Ridwan Al Aziz¹, Hasan Symum^2,, Nafisa Mahbub³, Abdullahil Azeem¹

Most studies on healthcare process monitoring focus on surveillance end-stage clinical outcomes and thus may ignore interaction from earlier stages. For a complex multi-stage healthcare process such as maternal delivery in hospital settings, ignoring what occurs in earlier stages of the procedure may inadvertently lead to poorer outcomes, if continued. Therefore, this study employs a holistic approach to developing a Bayesian network-based monitoring process by considering interaction among multiple maternal healthcare stages. A total of fourteen outcome variables, three process variables, and eight risk factors from antepartum and subsequent periods are considered. Data are obtained from the Dhaka Medical College Hospital and Rajshahi Medical hospitals of Bangladesh from March 2017 to May 2017. The Bayesian network parameters were estimated using logistic regression models for each outcome variable and then simulated for monitoring by multi-stage exponentially weighted moving average control charts. Our findings demonstrate that different variables of the antepartum period and other new variables incorporated in this paper are crucial in evaluating the risk for pregnant mothers and infants. This corroborates the significance of tracing back to earlier stages to identify the causes for an abnormal outcome and provides a clearer understanding of the potential sources of excess variation that lead to a shift in final stage outcomes.

maternal health, childbirth, predictive modeling, Bayesian network, control chart