All 10 studies (N=2,230) that were located were included.
Heterogeneity: the chi-squared test showed that the effectiveness values were heterogeneous for both informant questionnaires and cognitive tests. The only intervention for which the effectiveness values were homogeneous was the MMSE, for which a fixed-effect model was used.
The weighted mean effectiveness value for the informant questionnaires was 1.74 (95% CI: 1.39, 2.09), for the cognitive tests it was 1.48 (95% CI: 1.23, 1.73). When only the seven IQCODE versus MMSE studies were considered, weighted mean effectiveness was 1.75 (95% CI: 1.36, 2.14) for the IQCODE compared with 1.43 (95% CI: 1.25, 1.61) for the MMSE. For all 8 studies using the IQCODE the weighted mean effectiveness was 1.82 (95% CI: 1.43, 2.21), while for all 8 MMSE studies it was 1.53 (95% CI: 1.36, 1.70).
The mean effectiveness measures were translated into sensitivity and specificity values. Using a cut-point to balance sensitivity and specificity, informant questionnaires would have sensitivity and specificity of 0.83, compared with 0.79 to 0.80 for cognitive tests. If specificity is held to a fixed value of 0.80, the sensitivity of informant questionnaires is 0.86 compared with 0.79 for cognitive tests.