MyHealthyPregnancy mobile health app: Combining behavioral science and machine learning for risk communication during the peripartum period

Project Summary Abstract BackgroundPreterm birthsthose that occur prior toweeks of gestationare the leading direct cause of neonatal mortality and morbidityMore thaninbirths in the U Sare pretermwith rates that are disproportionately high among African Americans and families living in povertyregardless of raceAddressing the problem of preterm birth requires both accurate identification of the factors that put women at riskand communication of that risk to women and their healthcare providersThe reduction in preterm births is a core mission of the NICHD Pregnancy and Perinatology BranchStudy aimsApply our novel machine learning algorithmKCI neighborsto a large prospective cohort study to model adverse pregnancy outcomes among women with low income and English literacyuse mixed methods researchgrounded in decision science techniquesto tailor the MyHealthyPregnancyMHPapp to the specific needs of low income and low English literacy patientsandconduct smallscale usability testing of the modified technology with a sample of peripartum low incomelow literacy patients to determine the appandapos s acceptability and interested in targeted intervention strategiesInnovationMHP is the first mobile health app that combines machine learningexpert modelsand behavioral decision research to provide pregnant women and their providers scientifically soundhighly personalizedand actionable feedback on individual level riskThe machine learning algorithms are designed to learn from users as the app is more widely deployedallowing for the identification of new causal pathways linking risk factors to adverse pregnancy outcomesMHP targets specific engagement metricse gappointment attendanceto meet health system stakeholdersandaposgoalsenabling healthcare systems to meet performance targetsas well as decrease costs through reduced adverse outcomesMethodology and expected resultsWe will employ statistical machine learning to model the risk of adverse pregnancy outcomescomplemented by qualitative mixed methods research to identify the most important measures to include in the MHP appWe anticipate that usability testing will show an engaging appcapable of capturing and communicating risks in our target populationPotential impactThis work will advance scientific understanding of the risk factorsneedsand implementation science required to reach pregnant women who experience the most difficulty engaging in the healthcare systemMoreoverit will help improve clinical practice through development of a toolMHPthat can detect and communicate preterm birth precursor risk to both patients and providers Project Narrative Preterm birth affects millions of families in the United Stateswith severe health consequences for surviving infants and their familiesDetermining the factors that put individual women at greatest riskand communicating those factors in a decision relevant mannerrequires a deep understanding of the patient populationThe proposed research integrates individual level interviews from women in a low income and low English literacy populationwith novel statistical machine learning algorithms and a user centered design approachto deliver a personalizedengagingand effective pregnancy risk communication mobile health applicationMyHealthyPregnancy