Modelling improved efficiency in healthcare referral systems for the urban poor using a geo-referenced health facility data: the case of Sylhet City Corporation, Bangladesh

BACKGROUND: An effective referral system is critical to ensuring access to appropriate and timely healthcare services. In pluralistic healthcare systems such as Bangladesh, referral inefficiencies due to distance, diversion to inappropriate facilities and unsuitable hours of service are common, particularly for the urban poor. This study explores the reported referral networks of urban facilities and models alternative scenarios that increase referral efficiency in terms of distance and service hours. METHODS: Road network and geo-referenced facility census data from Sylhet City Corporation were used to examine referral linkages between public, private and NGO facilities for maternal and emergency/critical care services, respectively. Geographic distances were calculated using ArcGIS Network Analyst extension through a “distance matrix” which was imported into a relational database. For each reported referral linkage, an alternative referral destination was identified that provided the same service at a closer distance as indicated by facility geo-location and distance analysis. Independent sample t-tests with unequal variances were performed to analyze differences in distance for each alternate scenario modelled. RESULTS: The large majority of reported referrals were received by public facilities. Taking into account distance, cost and hours of service, alternative scenarios for emergency services can augment referral efficiencies by 1.5–1.9 km (p < 0.05) compared to 2.5–2.7 km in the current scenario. For maternal health services, modeled alternate referrals enabled greater referral efficiency if directed to private and NGO-managed facilities, while still ensuring availability after working-hours. These referral alternatives also decreased the burden on Sylhet City’s major public tertiary hospital, where most referrals were directed. Nevertheless, associated costs may be disadvantageous for the urban poor. CONCLUSIONS: For both maternal and emergency/critical care services, significant distance reductions can be achieved for public, NGO and private facilities that avert burden on Sylhet City’s largest public tertiary hospital. GIS-informed analyses can help strengthen coordination between service providers and contribute to more effective and equitable referral systems in Bangladesh and similar countries.