Identification of Subgroups with Nondifferential Exposure Misclassification

Because the effects of errors in exposure that are the same for diseased and non-diseased individuals tend to produce more predictable errors, often leading to bias toward the null value, there is value in trying to create strata in which the error is likely to be nondifferential. It would be preferable, of course, to avoid error altogether, but stratification to achieve nondifferential misclassification is still of benefit.

When a determinant of exposure accuracy, for example, educational level, is also associated with disease, the results will be distorted due to differential mis-classification of exposure if education is ignored. In other words, the imbalance in educational levels among those with and without disease, and the association between educational level and accuracy of reported exposure, results in differ-

Table 8.8. Standardized Incidence Ratios for All Types of Cancer Combined and for Tobacco-Related Cancers, by Diagnostic Group, in Patients Hospitalized with an Effective Disorder in Denmark, 1969-1993

Portion of Follow-up Period

Total First Year of Follow-up 1-9 Years of Follow-up > 10 Years of Follow-up

Table 8.8. Standardized Incidence Ratios for All Types of Cancer Combined and for Tobacco-Related Cancers, by Diagnostic Group, in Patients Hospitalized with an Effective Disorder in Denmark, 1969-1993

Total First Year of Follow-up 1-9 Years of Follow-up > 10 Years of Follow-up

DIAGNOSIS

OBS

SIR*

95% CI

OBS

SIR

95% CI

OBS

SIR

95%CI

OBS

SIR

95% CI

Total Cohort

9922

1.05

1.03,

1.07

654

1.19

1.11,

1.29

4655

1.02

0.99,

1.05

4613

1.07

1.03, 1.10

Tobacco-related cancers^

2813

1.21

1.16,

1.25

182

1.37

1.18,

1.59

1224

1.09

1.03,

1.16

1407

1.30

1.23, 1.37

Non-tobacco-related cancers

7109

1.00

0.98,

1.02

472

1.14

1.04,

1.25

3431

1.00

0.97,

1.03

3206

0.99

0.95, 1.02

Brain cancer

277

1.18

1.00,

1.32

46

3.27

2.39,

4.36

142

1.24

1.04,

1.46

89

0.84

0.67, 1.03

Other

6832

0.99

0.97,

1.02

426

1.06

0.97,

1.17

3289

0.99

0.96,

1.02

3117

0.99

0.96, 1.03

Diagnostic Levels

Bipolar psychosis

1217

0.99

0.93,

1.03

62

1.16

0.89,

1.49

557

1.00

0.92,

1.09

598

0.94

0.87, 1.02

Tobacco-related cancers^

292

0.92

0.82,

1.04

18

1.30

0.77,

2.06

133

0.94

0.78,

1.11

141

0.88

0.74, 1.04

Non-tobacco-related cancers

925

1.00

0.93,

1.06

44

1.11

0.80,

1.48

424

1.02

0.93,

1.12

457

0.97

0.88, 1.06

Brain cancer

26

0.82

0.53,

1.20

4

2.72

0.73,

6.97

12

0.81

0.42,

1.42

10

0.64

0.31, 1.17

Other

899

1.00

0.94,

1.07

40

1.04

0.75,

1.42

412

1.03

0.93,

1.13

447

0.98

0.89, 1.07

Unipolar Psychosis

4345

0.98

0.95,

1.01

290

1.00

0.89,

1.12

2176

0.94

0.90,

0.98

1879

1.03

0.99, 1.08

Tobacco-related cancers^

1144

1.05

0.99,

1.11

76

1.07

0.84,

1.34

538

0.95

0.87,

1.03

530

1.18

1.08, 1.29

Non-tobacco-related cancers

3201

0.96

0.93,

1.00

214

0.98

0.85,

1.12

1638

0.94

0.90,

0.99

1349

0.98

0.93, 1.04

Brain cancer

119

1.19

0.99,

1.43

21

3.10

1.92,

4.75

58

1.11

0.84,

1.43

40

0.99

0.70, 1.34

Other

3082

0.96

0.92,

0.99

193

0.91

0.79,

1.05

1580

0.94

0.89,

0.99

1309

0.98

(continued)

Table 8.8. Standardized Incidence Ratios for All Types of Cancer Combined and for Tobacco-Related Cancers, by Diagnostic Group, in Patients Hospitalized with an Effective Disorder in Denmark, 1969-1993 (continued)

Portion of Follow-up Period

Table 8.8. Standardized Incidence Ratios for All Types of Cancer Combined and for Tobacco-Related Cancers, by Diagnostic Group, in Patients Hospitalized with an Effective Disorder in Denmark, 1969-1993 (continued)

Portion of Follow-up Period

DIAGNOSIS

Total

First Year of Follow

-up

1-9 Years of Follow-up

> 10 Years of Follow-up

OBS

SIR*

95% CI

OBS

SIR

95% CI

OBS

SIR

95%CI

OBS

SIR

95% CI

Reactive Depression

2075

1.13

1.08,

1.18

184

1.62

1.39,

1.87

997

1.12

1.05, 1.19

894

1.07

1.00, 1.14

Tobacco-related cancers^

663

1.41

1.30,

1.52

58

2.03

1.54,

2.62

297

1.32

1.17, 1.48

308

1.43

1.27, 1.59

Non-tobacco-related cancers

1412

1.03

0.98,

1.09

126

1.48

1.23,

1.76

700

1.06

0.98, 1.14

586

0.95

0.87, 1.02

Brain cancer

59

1.20

0.92,

1.55

14

4.55

2.48,

7.63

29

1.21

0.81, 1.74

16

0.73

0.42, 1.18

Other

1353

1.03

0.97,

1.08

112

1.36

1.12,

1.64

671

1.05

0.97, 1.13

570

0.95

0.88, 1.03

Dysthymia

2285

1.18

1.13,

1.23

118

1.32

1.09,

1.58

925

1.15

1.08, 1.23

1242

1.19

1.13, 1.26

Tobacco-related cancers^

714

1.56

1.45,

1.68

30

1.58

1.07,

2.26

256

1.40

1.24, 1.59

428

1.68

1.52, 1.84

Non-tobacco-related cancers

1571

1.06

1.01,

1.12

88

1.25

1.00,

1.54

669

1.07

0.99, 1.16

814

1.04

0.97, 1.11

Brain cancer

73

1.34

1.05,

1.68

7

2.54

1.02,

5.24

43

1.80

1.30, 2.43

23

0.82

0.52, 1.23

Other

1498

1.05

1.00,

1.11

81

1.19

0.95,

1.48

626

1.04

0.97, 1.13

791

1.04

0.97, 1.12

♦Observed number of cases/expected number of cases. The expected number of cases was the number of cancer cases expected on the basis of age-, sex-, and calendar-year-specific incidence rates of first primary cancers in Denmark.

fCancers of the buccal cavity, larynx, lung, esophagus, pancreas, kidney, and urinary bladder.

¿Bipolar psychosis: ICD-8 codes 296.39, 296.19, and 298.19; unipolar psychosis: ICD-8 codes 296.09, 296.29, 296.89, and 296.99; reactive depression: ICD-8 code 298.09, dysthymia: ICD-8 codes 300.49 and 301.19.

Obs, observed; SIR, standardized incidence ratio; CI, confidence interval; ICD-8, International Classification of Diseases, Eighth Revision. Dalton et al., 2002.

♦Observed number of cases/expected number of cases. The expected number of cases was the number of cancer cases expected on the basis of age-, sex-, and calendar-year-specific incidence rates of first primary cancers in Denmark.

fCancers of the buccal cavity, larynx, lung, esophagus, pancreas, kidney, and urinary bladder.

¿Bipolar psychosis: ICD-8 codes 296.39, 296.19, and 298.19; unipolar psychosis: ICD-8 codes 296.09, 296.29, 296.89, and 296.99; reactive depression: ICD-8 code 298.09, dysthymia: ICD-8 codes 300.49 and 301.19.

Obs, observed; SIR, standardized incidence ratio; CI, confidence interval; ICD-8, International Classification of Diseases, Eighth Revision. Dalton et al., 2002.

ential misclassification of exposure. The solution in this example is simple: stratify on educational level and examine the exposure-disease association within education strata. Within the strata, the accuracy of reporting exposure is more homogeneous in that one source of variability has been controlled, so that remaining errors are more nondifferential in character. There may be other contributors to differential exposure accuracy that remain or educational level may have been the sole basis for the differential. Regardless, stratification by education will be beneficial.

Note that education, in this example, is a marker of accuracy, not a confounder, and should therefore not be adjusted. In fact, under the assumption that more highly educated respondents provide more accurate data, the expectation would be that the most accurate results would be obtained in the high education stratum (less nondifferential misclassification). In order to implement this strategy, the markers of classification accuracy must be known, measured, and not too strongly related to disease status. At the extreme, if the marker of accuracy is perfectly correlated with disease status, it cannot be separated from disease in the analysis.

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook


Post a comment