Background and Objectives: Telemedicine holds the promise of increasing access-to-care at a lower cost. Yet, for years, the evidence of telemedicine's cost-effectiveness was scarce. Faced with a rapidly expanding literature, we conduct both manual and systematic selection of the literature, and analyzed the data to determine: (1) the characteristics of economic evaluations of telemedicine, and (2) the determinants of economically efficient telemedicine interventions. Methods: We reviewed all published economic evaluations of telemedicine in Cochrane, Embase, and Pubmed from 2008 to 2018. Articles were screened by two researchers first on title and abstract (Stage 1), then on full article (Stage 2), (protocol available on PROSPERO, ref. CRD42019143032). We proposed an alternative method for screening articles using machine learning based on textual classification and compared these two approaches. We constructed an exclusive dataset on the characteristics of the selected articles and enriched it using OECD data at the country level. We identified the determinants of efficient telemedicine interventions using multiple logit models. Results and Conclusion: We included 156 articles out of 2,639. Most economic studies of our sample regard telemonitoring. A majority (73.7%) of studies found that telemedicine intervention is efficient, regardless of the medical domain. Articles with higher standards of economic evaluation (cost-effectiveness analysis, randomized trials with high sample size) were less likely to report an efficient intervention. We found no effect of the publication year, signifying that the nature of the evidence has not changed over time.
Keywords: costing; economic evaluation; systematic review; telemedicine; textual analysis.