Machine Learning-Based Prediction of Unmet Need for Family Planning Among Reproductive-Age Married Women in Nigeria
- Esther I. Aikulola-Bolaji1 & Ayo S. Adebowale2
- DOI: 10.5281/zenodo.19711699
- ISA Journal of Medical Sciences (ISAJMS)
Background: Unmet
Need for Family Planning (UMFP) remains a public health concern in developing
countries, particularly in Nigeria, where women who want to stop or delay
childbearing are not using contraception. Although, previous studies identified
factors influencing UMFP, limited evidence exists on how these factors are
prioritised and used to predict the probability of UMFP. This study prioritised
factors influencing UMFP and predicted its probability among married women of
reproductive age in Nigeria.
Methodology: This
cross-sectional study used 2003, 2008, 2013, 2018 and 2024 rounds of Nigerian
Demographic and Health Survey (NDHS), which applied two-stage cluster sampling.
Sub-samples extracted were 3,651 (2003), 17,316 (2008), 18,600 (2013), 19,318
(2018) and 16,111 (2024) currently married women of reproductive age. Logistic
regression identified significant factors influencing UMFP,while dominance
analysis (ΔR² ≥0.10) priortised their relative contributions. Decision trees,
random forest, support vector machine and KNN models were trained on
2003-2018data and tested on 2024. Model performance was assessed using
F1-score, precision, accuracy, recall and AUC. Decomposition analysis examined
contributors to changes in UMFP. All analyses were conducted at α0.05.
Result: Standardised
prevalence of UMFP increased from 29.3% (2003) to 36.0% (2024). Of 21
identified significant factors, dominant factors were Number of Surviving
Children (NSC) with ΔR² ranging from 0.43-0.49, age (ΔR²: 0.33- 0.39), ideal
number of children (ΔR²: 0.17-0.33), religion (ΔR²: 0.16-0.33), and wealth
index (ΔR²: 0.10-0.19). Random Forest (AUC=76.2%; 95% CI=75.0-77.1). Highest
predicted probability of UMFP was observed among women who live in the urban
areas, belong to Yoruba ethnic group, in the rich wealth quintile, and
practicing Christianity (p2003/2024=81.2%, p2008/2024=90.0%,
p2013/2024=81.2%, and p2018/2024=97.2%).The rise in UMFP
(rate=1.43) was mostly attributed to women’s NSC and religion.
Conclusion: Targeted, parity- sensitive and religiously responsive interventions are essential t reduce UMFP in Nigeria.
