The relative burden of community-acquired pneumonia hospitalizations in older adults: a retrospective observational study in the United States

This retrospective cohort study utilized administrative healthcare claims data from Optum’s Clinformatics™ Data Mart (CDM) between 2014 and 2015 in a study population comprised of Medicare Advantage with Prescription Drug Plan (MAPD) beneficiaries from a large national managed care company affiliated with Optum. MAPD plans are alternative plans to “traditional” Medicare insurance coverage, often call “Part C,” where federal payments are provided to private insurance companies. These organizations provide an insurance benefit which includes inpatient, outpatient, prescription coverage, and often additional dental, vision, or other benefits usually within Health Maintenance Organization (HMO) or Preferred Provider Organization (PPO) structures. Enrollment into MAPD is optional and roughly one-third of Medicare beneficiaries enroll in MAPD coverage [11]. Compared to traditional Medicare enrollees, MAPD members tend to be older, more racially diverse, and have lower socioeconomic status [12].

The Optum Clinformatics™ Data Mart is statistically de-identified under the Expert Determination method consistent with HIPAA and managed according to Optum customer data use agreements. These administrative claims submitted for payment by providers and pharmacies are verified, adjudicated, adjusted, and de-identified prior to inclusion. The population is geographically diverse, spanning all 50 states in the U.S.. In addition to medical claims and pharmacy claims, the data include tables with member eligibility and demographic information. The data also include standard pricing for all medical claims, pharmacy claims, and inpatient confinements. The data use an encrypted member identification number to longitudinally track members across years. Information on the month and year of death are appended to the database by Optum prior to de-identification using the Social Security Administration Death Master File.

Adults aged 65 to 89 years with continuous MAPD enrollment from January 1, 2014 through December 31, 2015 were included in this study. Censoring due to death in 2015 was permitted since it was an outcome of interest. Hospitalization for CAP, MI, stroke, and OF were identified during the 2015 calendar year. The 2014 calendar year was used to assess cohort clinical characteristics as well as to capture preventive measures for each of the four conditions. Subjects who had a hospitalization for CAP, MI, stroke or OF or a cumulative stay in a skilled nursing facility for ≥90 days during 2014 were excluded to create a cohort that was community-dwelling and free of the hospitalizations of interest at the start of follow-up in 2015.

Age, gender, region, and low income subsidy status were assessed as of January 1, 2015 from the enrollment data. Individuals were categorized into age groups: 65–69, 70–74, 75–79, 80–84, and 85–89 years; geographic region was based on Census regions (North Central, Northeast, South, West, Unknown). Key comorbidities were identified using ICD-9 and corresponding ICD-10 codes. Deyo-Charlson comorbidities were calculated as a weighted summary score for each individual [13, 14]. Coding algorithms of the comorbidity classifications are summarized in Appendix 1 Table 4.

CAP, MI, stroke, and OF hospitalizations were identified beginning on January 1, 2015 in the medical claims using diagnosis codes listed in Appendix 1 Table 4. Only hospital admissions with a primary diagnosis of one of these four conditions were included. Hospitalizations for CAP were identified based on prior work [1], with slight modification to use a primary diagnosis for pneumonia or a primary diagnosis of septicemia or respiratory failure with a diagnosis of pneumonia during the same hospital visit. To distinguish CAP from potential hospital-acquired pneumonia, index CAP hospitalizations with claims indicating other hospitalization, skilled nursing facility, or hospice care in the 14 days preceding a pneumonia hospitalization were excluded.

A 30-day post-discharge period was used to assess hospital readmission for any condition, similar to “all cause readmission” metrics already in use as measures of care quality after a hospitalization [15]. After a 30-day period had elapsed with no readmissions, new hospitalizations for any study condition were counted. All-cause deaths after the index hospitalization were also recorded. However, due to data privacy requirements, dates were restricted at the month level. Thus, we recorded deaths that occurred in the same month or the following month of index admissions of interest.

Each hospitalization and readmission included the length of stay (LOS) and the sum of the amounts paid by the individual and the health plan. Total episode costs included the costs of the index hospitalization and any readmissions observed during the 30-day window. Costs for preventive measures were summarized as costs to the health insurer, member out-of-pocket (OOP) cost, total costs (health insurer plus member OOP), and summarized as per member per year (PMPY) expenditures based on the total population.

Preventive measures for CAP included influenza and pneumococcal vaccinations, consistent with Infectious Diseases Society of America and the American Thoracic Society Consensus Guidelines on the Management of Community-Acquired Pneumonia [16]. Preventive medications for MI and stroke comprised of antihypertensives, statins, aspirin, anticoagulants, and antiplatelet medications. OF preventive measures included bisphosphonates, selective estrogen receptor modulators, and injectable therapies (denosumab and teriparatide). All medications were identified by Medi-Span Generic Product Identifier (GPI) codes in the pharmacy claims. Injectable medications and vaccinations were also identified by Current Procedural Terminology codes and Healthcare Common Procedure Coding System codes in the medical claims. Codes for preventive medications are available in Appendix 2 Table 5.

Total summed costs for hospitalizations, readmissions, and total combined costs were described as well as per event costs. Admission, re-admission, and mortality rates were standardized to events per 100,000 life-years to allow direct comparison between conditions within the cohort. Data management and analysis were conducted using SAS Enterprise Guide 5.1 (SAS Institute, Cary, NC). This manuscript was drafted in accordance with STROBE reporting guidelines (http://​www.​strobe-statement.​org/​) for observational studies.

Limitations common to studies using administrative claims data apply to this study [44‐46]. These include a lack of certain information in the database (e.g. lab results, smoking status, weight, and health behavior information) and errors or omissions in claims coding. Data were from a single national MAPD health plan with a diverse population of older adults residing in broad geographic regions. Because this study uses data from a single insurer’s members only, the results may not be generalized to the general population. This study limited the cohort to members who did not have one or more hospitalizations in the 2014 pre-index period to focus on primary prevention costs rather than in a population with previous disease. This underestimates the burden of all of the disease states from a health plan perspective but does not bias the comparison between the disease states. The current study underestimated the burden of CAP since only inpatient CAP episodes, which have been shown to include only 40% of cases in the older population, were considered [1, 8]. Additionally, the study only considered primary hospital diagnosis. In cases of pneumonia, it is possible that hospitals code a different cause as the primary diagnosis if an underlying condition is exacerbated that requires serious management – thus further underestimating our estimated burden of CAP episodes. Medications or vaccinations obtained without utilizing an insurance benefit were unobserved in this study. This may include over-the-counter medications (e.g. aspirin to treat MI or stroke), cash payments for medications or vaccines, medications filled using low-cost generic programs [47], and vaccinations received through employers or other health fairs. This study also grouped together medication classes commonly used for primary prevention for MI and stroke. However, indications for each medication cannot necessarily be confirmed and may have been used to treat other conditions (e.g. anticoagulants prescribed for venous thromboembolism). Lastly, the data may have underestimated death as this was based on the Social Security Administration Death Master File for which some states ceased providing data in 2011. Also, discharge status of “expired” is not available given discharge date would represent the date of death which may jeopardize data de-identification.