sent home during the deadliest window

Problem statement

One-third of maternal deaths and 50% of maternal sepsis cases occur in the postpartum period — after hospital discharge when clinical monitoring stops. No validated wearable or remote monitoring tool exists for detecting postpartum sepsis, despite clear physiological markers (temperature, heart rate, respiratory rate) that could signal early infection. Standard sepsis screening tools (SIRS criteria, qSOFA) use vital sign thresholds developed for non-pregnant populations that are inappropriate for postpartum physiology, where baseline heart rate, temperature, and white blood cell counts are naturally elevated for weeks after delivery.

Why this matters

Maternal mortality in the US reached 32.9 deaths per 100,000 live births in 2021, with sepsis among the leading preventable causes. Most postpartum sepsis deaths are preventable with early detection and antibiotic treatment, but the current care model discharges mothers 24–72 hours after delivery with a single follow-up visit at 6 weeks — creating a surveillance gap during the highest-risk window. Disparities compound the problem: Black women are three times more likely to die from pregnancy-related causes, and rural mothers may be hours from the nearest obstetric facility.

What’s been tried and why it hasn’t worked

General-purpose wearable health monitors (Fitbit, Apple Watch) track heart rate and activity but are not validated for postpartum sepsis detection and do not incorporate pregnancy-adjusted vital sign thresholds. A resting heart rate of 100 bpm that would trigger an alert in a non-pregnant adult is within normal postpartum range. Hospital-based early warning scores (MEOWS — Modified Early Obstetric Warning System) exist but are designed for inpatient use and cannot be applied to home monitoring because they require clinical interpretation and vital sign equipment. Research studies on postpartum remote monitoring are in early pilot phases with no validated algorithm for sepsis detection specifically.

What would unlock progress

Pregnancy-and-postpartum-adjusted vital sign algorithms that define sepsis alert thresholds calibrated to the physiological changes of the postpartum period. Clinical validation studies establishing the positive predictive value of wearable-detected physiological changes for postpartum sepsis. Integration pathways that connect home monitoring alerts to clinical teams without overwhelming provider workflows or generating excessive false alarms — the alert must be specific enough to trigger action.

Entry points for student teams

A student team could build the pregnancy-adjusted threshold logic out of the published record instead of out of patient data: several MEOWS validation studies report their trigger parameters together with sensitivity and specificity in obstetric cohorts (Singh et al., Anaesthesia 2012;67(1):12–18; Ryan et al., J Obstet Gynaecol Can 2017;39(9):728–733; Singhal et al., J Family Med Prim Care 2022;11(12):7644–7651; Kefeli Çelik et al., Ulus Travma Acil Cerrahi Derg 2024;30(9):635–643; Werlang et al., J Obstet Gynaecol Can 2026;48(4):103248), so a team can work out which parameters carry the discrimination, re-specify them for the weeks at home rather than the inpatient ward, and test the resulting rule against published case series of postpartum sepsis presentations. Alternatively, teams could design a prototype wearable alert system using commercially available sensors (temperature, pulse, respiratory rate) with postpartum-specific detection algorithms — and because a real accuracy study needs enrolled postpartum patients, the semester deliverable is the validation trial's design rather than its execution: sampling frame, reference standard for sepsis, the alert-to-clinician escalation path, and a pre-registered analysis plan sized to the event rate, handed to a clinical partner. Analyzing hospital continuous postpartum vital-sign records is a real third door but not an open one: those are identifiable clinical records rather than a public dataset, and reaching them takes an IRB protocol plus a sponsoring obstetric service or hospital-informatics group that already owns the monitoring archive. Relevant disciplines: biomedical engineering, clinical informatics, data science, obstetric medicine.