Skip to main content

Table 2 Comparison of classification methods for estimating BSI rate and difference in adjusted rates with 10% identifier error (simulated datasets 10-12)

From: Evaluating bias due to data linkage error in electronic healthcare records

Classification

N Infections

Crude rate (95% CI) N Ă— 1000 54 , 826 bed days

Standard error

% bias

Difference in adjusted rates (95% CI)

Standard error

% bias

Gold standard

1000

18.24

0.577

 

5.514

1.214

 

10: Random error

HW

869

15.84 (14.79, 16.90)

0.646

-13.1

4.55 (2.33, 6.76)

1.129

-17.5

PII MPD = 0.1

1038

18.94 (17.67, 20.21)

0.646

3.8

5.32 (2.75, 7.89)

1.313

-3.5

PII MPD = 0.9

860

15.69 (14.61, 16.77)

0.551

-14.0

4.45 (2.18, 6.72)

1.160

-19.3

11: Non-random error; by covariate

HW

886

16.15 (15.09, 17.21)

0.543

-11.4

10.93 (8.61, 13.24)

1.183

98.2

PII MPD = 0.1

1010

18.41 (17.21, 19.62)

0.614

1.0

11.69 (9.12, 14.26)

1.311

111.9

PII MPD = 0.9

858

15.65 (14.57, 16.72)

0.548

-14.2

11.454 (9.09, 13.82)

1.208

107.7

12: Non-random error; by outcome

HW

364

6.65 (5.98, 7.32)

0.343

-63.6

1.94 (0.53, 3.35)

0.720

-64.9

PII MPD = 0.1

684

12.48 (10.87, 14.09)

0.822

-31.6

3.36 (0.51, 6.20)

1.453

-39.1

PII MPD = 0.9

217

3.95 (3.39, 4.51)

0.287

-78.3

1.20 (0.00, 2.39)

0.610

-78.3