Table 4. Results from logistic regression models to evaluate exposure length variables in investigation of invasive extrapulmonary NTM infections among patients who underwent cardiothoracic surgery, York, Pennsylvania, USA, 2015*.

Regression models	Odds ratio (95% CI)	p value
Univariable models†
Patient characteristics
Age, per 1-y increase	1.0 (1.0–1.1)	0.4
Male sex	3.0 (0.7–16.8)	0.2
White race	1.8 (0.2–244.3)	0.7
Chronic lung disease: yes	2.9 (0.7–11.4)	0.1
Diabetes: yes	0.3 (0.03–1.5)	0.2
Immunocompromised: yes	1.5 (0.3–6.0)	0.6
Length of surgical exposure
Surgical time with HCD, h
0	Referent
>0 to <4	1.8 (0.2–15.4)	0.6
≥4 to <5	2.6 (0.2–23.6)	0.4
≥5	15.6 (3.2–103.9)	0.002
Time on bypass, h
0	Referent
>0 to <2	1.7 (0.2–12.2)	0.6
≥2	19.0 (3.7–133.0)	0.001
Final logistic model‡
Model 1.1, only surgical time with HCD retained in the final model: Surgical time with HCD >5 h	13.2 (3.2–62.9)	0.0008
Model 1.2, includes surgical time with HCD and immunocompromised status
Surgical time with HCD >5h	13.6 (3.3–68.8)	0.001
Immunocompromised: yes	2.2 (0.4–12.2)	0.4
Model 2.1, only time on bypass retained in the final model: Time on bypass >2 h	16.5 (3.8–84.0)	0.0004
Model 2.2, includes time on bypass and immunocompromised status
Time on bypass >2h	16.6 (3.8–88.4)	0.0006
Immunocompromised: yes	2.1 (0.3–12.0)	0.4

*HCD, heater–cooler device; NTM, nontuberculous mycobacteria.
†Regression models for factors considered in the multivariable model; each regression model used Firth’s penalized maximum likelihood method.
‡Each final regression model used Firth’s penalized maximum likelihood method; other variables (sex, age, immunocompromised status, chronic lung disease) were removed because of nonsignificance.