Modeling the Risk Factors of Miscarriage Using Censored Quantile Regression in Kenyan Cohort

John Olives Chibayi *

Department of Mathematics and Statistics, Kaimosi Friends University, Kenya.

*Author to whom correspondence should be addressed.


Abstract

Background: Numerous researches have been undertaken to delineate the significant risk factors linked to miscarriage. The etiology of miscarriage has presented challenges in its elucidation, necessitating a comprehensive understanding of the elements that are linked to this phenomenon. The examination of these aspects has been conducted utilizing diverse methodological techniques and analytical procedures. The Cox's proportional hazards model and the accelerated failure time models are commonly utilized in the field of survival analysis. While these models provide a certain level of flexibility, they do not allow for the differentiation of predictor indicators between high-risk and low-risk subpopulations. In contrast, quantile regression allows for the assessment of separate effects of covariates at different quantiles of the conditional distribution of miscarriage.

Objective: This study assessed the risk factors for miscarriage using the censored quantile regression model and compared the results to that of Cox proportional hazard model.

Methodology: A secondary data from 6077 records of pregnant women with recognized pregnancy and enrolled in the pre-natal care in Kakamega County General Teaching and Referral Hospital (KCGTRH) in Kakamega county, western Kenya during the period from 1 January, 2019 up to 31 October, 2020 were recruited into the study. The study used chi-squared tests for descriptive analysis. Cox proportional hazards and censored quantile regression were used for modeling. Analysis was done using the R package and 5% was the level of significance.

Results: Censored quantile regressions were conducted to estimate the parameters for the quantiles, \(\tau\) = 0.05, 0.1, 0.15, 0.2. The results showed that ethnicity had a positive effect and was statistically significant at \(\tau\) = 0.1 (p = 0.0023) and number of previous stillbirths had a negative effect and was significant at \(\tau\) = 0.15 (p = 0.032) and at \(\tau\) = 0.2 (p = 0.036).

Graphical comparison of censored quantile model effects to that of cox PH model revealed that the variables; prior miscarriages, prior stillbirths, place of residence, and number of ANC visits have a statistically significant adverse effect on the survival time of miscarriages in the initial phases of pregnancy, contrary to cox PH model.

Conclusion: No variable had a statistically significant influence across all quantiles. The results depicts that at a certain moment in the monitoring procedure, they will have a significant impact on the length of survival for miscarriages. A quantile regression model have the capacity to incorporate cross-over effects, whereby the influence of a covariate can manifest as either positive or negative. In contrast, Cox estimates do not allow for the potential scenario when a covariate initially increases the hazard for a specific time period and subsequently decreases it.

Keywords: Quantile regression, miscarriage, modeling, statistically


How to Cite

Chibayi, John Olives. 2024. “Modeling the Risk Factors of Miscarriage Using Censored Quantile Regression in Kenyan Cohort”. Asian Journal of Probability and Statistics 26 (11):124-43. https://doi.org/10.9734/ajpas/2024/v26i11676.