Methodological quality of the review: Medium confidence
Author: Zhang X, Norris SL, Saadine J, Chowdhury FM, Horsley T, Kanjilal S, Mangione CM, Buhrmann R.
Region: United States of America (USA), Australia, United Kingdom (UK), Israel and Saudi Arabia.
Sector: Diabetic retinopathy screening, diabetes.
Sub-sector: Service uptake
Equity focus: None Specified
Review type: Effectiveness review
Quantitative synthesis method: Meta-analysis of randomized controlled trials and narrative analysis of the remainder.
Qualitative synthesis methods: Not applicable
Early detection and timely treatment are key strategies for reducing the burden associated with diabetic retinopathy (DR) and resulting visual impairment. Medical organizations recommend regular retinal screening for preventing blindness related to DR. Despite the evidence supporting the effectiveness of periodic screening for DR, screening rates consistently fall far below recommended levels.
To assess the effect of interventions to increase the use of retinal screening among people with diabetes.
The authors identified 48 eligible studies of interventions to promote screening for DR, including 12 randomized controlled trials (RCTs), four non-randomized studies and 32 pre-post studies. Of these, 32 studies were conducted in the USA, four in Australia, three in the UK, three in Israel, one in Saudi Arabia and the remaining ones in other developed countries.
The authors explained: ‘Four of five RCTs focusing on patients demonstrated that interventions increased screening significantly, with relative risk ranging from 1.05 (95% confidence interval [CI] _1.01–1.08) to 2.01 (95% CI_1.48 –2.73). Five RCTs with a focus on the system all demonstrated significant increases in screening with relative risk ranging from 1.12 (95% CI_1.03–1.22) to 5.56 (95% CI_2.19 –14.10). Thirty-six non-RCTs, which included interventions with single or multiple foci, also generally demonstrated positive effects.’
‘Further research should explore strategies for increasing the rate of retinal screening among diverse or disadvantaged populations and the economic efficiency of effective interventions in large community populations.’
The review protocol was developed using the methods of the Cochrane Collaboration.
The following databases were searched between 1980 and 2005 for relevant reviews of interventions to promote screening for DR: MEDLINE, EMBASE, CINAHL, Web of Science, Cochrane Library and Cochrane Controlled Trials Registers (including DARE). Relevant journals and references lists of included articles were also manually searched. Authors of original articles were contacted when data were unclear or missing. No language restrictions were applied. Studies with the following study designs were included: RCTs, controlled clinical trials, controlled before-and-after trials, interrupted time series and pre-post studies.
One author reviewed each article for inclusion, and a second author reviewed the articles only if there was uncertainty as to include it. Data was extracted by two different authors and all data were reviewed again by a third author. Authors followed Cochrane Collaboration methods and quality assessment.
A sufficient number of RCTs were identified and therefore were pooled for analysis. Relative Risk (RR) was calculated as the primary measure of effect for both RCTs and other studies. Narrative analysis was also done for the latter. A sensitivity analysis was conducted to compare results between studies with high and poor risk of bias. A funnel plot was performed to assess publication bias within the review. To test for statistical heterogeneity, the chi-square test was used.
Authors mentioned that due to the diversity of the health care characteristics of the included studies, specific interventions may only be applicable to certain countries, settings or regions.
This review focuses on all countries; however, authors did not discuss results from low- and middle-income settings separately from the high-income countries.
Overall, there was medium confidence in the conclusions about the effects of this study. Authors performed a thorough literature search in order to ensure the inclusion of all relevant articles, although only one author reviewed each article for inclusion, which may contribute to the presence of selection bias/publication bias. Quality assessment of each included articles were appropriately assessed, but a table or summary of the assessment of each included study for each criterion was not reported.
Authors took into account the potential for bias from cofounding and secular trends in studies without randomization; these were analysed separately from randomized studies.
Authors acknowledged that due to the heterogeneity among the studies, types of interventions, the length of follow-up, and the characteristics of participants, a quantitative synthesis and between-study made comparisons difficult. All studies were stratified into three intervention groups; however, there was considerable diversity of interventions within each group. Publication bias was likely to be present within the review, which could have potentially affect the conclusions of the review.