JMIR Publications

ABSTRACT

Background:

Patient payment queuing at hospital cashier desks is a persistent operational challenge in resource-limited healthcare settings. In October 2025, the University Teaching Hospital of Butare (CHUB), a public tertiary referral centre in Rwanda, integrated an asynchronous Mobile Money payment module with its OpenClinic GA hospital management information system (HMIS) through the Irembo payment gateway, as an unfunded quality improvement initiative.

Objective:

To describe the integration architecture and quantify its impact on patient payment cycle time, channel adoption, and transaction reliability over eight months of routine operation.

Methods:

The architecture is described from direct analysis of the deployed production source code. Performance was evaluated by retrospective analysis of all closed invoices for prepaid services (laboratory, imaging, pharmacy) from 1 October 2025 to 31 May 2026 (n=146,203 episodes). Payment cycle time was defined as the interval from bill creation to payment confirmation. The primary channel comparison used a 1,440-minute (24-hour) analytic cutoff; a sensitivity analysis using Kaplan-Meier and Cox proportional-hazards regression on the full, untruncated duration distribution, and a calendar-month-adjusted quantile regression, tested whether this cutoff affected the channel comparison. Bootstrap 95% confidence intervals were computed for median cycle times.

Results:

The integration implements a fully asynchronous, non-blocking model with dual trigger pathways (patient-initiated SMS and staff-initiated HMIS form) and a four-step Irembo API workflow. Across 146,203 invoiced episodes, Mobile Money completion grew from 3.8% to 20.9% of monthly episodes. Among 131,037 within-24-hour transactions, median payment cycle time was 5 minutes (95% CI 5-5) for Mobile Money versus 72 minutes (95% CI 71-72) for cash, a difference of 67 minutes (95% CI 66-67) associated with payment channel. This association was materially unchanged after adjustment for calendar month (quantile regression coefficient -67.0 minutes, 95% CI -67.7 to -66.3; P<.001) and in a Kaplan-Meier/Cox sensitivity analysis using the full, untruncated duration distribution (median 5 versus 88 minutes; Cox hazard ratio 3.88, 95% CI 3.82-3.95, adjusted for month; log-rank P<.001). The Mobile Money transaction system processed 44,216 unique transactions with a 97.9% completion rate (43,278/44,216); 18,935 of 146,203 episodes (12.9%) were ultimately paid via Mobile Money. The per-attempt failure rate was 3.6% (3,167/87,269 gateway attempts); 329 transactions (0.7%) failed permanently. Insurer paper-documentation requirements were the principal structural constraint on adoption identified.

Conclusions:

Asynchronous Mobile Money integration was associated with a substantially shorter patient payment cycle time and provides a reproducible reference architecture for sub-Saharan African hospitals. Because payment channel was self-selected by patients and no pre-implementation baseline was available, these findings should be interpreted as an associational, quality-improvement-level evaluation of cycle time rather than a causal demonstration of queue elimination.