The sensitivity and specificity of pain response to activity and position in categorizing patients with low back pain.

For persons under the age of 45 years, low back pain (LBP) is the most common cause of activity limitation.[1] Prolapsed disk is one of most frequent conditions--second only to schizophrenia--for which worker disability payments are allowed.[2] Three months after onset of the condition, only 5% of patients have persisting symptoms; however, these patients account for 85% of the LBP costs associated with compensation and time lost from work.[3] Only 50% of employees with back problems who are absent from work for more than 6 months ever return to work.[4]

Low back pain is not a diagnosis. It is a symptom that occurs in a wide variety of medical, musculoskeletal, and neurological conditions. The specific cause of LBP is difficult to establish in most patients at the time of the initial episode.[5] A structural diagnosis is possible in about 70% of patients with chronic pain if those patients with documented psychological aggravation of their symptoms are excluded.[3] Much effort has been devoted to developing diagnostic categories for LBP to guide treatment and inform prognosis. The bases for these diagnostic classification systems, however, have been highly variable.[3,6-10] Some diagnostic classification systems are based on anatomical structures. Many structure-based LBP diagnoses (eg, compression fracture, disk herniation, spinal stenosis) are widely accepted, whereas others (eg, facet syndrome) are more controversial.[6,7] Structure-based diagnostic systems operate by correlating the patient's symptoms and signs with the results of diagnostic imaging tests.[3] Other LBP classification systems such as the system devised by the Quebec Task Force on Spinal Disorders are based on a combination of time since onset and symptoms.[8] Still other diagnostic systems have attempted to classify patients based on symptoms and response to treatment.[9,10]

Any system for classifying LBP benefits from having tests that can be used to screen patients for inclusion in or exclusion from a particular category. The validity of a screening test is determined by how well it correctly classifies individuals with and without a particular disease.[1,12] Four concepts are used to describe the validity of a screening or diagnostic test: (1) sensitivity, (2) specificity, (3) positive predictive value, and (4) negative predictive value (Tab. 1). Sensitivity, or the proportion of true positives, is defined as the proportion of patients with a disease who test positive.[11,12] A test that correctly identifies every person who has a disease has a sensitivity of 1.0. Specificity, or the proportion of true negatives, is defined as the proportion of patients who do not have a disease who test negative.[11,12] A test that correctly identifies as negative every person who does not have a disease has a specificity of 1.0. The positive predictive value of a test is defined as the proportion of patients with a positive test who actually have the disease.[11,12] If everyone who tested positive actually had the disease, the positive predictive value would be 1.0. Similarly, the negative predictive value of a test is defined as the proportion of patients with a negative test who really do not have the disease.[11,12] The positive and negative predictive values of screening tests are affected by the sensitivity and specificity of the test, as well as the prevalence of the disease in the group of patients being screened. As the prevalence of a diagnosis decreases, the positive predictive value also decreases but the negative predictive value increases[11,12]

Table 1.
Sensitivity, Specificity, and Predictive Values(a)

                                Subjects
               Subjects         Without
               With Disease     Disease         Total

Subjects
   who test    a                b
   positive    True positive    False positive   a+b
Subjects
   who test    c                d
   negative    False negative   True negative    c+d
Total          a+c              b+d              a+b+c+d

"Sensitivity=Subjects with the condition who test positive/All subjects with the condition=a/a + c '

Specifity=Subjects without the condition who test negative/Specifity = All subjects without the condition=d/b + d '

Positive predictive value=Subjects with the condition who test positive All subjects who test positive=a/a + b '

Negative predictive value=Subjects without the condition who test negative/All subjects who test negative=d/c + d.

Because the patients participating in this study had either recurrent or chronic LBP, we used a structure-based diagnostic classification system. The purposes of this study were (1) to develop screening tests based on patient reports of changes in LBP in response to activity and posture that could be used to place patients into diagnostic categories and (2) to test the sensitivity, specificity, positive predictive value, and negative predictive value of these screening tests.

Method

Subjects

A sample of 106 consecutive patients with complaints of LBP who were seen at several university-based outpatient orthopedic spine clinics were recruited to participate in this study. The sample represented all patients who met inclusion and exclusion criteria during the time frame of the study. Patients were eligible if they were first-time patients in these offices or if they had not been treated for back pain at these clinics during the 12 months before they entered the study. Patients were excluded as subjects if they did not have complaints of pain in the lumbar area, had a history of back surgery, did not have a confirmed diagnosis at the end of the study period (n=3), or were unable to speak or understand English. All patients signed an informed consent statement prior to participating in the study.

Materials

The patient self-report instrument used in this study, the Pain Response to Activity and Position Questionnaire (PRAP), consists of items selected to represent positions and activities that most individuals experience on a daily basis and that place a variety of biomechanical demands on the lumbar spine. Test-retest reliability was examined in a previous study, and the PRAP demonstrated acceptable reliability.[13] Previous research[14] suggests that this type of instrument can be used to classify LBP as serious or benign. Items on the PRAP are divided into two sections. Fifteen items relate to LBP, and 15 items relate to leg pain. In both sections, subjects are asked to indicate which response best describes their pain when performing an activity: "no pain," "better," "same," or "worse" (Appendix 1). Data were also collected on demographic characteristics, such as gender and age, past medical history, and the presence of radiating leg pain. We also noted the presence of pseudoclaudication (intermittent neurogenic claudication), which is leg pain during walking of neurologic rather than peripheral vascular origin. The presence of pseudoclaudication was determined by the patients' response to an item on the intake questionnaire. Unlike peripheral vascular claudication, the leg pain associated with neurogenic claudication is relieved during ambulation by flexing the lumbar spine and is absent when lower-extremity activity is performed in a seated position, such as while riding a bicycle.[15]

Procedure

Data were collected over the course of two consecutive summers at several university-based outpatient orthopedic spine clinics. The attending physicians at these clinics were medical school faculty and orthopedic surgeons specializing in the treatment of the spine. Many of the patients seen at these clinics are referred by other physicians.

After giving informed consent, subjects completed the PRAP, with assistance as required, but before the initial physician visit. The final LBP diagnosis was determined by the physician by correlating the patient's complaints and results of the physical examination, diagnostic radiologic tests (including computed axial tomography scans and magnetic resonance imaging), laboratory work, consultations, and any other diagnostic tests. The LBP diagnoses used in this study represent diagnoses based on the clinical opinions of the examining physicians. No attempt was made to structure or standardize the diagnostic process. The results of the screening tests were unknown to the physician recording the LBP diagnosis in the medical record and to the researcher extracting the diagnosis from the medical record. We waited at least 1 month from the time of the initial visit before recording the LBP diagnosis from the medical record.

This approach ensured that the results of the diagnostic tests and follow-up visit were available to determine the diagnosis. The LBP diagnoses were placed into one of four categories: (1) benign back pain, (2) disk disease, (3) spinal stenosis, and (4) disk disease with spinal stenosis. Seven subjects did not fit into any of these four categories and were classified as having "other diagnoses." Consistent with previous studies,[14,16-18] benign LBP was defined as nonspecific LBP lacking positive radiographic findings. A diagnosis of spinal stenosis required radiographic evidence of the narrowing of the spinal canal or foramina of either developmental or degenerative origin, although no exact degree of narrowing was specified.[15-18] The diagnosis of disk disease required radiographic evidence of a rupture or displacement of the intervertebral disk material.[16-18] Patients were given the diagnosis of disk disease with spinal stenosis if they had radiographic evidence of rupture or displacement of the intervertebral disk material along with narrowing of the spinal canal or foramina of either developmental or degenerative origin. Patients classified as having "other diagnoses" included those with spinal infection, tumor, fracture, and Reiter's disease.

Development of the Screening Tests

A principal-components analysis with Varimax rotation was performed using items from the PRAP to identify those activities and positions that tended to group together (Tabs. 2 and 3). The sample size of 106 met the criteria of at least 100 subjects, with a minimum of 5 subjects per variable required for replicability of factors.[19] A minimal eigenvalue of 1 was used for this analysis. Following the component analysis, items from the PRAP were individually cross tabulated with each of the diagnoses to determine which items were associated with each of the diagnostic categories. The results of the principal-components analysis and the cross tabulations were combined with clinical judgment to construct screening test algorithms for each of the four diagnostic categories (Appendix 2). The researchers' final decision during data analysis to include an item in the screening test algorithm was based on the effect of including that item on the sensitivity and specificity of the screening test algorithm.

Table 2. Principal-Component Analysis With Varimax Rotation for Low Back Pain Symptoms

Component                                       Loading
            Items Grouping Together             Weight
1
            Sitting more than 5 minutes          .7327
            Bending forward                      .5399
            Driving a car                        .8061
            Lying on back                        .7138
            Lying an side                        .7509
            Lying on stomach                     .6935
            Coughing or bearing down             .5229
2           Standing more than 5 minutes         .8553
            Walking a block or more              .8689
            First thing in the morning           .5711
            At the end of the day                .6708
            Moving from a sitting to a
            standing position                    .6388
3           Lifting heavy objects                .9129
            Carrying heavy objects               .9056
            Pushing heavy objects                .5567

Table 3. Principal-Component Analysis With Varimax Rotation for Leg Pain Symptoms

                                                 Loading
Component   Items Clustering Together            Weight

1           Standing mare than 5 minutes         .7485
            Walking more than a block            .7983
            Bending forward                      .6354
            Lying on stomach                     .5774
            First thing in the morning           .7646
            At the end of the day                .7467
            Moving from a sitting to a
            standing position                    .7165
            Pushing heavy objects                .8357
            Lifting heavy objects                .8385
            Carrying heavy objects               .8394

2           Sitting for more than 5 minutes      .7939
            Driving a car                        .8321
            Lying on back                        .6913
            Lying on side                        .7921
            Coughing                             .6998

Analysis of Reliability and Validity of the Screening Test Algorithms

The data were analyzed on a VAX mainframe computer using SAS version 6.08 statistical software. (*)[20] The reliability of the screening test algorithms was examined using data that were collected on 56 subjects to examine the test-retest reliability of the PRAP. Cohen's kappas were calculated for each screening test algorithm.

Validity of a screening test is described in terms of sensitivity and specificity.[11,12] Computer algorithms were written to classify subjects as testing positive for each of the screening tests based on whether they met the criteria for a particular algorithm. Subjects were cross classified based on their responses to the screening test for a diagnosis and whether they actually had that diagnosis. Sensitivity, specificity, and positive and negative predictive values were calculated from these cross-classification tables. For our study, sensitivity was defined as the proportion of subjects with a particular LBP diagnosis who tested positive on the screening test algorithm for that same diagnostic category.[11,12] Specificity was defined as the proportion of subjects who did not have a particular LBP diagnosis and who tested negative on the screening test algorithm for that diagnostic category.[11,12] The positive predictive value of a screening test was defined as the proportion of subjects with a positive screening test for a particular diagnostic category who actually had that LBP diagnosis.[11,12] Similarly, the negative predictive value of a screening test was defined as the proportion of subjects with a negative screening test for a particular diagnosis who did not have that diagnosis.[11,12]

Results

Of the 106 subjects who participated in the study, 7 subjects were excluded due to previous back surgery or a nonmusculoskeletal source of LBP. An additional 3 subjects were excluded because no diagnosis was available at the end of the study period. Subjects (51.2% male and 48.8% female) had a mean age of 54.6 years (SD=17.9, range=19-88). Eighty-four percent of the subjects reported that their current episode of LBP had lasted more than 7 weeks. Ninety-six percent of the subjects reported no current involvement in a lawsuit, and 91% reported that they had not filed for Workers' Compensation due to their back problem. Subjects having other diagnoses had the oldest mean age (X=70 years), and subjects in the benign back pain category had the youngest mean age (X=46 years) (Tab. 4).

Table 4.
Demographic Information by Low Back Problem Category

                                      Other Low    Benign Low
Characteristic          Total Group   Back Pain    Back Pain

Sample size             106             7           27
Age (y)
   X                     54.6           70.3        46.3
   SD                    17.9           12.8        20.7
   Range                 19-18          44-81       19-84

Weight (lb)
   X                    160.6          156.8         154.1
   SD                    34.9           45.3          29.1
   Range                 90-270         90-225       106-215

Height (in)
   X                     66.6           66.2          66.5
   SD                     4.5            4.9           4.8
   Range                 55-75          60-73         55-75

Percentage with
duration of LBP
of >7 wk                 84.2            83.3          88.0

Percentage male          51.2            63.6          51.8

Percentage receiving
   Workers'
   Compensation           8.8             0.0          17.4

                                                       Disk Disease
                                                       With Spinal
Characteristic       Disk Disease    Spinal Stenosis    Stenosis

Sample size            31                24               17
Age (y)
   X                  51.1               65.5             52.4
   SD                 11.1               14.6             19.2
   Range              59-75              35-88            24-83

Weight (lb)
   X                  168.9              153.0            166.9
   SD                  42.5               29.4             28.5
   Range               119-270  110-225           132-230

Height (in)
   X                    67.4      65.8             66.9
   SD                    4.4       4.0              4.8
   Range                59-75     60-73             56-74

Percentage with
duration of LBP
of >7 wk                79.3      95.0              73.3

Percentage male         51.6      45.8              58.8

Percentage receiving
   Workers'
   Compensation          3.4       5.0              14.3

The kappas for the test-retest reliability of the screening algorithms were .57 for benign back pain, .89 for disk disease, .58 for stenosis, and .91 for disk disease with stenosis.

The presence of leg pain had good sensitivity for both disk disease (.81) and disk disease with spinal stenosis (.94); however, the specificity of this symptom was low for both diagnoses, and the positive predictive value for leg pain was .29 for disk disease and .19 for disk disease with spinal stenosis. The presence of pseudoclaudication had relatively good sensitivity (.63) and specificity (.71), with a positive predictive value of .40 and a negative predictive value of .86 (Tab. 5).

Table 5 Sensitivity and Specificity of a History of Radiating Leg Pain or Pseudoclaudication

                                                    True Positive
                                                    Rate of
Category of Low Back Problem   Test                 Sensitivity

Disk disease                   Radiating leg pain    .81
Spinal stenosis                Pseudoclaudication    .63
Disk disease with spinal       Radiating leg pain    .94
stenosis
Disk disease with spinal       Pseudoclaudication    .47
stenosis

                              True Negative   Positive    Negative
                              Rate of         Predictive  Predictive
Category of Low Back Problem  Specificity     Value        Value

Disk disease                   .18              .29         .68
Spinal stenosis                .71              .40         .86
Disk disease with spinal       .21              .19         .95
stenosis
Disk disease with spinal       .64              .18         .88
stenosis

The sensitivity and specificity of the screening tests varied widely (Tab. 6). The algorithm for disk disease with spinal stenosis achieved the highest sensitivity (.81) and negative predictive value (.94). The algorithm for benign back pain had the highest level of specificity (.71) and the highest positive predictive value (.40).

Table 6. Sensitivity and Specificity of Pain Response to Activity and Position Algorithms

                                True Positive    True Negative
                                Rate of          Rate of
Category of Low Back Problem    Sensitivity      Specificity
Benign low back pain            .56                  .71
Disk disease                    .65                  .49
Spinal stenosis                 .52                  .74
Disk disease with spinal        .81                  .54
stenosis

                                Positive        Negative
                                Predictive      Predictive
Category of Low Back Problem    Power           Power
Benign low back pain            .40             .82
Disk disease                    .35             .77
Spinal stenosis                 .36             .85
Disk disease with spinal        .24             .94
stenosis

Discussion

Because our study sample was drawn from tertiary care, university-based spine clinics, it contained an unusually large proportion of subjects who had LBP for a long time and who had radiographic findings of serious LBP disorders such as spinal stenosis and disk disease. The unusually high prevalence of serious LBP disorders may have affected the results of our study in several ways. Our findings concerning the sensitivity and specificity of pseudoclaudication and radiating leg pain are largely consistent with the findings of previous studies.[6,21] The one exception is our finding of a specificity of .18 for radiating leg pain in subjects with disk herniation. This finding stands in contrast to a previously reported specificity of .88 for this symptom.[6] Unlike previous studies, our sample was on average older and included a large proportion of subjects with spinal stenosis or disk disease with spinal stenosis. Pain radiating below the knee is a symptom of both of these diagnoses. Therefore, a large number of subjects without disk disease may have reported this symptom. These subjects would be classified as false positives, and a large proportion of false positives produces a low level of specificity.

The screening algorithms are largely compatible with clinical expectations for these LBP diagnoses. The algorithm for disk disease was consistent with the typical clinical picture of disk disease, in which flexion activities exacerbate pain and extension activities relieve it.[15,18] Likewise, the algorithm for spinal stenosis was consistent with the clinical literature, which describes patients who experience relief from pain with flexion activities and exacerbation of pain with extension activities.[15,17,18] The screening algorithm for disk disease with spinal stenosis includes both flexion and extension activities that are thought to increase back or leg pain in patients with disk disease or with spinal stenosis, respectively.

Ideally, a screening algorithm would have 100% sensitivity and specificity. Unfortunately, sensitivity and specificity tend to be inversely related.[11,12] Because clinicians must often select either a sensitive test or a specific test, it is important to understand the implications of each test. Highly sensitive tests are, by definition, usually positive in the presence of a disease. Sensitive tests are helpful when there would be serious consequences associated with failing to diagnose a disease correctly.[11,12] Therefore, highly sensitive tests can be used to rule out a disease. In our study, the screening test for disk disease with spinal stenosis had good sensitivity in that 81% of the patients in this diagnostic category screened positive for this diagnosis. If a test is highly sensitive, it is most useful to the clinician when the result is negative. For example, if a patient did not screen positive for disk disease with spinal stenosis, it would be unlikely that the patient had this diagnosis. The effect of a highly sensitive test is reflected by the finding that the negative predictive value of the screening test for disk disease with spinal stenosis was 94% in a clinic population in which this diagnosis was relatively common. In a primary care setting where far less than 16% of the patients would be likely to have this diagnosis, the negative predictive value of this test would be even higher.

Highly specific tests are rarely positive in the absence of disease[12] and may therefore be used to rule in a suspected diagnosis. Highly specific tests are required when false positive results have serious medical, financial, or emotional consequences. In our study, the algorithm for benign back pain had a specificity of .71. Highly specific tests are most helpful when they are positive. The positive predictive value of this test was 40% in a clinic population in which only 25% of the patients had benign back problems. Because of the influence of disease prevalence on the positive predictive power of a test, in a more typical primary care setting in which 90% of the patients have benign back problems, the positive predictive value of this test would be 96%.

Our study has several limitations. There is no universally accepted diagnostic classification system for LBP.[6-10] The sensitivity and specificity of these screening test algorithms were calculated using a structure-based clinical diagnosis as the criterion for comparison. Although this approach is consistent with that of other studies,[6,21] it leaves open questions concerning the reliability and validity of the clinical diagnosis itself. Although those issues were beyond the scope of our study, they may affect the generalizability of our findings. Although we firmly believe that the LBP diagnoses used in this study are consistent with standard medical practice for determining a structural LBP diagnosis, it is possible that the sensitivity and specificity of the screening test algorithms would be somewhat different if the clinical diagnoses had been determined by another group of physicians. The use of symptom-based screening test algorithms for other diagnostic classification systems remains to be tested.

Reports of changes in pain intensity in response to activity and position are subjective and may vary according to the patient's pain awareness, threshold, perception, and knowledge.[22] Medication use may also confound self-reports of pain by masking pain produced by an activity, decreasing the patient's awareness of fluctuations in pain, and altering cognitive ability to report past pain experience. Fifty percent of the subjects in this study reported medication use.

The PRAP responses were based on the subjects' recent experience with the stated activity or position and were therefore influenced by all the factors that affect pain recall. If subjects were asked to report changes in pain intensity in response to activities and positions performed during the course of a clinical examination, the findings might be somewhat different. Whether changes in pain produced during a clinical examination are characteristic of those experienced during daily activities, however, is unclear.

Inaccurate pain recall is not the only source of variability in reported fluctuations in pain in response to the activities and positions encountered in the course of daily activities. A previous study examining test-retest reliability of the PRAP showed that the patient reports of changes in back pain when sitting were only moderately reliable (kappa=.46).[9] This finding suggests that there is substantial variability in pain response to this activity. Whether sitting increases or decreases back pain may depend on the type of surface on which the subject is seated. Patients might give more consistent responses if the activities were standardized. The test-retest reliability of the screening tests was fair to good over a short time frame. Consistency over a period of days or weeks is unknown.

Biobehavioral factors may have influenced patient pain reports, thus limiting the accuracy of findings obtained with the algorithms.[22] Only a small percentage of the subjects in our study were involved in Workers' Compensation cases (8.8%) or litigation (3.7%). Within the four diagnostic categories, the percentage of subjects involved in Workers' Compensation claims was highest in the mechanical LBP (17.4%) and disk disease with spinal stenosis (14.3%) categories. The possible influence of these variables should not be overlooked. These variables probably had a negligible impact on the results of this study, however, given the small number of subjects involved in litigation or Workers' Compensation claims.

Conclusion

Our study provides evidence that screening test algorithms, utilizing patient reports of pain responses to selected postures and activities, may be useful in placing patients into diagnostic categories. Pain reports by themselves, however, are unlikely to provide sufficient information to place a patient in a diagnostic category. All of the algorithms led to misclassification of a substantial number of patients. The misclassification of subjects by diagnostic tests, however, is not limited to these screening test algorithms. Diagnostic imagery, such as radiography, magnetic resonance imaging, and computed tomography scans, produces substantial numbers of false positive findings, indicating herniated disks and spinal stenosis in asymptomatic subjects.[21] A thorough patient history and a detailed physical examination, followed, at times, by diagnostic imagery, are required to confirm or deny the LBP diagnosis suggested by the screening test algorithm. These screening test algorithms can be used as a first step in the diagnostic process to focus and direct the clinical examination.

(*) SAS Institute Inc, PO Box 8000, Cary, NC 27511.

References

[1] Verbrugge LM. From sneezes to adieux: stages of health for American men and women. In: Ward RA, Tobin SS, eds. Health in Aging: Sociological Issues and Policy Directions. New York, NY: Singer Publishing Co; 1987:17-57.

[2] Kelsey JL, White AA. Epidemiology and impact of low-back pain. Spine. 1980;5: 133-142.

[3] Frymoyer JW. Back pain and sciatica. N Engl J Med. 1988;318:291-298.

[4] McGill CM. Industrial back problems: a control problem. J Occup Med. 1968;10:174-178.

[5] Dillane JB, Fry J, Kalton G. Acute back syndrome: a study from general practice. BMJ. 1966;2:82-84.

[6] Bigos S, Bower O, Braen G, et al. Acute Low Back Problems in Adults: Clinical Practice Guideline, Quick Reference Guide Number 14. Rockville, Md: US Department of Health and Human Services, Public Health Service, Agency for Health Care Policy and Research; December 1994. AHCPR Publication No. 95-0643.

[7] Binkley J, 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] Quebec Task Force on Spinal Disorders. Scientific approach to the assessment and management of activity-related spinal disorders--a monograph for clinicians: report of the Quebec Task Force on Spinal Disorders. Spine. 1987;12(suppl 7):S1-S59.

[9] 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.

[10] 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;75:470-485.

[11] Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical Epidemiology: A Basic Science for Clinical Medicine. Boston, Mass: Little, Brown and Company Inc; 1991:69-152.

[12] Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology: The Essentials. Baltimore, Md: Williams & Wilkins; 1988:42-75.

[13] Roach KE, Brown MD, Dunigan KM, Walas M. Test-retest reliability of a low back pain questionnaire. Phys Ther. 1994;74(suppl):S56. Abstract.

[14] Roach KE, Brown MD, Ricker E, et al. The use of patient symptoms to screen for serious back problems. J Orthop Sports Phys Ther. 1995;21:2-6.

[15] Anderson GBJ, McNeill TW. History and physical examination in spinal stenosis. In: Anderson GBJ, McNeill TW, eds. Lumbar Spinal Stenosis. St Louis, Mo: Mosby-Year Book; 1992:227-234.

[16] Mann NJ, Brown MD. Artificial intelligence in the diagnosis of low back pain. Orthop Clin North Am. 1991;22:303-314.

[17] Lipson SJ. Low back pain. In: Kelley WN, Harris ED, Ruddy S, Sledge CB, eds. Textbook of Rheumatology. Philadelphia, Pa: WB Saunders Co; 1991:441-458.

[18] Hu SS, Bueff HU, Tribus CB. Disorders, diseases, and injuries of the spine. In: Skinner HB, ed. Current Diagnosis and Treatment in Orthopedics. East Norwalk, Conn: Appleton & Lange; 1995:165-235.

[19] Gorsuch RL. Factor Analysis. Hillsdale, NJ: Lawrence Erlbaum Associates Inc; 1983:328-349.

[20] SAS/STAT User's Guide, Version 6, Volumes 1 and 2. 4th ed. Cary, NC: SAS Institute Inc; 1989.

[21] Deyo RA, Rainville J, Kent DL. What can the history and physical examination tell us about low back pain?JAMA. 1992;268:760-765.

[22] Feuerstein M, Beattie P. Biobehavioral factors affecting pain and disability in low back pain: mechanisms and assessment. Phys Ther. 1995;75: 267-280.

Appendix 1. Sample Questions From the Pain Response to Activity and Position Questionnaire (PRAP)(a)

Positions That Affect Pain                      No Pain
   1. When you stand for more than 5 minutes       N
   2. When you walk for a block or more            N
   3. When you sit for more than 5 minutes         N
   4. When you bend forward over a sink            N
   5. When you drive a car                         N
   6. When you lie on your back                    N
   7. When you lie on your side                    N
   8. When you lie on your stomach                 N
   9. When you cough or bear down                  N
   10. When you get up in the morning              N
   11. At the end of the day                       N
   12. When you move From a sitting to a
   standing position                               N
   13. When you push something heavy
   like a vacuum cleaner or lawn mower             N
   14. When you lift heavy objects
   from the floor                                  N
   15. When you carry heavy objects                N

Positions That Affect Pain                      Better     Same
   1. When you stand for more than 5 minutes    B          S
   2. When you walk for a block or more         B          S
   3. When you sit for more than 5 minutes      B          S
   4. When you bend forward over a sink         B          S
   5. When you drive a car                      B          S
   6. When you lie on your back                 B          S
   7. When you lie on your side                 B          S
   8. When you lie on your stomach              B          S
   9. When you cough or bear down               B          S
   10. When you get up in the morning           B          S
   11. At the end of the day                    B          S
   12. When you move From a sitting to a
   standing position                            B          S
   13. When you push something heavy
   like a vacuum cleaner or lawn mower          B          S
   14. When you lift heavy objects
   from the floor                               B          S
   15. When you carry heavy objects             B          S

Positions That Affect Pain                   Worse      N/A
   1. When you stand for more than 5 minutes    W       N/A
   2. When you walk for a block or more         W       N/A
   3. When you sit for more than 5 minutes      W       N/A
   4. When you bend forward over a sink         W       N/A
   5. When you drive a car                      W       N/A
   6. When you lie on your back                 W       N/A
   7. When you lie on your side                 W       N/A
   8. When you lie on your stomach              W       N/A
   9. When you cough or bear down               W       N/A
   10. When you get up in the morning           W       N/A
   11. At the end of the day                    W       N/A
   12. When you move From a sitting to a
   standing position                            W       N/A
   13. When you push something heavy
   like a vacuum cleaner or lawn mower          W       N/A
   14. When you lift heavy objects
   from the floor                               W       N/A
   15. When you carry heavy objects             W       N/A

(A) The PRAP consists of 30 questions. The 15 questions presented here are asked one time referring to the low back, and the same questions are asked a second time referring to the legs.

Appendix 2. Low Back Pain Classification Algorithms

Criteria for Spinal Stenosis Classification

If at least one of these five symptoms is present

1. Back pain is worse when standing for more than 5 minutes

2. Back pain is worse when walking a block or more

3. Leg pain is worse when carrying heavy objects

4. Leg pain is worse when pushing heavy objects

5. Leg pain is worse when standing for more than S minutes And at least one of these two symptoms is present

1. Back pain is absent, better, or unchanged when sitting for more than 5 minutes

2. Leg pain is absent, better, or unchanged when sitting for more than 5 minutes And at least one of these two symptoms is present

1. Leg pain is absent, better, or unchanged first thing in the morning

2. Leg pain is absent, better, or unchanged at the end of the day Then the subject is placed in the spinal stenosis category.

Criteria for Disk Disease Classification

If at least one of these four symptoms is present

1. Back pain is worse when driving a car

2. Back pain is worse when sitting for more than 5 minutes

3. Leg pain is worse when driving a car

4. Leg pain is worse when sitting for more than 5 minutes And at least one of these three symptoms is present

1. Leg pain is absent, better, or unchanged when pushing heavy objects

2. Leg pain is absent, better, or unchanged when standing for more than 5 minutes

3. Leg pain is absent, better, or unchanged when lying on stomach Then the subject is placed in the disk disease category.

Criteria for Disk Disease With Spinal Stenosis Classification

If at least one of these three symptoms is present

1. Back pain is worse when pushing heavy objects

2. Leg pain is worse when carrying heavy objects

3. Leg pain is worse when standing for more than 5 minutes And at least one of these two symptoms is present

1. Back pain is worse first thing in the morning

2. Leg pain is worse when driving a car Then the subject is placed in the disk disease with spinal stenosis category.

Criteria for Benign Low Back Pain Classification

If at least one of these three symptoms is present

1. Back pain is worse sitting for more than 5 minutes

2. Back pain is worse walking more than a block And at least one of these three symptoms is present

1. Back pain is absent, better, or unchanged when moving from a sitting to a standing position

2. Leg pain is absent, better, or unchanged when moving from a sitting to a standing position

3. Leg pain is absent, better, or unchanged when walking more than a block And at least one of these three symptoms is present

1. Leg pain is absent, better, or unchanged when sitting for more than 5 minutes

2. Leg pain is absent, better, or unchanged when lifting heavy objects

3. Leg pain is absent, better, or unchanged when bending forward Then the subject is placed in the benign low back pain category.

KE Roach, PhD, PT, is Assistant Professor, Division of Physical Therapy, Department of Orthopaedics and Rehabilitation, University of Miami School of Medicine, 5915 Ponce de Leon Blvd, Plumer Building, 5th Floor, Coral Gables, FL 33146 (USA) ([e-mail]) Address all correspondence to Dr Roach.

MD Brown, MD, PhD, is Professor and Chair, Department of Orthopaedics and Rehabilitation, University of Miami School of Medicine.

RD Albin, PT, was a graduate student in the Division of Physical Therapy, Department of Orthopaedics and Rehabilitation, University of Miami School of Medicine, at the time this study was conducted.

KG Delaney, PT, was a graduate student in the Division of Physical Therapy, Department of Orthopaedics and Rehabilitation, University of Miami School of Medicine, at the time this study was conducted.

HM Lipprandt, PT, was a graduate student in the Division of Physical Therapy, Department of Orthopaedics and Rehabilitation, University of Miami School of Medicine, at the time this study was conducted.

D Rangelli, PT, was a graduate student in the Division of Physical Therapy, Department of Orthopaedics and Rehabilitation, University of Miami School of Medicine, at the time this study was conducted.

This study was approved by the University of Miami Medical Sciences Subcommittee for the Protection of Human Subjects in Research.

The results of this study were presented in abstract format at the 12th International Congress of the World Confederation for Physical Therapy; June 25-30, 1995; Washington, DC.

This article was submitted January 16, 1996, and was accepted March 6, 1997