Abstract
Background Out-of-pocket (OOP) expenditures are incurred as insurers and employers shift some of the burden of health care costs onto consumers. As cost-sharing increases, OOP expenditures could be a barrier to physical therapy care.
Objective The purposes of this study were: (1) to identify factors associated with any OOP physical therapy spending and (2) to identify factors associated with higher spending among individuals incurring OOP costs.
Design The study was a retrospective analysis using the 4 most recently available panels of data from the Medical Expenditure Panel Survey (MEPS) encompassing 2008–2012.
Methods A data file containing episodes of physical therapy care for 2,189 people was created. Logistic regression was used to identify factors related to having an OOP expenditure. A multivariable generalized linear model was used to identify factors related to mean OOP expenditures.
Results On average, an episode of care encompassed 9.9 visits, with mean total expenditures of $1,708 (median: $792). Fifty-four percent of episodes of care had an OOP expenditure. For individuals with OOP expenditures, the mean OOP expenditure for an episode of care was $351 (median: $144). Being female or non-Hispanic and having a higher income were associated with higher odds of incurring an OOP expenditure, whereas being in worse general health, >65 years of age, or nonwhite and having public funding were associated with lower odds of incurring an OOP expenditure. Amounts of OOP spending were higher in urban areas and in all census geographic regions relative to the Northeast region.
Limitations Estimates are based on household-reported survey data, limited to ambulatory care, and do not include institutionalized individuals.
Conclusions At 54%, the proportion of individuals with OOP expenditures for physical therapy is lower than for general medical care. Several predictors were found of having OOP expenditures and of the magnitude of those expenditures.
Controlling health care expenditures is a national priority and a focus of substantial legislative and regulatory activity.1 Although the rate of growth in health care expenditures in the United States has slowed, out-of-pocket (OOP) expenditures are accelerating.2 Cost-sharing mechanisms such as copayments and deductibles shift some of the burden for health care expenditures onto consumers.3 These mechanisms are quite common, with only 12% of US health care seekers paying nothing out of pocket toward their health care expenditures.4 In 1996, the average OOP expenditure for a US family was $459; that figure rose to $795 by 2009, an increase of 73%.5 Many American families have difficulty paying medical bills and, as a result, are cutting back on their use of health care.5,6 Increasing the OOP burden on consumers creates a possible barrier to care and a greater financial burden for those with less economic means that could result in poorer health status.7 This barrier may be amplified in a service such as physical therapy that involves episodic care with multiple visits in an episode.
Demographic variations in OOP expenditures have been seen in preventive health services such as mammography screening.8 Disparities in OOP expenditures are evident in populations with disabilities; families of children with activity limitations have higher OOP expenditures,9 with more burden falling on the lower income families.10 The prevalence of chronic conditions is rising, and there is evidence of a disproportionate burden of OOP expenditures among people who have these conditions.11 The financial burden created by cost-sharing is of concern because it has the potential to be a barrier to care and could result in poorer health status in at-risk populations who avoid these additional costs.
Recently, a number of American Physical Therapy Association (APTA) state chapters have initiated state-level legislative activity directed at limiting the impact of OOP expenditures that take the form of copayments. The APTA has drafted model legislation for fair copayments, and there are currently 5 states that have passed laws that limit the copayment for physical therapy.12 The data that support these legislative efforts come from state and federal surveys and indicate that nationally there is wide variation in copayments per visit, ranging from $10 to $40.13–16
Our study examined the burden of OOP expenditures on patients in the United States who report the use of physical therapist services. We used Andersen's “behavioral model”17,18 as a conceptual framework to identify the independent variables (ie, predictors) for having any OOP expenditure and for expenditure magnitude. The behavioral model was developed to model health care utilization, and we used it to frame our research because factors associated with utilization should align with factors associated with the expenditures incurred due to utilization. In the behavioral model, the 3 constructs associated with utilization at the individual level are need factors, predisposing factors, and enabling factors. Need factors are the variables that describe perceived health status and health conditions. When access to care is equitable, need factors are the strongest predictors of utilization.17 Predisposing factors are demographic descriptors and variables related to socioeconomic status. Enabling factors are the resources needed to pursue care such as insurance, income, and having a usual source of care. Predisposing and enabling factors may predominantly influence utilization when there are inequities of access to care.17
Based on previous studies of OOP expenditures, we hypothesized that having any OOP expenditure would be associated with enabling factors such as income and health insurance coverage.19 We also hypothesized that predictors of the magnitude of OOP expenditures for physical therapy would be associated with the enabling factor of geographic variation20 and that higher income would be associated with higher OOP expenditures.19
Method
Data Source
This study used data from the Medical Expenditure Panel Survey (MEPS) Household Component. The MEPS is a survey sponsored by the Agency for Healthcare Research and Quality (AHRQ) that provides data drawn from a nationally representative sample of the US civilian noninstitutionalized population. Detailed information and survey methodology for the MEPS are available through the AHRQ MEPS website.21 The MEPS is designed, in part, to examine issues of utilization and expenditure. Machlin et al22 used the MEPS for studying utilization and expenditures for episodes of ambulatory physical therapy. The MEPS also has been used as a data source to examine OOP expenditures for mammography screening as a specific health service8 and OOP expenditures associated with various health conditions, including psoriasis,23 inflammatory bowel disease,24 osteoarthritis,25 and other chronic health conditions.11
To construct the analytic file, we used the longitudinal data files and the annual office-based and hospital outpatient event files from the most recently published publicly available data sets. Two years of data were available for each of the 4 MEPS panels we used, which were: panel 13 (2008–2009), panel 14 (2009–2010), panel 15 (2010–2011), and panel 16 (2011–2012). By pooling multiple panels, we were able to generate sample sizes that were sufficient for nationally based estimates and that reduced sampling error for these estimates.
Participants
The unit of analysis for this study was the individual participant's episode of physical therapy care. To be included in the study, the individual had to have participated in the entire 2-year survey cycle and to have had at least one visit to a physical therapist. Physical therapy visits were identified using the “type of provider seen” variable in the office-based and hospital outpatient events files. Episodes of care were constructed in a manner similar to the study by Machlin et al22 such that an episode was a group of visits with a gap of no more than 90 days between consecutive visits. Episodes with visits occurring in the first or last 60 days of the 2-year cycle of empanelment were excluded because we were unable to determine when an episode of care started or stopped for these visits. We used a combination of 60-day and 90-day windows to balance our desire to maximize our sample size while limiting the extent to which we would underestimate visit counts due to censoring at the beginning and end of each panel and the extent to which we would overestimate visit counts per episode with a shorter time interval. As the time window to avoid censoring had a much greater impact on sample size, we elected to use a less conservative cutoff of 60 days rather than 90 days.
For any participant with more than one episode of care in the 2-year period, only the first episode of care was used. The final analytic file included 2,189 participants who had an episode of physical therapy care (Fig. 1). Of the 2,189 participants, only 229 had more than one episode of care, with 211 having 2 episodes and 18 having 3 episodes.
Sample construction and composition from Medical Expenditure Panel Survey (MEPS) public use files to a single analytic file. PT=physical therapy.
Variables
In the analysis of factors related to having an OOP expenditure, the dichotomous dependent variable was defined as having $0 or greater than $0 in OOP expenditures across the episode of physical therapy care. All expenditure data and visit count data were derived directly from the 2 events files and summed across all visits in an episode of care. Expenditures were adjusted to 2012 dollars using the Consumer Price Index.26,27 Total expenditures for an episode represented both payments by insurers and other sources and OOP expenditures by the participant. The OOP expenditures included copayments and any other direct payments made by the individual participant for care. A variable was generated to mark all episodes associated with any OOP expenditure.
Need factor variables were: number of visits in the episode of care, diagnosis, perceived health status, mental health status, and the presence of a functional limitation. The cumulative count of visits to physical therapy providers associated with the episode was derived from the MEPS events files to represent the number of visits in the episode of physical therapy care. The diagnosis variable was created by classifying the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code of the primary household-reported condition associated with the visits as either being musculoskeletal or any other condition.22 The perceived health status and mental health status variables originated from a MEPS question in which the individual was asked to rate his or her health or mental health as excellent, very good, good, fair, or poor. We collapsed these response categories to very good to excellent, good, and poor to fair to ensure sufficient sample sizes in each category. The variable representing the presence of a functional limitation was derived from a MEPS question that asked if the individual had “difficulties walking, climbing stairs, grasping objects, reaching overhead, lifting, bending or stooping, or standing for long periods of time.”28 The health status and functional limitations questions were asked of participants during each of the MEPS interviews, and in our analysis were matched to the time period of the MEPS when the episode of physical therapy care took place.
Predisposing factor variables were: age, sex, race, Hispanic ethnicity, and education. Age was divided into 3 categories: 0–17, 18–64, and >65 years. Race was classified as white/nonwhite, and ethnicity was classified as Hispanic/non-Hispanic. Each participant's highest completed years of education were divided into 3 categories: <12, 12, and >12 years.
Enabling factor variables were: geographic region, metropolitan statistical area (MSA) status (differentiating urban from rural geographic areas), income, insurance, and setting in which care was provided. Geographic region was based on the 4 US census regions of Northeast, Midwest, South, and West. Metropolitan statistical area status classifies participants as living in urban or rural areas. Three income categories were used in relation to poverty status, with low income being defined as less than 200% of poverty level, middle income defined as 200% to 399% of the poverty level, and high income defined as 400% of the poverty level or higher. We used the summary insurance coverage indicator provided by the MEPS that has 3 categories of insurance: private insurance indicates the person had any type of private coverage (including TRICARE/CHAMPVA), public insurance coverage only that includes Medicare and Medicaid, and uninsured. Unlike Machlin et al,22 we did not use a blended age/insurance variable because age and public insurance were not highly correlated. Among the subgroup of adults aged ≥65 years, 78% were identified as having private insurance, indicative of individuals with supplemental insurance in addition to Medicare. The setting in which care was provided is dichotomized in the MEPS as either a hospital-based outpatient clinic or an office-based clinic.
Data Analysis
We used the “SVY” survey estimation commands in Stata version 13.1 (StataCorp LP, College Station, Texas) to utilize longitudinal weights, strata, and cluster variables supplied by the MEPS to compute nationally representative descriptive statistics for all episodes of physical therapy care. To examine predictors of any OOP expenditures among all people who had an episode of physical therapy care, we first examined unadjusted, bivariable relationships using chi-square tests for categorical variables and t tests for continuous variables. To adjust for multiple factors, we used multivariable logistic regression analysis.
To address our second aim of identifying predictors of the magnitude of OOP expenditures, we used a generalized linear model with gamma distribution and log link in the subsample that had any OOP expenditure. Expenditure data are typically skewed because of the small percentage of patients in the tail of the distribution with larger values. Linear regression models of expenditure data can produce biased results, which even with transformed expenditure data cannot be easily interpreted. The generalized linear model accommodates skewness and provides reliable estimates of the mean and variance relationship of the data.29 The appropriateness of specifying the model with a gamma distribution and log link was verified by a modified Park test.30
Results
Characteristics of the Study Sample
The sample contained 2,189 participants who had an episode of physical therapy care that took place in 2008–2012 and were part of 1 of 4 MEPS panels. The sample with population weights applied represents 53 million episodes in the United States across the 5 years of study. Among these episodes, 53.7% (95% confidence interval [CI]=51.3, 56.0) of patients incurred at least $1 in OOP spending. Table 1 provides the weighted descriptive information on the entire sample. On average, there were 9.9 (95% CI=9.1, 10.6) visits per episode of physical therapy care, with mean total expenditures of $1,709 (95% CI=$1,498, $1,919) and a median value of $792. The full sample contained predominantly working age adults (65.5% between the ages of 18 and 64 years) with musculoskeletal conditions (78.1%; 95% CI=75.9, 80.2). The sample was 60.6% (95% CI=58.5, 62.8) female, and 88.4% (95% CI=86.4, 90.1) were reported as white and only 7.3% as Hispanic (95% CI=6.1, 8.7). Most care was provided to patients who live in urban areas (85.1%; 95% CI=80.8, 88.6) and in office-based settings (88.9%; 95% CI=87.2, 90.5). Both the full sample and the subsample with OOP expenditures were fairly evenly divided among the 4 MEPS panels (χ2=61.12, P=.70).
Sample Characteristics and Expenditures for an Episode of Physical Therapya
Predictors of OOP Expenditures Versus No OOP Expenditures
Among patients with some amount of OOP spending, mean OOP expenditures were $351.43 (95% CI=$299.03, $403.83) per episode of care, with a median of $144.00 (Tab. 2). When OOP expenditures occurred, they were 19.5% of the total expenditures over the course of an episode of physical therapy care. The OOP expenditures equated to a mean value of $44.73 expended per visit and a median value of $22.96 per visit. The odds of incurring an OOP expenditure were significantly higher for women, non-Hispanic individuals, and those with higher incomes and significantly lower for those in worse general health, nonwhites, those with publicly funded insurance coverage, and adults aged >65 years (Tab. 3).
Characteristics of Episodes of Care With OOP Expendituresa
Factors Associated With OOP Expenditures for an Episode of Physical Therapya
Characteristics of the Subsample With OOP Expenditures
The distribution of the sample by the magnitude of OOP expenditures per episode of care is illustrated in Figure 2. The top 1% of those with an OOP expenditure accounted for 14.6% of all OOP expenditures. The top 5% accounted for 38.2% of the total. Overall, the top 50% accounted for 90.6% of all OOP expenditures, with a mean OOP expenditure per episode of $638.50 or $67.79 per visit.
Distribution of out-of-pocket expenditures for episodes of physical therapy (PT) 2008–2012. The first column represents all episodes with an out-of-pocket expenditure, and the second column represents all the out-of-pocket expenditures made for PT episodes of care. The shading shows proportionally how the episodes correspond to a share of the total out-of-pocket expenditures made for episodes of PT. For example, 1% of the PT episodes correspond to 14.6% of the total expenditures, and this 1% has a mean per episode expenditure of $5,246 and expenditure per visit of $304.
Predictors of Higher OOP Expenses
In the generalized linear model, after controlling for the number of visits in an episode and other factors, the enabling factors related to a participants living in other geographic locations were associated with higher OOP expenditures relative to those living in the Northeast census region and in those living in urban versus rural areas (Tab. 3).
Discussion
Our research provides a comprehensive assessment of the financial impact of OOP expenditures on patients in the United States who receive ambulatory care from a physical therapist. We identified predictors of incurring OOP expenditures, the absolute amounts of OOP expenditures, and the relative burden these costs impose on patients. Approximately 54% of episodes of physical therapy care are partially or fully funded by the patient through OOP expenditures that are, on average, $45 per visit. Women, non-Hispanics, and individuals with a higher income are more likely to have OOP expenditures. Individuals in fair to poor health, older adults, nonwhite people, and those insured through public programs are less likely to have OOP expenditures. After controlling for the number of visits in an episode, the strongest predictors of the magnitude of expenditures for an episode of physical therapy care were living in areas other than the Northeast region and living in an urban area.
This study expands on work by Machlin et al,22 who used MEPS data from the time period of 2004–2007 to study the determinants of utilization and total expenditures for an episode of physical therapy care. The characteristics of the participants and the episodes of care in our more current MEPS panels are quite similar to those in the study by Machlin et al,22 who provided the first benchmark data, with an average of 9.6 visits per episode and average expenses adjusted to 2012 dollars of $1,311. Comparably, the average number of visits per episode in our sample was 9.9, and the increase in expenses per episode was $1,709, indicating a growth in episode-level spending on physical therapy. Machlin et al22 did not report OOP costs for physical therapy episodes of care, so we cannot evaluate whether the share of expenditures paid directly by patients has changed between the time periods covered in the 2 studies. In addition, we included all episodes of care, including not only care for adults but also care for children.
Data from 2011 for the United States indicate that 12% of people pay no OOP expenditures for health services,4 but our analysis shows that 46% of physical therapy episodes of care involve no OOP expenditure. It would appear that, as of 2012, cost-sharing in the form of an OOP expenditure is not consistently a standard approach to funding physical therapy care. Our analysis of the determinants of having OOP expenditures suggests that the burden of those expenditures is related not only to the enabling characteristics of our conceptual model but also to elements of the need and predisposing characteristics. In terms of the predisposing characteristics and patient demographics, women and non-Hispanic individuals are more likely to have an OOP expenditure, whereas nonwhites and older adults are less likely to incur OOP expenses. A series of studies on condition-based OOP expenditures similarly showed a relationship between having an OOP expenditure and sex, race, and ethnicity.23–25 Although these groups are more likely to have OOP expenditures, the average amount expended per episode of care did not differ significantly among individuals with these characteristics.
Among the need characteristics, being in poorer health resulted in a 32% lower odds of having an OOP expenditure. As the need for physical therapy may be higher among those in poorer health, this finding indicates that this subgroup is less likely to have OOP expenditures as a barrier to physical therapy access. Our analysis is limited in that we cannot determine if some participants did not receive physical therapist services because OOP expenditures were so great that access to physical therapy was never achieved.
As we had anticipated in the conceptual model, the enabling construct variables of insurance and income were both related to having an OOP expenditure. However, geographic variables were not significant in the model. The odds of incurring OOP costs were 31% lower for individuals with public insurance relative to individuals with private insurance, indicating that Medicaid (for younger individuals primarily) and Medicare programs have alleviated much of the OOP cost burden for physical therapy care for its beneficiaries. Individuals of higher income categories are more than twice as likely to have an OOP expenditure compared with individuals in the lowest income category even after controlling for insurance. From these results, we conjecture that the concept of cost-sharing as represented by OOP expenditures primarily targets higher income and privately insured individuals.
Given the amounts of OOP expenditures per episode of care and the distribution of total expenditures, it might appear that a number of individuals with higher incomes may be paying for the entire episode OOP. In the model for the magnitude of OOP expenditures after controlling for visits in an episode, living in an urban area was associated with 72% greater OOP expenditures, and in terms of geographic regions, the South region is associated with 44% greater and the West region with 29% greater expenditures compared with people in the Northeast region. The geographic results are consistent with current APTA lobbying efforts at limiting patient expenditures through the fair copayment model legislation, which are focused in these same geographic regions.12
In our sample, only 10.8% of episodes of care were completely self-paid. Episodes that were funded completely OOP were evenly distributed across the 3 income levels and across all quintiles of the OOP expenditure distribution. The highest burden of OOP expenditures for physical therapy is disproportionately concentrated; 25% of patients who incur any OOP expenditure accounted for 75% of all OOP expenditures. Fifty percent of patients with OOP expenditures accounted for only 10% of the total OOP expenditures on physical therapy episodes of care. This disproportionate pattern with a concentration of spending is typical in spending for all health services and for OOP expenditures for all medical care.31,32 We believe that if there is a desire to reduce the level of OOP expenditures, attention to the small proportion of higher expenditure episodes could result in the largest reduction of OOP expenditures for physical therapy.
We undertook this study to understand the nature of OOP expenditure and to ascertain if there were clear disparities in who incurred these expenditures for physical therapy. Although the mean value of $351 for an episode of physical therapy care may not appear to be excessive, using Machlin and Carper's4 annualized mean of $703 indicates that for those individuals the physical therapy episode of care encompassed half of their OOP expenditures for the year. We believe that for some individuals the OOP expenditure could be a barrier to obtaining physical therapy care and that the policy implications could include advocacy for a limit on OOP expenditures for physical therapy services as a smaller proportion of annual OOP health spending. At the least, our findings indicate that the profession should continue to monitor the proportion of those who have OOP expenditures and the amount incurred.
Our study had several strengths, but also some important limitations. The use of the MEPS, a population-based data source, and the pooling of multiple years are strengths of our analysis. Although the MEPS collects information on the cost of insurance premiums, the survey is not designed to provide specific information about the proportion allocated to cover specific care types, including physical therapy. Nevertheless, rising insurance premiums contribute to the financial burden of medical care for many patients. Additional limits of the data source include that the MEPS does not incorporate institutionalized individuals and that the information in the files is self-reported and based on household responses to the repeated surveys. It is possible that respondent recall bias is a limitation of the data. The AHRQ does not validate the household or individual response data and validates only a proportion of the expenditures. Furthermore, we recognize that these estimates are limited to only physical therapy provided in ambulatory care settings and that both unmeasured factors and possible selection bias among those who receive physical therapy care may explain some of the observed differences in OOP expenditures. Future research should address OOP expenditures for institutional care to gain an understanding of their impact.
Our analysis was restricted to the first episode of physical therapy care during the period of participant empanelment. For 90% of the participants, this was the only episode of care; however, not including all episodes may be viewed as a limitation of this study. The insurance coverage variable we used from the MEPS was restricted to 3 broad categories that resulted in a limited analysis of insurance type on OOP expenditures. The insurance variable did not clearly indicate people receiving Medicare or whether people with Medicare also had supplemental coverage. This lack of information on supplemental coverage certainly influenced our findings for those with Medicare coverage.
Our analysis is limited to individuals who had at least one physical therapy visit, and we could not discern to what extent cost-sharing in the form of OOP expenditures may influence a patient's decision to initiate an episode of physical therapy. Finally, although we used the most recently available MEPS data, it is possible that OOP expenditures for physical therapist services after 2012 have changed relative to that reported in our study. Recently published trend analyses show growth in patients' health care financial burden.33,34 Cost-sharing mechanisms, including copayments and high deductible plans, are anticipated to increase; thus, OOP expenditures for physical therapy also are likely to increase over time.
This study offers practitioners and policy makers a US population-based analysis of the burden of OOP expenditures placed on patients who receive physical therapist services. These data may assist legislative efforts to alter the landscape of cost-sharing and OOP expenditures for physical therapy care. We provide national estimates that describe those who incur OOP expenditures in an episode of ambulatory physical therapy care and the amounts incurred. We found several predictors of having an OOP expenditure and of the magnitude of that expenditure. The proportion of those incurring OOP expenditures for physical therapy is lower than for general medical care.
Footnotes
Dr Chevan and Dr Riddle provided concept/idea/research design and writing. Dr Chevan provided data analysis and project management. Dr Riddle and Dr Reed provided consultation (including review of manuscript before submission).
- Received January 16, 2015.
- Accepted June 11, 2015.
- © 2015 American Physical Therapy Association