Abstract
Objective. With the expected rise in the arthritis population, information is required regarding trends of healthcare expenditures among individuals with arthritis in the United States. We examined temporal trends in direct and out-of-pocket (OOP) healthcare expenditures among individuals with arthritis using a nationally representative database, the Medical Expenditures Panel Survey.
Methods. The study population was composed of cross-sectional cohorts of individuals aged ≥ 18 years from 2008 to 2014. Two-part models were used to estimate the incremental total and types of annual direct and OOP healthcare expenditures (adjusted to 2014 US dollars) for arthritis, after controlling for predisposing, enabling, need, personal health practice, and external environmental factors, as per the Anderson Healthcare Behavioral Model.
Results. An annual weighted arthritis population rose from 56.1 million in 2008 to 65.1 million in 2014. Among individuals with arthritis, the annual average direct and OOP expenditure was $10,424 [standard error (SE) = $345, aggregate = $584.8 billion] and $1493 (SE = $50, aggregate = $83.8 billion) in 2008, respectively, and $910 (SE = $279, total = $645.1 billion) and $1099 (SE = $36, aggregate = $71.5 billion) in 2014, respectively. In the fully adjusted model, individuals with arthritis had significantly greater total and OOP expenditures from 2008 to 2014; however, the magnitude of incremental OOP expenditure declined from 2008 to 2014.
Conclusion. Although the annual direct healthcare expenditures per person remained stable over the years, the rise in proportion of the arthritis population led to a huge increase in aggregate economic burden to the US healthcare system.
Arthritis, the leading cause of disability among adults in the United States1, affects about 52.5 million US individuals2,3. The significant disease burden of arthritis could translate into a huge economic burden4,5. Although previous studies have determined healthcare expenditures among individuals with arthritis in the United States3–11, estimates of healthcare expenditures from these studies were obsolete3–11, limited to a particular type of arthritis6,7,8, restricted to direct with limited information on out-of-pocket (OOP) expenditures7,8, or included people having arthritis combined with joint pain or other rheumatic conditions10,11. Except for a study by Cisternas, et al4, none of the studies provided information on changes in total expenditure and types of healthcare expenditures across years among individuals with arthritis. Cisternas, et al estimated direct healthcare expenditures of $252 billion (2005 US$) in 1997 among individuals with arthritis, and the figure rose to $353 billion in 20054. Hootman, et al projected an increase in the number of individuals with arthritis to 78 million by 2040 in the United States3; therefore, it is important to determine trends in healthcare expenditures among these individuals to guide health policy makers in allocating budgets and designing interventions to efficiently manage a highly prevalent and priority health condition.
To our knowledge, there has been no updated information in the last 10 years about current trends in healthcare expenditures among individuals with arthritis from a nationally representative database. Understanding trends in healthcare expenditures among these individuals will assist in developing strategies to curb excessive healthcare expenditures. Therefore, the primary objective of our study was to assess trends in direct and OOP annual healthcare expenditures among individuals with and without arthritis using nationally representative data of the US population from 2008 to 2014.
MATERIALS AND METHODS
Study design and data source
We used a retrospective cross-sectional trend design to estimate healthcare expenditures among individuals with and without arthritis using the Medical Expenditure Panel Survey (MEPS) data. MEPS provides nationally representative estimates of healthcare use, expenditures, prescription medications, source of payment, health status, and health insurance coverage from the US noninstitutionalized civilian population12 (Supplementary Data 1, available with the online version of this article). We used 2 data files: a full-year household consolidated (HC) data file, and a medical condition data file for each year from 2008 to 2014. The HC data file has information on healthcare expenditures, diagnoses of priority conditions, and demographic, socioeconomic, health, and personal healthcare practice. The medical condition data file contains information on 3-digit codes from the International Classification of Diseases-9-Clinical Modification (ICD-9-CM), as well as clinical classification codes. Survey respondents reported the presence of conditions. Once clinical conditions were reported in exactly the same words as were used originally, professional coders transferred the responses to 5-digit ICD-9 codes, but released only 3-digit ICD-9 codes, to preserve participant confidentiality12,13,14. Survey response for annual files ranged from 60% in 2008 to 50% in 201413. Because our study was conducted using a publicly available unrestricted and deidentified survey database, it was considered exempt by the University of Rhode Island Institutional Review Board.
Study cohort
The study cohort was composed of noninstitutionalized US individuals age ≥ 18 years with and without arthritis alive during each calendar year from 2008 to 2014.
Dependent variables
Total direct healthcare expenditures per person per year was the dependent variable of interest. These expenditures referred to payments for medical events, including inpatient visits, emergency room (ER) visits, outpatient services including office-based physician visits, home healthcare visits, prescription drugs, and other associated costs. Payment referred to direct payment from individuals (OOP), private insurance, Medicaid, Medicare, and other sources for each calendar year. In addition, expenditures were also classified according to cost types, such as inpatient, outpatient, prescription, ER, and “other” categories. “Other” expenditures included home health, vision, durable medical equipment, ambulance services, orthopedic items, hearing devices, prosthesis, bathroom aids, medical equipment, disposable supplies, and other miscellaneous items or services rented or purchased during the calendar year. Total OOP expenditures comprised self-reported payments for coinsurance, deductibles, cash outlays for services, supplies, and other items not covered by health insurance. All direct and OOP expenditures were inflated to 2014 constant dollars using the annual consumer price index for medical care obtained from the Bureau of Labor Statistics15. Additionally, based on research on inflation indices by Dunn, et al16, we presented aggregate total direct and OOP expenditures using Personal Care Expenditure-Health (PCEH) index with a sensitivity analysis.
Arthritis as key independent variable
Self-reported arthritis was the key independent variable. The Agency for Healthcare Research and Quality has specified a group of medical conditions, including arthritis, as priority conditions. Therefore, MEPS collects detailed information on the list of priority conditions14. We defined presence of arthritis if a survey participant responded with “Yes” to the following survey question: “(Have/Has) (person) ever been told by a doctor or other health professional that (person) had arthritis?” The self-reported measure of arthritis from the MEPS survey data is a commonly used measure of treated prevalence of the condition17. Based on this definition, an annually weighted population of 56.1 million (weighted % = 24.8%, unweighted N = 4967) reported presence of arthritis in 2008, which rose to 65.1 million (weighted % = 27.0%, unweighted N = 5779) in 2014 (Table 1A).
Other independent variables
The Anderson Behavioral Model of Health Services Utilization was used to identify the predictors of direct healthcare expenditures among individuals with and without arthritis. According to this model, predisposing, enabling, need-related, personal health practice, and external environmental healthcare factors influence health services use18 (Supplementary Data 2, available with the online version of this article). Predisposing factors consisted of age, sex, and race/ethnicity. Enabling resources were marital status, income, health insurance, and employment status. Need-related factors consisted of perceived physical and mental health status, and co-occurring chronic conditions categorized into total number of physical and mental health conditions. Self-reported perceived mental and physical health status were identified during multiple rounds in a given calendar year. In that case, we did a hierarchical approach by identifying values of responses in the last round during a particular year, followed by responses for the previous rounds in cases of missing values in the last rounds. We determined the list of conditions based on the guidance document from the Office of the Assistant Secretary of Health19, as well as common chronic conditions identified among individuals with arthritis using all clinical classification codes. Applying these criteria, we included all MEPS priority conditions (high blood pressure, heart disease, stroke, emphysema or chronic bronchitis, high cholesterol, cancer, diabetes, and asthma), gastrointestinal disorders, endocrinology or metabolic disorders, kidney diseases, and eye disorders. In addition, we defined the number of mental health conditions by summing up the presence of any of anxiety, depression, schizophrenia, substance abuse, or dementia. We also identified the functional limitations if an individual had restricted ability to work because of experienced pain or a disability, or had self-reported joint pain20. In addition, we identified common acute conditions, such as infections or acute injury. Personal health practice–related factors consisted of body mass index, current smoking status, and exercise. External environmental factors consisted of census-defined regions (Supplementary Table 1, available with the online version of this article).
Statistical analysis
Separate chi-square statistics were used to examine differences in all the independent characteristics according to the presence of arthritis for each year. We also reported difference in prevalence of all independent characteristics from 2008 to 2014, along with the respective 95% CI and p values to reflect the changes in the composition of individuals with and without arthritis from 2008 to 2014. To account for individuals with zero expenditures, a 2-part model known as mixed discrete-continuous variable regression was used to estimate incremental healthcare expenditures among individuals with arthritis compared to those without21. In the first part, a probit model was used to determine the probability of zero expenditure versus positive healthcare expenditures. In the second part, a generalized linear model with log link and gamma distribution was used to determine conditional expenditures for those with positive healthcare expenditures22. Postregression predicted estimates (i.e., margins), and provided unadjusted and adjusted incremental healthcare expenditures among individuals with arthritis compared to those without arthritis from 2008 to 2014. We reported estimates for 4 adjusted models with the following covariates: (1) unadjusted as no covariate except presence of arthritis; (2) Model 1 plus predisposing factors; (3) Model 2 plus enabling factors; and (4) Model 3 plus need-related, personal healthcare practice, and external environmental factors. Delta method was used to estimate 95% CI. We confirmed absence of multicollinearity among covariates using variance inflation factor, tolerance, and eigenvalues. We considered presence of significant collinearity with variance inflation factor ≥ 5, tolerance < 0.20, or relatively high eigenvalues with condition index > 30. All the analyses were performed using SAS 9.4 and Stata 14, and accounted for complex survey design.
RESULTS
Description of study cohort
Table 1B describes distributions of predisposing, enabling, need, lifestyle, and external environmental factors according to the presence of arthritis. There were significant differences in all characteristics between individuals with and without arthritis across all years, expressed in range. Across 2008 to 2014, individuals with arthritis were mostly middle aged (45.0%–46.1%) or elderly (39.3%–42.5%), female (59.5%–61.8%), white (74.4%–78.1%), married (54.5%–57.3%), with high income (36.1%–40.6%), and with excellent or very good perceived physical health status (37.7%–39.7%) and mental health status (52.9%–54.7%), with limitation in doing routine physical activity (29.1%–34.1%) and pain limiting ability to work (41.5%–45.3%). They were also past smokers/nonsmokers (79.8–83.2%), with limited physical activity (50.3%–59.9%), obese (38.0%–41.4%), and they resided in the South (37.0%–38.7%). The proportion of individuals with ≥ 2 physical chronic conditions ranged from 42.6% to 44.7%, and those with ≥ 1 mental health condition ranged from 26.5% to 32.4%. Further, the most common chronic physical health conditions were hypertension (58.0%–60.6), hyperlipidemia (53.1%–55.6%), heart disease (28.4%–31.0%), cancer (20.0%–21.8%), and diabetes (17.2–20.2%), while depression (16.8%–20.5%) and anxiety (12.7%–18.3%) were the most common mental health conditions among individuals with arthritis (Supplementary Table 2, available with the online version of this article).
Regarding changes between 2008 and 2014 in the composition of characteristics among individuals with arthritis, we observed an increase in proportion of individuals who were Hispanic (6.9% vs 9.0%, p < 0.001), were poor (11.7% vs 13.7%, p < 0.05), had activity limitations (29.1% vs 32.3%, p < 0.01), and were nonsmokers or past smokers (79.8% vs 83.2%, p < 0.001). On the other hand, there was a decline in proportion of individuals with arthritis who were white (78.0% vs 74.4%, p < 0.001), widowed (15.7% vs 13.4%, p < 0.01), had middle income (28.4% vs 25.9%, p < 0.05), and had pain limiting ability to work (44.1% vs 41.5%, p < 0.01). Similar trends were observed for individuals without arthritis. Further, from 2008 to 2014, we noted an increase of 6 percentage points in the proportion with ≥ 1 mental health condition among those with arthritis (26.5% vs 33.0%, p < 0.01), and an increase of 2 percentage points among those without arthritis (13.3% vs 15.8%, p < 0.01).
Average annual healthcare expenditures for arthritis versus non-arthritis
Among individuals with arthritis, average annual total healthcare expenditures declined from $10,424 in 2008 to $9910 in 2014 (Table 2). The top 3 ranked expenditure categories were outpatient (32.6%), inpatient (29.0%), and prescription drug costs (24.7%) in 2008, and changed to outpatient (33.8%), prescription drug (26.8%), and inpatient costs (26.4%) in 2014 (Figure 1). Average annual total OOP was $1493, representing 14% of total healthcare expenditures in 2008, which declined to $1099 in 2014 (Table 2 and Supplementary Table 3, available with the online version of this article). Similar trends were observed among individuals without arthritis.
To provide insight through 3 key predisposing factors, we estimated total direct and OOP healthcare expenditures by age, sex, and race/ethnicity among individuals with arthritis (Figure 2). There was a slight decline in total direct healthcare expenditures among elderly individuals (≥ 65 yrs) with arthritis from $11,200 in 2008 to $10,483 in 2014, while the expenditures among non-elderly (18–64 yrs) individuals with arthritis remained stable. Men and women had relatively similar total direct and OOP expenditures. Whites had consistently higher total direct and OOP healthcare expenditures compared to African Americans or other races.
Average incremental healthcare expenditures: arthritis versus non-arthritis
Without adjustment of covariates, individuals with arthritis had $6974 and $6318 higher healthcare expenditures compared to those without arthritis in 2008 and 2014 respectively (Table 3). Magnitude of incremental total healthcare expenditures among individuals with arthritis declined to $4116 (about 59% of incremental expenditures from model 1) after adjusting for predisposing factors in model 2, $4123 (59% of incremental expenditures from model 1) after adjusting for predisposing and enabling factors in model 3, and $2000 (29% of incremental expenditures from model 1) after adjusting predisposing, enabling, need, lifestyles, and external environment in model 4 for the year 2008. Almost 71% of the difference in total healthcare expenditures between individuals with and without arthritis was explained by the differences (expressed as %) in predisposing factors (41%), enabling factors (about 0%), need, lifestyle, and external environmental factors (30%). Those factors again explained 72%, 77%, 70%, 62%, 67%, and 74%, respectively for each year from 2009 to 2014, in incremental difference in healthcare expenditures between individuals with and without arthritis. With respect to individual covariates, differences in age, perceived health status, and number of physical and mental health conditions explained a majority of incremental total healthcare expenditures between individuals with and without arthritis (data not shown).
Regarding expenditures by types of specific services, individuals with arthritis had significantly higher inpatient, outpatient, emergency, prescription drug, and OOP expenditures compared to those without arthritis across years in adjusted analyses. After adjusting for need-related and personal health practice factors in model 4, incremental expenditures because of inpatient services were significantly greater for 2008 and 2012, and were greater for outpatient services and prescription drugs for all the years among individuals with arthritis compared to those without. With respect to magnitude of incremental expenditures over the years, individuals with arthritis had greater total incremental expenditures for 2012 and 2013 ($2654 and $2284), which corresponded to greater outpatient expenditures for those years ($935 and $980, respectively) in fully adjusted model 4.
Estimated direct aggregate expenditures for individuals with arthritis
Aggregate burden estimates were calculated by multiplying a weighted population of arthritis with per patient estimate of healthcare expenditures. For individuals with arthritis, aggregate (sum of) annual direct expenditures increased from $584.8 billion in 2008 to $645.1 billion in 2014, whereas OOP declined from $83.8 billion in 2008 to $71.5 billion in 2014 for individuals with arthritis (Table 4). Sensitivity analyses using PCEH-adjusted expenditures is shown in Supplementary Table 4A and 4B (available with the online version of this article). After adjustment for all independent characteristics, aggregate annual direct expenditures rose from $112.2 billion in 2008 to $160.3 billion in 2012, then declined to $107.7 billion in 2014. However, aggregate OOP expenditures increased from $7.4 billion to $12.3 billion until 2012, then declined to $10.9 billion in 2014 (Table 4). The top 2 expenditure categories were outpatient and inpatient in 2008 and outpatient and prescription drugs in 2014.
DISCUSSION
Our study aimed to provide insights on the trends in direct and OOP healthcare expenditures among individuals with arthritis using nationally representative data for 2008 to 2014. About 1 in 5 adults had arthritis, consistent with the estimates from the US Centers for Disease Control and Prevention2. From 2008 to 2014, there was an increase in the proportion of Hispanics, poor, obese, and individuals with activity limitations and mental health conditions among those with arthritis as well as among those without arthritis. These findings are indicative of a change in the composition of the US population with arthritis. Further, hypertension, hyperlipidemia, and heart disease were the most common chronic conditions among individuals with arthritis and remained relatively similar across 2008 to 2014. In multivariable models, we found that incremental healthcare expenditures among individuals with arthritis were mainly driven by difference in age, health status, and chronic conditions between individuals with and without arthritis. Our findings suggest the needs of managing multiple chronic physical and mental health conditions among individuals with arthritis.
From 2008 to 2014, we noted a trend toward nonsignificant increase in unadjusted incremental total healthcare expenditures among individuals with arthritis in 2011, 2012, and 2013, and a decline in 2014. Such trends remained with the adjusted incremental total healthcare expenditures. Similarly, a previous study by Cisternas, et al reported a nonsignificant increase in total healthcare expenditures (2014 US dollars) among individuals with arthritis from $9223 in 1997 to $10,578 in 20054. Further, for expenditures by types of services, the order of top 3 categories was outpatient, inpatient, and prescription drugs in 2008, which is similar to that reported by Cisternas, et al4. In 2005, however, we noted a shift in top 3 categories in 2014 to outpatient, prescription drugs, and inpatient services. There could be many plausible explanations for these trends.
First, it is highly likely that the outpatient procedures such as orthopedic surgeries are becoming a common management option for arthritis. In fact, literature on outpatient surgical procedures suggests an increase in uptake of outpatient orthopedic surgeries such as knee or hip arthroplasty in the US from 1995 to 200523. Further, outpatient medication services also include the administration of injectable medications such as biologics. Because biologics are administered as a part of outpatient medication services, the use of outpatient services may have been sustained over a period of time as a management approach among individuals with arthritis.
Second, prescription drug expenditures were also sustained and grew marginally over 2008 to 2014. After the Medicare Part D era, affordability to prescription drugs and off-patent prescription drugs (generics) could have led to improved access, reflecting a relatively stable prescription drug expenditure throughout 2008 to 2014. For example, a study by Polinksi, et al found an increase in biologics use among Medicare beneficiaries with rheumatoid arthritis after the Part D Medicare coverage, and a decline in OOP expenditures with cost sharing24. Our study also noted a decline in unadjusted OOP expenditures from 2008 to 2014, mainly driven through decline in prescription drug OOP (Supplementary Table 3, available with the online version of this article).
Third, we found relatively stable unadjusted and adjusted inpatient expenditures from 2008 to 2013, with a decline specifically in 2014. Our findings are consistent with studies assessing the effect of the Affordable Care Act, such as introduction of the Hospital Readmission Reduction Program. Starting in October 2012, hospitals with excessive 30-day readmissions and hospital-acquired infection-related hospitalizations were penalized. Studies have shown a decline in 30-day readmission rate from 19.0% to 18.0% among Medicare beneficiaries, and 17% decline in hospital-acquired conditions from 2010 to 201325. In our study, we also noted a decline in inpatient hospitalization from 15% in 2008 to 13% in 2014 (Supplementary Figure 1, available with the online version of this article), and observed a decline in the magnitude of incremental healthcare expenditures.
Last, we noted increasing trends in aggregate expenditures for the population of arthritis from $584.8 billion in 2008 to $645 billion in 2014. Total aggregate healthcare expenditures for individuals with arthritis accounted for about 4.0% of the US gross domestic product in 2008 and 2014. The increase in aggregate unadjusted total direct healthcare expenditures was mainly driven through an increase in weighted population of individuals with arthritis. On a positive side, we noted a slowdown in the intensity of adjusted incremental expenditures and OOP specifically from 2013, which led to a huge decline in the aggregate direct healthcare expenditures in 2014.
Our study provides substantially crucial and up-to-date information about total and OOP expenditures among individuals with arthritis in the US using the recently available nationally representative MEPS data. A robust and widely used method of addressing the issue of zero expenditures and skewed expenditure data to provide robust estimates of healthcare expenditures was used in the analyses. However, there are a few limitations worth noting. Because the MEPS data are based on self-reported data, treated prevalence of arthritis and other chronic conditions may affect estimates; however, prevalence of arthritis using self-reported data from MEPS has been consistent with the estimates obtained from the other survey data17. Our study estimates have not included indirect expenditures. Arthritis is also prevalent among individuals residing in nursing homes, but because the MEPS covers non-institutionalized individuals, our findings cannot be generalizable to institutionalized individuals with arthritis. Disease progression and severity, and types of medications under outpatient services, cannot be collected using the MEPS data.
Our study quantifies incremental healthcare expenditures among individuals with arthritis compared to those without arthritis in the US and provides crucial information to providers, payers, and policy makers on how the dollar amount is being spent in treating patients with arthritis.
ONLINE SUPPLEMENT
Supplementary material accompanies the online version of this article.
Acknowledgment
Partial results from our study were submitted for poster presentation at the 22nd International Society for Pharmacoeconomics and Outcomes Research Annual International Meeting, May 2017.
- Accepted for publication October 24, 2017.