This study forms part of the “Risk factors of hospitalization due to influenza A (H1N1) 2009 and effectiveness of pharmacological and non-pharmacological interventions in its prevention. Case–control study project”, led by the Centres for Biomedical Research in Epidemiology and Public Health (CIBERESP) and Respiratory Diseases (CIBERES).9 The study was conducted in 29 hospitals in 7 Spanish regions (Andalusia, the Basque Country, Castile and Leon, Catalonia, Madrid, Navarre and Valencia Community).

Cases and controls were recruited between July 2009 and April 2011. We selected cases among patients hospitalized for >24h with influenza-like illness, acute respiratory infection, septic shock or multiple organ failure in whom influenza virus A (H1N1) 2009 infection was confirmed by real-time reverse-transcription polymerase chain reaction (RT-PCR).10 Participating hospitals had protocols for systematic swabbing of patients admitted with influenza-like illness, defined as the sudden onset of any general symptom (fever or febrile symptoms, headache, myalgia), in addition to any respiratory symptom (cough, sore throat, shortness of breath).11 We excluded patients with nosocomial infection, defined as influenza appearing 48h or more after admission for another cause.

For each hospital case, we selected two matched controls: one patient consulting for influenza confirmed by RT-PCR (outpatient case) and another patient consulting for any reason other than influenza-like illness or acute respiratory infection (outpatient control).

Both outpatient cases and outpatient controls were recruited in primary care centers of the same geographical area as the hospitalized cases, from among the population of patients attending the primary care center for reasons other than acute respiratory infection or influenza-like illness. No cases were recruited from hospital-based emergency departments.

Controls were matched with each case according to age (±5 years), date of hospitalization (±10 days) and province of residence. Occasionally, more than one patient who met the criteria for pairing with the same case was recruited.

Patients were recruited prospectively and retrospectively. If no admissions were made in a hospital center after the beginning of the study, cases and controls were recruited retrospectively with the same criteria as for prospectively recruited patients.

Patients who did not give written consent were excluded and for this paper, we excluded patients <18 years old, BMI<18.5kg/m2, and pregnancy.

The study was approved by the Ethics Committee of the hospitals involved. Written informed consent was obtained from all patients included in this study.

Specifically trained health professionals used a structured questionnaire to collect information on cases and controls by interview and review of clinical charts. This information included sociodemographic variables, current smoking habit, alcohol abuse (defined as consumption was greater than 4 standard units a day for women and 6 for men), pregnancy, major chronic conditions, unplanned hospitalization in the previous 12 months, confinement to bed in the 7 days before the hospitalization or visit, and other pre-existing risk factors for complicated influenza. The following major clinical conditions were recorded: chronic respiratory disease, asthma, cardiovascular disease, chronic renal disease, diabetes mellitus, immunodeficiency, neurological disease and disability, and cancer. Other risk factors for influenza complications previous to hospitalization or the medical visit were collected: chronic respiratory failure, chronic heart failure, pneumonia in the previous 2 years, systemic corticoids, and inhaled corticoid treatments. Information on the monovalent influenza A (H1N1)pdm09, 2010–2011 seasonal influenza and pneumococcal vaccination status was obtained from medical records or vaccination cards. Cases were considered vaccinated with the seasonal or pandemic influenza vaccines if they had received a dose of the vaccine at least 14 days before symptom onset. Controls were considered vaccinated if they had received the dose at least 14 days before symptom onset in the matched case.

BMI was obtained from direct or telephonic interview with the patients by healthcare workers. BMI was used as a continuous and categorical variable (normal weight: BMI 18.5–24.9kg/m2; overweight: 25.0–29.9kg/m2; obesity grade I: 30–34.9kg/m2; obesity grade II: 35–39.9kg/m2, and obesity grade III: BMI≥40kg/m2).1,12

Hospital cases were compared with outpatient cases and outpatient controls to estimate the risk of influenza-associated hospitalization. A bivariate comparison of demographic variables, medical conditions and BMI was made between cases and controls using the χ2 test for categorical variables and the Student's t test for continuous variables. The odds ratios (OR) for BMI, with 95% confidence intervals (CI), were calculated using conditional logistic regression models adjusted for potential confounders. The comparison was adjusted by sex, age (≥65 vs <65 years), ethnicity (Caucasian vs non-Caucasian), studies (primary education or less vs education level higher than primary), alcohol abuse (yes or no), smoking status (current vs not current smoker), history of pneumonia in the two previous years (yes or no), influenza vaccination, and risk factors (yes or no). One or more of the following problems were considered risk factors: lung diseases, cardiovascular diseases, kidney diseases, immunosuppressive diseases, neurological diseases, diabetes, immunosuppressive treatment or acetylsalicylic acid treatment.

We performed restriction analysis for the following subgroups: age (<65 years), influenza vaccination (unvaccinated subjects) and season (2009–2010 and 2010–2011). The odds ratios (OR) for BMI (continuous and categorical), with their 95% confidence intervals (CI), were calculated using conditional logistic regression, aOR were calculated using conditional logistic regression models adjusted for those all variables (Table 1).

The population attributable fraction (PAF) was calculated using the formula:

Fig. 1 shows the selection algorithm for cases and controls for this study, selected from the population which participated in the “Risk factors of hospitalization due to influenza A (H1N1) 2009 and effectiveness of pharmacological and non-pharmacological interventions in its prevention. Case–control study project”. According to the selection criteria of this study, 27.6% of hospitalized cases, 32.8% of outpatient cases, and 22.5% of outpatient controls were excluded. No information on height and weight was available for 29.6% of the hospitalized cases, 12.9% of the outpatient cases, and 11.1% of the outpatient controls. A total of 755 hospitalized patients, 783 outpatient cases, and 950 outpatient controls met the selection criteria and provided height and weight data. Finally, conditional logistic regression analyses were performed on 521 hospitalized cases, and 521 paired outpatient cases, and on 625 hospitalized cases and 625 paired outpatient controls. Table 1 shows the distribution of the study variables among the study subjects. Hospitalized cases were mainly men (60.1%), 22% were 65 years of age or older, 40.9% had primary level education or less, 30.4% were current smokers, 12.5% at risk drinkers, 10.6% had a history of pneumonia in the previous 2 years, 69% had some risk factor or risk treatment for developing influenza, and 11.5% had been vaccinated against influenza. Outpatient cases were 49% men, 7.9% aged 65 years or older, 19.8% had primary level education or less, 26.8% were current smokers, 4.6% at risk drinkers, 5.6% had a history of pneumonia in the previous 2 years, 28% had some risk factor or risk treatment for developing influenza, and 6.3% had been vaccinated against influenza. Outpatient controls were 43% men, 18.4% aged 65 years or older, 31.3% had primary level education or less, 23.2% were current smokers, 4.9% at risk drinkers, 2.8% had a history of pneumonia in the previous 2 years, 31.7% had some risk factor or risk treatment for developing influenza, and 16.3% had been vaccinated against influenza.

Fig. 1.

Algorithm missing in hospitalized cases, outpatient cases and ambulatory controls.

(0.34MB).

The prevalence of obesity and BMI values averages were 27.7% and 27.8kg/m2; 14.3% and 25.4kg/m2, and 17.8% and 26.3kg/m2 in hospitalized cases, outpatients cases and outpatients controls, respectively.

The causes of hospital admission were: pneumonia, 192 cases (25.4%); acute respiratory distress syndrome, 115 cases (15.2%); sepsis, 72 cases (9.5%); and influenza, 376 cases (49.8%).

Hospital cases vs outpatient cases: There were significant differences in the BMI distribution between hospitalized and outpatient cases (28.1kg/m2 vs 25.2kg/m2; P<0.0001), with a higher prevalence of obesity in hospitalized cases (28.1% vs 12.9%; P<0.0001) (Table 2).

The odds of hospitalization in patients diagnosed with influenza (comparison between hospitalized and outpatient cases) was significantly higher in patients with BMI 30–34.9kg/m2 (aOR=1.85), 35–39.9kg/m2 (aOR=5.24), and ≥40kg/m2 (aOR=44.36), compared with patients of normal weight (Table 2). In the case of BMI as a continuous variable, a greater risk of hospitalization was observed with a higher BMI (aOR=1.11; 95% CI=1.07–1.15) (Table 2).

Hospitalized cases vs outpatient controls: There were significant differences in the BMI distribution between hospitalized cases and outpatient controls (27.7kg/m2 vs 26.2kg/m2; P<0.0001), with a higher prevalence of obesity in hospitalized cases (26.0% vs 16.3%; P<0.0001) (Table 3).

The odds of being hospitalized was significantly higher in patients with BMI 35–39.9kg/m2 (aOR=4.37) and ≥40kg/m2 (aOR=4.95), compared with normal weight (Table 3). When BMI was taken as a continuous variable, a greater risk of hospitalization was observed with a higher BMI (aOR=1.04; 95% CI=1.01–1.07) (Table 3).

Restriction analysis: BMI was associated with an increased risk of hospitalization in patients not vaccinated against flu compared with both outpatient cases (aOR=1.11) and outpatient controls (aOR=1.06) (Tables 4 and 5). Compared with normal weight, the risk of hospitalization was increased with each BMI category from obesity grade I (BMI≥30kg/m2; aOR 1.87) to BMI≥40kg/m2 (aOR=47.51) in both comparison groups (Tables 4 and 5).

BMI as a quantitative variable was associated with an increased risk of hospitalization in patients younger than 65 years in both comparison groups (aOR=1.14 and aOR=1.04) (Tables 4 and 5). Compared with normal weight, a risk of hospitalization was observed in obesity grade II (aOR=14.41; aOR=3.89) in both comparison groups, compared with outpatient cases also in the category of BMI 30–34 (aOR=2.72).

Similarly, in the two seasons studied, BMI was associated with a risk of hospitalization in both comparison groups (aOR=1.12 and aOR=1.14; aOR=1.05 and aOR=1.04). Compared with normal weight, BMI≥35kg/m2 was associated with an increased risk of hospitalization (Tables 4 and 5) in both seasons.

Fraction of hospitalized cases of influenza attributable to BMI: In patients infected by influenza, comparison of hospitalized cases and outpatient cases showed that the PAF of hospital admission for BMI was 21.9%. Comparison of hospitalized cases and outpatient controls showed a PAF of hospitalization for influenza for BMI of 8.5%. In patients unvaccinated against influenza, the contribution of BMI to hospitalization for influenza ranged between 20.5% and 16.9% depending on the type of comparison.