학술논문

Modelling Methods of Economic Evaluations of HIV Testing Strategies in Sub-Saharan Africa: A Systematic Review
Document Type
redif-article
Source
Springer, Applied Health Economics and Health Policy. 21(4):585-601
Subject
Language
English
Abstract
Background and Objective Economic evaluations, a decision-support tool for policy makers, will be crucial in planning and tailoring HIV prevention and treatment strategies especially in the wake of stalled and decreasing funding for the global HIV response. As HIV testing and treatment coverage increase, case identification becomes increasingly difficult and costly. Determining which subset of the population these strategies should be targeted to becomes of vital importance as well. Generating quality economic evidence begins with the validity of the modelling approach and the model structure employed. This study synthesises and critiques the reporting around modelling methodology of economic models in the evaluation of HIV testing strategies in sub-Saharan Africa. Methods The following databases were searched from January 2000 to September 2020: MEDLINE, Embase, Scopus, EconLit and Global Health. Any model-based economic evaluation of a unique HIV testing strategy conducted in sub-Saharan Africa presenting a cost-effectiveness measure published from 2013 onwards was eligible. Data were extracted around three components: general study characteristics; economic evaluation design; and quality of model reporting using a novel tool developed for the purposes of this study. Results A total of 21 studies were included; 10 cost-effectiveness analyses, 11 cost-utility analyses. All but one study was conducted in Eastern and Southern Africa. Modelling approaches for HIV testing strategies can be broadly characterised as static aggregate models (3/21), static individual models (6/21), dynamic aggregate models (5/21) and dynamic individual models (7/21). Adequate reporting around data handling was the highest of the three categories assessed (74%), and model validation, the lowest (45%). Limitations to model structure, justification of chosen time horizon and cycle length, and description of external model validation process were all adequately reported in less than 40% of st