Fig. 4

The proportion (95% Confidence Interval) of studies using each diagnostic, by propensity score method. Assessment not reported’ refers to papers which did not specify whether propensity scores were assessed; ‘Diagnostic not reported’ refers to papers which reported that assessment took place, but not how; ‘Other’ includes: absolute differences, graphical approaches, post-matching c-statistic, regression, standardised bias, and variance ratios