Effectiveness of Telemedicine Strategies for Diabetic Retinopathy Screening: A Systematic Review

Abstract

 

Purpose

 

Telemedicine has been described as a promising alternative that facilitates compliance to healthcare guidelines regardless of geographic constraints, promoting consistency of health services delivery. To evaluate whether the existing evidence justifies the implementation of telemedicine programs for diabetic retinopathy screening, we conducted a systematic review of diagnostic accuracy of telemedical tools for diabetic retinopathy screening in adults.

 

Methods

 

Search strategy A structured search was conducted among six databases (Medline, EMBASE, CINAHL, BIOSIS, Web of Science, Cochrane library) from January 1998 to July 2012, with no language or study design restrictions. Subject headings and free text terms were tailored to each of the electronic databases. Searches were supplemented by handsearching the bibliographies of included studies; grey literature was also consulted.Screening & data collection First, titles and abstracts were reviewed to identify potentially relevant articles. Next, full-text articles from included citations were retrieved to closely assess inclusion and exclusion criteria. Two reviewers independently assessed all citations and disagreements were solved by discussion. Remaining discrepancies were reconciled by an ophthalmologist. A single reviewer performed the data collection using predefined data fields, and data abstraction was audited by a second reviewer. QUADAS-2 guidelines were used to assess methodological quality of included studies.

 

Results

 

1060 articles were screened, from which 156 were shortlisted for full-text assessment. Using pre-specified inclusion and exclusion criteria, 27 studies were deemed relevant for this review. Data were divided into two subgroups according to detection criteria (any retinopathy or referable retinopathy). For detection of any retinopathy, sensitivity and specificity were 84.7% (95% CI: 79%-90%) and 93.1% (95% CI: 89.7%-96.4%), respectively. For detection of referable retinopathy, values were 86.4% (95% CI: 83.1%-89.6%) and 92.1% (95% CI: 89.5%-94.7%), respectively.

 

Conclusions

 

Diagnostic accuracy of telemedicine strategies for diabetic retinopathy screening meets British Diabetic Association recommendations (sensitivity 80%, specificity 95%), although specificity was slightly lower. Results from the referable retinopathy subgroup should be cautiously interpreted, since studies used diverse grading guidelines.

 

Keywords: 499 diabetic retinopathy • 496 detection • 459 clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology