How might improved estimates of HIV programme outcomes influence practice? A formative study of evidence, dissemination and response

BACKGROUND: While HIV programmes have started millions of persons on life-saving antiretroviral therapy in Africa, longitudinal health information systems are frail and, therefore, data about long-term survival is often inaccurate or unknown to HIV programmes. The ‘Better Information for Health in Zambia’ (BetterInfo) Study – a regional sampling-based survey to assess retention and mortality in HIV programmes in Zambia – found both retention and mortality to be higher than prevailing estimates from national surveillance systems. We sought to understand how Zambian health decision-makers at different health system levels would respond to these new data, with a view to informing research translation. METHODS: We interviewed 25 purposefully sampled health decision-makers from community, facility, district, provincial and national levels. During the interviews, we shared retention and mortality estimates from both routine programme surveillance and those generated by the study. Transcripts were analysed for inductive and deductive themes, the latter drawing on Weiss’s framework that policy-makers interpret and apply evidence as ‘warning’, ‘guidance’, ‘reconceptualisation’ or ‘mobilisation of support’. FINDINGS: All decision-makers found study findings relevant and important. Decision-makers viewed the underestimates of mortality to be a warning about the veracity and informativeness of routine data systems. Decision-makers felt guided by the findings to improve data monitoring and, acknowledging limitations of routine data, utilised episodic patient tracing to support improved data accuracy. Findings catalysed renewed motivation and mobilisation by national level decision-makers for differentiated models of HIV care to improve patient outcomes and also improved data management systems to better capture patient outcomes. Inductive analysis highlighted a programmatic application data interpretation, in which study findings can influence facility and patient-level decision-making, quality of care and routine data management. CONCLUSIONS: New epidemiological data on patient outcomes were widely seen as informative and relevant and can potentially catalyse health system action such as using evaluations to supplement electronic medical record data to improve HIV programmes. Formative evidence suggests that targeting research dissemination at different levels of the health system will elicit different responses. Researchers supporting the translation of evidence to action should leverage all relevant levels of the health system to facilitate both policy and programmatic action.