Antenatal Follow-Up Prioritiser: Which Pregnant Women a Health Worker Should Call This Week
A health worker with 150 pregnant women on her list cannot call all of them every week. Build a mobile app and ranking model that suggests who to contact first, based on missed visits, risk factors and engagement history, and measures whether the ranking finds women who would otherwise drop out of antenatal care. The project is about fair, explainable prioritisation under a fixed call budget, not about diagnosing anyone.
Intermediate
A small team, or one strong student willing to learn something new.1 semester · 2 students
7 to ship
4 optional extrasSuggested stack
What you should ship
- A synthetic but realistic dataset of pregnancies, scheduled visits, missed visits and calls, with documented assumptions
- A phone app listing a worker's caseload with due and overdue visits
- A weekly call list ranked by a simple rule baseline and by a learned model, with the reason for each name
- Call logging with outcome (reached, not reached, visit booked) feeding back into the ranking
- An evaluation comparing rules, the model and random choice at the same weekly call budget
- A fairness check across age, parity and distance to the clinic
- A short guide for supervisors explaining what the ranking does and does not do
If you have time left
- Voice message reminders in Urdu, Bangla, Hindi or Swahili for women who prefer them
- A bandit approach that learns which women benefit most from a call
- Integration with an existing health record system through its API
- A household visit planner that groups nearby women
The problem: too many names, too few calls
Antenatal care visits catch problems such as high blood pressure, anaemia and poor growth early. Many women in South Asia and sub-Saharan Africa start late or stop attending after one or two visits. Community health workers and helpline staff are asked to follow up, but they have long lists and limited time and phone credit. Usually they call whoever is at the top of the register, or whoever they remember.
Some women would attend anyway. Others will drop out unless someone checks in. A good weekly list puts the second group first and tells the worker why.
What already exists and the gap
- MomConnect in South Africa and Kilkari in India send pregnancy health messages to millions of women.
- mMitra, run by the Indian NGO ARMMAN, sends voice calls to pregnant women and has health workers follow up with those who stop listening.
- Health worker apps on CommCare or DHIS2 record visits and show overdue lists.
Overdue lists are simple and useful. The gap is a ranking that uses more than one date, explains itself and is tested against the budget a real worker has.
Inspiration
ARMMAN worked with Google Research on exactly this question for mMitra: with limited health worker time, which beneficiaries should get a live call. Their AAAI 2022 paper, "Field Study in Deploying Restless Multi-Armed Bandits: Assisting Non-Profits in Improving Maternal and Child Health" by Aditya Mate and colleagues, reports a large field study. Your project builds a smaller, explainable version you can test on simulated data.
What you build
- Health workers open the app on Monday and see this week's list, each name with a short reason ("missed visit 2 weeks ago, first pregnancy, not reached last call").
- Supervisors see how many women were reached and how many came back to care.
- Programme managers can compare ranking methods before changing anything in the field.
The technically interesting part
- Simulation. You will not get real records, so build a simulator with different kinds of women: steady attenders, early drop-outs, women who respond to calls, women who do not. Make the rule for who benefits from a call explicit, and change it to test your model.
- Ranking under a budget. The metric is not accuracy, it is how many extra visits happen when the worker makes N calls a week.
- Explanations a worker trusts. Show two or three plain reasons, not a score.
- Fairness. Check the list does not ignore women who live far away or who never answer, which is easy for a model to learn.
How to evaluate it
- Extra visits per 100 calls for random, rule-based and model-based lists in the simulator.
- Robustness: results when the simulator's behaviour rules change.
- Coverage: share of women contacted at least once in a month.
- Usability: a task test with five people playing health workers.
Suggested plan
- Weeks 1 to 3: read the papers and design the simulator.
- Weeks 4 to 7: the app and backend with caseload and call logging.
- Weeks 8 to 11: rule baseline, model, explanations.
- Weeks 12 to 15: evaluation, fairness check, report and demo.
Path to a product
NGOs and health departments running maternal health programmes need tools like this, and the same pattern applies to vaccination follow-up, TB treatment and school drop-out. A tested ranking module that plugs into an existing health worker app is a realistic offering, and the project shows you can apply machine learning carefully where people's health is involved.
Scope warning
The ranking only decides call order. It must never tell a worker to skip someone. Do not use real patient data without ethics approval and a partner. Keep medical advice out of the app.
Learn what you need on Sefism
- Mobile roadmap: offline lists, local storage and sync.
- Machine learning roadmap: ranking, imbalanced data and evaluation.
- Data science roadmap: simulation and honest comparisons.
- scikit-learn User Guide: the baseline models.
- Flutter documentation: the phone app.
Get help with this idea
The hard part is a simulator realistic enough that the comparison means something. Tauseef can help you design it and choose a metric that matches a worker's real week. If you are new here, create a free account first.
Ideas and guidance, not finished projects
These are project ideas and scoping guidance, published free for students to use as a starting point. I do not build, write, or sell final-year projects, and I do not complete coursework for anyone. Take an idea, make it yours, and build it.
Not sure this one fits you?
I help students pick an idea that matches their skills and their deadline, then scope it down until it is finishable. Guidance only, never done for you.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Sessions are free for approved Sefism members, and a few slots open each week.
Follow along
New writing, resources and project ideas land here first.