A comparison of resource use and cost in direct access versus physician referral episodes of physical therapy.

Key Words: Direct access, Episode of care, Physical therapy, Physician referral.

In many states, the practice of physical therapy is contingent on the prescription or referral by a physician, a requirement that effectively limits access to physical therapy services. Other states have enacted legislation permitting direct access--the ability of a health care consumer to freely visit a physical therapist without first securing referral from a physician. In these states, licensed therapists may evaluate patients without referrals and make autonomous decisions about subsequent clinical management.[1]

Although direct access in the United States dates back to 1957, the majority of states with direct access statutes have permitted physical therapists to treat and evaluate patients without physician referral only since the 1980s.[2] No published research has evaluated the impact of physician referral versus direct access on utilization and costs of care for persons undergoing physical therapy. This exploratory study compared the utilization of health care resources and third-party medical expenditures for persons receiving physical therapy under direct access versus those referred for such services by a physician.

We begin this report by providing some background on direct access to physical therapy. Next, we describe a study method based on the analysis of episodes of physical therapy created using Blue Cross-Blue Shield claims data. The final section discusses empirical results study limitations, and implications for public policy.

Background

Thirty states allow physical therapists to treat and evaluate patients without physician referral, and an additional 14 states allow physical therapists to evaluate, but not treat, patients without referral.[2] Twenty states and the District of Columbia require physician referral as a prerequisite for treatment by a physical therapist.[1]

Advocates for physical therapists to have direct access argue that direct access extends consumers' choice of health care providers, improves access to services that promote prevention and rehabilitation, and reduces delays before commencing therapy. Proponents further argue that direct access may result in cost savings by avoiding the referring physician's fees and related ancillary services (eg, roentgenograms, laboratory tests). Supporters of direct access also point out that other nonphysician providers, such as chiropractors and clinical psychologists, do not require physician referrals or screening evaluations.[1,3]

Critics of direct access argue that physical therapists may overlook serious medical conditions and for this reason contend that all patients should be screened initially by physicians.[1,3] The American Medical Association (AMA) contends that although allied health care professionals are useful as physician extenders, they would not serve the public as well in an autonomous role.[4] The AMA and the American Academy of Orthopaedic Surgeons oppose independent practitioner status for physical therapists because of concerns about improper diagnosis, inappropriate care, and the potential for increased costs.[5] State medical societies and chiropractic groups have also been major adversaries of direct access. A common concern is that direct access legislation may lead therapists to diagnose and treat beyond their level of competency, thus erroneously assuming the role of physician.[6,7]

Previous research on direct access to physical therapy has considered the incidence of direct access practice,[1,8,9] patient and provider satisfaction with physical therapy received under direct access,[2,10] and physical therapist and patient opinions about direct access to physical therapy.[11-13] The limited available evidence from these published studies indicates that direct access has had only a minimal impact on physical therapy practice.[1,8-10] In some of these studies,[1,10] however, physical therapists expressed greater job satisfaction and patients preferred the more expeditious treatment received.[1,10]

Method

The Data

The study is based on health insurance claims data furnished by Blue Cross-Blue Shield of Maryland. This insurer has been reimbursing for physical therapy provided under direct access since 1986, so the coverage is well established. Group insurance paid claims represent a broad cross section of the employed population and their dependents. Because these individuals obtained health insurance through employer-sponsored plans, the effect of adverse selection, which characterizes persons with individual policies (or no insurance), is minimized. Although the data encompassed a number of different employer groups, the range of services covered and the level of reimbursement among groups in the sample were virtually identical. The plans covered only working-age adults and their children; persons eligible for Medicare (age 65 years and over) were not examined.

The data set included all paid claims for the calendar years 1989 through mid-1993. The initial file contained 1.7 million claims in four categories: professional fees, outpatient services (de, radiology, laboratory, and ancillary services), prescription drugs, and hospitalization. Each record contained a unique beneficiary identification number, date of service, type of service, submitted charge, amount reimbursed by Blue Cross-Blue Shield, and subscriber copayment amount. Claims for professional services also included a designation of clinical specialty (eg, licensed physical therapist, orthopedic physician, chiropractor), Current Procedural Terminology (CPT) code for type of service, and ICD-9-CM (International Classification of Diseases, 9th Revision, Clinical Modification) diagnostic code for the condition.

Analytical Framework--Episodes of Physical Therapy

Health insurance claim files comprise a series of discrete transactions that document beneficiary encounters with the medical care system. Claims records can be grouped sequentially to construct episodes of care that encompass a series of temporally contiguous health care services related to treatment of a specific illness or health condition.[14] Recent studies have used the episodes framework to examine the decision to seek medical care, subsequent utilization of services, and expenditures.[14-20]

The main advantage of using claims data for health services research is that observations on a large number of individuals over an extended period of time can be obtained at relatively low cost. When compared with audits of medical records, this method for assessing medical care has limitations. First, only sparse information is available for each encounter, and this information has been collected for administrative rather than clinical purposes. Second, the validity of episode construction is contingent on algorithms created by the investigator. Error may arise from either the inclusion of irrelevant transactions or the exclusion of transactions actually related to the condition of interest. Third, a subject s health history and clinical status at the start of an episode must be inferred from the pattern of prior claims. Similarly, outcome of treatment following an episode must also be deduced from the presence (or absence) of subsequent claims. Finally, medical expenditures paid directly by the patient, such as charges for over-the-counter drugs, are not documented (although this is also true of medical records).

Episodes of physical therapy were constructed with guidance from an advisory panel of five licensed health care professionals practicing in Maryland. Panel members were selected from a list of candidates provided by the Maryland Physical Therapy Association in response to a request for names of active practitioners specializing in physical therapy and orthopedic medicine. The panel consisted of three physical therapists and two physicians (an orthopedic surgeon and a physical medicine/ rehabilitation specialist). Additional insight on the idiosyncracies of the claims data was provided by the medical director of Maryland Blue Cross-Blue Shield. Panel functions were to develop criteria for constructing episodes of care and to establish rules for classifying episodes as either direct access or physician referral.

An episode of physical therapy should encompass all services provided in relation to a specific illness or condition during a suitable time period.[14] At the time of this study, physical therapy performed by a licensed physical therapist was billed under "physical medicine" procedure (CPT) codes. Other health care professionals such as physicians and chiropractors also utilize these CPT codes for services performed, even though they are not licensed physical therapists and thus may not be performing identical services. For purposes of classification, we refer to episodes of care defined by physical medicine procedures as physical therapy, irrespective of the health care provider who rendered the service.

We first identified all individuals who had at least one physical therapy claim during period January 1990 through December 1991. Approximately 11,600 individuals met this criterion. We then sorted each individual's claims for the period 1989 through 1993 in chronological order by date of service and created a window of observation extending from 12 months prior to the date of the first physical therapy service to 12 months after the last physical therapy service. This window contained all or part of one or more episodes of care.

Criteria for marking an episode's beginning and end points were devised by the advisory panel. We examined the 30-day period prior to the first physical therapy claim that occurred during the period January 1990 through December 1991. If no physical therapy claim occurred during the 30 days preceding the first physical therapy service, this date marked the beginning of an episode of care. If a physical therapy claim did occur within that 30-day period, the next 30-day period prior to that claim was reviewed. This process was repeated for each preceding 30-day period until reaching the initial transaction in the data set (January 1, 1989).

We then identified the last physical therapy service that occurred during the period January 1990 through December 1991. The panel recommended examining a 45-day period subsequent to this encounter. If no physical therapy claims were recorded during this 45-day period, then the last physical therapy service marked the end of the last episode. Alternatively, if a physical therapy service was recorded during this subsequent 45 days, the episode was deemed incomplete and the next 45-day period following the physical therapy service was examined. Again, this procedure was repeated until reaching the end of the data set (December 31, 1992). Using this approach, we created a new file containing observations on approximately 3,500 persons who had at least one episode of physical therapy that began and ended during the period 1989 through 1992.

These beginning and end points could actually mark different episodes. For this reason, we next examined the 45-day period occurring after the date established as the commencement point of the episode denoted by the first physical therapy service in order to distinguish among multiple episodes. If a physical therapy encounter occurred within 45 days after the commencement of an episode, the two encounters were considered part of a single episode. This procedure was repeated for all subsequent physical therapy services. If a period of 45 days occurred in which there was no physical therapy service, then the date of the last physical therapy service prior to the 45 days in which no physical therapy services were rendered marked the end of the episode. If another physical therapy service was observed beyond this 45-day posttreatment period, then this date marked the commencement point of another episode.

Classification of Episodes

After creating episodes of physical therapy, the next task was to classify episodes as either direct access or physician referral. Because claims data do not differentiate direct access episodes from those that were referred, we adopted decision rules recommended by the advisory panel. The classification algorithm, depicted in the Figure, differentiated eight categories of episodes.

We first examined the 30-day period prior to the first physical therapy service within each episode to determine whether there was a claim for a physician service with either ICD-9-CM codes or CPT codes indicating a condition that could reasonably lead to the provision of physical therapy. The panel recommended a focus on only acute and sporadic musculoskeletal-related disorders (ICD-9-CM codes 710-739 and 840-848). The 30-day period was deemed conservative because a typical person receiving a prescription for physical therapy could likely schedule an initial appointment within 2 weeks. We then determined whether claims for physical therapy services within the episode were rendered by a licensed physical therapist in order to exclude physical therapy services rendered by other providers (eg, chiropractors). If these criteria were met, the episode was classified as a physician referral (category 3).

Episodes for which there was no indication that a physician encounter occurred in the 30-day period preceding the first physical therapy service were then examined to determine whether services were provided by a licensed physical therapist. Category 7 contained episodes in which claims for diagnostic or evaluation procedures were recorded for the first encounter with the physical therapist. Criteria for category 8 were identical to those for category 7 except that no initial claims for diagnostic evaluation were observed. Categories 7 and 8 were grouped together and comprise the direct access episodes. Other categories (1, 2, 4, 5, and 6) did not meet the criteria for either direct access or physician referral and were excluded from the analysis.

We then visually inspected the set of transactions comprising episodes in categories 3, 7, and 8. Following recommendations of the advisory panel, we excluded episodes that involved claims for chronic musculoskeletal conditions (eg, arthritis, cancer, multiple sclerosis, osteoporosis). We also excluded episodes in which the patient appeared to have multiple comorbidities. These episodes tended to contain visits to a number of different providers for a range of health problems, making it impossible to determine whether physical therapy received by the patient represented treatment for the initial encounter with a musculoskeletal diagnosis. The final analysis file comprised 252 direct access and 353 physician referral episodes.

Statistical Analyses

We first compared the mean values of utilization and cost variables for direct access versus physician referral episodes using a two-tailed test for differences between means, with a null hypothesis of no difference (Tab. 1). Because simple comparisons do not control for confounding factors, we also used multiple regression analysis to compare direct access and physician referral episodes with respect to utilization (number of physical therapy visits) and costs. Definitions of variables used in the analysis are presented in Table 2. Summary statistics for the dependent and explanatory variables follow each definition.

Table 1 Comparison of Mean Values for Resource Utilization and Cost in Direct Access Episodes Versus Physician Referral Episodes

                                    Direct        Physician
                                    Access        Referral
Variable                            (n = 252)     (n = 353)

Physical therapy claims                20.2         33.6
                                      (82.9)       (39.0)
Physical therapy office visits          7.6         12.2
                                       (9.1)        (12.8)
Physical therapy claims paid ($)      566           890
                                     (716)         (941)
Drug claims                             1.47          3.13
                                       (4.0)         (7.72)
Drug claims paid ($)                   36            78
                                     (109)         (223)
Radiology claims                        0.32          1.02
                                       (1.03)        (1.86)
Radiology claims paid ($)              44           175
                                     (190)         (541)
Hospital admissions                     0.25          0.64
                                       (0.80)        (1.17)
Hospital admissions paid ($)           83           397
                                     (402)       (1,003)
Total claims paid ($)               1,004         2,236
                                   (2,030)       (2,827)

Variable                            Differences(a)

Physical therapy claims                  13.4

Physical therapy office visit             4.6

Physical therapy claims paid ($)        324

Drug claims                                1.66

Drug claims paid ($)                      42

Radiology claims                           0.70

Radiology claims paid ($)                131

Hospital admissions                        0.39

Hospital admissions paid ($)             315

Total claims paid ($)                  1,232

(a) P < .01.

Table 2.
Definitions of Variables Used in Regression
Analyses

Variable        Definition

Dependent
Logarithm-physical therapy visits    Natural logarithm of the count
                                     of physical therapy office
                                     visits during the episode (X=
                                     1.78 SD=1.12)
Logarithm-physical therapy paid      Natural logarithm of total
                                     dollar amount reimbursed
                                     by Blue Cross-Blue Shield
                                     for physical therapy
                                     services received by patient
                                     during the episode (X=6.03,
                                     SD=1.26)
Logarithm-total paid                 Natural logarithm of total
                                     dollar amount reimbursed by
                                     Blue Cross-Blue Shield for
                                     all services
                                     received by patient during
                                     the physical therapy episode
                                     (X=6.61, SD= 1.48)
Independent
Direct access                        Dichotomous variable: 1 if
                                     episode was direct access
                                     (category 7 or 8), 0 if episode
                                     was physician referral
                                     (category 3) (X=0.58, SD=0.49)
Female                               Dichotomous variable: 1 if the
                                     beneficiary gender was female,
                                     0 if male (X=0.63, SD=0.48)
Age                                  Beneficiary age (in years)
                                     (X=42.19, SD=12.5)
Drugs                                Dichotomous variable: 1 if
                                     the episode contained any
                                     claims far prescription
                                     drugs, 0 if otherwise
                                     (X=0.42, SD=0.49)
Hospital                             Dichotomous variable: 1 if the
                                     episode contained any claims
                                     for inpatient or outpatient
                                     services provided by an acute
                                     care general hospital, 0 if
                                     otherwise (X=0.25, SD=0.44)
Radiology                            Dichotomous variable: 1 if
                                     the episode contained any
                                     claims for diagnostic radiology
                                     services provided by a
                                     physician or freestanding
                                     imaging center, 0 if otherwise
                                     (X=0.29, SD=0.46)
Direct access-drugs                  Interaction of "direct access"
                                     and "drugs": 1 if a direct
                                     access episode contained
                                     prescription
                                     drug claims; 0 if otherwise
                                     (X=O.12, SD=0.32)
Direct access-hospital               Interaction of "direct access"
                                     and "hospital": 1 if a direct
                                     access episode contained claims
                                     for hospital services; 0 if
                                     otherwise (X=0.55, SD=0.23)
Direct access-radiology              Interaction of "direct access"
                                     and "radiology": 1 if a direct
                                     access episode contained
                                     diagnostic radiology claims
                                     performed at a physician office
                                     or freestanding imaging center,
                                     0 if otherwise (X=0.55, SD=
                                     0.23)

The total cost of each episode of physical therapy was computed as the sum of all paid claims for services and drugs provided during the episode. A logarithmic transformation was performed on the dependent variables to adjust for observed right-skewed distribution, which is typical of medical utilization and expenditure data.[21] The primary explanatory variable of interest was referral status. The dichotomous variable "direct access" identified episodes in categories 7 and 8 while category 3 (physician referral) served as the reference category. Three dichotomous variables were constructed to identify episodes that contained any claims for hospital services (hospital), pharmaceuticals (drugs), and diagnostic imaging rendered via a physician's office or freestanding center (radiology). All three categories of service must be prescribed by a physician and thus suggest greater severity of illness than episodes not including these services. To further distinguish episodes involving any or all of these services by referral status, we constructed interaction terms. These terms identified direct access episodes that involved claims for hospital services (direct access-hospital), pharmaceuticals (direct access-drugs), and imaging procedures (direct access-radiology). Additional variables controlled for age and gender.

Results

Table 1 shows simple comparisons using tests for differences between means. Physician referral episodes were characterized by 13.4 (67%) more physical therapy claims and 4.6 (60%) more office visits than direct access episodes (P<.0001). Reimbursements for physical therapy services were, on average, $324 (57%) more expensive for physician referral episodes when compared with direct access episodes (P<.0001). Total paid claims averaged $2,236 for physician referral episodes and $1,004 for direct access episodes; this $1,232 difference signifies that the cost to Blue Cross-Blue Shield for physician referral episodes exceeded the cost for direct access episodes by about 123% (P<.001).

Table 3 displays the results of regressions where the dependent variables were the number of physical therapy visits, paid claims for physical therapy services, and total paid claims for all services and drugs. In each case, the dependent variable has been transformed and is expressed as its natural logarithm. Adjusted multiple regression ([R.sup.2]) values indicate that models account for about 25% of the variation in the logarithm of physical therapy visits and for about 21% for the logarithm of physical therapy claims. The regression explains 48% of the variation in total paid claims for all services and drugs.

Table 3.
Regression Estimates for Number of Physical Therapy Visits, Paid Claims for
Physical Therapy Services, and Paid Claims for All Services

Independent                  Number of Physical
Variable                     Therapy Visits (Log)

Direct access(b)             -0.503(***)
                             (0.111)
Drugs                         0.361(**)
                             (0.10488)
Hospital                      0.268(*)
                             (0.121)
Radiology                     0.479(**)
                             (0.117)
Direct access-hospital(c)     0.127
                             (0.251)
Direct access-drugs(c)        0.601(**)
                             (0.178)
Direct access-radiology(c)   -0.298
                             (0.248)
Female(b)                     0.112
                             (0.083)
Age                          -5.643
                             (0.003)
Constant(c)                   1.504(**)
                             (0.155)

Adjusted [R.sup.2]             .247
F statistic                  22.94

Independent                  Paid Claims for Physical
Variable                     Therapy service (Log)

Direct access(b)             -0.519
                             (0.134)
Drugs                         0.346(**)
                             (0.124)
Hospital                      0.274(*)
                             (0.142)
Radiology                     0.534(**)
                             (0.138)
Direct access-hospital(c)     0.106
                             (0.295)
Direct access-drugs(c)        0.644(*)
                             (0.210)
Direct access-radiology(c)   -0.107
                             (0.292)
Female(b)                     0.161
                             (0.098)
Age                          -0.002
                             (0.004)
Constant(c)                   5.756(**)
                             (0.184)
Adjusted [R.sup.2]             .212
F statistic                  17.34

Independent                  Total Paid Claims for All
Variable                     Services and Drugs (Log)

Direct access(b)             -0.864(**)
                             (0.125)
Drugs                         0.425(**)
                             (0.116)
Hospital                      0.934(**)
                             (0.134)
Radiology                     0.853(**)
                             (0.130)
Direct access-hospital(c)     0.133
                             (0.269)
Direct access-drugs(c)        0.685(**)
                             (0.198)
Direct access-radiology(c)    0.249
                             (0.272)
Female(b)                     0.149
                             (0.092)
Age                          -0.002
                             (0.004)
Constant(c)                   6.191(**)
                             (0.173)
Adjusted [R.sup.2]             .479
F statistic                  61.79

(a) Standard errors of regression coefficients are in parentheses. Single asterisk (*) indicates P < .05, double asterisk(**) indicates P < .01. (b) Reference category for "direct access" is "physician referral"; reference category for "female" is "male." (c) Interaction term between "direct access" and named variable.

In each model, the coefficient for the variable "direct access" was negative (P<.01), implying that episodes of physical therapy classified as direct access involved fewer visits and lower costs relative to episodes classified as physician referral. Coefficients for the variables identifying episodes of physical therapy that included claims for drugs, hospitalizations, or radiology were positive and significant at P<.01. These findings imply that physician referral episodes with claims for any or all of these services are characterized by more physical therapy visits, higher paid claims for physical therapy services, and higher total costs per episode relative to physician referral episodes that do not involve drugs, hospitalizations, or imaging procedures.

Interaction terms that identified direct access episodes involving hospital inpatient services or imaging were not significant, implying that such services have little bearing on use of physical therapy or episode costs. By contrast, direct access episodes that contained one or more claims for pharmaceuticals were associated with more physical therapy visits, higher paid claims for physical therapy, and higher total episode costs. The variables controlling for gender and age had negligible effects on both utilization and costs.

Because log-transformed results cannot be interpreted directly, the coefficients for the direct access variables have been converted to percentages (Tab. 4). Relative to physician referral episodes, those episodes classified as direct access involved 65% fewer physical therapy visits and 68% lower paid claims for physical therapy services.

Table 4.
Percentage of Difference in Utilization and Cost for Direct Access
Episodes Relative to Physician Referral Episode

                                              Difference Relative
                                              to Physician
Model Dependent Variable                       Referral Episode

Number of physical therapy visits                   -65%

Paid claims for physical therapy services           -68%

Total paid claims for all services and drugs       -137%

(a) Based on regression results shown in Table 3.

The lower utilization rates for all services that characterized direct access episodes is best seen by examining total episode costs. When measured in terms of paid claims, direct access episodes were 137% less expensive than those classified as physician referral.

Discussion

Thirty states have legislation enabling patients to obtain physical therapy services without physician referral (direct access). The public policy objective for direct access statutes is to give the consumer the ability to select the most appropriate source of care. Consumers, however should be protected against underprovision of care that could occur if physician services were not provided when medically necessary.

Using Blue Cross-Blue Shield claims data from Maryland (a state with direct access statutes), we compared episodes of physical therapy categorized as direct access relative to those classed as physician referral and found substantial differences. Direct access episodes were shorter, encompassed fewer numbers of services, and were less costly than those classed as physician referral. Some direct access episodes included claims for inpatient hospital care, drugs, or outpatient radiology--all services requiring physician prescription. The use of hospital services or imaging procedures during direct access episodes had a negligible relationship with the number of physical therapy visits or episode costs. In contrast, direct access episodes that contained claims for drugs were associated with greater use of physical therapy and higher costs. Physician referral episodes that included any or all of these three items were associated with higher utilization and costs.

Because our study was based on health insurance claims data, these findings must be interpreted with caution. The method relied on sorting algorithms to identify episodes of care and to distinguish direct access from physician referral. We cannot be certain that resource use attributed to episodes and their classification accurately identified each patient's course of therapy. In addition, we have no way of knowing whether the lower cost of direct access episodes was due to underprovision of care or whether the greater resource intensity and cost of physician referral episodes reflects overprovision of care.

Conclusions

We conclude that direct access episodes, on average, are short in duration and relatively inexpensive. Potential explanations why this may be the case include lower severity of the patient's condition, overutilization of services by physicians, and underutilization of services by physical therapists. Concern that direct access will result in overutilization of services or will increase costs appears unwarranted. The fact that some direct access episodes included physician-prescribed services indicates that physical therapists are making referrals to physicians. Thus, our study offers evidence that public policy objectives for direct access to physical therapy services are being achieved.

Acknowledgments

We acknowledge the invaluable guidance provided by advisory panel members Richard Hinton, MD, Cindy Juris, MD, Annette Iglarsh, PhD, PT, Rod Schlegel, PT, and Mark Valente, PT. Insight on use of Maryland Blue Cross-Blue Shield claims data was provided by Alan Wright, MD. Chuanfa Guo provided expert computer programming in construction of episode-of-care files. Emily Tobias Shumsky assisted in the detailed inspection of final-analysis files. Comments on an earlier version of this manuscript were provided by Jack Hadley, Vivian Hamilton, and Robert Hurley.

[Figure 1 ILLUSTRATION OMITTED]

References

[1] Domholdt E, Durchholz AG. Direct access use by experienced therapists in states with direct access. Phys Ther. 1992;72:569-574.

[2] Physical Therapy Practice Without Referral: Direct Access. Alexandria, Va: American Physical Therapy Association, Department of Government Affairs; July 1992.

[3] Overman SS, Larson JW, Dickstein DA, Rockey PH. Physical therapy care for low back pain: monitored program of first-contact nonphysician care. Phys Ther. 1988;68:199-207.

[4] Physical therapists and patient access without physician referral. Medical World News. 1987;28(17).

[5] Barr J. It's time to be direct about direct access. Physical Therapy Today. Spring 1991.

[6] Durant TL, Lord LJ, Domholdt E. Outpatient views on direct access to physical therapy in Indiana. Phys Ther. 1989;69:850-857.

[7] Dorste T. Physical therapists: green light on direct access. Hospitals. 1987;61(15):113.

[8] Jette AM, Davis KD. A comparison of hospital-based and private outpatient physical therapy practices. Phys Ther. 1991;71:366-381.

[9] Dennis JK. Decisions made by physiotherapists: a study of private practitioners in Victoria. Australian Journal of Physiotherapy. 1987;33: 181-191.

[10] James IJ, Stuart RB. Expanded role for the physical therapist: screening musculoskeletal disorders. Phys Ther. 1975;55:121-132.

[11] Domholdt E, Clawson A, Flesch P, Taylor T. Direct Access to Physical Therapy. Alexandria, Va: American Physical Therapy Association; 1991: chap 4.

[12] Hamouz l., Barron A, Porter RE. Indiana physical therapists' attitudes toward direct access legislation. Phys Ther. 1989;69:393. Abstract.

[13] LeMasters A, Domholdt E. Direct access opinions of physical therapy students. Phys Ther. 1989;69:392-393. Abstract.

[14] Hornbrook MC, Hurtado AV, Johnson RE. Health care episodes: definition, measurement, and use. Medical Care Review. 1985;42(2): 163-218.

[15] Hillman BJ, Joseph CA, Mabry MA, et al. Frequency and costs of diagnostic imaging in office practice a comparison of self-referring and radiologist-referring physicians. N Engl J Med 1990;323:1604-1608.

[16] Hillman BJ, Olson GT, Griffith PE, et al. Physicians' utilization and charges for outpatient diagnostic imaging in a Medicare population. JAMA. 1992;268:2050-2054.

[17] Keeler EB, Rolph JE. The demand for episodes of treatment in the health insurance experiment. Journal of Health Economics. 1988;7:337-367.

[18] Garnick DW, Luft HS, Gardner LB, et al. Services and charges by PPO physicians for PPO and indemnity patients: an episode of care comparison. Med Care. 1990;28:894-906.

[19] Keeler EB, Manning WG, Wells KB. The demand for episodes of mental health services. Journal of Health Economics. 1988;7:369-392.

[20] Haas-Wilson D, Cheadle A, Scheffler R. Demand for mental health services: an episode of treatment approach. Southern Economic Journal. 1989;56:219-232.

[21] Manning WG, Newhouse JP, Duan N, et al. Health insurance and the demand for medical care. American Economic Review. 1987;77:251-277.

JM Mitchell, PhD, is Associate Professor, Graduate Public Policy Program, Georgetown University, 3600 N St NW, Room 105, Washington, DC 20007 (USA) ([e-mail]). Address all correspondence to Dr Mitchell.

G de Lissovoy, PhD, MPH, is Vice President, MEDTAP International, 7101 Wisconsin Ave NW, Suite 600, Bethesda, MD 20814.

An earlier version of this article was presented at the Winter Meeting of the Econometric Society; Washington, DC; January 6, 1995.

This research was supported by the American Physical Therapy Association.

This article was submitted November 30, 1995, and was accepted September 23, 1996.