Table 4.

Relationship between smoking status and acute myocardial infarction as shown using three different methods to account for missing values and analyzed using conditional logistic regression (odds ratio, 95% confidence interval). The never smoking group was used as the reference.

Smoking status	Case Deletion	Indicator Variable	Multiple Imputation (10 copies)
Crude Estimates of Effect
Ever	1.92 (1.84–2.00)	1.92 (1.84–2.00)	1.90 (1.84–1.97)
Never	Reference	Reference	Reference
Missing	n/a	0.88 (0.82–0.95)	n/a
Adjusted* Estimates of Effect
Ever	1.83 (1.75–1.91)	1.83 (1.75–1.91)	1.81 (1.75–1.87)
Never	Reference	Reference	Reference
Missing	n/a	0.86 (0.79–0.94)	n/a

n/a: not applicable.

* matched for age, GPRD practice and index date and adjusted for age, sex, heavy alcohol use, smoking and number of hospitalizations in the past year.