An acute care physical therapy clinical practice database for outcomes research.

Key Words: Computers, Database management systems, Outcome and process assessment (health care), Physical disability.

In December 1986, the Department of Physical Therapy at the University of Iowa Hospitals and Clinics embarked on a venture to establish standards for physical therapy practice. The need for these standards became apparent when attempts were made to study outcomes of various therapists' treatment approaches. Not only was it difficult to determine an individual patient's treatment outcome, but it was nearly impossible to compare patient treatment outcomes among therapists using various treatment approaches. Physical therapy evaluation and treatment standards enable routine monitoring of patient outcomes and enhance the feasibility that effective physical therapy interventions can be detected.

One strategy we chose to help standardize our practice and facilitate outcome comparisons was to establish a physical therapy computerized medical record (PTCMR). This system would gather incoming physical therapy information obtained during the routine evaluation and treatment of all patients. The proposed system would be designed to minimize free text entry and require forced choice selections. This system, at the very least, would standardize some of the varied information that frequently comprises routine clinical assessments. In addition, administrative checks verifying user compliance could be routinely performed, a task that proved extremely tedious with our previous paper medical record. Our major interest in developing a PTCMR, however, was to establish a physical therapy database system that could ultimately be used for clinical research. Specialty system databases such as the Duke Cardiovascular Database (DCD)[1-5] and the American Rheumatism Association Medical Information Database (ARAMIS)[5,6] were developed to gather information generated during the routine delivery of patient care. These databases were instrumental in establishing optimal methods of managing atherosclerosis (DCD) and connective tissue diseases ARAMIS).[5,6]

Although standard entry and efficient retrieval are extremely important components of an observational computerized database, these elements alone will not quarantee high-quality clinical outcome data.[4,7] Consequently, we needed to operationally define measurement terms and establish mechanisms to routinely study the quality of the clinical measures entering the database.

Although the PTCMR was designed to be comprehensive, thus including all standard physical therapy measures of impairment (eg, range of motion, strength, pain), our data quality studies were prioritized to focus on outcome measures necessary to assess function. We discovered the terms most commonly used to assess function were also the terms that lacked operational definitions among our practicing therapists. Terms such as "independence," "standby assistance," "minimal assistance," "moderate assistance," and "maximal assistance" were routinely used to reflect a patient's ability to complete a functional task, although these terms lacked standard definitions. Grading functional activities with these terms was the criterion that therapists used to determine when patients would be discharged from physical therapy and that determined the patients' aftercare plans. Consequently, operationally defining these terms and instituting quality assessments were and continue to be an integral part of this system.

The use of clinical practice databases to determine treatment effectiveness is not without controversy.[2-7] The inability to have appropriate control groups and the overall quality of the data continue to spark extensive debate about the meaningfulness of treatment effectiveness studies using a clinical database.[4,7] Other important uses of the database, however, can assist in the development of a comprehensive goal-oriented research program. These uses include stimulating clinical hypotheses,[8] targeting high-priority areas for controlled randomized clinical trials,[2,5] quantifying descriptive outcome trends,[2,9] and deriving an estimate of the variability of a given patient population to better predict sample sizes needed for randomized clinical trials.[5]

The purposes of this article are (1) to describe the University of Iowa Physical Therapy Acute Care Computerized Data Base System developed for use in outcomes research, (2) to describe the development of functional measures entering the database, and (3) to illustrate important uses of the clinical practice database through demonstration projects.

Development and Description of the Physical Therapy Computerized Medical Record Database System

Developing a PTCMR requires extensive support and compliance from all therapists involved. Thus, the first step in developing this system was to establish the Clinical Standards of Practice Committee (begun in 1986) with representation from each specialty division (cardiopulmonary, musculoskeletal, neuromuscular), the director, and research coordinator. This committee served an important role in operationally defining terms and developing standards for the department. Equally important was the role the committee played in cultivating staff support in preparation for the complete conversion to the computerized medical record, The single most common reason for failure when designing a disability outcome database system is a lack of compliance among the staff.[10] Consequently, the development of this committee was an integral part of the methods necessary to develop a clinical practice database. The physical therapy medical record to be described has been operational since August 1991. The ability to retrieve data from the database has been available since July 1993.

Following extensive discussion and debate, a problem-oriented physical therapy medical record was determined to be the most appropriate design for the PTCMR software. This design was in contrast to the diagnosis-based approach used by the medical community. Representatives from the cardiopulmonary, neuromuscular, and musculoskeletal specialty divisions developed physical therapy problems frequently seen in their respective areas. It became apparent that across all patient types, similar problems were being addressed. This problem-oriented design, therefore, served as the common screen from which all therapists would begin the medical record. From this problem-oriented screen, all other screens would emanate. Multiple drafts of possible computer screens were simulated on paper for over 2 years. Only after a paper system had debugged multiple problems was the computerization phase implemented. Once the software was complete, training programs were provided for all staff. Following training, on a specific date (August 1991), the entire department converted to a computerized medical record.

The PTCMR problem list is the backbone of the computerized data entry system (Tab. 1). After entering patient demographic data, the problem list screen is the main screen by which all associated nested screens emanate. Screens providing specific forced-choice tests and measures accompany each problem selected.

For ease of description, associated screens for assessing function and disability will be described. For example, if the problem selected is decreased functional abilities (problem 8), then the computer loops to another screen of functional items and the amount of assistance (determined by The University of Iowa Level of Assistance Scale described later) required by the patient to perform these functional tests must be documented by the therapist. In addition, the accepted operational definitions for all terms are readily available on help screens. Items in the decreased functional abilities screen include activities such as coming from a supine to a sitting position, coming from a sitting to a standing position, standing, single-limb stance, ambulation over a therapist-specified distance, tandem walking, side stepping, crossover stepping, stair climbing, bathroom transfers, car transfers, and certain activities of daily living. The context with which these activities are performed can be duly noted (type of chair, mat versus a bed, distances, ramps, gym activities versus home and community activities). In addition, an ordinal scale (0-5) denotes the type of assistive device used to complete the functional activities.

The problem category "disability surveys" (problem 14 in the problem list shown in Tab. 1) provides a list of health-related status measures (Quality of Well Being Index, Sickness Impact Profile, Short Form-36, and Functional Status Questionnaire)[11-15] that allows the score for one or more of these surveys to be recorded. Gathering this information provides the potential to relate many of the typical physical therapy measures of impairment and function to these more recently accepted global measures of disability. Typical physical therapy measures of impairment (eg, pain, range of motion, strength, sensation) are included in the database and emanate from the corresponding categories in the PTCMR problem list presented in Table 1.

Table 1. Physical Therapy
Computerized Medical Record
Problem List[a]
 1. Decreased ability to ambulate
 2. Decreased strength
 3. Decreased range of motion
 4. Pain
 5. Impaired mental status
 6. Impaired sensation
 7. Impaired motor control
 8. Decreased functional abilities
 9. Decreased endurance
10. Decreased respiratory function
11. Wound
12. Joint instability
13. Decreased gross motor skills
14. Disability surveys
15. Other
(a) This screen allows the therapist to select the
problems identified for the patient. The system
will loop to associated screens set up with
standard inputs for each problem listed.

All information entering the PTCMR is downloaded monthly from the university hospital's mainframe computer interfaced to a physical therapy network of microcomputers. Currently, all year-to-date physical therapy information can be downloaded into 67 files that are accessed by a database software package within the microcomputer network. This database currently constitutes over 290,000 patient entries representing the last 2 years. Common outcome measures routinely calculated include days to discharge, treatment charges, number of visits, patient demographics, and comorbidity factors. Patients can be sorted by therapist, physical therapy diagnostic codes, diagnostic-related grouping codes, standard hospital-based ICD-9 codes, or any other designated variables.

Development of Functional Measures

As previously stated, the methods used to assess function received top priority for development so that meaningful outcome information was entering the database. We chose to develop our own definitions of level of assistance for grading various functional activities to be used in the database instead of adopting previously reported scales[16-22] for several reasons. The Functional independence Measure (FIM), the most well known of these scales, defines, as a percentage, the level of assistance required to complete a battery of functional tests. The battery of tests comprising the FIM were not items routinely assessed by our acute care therapists (eg, self-care, communications, cognitive skills, sphincter control). We could not achieve staff consensus on what constituted 25%, 50%, 75%, or 100% effort as is required for the FIM. Furthermore, a full report of the interrater reliability of the FIM had not been published at the time and, to date, has been presented only in abstract form[20] or alluded to briefly in a separate report.[23]

We had preliminary data supporting the reliability of our scale[24,25] and full compliance among our staff in using our scale. We therefore decided to continue to develop our own definitions of level of assistance to be used in the physical therapy medical record and database. Care was taken during this development because we believed these measures provided the most important data that acute care therapists used to make clinical decisions about their patients' discharge status. A training manual was developed, and periodic written examinations were given to ensure the competence of all users of this scale. This measurement standard, as well as other measurement standards, has become a routine component of the orientation procedure for all new staff therapists.

The University of Iowa Level of Assistance Scale, used to grade various functions emanating from the decreased functional abilities problem, is shown in Table 2. In general terms, independence indicates the therapist could leave the room and the patient could safely perform the activity being assessed. Standby assistance indicates that the therapist would not feel comfortable leaving the patient, but provides no physical assistance. Minimal assistance indicates the therapist provides one point of contact to the patient, moderate assistance indicates the therapist applies two points of contact, and maximal assistance indicates the therapist or therapists are applying a total of three or more points of contact. An activity that is attempted, but not completed, with maximal assistance indicates that the patient has failed maximal assistance. A patient not tested for reasons of safety, as determined by the therapists' judgment, is graded as not tested.

[TABULAR DATA 2 OMITTED]

Demonstration of Preliminary Database Projects

Data Quality Studies

Routine assessments of data quality are an integral part of the database system. Frequently, assessments are videotaped and discussed by therapists to encourage consistent use of the functional measures. In addition, structured studies of data quality (reliability, validity, responsiveness) are ongoing.

For example, the interrater and intrarater reliability of the Iowa Level of Assistance Scale, used to grade five functional activities, was established after 288 observations of patients with total joint replacement (hips and knees). These studies were carried out among four therapists.

An index of the total score for all five functional activities assessed revealed an intraclass correlation coefficient (ICC [2,K]) of .88,[26] In addition, intertester weighted Kappa statistics for each item ranged from .41 for ambulation to .80 for stair climbing. The intratherapist weighted Kappa statistics ranged from .79 to .99 among four therapists. The knowledge gained from each data quality study facilitates the development of meaningful tests and measurements entering the clinical practice database. A report of the data quality studies for the patients with total joint replacement will be published separately.

Preliminary projects demonstrating how the database efficiently retrieves patients' summary data will be presented in the following sections,

Clinical Hypothesis Generation

In the first preliminary project, over 470 patients treated for pain were evaluated following transcutaneous electrical nerve stimulation (TENS) therapy. The standard measurement tool used to assess pain was the visual analog scale, which was administered before and after treatment. The high (10) and low (0) ends of the visual analog scale were defined as "the worse possible pain ever experienced" and "no pain," respectively. The change in pain after each TENS trial was calculated and summarized as an average change in pain. Monthly reports for each therapist were summarized for immediate feedback. In addition, the relative change in each patient's pain was determined if a second trial of TENS was administered. Consequently, the database was used, in a very general way, to assess the average change in pain between the first and second trials of TENS therapy.

Figure 1 shows the average change in the visual analog pain ratings of patients treated with TENS. On average, there was approximately a 32% reduction in pain after the use of TENS. Our therapists, however, routinely performed a second trial of TENS with different stimulation settings if there was little relief from the first trial. Figure 2 shows the percentage of change in pain between the first and second trials when two trials of TENS were administered. The percentage of pain suggests that patients who do not report a reduction in pain from the initial trial will, in general, show a minimal reduction with a second trial. As shown in Figure 2, there was, on average, an increase in pain with a second trial during the several months analyzed. This finding is an example of the database stimulating a clinical hypothesis. For example, a randomized controlled clinical trial evaluating whether a second trial of TENS is truly warranted would appear to be an important study to pursue.

Clinical Management Tool (Decreased Ambulation)

In the second preliminary project, 1,371 patients treated over a 1-year period (November 1992-October 1993) for decreased ambulation ability (indicating they could not walk independently) were quickly retrieved via the database to determine the average change in the amount of assistance needed between their initial evaluation and their final discharge evaluation. This project was used to demonstrate the frequency that therapists chose the problem "decreased ability to ambulate" over a 1-year period and to demonstrate the large number of patients that could be routinely retrieved from a database organized by physical therapy problems. In addition, general trends depicting the relative change in level of assistance between the initial evaluation and discharge from physical therapy could be determined.

Figure 3 shows the average change in the level of assistance provided for the 1,371 patients seen for decreased ambulation ability over the period November 1992 through October 1993. These data suggest that, on average, the patients required moderate assistance at the initial evaluation but only minimal assistance by discharge. This is an example of how immediate feedback of descriptive trends can be quickly determined via a database. These data can be further sorted by patient type, patients that required no assistance at discharge, therapists, patient demographics, or other comorbidity factors.

Clinical Measurement Development (Ambulation Distance)

In the third preliminary project, 638 patients evaluated for decreased ambulation ability were retrieved for a 6-month period (May 1993-October 1993). Two hundred ninety-six of these patients were seen for decreased ambulation due to a musculoskeletal problem, 243 patients were seen for decreased ambulation due to a neuromuscular problem, and 99 patients were seen for decreased ambulation due to a cardiopulmonary problem. The average distance that therapists from these respective specialties walked their patients before discharge was assessed.

Table 3 shows how far patients ambulated before discharge when their gait problems were predominantly caused by musculoskeletal problems, cardiopulmonary problems, and neuromuscular problems. These data, at the very least, suggest differences in distances walked for patients with these various problems. This information provided the authors with some guidance in establishing a standard distance required to test ambulation for various patients being seen in these specialty areas. In addition, specific practice patterns within each specialty division were better defined.

Table 3. Descriptive Statistics for
Distance Walked in Patients With
Musculoskeletal, Neuromuscular, and
Cardiopulmonary Causes for Decreased
Ambulation
             Musculo-    Neuro-        Cardio-
             skeletal    muscular      pulmonary
N            296         243           99
[bar]X(a)    45.37       109.46        50.66
SD(a)        37 07       70.68         32.10
(a) In feet (1 ft=0.3048 m).

Pilot Research (Patient Sample Variability)

Finally, in the last example, 43 individuals receiving a total knee replacement over the period November 1992 through October 1993 were retrieved from the database for their knee flexion range of motion initially and at discharge. Given an estimate of their variability at a known power, the sample size needed to detect an important clinical change was estimated to assist in designing a randomized clinical trial.

The average initial knee flexion range of motion was 65 degrees (SD= 17[degrees]), whereas the average discharge range of motion was 83 degrees (SD = 11[degrees]). Given these estimates of variability and a clinically important change in knee flexion of 10 degrees, approximately 70 patients would be considered necessary to detect this 10-degree change with a power of .80 and a .05 probability of a Type I error. Through the database, a source of valuable information is available when designing randomized clinical trials. The database also provides an efficient projection of the patients available within an institution when proposing clinical research trials.

Discussion

Implications for the Physical Therapy Computerized Database

The PTCMR has been operational since August 1991. All staff (37 therapists) have been in full compliance with the computerized system. We believe this compliance is particularly due to the therapists' involvement during the developmental stages. Studies indicate that the computerized physical therapy record takes approximately 30% less time for providing documentation than did the previously handwritten notes (Shields RK, Leo KC, Miller B; unpublished report; 1989). In addition, legible notes and discharge summaries assist third-party payers in comprehending patient status. Our database information provides routine administrative information regarding therapists' work load, billing information, and length of hospital stay and helps prioritize the problems therapists most frequently select for physical therapy intervention.

The system, however, is in its infancy, frequently undergoing scale and definition changes as dictated from our research studying data quality. This raises an important concern regarding the future of physical therapy database information. With the growth in microcomputer storage and retrieval capabilities, we believe physical therapy clinical practice databases will soon be the norm instead of the exception. Summary data for measures of impairment, functional status, or disability are only as good as the quality of the incoming measures. Consequently, we must not assume objectivity because a computer stores and summarizes numerical data. A database without structured checks of the data quality raises concern about the credibility of the data.

Functional Measurements Within the Database

Five activities graded with the University of Iowa Level of Assistance Scale appeared to have good between- and within-tester reliability in studies completed thus far.[25] In addition, these measurements appear to be valid and responsive for patients with total joint replacement.[25]

Two important issues regarding data reliability with respect to databases warrant discussion. First, it is extremely difficult to blind therapists (and patients) when repeated tests are performed on the same patients during between-therapist reliability studies. Consequently, many reliability studies may not reflect the true error that may exist during the routine administration of tests and measures in physical therapy (clinical practice database). Measurements obtained with repeated tests, however, do provide an index of measurement reliability under ideal conditions, which, if poor, suggests that the reliability of these tests during routine clinical assessments may also be unsatisfactory.

We discovered that therapists using the University of Iowa Level of Assistance Scale as a test in a reliability study appeared to provide less assistance when assessing a task than when they used the scale during routine clinical application. Support for this notion surfaced when we discovered the average length of hospital stay was 6.4 days (SD = 1.4) for patients with a total hip replacement (receiving standard physical therapy) participating in a randomized clinical trial. However, patients retrieved through the database who met the study criteria (treated by the same therapists), but who did not participate in the trial, were hospitalized on average 9.7 days (SD=2.5). These findings suggest that therapists apply these functional tests differently when used for a study as compared with routine clinical practice.

One explanation for this discrepancy is that during routine clinical practice, therapists are not merely instruments assessing a given function, but instead weigh many factors (psychological, emotional, physiological) when deciding how much assistance to provide a patient for a given task. Additionally, some therapists may decide prematurely to provide a certain amount of assistance instead of testing how much the patient can contribute to the task before providing assistance. Because we discovered this discrepancy between our routine clinical tests and our research clinical tests, our department has adopted the philosophy that we strive to provide the least amount of assistance possible during a defined clinical test to maximize finding the true capabilities of the patient. This level of assistance is in contrast to the amount of assistance a therapist may provide during a given treatment. We believe disciplined application of standardized clinical tests, whether using our scale (University of Iowa Level of Assistance Scale) or others (FIM),[16,17] is essential to the future success of outcome research using a clinical practice database.

Second, because repeated measures must be completed on the same patient during interrater reliability studies, some error may be attributed to true change in patient status (fatigue or learning) between two trials. This may be especially true with patients in acute care hospitals. Thus, the error may be overestimated during repeated reliability tests involving patients. This specific issue has been studied and preliminary data have been presented elsewhere.25 A full discussion of this issue will be published in a separate report.

Studies addressing measurement reliability, validity, and responsiveness must be an integral part of a physical therapy clinical practice database. Other measurement issues such as using arithmetic functions on ordinal data, item scaling, and rater and test item calibration may need to be addressed using methods of test development theory such as Rasch analysis.[23]

Database Demonstration Projects

Large quantities of data can be quickly accessed and summarized from the database. The projects demonstrate that the database can assist in generating clinical hypotheses, describing outcome trends, and estimating patient variability. The DCD successfully generated risk profiles that accurately predicted patients at risk for a myocardial infarction.[1,5] Patient profiles established from a clinical physical therapy database may someday assist in accurately predicting patients who will benefit from physical therapy intervention.

Our preliminary data suggests that patients seen in an acute care setting for decreased ambulation abilities on average improved in how much assistance they required to ambulate. Patient change, however, cannot automatically be assumed to be attributable to the physical therapy intervention that is being administered. This is especially true in the acute care population, in which spontaneous recovery is expected at the same time that the patient is treated by the physical therapist (ie, for stroke, total joint replacements, and so forth) (see article by Kane in this issue).

Database Methodological Issues

Database methodological concerns for research are frequently summarized under the categories of data issues[2-5] and bias issues.[2-4] The major data issue that plagues clinical databases is that of missing values.[2-5] Missing values can occur by error or can be attributed to the fact that the patient is better and has not returned or the patient is being seen elsewhere. The issue of bias suggests that patient treatments are selected based on an undefined characteristic of the patient. Byar[4] argues that the lack of standard definitions for diagnoses and the lack of standard treatment approaches across medical practitioners limits the ability to assess treatments, via a database, due to bias.

Consequently, there continues to be considerable controversy regarding the use of clinical databases in the determination of treatment efficacy.[2-7] There is little question, however, that the future of health care has and will continue to be strongly influenced by databases developed by large reimbursement agencies as well as by clinical practice databases in other medical specialties.[1-5] Health care reform requires answers today regarding the effectiveness of our physical therapy interventions. Through the use of high-quality clinical practice databases, large volumes of information can be summarized in relatively short periods of time, which increases the attractiveness of this alternative research method. Conversely, randomized clinical trials are extremely expensive and time consuming and have been accused of assessing treatment efficacy and not treatment effectiveness.[4] physical therapy practice database, however, cannot survive without appropriately timed randomized clinical trials. The physical therapy practice database merely provides an additional dimension that can assist the overall research mission. When treatment effectiveness studies are pursued within a database, randomized clinical trials should be developed to further elaborate on the database findings.[2]

The combined effort of developing a physical therapy clinical practice database in association with studies of measurement integrity is a difficult, but important, undertaking. Through these efforts, our hope is that the meaningfulness of health status surveys, functional assessments, and measures of impairment can be systematically delineated. Standard methods of determining patient outcome from physical therapy intervention will be an integral part of the future health care system. The clinical practice database is just one of many important methods needed to develop standards to better measure physical therapy (disability) outcomes.

Acknowledgments

We recognize the assistance provided by Lori Enloe and Kent Smith and the Clinical Standards of Practice Committee members during the development of this system. We also recognize the invaluable support provided by the University of Iowa Hospitals and Clinics physical therapy staff and the University of Iowa Hospital Informational Services Department.

References

[1] Rosati RA, McNeer JF. A new information system for medical practice. Arch Intern Med. 1975;135:1017-1024. [2] Pryor DB, Kerry LL. Methods for the analysis and assessment of clinical data bases: the clinician's perspective. Stat Med. 1991; 10:617-628. [3] Hlatky MA. Using data bases to evaluate therapy. Stat Med. 1991; 10:647-652. [4] Byar DP. Problems with using observational data bases to compare treatment. Stat Med. 1991;10:663-666. [5] Tierney WM, McDonald CJ. Practice data bases and their uses in clinical research. Stat Med. 1991;10:541-557. [6] Fries JF, McShane DJ. ARAMIS: a prototypical national chronic-disease databank in medical information. West J Med. 1986;145:798-804. [7] Moses LE. Innovative methodologies for research using data bases. Stat Med. 1991; 10:629-633. [8] Safran C. Using routinely collected data for clinical research. Stat Med. 1991;10:559-564. [9] Enloe LJ, Shields RK. Standardization of physical therapy for patients with total hip and knee arthroplasty. Phys Ther. 1992;72:58. Abstract. [10] Farer S. Outcome analysis for program service management. In: Furher MJ, ed. Rehabilitation Outcomes Analysts and Measurement. Baltimore, Md: Paul H Brookes Publishing Co Inc; 1984:119-121. [11] Anderson JP, Kaplan RM, Berry CC, Bush JW. Interday reliability of function assessment for a health status measure: the quality of well being scale. Med Care. 1989;27:1076-1084. [12] Katz JN, Larson MG, Phillips CB, Fossel AH. Comparative measurement sensitivity of short and longer health status instruments. Med Care. 1992;30:917-925. [13] Bergner M, Bobbit RA, Carter WB, Gilson BS. The sickness impact profile: development and final revision of a health status measure. Med Care. 1981;19:787-805. [14] Ware JE, Wherbourne CD. The MOS 36-item short form health survey (SF-36): conceptual framework and item selection. Med Care. 1992;30:473-483. [15] Jette AM, Cleary PD. Functional disability assessment. Phys Ther. 1987;67:1854. [16] Keith RA, Granger CV. The functional independence measure: a new tool for rehabilitation. In: Eisenberg MG, Gryesiak RC, eds. Advances in Clinical Rehabilitation. New York, NY: Springer Publishing Co Inc; 1987:6-18. [17] Granger CV, Hamilton BB. The uniform data system for medical rehabilitation: report of first admissions for 1991. Am J Phys Med Rehabil. 1993;72:33-38. [18] Granger CV, Cotter AC, Hamilton RB, Fiedler RC. Functional assessment scales: a study of persons after stroke. Arch Phys Med Rehabil. 1993;74:133-138. [19] Granger CV, Hamilton BB. Measurement of stroke rehabilitation outcome in the 1980s. Stroke. 1990;21:1146-1147. [20] Hamilton BB, Laughlin JA, Granger CV. Interrater agreement of the seven level functional independence measure (FIM). Arch Phys Med Rehabil, 1991;72:790. Abstract. [21] Gresham GE, Labi ML. Functional assessment instruments currently available for documenting outcomes in rehabilitation medicine. In: Granger CV, Gresham GE, eds. Functional Asset in Rehabilitation Medicine Baltimore, Md: Williams & Wilkins; 1984:65-85. [22] Roach KE, Van Dillen LR. Development of an acute care index of functional status for patients with neurologic impairment. Phys Ther. 1988;68:1102-1108. [23] Heinemann AW, Linacre JM, Wright BD, et al. Relationships between impairment and physical disability as measured by the Functional Independence Measure. Arch Phys Med Rehabil. 1993;74:566-573. [24] Shields RK, Dostal WF. Inter-rater reliability of the standing test in neurologically impaired patients. In: Abstracts of Platform and Poster Presentations, June 27, 1990; Anaheim, Calif. Alexandria, Va: American Physical Therapy Association; 1990:99. [25] Shields RK, Enloe LJ, Evans R, et al. Analysis of the reliability of clinical functional tests in total hip replacement patients. Phys Ther. 1992; 72:S113. [26] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1970;86:420-428. [27] Guyatt GH, Walter S, Norman G. Measuring change over time: assessing the usefulness of evaluative instruments. J Chronic Dis. 1987;40: 171-178,

RK Shields, PhD, PT, is Assistant Professor and Clinical Research Coordinator, Physical Therapy Graduate Program, College of Medicine, The University of Iowa and University of Iowa Hospitals and Clinics, 2600 Steindler Bldg, Iowa City, IA 52242 (USA). Address all correspondence to Dr Shields.

KC Leo, PT, is Director, Department of Physical Therapy, University of Iowa Hospitals and Clinics.

B Miller, PT, is Supervisor,. Department of Physical Therapy, University of Iowa Hospitals and Clinics.

WF Dostal, PhD, PT, is Supervisor, Department of Physical Therapy, University of Iowa Hospitals and Clinics.

R Barr, PT, is Supervisor, Department of Physical Therapy, University of Iowa Hospitals and Clinic

Major portions of this project were part of the Clinical Research Center grant funded by the Found for Physical Therapy Inc.