RESUMO
SETTING: Human immunodeficiency virus (HIV) clinics in five hospitals and five health centres in Lusaka, Zambia, which transitioned from daily entry of paper-based data records to an electronic medical record (EMR) system by dedicated data staff (Electronic-Last) to direct real-time data entry into the EMR by frontline health workers (Electronic-First). OBJECTIVE: To compare completeness and accuracy of key HIV-related variables before and after transition of data entry from Electronic-Last to Electronic-First. DESIGN: Comparative cross-sectional study using existing secondary data. RESULTS: Registration data (e.g., date of birth) was 100% complete and pharmacy data (e.g., antiretroviral therapy regimen) was <90% complete under both approaches. Completeness of anthropometric and vital sign data was <75% across all facilities under Electronic-Last, and this worsened after Electronic-First. Completeness of TB screening and World Health Organization clinical staging data was also <75%, but improved with Electronic-First. Data entry errors for registration and clinical consultations decreased under Electronic-First, but errors increased for all anthropometric and vital sign variables. Patterns were similar in hospitals and health centres. CONCLUSION: With the notable exception of clinical consultation data, data completeness and accuracy did not improve after transitioning from Electronic-Last to Electronic-First. For anthropometric and vital sign variables, completeness and accuracy decreased. Quality improvement interventions are needed to improve Electronic-First implementation.