Physical therapy use by community-based older people.

Community-based older people use a variety of health care providers to maintain or improve their health. (1-3) Physical therapists are one group of providers who primarily focus on improving, maintaining, or limiting decline in the physical function of the older person. Physical therapy may be provided in inpatient, outpatient, or home settings. National estimates from the Medicare Current Beneficiary Survey (MCBS) indicate that in 1998 more than 1.4 million community-based older people received physical therapy in one or more of these settings. (4) Estimates from the 1998 National Ambulatory Medical Care Survey and the National Hospital Ambulatory Medical Care Survey indicate that more than 3.8 million visits made to physicians by people 65 years of age and older included a prescription for physical therapy). (5,6) In 1996, on an average day, approximately 350,000 people 65 years of age and older were receiving physical therapy in their homes. (7)

Despite the substantial use of physical therapy by community-based older people and its potential for improving outcomes, (8-16) information on the characteristics of people who use physical therapy and factors associated with its use are limited. The few studies that have considered physical therapy in their analyses are outdated, examined physical therapy use in a cursory manner, or included physical therapy in an aggregate measure of health care use. (17-22) White-Means, (22) for example, in her analysis of health care use by older people with disabilities, created one variable to represent the number of visits to a physical therapist, occupational therapist, speech therapist, or hearing therapist. Because the care provided by these therapists is different, grouping the data in this manner precludes any meaningful conclusions about physical therapy.

Identifying the characteristics of people who use physical therapy and the factors associated with its use is a useful first step in determining whether disparities exist in physical therapy use. Variation in physical therapy use, explained by factors other than health and need, would suggest that disparities may he present. Racial, ethnic, socioeconomic, and geographic disparities in health care delivery have been well documented in the literature. Numerous studies suggest that people of a lower socioeconomic status (SES), racial and ethnic minorities, and certain geographic groups are not receiving necessary care or are receiving care of a lower quality. (23,24) More recent studies also have raised questions about disparities in health care provided to women, children, elderly people, and those with chronic illnesses. (23)

Disparities in health care use are coupled with disparities in health. People of a lower SES and racial and ethnic minorities tend to be in poorer health than people of a higher SES and the majority Caucasian population. (23,24) Demographic trends also indicate that the gap between the richest and poorest households in America is widening and the rate of growth of some racial and ethnic minorities exceeds that of the rest of the population. (23) Disparities in our health care system are pervasive, and eliminating these disparities is a major priority of the US Department of Health and Human Services. (23)

Although numerous studies have identified disparities in the use of a variety of health care services, whether or to what extent these disparities actually exist for physical therapy is largely unknown. Two studies that examined use of home health by older people suggest that living in a rural area may be a barrier to physical therapy access. (19,21) White-Means (22) also found that older African Americans with disabilities reported more incidences of medical conditions and disabilities, but had fewer visits to at least one type of therapist (ie, physical therapist, occupational therapist, speech therapist, or hearing therapist) than older Caucasian people with disabilities.

Disparities or differences in the use of health care may not always be indicative of underuse of services. Regional differences in health care use and spending by Medicare beneficiaries, for example, are considered to be indicative of overuse of services in some parts of the United States. Fisher et al (25) found that regional differences in Medicare spending, alter controlling for differences in cost and illness severity, are largely explained by differences in the use of inpatient services and specialists. Medicare beneficiaries who lived in higher-spending regions of the country used more inpatient services and saw more specialists. Quality of care, however, was no better for these beneficiaries relative to Medicare beneficiaries who lived in lower-spending regions. Findings such as these have implications in regard to controlling health care costs.

The objectives of this study were: (1) to identify factors associated with physical therapy use by community-based older people and (2) to identify factors associated with the amount of physical therapy received. We considered these objectives an important first step in determining whether disparities in physical therapy use exist. Variation in physical therapy use and in the amount of physical therapy received, explained by factors other than health and need, would suggest potential underuse of services (ie, not getting necessary physical therapy) or overuse of services. Both situations can be detrimental--the former because quality of care is compromised, and the latter because of unnecessary health care costs. If disparities in physical therapy use are present, future endeavors should determine the underlying reasons for these disparities so that efforts to eliminate these disparities can be appropriately targeted.

Method

Data Source

Data for this study were obtained from the MCBS. (26) The MCBS, sponsored by the Centers for Medicare and Medicaid Services (CMS), is an ongoing, longitudinal survey of a nationally representative cohort of Medicare recipients that was started in 1991. The primary goals of the MCBS are to determine costs for all health care services used by Medicare beneficiaries, including copayments, deductibles, and noncovered services; to determine all types of health insurance coverage and to relate coverage to payment for health care services; and to trace processes over time, such as changes in health status and the effects of Medicare program changes. Since 1994, the survey has been conducted using a rotating panel design with a target sample size of 12,000 people per year. Each fall, a new panel or group of beneficiaries is admitted to the survey and followed over the next 3 calendar years.

The sample for the MCBS is drawn from CMS's Medicare enrollment files using a multistage sampling strategy, which consists of first selecting 107 geographic areas of the United States, including Puerto Rico, that are representative of the nation. These geographic areas or primary sampling units (PSUs) consist of groups of counties and are used in national surveys to reduce the cost of traveling while maintaining national representation. Beneficiaries residing in these PSUs are then selected by systematic random sampling within the following age strata: 0 to 44 years, 45 to 64 years, 65 to 69 years, 70 to 74 years, 75 to 79 years, 80 to 84 years, and 85 years or over. Sampling rates vary by age in order to overrepresent disabled people (<65 years of age) and the oldest-old ([greater than or equal to]85 years of age). Because beneficiaries are selected without regard to type of residence, beneficiaries living in the community and long-term care facilities are represented.

Sampled beneficiaries or appropriate proxies are interviewed in-person using computer-assisted personal interviewing. The survey instruments for community-based and facility-based beneficiaries differ to some degree. Because our study focused on community-based beneficiaries, the following description of the interview process is specific to community-based respondents.

The introductory interview, which occurs in the fall, introduces the respondents to the survey. During this interview, respondents are provided with a calendar to record details of health care use (inpatient, outpatient, home health) for the following year. They are encouraged to collect their Medicare and insurance statements, supporting bills, receipts, and prescriptions in preparation for the next interview. Baseline demographic information also is gathered during the first interview, along with information about health insurance, health status, functioning, access to care, and attitudes about medical care. Respondents are then interviewed 3 times per year for the next 3 calendar years to obtain detailed information on insurance coverage and the use and cost of all health care services received, including services not covered by Medicare. The calendar and the receipts and bills collected by the respondent are specifically reviewed as part of the interview process. On a recurring annual basis each fall, information on health status, functioning, access to care, and attitudes about medical care is gathered.

The MCBS interview data are linked to Medicare claims and administrative data to increase the analytic power of the data set. Combining the survey and claims data provides a more complete picture of health care use, costs, and sources of payment than either source alone. Because there are differences in the ways that medical goods and services are characterized in the survey and in the Medicare claims records, some health care events are recorded in the survey only, some are represented in the claims data only, and some are represented in both. The MCBS staff has developed an elaborate set of reconciliation and imputation rules to present the most accurate picture possible of the use of and cost of health care services during the year.

Public-use data sets from the MCBS are issued on a calendar-year basis and include the Access to Care file and the Cost and Use file. The Access to Care file contains annual information on beneficiaries' demographics, insurance, health status, function, access to care, and attitudes about care. The file is augmented with Medicare claims data on the use and cost of Medicare services. The Cost and Use file, which also is augmented with Medicare chums data, provides complete expenditure and source of payment data on all health care services used by the beneficiary, including those not covered by Medicare. This file also contains information on demographics, insurance, health status, and function. Adler (27) and the CMS (26) provided a comprehensive profile of the MCBS, including more specific information on the sampling strategy, survey questions, and timing of data collection.

Sample

The analyses presented here are based on MCBS data for 1994-1998. The response rate over that 5-year period was approximately 70%. Tim sample was identified using the Cost and Use files and was restricted to people living in the United States who were 65 years of age or older and who were solely community-based (N=43,889). Because we were interested in the respondents' use of physical therapy during a period defined by the calendar year, we excluded respondents who died during the calendar year (n=1,787). Although health care use and expenditures tend to be higher in the last year of life, (28) we considered the subset of beneficiaries who died during the calendar year to differ in enough ways from individuals who did not die during the calendar year to justify their exclusion.

We also eliminated individuals who did not have any type of physician encounter (ie, in any type of inpatient or outpatient setting) during the calendar year (n=1,790). We limited the sample to people who had at least one physician encounter, because Medicare requires a physician referral for physical therapy and because the physical therapist practice acts of many states require that physical therapists only treat patients referred by a physician. Because we were specifically interested in identifying factors associated with physical therapy use, and not physician use, we reasoned that including respondents who had no physician encounters during the calendar year would confound the interpretation of our results. This was confirmed by the fact all subjects in the sample who had one or more physical therapy events also had one or more physician encounters. Approximately 5% of the records were eliminated because of missing data on demographic or health-related characteristics, for a final sample of 38,312 person-years.

The term "person-years" can be used to indicate that the same individuals are represented in more than 1 year. Because the MCBS follows beneficiaries over a 3-year period and because the time frame for our analysis was the calendar year, some respondents were included in the sample once (n=8,026), some were included twice (n=6,317), and some were included 3 times (n=5,884). Our sample of 38,312 person-years, therefore, came from data on 20,227 individuals. We chose to include subjects more than once to increase our sample size and because we could account for this in our data analysis. Furthermore, because the MCBS sample for each calendar year is nationally representative, inclusion of the same subject for more than 1 year is appropriate.

Analytic Framework

The analytic framework for this study derives from Andersen and Newman's behavioral model of health care use. (29) This model is the most widely adopted framework for studying health care use and is the most amenable conceptualization for framing secondary analyses. (30) The model views the use of health care as a function of the predisposing, enabling, and need characteristics of the individual. The predisposing component reflects the fact that some individuals have a greater propensity to use health care than others. Predisposing characteristics include sociodemographic characteristics and attitudes and beliefs about health care. Predisposing characteristics, in and of themselves, are not directly responsible for health care use. For example, race is not considered a reason for seeking health care. Rather, people of different races have different experiences, beliefs, and attitudes that affect their health care use. Although individuals may be predisposed to use health care services, they must have some means for obtaining them. This is reflected by the enabling component, which includes family resources, such as income and insurance coverage, and community resources, such as the supply of health care providers.

Assuming the presence of predisposing and enabling characteristics, the individual must perceive illness or the probability of its occurrence for the use of health care services. This is reflected in the need component. The need component is considered the most immediate cause of health care use and can include a variety of measures that reflect the individual's health (eg, self-reported function, comorbidities, symptoms). The specific variables we chose to represent the predisposing, enabling, and need characteristics of subjects are presented in the next section. Our choice of variables was based on Andersen and Newman's model, the data available in the MCBS, the designs of previous studies that used secondary data to examine factors associated with the health care use, our experience and knowledge in the area of health care use and access, and the primary author's (JKF's) clinical experience.

Study Variables

Descriptive statistics for tire study variables are presented in Table 1. A majority of the variables were dichotomous or categorical. All of the data, with the exception of data for 5 variables, were extracted from the Cost and Use files.

Dependent variables. The 3 dependent variables were: (1) any physical therapy use during the year (yes/no), (2) number of physical therapist visits for the year, and (3) cost of physical therapy for the year. The physical therapist visits for the year represent inpatient (acute and subacute), outpatient, and home health visits. The physical therapy cost variable was the sum of fees that were paid by insurance, paid by the respondent out-of-pocket, and not paid for each physical therapist visit for the year. Physical therapy cost data were missing for 6% of the physical therapist visits.

Predisposing characteristics. Predisposing characteristics were represented by age, sex, race, ethnicity, and amount of education. Age was treated as a categorical variable. Variables that represented subjects' attitudes about medical care in the fall prior to the study year also were included to represent attitudes that may predispose the subject to use more or less health care during the study year. Each year, in the fall round, subjects are asked to respond to 6 statements regarding their satisfaction with medical services received from all physicians and hospitals over the past year. They also are asked to respond to 6 statements regarding the care provided by their usual health care provider. These statements are presented in the Appendix.

A principal components factor analysis with varimax rotation (31) was conducted to determine if the statements regarding satisfaction with medical services and attitudes about usual care represented 9 distinct constructs. Based on the eigenvalue greater than one rule and examination of the scree plot, (31) 2 factors were retained. Statements regarding satisfaction with medical services loaded most heavily on one factor, with all factor loadings being [greater than or equal to]0.72. Statements regarding usual care loaded most heavily on the other factor, with all factor loadings being [greater than or equal to]0.58. The factor loadings are presented in Table 2.

Based on the results of the factor analysis, and for sake of parsimony, the statement that loaded most heavily on each factor was chosen to represent that construct. Satisfaction with medical services was represented by the subjects' response to the following: "Tell me how satisfied have you been with the information given to you about what was wrong with you." Ordinal responses ranged from "very satisfied" (1) to "very unsatisfied" (4). This statement had a factor loading of 0.79. Attitudes about usual care were represented by the subjects' response to the following: "Tell me how strongly you agree or disagree with this statement: The doctor(s) answers all of your questions." Ordinal responses ranged from "strongly agree" (1) to "strongly disagree" (4). This statement had a factor loading of 0.78.

Enabling characteristics. Enabling characteristics included income, which was treated as a categorical variable, and several dichotomous variables to represent insurance coverage, whether the subject had a usual source of care, whether someone accompanied the subject on physician visits, and whether the subject had a problem and did not see a physician. The latter variable was created based on the subjects' response (yes/no) to the following question: "During this year, did you ever have any health problem or condition about which you think you should have seen a doctor or other medical person, but did not?" Other enabling characteristics included a dichotomous variable to indicate whether the subject lived in a metropolitan area and a categorical variable to indicate the census division in which the subject lived. Numerous studies have documented variation in health service use based on geographic location. (32) A measure of physical therapist supply in the county where the subject lived also was included. This variable was created using data from the 1997 Area Resource File. (33) A long history of health services research has shown a positive correlation between resource availability and use of services. (34)

Need characteristics. Need characteristics included measures of comorbidity, general health, difficulty with physical function, and difficulty with activities of daily living (ADL) and instrumental activities of daily living (IADL). For a majority of the subjects, these data were collected in the latter 4 months of each year. These measures, therefore, may have been obtained before physical therapy started, during the course of physical therapy, or after physical therapy had ended. Although we had the option of using data from the previous fall (ie, using measures of health and function in the year prior to physical therapy use) these data are limited because they are likely reflective of the subjects' health prior to some change in health that would have led to physical therapy use. We considered the measures of health and function obtained in the same year as physical therapy use to be the "better" indicators of need, recognizing the limitations of these measures due to the potential temporal problems.

The MCBS survey asks whether the subject has a history of any of the following comorbidities: stroke, hip fracture, arthritis, osteoporosis, partial paralysis, atherosclerosis, hypertension, heart condition, cancel (excluding skin cancer), diabetes, Alzheimer disease, Parkinson disease, respiratory problems, and extremity amputation. One variable represented the total number of comorbidities the subject had out of the 14 listed. Dichotomous variables also were included to indicate whether the subject had a history of stroke, hip fracture, arthritis, osteoporosis, or partial paralysis. These comorbidities were chosen because they are conditions that are often managed by a physical therapist. Dichotomous variables for Parkinson disease and amputation were not included because less than 1% of the subjects had either of these comorbidities.

Subjects rated their general health by responding to the following question: "In general, compared with other people your age, how would you rate your health?" Response choices were "excellent," "very good," "good," "fair," and "poor." Based on the distribution of responses, this variable was dichotomized into "'general health good or better" and "general health fair or poor."

Physical function was assessed by asking subjects how much difficulty they had, on average, with the following 4 activities: (1) stooping, crouching, or kneeling, (2) lifting or canting objects as heavy, as 10 lb (4.5 kg), (3) reaching or extending arms above shoulder level, and (4) walking 2 to 3 blocks. Response choices were "no difficulty" (1), "little difficulty" (2), "some difficulty" (3), "lot of difficulty" (4), and "unable to do" (5). Walking 2 to 3 blocks was excluded from the analyses because this variable was highly correlated with the stooping, crouching, or kneeling variable (r = .70) and because the MCBS addresses walking in questions about ADL.

The MCBS asks a series of questions about ADL tasks (bathing, dressing, eating, getting in or out of a chair, walking, using the toilet) and IADL tasks (using telephone, doing light housework, doing heavy housework, preparing meals, shopping, managing money). The questions about ADL tasks determined whether the subject had any difficulty with the activity, whether the subject used any equipment to assist with the activity, and whether the subject required any help with the activity. Two mutually exclusive, dichotomous variables were created to summarize the responses to the ADL questions. One variable indicated whether the subject received help with one or more ADL tasks. The second variable indicated whether subjects who did not receive help with any ADL tasks had difficulty with one or more ADL tasks, Difficulty was defined as subject-reported difficulty or use of equipment for the activity. Two similar dichotomous variables were created to summarize responses to the IADL questions. One variable indicated whether the subject received help with one or more IADL tasks. The second variable indicated whether subjects who did not receive help with IADL tasks had difficulty with one or more IADL tasks. Difficulty was defined as subject-reported difficulty only because there were no questions regarding use of equipment for IADL tasks. The ADL and IADL data were coded in this manner based on the distribution of responses and because there are data to suggest a hierarchical association with loss of abilities with ADL and IADL tasks (ie, subjects first have difficulty with an activity and then require help with the activity). (35)

Finally, 3 dichotomous variables were created to indicate whether a subject had one or more inpatient admissions, one or more home health events, or one or more subacute admissions (skilled nursing facility, long-term care facility, or rehabilitation hospital) during the study year.

Data Analysis

A total of 3 analyses were conducted, a logistic regression analysis and 2 ordinary least squares (OLS) regression analyses. First, a logistic regression analysis was conducted to identify predisposing, enabling, and need characteristics associated with any physical therapy use during the year (N = 38,312 person-years). The significance level fin the analysis was set al P [less than or equal to] .05. For the subsample of people who had at least one physical therapist visit during the year (n = 1,840 person-years), 2 OLS regression analyses were conducted to identify predisposing, enabling, and need characteristics associated with amount of physical therapy received. The dependent variables for these analyses were: (1) the natural log of the total number of physical therapist visits during the year and (2) the natural log of the total cost of physical therapy for the year. The physical therapist visits and physical therapy cost variables were log-transformed to decrease the influence of outliers. Prior to transformation, total cost of physical therapy was adjusted to 1996 costs using the Consumer Price Index. (36) Because the OLS regression models lacked power due to the small sample size, statistical significance was set at P <. 10. The sequential approach of first modeling the use of physical therapy and then modeling the intensity of use conditional on any physical therapy is sometimes referred to as a "hurdle model" and is a relatively common approach for modeling health care utilization. (37)

As noted previously, a majority of the independent variables were dichotomous or categorical in nature. When conducting multivariate analyses with such variables, interpretation of the results is based on a referent. For dichotomous variables with a "yes/no" response, such as history of arthritis, the referent is those individuals without the characteristic (eg, those without arthritis). For categorical variables, the referent must be chosen by the analyst. For the categorical independent variables that were ordinal (ie, age, income), we chose the lowest category (ie, 65-69 years of age, [less than or equal to] $10,000) as our referent for all analyses. For categorical variables that were not ordinal (ie, census division, race), we chose the category with the greatest representation as the referent (ie, South Atlantic division, Caucasian).

Dummy variables were included in all 3 models to account for year effects and missing data on the satisfaction and access to care variables. All analyses were conducted in Stata (version 8.0) * using the survey commands, (38) which accounted for the sampling weights, the clustering of observations within PSUs, and the sampling of subjects within age strata. With population-based surveys such as the MCBS, sampling weights, clustering, and stratification must be accounted for in order to obtain accurate point estimates and standard errors.

Results

Logistic Regression Analysis

The results of the logistic regression analysis are presented in Table 3. Twelve of the 17 need characteristics had statistically significant odds ratios (ORs) (P [less than or equal to] .05). Several of the enabling and predisposing characteristics also had significant ORs. An OR greater than 1 implies that the event, in this case the use of physical therapy, is more likely to occur far subjects who have that characteristic than for those who do not have that characteristic. That is, the characteristic is positively associated with the occurrence of the event. An OR of less than 1 implies that the event is less likely to occur for subjects who have the characteristic than for those who do not have that characteristic or that the characteristic is negatively associated with the occurrence of the event. For example, history of arthritis was positively associated with physical therapy use (OR = 1.73), whereas number of comorbidities was negatively associated with physical therapy use (OR = 0.94). When the occurrence of the event is infrequent (ie, <10%), the OR can be interpreted as a relative risk. Considering the OR for people with a history of arthritis, we interpret this as meaning that people with a history of arthritis were 73% more likely to use physical therapy relative to people without a history of arthritis. An OR of 0.94 for number of comorbidities can interpreted as follows: far each one-unit increase in the number of comorbidities subjects had, the likelihood of physical therapy use decreased by 6%.

Ordinary Least Squares Regression Analyses

The results of the OLS regression analyses, which explored factors associated with the amount of physical therapy received for those who received any physical therapy (n = 1,840 person-years), are presented in Table 4. Because there were missing cost data, the analysis with cost as the dependent variable was based on a sample size of 1,737 person-years. Positive coefficients mean that the amount of physical therapy received or the cost of physical therapy was greater for the given characteristic. Negative coefficients mean that the amount of physical therapy received or the cost of physical therapy was lower for the given characteristic. Only one need characteristic was statistically significant in these analyses (P [less than or equal to] .05). People who needed help with ADL tasks had higher physical therapy costs compared with those who did not need help. The beta coefficient for this variable was .27. Because the dependent variable was the natural log of physical therapy costs, this coefficient can be interpreted as follows: people who needed help with ADL tasks had physical therapy costs that were 27% higher than the physical therapy costs of those who did not need help.

Several enabling characteristics were significantly associated with the number of physical therapist visits or total cost of physical therapy (P [less than or equal to] .05). In regard to predisposing characteristics, African-American race was the only variable significantly associated with physical therapy costs and physical therapist visits.

Discussion and Conclusions

This is the first study that has identified factors associated with physical therapy use in a nationally representative sample of community-based older people. As we expected, several need characteristics were associated with any physical therapy use during the year. These findings are consistent with Andersen and Newman's model of health care use, which identifies need characteristics as the most immediate cause of health care use. (29) Having one or more inpatient, subacute, or home health events had the strongest associations with physical therapy use, with ORs ranging from 1.73 to 2.05. This finding is not surprising because these variables are proxies for a change in health status that may lead to the need for physical therapy.

The hierarchical nature of the relationships between the ADL and IADL variables also seemed appropriate. People who had difficulty or received help with one or more ADL tasks were 52% to 53% more likely to use physical therapy relative to people who did not have difficulty or receive help (Ors = 1.52 and 1.53, respectively). In addition, people who needed help with one or more IADL tasks were 23% more likely to use physical therapy relative to those who did not need help (OR = 1.23). The relationship between physical therapy use and difficulty with physical function also was what we expected to find. People with more difficulty stooping or reaching over their head were more likely to use physical therapy compared with those who had little difficulty stooping or reaching over their head (Ors = 1.17 and 1.12, respectively). Amount of difficulty lifting 10 lb was not associated with physical therapy use. One likely explanation for this finding is that difficulty lifting was correlated with difficulty reaching overhead (Pearson r = .57). When 2 independent variables are correlated or collinear, the effect of one may be masked by the effect of the other in multiple regression analyses.

Although number of comorbidities was negatively associated with any physical therapy use, people with a history of arthritis, osteoporosis, or partial paralysis--diagnoses commonly treated by physical therapists--were more likely to have used physical therapy during the year relative to individuals without a history of these diagnoses. Hip fracture and stroke, however, were not associated with physical therapy use. One explanation for these findings is that the questions on comorbidities are not specific to the study year. For example, in regard to hip fracture, subjects were asked if a physician has ever told them they had a broken hip. Strokes also can range in severity from minor events that may not require physical therapy to severe, disabling events that do require physical therapy.

Only one need characteristic was associated with amount of physical therapy received as measured by physical therapy costs. People who needed help with ADL tasks had higher physical therapy costs relative to people who did not need help with ADL tasks. Guralnik and colleagues (39) found a substantial increase in the cost of health care for community-based Medicare beneficiaries transitioning from independence with ADL tasks to requiring help with one or more ADL tasks. Chan et al (40) also found that total median health care costs for Medicare beneficiaries increased as ADL limitations increased and were due to an increase in the frequency of health care events (ie, inpatient, outpatient, or home health). In our study, physical therapy costs were positively associated with needing help with ADL tasks, but there was no association between ADL help and number of physical therapist visits.

People who were more socioeconomically advantaged were more likely to receive physical therapy than people who were less socioeconomically advantaged. Patients with incomes over $50,000, for example, were more than 100% more likely to receive physical therapy than patients with incomes of $10,000 or less (OR = 2.05), and each additional year of education increased the likelihood of physical therapy use by 5% (OR = 1.05). Patients with supplemental private insurance were 31% more likely to receive physical therapy than patients without such insurance (OR = 1.31). Income and supplemental insurance also were positively associated with number of physical therapist visits or physical therapy costs. Other researchers who have examined the influence of socioeconomic factors on health care use by older people have reported similar findings. (17,41-44)

Although managed care is often considered a barrier to health care access, subjects who participated in a managed care plan were 17% more likely to use physical therapy than subjects who did not participate in a managed care plan. Subjects who participated in managed care plans, however, incurred lower physical therapy costs than subjects who did not participate in managed care plans. The number of physical therapist visits was not affected by participating in a managed care plan.

Physical therapist supply was positively associated with physical therapy use and had a weaker, but still positive, association with amount of physical therapy received. A possible explanation for these findings is that a greater local availability of physical therapists may lead to shorter waits and greater flexibility in scheduling, which could potentially increase the use of physical therapy. Physicians also may be more likely to refer marginal cases (ie, patients who may or may not get better with physical therapy) if physical therapist supply in the area is sufficient. Hunter (45) argued that supply determines utilization and demand. That is, if physical therapists are available, then patients will be found to see them. Freburger et al, (46) using a crude measure of physical therapist supply based on census region and metropolitan status, found that physical therapist supply was positively associated with the likelihood of physicians making physical therapist referrals for adults with musculoskeletal disorders. Joling et al (47) reported the curious finding of a negative association between physical therapist supply and physical therapy use in their analysis of people with musculoskeletal complaints. Their measure of physical therapist supply, however, was based on number of physical therapists per square kilometer and did not take into account the number of people living in the area.

Having a usual source of care did not affect whether a subject saw a physical therapist, but having someone accompany the subject to a physician did. Subjects who were accompanied on physician visits were 2,5% less likely to receive physical therapy than subjects who were not accompanied on physician visits. One possible explanation for this finding is that subjects who rely on others for transportation may have to forgo less critical services such as physical therapy.

Metropolitan status was not related to any physical therapy use during the year, but people living in a metropolitan area who received physical therapy had a greater number of visits and incurred greater costs than people living in nonntetropolitan areas. These findings are similar to those reported by other researchers who have examined the effect of metropolitan status on health care use among older people. (17,20,41,42,48)

There was some geographic variation in physical therapy use and amount of physical therapy received (as measured by cost and number of visits). Relative to the South Atlantic census division, physical therapy use was less in the East South Central census division and greater in the Pacific census division, and amount of physical therapy received was less in the West North Central and West South Central census divisions. Two recent studies on nationally representative samples of Medicare beneficiaries indicated that geographic variation in health care use and spending continues to be present and is not entirely explained by differences in illness severity or cost of medical services. (25,49) Some of the literature on geographic variation in health care use also suggests that variation is particularly high for more discretionary treatments (eg, elective surgery) that lack strong scientific evidence of efficacy. (25,32,50) For such treatments, physicians' preferences, attitudes, and past experiences may influence whether they offer it to their patients. In many cases, physical therapy can be considered a discretionary treatment that lacks strong scientific evidence of efficacy. Access to physical therapy, therefore, is likely influenced by physicians' attitudes and preferences, especially because Medicare reimburses only for physical therapy prescribed by a physician.

In addition to education, which was positively associated with physical therapy use, age was a predisposing characteristic inversely related to physical therapy use. Relative to people 65 to 69 years of age, people 80 years of age and older were less likely to use physical therapy. Although health care use and disability increase with age, (1,51,52) these variables were controlled for, to some degree, in the analyses. For example, we had a variable to represent whether subjects had one or more inpatient admissions and a variable to represent difficulty with ADL tasks. Age, therefore, may be a proxy for some other unmeasured need characteristic. Male sex also approached statistical significance and was negatively associated with physical therapy use (OR = 0.86).

We found no evidence of disparities in physical therapy use due to minority status (ie, being African American or Hispanic). In many areas of health care, even after controlling for SES, racial and ethnic disparities in health care use remain. (23) Being African American, however, was associated with an increase in the amount of physical therapy received for both number of physical therapist visits and physical therapy costs. This finding may be related to unmeasured need characteristics. Some chronic diseases (eg, diabetes) are more prevalent in African-American adults than in Caucasian adults. (53) There also was no evidence of a relationship between satisfaction with health care and physical therapy use. This finding may be due to the generally high levels of satisfaction reported by subjects when asked about the health care they received and their usual health care provider. For example, the mean response to the satisfaction question included in our analysis was 1.87 (between 1 ["very satisfied"] and 2 ["satisfied"]). The mean values for the other satisfaction questions (Appendix) were similar.

Limitations

This study has several limitations. First, the analyses were limited by the data available. There are likely other unreported characteristics that would explain variation in physical therapy use. Personal preferences, for example, were not represented in the analyses, nor were more specific measures of need such as the ability to do recreational activities. Many of the subjects in the sample had little to no difficulty with ADL tasks, IADL tasks, and physical function. A second limitation is that much of the data was based on subject reports, which may be affected by recall bias. Several strategies are used to improve subjects' recall on health care use, including: interviewing subjects 3 times per year, providing subjects with a calendar to record health care use, and encouraging subjects to collect insurance statements and bills in preparation for the next interview. (26) Data on survey-reported health care use also are matched to administrative records to correct for underreporting and other errors. (54) A third limitation is the cross-sectional nature of the analyses. The timing of physical therapy use relative to the time at which ADL, IADL, and physical function abilities were reported was not accounted for in the analyses. Physical therapy use may have occurred before, during, or after data on the subjects' ADL, IADL, and physical function abilities were collected. Despite this limitation, we conclude that, on average, older people who were more likely to require help or have difficulty with ADL and IADL tasks were more likely to use physical therapy than older people who did not require help or have difficulty with ADL and IADL tasks.

The analyses were limited to community-based older people. Use of physical therapy may be very different for older people living in long-term care facilities. In addition, our analyses were limited to people who had at least one physician encounter. Only a small number of subjects did not have a physician encounter during the year (n= 1,790 person-years), and a post hoc analysis including these people did not alter our findings. A final limitation was that physical therapy use was an infrequent event among the sample, which decreased the statistical power and limited the findings in the OLS regression analyses.

Limitations notwithstanding, we believe that our study makes a valuable contribution by addressing an area of health care use that has received little attention. As has been reported in many other areas of health care, the results of our study suggest there may be disparities in access to and use of physical therapy. In an ideal world with no barriers to health care access, appropriate provision of services by providers, and assuming patient preferences were not a factor, need and need alone should explain variation in health care use. People with more severe diagnoses or illnesses would use more health care than people without known disabilities or pathology or people with less severe diagnoses or illnesses. Although several of the need characteristics were associated with physical therapy use, several enabling and predisposing characteristics were associated with physical therapy use or amount of physical therapy received. These findings suggest potential underuse or overuse of physical therapy by community-based older people.

Suggestions for Future Research

Our findings can serve as a point of departure for future efforts and studies examining issues related to physical therapy use and access. Even with the "universal coverage" of Medicare, we found that people with a lower income, people who were less educated, and people without supplemental private insurance were less likely to use physical therapy. Because people of a lower SES tend to be in poorer health, (23,24) data indicating they use less health care than people of a higher SES are typically considered to be indicative of problems with access (ie, not receiving necessary care). Initial efforts to address disparities in physical therapy access for Medicare beneficiaries 65 years of age and older should focus on beneficiaries in lower socioeconomic brackets and should attempt to identify the underlying reasons why these beneficiaries may use less physical therapy.

Other areas that may be of particular relevance based on the results of our study and the literature are the effects of metropolitan status and Medicare managed care on physical therapy use. Variation in physical therapy use by geographic location and physical therapist supply also should be explored further. Geographic variation in health care use and variation due to health care provider supply has prompted payers and policymakers to question whether there is overuse or underuse of health care in certain parts of the country and have led to efforts such as clinician profiling and clinical practice guidelines. (32) One of the biggest challenges in improving health care delivery and quality is determining the "right" amount of health care to deliver (ie, distinguishing clinically indicated and efficacious services from those that are not). Only then can it be determined whether people face barriers to receiving services that are medically necessary. Future studies, therefore, should begin to determine what is considered the appropriate amount and type of physical therapy for given diagnoses.

In their first National Healthcare Disparities Report published in July 2003, the Agency for Healthcare Research and Quality reported that the data strongly indicate that disparities exist in many areas of health care delivery. (23) Knowledge of why these disparities exist, however, is quite limited. (23) Our study provides initial data to suggest that disparities in the use of physical therapy are present. Our study also provides information that can be used to target future efforts to identify and understand why these disparities exist, with the ultimate goal of eliminating these disparities.

Appendix.

Medicare Current Beneficiary Survey (MCBS)

MCBS statements about satisfaction with medical services We're interested in how you feel about the medical services you received over the past year from medical doctors and hospitals. Please tell me how satisfied you have been with the following (response categories are very satisfied, satisfied, unsatisfied, very unsatisfied):

1. The overall quality of medical services you have received in the last year. (mcqual)

2. The ease and convenience of getting to a doctor from where you live. (mcease)

3. The out-of-pocket costs you paid for medical services. (mccosts)

4. The information given to you about what was wrong with you. (mcinfo)

5. The follow-up care you received after an initial treatment or operation. (mcfolup)

6. The concern of doctors for your overall health rather than just for an isolated symptom or disease. (mcconcrn)

MCBS statements about usual care provider

Think about the care you receive from your usual health care provider. For each Statement tell me whether you strongly agree, agree, disagree, or strongly disagree.

1. The doctor(s) often seems to be in a hurry. (ushurry)

2. The doctor (s) often does not explain your medical problem to you. (usexp)

3. The doctor(s) often acts as though he is doing you a favor by talking to you. (usfavor)

4. The doctor(s) tells you all you want to know about your condition or treatment. (ustell)

5. The doctor (s) answers all your questions. (usans)

6. You have great confidence in your doctor(s). (usconf)

Table 1.

Descriptive Statistics on Study Variables (N=38,312)

                                                 N (%) or
Variable                                         [bar.X] (SD)

Dependent variables
    Any physical therapy use during the
      year?
    Yes                                          1,840 (5%)
    No                                           36,472 (95%)
  No. of physical therapist visits during        10 (15)
    the year (n=1,840)
  Total cost of physical therapist visits        768 (1,449)
    for the year (n=1,737)
Predisposing characteristics
  Sex
    Male                                         15,696 (41%)
    Female                                       22,616 (59%)
  Race
    Caucasian                                    33,986 (89%)
    African American                             3,238 (8%)
    Other                                        1,088 (3%)
  Age (y)
    65-69                                        7,262 (19%)
    70-74                                        9,015 (24%)
    75-79                                        8,088 (21%)
    80-84                                        7,810 (20%)
    [greater than or equal to] 85                6,137 (16%)
  Education (no. of years)                       10 (4)
  Physician answers all questions (a,b)          1.83 (0.55)
    (1=strongly agree-4=strongly
    disagree)
  Satisfied with information given (a,c)         1.87 (0.51)
    (1=Very satisfied-4=very
    unsatisfied)
Enabling characteristics
  Income
    [less than or equal to] $10,000              9,855 (26%)
    $10,001-$20,000                              12,704 (33%)
    $20,001-$30,000                              7,601 (20%)
    $30,001-$40,000                              3,548 (9%)
    $40,001-$50,000                              1,918 (5%)
    >$50,000                                     2,686 (7%)
  Has Medicaid coverage                          4,233 (11%)
  Has supplemental private insurance             29,735 (78%)
  In a managed care plan                         7,678 (20%)
  Has a usual care provider (a)                  36,513 (95%)
  Is accompanied on physician visits (a,d)       15,313 (40%)
  Had a problem and did not see a
    physician (a)                                3,191 (8%)
  Lives in a metropolitan area
  Census division                                27,599 (72%)
    New England                                  1,194 (3%)
    Middle Atlantic                              6,450 (17%)
    East North Central                           6,710 (18%)
    West North Central                           2,623 (7%)
    South Atlantic                               7,434 (19%)
    East South Central                           2,131 (6%)
    West South Central                           4,115 (11%)
    Mountain                                     2,444 (6%)
    Pacific                                      5,211 (14%)
  Physical therapist supply (physical            3.54 (2.10)
    therapists/10,000 people)
Need characteristics
  History of stroke
  History of arthritis                           4,499 (12%)
  History of osteoporosis                        24,824 (65%)
  History of hip fracture                        5,025 (13%)
  History of partial paralysis                   1,804 (5%)
  Total no. of comorbidities                     1,881 (5%)
  General health good or better                  2.34 (1.53)
  Difficulty with physical function (1=little    29, 259 (76%)
      difficulty-5=unable to do)
    Stooping, kneeling, crouching                2.62 (1.39)
    Lifting 10 lb (4.5 kg)                       1.95 (1.39)
    Reaching over head                           1.59 (1.07)
  Has difficulty with one or more IADL           2,603 (7%)
    tasks (e)
  Requires help with one or more IADL            12,458 (33%)
    tasks (e)
  Has difficulty with one or more ADL            7,121 (12%)
    tasks (f)
  Requires help with one or more ADL             4,454 (12%)
    tasks (f)
  Had one or more inpatient admissions           7,364 (19%)
  Had one or more subacute admissions            1,027 (3%)
  Had one or more home health events             6,025 (16%)

(a) Extracted from Access to Care file.

(b) n=35,759.

(c) n=36,731.

(d) n=36,048.

(e) IADL=instrumental activities of daily living.

(f) ADL=activities of daily living.

Table 2.

Results of Principal Components Factor Analysis With Varimax
Rotation (a)

               Rotated Factor
                  Loadings

Variable    Factor 1    Factor 2

usconf        0.723       0.282
usans         0.780       0.247
ustell        0.760       0.239
ushurry      -0.718      -0.070
usexp        -0.772      -0.122
usfavor      -0.737      -0.136
mcqual        0.168       0.726
mcease        0.102       0.641
mccosts       0.043       0.584
mcinfo        0.223       0.789
mcfolup       0.190       0.778
mcconcrn      0.284       0.749

(a) See Appendix for definitions of variables.

Table 3.

Adjusted Odds Ratios (95% Confidence Interval [CI]) for Use of Physical
Therapy by Need, Enabling, and Predisposing Characteristics (a)

                                        Odds
Variable                                Ratio      95% CI       P

Need characteristics
  History of stroke                     0.93       0.77,111      .41
  History of hip fracture               1.13       0.89,143      .31
  History of arthritis                  1.73 **    1.49,201     <.01
  History of osteoporosis               1.43 **    1.20,1.71    <.01
  History of partial paralysis          1.33 *     1.04,169      .02
  Total no. of comorbidities            0.94 *     0.89,100      .05
  General health good or better         1.12       0.96,130      .16
  Amount of difficulty stooping,        1.17 **    1.11,123     <.01
    kneeling, crouching
  Amount of difficulty reaching         1.12 **    1.07,119     <.01
    overhead
  Amount of difficulty lifting 10 lb    0.98       0.93,103      .42
    (4.5 kg)
  Requires help with ADL (b) tasks      1.52 **    1.21,1.91    <.01
  Has difficulty with ADL tasks         1.53 *     1.31,1.79    <.01
  Requires help with IADL (c) tasks     1.23 *     1.04,1.46     .02
  Has difficulty with IADL tasks        1.12       0.85,1.48     .43
  Had one or more inpatient             2.05 **    1.79,2.34    <.01
    admissions
  Had one or more subacute              1.73 **    1.41,2.11    <.01
    admissions
  Had one or more home health           1.88 **    1.62,2.20    <.01
    visits
Enabling characteristics
  Income
    Income [less than or equal to]
      $10,000 (reference)               1.00
    Income $10,001-$20,000              1.11       0.93,1.33     .27
    Income $20,001-$30,000              1.09       0.92,1.29     .35
    Income $30,001-$40,000              1.39 **    1.12,1.72    <.01
    Income $40,001-$50,000              1.43 **    1.10,1.84     .01
    Income >$50,000                     2.05 **    1.63,2.59    <.01
  Has Medicaid coverage                 0.82       0.64,1.04     .11
  Has supplemental private              1.31 **    1.12,1.54    <.01
    insurance
  In some type of a managed care        1.17 *     1.00,1.36     .05
    plan
  Has a usual care provider             1.09       0.81,1.47     .58
  Is accompanied on physician           0.75 **    0.65,0.87    <.01
    visits
  Had a problem and did not see         1.02       0.86,1.20     .84
    a physician
  Lives in a metropolitan area          1.02       0.82,1.27     .89
  Census division
    South Atlantic (reference)          1.00
    New England                         1.07       0.79,1.44     .68
    Middle Atlantic                     1.26       0.99,1.61     .06
    East North Central                  1.16       0.90,1.50     .24
    West North Central                  1.25       0.84,1.86     .27
    East South Central                  0.64 *     0.44,0.93     .02
    West South Central                  0.79       0.59,1.04     .09
    Mountain                            1.09       0.83,1.42     .55
    Pacific                             1.35 *     1.05,1.75     .02
  Physical therapist supply             1.04 *     1.01,1.08     .02
    (physical therapists/10,000
    people)
Predisposing characteristics
  Male                                  0.86       0.74,1.00     .06
  Age (y)
    65-69 (reference)                   1.00
    70-74                               1.04       0.87,1.23     .67
    75-79                               1.02       0.84,1.25     .82
    80-84                               0.78 **    0.65,0.93     .01
    [greater than or equal to] 85       0.59 **    0.49,0.73    <.01
  Race
    Caucasian (reference)               1.00
    African American                    0.93       0.72,1.21     .61
    Other                               1.39       0.96,2.01     .08
  Hispanic ethnicity                    0.77       0.52,1.16     .21
  Education (no. of years)              1.05 **    1.03,1.07    <.01
  Physician answers all questions       0.93       0.83,1.03     .16
  Satisfied with information given      1.02       0.92,1.13     .70

(a) Controlled for year effects, missing variables.

(b) ADL=activities of daily living.

(c) IADL=instrumental activities of daily living.

* Significant at P [less than or equal to] .05.

** Significant at P [less than or equal to] .01.

Table 4.

Results of Linear Regression Analysis (a)

                                            Log of Physical Therapist
                                                     Visits

                                          Coeffi-
Variable                                   cient      95% CI (b)     P

Need characteristics
  History of stroke                       -0.10       -0.28,0.08    .28
  History of hip fracture                  0.04       -0.19,0.28    .73
  History of arthritis                     0.00       -0.15,0.14    .97
  History of osteoporosis                 -0.11       -0.25,0.04    .17
  History of partial paralysis             0.01       -0.24,0.25    .97
  Total comorbidities                      0.02       -0.03,0.08    .39
  General health good or better            0.00       -0.15,0.15    .97
  Amount of difficulty stooping,           0.03       -0.03,0.09    .35
    kneeling, crouching
  Amount of difficulty reaching            0.01       -0.04,0.06    .64
    overhead
  Amount of difficulty lifting 10 lb      -0.02       -0.07,0.03    .36
    (4.5 kg)
  Requires help with ADL (d) tasks         0.18       -0.05,0.40    .12
  Has difficulty with ADL tasks            0.06       -0.10,0.21    .48
  Requires help with IADL (e) tasks       -0.02       -0.20,0.16    .84
  Has difficulty with IADL tasks          -0.15       -0.37,0.08    .20
  One or more inpatient admissions         0.08       -0.03,0.19    .16
  One or more subacute admissions         -0.01       -0.16,0.15    .92
  One or more home health visits           0.09       -0.04,0.23    .18
Enabling characteristics
  Income
    [less than or equal to] $10,000
      (reference)
    $10,001-$20,000                        0.04       -0.12,0.20    .62
    $20,001-$30,000                        0.18 *     -0.01,0.37    .07
    $30,001-$40,000                        0.07       -0.15,0.30    .52
    $40,001-$50,000                        0.19 *     -0.03,0.41    .09
    >$50,000                               0.22 *     -0.02,0.45    .07
  Medicaid coverage                        0.11       -0.10,0.33    .31
  Has supplemental private insurance       0.15 **     0.00,0.29    .05
  In a managed care plan                  -0.12       -0.26,0.03    .11
  Has a usual care provider               -0.25       -0.56,0.06    .12
  Accompanied on physician visits         -0.02       -0.14,0.10    .76
  Had a problem and did not see a         -0.11        0.02,0.32    .22
    physician
  Lives in a metropolitan area             0.17 **     0.02,0.32    .02
  Census division
    South Atlantic (reference)
    New England                            0.05       -0.29,0.38    .78
    Middle Atlantic                        0.11       -0.07,0.29    .22
    East North Central                    -0.11       -0.29,0.07    .22
    West North Central                    -0.14       -0.35,0.07    .20
    East South Central                     0.18       -0.23,0.59    .38
    West South Central                     0.26 **     0.02,0.49    .03
    Mountain                               0.17       -0.12,0.46    .25
    Pacific                               -0.15       -0.34,0.04    .13
Physical therapist supply (physical        0.03 *      0.00,0.06    .08
  therapists/10,000 people)
Predisposing characteristics
  Male                                     0.03       -0.09,0.15    .62
  Age (y)
    65-69 (reference)
    70-74                                  0.08       -0.11,0.26    .42
    75-79                                 -0.05       -0.22,0.11    .53
    80-84                                  0.00       -0.18,0.18    .98
    [greater than or equal to] 85         -0.11       -0.32,0.10    .30
  Race
    Caucasian (reference)
    African American                       0.22 *     -0.01,0.45    .06
    Other                                 -0.03       -0.41,0.36    .90
  Hispanic ethnicity                       0.25       -0.08,0.58    .14
  Education (no. of years)                 0.00       -0.02,0.02    .92
  Physician answers all questions         -0.01       -0.14,0.09    .88
  Satisfied with information given         0.02       -0.08,0.12    .67

                                             Log of Physical Therapy
                                                    Costs (c)

                                          Coeffi-
Variable                                  cient       95% CI (b)     P

Need characteristics
  History of stroke                       -0.11       -0.35,0.14    .39
  History of hip fracture                  0.01       -0.29,0.32    .93
  History of arthritis                     0.03       -0.15,0.21    .74
  History of osteoporosis                 -0.08       -0.29,0.13    .45
  History of partial paralysis             0.01       -0.29,0.32    .94
  Total comorbidities                      0.03       -0.04,0.09    .44
  General health good or better            0.01       -0.18,0.19    .96
  Amount of difficulty stooping,           0.06       -0.02,0.13    .14
    kneeling, crouching
  Amount of difficulty reaching           -0.01       -0.07,0.05    .82
    overhead
  Amount of difficulty lifting 10 lb      -0.03       -0.09,0.03    .32
    (4.5 kg)
  Requires help with ADL (d) tasks         0.27 **     0.02,0.52    .03
  Has difficulty with ADL tasks            0.11       -0.09,0.32    .27
  Requires help with IADL (e) tasks       -0.09       -0.31,0.14    .47
  Has difficulty with IADL tasks          -0.15       -0.43,0.14    .30
  One or more inpatient admissions         0.11       -0.04,0.26    .17
  One or more subacute admissions          0.06       -0.18,0.29    .64
  One or more home health visits           0.14       -0.03,0.32    .12
Enabling characteristics
  Income
    [less than or equal to] $10,000
      (reference)
    $10,001-$20,000                        0.12       -0.09,0.33    .26
    $20,001-$30,000                        0.14       -0.10,0.37    .26
    $30,001-$40,000                        0.18       -0.13,0.48    .26
    $40,001-$50,000                        0.15       -0.17,0.47    .36
    >$50,000                               0.36 **     0.03,0.69    .03
  Medicaid coverage                       -0.08       -0.37,0.21    .60
  Has supplemental private insurance       0.10       -0.13,0.33    .40
  In a managed care plan                  -0.25 **    -0.46,-0.4    .02
  Has a usual care provider               -0.24       -0.64,0.16    .23
  Accompanied on physician visits          0.02       -0.15,0.19    .79
  Had a problem and did not see a         -0.09       -0.34,0.16    .48
    physician
  Lives in a metropolitan area             0.26 **     0.04,0.49    .02
  Census division
    South Atlantic (reference)
    New England                           -0.08       -0.58,0.42    .75
    Middle Atlantic                        0.10       -0.12,0.33    .38
    East North Central                    -0.14       -0.40,0.11    .26
    West North Central                    -0.32 *     -0.67,0.04    .08
    East South Central                    -0.12       -0.60,0.37    .64
    West South Central                     0.23 *     -0.04,0.50    .10
    Mountain                               0.27       -0.08,0.62    .13
    Pacific                               -0.19       -0.42,0.05    .13
Physical therapist supply (physical        0.04 **     0.00,0.08    .05
  therapists/10,000 people)
Predisposing characteristics
  Male                                     0.04       -0.12,0.20    .64
  Age (y)
    65-69 (reference)
    70-74                                  0.11       -0.14,0.36    .40
    75-79                                 -0.04       -0.27,0.18    .71
    80-84                                 -0.02       -0.26,0.23    .90
    [greater than or equal to] 85         -0.12       -0.41,0.17    .42
  Race
    Caucasian (reference)
    African American                       0.36 **     0.06,0.67    .02
    Other                                 -0.16       -0.73,0.41    .58
  Hispanic ethnicity                       0.26       -0.20,0.72    .26
  Education (no. of years)                -0.01       -0.03,0.02    .54
  Physician answers all questions          0.06       -0.09,0.20    .45
  Satisfied with information given        -0.02       -0.15,0.11    .76

(a) Controlled for year effects, missing variables.

(b) CI=confidence interval.

(c) n=1,737.

(d) IADL=activities of daily living.

(e) IADL=instrumental activities of daily living.

* Significant at P [less than or equal to] 10.

** Significant at P [less than or equal to] 05.

* Stata Carp, 709 University Dr E, College Station, TX 77840.

References

(1) Liu K, Wall S, Wissoker D. Disability and Medicare costs of elderly persons. Milbank Q. 1997;75:461-493.

(2) Komisar HL, Hunt-McCool J, Feder J. Medicare spending for elderly beneficiaries who need long-term care. Inquiry. 1997-1998;34 (winter) : 302-310.

(3) Friedman L, Droge JA, Rabin DL. Functional limitations among home health care users ill the National Health Interview Survey supplement on aging. Gerontologist. 1992;32:641-646.

(4) 1998 Medicare Current Beneficiary Survey. Baltimore, Md: Centers for Medicare and Medicaid Services; 2001.

(5) National Center for Health Statistics. 1998 National Ambulatory Medical Care Survey. Available at: http://www.cdc.gov/nchs/about/major/ ahcd/ahcdl.htm. Accessed December 2003.

(6) National Center for Health Statistics. 1998 National Hospital Ambulatory Medical Care Survey Available at: http://www.cdc.gov/nchs/about/ major/ahcd/nhamcsds.htm. Accessed December 2003.

(7) Munson ML. Characteristics of elderly home health care users: data from the 1996 National Home and Hospice Care Survey. Advance Data From Vital and Health Statistics. 1999; no. 309.

(8) Outpatient Service Trialists. Therapy-based rehabilitation services for stroke patients at home. Cochrane Database Syst Rev. 2003; (1):CD002925.

(9) Fransen M, Crosbie J, Edmonds J. Physical therapy is effective for patients with osteoarthritis of the knee: a randomized controlled clinical trial. J Rheumatol. 2001;28:156-164.

(10) Langhorne P, Duncan P. Does the organization of postacute stroke care really matter? Stroke. 2001;32:268-274.

(11) Mulrow CD, Gerety MD, Kantend D, et al. A randomized trial of physical rehabilitation for very frail nursing home patients. JAMA. 1994;271:519-524.

(12) Harada N, Chiu V, Fowler E, et al. Physical therapy to improve functioning of older people in residential care facilities. Phys Ther: 1995;75:830-838.

(13) Przybylski BR, Dumont Ed, Watkins ME, et al. Outcomes of enhanced physical and occupational therapy service in a nursing home setting. Arch Phys Med Rehabil. 1996;77:554-561.

(14) Freburger JK. An analysis of the relationship between the utilization of physical therapy services and outcomes of care for patients after total hip arthroplasty. Phys Ther. 2000;80:448-458.

(15) Freburger JK. Analysis of the relationship between the utilization of physical therapy services and outcomes tin" patients with acute stroke. Phys Ther. 1999;79:906-918.

(16) Tinetti ME, Baker DI, McAvay G, et al. A multifactorial intervention to reduce the risk of falling among elderly people living in the community. N Engl J Med. 1994;33:821-827.

(17) Dansky KM, Brannon D, Shea DG, et al. Profiles of hospital, physician, and home health service use by older persons in rural areas. Gerontologist. 1998;38:320-330.

(18) Branch LG, Goldberg HB, Cheh VA, et al. Medicare home health: a description of total episodes of care. Health Care Financ Rev. 1993;14: 59-74.

(19) Cheh V, Phillips B. Adequate access to posthospital home health services: differences between urban and rural areas. J Rural Health. 1993;9 (fall):262-269.

(20) Mauser E, Miller NA. A profile of home health users. Health Care lqnanc Rev. 1994;16:17-33.

(21) Kenney GM. Rural and urban differentials in Medicare home health use. Health Care Financ Rev. 1993;14:39-57.

(22) White-Means SI. Racial patterns in disabled elderly persons' use of medical services. J Gerontol B Psychol Sci Soc Sci. 2000;55:S76-S89.

(23) Agency for Healthcare Research and Quality. National Healthcare Disparities Report. Rockville, Md: US Dept of Health and Human Services; July 2003.

(24) Institute of Medicine. Unequal Treatment: Confronting Racial and Ethnic Disparities in Healthcare. Washington, DC: National Academies Press; 2003.

(25) Fisher ES, Wennberg DE, Stukel TA, et al. The implications of regional variations in Medicare spending, part 1: the content, quality, and accessibility of care. Ann Intern Med. 2003;138:273-287.

(26) Centers for Medicare and Medicaid Services. The Medicare Current Beneficiary Survey: Survey Overview. Available at: http://www.cms.hhs. gov/MCBS/Overview.asp. Accessed December 20113.

(27) Adler G. A profile of the Medicare Current Beneficiary Survey. Health Care Financ Rev. 1994;15:153-163.

(28) Hoover DR, Crystal S, Jumar R, et al. Medical expenditures during the last year of life: findings from the 1992-1996 Medicare current beneficiary survey. Health Serv Res. 2002;37:1625-1642.

(29) Andersen R, Newman JF. Societal and individual determinants of medical care use. Milbank Q. 1973;51:95-107.

(30) Wolinsky, FD. Health services utilization among older adults: conceptual, measurement, and modeling issues in secondary analysis. Gerontologist. 1994;34:470-475.

(31) Sharma S. Factor analysis. In: Applied Multivariate Techniques. New York, NY: John Wiley & Sons Inc; 1996:90-143.

(32) Wennberg DE. Variation in the delivery of health care: the stakes are high. Ann Intern Med. 1998;128:866-868.

(33) Bureau of Health Professions Area Resource bile, 1997. Rockville, Md: US Dept. of Health and Human Services, Health Resources and Services Administration, Bureau of Health Professions, Office of Data Analysis and Management; 1997.

(34) Folland S. The Economics of Health Care. 4th ed. Upper Saddle River, NJ: Pearson Prentice Hall; 2004.

(35) Gill TM, Robison .IT, Tinetti ME. Difficulty and dependence: two components of the disability continuum among community-living older persons. Ann Intern Med. 1998;128:96-101.

(36) US Bureau of Labor Statistics. Consumer Price Index Series 2003. Available at: http://www.bls.gov/cpi/home.htm. Accessed December 2003.

(37) Cragg JG. Some statistical models for limited dependent variables with application to the demand for durable goods. Econmetrica. 1971; 39:829-844.

(38) Overview of survey estimation. In: User's Guide: STATA 8. College Station, Tex: Stata Press; 2001:343-358.

(39) Guralnik JM, Alexxih L, Branch LG, et al. Medical and long-term care costs when older persons become more dependent. Am J Public Health. 2002;92:1244-1245.

(40) Chan L, Beaver S, MacLehose RF, et al. Disability and health care costs in the Medicare population. Arch Phys Med Rehabil. 2002;83: 1196-1201.

(41) Blustein J, Weiss LJ. Visits to specialists under Medicare: socioeconomic advantage and access to care. J Health Care Poor Underserved. 1998;9:153-169.

(42) Stearns SC, Slifkin RT, Edin HM. Access to care for rural Medicare beneficiaries. J Rural Health. 2000;16(winter):31-42.

(43) Grana J, Stuart B. The impact of insurance on access to physician services for elderly people with arthritis. Inquiry. 1996;33:326-338.

(44) Gornick ME, Eggers PW, Reilly TW, et al. Effects of race and income on mortality and use of services among Medicare beneficiaries. N Engl J Med. 1996;335:791-799.

(45) Hunter D.J. Getting a grip on clinical variations in hospital services. BMJ. 2004;328(7740):610.

(46) Freburger JK, Holmes GM, Carey TS. Physicians' referral to physical therapy for the treatment of musculoskeletal conditions. Arch Phys Med Rehabil. 2003;84:1839-1849.

(47) Joling C, Groot W, Janssen PP. Taking a new course in research on the effectiveness of interventions: a study of physiotherapy interventions for people with musculoskeletal complaints, Int J Rehabil Res. 2002;25:33-46.

(48) Blazer DG, Landerman LR, Fillenbaum G, et al. Health services access and use among older adults in North Carolina: urban vs rural residents. Am J Public Health. 1995;85:1384-1390.

(49) Skinner J, Weinstein JN, Sporer SM, et al. Racial, ethnic, and geographic disparities in rates of knee arthroplasty among Medicare patients. N Engl J Med. 2003;349:1350-1359.

(50) Birkmeyer JD, Sharp SM, Finlayson SRG, et al. Variation profiles of common surgical procedures. Surgery. 1998;124:917-923.

(51) Fried TR, Bradley EH, Williams CS, et al. Functional disability and health care expenditures for older persons. Arch Intern Med. 2001;161: 2602-2607.

(52) Ostchega Y, Harris TB, Hirsch R, et al. The prevalence of functional limitations and disability in older persons in the US: data from the National Health and Nutrition Examination Survey III. J Am Geriatr Soc. 2000,48:1132-1135.

(53) Pleis JR, Coles R. Summary health statistics for US adults: National Health Interview Survey, 1999. National Center for Health Statistics. Vital Health Stat. 2003;10(212).

(54) Medicare Current Beneficiary Survey 1997 Cost and Use bile Documentation, Section 5: Filling in the Gaps. Baltimore, Md: Health Care Financing Administration, Office of Strategic Planning; May 2000.

JK Freburger, PT, PhD, is Research Associate and Fellow, Cecil G Sheps Center for Health Sciences Research, University of North Carolina, Chapel Hill, NC 27599 (USA) ([e-mail]). Address all correspondence to Dr Freburger.

GM Holmes, PhD, is Research Associate and Fellow, Cecil G Sheps Center for Health Sciences Research, University of North Carolina.

Both authors provided concept/idea/research design, writing, data analysis, and fund procurement.

This research was presented at the Combined Sections Meeting of the American Physical Therapy Association; February 4-8, 2004; Nashville, Tenn.

This research was supported by the National Institute on Aging (grant no. R03-AG20065-01) and by the Agency for Healthcare Research and Quality (National Research Service Award Postdoctoral Traineeship sponsored by the Cecil G Sheps Center for Health Services Research, grant no. T32-HS00032).

This article was received December 9, 2003, and was accepted July 16, 2004.