Association between study characteristics and risk of bias
| Characteristic | All studies (n) | Studies in high risk of bias in at least one domain number (%) | p Value* |
|---|---|---|---|
| Year of publication | 0.37 | ||
| Pre-2002 | 37 | 17 (46%) | |
| 2002–2006 | 46 | 20 (44%) | |
| 2007–2010 | 59 | 32 (54%) | |
| Type of diabetes | 0.11 | ||
| Type 1 | 9 | 3 (33%) | |
| Type 2 | 80 | 36 (45%) | |
| Both | 34 | 16 (47%) | |
| Unclear | 19 | 14 (74%) | |
| Unit of allocation | 0.24 | ||
| Patient | 94 | 49 (52%) | |
| Cluster (eg, provider/clinic) | 48 | 20 (42%) | |
| Country/setting | 0.62 | ||
| The USA or Canada | 79 | 41 (52%) | |
| The UK or Western Europe | 40 | 17 (43%) | |
| Other | 23 | 11 (48%) | |
| Journal impact factor | 0.87 | ||
| Greater than 3 (median) | 71 | 34 (47.9%) | |
| Less than 3 (median) | 71 | 35 (49.3%) | |
| Effective sample size | 0.87 | ||
| Greater than 154 (median) | 71 | 35 (49.3%) | |
| Less than 154 (median) | 71 | 34 (47.9%) | |
| Intervention type | 0.17 | ||
| Multifaceted (featuring more than one QI strategy) | 124 | 63 (51%) | |
| Single intervention | 18 | 6 (33%) |
*Comparing proportion of studies with at least one domain at high risk of bias against studies no domains at high risk of bias. For year of publication, Cochran-Armitage test for trend was conducted. For other study characteristics, χ2 (or Fisher's exact) tests for categorical and Wilcoxon signed-rank tests for continuous variables were used.
QI, quality improvement.