IN A LOT OF STRONG DISCUSSION AND 1909 01:22:41,800 --> 01:22:46,520 CONFLICT AND WORKING THROUGH HOW TO MAKE 1910 01:22:46,520 --> 01:22:54,600 IT ALL WORK AND REALLY FOCUS ON SO TO 1911 01:22:54,600 --> 01:22:56,040 MAURICE, TO REGINA, AND THE WHOLE TEAM, 1912 01:22:56,040 --> 01:23:00,240 I WANT TO THANK TAYLOR AND BOB, BOTH 1913 01:23:00,240 --> 01:23:02,040 FROM THE CHALLENGE OFFICE AT NIH, YOU 1914 01:23:02,040 --> 01:23:05,600 WORKED THROUGH THIS AS WELL AS OUR THE 1915 01:23:05,600 --> 01:23:09,520 LAWYERS FROM THE COUNSEL AND NICHD 1916 01:23:09,520 --> 01:23:10,680 FINANCIAL MANAGEMENT BRANCH, OUR DASH 1917 01:23:10,680 --> 01:23:17,400 TEEM AND OUR PARTNERS, AND HONESTLY, THE 1918 01:23:17,400 --> 01:23:18,800 INVESTIGATOR WHO IS PROVIDED THIS DATA 1919 01:23:18,800 --> 01:23:21,440 TO MATCH THAT MADE THIS CHALLENGE 1920 01:23:21,440 --> 01:23:21,800 POSSIBLE. 1921 01:23:21,800 --> 01:23:24,200 JUST DIDN'T REALIZE HOW MANY PEOPLE WERE 1922 01:23:24,200 --> 01:23:27,480 BEHIND THE SCENES SORT OF MAKE 1923 01:23:27,480 --> 01:23:29,440 THANKSGIVING ALL GO. 1924 01:23:29,440 --> 01:23:31,880 AND THEY REALLY DID MAKE IT ALL GO SO I 1925 01:23:31,880 --> 01:23:35,320 WANT TO THANK THEM AND I WANT TO THANK 1926 01:23:35,320 --> 01:23:37,360 OUR COMMUNICATIONS TEAM FOR ALL OF THE 1927 01:23:37,360 --> 01:23:39,120 WEB PLATFORM AND ALL THE INFORMATION 1928 01:23:39,120 --> 01:23:40,160 THROUGHOUT THE PROCESS AND THANK TO YOU 1929 01:23:40,160 --> 01:23:42,480 RECEIVE THAT AND ALL OF THE TEAM AND 1930 01:23:42,480 --> 01:23:42,920 COMMUNICATIONS TEAM. 1931 01:23:42,920 --> 01:23:45,720 I ALMS WANT TO THANK OW PARTNERS AT 1932 01:23:45,720 --> 01:23:47,800 NASA, AND THE COLLABORATIVE CENTER OF 1933 01:23:47,800 --> 01:23:50,120 EXCELLENCE, THEY WERE TRULY HELPING US 1934 01:23:50,120 --> 01:23:51,800 THE WHOLE WAY THROUGH WITH OUR FIRST 1935 01:23:51,800 --> 01:23:53,400 CHALLENGE AND IT WAS COMPLICATED AND 1936 01:23:53,400 --> 01:23:56,320 THEY REALLY DID AN OUTSTANDING JOB OF 1937 01:23:56,320 --> 01:23:59,440 HOLDING OUR HANDS ESPECIALLY 1938 01:23:59,440 --> 01:24:02,320 [INDISCERNIBLE] AND OUR PARTNERS AT 1939 01:24:02,320 --> 01:24:03,480 FREELANCER AND [INDISCERNIBLE], I CANNOT 1940 01:24:03,480 --> 01:24:05,480 SAY ENOUGH ABOUT HOW PATIENT THEY WERE 1941 01:24:05,480 --> 01:24:07,920 WITH US, ESPECIALLY WITH WE WERE 1942 01:24:07,920 --> 01:24:08,400 FLAILING. 1943 01:24:08,400 --> 01:24:10,280 SO, WHAT'S GREAT IS LOOKING AT THE END 1944 01:24:10,280 --> 01:24:12,880 RESULT AND I WILL SAY THE INVESTIGATORS 1945 01:24:12,880 --> 01:24:14,440 WHO PRESENTED TODAY, I WANT TO SAY THANK 1946 01:24:14,440 --> 01:24:17,520 YOU FOR TAKING THIS CHANCE AND DOING 1947 01:24:17,520 --> 01:24:19,400 THIS WORK AND NONAPOPTOTIC TD NOT 1948 01:24:19,400 --> 01:24:22,040 GETTING PAID TO DO IT UNTIL THE END. 1949 01:24:22,040 --> 01:24:24,800 LOTS OF PEOPLE SORT OF WAIT AROUND BUT 1950 01:24:24,800 --> 01:24:27,440 YOU KNOW YOU ALL TOOK THIS DATA AND SAID 1951 01:24:27,440 --> 01:24:29,360 HOW CAN WE MAKE A DIFFERENCE IN PUBLIC 1952 01:24:29,360 --> 01:24:32,760 HEALTH AND SO, YOU KNOW DID WE GET AS AN 1953 01:24:32,760 --> 01:24:34,560 INSTITUTE WHAT WE THOUGHT WE WOULD GET 1954 01:24:34,560 --> 01:24:35,840 OUT OF THIS CHALLENGE? 1955 01:24:35,840 --> 01:24:37,480 WELL AFTER WATCHING THE PRESENTATIONS 1956 01:24:37,480 --> 01:24:40,560 TODAY, YOU KNOW NUMBER 1 DID WE ENGAGE 1957 01:24:40,560 --> 01:24:42,160 NEW PEOPLE FROM NEW PLACES WITH THIS 1958 01:24:42,160 --> 01:24:42,800 DATA? 1959 01:24:42,800 --> 01:24:46,440 I WILL SAY RESOUNDINGLY YES, WE DID. 1960 01:24:46,440 --> 01:24:48,080 AND I JUST REALLY WAS EBS SIGHTED TO SEE 1961 01:24:48,080 --> 01:24:51,080 SOME OF OUR PUBLIC AND CORPORATE 1962 01:24:51,080 --> 01:24:53,040 PARTNERS COMING IN AND LOOKING AND 1963 01:24:53,040 --> 01:24:55,280 SAYING WOW, YOU HAVE DATA WE CAN USE TO 1964 01:24:55,280 --> 01:24:56,040 SOLVE THE PROBLEMS. 1965 01:24:56,040 --> 01:24:58,240 SO THIS WAS A GREAT RESULT. 1966 01:24:58,240 --> 01:25:02,400 DID WE SEE NEW MODELS AND NEW 1967 01:25:02,400 --> 01:25:03,440 APPROACHES, WE DID. 1968 01:25:03,440 --> 01:25:07,120 I WILL SAY THAT SOME OF THESE I THINK 1969 01:25:07,120 --> 01:25:08,640 REALLY COULD ADVANCE OUR EXPERIENCE IF 1970 01:25:08,640 --> 01:25:10,920 WE LOOK AT NEW DATA MODELS AND NEW 1971 01:25:10,920 --> 01:25:14,040 APPROACHES TO THESE DATA THAT HELP US 1972 01:25:14,040 --> 01:25:16,640 UNDERSTAND THINGS LIKE, 1973 01:25:16,640 --> 01:25:18,040 INTERSECTIONALITY OR CLUSTERED RISK. 1974 01:25:18,040 --> 01:25:21,640 AND HELP US MAKE PREDICTIVE MODELS THAT 1975 01:25:21,640 --> 01:25:23,480 BETTER IDENTIFY WOMEN WHO ARE GOING TO 1976 01:25:23,480 --> 01:25:27,080 BECOME VERY SICK DURING PREGNANCY OR 1977 01:25:27,080 --> 01:25:29,640 PEOPLE WHO MAY BECOME VERY SICK DURING 1978 01:25:29,640 --> 01:25:30,000 PREGNANCY. 1979 01:25:30,000 --> 01:25:34,040 AND I THINK FINALLY, DID WE SEE PEOPLE 1980 01:25:34,040 --> 01:25:35,360 APPROACH, QUESTIONS AROUND ETHNIC 1981 01:25:35,360 --> 01:25:39,200 DISPARITIES AND THE ANSWER WAS YES, 1982 01:25:39,200 --> 01:25:40,160 HEALTH DISPARITIES DID GET GOOD 1983 01:25:40,160 --> 01:25:42,600 ATTENTION AND I WAS REALLY EXCITED TO 1984 01:25:42,600 --> 01:25:45,240 SEE SOME OF YOU COMBINING THINGS AND ON 1985 01:25:45,240 --> 01:25:47,040 THE OTHER HANDING INTERSECTIONALITY AS 1986 01:25:47,040 --> 01:25:50,360 IT RELATES TO SOCIAL DETERMINANTS OF 1987 01:25:50,360 --> 01:25:52,640 HEALTH THAT UNDERLIE MATERNAL MORBIDITY. 1988 01:25:52,640 --> 01:25:55,560 SO, YOU KNOW DID WE MEET THE OBJECTIVES 1989 01:25:55,560 --> 01:26:01,000 THAT WE SET OUT WHEN WE SAID WE WOULD 1990 01:26:01,000 --> 01:26:01,440 HAVE THIS CHALLENGE? 1991 01:26:01,440 --> 01:26:03,200 I THINK WE DID AND WE PROVED THAT YOU 1992 01:26:03,200 --> 01:26:07,080 CAN USE DATA USING SECONDARY ANALYSIS 1993 01:26:07,080 --> 01:26:08,960 THAT ARE CONSENT TO DATA THAT YOU CAN 1994 01:26:08,960 --> 01:26:15,400 SUH USE THOSE TO REALLY GET INNOVATIVE 1995 01:26:15,400 --> 01:26:15,640 MODELS. 1996 01:26:15,640 --> 01:26:16,920 SO, I CANNOT SAY ENOUGH ABOUT OUR TEAM 1997 01:26:16,920 --> 01:26:18,800 AND WHAT THEY DID AND CANNOT SAY ENOUGH 1998 01:26:18,800 --> 01:26:20,840 ABOUT THE INVESTIGATOR GROUP THAT 1999 01:26:20,840 --> 01:26:21,240 COMPETED IN THIS. 2000 01:26:21,240 --> 01:26:24,120 AND I JUST WANT TO SAY THANK YOU FROM 2001 01:26:24,120 --> 01:26:27,160 OUR INSTITUTE AND FROM THE NIH 2002 01:26:27,160 --> 01:26:32,000 OVERARCHING THERE WAS AN EXCITING 2003 01:26:32,000 --> 01:26:32,320 PROJECT. 2004 01:26:32,320 --> 01:26:34,000 AND A REAL ACTUAL CHALLENGE TO US AND 2005 01:26:34,000 --> 01:26:35,200 WHAT WE DO. 2006 01:26:35,200 --> 01:26:38,600 SO, AGAIN, THANKS SO MUCH TO EVERYBODY 2007 01:26:38,600 --> 01:26:40,840 AND MAURICE, I WILL TURN IT BACK OVER TO 2008 01:26:40,840 --> 01:26:41,040 YOU. 2009 01:26:41,040 --> 01:26:43,240 >> THANK YOU VERY MUCH. 2010 01:26:43,240 --> 01:26:48,400 I WOULD ALSO LIKE TO ECHO ALLISON'S 2011 01:26:48,400 --> 01:26:53,000 SENTIMENT SAYING THANK TO YOU THE ENTIRE 2012 01:26:53,000 --> 01:26:55,800 TEAM FOR EVERYTHING YOU'VE DONE. 2013 01:26:55,800 --> 01:27:02,400 ALSO A SPECIAL THANK YOU TO OUR WEBINAR 2014 01:27:02,400 --> 01:27:04,040 PARTICIPANTS WHO ACTUALLY HELPED PUT 2015 01:27:04,040 --> 01:27:06,240 THIS TOGETHER AS WELL AS OUR SIGN 2016 01:27:06,240 --> 01:27:08,840 LANGUAGE INTERPRETERS AS WELL. 2017 01:27:08,840 --> 01:27:11,240 AND THANKS EVERYONE FOR ATTENDING. 2018 01:27:11,240 --> 01:27:12,480 THERE IS 1 QUESTION THAT JUST CAME IN 2019 01:27:12,480 --> 01:27:18,680 AND I WANT TO TAKE A LOOK BEFORE I 2020 01:27:18,680 --> 01:27:20,080 CONCLUDE OUR MEETING, AND I'M NOT SURE 2021 01:27:20,080 --> 01:27:25,240 WHO THIS QUESTION IS DIRECTED TO BUT IF 2022 01:27:25,240 --> 01:27:27,920 ANY OF OUR PRESENTERS WOULD LIKE TO 2023 01:27:27,920 --> 01:27:29,360 ANSWER THIS QUESTION, THAT WOULD BE 2024 01:27:29,360 --> 01:27:29,640 GREAT. 2025 01:27:29,640 --> 01:27:32,640 DO HAVE YOU ANY LESSONS LEARNED FROM 2026 01:27:32,640 --> 01:27:39,360 DATA MANAGEMENT FROM THE DATA SET WOULD 2027 01:27:39,360 --> 01:27:40,120 BE HELPFUL? 2028 01:27:40,120 --> 01:27:43,320 >> I'M HAPPY TO SAY SOMETHING? 2029 01:27:43,320 --> 01:27:44,960 >> A REPRODUCIBLE FLOW OF DATA 2030 01:27:44,960 --> 01:27:47,640 MANAGEMENT IS ABSOLUTELY ESSENTIAL. 2031 01:27:47,640 --> 01:27:49,240 THIS IS A LOT OF LITTLE PIECES OF DAILY 2032 01:27:49,240 --> 01:27:50,680 BASIS THEA SET THAT HAD TO BE PUT 2033 01:27:50,680 --> 01:27:54,360 TOGETHER AND SO BEING ABLE TO DO THAT IN 2034 01:27:54,360 --> 01:27:55,560 A REPRODUCIBLE MANNER WAS SUPER HELPFUL 2035 01:27:55,560 --> 01:27:59,280 TO BE ABLE TO SEE IT ALL SO THAT WAS OUR 2036 01:27:59,280 --> 01:28:01,440 BIG TAKE AWAY. 2037 01:28:01,440 --> 01:28:01,720 >> GREAT. 2038 01:28:01,720 --> 01:28:05,440 THANK YOU SO MUCH FOR THAT DR. CARLSON. 2039 01:28:05,440 --> 01:28:06,840 AGAIN, THANKS EVERYONE FOR 2040 01:28:06,840 --> 01:28:09,480 PARTICIPATING, WE ARE ENDING A LITTLE 2041 01:28:09,480 --> 01:28:12,360 EARLY BUT I THINK YOU CAN TAKE ADVANTAGE 2042 01:28:12,360 --> 01:28:14,920 OF THE NICE WEATHER THAT WE HAVE AT 2043 01:28:14,920 --> 01:28:17,360 LEAST IN THE D. C. METROAREA, WE HAVE 2044 01:28:17,360 --> 01:28:20,400 NICE WEATHER HERE SO THANK YOU EVERYONE 2045 01:28:20,400 --> 01:28:24,280 FOR PARTICIPATING AND WE WILL--THIS WILL 2046 01:28:24,280 --> 01:28:27,400 CONCLUDE OUR WINNERS WEBINAR FOR THE 2047 01:28:27,400 --> 01:28:27,760 DATA CHALLENGE. 2048 01:28:27,760 --> 00:00:00,000 THANK YOU VERY MUCH.