Transferability of Indirect Cost of Chronic Disease: A Systematic Review and Meta-Analysis

Background Indirect cost is an important component in economic evaluations. The variation in the magnitude of indirect costs across studies and countries is substantial and affects the transferability of results across jurisdictions. Objective This study explored the factors involved in the variation of reported indirect cost and investigated the feasibility of transferring indirect costs across settings. Methods A systematic literature review was conducted to identify studies estimating indirect costs for four selected chronic diseases, namely, asthma (AS), diabetes (DI), rheumatoid arthritis (RA) and schizophrenia (SC). A multiple linear regression analysis was run to identify the factors that potentially explain the variation in reported indirect costs. Parametric (fixed- and random-effect models) and non-parametric (bootstrapping method) meta-analyses were applied to local gross domestic product (GDP)/capita-adjusted indirect costs for each disease. Results from the three different analytical methods were compared to ascertain the robustness of estimation. Results The systematic literature review identified 77 articles that reported indirect costs of AS (n = 18), DI (n = 20), RA (n = 25) and SC (n = 14) for literature synthesis. Substantial inter- and intra-disease variations among the indirect cost studies were observed with respect to geographic distribution, methodology and magnitude of cost estimation. Regression analysis showed that disease categories and local GDP/capita significantly (p < 0.001) contributed to the variance of indirect cost. The range of intra-disease variation in indirect costs was substantially reduced after adjusting by and expressing values as local GDP/capita. The GDP-adjusted indirect cost in terms of percentage of local GDP/capita of AS was the lowest and that of SC was the highest. Bootstrapping estimation was relatively conservative, with slightly wider confidence intervals (CIs) than the parametric method, with a mean (95 % CI) of 2.12 % (1.4089-2.9332) for AS, 10.65 % (7.215-14.7438) for DI, 21.98 % (17.4360-27.0631) for RA, and 79.19 % (52.4243-117.833) for SC. Conclusion It would be convenient and feasible to construct a universal reference range of indirect cost for a specific disease based on existing data and present this as a percentage of local GDP to assist local decision making in jurisdictions where indirect cost data are not available.