Evaluation of the electronic Early Warning and Response Network (EWARN) system in Somalia, 2017–2020

Abstract Background In 2008, Somalia introduced an electronic based Early Warning Alert and Response Network (EWARN) for real time detection and response to alerts of epidemic prone diseases in a country experiencing a complex humanitarian situation. EWARN was deactivated between 2008 to 2016 due to civil conflict and reactivated in 2017 during severe drought during a cholera outbreak. We present an assessment of the performance of the EWARN in Somalia from January 2017 to December 2020, reflections on the successes and failures, and provide future perspectives for enhancement of the EWARN to effectively support an Integrated Disease Surveillance and Response strategy. Methods We described geographical coverage of the EWARN, system attributes, which included; sensitivity, flexibility, timeliness, data quality (measured by completeness), and positive predictive value (PPV). We tested for trends of timeliness of submission of epidemiological reports across the years using the Cochran-Mantel–Haenszel stratified test of association. Results By December 2020, all 6 states and the Banadir Administrative Region were implementing EWARN. In 2017, only 24.6% of the records were submitted on time, but by 2020, 96.8% of the reports were timely (p < 0.001). Completeness averaged < 60% in all the 4 years, with the worst-performing year being 2017. Overall, PPV was 14.1%. Over time, PPV improved from 7.1% in 2017 to 15.4% in 2019 but declined to 9.7% in 2020. Alert verification improved from 2.0% in 2017 to 52.6% by 2020, (p < 0.001). In 2020, EWARN was enhanced to facilitate COVID-19 reporting demonstrating its flexibility to accommodate the integration of reportable diseases. Conclusions During the past 4 years of implementing EWARN in Somalia, the system has improved significantly in timeliness, disease alerts verification, and flexibility in responding to emerging disease outbreaks, and enhanced coverage. However, the system is not yet optimal due to incompleteness and lack of integration with other systems suggesting the need to build additional capacity for improved disease surveillance coverage, buttressed by system improvements to enhance data quality and integration.