Applying Survival Analysis in Adolescent Pregnancy Risk Assessment

While scrolling through the Philippine News Agency on a post-Valentine weekend, I came across a report that caught my attention - the government was alarmed over the rising cases of teenage pregnancy in the country. The article raised concerns about the effectiveness of existing programs and called for data-driven solutions to understand and mitigate the problem.

Given that Valentine's Day had just passed, and with this concern fresh in my mind, I decided to play with the data. I started examining trends using the Demographic and Health Survey (DHS) dataset, trying to frame a research question that could shed light on the patterns and predictors of adolescent pregnancy risk.

After an initial exploratory analysis, I decided that a Survival Analysis approach would be best suited for this study. Survival models allow us to analyze the timing of pregnancy and estimate how long adolescents remain pregnancy-free, while also identifying risk factors that accelerate early pregnancy.

Statistical Methods

To systematically analyze adolescent pregnancy risks, I propose using the following statistical methods:

1) Kaplan-Meier Survival Analysis to help visualize the probability of remaining pregnancy-free over time. It will allow us to compare different groups (e.g. wealth quintiles, education levels) and see disparities in survival curves.

2) Cox Proportional Hazards Model to identify the key socioeconomic and demographic predictors influencing early pregnancy risk. Hazard ratios (HR) will provide insights into how variables such as marital status, education level, and household wealth contribute to pregnancy timing.

3) Competing Risks Model to consider a competing risks framework to separate the risks of pregnancy within marriage versus outside of marriage (given that marriage is a strong predictor of adolescent pregnancy).

Data Preparation and Study Design

The analysis will use the Philippine DHS dataset focusing on adolescent females. Right-censoring will be accounted for, as not all adolescents in the dataset may have experienced pregnancy at the time of data collection. Sensitivity analyses will be conducted to check model assumptions and robustness.

Next Steps: Collaboration and Refinement

Now that the statistical framework is in place, I am actively looking for co-authors who are interested in refining the research questions, discussing policy implications, and contributing to the literature review.

If you are interested in co-authoring, providing expertise, or collaborating on this study, feel free to reach out!

Email me

Whether you have a request, a query, or want to work with me, use the form below to get in touch.