Predictors of influenza severity among hospitalized adults with laboratory confirmed influenza: Analysis of nine influenza seasons from the Valencia region, Spain

Purpose Influenza hospitalizations contribute substantially to healthcare disruption. We explored the impact of ageing, comorbidities and other risk factors to better understand associations with severe clinical outcomes in adults hospitalized with influenza. Methods We analysed multi‐season data from adults ≥18 years, hospitalized with laboratory‐confirmed influenza in Valencia, Spain. Severity was defined as intensive care unit (ICU) admission, assisted ventilation and/or death. Generalized estimating equations were used to estimate associations between risk factors and severity. Rate of hospital discharge was analysed with a cumulative incidence function. Results Only 26% of influenza patients had their primary discharge diagnosis coded as influenza. Comorbidities were associated with severity among adults aged 50–79 years, with the highest odds ratio (OR) in patients with ≥3 comorbidities aged 50–64 years (OR = 6.7; 95% CI: 1.0–44.6). Morbid obesity and functional dependencies were also identified risk factors (ORs varying from 3 to 5 depending on age). The presence of increasing numbers of comorbidities was associated with prolonged hospital stay. Conclusions Influenza clinical outcomes are aggravated by the presence of comorbidities and ageing. Increased awareness of influenza among hospitalized patients could prompt clinical and public health interventions to reduce associated burden.


| Data source and study protocol
We analysed hospital-based influenza surveillance data collected from the Valencia region of Spain during influenza seasons 2010/2011 through 2018/2019, collected following a standard prospective active surveillance study protocol as previously described. 16,17 Briefly, patients presenting with a protocol-defined respiratory, cardiovascular or other specified complaints, hospitalized for minimum one night, non-institutionalized and resident in the hospital catchment area were eligible for screening. Those presenting with at least one respiratory and one systemic symptom with an onset of <7 days, as per the European Centre for Disease Prevention and Control influenza-likeillness (ILI) case definition, were invited to join the study. 18 After informed consent, detailed clinical and demographic data were gathered through patient interview and medical record abstraction into a study database. Pharyngeal and nasopharyngeal swab samples from all study participants were collected and tested for influenza by reverse transcription polymerase chain reaction (RT-PCR). Patients were followed up during hospitalization with collection of data on clinical progression, treatments and discharge.
To analyse the impact of comorbidities (listed in Table 1), patients were categorized into (i) those with no underlying chronic medical conditions and (ii) those with one, (iii) two and (iv) three or more conditions. Selected comorbidities were also evaluated separately as risk factors by comparing patients with each comorbidity to those without.
Primary discharge diagnoses associated with each hospitalization were described based on International Classification of Diseases (ICD) codes. Because discharge diagnoses were recorded in ICD-9 and ICD-10 codes depending on the season, diagnoses in ICD-9 were converted to corresponding ICD-10 diagnostic groups for analysis (Table S1).
We defined severe clinical outcomes using a binary composite indicator based on feedback from site investigators and review of the literature. The indicator included either ICU admission, 3,5,21 mechanical ventilation or extracorporeal membrane oxygenation (ECMO) [21][22][23] or death at any time during the patient's hospitalization. 21,22 2.4 | Severity risk factor analysis Univariate associations between potential risk factors and influenza severe outcome, as defined by the indicator above, for the entire study population were assessed using Pearson's chi-squared test.
Adjusted multivariable associations were estimated using generalized estimating equations (GEE). 24 Our model assumed both a correlation within hospitals (different clinical practices) and within seasons (different strain circulation and severity). 24 To analyse the contribution of comorbidities independently from age, we developed parsimonious models for each age group to estimate adjusted odds ratios (OR) for severity and 95% confidence intervals (CI) using a robust covariance estimator. Model selection was performed, explicitly retaining information on the number of comorbidities as the primary research question, by removing variables if (i) they were not confounding the relationship between severity and comorbidity (i.e. did not change the

| Length of stay analysis
We also explored length of hospital stay as a proxy for severity. Association between median length of hospitalization with increasing age (age groups) was compared with the non-parametric test for trend across ordered groups. A cumulative incidence function was  Of those, 3180 were adults with laboratory-confirmed influenza that were further included in this analysis. Baseline characteristics of study population overall and by severity are described in Table 1

| Primary discharge diagnoses of hospitalized patients
Information on primary diagnoses was available for 3087 subjects (>97% of the study population), described in Table 3. Most were of respiratory cause (86%), of which influenza was the most common (920 subjects, 30%). However, only 26% of the laboratory confirmed influenza patients had their primary discharge diagnosis coded as influenza. Pneumonia was recorded for 470 subjects (15%), whereas 456 (15%) and 586 patients (19%) were recorded with chronic respiratory and other respiratory diseases, respectively. Non-respiratory diagnoses were less frequent: 142 patients (5%) had recorded circulatory events including heart attack and cardiac insufficiency (Tables 3   and S1). No substantial differences in discharge diagnosis frequency were observed by age group/severity.

| Multivariate risk factor analysis
Model selection strategy was based on four GEE models (one per age group

| Length of stay analysis
The median length of hospitalization was significantly longer for older than younger patients (test-for-trend across groups: P < 0.001; Table 2). Patients with ≥3 comorbidities were discharged later than those with fewer comorbidities irrespective of age, though these rela-  Neoplasms 11 (0) 0 (0) 1 (0) 6 (1) 4 (0) 9 (0) 2 (1) Endocrine system diseases 20 (1)  5 (2) 3 (1) 6 (1) 6 (1) 16 (1)  4 (2) Circulatory system diseases 142 (5)  4 (1) 18 (4) 55 (5) 65 (6) 127 (4) 15 (6) Mental disorders Musculoskeletal and connective tissue diseases  We identified associations between the prevalence of comorbidities and the frequency of severe outcomes among influenza hospitalized patients 50-79 years in Spain. BMI, virus strain, smoking and increased levels of functional dependency among the older population were also associated with influenza severity, albeit at different magnitudes across age groups. Patients with comorbidities and of older age also experienced longer hospitalization. Although respiratory outcomes were the most common discharge diagnoses across age groups, a considerable proportion of hospitalized patients with influenza had non-respiratory outcomes as their primary discharge diagnosis, indicative of the range of clinical outcomes associated with influenza. Presence of comorbidity could not entirely explain disease severity in the young age group (18-49 years). Nonetheless, despite lack of risk factors, this group was vulnerable to severe disease outcomes due to H1N1pdm09 virus infection, with 20% having pneumonia as their main discharge diagnosis. This is a reminder that influenza virus infection can also lead to severe disease in otherwise healthy young adults who could benefit from influenza vaccination-which is not currently recommended for this age group in European countries.
We showed that severity increased steadily with age, from 6% among 18-49 years to 50% among those ≥80 years, with 26% of inhospital influenza death among those 65-79 years and 68% among ≥80 years (despite lower ICU admission rates in the latter age group, possibly a consequence of hospital care management decisions). Similarly, other studies have described that laboratory-confirmed influenza cases have mortality rates increasing with ageing, although ICU admission would be requested more often for 40-to 79-year age group than those ≥80 years. 27 Older age, in its own right, is associated with deterioration of the immune system in producing an efficient response to infections or to developed immunity after vaccination, both of which are associated with mortality. 28 However, the most challenging expression of population ageing is the clinical condition of frailty. 29 It is estimated that as many as 50% of people ≥85 years are frail, 30 which strongly predicts not only mortality but also cognitive decline, disability and institutionalization. 31 Indeed, in our findings, in which we used functional dependency as a proxy for frailty, the odds of severe outcomes increased over threefold in the most frail individuals aged 65-79 and ≥80 years compared with those who were functionally independent at admission. Annual influenza vaccination can provide protection from severe influenza-associated outcomes among older adults, including hospitalizations, ICU admissions and death even if the vaccine does not protect from infection. 32 The availability of more immunogenic vaccines could have an even greater impact.
Comorbidities as predictors of influenza severity have been the subject of previous research, with reported higher risk of ICU admission and death in patients with specific comorbidities in the United States, mitigated by antiviral treatment or vaccination. 6,32 In our study, presence of comorbidity was associated with prolonged hospitalization, which drives influenza-associated healthcare costs. 33,34 Another study of hospitalized influenza, using a similar severity definition, identified comorbidities such as diabetes and obesity to predispose complications in young adults aged 15-49 years; F I G U R E 1 Multivariate risk factor analysis results from the four final GEE models. Odds ratios for severe influenza are represented on the xaxis (Note: Some confidence intervals exceed the x-axis scale). Study exposures are listed in the y-axis but these were less apparent in older adults. 7 39 In our study, though we included influenza vaccination status in our severity risk analysis, this did not survive model selection.
Furthermore, we could not explore the effect of vaccination on severity of disease due to the potential for indication and health-user biases that could not be addressed in the analysis.
Our study had some limitations. Despite the large dataset, severe outcomes are rare, which affected analytical power; moreover, deaths occurring after discharge were not captured. The final sample of severe laboratory-confirmed influenza cases was therefore small, reducing the statistical power of the analysis and hindering our ability to assess further any potential impact of confounders. This was especially evident in the younger adult population, where case numbers were smallest. Our analysis was not designed to assess the impact of vaccination on severity, for which a different modelling strategy would have been required. Our modelling approach involves assumptions around data correlation, but results were qualitatively robust to sensitivity analyses using logistic or random effects models, increasing our confidence in reported results. Observational studies are affected by biases; for example, the decreased odds of severe outcomes in some group with one comorbidity may be a consequence of a lower clinical threshold for hospitalizing patients with comorbidities versus those with none, and interpretation of results needs to bear unmeasured confounders. Many individuals had more than one comorbidity, and we did not consider the impact of specific medical history on severity and could not assess whether some patients' comorbidities were better controlled than others', which could further affect interpretation.
In conclusion, our results confirm that influenza is an underappreciated disease that can cause severe clinical outcomes in adults of all ages, being further impacted by the presence of comorbidities and ageing. Increased awareness of influenza among hospitalized patients may have important impact for patients who could benefit from early antiviral therapy. Furthermore, availability of influenza vaccine formulations affording improved protection for adults, especially older adults and those with comorbidities, could minimize associated burden and healthcare resource consumption.