Modeling the Risk Factors of Miscarriage using Advanced Survival Analysis Techniques: Cure Model

Background: A miscarriage is an abrupt and distressing occurrence that might have adverse consequences for the individual experiencing it. However the risk factors that lead normal pregnancy to a miscarriage are fairly well established. That is, its prevalence and causes is still a subject for continuing investigation. Cox's model and the accelerated failure time model are commonly utilized statistical models to assess the factors associated with spontaneous abortion. These models assume that all patients under surveillance would inevitably experience the event of interest, assuming they are monitored for a sufficiently long period of time. However, it is imperative to recognize that certain subjects may not undergo a comprehensive manifestation of the phenomenon, irrespective of the length of their investigation. Objectives: To estimate cure fraction of miscarriage and evaluate the effects of covariates on the cure rate, using cure models. Methods: This study used secondary data collected from 6077 pregnant women who were enrolled for antenatal care in Kakamega County General Teaching and Referral Hospital (KCGTRH) in Kakamega county, western Kenya. The study period was from 1 January, 2019 up to 31 October, 2020 and miscarriages were regarded as failures. Cox model and a proportional hazards mixture cure model (PHMC) were utilized to analyze the dataset. Results: The cure model showed that place of residency ( ), ethnicity: kalenjin ( ), kikuyu ( ) and luo ( ), number of prior miscarriages ( ), number of previous stillbirths ( ), and number of ANC visits ( ) statistically affect cure fraction. However these factors did not affect survival time, apart from the number of ANC visits ( ). The number of previous miscarriages, stillbirths, ANC visits, site of residency, and ethnic groups (Kalenjin, Kikuyu, and Luo) had cure probabilities of 54.9%, 47.1%, 76.1%, 87.2%, 39.3%, 44.9%, and 58.3%, respectively. All analysis was carried out using R software. The level of significance was 5%. Conclusion: The cure model in this study showed that these factors had effect on long-term survivors. On short-term trends there were little changes. Using the cure model to investigate miscarriage data provided more insights than Cox model analysis.