Predicting Unmet Need for Family Planning Among Women of Reproductive Age in Zambia Using Explainable Machine Learning: Evidence from the 2024 DHS
Using explainable machine learning on the 2024 Zambia Demographic and Health Survey, this study developed a leakage-aware ensemble model that effectively stratifies the risk of unmet family planning need among Zambian women, identifying work status, geography, and mobile phone ownership as key predictors to guide targeted reproductive health interventions.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
For millions of women across the globe, the ability to plan if and when to have children remains out of reach, not because they do not want it, but because the right information, services, or support are missing. This gap between a woman's desire to control her family size and her actual access to contraception is known as "unmet need." It is a silent barrier that shapes lives, affecting education, economic stability, and the health of mothers and children alike. In recent years, global health leaders have turned to data to solve this puzzle, hoping to find patterns that can guide resources to the women who need them most. The challenge lies in distinguishing between women who are simply not interested in family planning and those who want it but cannot get it. To do this effectively, researchers must look at a complex web of factors: where a woman lives, her job, her age, her marital status, and even her access to technology. The goal is to move beyond broad averages and identify specific groups of women who are most likely to be falling through the cracks of the healthcare system.
In Zambia, a team of researchers recently tackled this challenge using a massive, nationally representative survey conducted in 2024. They set out to build a digital tool capable of predicting which women of reproductive age are most likely to have an unmet need for family planning. Instead of relying on traditional statistical methods that often struggle with complex, real-world interactions, the team employed a sophisticated approach called machine learning. Think of this process as teaching a computer to recognize a specific pattern by showing it thousands of examples, rather than asking it to follow a single, rigid rule. The researchers fed the computer detailed information about nearly 10,000 women, including their age, education, whether they work, if they own a mobile phone, and their province of residence. Crucially, they designed the system to be "leakage-aware," meaning they strictly prevented the computer from using the very answers it was supposed to predict. This ensured the tool was learning to spot genuine risk factors rather than simply memorizing the definition of the problem.
The results of this study offer a clear, though modest, picture of how to improve family planning efforts in Zambia. The machine learning model successfully identified women with unmet needs with a level of accuracy that is reliable enough to be useful for public health planning, though it is not perfect. When tested on a fresh set of data it had never seen before, the model correctly flagged about 72 percent of the women who actually had an unmet need. While it also incorrectly flagged some women who did not need help, the researchers found that the benefits of catching the majority of those who do outweighed the cost of the false alarms. The study revealed that the most powerful indicators for predicting unmet need were surprisingly practical: whether a woman is currently working, where she lives, and whether she owns a mobile phone. Women who were not employed, those living in certain provinces like North Western and Northern, and those without a mobile phone were significantly more likely to be missing out on family planning services.
The researchers also mapped these risks across the country's ten provinces, uncovering a stark geographic divide. The burden of unmet need was highest in the North Western, Northern, and Lusaka provinces, where nearly one in four women faced this barrier. In contrast, the Central, Eastern, and Copperbelt provinces showed lower rates. This geographic variation suggests that a single, national strategy may not be enough; instead, health programs need to be tailored to the specific realities of each region. The study also highlighted that the problem extends beyond married women. By including sexually active unmarried women in their analysis, the researchers found that the unmet need is a widespread issue affecting a broad spectrum of the population, aligning with global goals to ensure universal access to reproductive health.
Perhaps the most important takeaway from this work is not just the list of who is at risk, but how the researchers arrived at that conclusion. They emphasized that their tool is designed for prioritization, not for determining the root causes of why a woman lacks access. The model acts as a spotlight, illuminating the groups that require immediate attention and resources, such as targeted outreach to non-working women or digital campaigns for phone owners. The authors caution that while the tool is effective at spotting patterns, it does not prove that being unemployed or living in a specific province causes the lack of access. Instead, it provides a trustworthy map for health officials to navigate their limited resources more efficiently. By combining these data-driven insights with on-the-ground knowledge, Zambia can better align its efforts with global health goals, ensuring that the women who need family planning the most are the ones who receive it.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.