Skip to main content

Advertisement

Table 2 Percent absolute relative bias (RB) and mean squared error (MSE) for computer simulation study

From: Combining population-based administrative health records and electronic medical records for disease surveillance

\( {prev}_{Y_1}, \) \( {prev}_{Y_2} \) \( {\overline{\rho}}_x \) RB; prev T = 20%
\( {\boldsymbol{\rho}}_{{\boldsymbol{Y}}_{\mathbf{1}}{\boldsymbol{Y}}_{\mathbf{2}}} \) = 0.85 \( {\boldsymbol{\rho}}_{{\boldsymbol{Y}}_{\mathbf{1}}{\boldsymbol{Y}}_{\mathbf{2}}} \) = 0.65
OR AND RSSA PSSA (16) PSSA (8) OR AND RSSA PSSA (16) PSSA (8)
18, 15% 0.00 9.5 47.5 7.5 9.5 11.3 23.1 59.4 1.3 48.3 49.5
0.20 9.0 47.7 7.9 2.1 7.4 22.9 59.4 1.5 41.7 54.3
0.50 10.1 47.2 7.0 24.3 21.1 23.7 59.1 0.9 99.0 78.6
18, 10% 0.00 0.3 58.6 18.2 1.1 3.0 10.8 67.1 12.8 28.9 37.5
0.20 0.9 58.9 18.7 5.9 5.9 10.5 67.2 13.0 31.1 54.3
0.50 0.2 58.3 17.7 48.8 26.1 11.2 66.8 12.4 108.8 90.1
15, 15% 0.00 4.1 49.2 11.5 3.7 4.3 17.6 61.7 5.8 41.3 42.5
0.20 3.6 49.4 12.0 3.1 3.6 17.2 61.9 6.1 37.6 50.1
0.50 4.8 48.7 10.9 20.5 12.5 18.1 61.5 5.3 102.0 70.7
\( {prev}_{Y_1}, \) \( {prev}_{Y_2} \) \( {\overline{\rho}}_x \) MSE; prev T = 20%
\( {\boldsymbol{\rho}}_{{\boldsymbol{Y}}_{\mathbf{1}}{\boldsymbol{Y}}_{\mathbf{2}}} \) = 0.85 \( {\boldsymbol{\rho}}_{{\boldsymbol{Y}}_{\mathbf{1}}{\boldsymbol{Y}}_{\mathbf{2}}} \) = 0.65
OR AND RSSA PSSA (16) PSSA (8) OR AND RSSA PSSA (16) PSSA (8)
18, 15% 0.00 0.04 0.90 0.02 0.06 0.47 0.22 1.41 < 0.01 0.99 1.76
0.20 0.03 0.91 0.03 0.02 0.52 0.21 1.41 < 0.01 0.82 2.25
0.50 0.04 0.89 0.02 1.06 1.47 0.23 1.40 0.00 4.68 4.46
18, 10% 0.00 < 0.01 1.37 0.13 0.02 0.69 0.05 1.80 0.07 0.40 1.70
0.20 < 0.01 1.39 0.14 0.06 1.16 0.05 1.80 0.07 0.70 3.48
0.50 < 0.01 1.36 0.13 2.28 2.32 0.05 1.79 0.06 5.36 6.16
15, 15% 0.00 0.01 0.97 0.05 0.03 0.46 0.13 1.53 0.01 0.74 1.62
0.20 0.01 0.98 0.06 0.02 0.72 0.12 1.53 0.02 0.74 2.20
0.50 0.01 0.95 0.05 1.03 1.29 0.13 1.51 0.01 4.84 3.96
\( {prev}_{Y_1}, \) \( {prev}_{Y_2} \) \( {\overline{\rho}}_x \) RB; prev T = 10%
\( {\boldsymbol{\rho}}_{{\boldsymbol{Y}}_{\mathbf{1}}{\boldsymbol{Y}}_{\mathbf{2}}} \) = 0.85 \( {\boldsymbol{\rho}}_{{\boldsymbol{Y}}_{\mathbf{1}}{\boldsymbol{Y}}_{\mathbf{2}}} \) = 0.65
OR AND RSSA PSSA (16) PSSA (8) OR AND RSSA PSSA (16) PSSA (8)
8, 7% 0.00 11.5 55.8 8.4 9.7 35.0 29.6 67.1 1.1 76.2 154.9
0.20 10.5 56.1 9.2 42.6 37.8 28.7 67.4 0.3 196.7 217.1
0.50 13.1 54.9 7.1 216.1 43.3 30.7 66.6 2.0 307.6 286.5
8, 5% 0.00 2.9 59.7 16.0 1.1 50.5 12.2 73.4 13.5 45.3 114.9
0.20 3.2 59.7 15.8 10.7 85.1 12.0 73.9 13.8 235.2 273.4
0.50 3.8 59.2 15.3 230.5 198.2 14.2 73.3 12.1 322.2 334.8
5, 5% 0.00 14.7 70.0 30.9 6.3 92.1 7.4 78.7 28.4 61.0 193.0
0.20 15.4 70.2 31.5 134.4 149.1 8.0 78.7 28.8 271.0 217.9
0.50 13.6 69.5 30.0 275.7 222.1 6.3 78.2 27.4 333.8 375.0
\( {prev}_{Y_1}, \) \( {prev}_{Y_2} \) \( {\overline{\rho}}_x \) MSE; prev T = 10%
\( {\boldsymbol{\rho}}_{{\boldsymbol{Y}}_{\mathbf{1}}{\boldsymbol{Y}}_{\mathbf{2}}} \) = 0.85 \( {\boldsymbol{\rho}}_{{\boldsymbol{Y}}_{\mathbf{1}}{\boldsymbol{Y}}_{\mathbf{2}}} \) = 0.65
OR AND RSSA PSSA (16) PSSA (8) OR AND RSSA PSSA (16) PSSA (8)
8, 7% 0.00 0.01 0.31 0.01 0.28 2.01 0.09 0.45 < 0.01 0.87 6.33
0.20 0.01 0.31 0.01 1.29 1.31 0.08 0.45 < 0.01 5.19 8.53
0.50 0.02 0.30 0.01 5.60 6.72 0.10 0.44 < 0.01 9.97 12.77
8, 5% 0.00 < 0.01 0.36 0.03 0.12 2.78 0.02 0.54 0.02 0.28 4.31
0.20 < 0.01 0.36 0.03 0.59 3.59 0.02 0.55 0.02 7.28 12.63
0.50 < 0.01 0.35 0.02 6.39 8.45 0.02 0.54 0.02 10.91 16.73
5, 5% 0.00 0.02 0.49 0.10 0.57 6.96 0.01 0.62 0.08 1.92 9.37
0.20 0.02 0.49 0.10 4.82 8.11 0.01 0.62 0.08 9.11 9.70
0.50 0.02 0.48 0.09 8.71 10.08 < 0.01 0.61 0.08 11.79 18.41
  1. OR Rule-based OR method, AND Rule-based AND method, RSSA Rule-based sensitivity-specificity adjusted method, PSSA Probabilistic-based sensitivity-specificity adjusted; prevT denotes true population prevalence; \( {prev}_{Y_1}, \) \( {prev}_{Y_2} \) denotes outcome prevalence; \( {\rho}_{Y_1{Y}_2} \) denotes correlation between data sources; \( {\overline{\rho}}_x \) denotes average correlation amongst disease markers using the exchangeable correlation pattern. * in PSSA(*) denotes the number of model markers (i.e., covariates) for PSSA method; we multiplied each MSE value by 100; The bolded simulation condition are consistent with the conditions observed for our numeric example of hypertension