Psychiatric problems surrounding parturition affect both the mother?s health and her child?s developmental trajectory. Peripartum depression (PPD), referring to a depressive episode occurring during pregnancy or after childbirth, is both common and morbid. PPD has been implicated in various short and long term adverse outcomes, including preterm delivery and heightened risk for mental illness in the adult offspring. In extreme cases, PPD can lead to maternal suicide and/or infanticide. Although an estimated 760,000 American women (and children) suffer from PPD each year and screening for PPD has been recommended by the USPTF, no accurate screening tool is available to adequately identify women at risk of PPD. This novel study will capitalize on the rich clinical, demographic, and laboratory information in patients? electronic medical reports (EMRs) to improve screening for PPD. We propose to implement advanced machine learning methods to build a model to optimize identification of women at risk for PPD. We we will adopt a psycho-social- biological approach of mental illness to prospectively explore the combined effect of various disease-related factors in improving the accuracy of PPD prediction. Our dataset will make use of a sample of 20,000 women who have been followed during their obstetrical care in two leading academic hospitals in Boston. We will gather information concerning socioeconomic factors, relevant obstetric factors, and mental and physical conditions in pregnancy and disease history, as derived from laboratory test results and the patient?s report. We expect our findings to advance scientific knowledge of women at risk for PPD. Our work may lead to the development of a screening protocol that is low-cost and easily performed by health providers in clinical settings. Early identification of women at risk could potentially allow targeted interventions to reduce the prevalence and morbidity of PPD in the US. This in turn could reduce treatment costs, avoid a potentially preventable disease, and improve the quality of care and health outcomes of mothers and their children. Our study accords with the NICD high priority area of research aimed at improving the health of women during and after pregnancy and improving pregnancy outcomes. The proposed project will further the NICHD mission that women suffer no harmful effects from reproductive processes, and that children achieve healthy and productive lives.
Peripartum depression (PDD) referring to the occurrence of depressive episode during pregnancy or following childbirth is a common and serious illness. While the US Preventive Services Task Force recommends universal screening for PPD, there is no accurate tool to predict women who would be at risk of suffering from PPD. The goal of this study is to test a machine learning predictive model of PPD using information in patients? medical records to decrease PPD?s prevalence and ultimately prevent it.
