Study design
This was a retrospective analysis of pneumonia-related healthcare utilization in the US. The analysis used data extracted from the Truven Health MarketScan® Commercial Claims and Encounters database [19]. The co-primary objectives were (a) to determine the annual and monthly frequency of pneumonia-related healthcare utilization from 2008 to 2014 and (b) to determine pneumonia-associated costs in the two most recent years (2013–2014) due to hospitalization, outpatient visits, and emergency department (ED)/urgent care (UC) visits.
Data source and extraction
The MarketScan database contains information on US individuals insured commercially (i.e. privately) or through the Medicare program [19]. The database collects information on paid claims from health plans, employers, and state-level Medicaid agencies using a nationwide convenience sample. It covers all census regions of the US, includes an average of 48,982,662 individuals per year, and has complete longitudinal records of patient demographics, outpatient services, inpatient services, long-term care, and prescription drug claims. The database is considered representative of the US population with employer-provided health insurance [1, 20] and is used extensively to understand the burden and healthcare utilization for different illnesses in the country. Records within the database are de-identified and fully compliant with US patient confidentiality requirements, including the Health Insurance Portability and Accountability Act of 1996. For this reason, ethical approval was not required for this study.
In this study, data were extracted from January 1, 2008 to December 31, 2014. Only data from individuals in the enrollment tables were included. Pneumonia episodes were identified as a consultation with a principal diagnosis of pneumonia or with a principal diagnosis of meningitis, septicemia, or empyema in addition to a diagnosis of pneumonia in another diagnostic field [21]. For outpatient visits, the principal diagnosis was considered based on International Classification of Diseases, 9th revision, Clinical Modification codes in the primary or the secondary position (see Additional file 1: Table S1). In addition, the first consultation had to be more than 28 days after any previous consultation that had the same diagnosis code. Index episodes were defined as the first episode of pneumonia occurring in the calendar year. Extracted data included the total enrollment numbers for each year; demographic data, including age, sex, geographic region (Northeast, North Central, South, West, or unknown), and insurance type (commercial, Medicare); and the amounts for adjudicated claims paid by health plans, insurers, and patients.
Outcome definitions and measures
Outcome measures included the number of pneumonia cases (overall and by geographical region and insurance type); index visit demographics (mean and median age, age range, and sex distribution); proportions of pneumonia cases for each setting (hospitalization, outpatient visits, and ED/UC visits) for all ages and each age group; annual frequency of pneumonia-related healthcare utilization overall and by setting of the index visit for all ages and each age group; and monthly frequency of pneumonia. For patients transferred to several services within the same day, the setting was defined as the most severe (i.e. hospitalization > ED/UC > outpatient). Pneumonia-associated costs during 2013 and 2014 were determined overall and by setting for the index visit and all follow-up visits that had the same diagnosis code occurring within 28 days. Costs were based on paid amounts of adjudicated claims, including health plan and insurer payments and patient cost-sharing (i.e., copayments, deductibles, and coinsurance). Total costs were estimated as the sum of all costs in each individual setting.
Statistical analysis
The annual frequency of pneumonia is given per 1000 person-years (PY) and was calculated as 1000 × [annual number of patients with a pneumonia episode] ÷ [annual number of total enrolled PY in the MarketScan databases]. Monthly frequencies for each year were calculated per 1000 person-months. Proportions were calculated as 100% × [number of index visits for each setting] ÷ [number of total index visits]. Costs related to the index episode were calculated as the mean payment for each episode during the 2 years from 2013 to 2014. The 95% confidence intervals (CIs) for frequencies of pneumonia and costs were calculated using a normal approximation.
The significance in the difference in frequencies of pneumonia between 2008 and 2009 and 2014 was assessed by a log-linked Poisson regression with the log of the number of PYs as an offset. The exposure variable was the number of patients with pneumonia for a given year and age group. Year, age group, and interaction between year and age group were included as predictors in the model. A p-value < 0.05 was considered to indicate a significant difference.
Analyses were performed using SAS® Enterprise Guide 7.1 (SAS Institute, Cary, NC, USA).