Reducing Coordination Failures in Antenatal Care: Evidence from Mozambique

We design an appointment scheduling system to address the coordination failure by
redistributing demand across hours and days. The goal is to stagger arrivals to reduce peak congestion, shrink
idle capacity, and smooth provider workloads. In partnership with the National Institute of Health, the National Directorate of Public Health,
and Provincial Health Authorities, we designed and implemented a low-cost appointment scheduling
system to improve the management of patient flows. Under the new system, providers maintained
an appointment book in each consultation room, dividing the day into one-hour blocks with a
maximum of six patients per block. At the end of each visit, the provider consulted the appointment
book, offered the patient a choice of available slots on their return date, and recorded the
appointment both in the book and on the patient’s own booklet. Patients were informed that they
would be seen within one hour of their scheduled time. To accommodate existing clinical practice,
first-time ANC patients could attend consultations during a reserved two-hour block each morning
(generally from 7 to 9 or 8 until 10 am).

2018-09-01

2024-09-01

Patient Retention in Care (number of antenatal consultations attended), quality of care (number of clinical procedures undertaken during consultation out of the health clinic's checklist), patient satisfaction (satisfaction with waiting time, satisfaction with the quality of care and satisfaction with providers), provider satisfaction (job satisfaction, burnout, stress levels)

We sampled 80 high patient volume public health clinics across four
southern provinces in Mozambique: Maputo City, Maputo Province, Gaza, and Inhambane. Clinics
were stratified by province and randomly assigned to implement the scheduling system (40
treatment clinics) or to continue the existing first-come, first-served system (40 control clinics).
To prevent contamination through patient self-selection into better-managed clinics, we imposed
a two-kilometer buffer zone between participating facilities. Switching across clinics was further
constrained by the fact that patient files were in physical records held at a single clinic and could
not be accessed elsewhere, making transfers extremely rare. Treatment and control clinics were balanced
on baseline characteristics including patient volume, degree of urbanization, infrastructure
quality, staffing levels, and organizational capacity.

Randomization done in computer

Health clinic

No

80

14000 patients surveyed through the audit survey to measure waiting time, 2000 patients surveyed in the exit surveys, 4000 observations from the registry book.

40 health clinics

0.05 standard deviations, number of visits, 80% power