Physical therapy and health outcomes in patients with spinal impairments.

Key Words: Back pain, Neck pain, Outcome and process assessment.

Musculoskeletal impairments account for a large percentage of conditions for which medical care is sought in the United States,[1] and a large proportion of the cost associated with this care is related to the disability caused by these impairments. In 1984, Cunningham and Kelsey[2] reported the overall prevalence of musculoskeletal impairments using data from the National Health and Nutrition Examination Survey (NHANES) I. In the United States, 32.6% of persons between the ages of 25 and 74 years were affected by some type of physician-observed musculoskeletal impairment, and 29.7% of the population had self-reported musculoskeletal impairments. Impairments related to the spine had the highest prevalence. In the National Medical Care Utilization and Expenditure Survey (NMCUES), about 20% of the 1980 noninstitutionalized population reported having a musculoskeletal problem involving the back or joints that resulted in some type of disability or use of the health care system.[3]

Conservative treatment of musculoskeletal impairments often includes physical therapy, and there may be a trend toward an increase in the use of physical therapy services.[4] Of all patients discharged from outpatient physical therapy practices in the United States, 25% were patients for whom low back pain was the primary problem, and a large majority of the remaining patients sought care for other musculoskeletal impairments.[5] Deyo and Tsui-Wu[6] reported that 40.5% of subjects with low back pain identified from NHANES II had used "exercises or physical therapy" for treatment. Other common interventions provided by physical therapists were traction, which had been used by 20.7% of the subjects, and diathermy or paraffin, which had been used by 16.7% of the subjects. The NMCUES indicated that 13.3% of total charges for treatment of musculoskeletal conditions was attributed to care given by health care professionals other than physicians, including physical therapists.3 The cost of health care related to the treatment of musculoskeletal impairments accounted for 8% of total health care expenditures in 1980, ranking third among health problems in terms of costs for civilian noninstitutionalized individuals.[3]

Little is known concerning the relationship of specific health care interventions to disability. Deyo and Diehl[7] found that demographic, psychosocial, and baseline health status, rather than impairment status or intervention, had an effect on functional outcomes in patients with low back pain. Although physical therapy is a common intervention and is often initiated with the belief that disability can be reduced or prevented, there is little evidence that suggests any particular treatment approach produces better outcomes than another. The recently published Acute Low Back Pain Problems in Adults: Clinical Practice Guideline No. 14,[8] developed by the federal Agency for Health Care Policy and Research (AHCPR), cites as "not recommended" several interventions commonly administered by physical therapists, including heat and cold modalities and stretching exercises for the low back. Other physical therapy interventions such as "low-stress aerobics" were recommended, noting, however, the limited research-based evidence for their inclusion.

The general goal of this study was to improve understanding of the pattern and magnitude of treatment outcomes following physical therapy for patients with spinal impairments. The specific purpose of the study was to examine the relationship of patient outcomes over a physical therapy episode of care to the characteristics of that episode of care in terms of the type and duration of treatment selected by the physical therapist, controlling for characteristics of the patient.

Method

Subjects

The data for the study were derived from a large database generated by the Focus on Therapeutic Outcomes (FOTO) network. The FOTO network was a privately funded consortium of five outpatient rehabilitation companies. The FOTO network was developed for the purpose of generating an outcome-oriented, standardized information management system for use in outpatient physical therapy settings. Companies were required to maintain an agreed-on data quality standard and record completion rate to remain in the FOTO network. The database contained information from 3,994 patients admitted for physical therapy care for lumbar and cervical impairments during a 1-year period beginning in July 1993. The sample for this study was 1,097 patients who filled out the requisite initial and discharge health status questionnaires and had a completed episode of care. The remaining patients either failed to make a final visit to allow for follow-up data collection or did not complete follow-up questionnaires at the final visit. These patients were treated in any of 68 physical therapy practices by 141 physical therapists. The project was approved by the Human Subjects Review Board of New England Research Institute, which was responsible for the design and administration of the database in 1993 and 1994.

Practices. The practices were located across the United States: 2% from the middle Atlantic states, 20% from the south Atlantic states, 9% from the south central states, 62% from the north central states, 7% from the mountain states, and less than 1% from the Pacific states. The practices ranged in size, with 24% having one full-time physical therapist, 38% having two full-time physical therapists, 15% having three full-time physical therapists, and 23% having more than three full-time physical therapists.

Physical therapists. The average age of the physical therapists was 32.6 years (SD=7.8, range=22-60). Seventy percent of the physical therapists were women. The highest credential of 84% of the physical therapists was a bachelor's degree. Nine percent of the physical therapists had an entry-level master's degree, and 7% of the therapists had an advanced master's degree. The average years of practice was 8.4 (SD=7.4, range= 1-33). Eighty-three percent of the therapists worked full time, treating an average of 50 patients per week.

Patients. Sixty-seven percent of the patients had impairments of the lumbar spine, and 33% of the patients had impairments of the cervical spine. Differences between patients with complete follow-up data and those with incomplete data were determined using analysis of variance. Patients with complete follow-up data were slightly older (41.1 years versus 40.0 years, F=5.12, P=.024) than those without follow-up data. Patients with incomplete data had lower initial health status in six health outcomes scales (energy/fatigue: 43.4 versus 46.1, F=11.45, P=.0007; general health perceptions: 67.4 versus 69.0, F=4.56, P=.033; mental health: 66.3 versus 69.3, F=17.55, P=.0001; physical function: 52.3 versus 55.4, F=9.97, P=.002; role limitation-emotional: 61.5 versus 65.8, F=7.78, P=.005; and social function: 54.1 versus 57.5, F=10.43, P=.001). Table 1 shows the characteristics of the patients by primary impairment.

Table 1.

Demographic and Financial Characteristics of Patients With Vertebral Impairments

                             Lumbar          Cervical
                             Impairment      Impairment
Characteristic               (n=739)         (n=358)
Age (y)
  [bar]X                     41.2            42.0
  SD                         13.9            13.4
Gender
  Female                     46%             67%
  Male                       53%             33%
Ethnicity
  White                      87%             83%
  Black                      10%             11%
  Native American            <1%             <1%
  Asian                      <1%              1%
  Hispanic                    2%              4%
Employment
  Full time                  34%             49%
  Light duty                 13%             10%
  Off because of health      35%             16%
  Retired                     8%             11%
  Unemployed                 10%             13%
Income
  <$15,000                   10%              8%
  $15,000-$25,000            18%             17%
  $26,000-$35,000            16%             22%
  $36,000-$45,000            15%             14%
  >$45,000                   26%             25%
  Refused to answer          15%             14%
Acuity
  Acute                      21%             19%
  Subacute                   49%             53%
  Chronic                    30%             28%
Surgery
  Yes                        15%              7%
  No                         85%             93%
Depressed
  Yes                        31%             31%
  No                         69%             69%
Comorbidities
  None                       58%             55%
  1 category                 27%             30%
  2 categories               13%             13%
  More than 2 categories     2%               2%

Treatments. Table 2 shows the percentages of patients receiving various physical therapy treatments over the episode of care. Ninety-six percent of the episodes included various combinations of these treatments. The most frequent combination was flexibility exercise, strength exercise, and heat.

Table 2.

Characteristics of Treatments(a) and Episode of Care for Patients With Spinal Impairments

                             Lumbar          Cervical
                             Impairment     Impairment
Characteristic                (n=739)        (n=358)
Manipulation/mobilization      39%             61%
Flexibility exercises          84%             81%
Strengthening exercises        81%             60%
Endurance exercises            52%             36%
Massage techniques             39%             65%
Heat modalities                81%             91%
Cold modalities                19%             12%
Episode length (d}
  [bar]X                       28.1            29.7
  SD                           21.7            23.5
  Range                        1-182           1-154

(a) Percentage of patients receiving this type of treatment.

Data Collection

Data for each patient in the database were obtained by the primary practitioner caring for the patient and from the patient and included age, gender, ethnicity, height, weight, educational level, income, employment status, comorbidity, duration of the spinal problem, surgical history, duration of the episode of care, and type of treatments provided. In addition, each patient completed the standard form of the Medical Outcome Study 36-Item Short Form Health Survey (SF-36)[9] and a disease-specific health outcomes questionnaire (Oswestry Low Back Pain Disability Index[10] or Neck Disability Index[11]) at initiation and completion of physical therapy. These questionnaires provided the major indicators of health outcome.

The SF-36 queries the patient concerning health over the past 4 weeks in eight different health outcome dimensions: energy/fatigue, general health perception, mental health, bodily pain, physical functioning, role limitation due to emotional problems, role limitation due to physical problems, and social functioning.[9] Acceptable reliability and validity of the SF-36 have been reported for its use in aggregate analyses.[12-15] The Oswestry Low Back Pain Disability Index is used to assess the status of patients with lumbar spine impairments.[10] The reliability and validity of measurements obtained with this instrument have been reported.[10] Ten areas are assessed: pain, personal care, lifting, walking, sitting, standing, sleeping, sex life, social life, and traveling. The Neck Disability Index is used to assess the status of patients with impairment of the cervical spine.[11] Ten areas are assessed: pain, personal care, lifting, reading, headaches, concentration, work, driving, sleeping, and recreation. Vernon and Moir[11] have reported on the reliability and validity of scores obtained with the instrument. For each of the SF-36 scales, responses to the questions were summarized and then transformed to provide scores ranging from 0 to 100, with 100 being the best possible score. The SF-36 provides a profile of health, with a score for each of the eight health dimensions. Each impairment scale provides one comprehensive score, with 0 being the best score and 100 being the worst score.

Data Analyses

The pattern of health status for patients with spinal impairments beginning physical therapy care was examined by comparing gender- and age-controlled SF-36 scores of patients at the time of their initial visit with the population norms. Norms were derived from a sample of 2,474 noninstitutionalized adults residing in the United States who responded to the National Survey of Functional Health Status in 1990.[16] The comparisons were made by computing standard scores, which were derived by subtracting scores for patients on each of the scales from the mean general population scores and dividing the results by the standard deviations of the general population scores.[14] Comparing scores in this way allows comparisons regarding the patterns of health status across groups with various demographic characteristics and health conditions.

The pattern and magnitude of outcomes for patients with spinal impairments were estimated by computing effect sizes for each of the SF-36 scales and the disease-specific scales. Effect size represents change in a scale as it relates to the standard deviation of scores for that scale, and it is computed by subtracting the initial score from the final score and dividing the result by the standard deviation for the initial score.[17,18] Demonstrating change in this way provides a standardized measurement of change to aid interpretation. The magnitude of change is described and, because it is standardized, can be compared with changes in other scales, with benchmarks that signify clinically meaningful changes, or with changes obtained with another treatment regimen or in another group of patients. Cohen[18] suggested the following interpretation of effect size: 0.2 to 0.4 is small, 0.5 to 0.7 is moderate, and 0.8 or greater is large.

To examine the relationship of the characteristics of the physical therapy episode of care to health outcomes, nine sets of multivariate analyses were conducted for each type of impairment, with follow-up scores on the SF-36 and disease-specific scales as the dependent variables and baseline scores as covariates. The major independent variables of interest consisted of treatment characteristics. In addition, relevant characteristics of the patient were included to control for potential confounding. Table 3 describes the variables in detail.

Initially, univariate analyses were used to determine possible confounding variables among the patient characteristics. Then, using general linear models and a backward deletion process, treatment variables were determined for models that included baseline scores and those patient variables that had a significant univariate correlation to the outcomes of interest. This is an interative process that begins with a regression model containing all independent variables. Subsequent steps involve decisions to retain or delete variables based on their contribution to the model and comparison of the Y partial F statistic with a preselected critical value. This multivariate analysis allows variation in the dependent variable to be attributed to a combination of independent variables, both continuous and categorical, and allows control for the baseline value of the dependent variable in a "pre-post measurement" design. An alpha level of.01 was used to determine the patient variables to include as possible confounders. An alpha level of .05 was used as the criterion for other variables to remain in a model controlled for baseline health status and the potentially confounding variables. Only data of patients with complete data for the independent variables of interest were included in the analyses. Due to some missing data for some variables, the number included in each model varied. All analyses were performed using SAS software.([dagger])

Results

The SF-36 can be seen as measuring two basic components of health: physical and emotional.[19] Patients with lumbar and cervical impairments during an initial physical therapy visit reported poorer physical and emotional health than did the general population (Fig. 1). The areas most notably affected were bodily pain and role limitation due to physical problems. In each scale, scores were nearly 2 standard deviations below scores for the general population. Patients with lumbar impairment also demonstrated striking problems with physical function ( 1.5 standard deviations below the norm). The pattern of loss of health was similar for patients with cervical and lumbar impairments except in the area of physical function, where lumbar impairment appeared to have a greater effect. The patient's age, gender, education, employment status, income, problem acuity, comorbid status, body mass index (BMI), surgical history, and depression were determinants of outcome in both physical and psychological health domains.

Health improved in the disease-specific scales and in all other areas except general health perceptions over a course of physical therapy care, which lasted, on average, 28 days (Fig. 2). Changes occurred most notably in the physical dimension of health as measured by bodily pain (approximately 0.8 standard deviation) and the disease-specific scales (approximately 0.7 standard deviation) for both lumbar and cervical impairments. Patients with lumbar impairments also demonstrated important changes in physical functioning (approximately 0.7 standard deviation). Patients with lumbar and cervical impairments demonstrated similar patterns of improvement except in physical functioning.

Controlling for relevant patient characteristics and baseline health, characteristics of the physical therapy episode of care were associated with outcome in six of the nine outcomes for patients with lumbar impairment and in four of the nine outcomes for patients with cervical impairments. Of all the treatments examined (Tab. 2), the use of treatments that included endurance exercise predicted better outcome most often in scales measuring both the physical and emotional aspects of health. Better outcome was associated with endurance exercise in energy/fatigue for patients with each type of spinal impairment, in physical function and social function for patients with lumbar impairments, and in general health perceptions for patients with cervical impairments. Treatment with mobilization or manipulation was related to better outcomes in general health perceptions and in the Neck Disability Index for patients with cervical impairments. Patients with cervical impairments also showed improved outcome in physical functioning with the inclusion of flexibility exercises and in the Neck Disability Index with inclusion of strength exercises as part of the treatment. For those patients with lumbar impairments, the inclusion of heat or cold modalities was associated with poorer outcomes in five scales.

The factors explaining better outcomes in the various health scales are presented in Tables 4 and 5. Significant models were generated for all scales ([R.sup.2]=.24-.56 and .23-.60 for lumbar and cervical impairments, respectively). In all cases, most of the variability in outcome was accounted for by baseline scores.

[TABULAR DATA 4 OMITTED]

Discussion

This study adds to our knowledge concerning the health effects of problems involving the spine and demonstrates the pattern and magnitude of outcomes achieved over a course of physical therapy. We believe that this is the first study to comprehensively examine whether the types of treatments provided by physical therapists are associated with outcomes of patients undergoing physical therapy. Although the majority of patients in our study had treatment that involved multiple approaches, we were able to show that outcomes were associated with the use of some types of treatments.

[TABULAR DATA 5 OMITTED]

Use of endurance exercises was the most consistent predictor of better outcome. The effect seemed to be in both the physical and emotional health dimensions. These results are supported by the findings of Lindstrom et al,[20] who found that a group of workers with low back pain who participated in an individualized exercise program that included some form of endurance exercise returned to work more quickly and had less long-term sick leave than a control group who did not have exercise instruction. Our findings also provide support for the AHCPR guideline for acute low back pain, which recommends endurance programs and low-stress aerobics.[8] Inclusion of endurance exercises in a physical therapy regimen may convince patients to try physical activities that they have avoided and may improve their aerobic capacity, allowing them to perform activities with less perceived exertion. Endurance exercise may also reduce sensitivity to pain, increase blood flow to painful muscles, and increase endorphin levels. In addition to endurance exercise, other treatments, including manipulation or mobilization, strengthening exercises, and flexibility exercises, are related to better outcomes for patients with cervical impairments. These findings suggest a multimodal approach to physical therapy care for patients with cervical impairments.

The inclusion of heat and cold modalities in a treatment episode was associated with poorer outcomes and may reflect less time during treatment spent on more active therapies such as exercise. This finding serves as support of the AHCPR guideline, which does not recommend physical agents provided by health care practitioners for the treatment of low back pain.[8]

This study confirms reports of the detrimental effects of low back pain on health.[12,21-23] Health profiles for patients with low back pain similar to those for our sample were reported by Stewart et al[23] using the SF-20 and by Lansky et al[22] and Garratt et al[14] using the SF-36. Deyo and Diehl[21] also showed similar disabilities as measured by the Sickness Impact Profile. In all cases, patients with back pain demonstrated the greatest disability in role functioning. The physical functioning and pain scales for the SF-36 and SF-20 in these studies were also considerably lower than for the general population.

Our findings show that cervical problems also affect role function related to physical health and that pain is a major problem. Unlike patients with lumbar impairment, however, patients with cervical impairment do not experience the same degree of loss of physical functioning. Interestingly, both lumbar and cervical problems appear to affect social functioning.

Over a course of physical therapy care, improvements in health occur in nearly all areas for patients with either type of vertebral problem. The design of the study precludes attributing these changes to the intervention and reported changes may well be due to the natural history of the problem or some characteristic of the patient (eg, patients receiving endurance exercises may be different from those receiving thermal modalities).

A comparison of the changes found in our study with those documented by Garratt et al[24] over a 1-year period and by Lansky et al[22] over a 90-day period provides some insight into the possible effects of history on low back pain problems. Whereas Garratt et al[24] and Lansky et al[22] showed the greatest amount of change in the bodily pain and physical role limitation scales, changes in our study were greatest in the bodily pain and physical functioning scales for patients with lumbar impairments. This discrepancy may suggest that early changes in health occur in physical functioning, with improvement in role functioning occurring later. Improvement in pain may occur early and be sustained. Physical therapists' understanding of pain as a major problem area can be inferred by the fact that reduction of pain was reported as a treatment goal for 90% of the patients treated by physical therapists for low back pain.[5] Goals related to physical function, role function, and social function, however, were not identified in that study. In spite of this lack of reported physical therapy goals for improved physical role and social functioning, such improvement appears to occur over a physical therapy episode of care. The improvement in function may be a reflection of implicit, rather than explicit, goals set with the physical therapist. Our study suggests a need for physical therapists to reconceptualize the scope of the goals set with the patient and to make functional goals more explicit.

Our findings also show that the degree of change in disease-specific scale scores is similar to the degree of change in pain scores of the SF-36 over the course of treatment. The degree of change in the Oswestry Low Back Pain Disability Index is also similar to that found in the physical functioning scale of the SF-36 and reflects their measurement of similar functions. The Neck Disability Index is not quite so clearly a measure of one aspect of health. These findings argue for the use of a generic health instrument, as it provides a more comprehensive measure of health than the disease-specific scales, with apparently similar responsiveness to change in some of its scales.

Outcomes for patients with vertebral impairments are only minimally to moderately predicted by an array of factors commonly noted during a physical therapy episode of care in spite of the fact that changes are moderate or better in most scales. Harada et al[25] found that they could explain 52% of the variability in functional change for patients with low back injury. These authors, however, forced all predictor variables into the model in a predetermined order. Our approach to the analyses used an elimination process that resulted in models that included only significant variables and, therefore, explained less variance. Deyo and Diehl[7] found that 17% of the variability in pain improvement in patients with low back pain could be accounted for by a model that included three patient-related factors.

According to our results, patients who might be expected to have better outcomes over an episode of physical therapy care for the treatment of either lumbar or cervical spine impairments are young, are not off work because of their health, have less comorbidity, and are not depressed. For patients treated for lumbar impairment, having a high-school diploma is an additional factor associated with better outcomes. Income in the range of $15,000 to $25,000 per year is associated with poorer outcomes in patients with lumbar impairments.

Younger age has been found to be associated with improvement in low back pain outcomes following surgery in at least one previous study.[26] Our finding that better outcome was associated with being younger held true in spite of controlling for comordities. This finding suggests reasons other than the increased rate of comorbidity with aging for older people not faring as well. Research reports consistently suggest that lower educational level has a negative effect on outcomes.[7,27] This finding may be a result of problems understanding and complying with instructions from health care providers or it may be due to the fact that lower education levels are generally associated with employment that requires physical labor that is difficult to resume with a low back impairment. The finding that women had better outcomes than did men is contrary to the reports of others who have found that women fare less well.[26,27] Interestingly, the scales in which women fared better represented the emotional domain of health.

Poor outcomes in both the physical and psychological domains of health were found to be related to being depressed at initiation of treatment. Previous investigations have not directly included a measure of depression in multivariate examinations of changes in health in patients with spinal impairments. Deyo and Diehl,[7] however, found that a report by patients of "always feeling sick" was a consistent predictor of poor outcome in patients with low back pain. They speculated that this variable was a marker for psychological state. In another study, the same authors found that patients with back pain who had psychiatric problems or were worried about serious illness had greater disability.[21]

The relationship of income to outcome in patients with lumbar impairments is most consistent in the emotional domain of health. This relationship may reflect problems associated with coverage for payment of care. Patients in this income bracket are likely to have low-paying employment that does not provide health insurance, yet too much income to qualify for government programs.

Limitations

This study was observational in nature and precludes conclusions about the effectiveness of physical therapy treatments. Patients were not randomized to treatments, and illness behavior may have influenced treatment choices made by the physical therapists. In addition, individual physical therapists indicated whether they had used particular treatment approaches. These approaches, for example, endurance exercise, were broadly defined, and no attempt was made to assess the reliability of therapists' entries into the database. The data also do not account for the subtle changes in patients' conditions over time and the treatment adjustments that occur in response to these changes.

The database used for this study lacked data concerning specific impairments of the musculoskeletal system that might be found in patients with vertebral impairments, for example, pain pattern, flexibility, and muscle strength. In addition, patients' problems were only grossly classified as lumbar or cervical. There is some evidence in the literature that specific impairments are related to disability.[27-29] Yet, other studies have not found any effects of physical examination findings when they were included with other psychosocial variables in multivariate analyses.[7,26]

The findings reported here are based on analysis of a clinical database. These data were generated for the major purposes of quality assurance and business decision making, and using them for predicting outcomes presents limitations. Because of these limitations, this study should be viewed as hypothesis generating rather than hypothesis testing.

The issues related to the use of clinical databases have been clearly outlined by Pryor and Lee[30] and include the use of data to answer questions not determined a priori, missing observations, and other biases that limit the external validity. These biases include selection bias and referral bias. In our database, only approximately 28% of the patients had completed episodes with complete data. Patients failing to complete data collection had poorer initial health status than those completing data collection in six of the outcome scales. If those patients did not complete an episode of care because of failure to improve or worsening, the inclusion of their data in the analysis would result in lower effect sizes and possibly different models showing the association of treatment to outcomes. This example demonstrates the problem that can occur when physical therapists fail to obtain or document data relevant to their practice and treatment of patients. The example also illustrates the necessity of quality control and rigorous auditing practices in designing, implementing, and maintaining databases that are used for any type of research.

In addition, this database included large numbers of variables that needed to be reduced in some way to facilitate interpretation of the predictive models. Such reduction results in the loss of some information and the increased chance of misclassification. Despite reducing variables to a manageable number and selecting variables that measured unique attributes, many predictor variables were examined. The large number of variables, defined in unique ways by the investigators, is likely to result in idiosyncratic models. Further studies are required to test the validity of the predictions.

In spite of these limitations, this study provides information not previously reported concerning the relationship of physical therapy to comprehensive health-related outcomes. It also adds to the body of knowledge concerning prognosis for patients with lumbar and vertebral impairments. Although observational studies using clinical databases cannot provide the same rigorous results as clinical trials concerning the effectiveness of treatments, they can provide valuable information for conceptualizing efficient and effective future trials.

Conclusion

Individuals with vertebral impairments demonstrate decrements in health status in both the physical and emotional dimensions of health. Both lumbar and cervical impairments substantially affect role function related to physical health, and pain is a major problem. Persons with cervical impairments, unlike those with lumbar impairments, do not experience a substantial loss of physical function. Over a course of physical therapy care, improvements in health occur in nearly all areas for those with either type of spinal problem. Physical therapists may, therefore, wish to expand the scope of the goals for treatment set with each patient. Inclusion of endurance exercise and the exclusion of heat or cold modalities in the treatment regimen were associated with better outcomes. Future studies are needed to validate the models reported here and to further examine the effect of physical examination findings and physical therapy interventions on outcomes in patients with spinal impairments.

[Figures 1 to ILLUSTRATION OMITTED]

([dagger]) SAS Institute Inc, SAS Campus Dr, Cary, NC 27513

References

[1] Kelsey JL, White AA, Pastides H, Bisbee GE. The impact of musculoskeletal disorders on the population of the United States. J Bone Joint Surg [Am]. 1979;61 :959-964.

[2] Cunningham LS, Kelsey JL. Epidemiology of musculoskeletal impairments and associated disability. Am J Public Health. 1984;74;574-579.

[3] Murt H, Parsons PE, Harlan WR, et al. Disability, utilization, and costs associated with musculoskeletal conditions. Natl Med Care Util Expend Surv C. 1986;5:1-64.

[4] Frymoyer JW, Cats-Baril WL. An overview of the incidences and costs of low back pain. Orthop Clin North Am. 1991;22:263-271.

[5] Jette AM, Smith K, Haley SM, Davis KD. Physical therapy episodes of care for patients with low back pain. Phys Ther. 1994;74:101-115.

[6] Deyo RA, Tsui-Wu Y. Descriptive epidemiology of low-back pain and its related medical care in the United States. Spine. 1987;12:264-268.

[7] Deyo RA, Dichl A1C Psychosocial predictors of disability in patients with low back pain. J Rheumatol. 1988;15:1557-1564.

[8] Bigos S, Bowyer O, Braen G, et al. Acute Low Back Pain Problems in Adults: Clinical Practice Guideline No. 14. Rockville, Md: Agency for Health Care Policy and Research, Public Health Service, US Department of Health and Human Services; 1994. AHCPR publication 95-0642.

[9] Ware JE, Sherbourne CD. The MOS 36-item short-form health survey (SF-36), 1: conceptual framework and item selection. Med Care. 1992; 30:473-483.

[10] Fairbank J, Couper J, Davies J, O'Brien J. The Oswestry low back pain disability index. Physiotherapy. 1980;66:271-273.

[11] Vernon H, Moir S. The neck disability index: a study of reliability and validity. J Manipulative Physiol Ther. 1991;14:409-415.

[12] Stewart AL, Hays RD, Ware JE. The MOS short-form general health survey. Med Care. 1988;26:724-732.

[13] Jenkinson C, Coulter A, Wright L. Short-form 36 (SF-36) health survey questionnaire: normative data for adults of working age. BMJ. 1993;306:1437-1440.

[14] Garratt AM, Ruta DA, Abdalla Ml, et al. The SF-36 health survey questionnaire: an outcome measure suitable for routine use within the NHS? BMJ. 1993;306:1440-1444.

[15] McHorney CA, Ware JE, Lu R, Sherbourne CD. The MOS 36-item short-form health survey (SF-36), III: tests of data quality, scaling assumptions, and reliability across diverse groups. Med Care. 1994;32: 40-66.

[16] Ware JE, Snow KK, Kosinski M, Gandek B. SF-36 Health Survey Manual and Interpretation Guide. Boston, Mass: The Health Institute, New England Medical Center; 1993.

[17] Kazis LE, Anderson JJ, Meenan RF. Effect sizes for interpreting changes in health status. Med Care. 1989;27:S178-S189.

[18] Cohen J. Statistical Power Analysis for the Behavior Sciences. New York, NY: Academic Press Inc; 1977.

[19] McHorney CA, Ware JE, Raczek AK. The MOS 36-item short-form health survey (SF-36), II: psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care. 1993;31: 247-263.

[20] Lindstrom I, Ohulnd C, Eek C, et al. The effect of graded activity on patients with subacute low back pain: a randomized prospective clinical study with an operant-conditioning behavioral approach. Phys Ther. 1992;72:279-290.

[21] Deyo RA, Dichl AK. Measuring physical and psychosocial function in patients with low-back pain. Spine. 1983;8:635-642.

[22] Lansky D, Butler JBV, Waller FT. Using health status measures in the hospital setting: from acute care to outcomes management. Med Care. 1992;30:MS57-MS73.

[23] Stewart AL, Greenfield S, Hays RD, et al. Functional status and well-being of patients with chronic conditions: results from the medical outcomes study. JAMA. 1989;262:907-913.

[24] Garratt AM, Ruta DA, Abdalla MI, Russell IT. SF-36 health survey questionnaire, II: responsiveness to changes in health status in four common clinical conditions. Quality in Health Care. 1994;3:186-192.

[25] Harada N, Sofaer S, Kominski G. Functional status outcomes in rehabilitation: implications for prospective payment. Med Care. 1993; 31 :345-357.

[26] Little DG, MacDonald D. The use of the percentage change in Oswestry disability index score as an outcome measure in lumbar spinal surgery. Spine. 1994;19:2139-2143.

[27] VonKorff M, Deyo RA, Cherkin D, Barlow W. Back pain in primary care: outcomes at one year. Spine. 1993;18:855-863.

[28] Hellsing AL, Linton SJ, Kalvemark M. A prospective study of patients with acute back and neck pain in Sweden. Phys Ther. 1994;74: 116-128.

[29] Waddell G, Somerville D, Henderson I, Newton M. Objective clinical evaluation of physical impairment in chronic low back pain. Spine. 1992;17:617-628.

[30] Pryor DB, Lee KL. Methods for the analysis and assessment of clinical databases: the clinician's perspective. Stat Med. 1991;10:617628.

Invited Commentary

This article contributes greatly to our understanding of the relationship between physical therapy interventions for neck and low back dysfunction and outcome. A differential effect of various interventions has been clearly demonstrated. This information adds to the existing body of knowledge that clinicians can draw on when selecting interventions for patients with neck or low back pain. In addition, it provides important information for planning and implementing prospective clinical trials to further elucidate the effects of our interventions in these patient groups.

With respect to the results of the study, it is encouraging that there are differences in patient outcomes that appear to be related to the intervention selected. It is evidence that, in general terms, we can add physical therapy to the list of factors known to affect outcome, such as educational level, income level, work status, and depression, as reported in this article and elsewhere.! Y Jette and Jette report that the inclusion of endurance exercises in the intervention was the most consistent predictor of better outcome for both the cervical and lumbar impairment groups. As noted by the authors, this finding is supported by other studies.[3,4] This finding suggests that physical therapy for patients with low back or neck pain should include at least a component of endurance exercise. In my opinion, an intensive exercise approach to the rehabilitation of patients with spinal problems, which includes an aerobic exercise component, represents a substantial shift from current clinical practice. There is mounting evidence supporting this approach that should not be ignored.

The authors have clearly outlined the limitations of the retrospective nature of the research design. Possible confounding variables among the patient characteristics, such as age, were controlled for in the multivariate analysis. As pointed out by the authors, one must bear in mind that the results may be influenced by factors that were not controlled and that affect treatment choices and application. It is possible that an unknown factor that affects outcome in one treatment group may be responsible, at least in part, for the results. For example, differing therapist approaches in explaining and implementing endurance exercises could account, in part, for the results in studies such as this one. The positive effect of enthusiastic endorsement of an intervention by the caregiver has been documented in the literature.[5] A difference in approach such as would occur if endurance exercises were presented in a more enthusiastic manner than heat or cold application were presented could affect the results. It is recognized that the methodology used precludes control of all potential factors that could account for the differences in outcome related to type of intervention. It is mentioned here only to highlight the difficulty in controlling all variables that may contribute to the findings presented here and in other studies, particularly in the area of low back pain.

The difficulty in reliably categorizing patients with low back pain into homogenous subgroups is well documented.[6-9] Differential treatment outcomes have been reported, however, related to diagnostic classification.[10,11] Is it possible that at least some of the differences noted between interventions can be accounted for by differences in the assignment of patients to treatment groups based on diagnosis? The interaction between diagnostic category and intervention will be an interesting area for further investigation.

This study provides researchers and clinicians with important data with respect to SF-36 and Oswestry Low Back Pain Disability Index and Neck Disability Index scores in a large sample of patients. In addition, information on sensitivity to change of these scales is reported using an effect-size calculation. Conclusions regarding relative sensitivity to change of these health status scales using effect size should be made with several caveats. As the design was a single-group before-after design and there was no external measure of change, one does not know whether the scales were measuring true, or valid, change or whether the scores would remain stable in patients who really did not change.[12] In addition, differences in the standard deviations of the initial scores among the three measures alter the effect-size calculation.[13] Both of the disease-specific Oswestry scales (low back and neck) and two dimensions of the SF-36, physical function and bodily pain, were found to be measures that demonstrated a large effect size over the course of treatment. The effect sizes of the other SF-36 subscales were not reported. The finding related to the Oswestry Low Back Pain Disability Index concurs with earlier reports of acceptable sensitivity to change.[14]

Individual questions and subscales that constitute larger health status measures have been shown to differ with respect to their capability of measuring change in an outpatient population.[15,16] Patrick et al[16] reported on the sensitivity to change of a modification of the Roland Morris low back functional scale and the SF-36 in a group of patients with low back pain. The modified Roland-Morris scale and SF-36 physical and pain dimension scores demonstrated the greatest sensitivity to change using an effect-size calculation, Guyatt's responsiveness index, and correlation with self-rated improvement. The other dimensions, including mental health, role-emotional, social function, and vitality scales of the SF-36, were substantially less responsive to change over time.[16] Ruta and colleagues[17] reported sensitivity to change for a condition-specific low back functional scale and the SF-36 in a group of patients with low back pain who reported to be worse at follow-up testing. Standardized response means indicated greater sensitivity to change for the condition-specific scale than for any of the SF-36 scales.[17] My interpretation of these findings is that although generic health status scales are important measures of overall health, there are limitations to generic scales when the goal is to measure change in an outpatient population with relatively low levels of disability. Inclusion of questions not germane to patients' functional problems, such as those forming the mental health and role-emotional dimensions, appears to render the overall scale less sensitive to change.

The authors suggest that the finding that two subscales of the SF-36 and both of the condition-specific Oswestry scales demonstrate similar sensitivity to change argues for the use of generic scales over condition-specific scales. My interpretation of this finding is somewhat different. That only two of the eight dimensions of the SF-36 are sensitive to change in persons with low back pain is of concern if the SF-36 is used in its entirety to document outcome in this population. The inclusion of the dimensions that we know are not sensitive to change will reduce the sensitivity to change of the instrument as a whole and make interpretation of the results difficult. The finding by these authors indicates that either the condition-specific questionnaire or only two of the SF-36 subscales should be used for evaluating outcome in patients with low back or neck pain. The use of only the bodily pain and physical function scales of the SF-36 transform the measure into a scale that resembles a condition-specific measure. Scales that are condition-specific and even patient-specific and that include functional and health items most relevant to the individual patient or population under question will be most sensitive to change.[13] Patient-specific health status mea-sures, in which patients identify their own functional limitations, have been shown to be equal to or more sensitive to change than even condition-specific scales.[19,20] I would suggest that carefully constructed condition-specific scales that include items meaningful to the population to which they are administered will demonstrate better sensitivity to valid change in an outpatient population than will generic health status measures.

There is a wealth of information contained in this database, and we will, no doubt, see subsequent publications pertaining to this database from this research team. It would be interesting to know summary statistics to assist with the following: (1) planning sample size for subsequent studies, (2) comparing the standard deviation of initial scores to determine implications for effect size calculation, and (3) comparing initial and follow-up mean scores and standard deviations with the results of future work with the SF-36 in this population. In addition, it would be interesting to know whether interaction terms were included in the initial regression model and whether treatment x covariate interactions occurred. This information may provide further insight into matching treatment to patients.

This article provides important information for a large sample of patients regarding the relationship between intervention and outcome. It is an innovative and efficient manner in which to begin to examine whether the types of interventions provided by physical therapists are indeed associated with different outcomes and, secondly, which interventions are related to better outcome. The methodology utilized in this study, that is, the application of initial and follow-up standardized health status measures in a group of diagnostically related patients, provision of standard forms of physical therapy, and retrospective analysis of differential effects of treatment, is an efficient and cost-effective method of gathering initial data with which to plan and implement further prospective studies. Randomized controlled trials, particularly in the area of low back pain, are expensive and fraught with methodological and clinical difficulties. Examples of the difficulties include the determination of homogenous subgroups for study; defining standard treatment guidelines; and determining outcome measures that are reliable, valid, and sensitive to valid change. In light of these methodological difficulties, and the large number of interventions used by physical therapists for patients with low back pain, preliminary studies such as this one are critical to guide clinical research. Time and effort can then be spent where the physical therapy profession will obtain the biggest return on its research investment.

Jill M Binkley, PT, COMP, FAAOMT Director of Clinical Research Rehab Management Systems Inc 1111 Mountain Dr Dahlonega, GA 30533-1906 ([e-mail])

References

[1] Deyo R, Dichl AK. Psychosocial predictors of disability in patients with low teach pain. J Rheumatol. 1988;15:1557-1564.

[2] VonKorff M, Deyo RA, Cherkin D, Barlow W. Back pain in primary care: outcomes at one year. Spine. 1993;18:855-863.

[3] Lindstrom I, Ohlund C, Eek C, et al. The effect of graded activity on patients with subacute low back pain: a randomized prospective clinical study with an operant-conditioning behavioral approach. Phys Ther. 1992;72:279-290.

[4] Cady LD, Bischoff DP, O'Connell ER, et al. Strength and fitness and subsequent back injuries in firefighters. J Occup Med. 1979;21 :269-272.

[5] Thomas KB. General practice consultations: Is there any point in being positive? BMJ 1987;294:1200-1202.

[6] Moffroid MT, Haugh LD. Henry SM, Short B. Distinguishable groups of musculoskeletal low back pain patients and asymptomatic control subjects based on physical measures of the NIOSH low back atlas. Spine. 1994; 12: 1350-1358.

[7] Binkley JM, Finch E, Hall J, et al. Diagnostic classification of patients with low back pain: report on a survey of physical therapy experts. Phys Ther. 1993;73:138-155.

[8] Riddle DL, Rothstein JM. Intertester reliability of McKenzie's classifications of the syndrome types present in patients with low back pain. Spine. 1993;18:1333-1344.

[9] Delitto A, Erhard RE, Bowling RW. A treatment-based classification approach to low back syndrome: identifying and staging patients for conservative treatment. Phys Ther. 1995;75:471-489.

[10] Delitto A, Cibulka MT, Erhard RE, et al. Evidence for use of an extension-mobilization category in acute low back syndrome: a prescriptive validation pilot study. Phys Ther. 1993;73:216-228.

[11] Di Fabio RP, Mackey G, Holte JB. Physical therapy outcomes for patients receiving workers' compensation following treatment for herniated lumbar disc and mechanical low back pain syndrome. J Orthop Sports Phys Ther. 1996;23:180-187.

[12] Stratford PW, Binkley JM, Riddle DL. Health status measures: strategies and analytic methods for assessing change scores. Phys Ther. In press.

[13] Kazis LE, Anderson JJ, Meenan RF. Effect sizes for interpreting changes in health status. Med Care. 1989;27:S178-S189.

[14] Stratford PW, Binkley JM, Solomon P, et al. Assessing change over time in patients with low back pain. Phys Ther. 1994;74:17-22.

[15] Stratford PW, Solomon P, Binkley JM, et al. Sensitivity of sickness impact profile items to measure change over time in a low back pain patient group. Spine. 1993;18:1723-1727.

[16] Patrick DL, Deyo RA, Atlas SJ, et al. Assessing health-related quality of life in patients with sciatica. Spine. 1995;20:1899-1909.

[17] Ruta DA, Garratt AM, Wardlaw D, Russell IT. Developing a valid and reliable measure of health outcome for patients with low back pain. Spine. 1994; 19:1887-1896.

[18] Guyatt G, Walter S, Norman G. Measuring change over time: assessing the usefulness of evaluative instruments. J Chronic Dis. 1987; 40:171-178.

[19] Stratford PW, Gill C, Westaway M, Binkley JM. Assessing disability and change on individual patients: a report of a patient-specific measure. Physiotherapy Canada. 1995;47:258-263.

[20] Tugwell P, Bombardier C, Buchanan WW, et al. The MACTAR patient preference disability questionnaire: an individual functional priority approach for assessing improvement in physical disability in clinical trials in rheumatoid arthritis. J Rheumatol. 1987;14:446-451.

Author Response

We would like to thank Ms Binkley for her commentary. As she correctly points out, there are certainly limitations to the methods we used in this study, and these limitations apply to a greater or lesser extent whenever a clinical database is used to examine predictors of outcome. On the other hand, as noted by Ms Binkley, randomized clinical trials in the study of low back pain have their own set of difficulties and can be prohibitively expensive. Ms Binkley identifies some of the variables that may have affected our results for which we have no data. Insofar as is possible, those designing clinical databases should consider including such information. In addition, our study highlights the need for complete data collection and follow-up, and the problems encountered in attaining this goal. Physical therapists participating in data-collection efforts must be accountable for obtaining all the necessary information from patients as well as collecting data from all patients to reduce the biases that we have noted.

One issue that we view differently than does Ms Binkley concerns the relative usefulness of generic and disease-specific scales in reporting outcomes. We reported the greatest effect sizes in the disease-specific scales and the bodily pain and physical functioning scales of the SF-36 in our sample. As Figure 2 illustrates, change occurred, though to a lesser extent, in five of the six scales of the SF-36. In using the generic instrument, therefore, we were able to determine where changes occurred, as well as where changes did not occur, and the extent of those changes. We argue, therefore, that use of the generic instrument allows a multidimensional view of the pattern of outcomes that the disease-specific instruments preclude. Because the SF-36 does not include an overall score, dimensions in which change does not occur do not reduce the effect found in other scales in which change does occur. In some disease-specific scales, exactly this problem may be encountered if they are designed to provide one score that considers all the important outcomes of a condition.

One problem in interpreting the effect sizes is that we cannot be sure whether the effect size we reported for each scale was related to how much "true" change occurred or to the degree of sensitivity of each scale to detect change. For example, Ms Binkley suggests that the mental health scale may not be sensitive in patients with spinal conditions. An equally plausible interpretation is that only a small change in mental health occurs over an episode of physical therapy. Without an additional measure against which to assess true change in mental health, this issue could not be addressed in our study.

As we had hoped, and as Ms Binkley's commentary suggests, our study demonstrates the issues that must be addressed in using clinical databases for research and raises interesting questions that we trust will be the basis of future studies.

Diane U Jette, DSc, PT Alan M Jette, PhD, PT

DU Jette, DSc, PT, is Associate Professor and Program Director, Graduate Program in Physical Therapy, Graduate School for Health Studies, Simmons College, 300 The Fenway, Boston, MA 02115 (USA) ([e-mail]), and Physical Therapist, Beth Israel HealthCare, 330 Brookline Ave, Boston, MA 02215. Address all correspondence to Dr Jette at the first address.

AM Jette, PhD, PT, is Professor and Dean, Sargent College of Allied Health Professions, Boston University, 635 Commonwealth Ave, Boston, MA 02215. He was Chief Research Scientist, New England Research Institutes Inc, 9 Galen St, Watertown, MA 02172, and Professor of Social and Behavioral Sciences, Boston University School of Public Health, 80 E Concord St, Boston, MA 02118, at the time of this study.

This study was approved by the Human Subjects Review Board of New England Research Institutes Inc.

This work was supported by a Mary Switzer Rehabilitation Research Fellowship awarded to DU Jette from the National Institute of Disability and Rehabilitation Research, US Department of Education (#H133F50022).

This article was submitted October 2, 1995, and was accepted March 27, 1996.