Resource utilization and cost of influenza requiring hospitalization in Canadian adults: A study from the serious outcomes surveillance network of the Canadian Immunization Research Network

Background Consideration of cost determinants is crucial to inform delivery of public vaccination programs. Objectives to estimate the average total cost of laboratory‐confirmed influenza requiring hospitalization in Canadians prior to, during, and 30 days following discharge. To analyze effects of patient/disease characteristics, treatment, and regional differences in costs. Methods Study utilized previously recorded clinical characteristics, resource use, and outcomes of laboratory‐confirmed influenza patients admitted to hospitals in the Serious Outcomes Surveillance (SOS), Canadian Immunization Research Network (CIRN), from 2010/11 to 2012/13. Unit costs including hospital overheads were linked to inpatient/outpatient resource utilization before and after admissions. Results Dataset included 2943 adult admissions to 17 SOS Network hospitals and 24 Toronto Invasive Bacterial Disease Network hospitals. Mean age was 69.5 years. Average hospital stay was 10.8 days (95% CI: 10.3, 11.3), general ward stays were 9.4 days (95% CI: 9.0, 9.8), and ICU stays were 9.8 days (95% CI: 8.6, 11.1) for the 14% of patients admitted to the ICU. Average cost per case was $14 612 CAD (95% CI: $13 852, $15 372) including $133 (95% CI: $116, $150) for medical care prior to admission, $14 031 (95% CI: $13 295, $14 768) during initial hospital stay, $447 (95% CI: $271, $624) post‐discharge, including readmission within 30 days. Conclusion The cost of laboratory‐confirmed influenza was higher than previous estimates, driven mostly by length of stay and analyzing only laboratory‐confirmed influenza cases. The true per‐patient cost of influenza‐related hospitalization has been underestimated, and prevention programs should be evaluated in this context.


| INTRODUCTION
Influenza is a common, highly communicable disease associated with febrile upper and lower respiratory tract infection that can result in serious complications particularly in young children, pregnant women, the elderly, and those with underlying medical conditions. 1 Influenza represents a substantial economic and clinical burden to the healthcare system, with a demonstrable relationship between the circulation of influenza virus and increased healthcare utilization. 2 Severe cases of influenza requiring hospital admission represent the largest component of healthcare costs in the management of influenza, 2 with up to 12 200 attributable hospitalizations annually in Canada. 3

Estimated hospitalization costs vary substantially between
Canadian provinces, possibly due to geographic differences in influenza strain circulation, standard practices for managing treatment, hospital unit costs, and methodologies used for allocating fixed costs. [4][5][6] Previous studies of hospitalization across Canada for patients with a diagnosis of influenza-like illness (ILI) have found lengths of stay and costs ranging from 3.7 days and costs of $2049 in Manitoba 6 to 5.9 days in hospital and costs of $7664 in Ontario, 4 and mean cost per case of $2145 across Canada as a whole. [4][5][6] Understanding cost determinants and their geographic variation is a key to informing the delivery of public vaccination programs. The SOS Network has collected patient demographics and clinical characteristics, as well as treatment and health resource use prior to hospital admission, during hospital stay, and over 30 days following discharge for patients hospitalized with laboratory-confirmed influenza. Data from the SOS Network demonstrated that the length of hospital stay (average 10.8 days) was substantially longer than that reported in prior Canadian studies of ILI diagnoses. 5,6 All regions of Canada are represented although the majority of cases were admitted in Ontario and Quebec.
The objective of this study was to estimate the average direct cost of hospitalization with laboratory-confirmed influenza in Canadian adults by assigning unit costs to detailed resource utilization data collected prior to, during, and for 30 days after hospitalization in a cohort of adults with laboratory-confirmed influenza admitted to participating hospitals of the SOS Network over 3 influenza seasons. Further objectives were to identify the influence of patient and disease characteristics, management, and outcomes on cost, and to explore variation in costs across geographic regions.

| METHODS
The SOS Network conducted active surveillance for influenza among patients aged 16 years and over admitted to participating hospitals with acute respiratory illness. This dataset is comprised of patients with laboratory-confirmed influenza admitted to the 17 participating SOS Network hospitals across 6 provinces and 24 associated sites of the Toronto Invasive Bacterial Disease Network (TIBDN) during the 2010/11, 2011/12, and 2012/13 influenza seasons. For each case, detailed demographic information, surgical history, medical comorbidities, details of hospital care, complications, and influenza outcomes were collected by interview and medical record review. 7 In addition, disease characteristics such as influenza type and subtype were identified by reverse transcription polymerase chain reaction (RT-PCR) on nasopharyngeal swab specimens. 7 The SOS Network also collected data on resource utilization prior to hospital admission, during hospitalization, and for 30 days following discharge. Information was collected on any physician or emergency department visit prior to hospitalization. In order to tabulate in-hospital resource use, general days-on-ward and intensive care unit (ICU) days-on-ward were calculated as the difference between admission and discharge dates. In all SOS Network sites, with the exception of 3 sites in Quebec, length of stay excluded days-on-ward designated as "alternate level of care" (ALC). Alternate level of care is a designation applied to on-ward days spent once patients have been deemed ready for discharge, but who remain in hospital for factors unrelated to the reason for acute hospital care. In all 3 seasons, details of antiviral and antibiotic use prior to and during hospitalization and duration of mechanical ventilation, occurrence of complications, and ICU stay were collected. Following discharge, general days-on-ward and ICU days-on-ward for subsequent hospitalizations within 30 days were collected. Additional detailed information about the types and number of diagnostic tests and procedures performed in the hospital setting were only collected in the 2010/11 and 2011/12 seasons. 8,9 This study linked the resource use of patients with laboratoryconfirmed influenza enrolled by the SOS Network to a single set of unit price weights for each case, regardless of the hospital where treatment was received. These costs for hospital resource use were obtained from Hamilton Health Sciences (HHS), a conglomerate of 7 hospitals in Ontario. Hospital costs were received in the form of unit prices of hospital care incorporating department overheads and fixed costs. Fees for physician services were obtained from the Ontario Schedule of Benefits. 10 Ward per-diem costs were sourced from HHS and included costs for mechanical ventilation and supplemental oxygen; as well as, procedures conducted at the bedside such as intubation, but excluded laboratory tests, diagnostics, and imaging; unit costs for these were provided by HHS separately. Costs for antivirals and antibiotics used during hospitalization were based upon unit prices provided by the Queen Elizabeth II Health Sciences Centre formulary in Halifax, Nova Scotia. 11 For pharmacy costs, expenditure related to the acquisition of medications was excluded and only components attributable to human resources and supplies were included. Costs of outpatient antiviral and antibiotic medications were based upon unit prices listed by the Ontario Drug Benefit (ODB) formulary. 12 Dosing information was not collected by the SOS Network; consequently, the lowest recommended dose from product monographs for severe cases of lower respiratory tract infections was assumed.

| Statistical methods
Surveillance data collected differed in detail for the included years. By Means and 95% confidence intervals of hospitalization cost were reported for different geographic regions, subgroups, and overall.
Costs between different subgroups were compared using t test and P-values. To calculate an overall P-value, an individual P-value on each of the 5 imputed datasets was calculated first. The test statistics were then combined to generate an overall P-value using Rubin's rule, the gold standard method to combine statistical tests using imputed data. 14 The confidence intervals were estimated using a similar approach. Linear regression with backward selection was used to identify significant predictors of the total cost. Imputed and collected costs were combined with demographic and clinical characteristics, and influenza outcomes to explore determinants of influenza cost and the variation across Canadian treatment settings.

| RESULTS
There were 2943 patients enrolled from 27 participating hospitals in 6 provinces over the included influenza seasons. The largest portion of the population was enrolled in Ontario (66.8%; 1966 patients) and Quebec (21.5%; 633 patients) with Eastern provinces (Nova Scotia and New Brunswick) contributing 197 patients (6.7%) and Western provinces (British Columbia, Alberta) contributing 147 patients (5.0%). The demographic and disease characteristics of the SOS Network study cohort show a relatively even male-to-female ratio (47.6%:52.4%) (see Table 1). Most patients had at least 1 comorbidity (90.3%), and nearly half the sample had chronic pulmonary illness (43.1%). Rates of past or current smoking were higher in the Eastern region than among the overall population enrolled (81.4% vs 50.5%). Patients hospitalized in the Western region were younger than the average age of the enrolled population (mean age 61.9 years vs 69.5 years).  Table 2). For the 14.4% of patients with an ICU stay, their average LOS in ICU was 9.8 days (95% CI: 8.6, 11.1). Total length of stay (LOS) and general ward LOS were consistent over the 3 influenza seasons considered. Mean total LOS was 12.6 days (95% CI: 11.  Table S2 Table 3). The higher cost in Western Canada was largely driven by higher rates of ICU admission and longer ICU stays among those requiring ICU admission (see Table   S2). Overall, a higher proportion of patients admitted in Western  Table 2, Table S2).
Univariate analysis found patients who experienced an ICU stay, renal comorbidity, or death had significantly higher hospitalization costs compared to the study average (see Table 3).

| DISCUSSION
The SOS Network dataset represents the most comprehensive national sentinel surveillance dataset available in Canada, providing prospectively collected data on health services utilization using standardized data collection tools across all SOS Network sites over 3 influenza seasons in a large cohort of adults admitted with laboratoryconfirmed influenza. The SOS Network provides the ideal dataset to explore the cost of laboratory-confirmed influenza requiring hospitalizations across the country using a microcosting methodology.
The study reflects costs of laboratory-confirmed influenza requiring hospitalizations and does not rely on non-specific respiratory illness diagnoses obtained from administrative discharge data. By applying a single set of unit price weights to prospectively collected cases, we were able to compare cost of influenza hospitalization between regions in Canada and among several clinical risk groups.
Differences in the cost to treat a case of laboratory-confirmed influenza requiring hospitalization were found across the Canadian provinces, and the source of this variation was explored. Hospitalization costs ranged from $13 711 in Ontario to $20 808 in Western Canada.
Understanding the cost drivers associated with influenza hospitalization can inform policy and decision making regarding publicly funded influenza immunization programs. 5 Table S1).
The length of stay and cost shown for the SOS Network surveillance study cohort were greater than that reported in other Canadian sources of cost for ILI. The OCCI found across ILI diagnoses mean LOS ranging from 3.4 to 9.0 days and costs from $4090 to $13 000 was observed. 4,16 Costs provided through CIHI hospitals reporting across Canada use a case-mix methodology, with weighting for resource intensity, based upon "most responsible diagnosis", which has most contributed to the patient's stay in hospital. 5 Costs in the Manitoba report were based upon the CIHI methodology, which may explain the similar values. The Manitoba case costing method also excludes physician services, ambulatory care, and hospital overheads such as administrative costs and capital costs of facilities. 6 The costs from the Manitoba study include pediatric cases and show a trend toward increasing costs for ILI hospitalization as patients get older, similar to the OCCI. Neither the CIHI nor Manitoba report required that influenza be confirmed by laboratory testing.
The clearest driver of the higher hospitalization costs from the SOS Network data is LOS (see Table S1). While the overall cost and LOS from the SOS Network are larger than the estimates from OCCI, the cost per day calculated from the SOS Network ($1254) is lower than the cost estimated from the OCCI data ($1338) 4 .
Patients in the SOS Network study were hospitalized for an average of 10.8 days, compared with a mean stay of 5.9 days as reported by the OCCI 4 . While the LOS from the current study is longer than the previously published Canadian estimates, the LOS of 10.8 days is consistent with estimates from the United States, where studies have reported LOS of approximately 10 days. 17,18 Length of stay and costs derived by the SOS Network may be higher for several reasons, mostly related to the differences in measurement of diagnosis, age cohorts included, and national vs regional representation of cases. Notably, LOS fluctuates by ILI diagnosis, reflecting differences in disease severity and outcomes depending on the underlying etiology of ILI. Patients in the SOS Network study had laboratory-confirmed influenza and are likely to be a more seriously ill population requiring longer mean LOS than the OCCI study which included patients with ILI without the requirement of laboratory confirmation of influenza. Differences in diagnosis classification may contribute to differences in LOS and cost. The CIHI report groups diagnoses by the case-mix group (CMG+) methodology, under which This study is the first to use resource use data and costs from Canadian adults with laboratory-confirmed influenza collected during active outcomes surveillance. The overall cost per influenza hospitalization calculated was higher than previously published national estimates 5 and provincial estimates from Ontario 4 and Manitoba. 6 The higher cost per hospitalization is largely driven by longer LOS than in previous estimates, rather than a higher cost per day. The cost per hospitalization calculated from this study reflects the cost of laboratory-confirmed influenza cases.
Previous estimates included all hospitalized cases with a clinical diagnosis, without requirement for laboratory confirmation, which likely contributed to overall shorter and less intensive hospitalizations and therefore lower costs. Influenza places a significant burden upon the healthcare system in Canada that has hitherto been underestimated.