Prevalent Statin Use in Long-Stay Nursing Home Residents With Life-Limiting Illness

Deborah S. Mack, MPH; Jennifer Tjia, MD, MSCE; Anne L. Hume, PharmD; Kate L. Lapane, PhD, MS


J Am Geriatr Soc. 2020;68(4):708-716. 

In This Article



This study used a cross-sectional design to evaluate statin use in US nursing homes (Figure S1). Prevalent statin use was estimated with prescription claims if date of fill plus supply overlapped with September 30, 2016. The Minimum Data Set (MDS) 3.0, a federally mandated nursing home assessment instrument, is updated October of each year. The specific date in September was selected to evaluate the most recent data before the annual MDS 3.0 changes were implemented. This study was approved by the University of Massachusetts Medical School internal review board.

Data Sources

Nursing home assessment data from the MDS 3.0 were linked to Medicare administrative claims Part A (inpatient), Medicare Part D (prescription drugs) files, and the Medicare Beneficiary Summary File for Medicare eligibility. The MDS 3.0 is a database composed of federally mandated assessments that collect sociodemographic and clinical information on nursing home residents residing in US Medicare- and Medicaid-certified facilities.[21,22] The assessment is administered in full at admission and annually, updated quarterly and with a change in clinical status.

Definition of Life-limiting Illness

Life-limiting illness was used to identify clinically complex long-stay nursing home residents near the end of life. These residents may face less favorable outcomes from use of statins in light of their decompensated state and limited life expectancy.[20] Life-limiting illness was used in other studies to describe patients near the end of life.[17,19] To define it formally, we use the term "life-limiting illness" to describe a "progressive life-limiting disease with a prognosis of months or less."[23] Active diagnoses (as of 7 days before assessment date) from the MDS 3.0 and International Classification of Diseases (ICD)-10 codes from Medicare Part A claims between March 31, 2016, and September 30, 2016, were used to define life-limiting illness operationally. Because life-limiting illness has not been defined operationally and validated in a nursing home population, we defined the condition in these four ways (Table S1):

  1. Life expectancy less than 6 months based on the definition of terminal illness in the US federal code, 42 USC Sec. 1395x[24] identified with the MDS variable J1400.

  2. Receipt of palliative care consultation based on ICD-10 diagnosis code Z51.5. Palliative care or comfort treatment measures are most often offered to those near the end of life. This hospital diagnosis code was shown to be valid,[25] with almost one-half of the population that received this service dying before discharge.

  3. Diagnosed with a Veterans Health Administration Palliative Care Index (PCI)[17,26,27] condition, indicated by ICD-10 codes for any of the following diagnoses: esophageal cancer, liver cancer, pancreatic cancer, broncho-tracheal-lung cancer, colon cancer, leukemia, Hodgkin's lymphoma, multiple myeloma, metastatic breast cancer, metastatic prostate cancer, acquired immunodeficiency syndrome (AIDS), heart failure, or chronic obstructive pulmonary disease (COPD). For heart failure and COPD, residents either had more than two hospitalizations, one hospitalization with an intensive care unit stay, or an active MDS diagnosis within the 6-month window.

  4. Diagnosed with a "serious illness" as recently defined by Kelley et al[28] indicated by the presence of any diagnosis in their validated list of ICD-10 codes for the following conditions: cancer, chronic kidney disease (stage 5)/end-stage renal disease, advanced liver disease/cirrhosis, diabetes with severe complications (ischemic heart disease, peripheral vascular disease, renal disease), AIDS, hip fracture, dementia, COPD or interstitial lung disease, heart failure, neurodegenerative diseases, and stroke.

Study Sample

Our population included long-stay nursing home residents near the end of life who resided in a Medicare- and Medicaid-certified nursing home facility on September 30, 2016 (target date) and met the following inclusion criteria (Figure 1): (1) current nursing home resident (not deceased or discharged) as of the target date; (2) long-stay resident (length of stay ≥90 d), indicated by one or more quarterly assessment between April 1 and September 30, 2016; (3) age 65 years or older; (4) continuous coverage of Medicare fee-for-service Parts A, B, and D in the 3 months preceding the target date (July, August, and September 2016); and (5) indication of limited life expectancy (<6 mo) or diagnosed with a life-limiting illness (Table S1). Residents on hospice were excluded from the sample due to their medication payment mechanism (medications become bundled into per diem rates) that excludes use of Medicare Part D claims (approximately 5% of residents).

Figure 1.

Flowchart of sample selection criteria relative to target date, September 30, 2016.

Prevalent Statin use

Prevalent oral statin use was determined if a prescription of any individual or combination of generic statins (date filled plus number of days' supply) overlapped with September 30, 2016. For descriptive purposes, we classified oral statin medications by generic type: atorvastatin, fluvastatin, lovastatin, pitavastatin, pravastatin, rosuvastatin, and simvastatin. The primary outcome of interest was statin use (yes/no). Guided by the American College of Cardiology/American Heart Association (ACC/AHA) guidelines,[29] we defined high-intensity statin therapy based on daily dose low-density lipoprotein-cholesterol (LDL-C) reduction more than 50%. This included atorvastatin 40 or 80 mg, rosuvastatin 20 or 40 mg, and simvastatin 80 mg. Of note, the US Food and Drug Administration (FDA) currently does not recommend simvastatin 80 mg due to an increased risk of myopathy.[29,30] Low-moderate-intensity statin use (daily dose LDL-C reduction <30%-50%) was defined as atorvastatin 10 or 20 mg; fluvastatin 20, 40, or 40 mg twice/day; fluvastatin XL 80 mg; lovastatin 10, 20, 40, or 80 mg; pitavastatin 1, 2, or 4 mg; pravastatin 10, 20, 40, or 80 mg; rosuvastatin 5 or 10 mg; and simvastatin 5, 10, 20, or 40 mg.

Resident Variables

Using the MDS 3.0 and Medicare Part A administrative claims (ICD-10 diagnosis codes), we included these variables: age, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic/Latino, other), and marital status (married vs not married). Age (65–75, >75 y) was classified based on common guideline stratification[31] including that of the ACC/AHA[29] and USPSTF.[14] Clinical information included cognitive impairment (none, mild, moderate, severe)[32] and activities of daily living (ADL) dependency status (independent to limited assistance required, extensive assistance required, dependent to total dependence).[33] The four operational definitions of life-limiting illness described earlier were also included as binary variables. Atherosclerotic cardiovascular disease (ASCVD) and its risk factors were also evaluated because these conditions are FDA indications for statins. ASCVD risk factors were defined as diabetes, hyperlipidemia, or hypertension. The variable for ASCVD risk was operationally defined as (1) an ASCVD diagnosis (coronary artery disease, stroke, or peripheral vascular disease/peripheral artery disease); (2) no ASCVD disease with one or more ASCVD risk factor; or (3) no ASCVD disease and no risk factors.[15,34]

Analytic Approach

All analyses were stratified by age group. Descriptive statistics were estimated for resident characteristics. The overall statin use was estimated as a point prevalence as of September 30, 2016, by any use, low-moderate, and high-intensity use. We addressed multicollinearity by examining variance inflation factors derived from a regression model using the binary variable for statin use as the outcome and the covariates of interest as the independent variables. Poisson models with facility clustering, an exchangeable correlation matrix, and robust confidence intervals (CIs) provided estimates of crude and adjusted prevalence ratios and 95% CIs of statin use by resident variables.