TABLE 2

Univariate Metaregression Analysis for Identifying Potential Sources of Heterogeneity in the Diagnostic Performance of CBCL-AP scale

Study characteristicStudies, nSensitivitySpecificity
Pooled estimate (95% CI)PPooled estimate (95% CI)P
Sample source1.001.00
 Clinic 80.78 (0.64–0.88)0.67 (0.54–0.78)
 Nonclinic70.76 (0.68–0.83)0.79 (0.66–0.88)
Study location.09.03
 United States80.70 (0.61–0.78)0.81 (0.73–0.87)
 Other countries80.83 (0.72–0.90)0.64 (0.49–0.77)
No. of participants.88.82
 ≥20070.77 (0.61–0.88)0.75 (0.63–0.84)
 <20090.75 (0.68–0.80)0.72 (0.57–0.84)
Cutoff value1.001.00
 ≥6580.69 (0.59–0.77)0.74 (0.60–0.84)
 <6560.86 (0.75–0.92)0.72 (0.53–0.86)
Study year.26.05
 After or during 2005110.80 (0.71–0.87)0.67 (0.55–0.78)
 Before 200550.71 (0.57–0.81)0.83 (0.76–0.89)
Age.91< .01
 ≥1170.77 (0.67–0.85)0.84 (0.75–0.90)
 <1190.77 (0.65–0.86)0.63 (0.52–0.72)
Female participants, %.35.04
 ≥3570.73 (0.66–0.79)0.83 (0.77–0.88)
 <3590.80 (0.67–0.88)0.64 (0.51–0.76)
Study quality
 Patient selection.13.17
  High risk80.83 (0.70–0.91)0.67 (0.53–0.78)
  Low risk80.73 (0.67–0.78)0.79 (0.67–0.87)
 Index test.61.95
  High risk100.74 (0.67–0.80)0.74 (0.60–0.84)
  Low risk60.79 (0.62–0.89)0.71 (0.58–0.82)
 Flow and timing.15.31
  High risk40.67 (0.52–0.79)0.80 (0.62–0.91)
  Low risk120.80 (0.72–0.87)0.71 (0.60–0.80)