Journal of Data Science

One of the main features of bipolar disorder is repletion of relapse overtime. Many studies have focused on time-to-first relapse using the most popular Cox proportional hazard model which discards subsequent information on recurrent relapses. The aim of this study was to identify some risk factors of time-to-recurrent relapses in bipolar disorder inpatients by using appropriate recurrent event model. Data on 206 inpatients, available at Amanuel mental specialized hospital, were collected by reviewing the medical records from September 11, 2013 to March 12, 2019. Different extended cox proportional hazard models including AG, PWP-TT, PWP-GT and semiparametric shared gamma frailty models were used. R package FrailtyEM package used to fit semi-parametric shared gamma frailty models through EM algorithm. The mean age of the patients was 33.33 years. Within the study time, a total of 418 inpatient admissions (relapses) were registered for 206 inpatients. Among these admissions, about 49.3% of the patients had first relapse and 50.7% of the patients had more than one relapses. The likelihood test results indicated that the appropriate model is the gap-time based semi-parametric shared gamma frailty model and the important risk factors that have effect on time since the end of the most recent relapse to the start of the next relapses are marital status, substance abuse, employment status and residence. Recurrent relapse may be reduced by giving more intensive forms of treatment and creating awareness on each risk factor.

Abstract: Observational studies of relatively large data can have potentially hidden heterogeneity with respect to causal effects and propensity scores–patterns of a putative cause being exposed to study subjects. This underlying heterogeneity can be crucial in causal inference for any observational studies because it is systematically generated and structured by covariates which influence the cause and/or its related outcomes. Addressing the causal inference problem in view of data structure, machine learning techniques such as tree analysis can be naturally necessitated. Kang, Su, Hitsman, Liu and Lloyd-Jones (2012) proposed Marginal Tree (MT) procedure to explore both the confounding and interacting effects of the covariates on causal inference. In this paper, we extend the MT method to the case of binary responses along with a clear exposition of its relationship with established causal odds ratio. We assess the causal effect of dieting on emotional distress using both a real data set from the Lalonde’s National Supported Work Demonstration Analysis (NSW) and a simulated data set from the National Longitudinal Study of Adolescent Health (Add Health).