Table 3.

Multivariable Analyses of Risk Factors associated with Severe Illness due to 2009 Pandemic Influenza A (H1N1) Virus Infection

Characteristic	OR (95% CI)	Pa
Risk factors associated with severe 2009 H1N1 illness
Among nonpregnant patients ≥ 2 years of age
Male sex (vs female sex)	1.19 (1.07–1.31)	.001
Delayed hospital admission (on symptom day ≥3 vs day <3)	1.27 (1.13–1.42)	<.001
Age and chronic medical conditionsb
2–17 years of age and without chronic medical conditions	Ref	Ref
18–59 years of age and without chronic medical conditions	2.12 (1.86–2.41)	<.001
≥ 60 years of age and without chronic medical conditions	3.49 (2.51–4.85)	<.001
2–17 years of age and with chronic medical conditions	3.93 (3.03–5.08)	<.001
18–59 years of age and with chronic medical conditions	5.37 (4.55–6.34)	<.001
≥ 60 years of age and with chronic medical conditions	6.88 (5.54–8.55)	<.001
Age and obesityb
2–17 years of age
Not obese	Ref	Ref
Obese	1.34 (1.10–1.63)	.004
18–59 years of age
Not obese	Ref	Ref
Obese	1.91 (1.57–2.31)	<.001
≥ 60 years of age
Not obese	Ref	Ref
Obese	0.68 (.37–1.25)	.211
Among female patients of reproductive age (15–49 years of age)
Age ≥35 years (vs 15–34 years)	1.40 (1.10–1.77)	.006
Delayed hospital admission (on symptom day ≥3 vs day <3)	1.80 (1.47–2.19)	<.001
Chronic medical conditions and pregnancyb
Without chronic medical conditions and nonpregnant	Ref	Ref
With chronic medical conditions and nonpregnant	3.62 (2.65–4.94)	<.001
Without chronic medical conditions but pregnant	3.30 (2.72–4.00)	<.001
With chronic medical conditions and pregnant	3.69 (2.15–6.31)	<.001
Effectiveness of early antiviral treatment on severe 2009 H1N1 illness among nonpregnant patients ≥ 2 years of agec
Started on symptom day 1–2	Ref	Ref
Started on symptom day 3–4	1.01 (.84–1.22)	.885
Started on symptom day ≥5	1.42 (1.20–1.67)	<.001

NOTE. CI, confidence interval; OR, odds ratio; Ref, reference.

^aMultivariable logistic regression analyses were performed. Age, obesity, pregnancy, and chronic medical conditions were not highly correlated with each other (Spearman's rank correlation coefficients were calculated; r<.38 for all), which indicated little evidence for taking account of collinearity in multivariate analyses.

^bChunkwise likelihood ratio tests suggested that there are statistical 2-way interactions between age and chronic medical conditions (P = .003), between age and obesity (P = .001), and between chronic medical conditions and pregnancy (P < .001) but no 2-way statistical interaction between chronic medical conditions and obesity (P = .073). Three-way interaction among age, chronic medical conditions, and obesity was evaluated as part of a model containing all possible 2-way interactions but was not statistically significant (P = .489). The final model was used to estimate OR within strata defined by interactions between age and chronic diseases and between age and obesity. Stratum-specific ORs equal e^β, where e denotes the base of natural logarithms and β denotes regression efficient. ORs were calculated as follows: eg, compared with reference group (those 2–17 years of age without chronic medical conditions), for those with chronic medical conditions and ≥ 60 years of age, β=β (≥ 60 years of age) + β (chronic medical conditions) + β (interaction term); here, the latter 3 β values were obtained from the multivariable logistic regression model.

^cOnly patients who received antivirus treatment and have had clinical outcome defined as discharged from the hospital or died during study period were included in the analysis.