Description of Session
Given the marked levels of neonatal mortality that occurs in the first 48h of life, improving the coverage and utilization of antenatal and postnatal care is critical. Several important rate-limiting steps prevent adequate and timely coverage including: non-systematic listing of eligible clients, poor definition of 'catchment areas', disproportionate client:worker ratios (1 worker serves 500-1000 households), poor supervision, absenteeism, vacant posts, and low demand / utilization of services. We have spent two years developing and testing disruptive innovations to some of these intractable challenges, using experimental designs to assess the feasibility of technology-enabled methods to strengthen worker performance. Three approaches were taken -- one focused on methods to identify gaps in coverage and timely service delivery, one on strengthening and facilitating supportive supervision by automating worker performance flagging -- leveraging AI/ML-based algorithms, and one that overturns the decades-old practice of catchment-area work assignment with a novel task-based allocation of public health services to clients. Although supportive supervision is critical to high-quality service delivery, we recognize that supervisors are strained to adequately digest and develop plans based on worker performance data. Supervisor shortages continue to be persistent – for both the health system and frontline workers. Our goal was to develop and test innovations that help workers better manage their own assigned tasks, while strengthening the capacity of managers to identify and support workers who need help the most. Worker allocations to service areas are often based on outdated household listings and incomplete census data, rarely using maps or GIS to identify households which may have been left out. The lack of reliable denominators have led to reliance on statistically-derived “targets” to project program success. We measure workforce coverage “footprints”, identify missed household clusters, identify undersupervised workers and detect worker-level variations.
