Diagnostic classification of patients with low back pain: report on a survey of physical therapy experts.

Key Words: Delphi technique, Diagnosis, Low back pain, Lumbar spine, Syndromes.

The percentage of the population that will experience back pain at some point in their lives is estimated to be between 51% and 70%.[1] Complaints of low back pain (LBP) are usually self-limiting, and only between 5% and 10% of persons with LBP remain symptomatic after 3 months.[1] Despite the self-limiting nature of back pain, the actual number of persons with back pain at a given point in time is high and, combined with a recurrence rate estimated at 60% to 85%, the social and economic costs of LBP are high.[1] The total health care costs for LBP in the United States has been estimated at over $12 billion per year.[1] This estimate is considered to be low and does not include similar figures for compensation costs and loss of work time.[1] In a recent survey of hospital and private inpatient, outpatient, and community physical therapy settings in an urban center, LBP accounted for the largest proportion, or 22%, of musculoskeletal conditions being treated and for 23% of physical therapists' musculoskeletal patient care time.[2] Despite the staggering socioeconomic impact of LBP and the large proportion of physical therapy practice devoted to its treatment, physical therapists lack an accepted standardized diagnostic classification. Diagnostic classification schemes have several important functions in health care. As outlined by Sahrmann,[3] diagnostic classifications clarify practice, provide an important means of communication with colleagues and consumers, classify and group conditions that can direct research and assessment of treatment effectiveness, and reduce the tendency toward practice based primarily on treatment approaches. For the purpose of this article, the terms diagnose and diagnosis will be used to describe the process and outcome, respectively, of categorizing patients by signs and symptoms.[4-6]

Physical therapy approaches to LBP classification have been developed based on anatomical and pathological rationales as well as clinical experience with a variety of clinical presentations. Subgroups of conditions of LBP may have a clearly identifiable anatomical or pathological cause, such as fracture or disk herniation. Often, however, a specific pathological or anatomical cause of LBP cannot be identified.[7] In this case, the clinical presentation is the basis for determining the management strategy. Clinical presentation is defined as a group of signs and symptoms, or objective data and subjective complaints, respectively. To determine the effectiveness of any management approach, signs and symptoms that identify homogeneous subgroups of patients must be determined. An effectiveness study that fails to specify subgroups of patients with LBP is likely to find a clinically and statistically less significant result when a treatment effect may be present in some patients. The improvement in patients benefiting from treatment may be masked by those in different subgroups in which treatment was not effective, or even harmful.

An evaluation system is being developed for patients with LBP.[8] The objectives of the system are to provide standardized documentation of clinical signs and symptoms and to assist physical therapists in categorizing patients into meaningful groups based on clinical presentations and measures of functional outcome. Patients are categorized on two dimensions. One dimension is the physical diagnosis, and the other is a "rehabilitation classification." The diagnostic dimension serves to categorize patients into diagnostic groups based on clinical signs and symptoms. The rehabilitation dimension is a scale used to rate degree of impairment, disability, and handicap.[8,9] We propose that both dimensions are pivotal in creating homogeneous subgroups that may benefit from similar physical therapy management strategies. We hypothesize that both diagnosis and factors such as degree of chronicity, long-term disability and handicap, and ability to cope are factors determining appropriate management and predicted treatment outcome. For example, a patient with an initial episode of disk pathology of 3 days' duration may have a different predicted outcome than would a patient with a similar diagnosis of disk pathology of 6 months' duration and long-term disability and handicap. In developing an evaluation system, we noted that there was a lack of a standardized classification system for patients with LBP. The purpose of this study was to begin development of the diagnostic dimension of the classification system by surveying a small group of identified expert orthopedic physical therapists.

Diagnostic Classification Systems

Many authors[6,10-15] categorize LBP on the basis of tissue pathology, such as disk, joint, or ligament abnormality, and by pathological process, such as inflammation, compression, or instability. There is little consensus on the combinations of tissue and pathological processes by which LBP is classified, as evidenced by the following review of the literature. The categories of disk pathology and spinal stenosis, for which clear radiological criteria exist, are exceptions and are included in each of the systems of classification reviewed.

Sypert[6] emphasizes the importance of diagnosing clinical syndromes to determine appropriate management strategies. Six LBP syndromes are described: neural compression, mechanical (instability), myofascial, inflammatory, neuropathic, and psychosocial-economic.[6] Mooney,[10] however, argues that the only reason to classify clinical entities is to establish prognosis and a rationale for management. In patients with LBP, it is difficult at present to distinguish between one clinical entity and another on the basis of pathology.[10] Observation of patterns of symptoms, therefore, is critical. Mooney suggests that diagnostic judgments are made on epidemiological data, symptom description, signs of dysfunction, and radiographic investigations. His framework for classifying patients with LBP consists of two categories: peripheral nerve dysfunction and instability. Peripheral nerve dysfunction consists of ruptured intervertebral disks and developmental and degenerative stenosis. Both are modes of nerve root impingement that have clearly understood pathological anatomy based on surgical and radiographic analysis. Instability consists of traumatic and degenerative spondylolisthesis, isolated disk disruption, and soft tissue incompetence. Soft tissue incompetence is stated to be the most common LBP syndrome and is characterized by normal radiographs and neurological examination findings. The term "soft tissue incompetence" is not defined. Mooney notes the difficulty in establishing clear pathological anatomy in the category of instability.[10]

Deyo[11] classifies patients with LBP into three categories according to the origin of the problem: mechanical, nonmechanical, and visceral. He states that it is usually not possible to clinically distinguish between the sources of mechanical LBP (eg, facet joint, ligament, paravertebral muscle] and suggests that these sources be considered one category, as the medical management of each is similar. Nonmechanical back pain includes neoplasm, inflammatory conditions, and infection, but Deyo estimates that these conditions are very rare, probably one case per 1,000 patients with LBP.[11]

Saunders[12] outlines a diagnostic classification system based on the anatomical structure involved. The major categories are muscle, joint, nerve root, and miscellaneous. He states that spinal pain of true muscle origin is not common. The joint category includes, for example, facet impingement, sprain, hypomobility, hypermobility, degenerative joint disease, and postural syndrome. The nerve root category consists of syndromes that cause impingement on nerve roots, including degenerative joint disease, disk herniation, and prolapse. Examples in the miscellaneous category are osteoporosis, spondylolisthesis, and sacroiliac syndromes. Saunders[12] outlines a clinical picture of each category, including history of onset and neurological, mobility, and radiographic findings. The categories suggested are not mutually exclusive; for example, degenerative joint disease falls into joint and nerve root categories.[12]

Porter[13] identifies four syndromes related to mechanical disorders of the lumbar spine: symptomatic disk protrusion, nerve root entrapment, segmental instability, and neurogenic claudication. Signs and symptoms relevant to each disorder are outlined. Fast[14] describes two broad categories of low back disorders of myofascial pain syndromes and sciatic syndromes. The sciatic syndromes include herniated disk, spinal stenosis, and facet syndrome.

Bernard and Kirkaldy-Willis[15] report on the classification of 1,293 cases of LBP in a retrospective review over a 12-year period. Syndromes are grouped into three categories: well recognized, less well recognized, and other. The well-recognized syndromes are based on clinical and radiological diagnoses. These diagnoses include herniated nucleus pulposus, lateral spinal stenosis, spondylolisthesis, and segmental instability. The less-well-recognized category consists of syndromes that cannot be verified radiologically. The sacroiliac joint, posterior joint, and myofascial syndromes are in this category, with the first two syndromes accounting for 23% and 22% of the total cases, respectively. These statistics indicate the magnitude of the problem of classifying patients in the absence of radiological evidence. The remaining syndromes were chronic pain, ankylosing spondylitis, tumor, and infection. The retrospective nature of the study suggests that control over diagnostic accuracy would be difficult, and the authors did not state whether the examination protocol was standardized. This study emphasizes that the largest proportion of patients with LBP fall into the less-well-recognized category, in which radiographic evidence of pathology is not available.[15]

In an extensive review of the literature in 1987, the Quebec Task Force on Spinal Disorders concluded that present knowledge does not support a diagnostic classification system based on the pathology of specific structures or on pathological processes.[7] The system concluded to be most valuable by the Task Force is one in which clinical signs and symptoms are the basis.[7] The Task Force documents a classification system based on pain location, neurological signs, and results of paraclinical investigations. Spinal stenosis, postsurgery, chronic pain, and "other" are separate categories. In addition, duration of symptoms and work status are considered when classifying patients.[7] Delitto et al[4] and McKenzie[5] also outline diagnostic classification systems based on clinical signs and symptoms, with less emphasis on pathological tissues. McKenzie[5] describes in detail the clinical presentation of three mechanical LBP syndromes: derangement, dysfunction, and postural syndromes. A specific pathology for each syndrome is postulated. Alterations in the position of the nucleus pulposus of the intervertebral disk are implicated in the derangement syndrome. Mechanical deformation of shortened soft tissue is the pathological and anatomical basis for the dysfunction syndrome, and prolonged stress on soft tissue attributable to postural positioning is the cause of pain in postural syndromes.[5] Data to support these hypotheses have not been presented.

Delitto et al[4] have developed an expert-based decision-support system for diagnosis and treatment of LBP. The goal of the work was to develop a treatment-oriented diagnostic scheme based on patients' signs and symptoms. Experts were defined as highly skilled clinicians who are well enough informed to understand and judge their clinical skills. A low back assessment battery of tests was developed. Expert opinion was used to establish protocols for determining when, in what order, and how tests should be administered. Signs and symptoms were then clustered, and critical signs and symptoms for three diagnostic categories were identified. These categories were extension principle, flexion principle, and mobilization principle. The first two diagnoses do not occur together, whereas the third is present with one of the first two categories. The authors concluded that with adequate training, physical therapists could reliably place patients into the three categories.4

A standardized system of classification of LBP should exhibit the following features: consistent terminology; mutually exclusive categories representing distinct, recognizable clinical presentations; and categories specific enough to guide patient management. In reviewing the literature, we found inconsistent terminology. An example of this inconsistency is the varied terminology used for pathology of disk origin, categorized as "derangement" by McKenzie,[5] "extension principle" by Delitto et al,[4] "herniated nucleus pulposus" by Bernard and Kirkaldy-Willis,[15] and "peripheral nerve dysfunction" by Mooney.[10] There are gaps in each classification system reviewed in which one or more of the classes described in the majority of the literature is not included. For example, Saunders[12] and Bernard and Kirkaldy-Willis[15] include sacroiliac syndromes, whereas the remaining literature reviewed does not. McKenzie[5] does not include instability or hypermobility as categories, although other classifications[6,11,13,15] contain this category. Only Sypert[6] included a psychosocial category of LBP: psychosocial-economic. McKenzie[5] and Saunders[12] appear to provide the most detailed descriptions of the clinical signs and symptoms associated with each proposed class. Several of the classification systems reviewed may be appropriate when the goal is to distinguish between LBP of spinal origin and LBP of visceral origin and between surgical and nonsurgical management options, but these classifications lack the specificity required to categorize and guide nonsurgical management of mechanical LBP.[6,7,11-13] In summary, a standardized classification system enabling categorization of all patients with LBP into specific, mutually exclusive classes by identifiable clinical presentation has not been described to date.

There is a need to develop a standardized system of classification for LBP. This system would allow discussion among health care professionals, would serve as a basis for management decision making, and would assist in the creation of homogeneous subgroups for intervention trials.

In the absence of methods with which to validate classifications of subgroups of LBP, a survey of expert opinion was selected as a method by which to begin development of a standardized classification system. The purposes of the survey were to measure the degree of expert agreement on diagnostic classes of LBP considered important to physical therapists and to identify key signs and symptoms essential to making the diagnosis.

Rationale for Modified Delphi Survey

In selecting a method of carrying out a survey to measure agreement on a diagnostic classification system for LBP, we reviewed the Delphi technique. This technique, traditionally used for forecasting events, is the best-known survey method for assessing expert opinion.[16] The purpose of this standardized approach is to gain convergence, or consensus, among a group of respondents.[16-21] The cardinal features of the Delphi technique are anonymity of response and iteration with feedback to respondents.[17,18]

The Delphi technique classically consists of three stages, or rounds.[17-19] Respondents to the survey are commonly experts in the field and are identified by the survey director. In round 1, a series of questions relevant to the overall survey topic are given to identified respondents. The responses, organized and processed by the survey director, become the basis for a second questionnaire in round 2. Items, or criteria, generated by the process in round 1 are collated, and rating scales for each criterion are included. Likert-type scales are most commonly used to rate Delphi questionnaire criteria.[19] In round 2, respondents are required to rate the items on the developed questionnaire, usually with respect to the relevance or importance of the item to the issue being surveyed.[17,19,21] Round 2 responses are analyzed by the survey director to evaluate the degree of agreement on items.

In round 3, respondents are again presented with the round 2 questionnaire but additionally receive the results of the descriptive analysis as a summary of the group response for each item. With this information and, in some cases, knowledge of their own previous responses, respondents are given the opportunity to rescore each questionnaire item.[17,18,20]

The advantages of the Delphi method for gaining consensus from a group of respondents on a specified topic are its repeatability and ease of administration.[18,19] Anonymity avoids direct personality influences among respondents. Such anonymity, however, may produce a lack of accountability among the respondents.[20]

Two potential shortcomings of the classic Delphi technique are (1) the potential to generate a vast amount of unusable data in round 1, when the topic is broad, and (2) the questionable value of round 3, in which conformity may be produced without a concomitant improvement in validity. In our survey, we addressed our concern regarding generating too much data by setting boundaries for round 1. Our item generation was done through a literature review. Our concerns regarding round 3 were based on literature that questions the validity of the conclusions drawn from such a Delphi survey.[19,21,22] The use of experts as respondents is common in the application of the Delphi technique as a method of increasing content validity.[20] There is, however, no guarantee that the final consensus achieved using the third round is close to the "correct" answer, and it is also difficult to estimate how far from the "truth" the result lies.[22] The effect of feedback to the respondents in round 3 was examined by giving false feedback to the respondents.22 It Was found that respondents moved closer to the mean, despite its being false. This finding raises a major concern that the Delphi technique, although moving respondents closer to conformity, does not necessarily move them closer to consensus on the "truth."[20] Romm and Hulka[21] have also reported that results of the third round of questionnaires to gain consensus had a minimal effect on responses from round 2. In our survey, the concerns regarding validity of round 3 were addressed by omitting this round and reporting the response distribution without seeking consensus.

Method

Selection of Experts

English-speaking physical therapists with expertise in assessment and treatment of spinal conditions and with a minimum of 5 years of experience in the field were surveyed. Experts were selected by consensus of four of the authors from physical therapists meeting one or more of the following criteria: (1) faculty membership in an undergraduate or postgraduate physical therapy program involved in teaching an orthopedic assessment and treatment course(s), (2) authorship on the topic of spinal orthopedics in a peer-reviewed journal, or (3) completion of a recognized orthopedic physical therapy specialization examination in their respective countries. These criteria were considered minimal, and we sought to include experts meeting as many criteria as possible.

It has been reported that the quality of information obtained by the Delphi technique is improved with successive numbers of respondents up to 13; thereafter, there are diminishing returns in information for each additional expert.[18] Based on the literature suggesting a minimum of 13 respondents,[17] aS Well as practical constraints, the survey was limited to 30 experts. The geographical distribution of experts selected were: Canada (n = 21), United States (n = 2), Australia (n = 4), New Zealand (n = 1), Britain (n = 1), and Norway (n = 1).

Experts were contacted initially in person, by phone, or by mail to obtain their consent to participate. Once consent was obtained, the survey questionnaire was either mailed to the experts or given to them in person. The experts were asked to return the questionnaires within 30 days. Follow-ups, consisting of mailings and phone calls, were made at 30-day intervals for a total of 3 months following the initial contact for individuals who had not responded.

Survey Development

Several modifications were made to the Delphi technique for this survey. Because of the breadth of content and terminology in the field of LBP, round 1 was replaced with item generation through the literature review. The use of alternatives to respondent-generated questionnaire items, such as the use of a literature review, has been described in the literature.[18,21) Compilation of the questions by the survey director, thus eliminating round 1, has the advantage of avoiding irrelevant responses and item generation, particularly where the topic is broad.[17] Round 3 was eliminated because of the questionable validity of the resulting data and the particular research question. The research objective was to measure agreement on diagnostic classification rather than necessarily coming to consensus on a single "answer" to each question.

Based on the review of the literature, 25 commonly described diagnostic classes of LBP and associated clinical findings for each class were identified. Each diagnostic class was rated on a five-point scale of appropriateness for inclusion as a category of LBP. Additional diagnostic classes not included in the survey were then requested. Relative "essentialness" of each clinical finding to the diagnosis was also rated on a five-point scale. "Essentialness" is a term used by Romm and Hulka[21] in a Delphi survey on clinical care for various medical conditions. For the purposes of our study, essentialness was defined as the extent to which the clinical finding is crucial for establishing the diagnosis. Additional essential findings were then requested for each diagnostic class. Finally, the usefulness of the survey was rated. An example of the survey questionnaire for 1 of the 25 diagnostic classes is presented in Appendix 1. The questionnaire was pilot tested on 5 physical therapists: 3 orthopedic clinicians, 1 orthopedic clinical researcher, and 1 researcher with expertise in clinical measurement. Minor revisions of the questionnaire were made as a result of the pilot testing.

Data Analysis

A survey of 10 practicing orthopedic physical therapists not involved as experts in the survey indicated that agreement of at least 75% of the experts would be required to have an impact on their clinical practice. These clinicians used our criteria for defining expert physical therapists when making this decision. We determined a priori that a rating of 3 or greater on the five-point scale would be considered agreement with an item. Respondents were not given information regarding this scoring interpretation prior to completing the survey. Five of the experts were randomly selected and contacted following the survey. These experts agreed retrospectively that a score of 3 or greater was indicative of agreement. In order for a diagnostic class or clinical finding to meet the criteria for agreement, therefore, a rating of 3 or greater by at least 75% of the experts was required.

Descriptive statistics (medians and ranges) were used to describe the responses across all respondents. The median is considered to be the best single statistic for describing group agreement, because response patterns tend to be bimodal.[17] The 25th and 75th percentiles are also commonly used as part of the analysis.[17] The 25th and 75th percentiles are the points at which one quarter of the responses fall below and above that level, respectively. This distribution is graphically displayed in Appendix 2. The range of responses for each expert was determined. Medians were used to determine central tendency, as the data were skewed. Quartiles and minimum and maximum values were used to examine the variability of the data. Differences in medians and ranges of responses relative to geographical location were examined. Medians, quartiles, and minimum and maximum values were summarized using a box-plots procedure and graphically displayed.[23] Guidelines for interpretation of box plots are included in Appendix 2. Responses were analyzed to determine whether each diagnosis and clinical finding met the a priori criteria for agreement. Diagnostic classes and clinical findings with a rating of 3 or greater by more than 75% of experts were retained.

Results

Profile of Respondents

Of the 24 completed survey questionnaires that were returned, 25% were from the United States, Australia, and New Zealand and 75% were from Canada. The geographical distribution of the experts responding was as follows: Canada (n = 17), Australia (n = 4), United States (n = 2), New Zealand (1), Britain (n = 0), and Norway (n = 0). Two experts responded on one questionnaire with consensus as a team and were included as a single expert's response. The specific criteria met by each respondent, as well as additional documentation of expertise, such as book publication, are detailed in the Table.

The following number of experts declined to participate in the study or did not return survey questionnaires: Canada (n = 3), Britain (n = 1), and Norway (n = 1). In one instance, when a selected expert did not respond to our initial request to participate, a second request for consent was mailed together with the survey questionnaire. This resulted in the only specific refusal to participate. The remaining nonrespondents had consented to participate initially but did not return survey questionnaires, despite follow-up procedures.

Diagnostic Classes

Nineteen of the 25 diagnostic classes met the criteria for agreement as appropriate for inclusion in a diagnostic classification scheme for LBP (Fig. 1). For example, on the fivepoint rating scale, 75% of the responses for the diagnostic class of fracture were 5 and the remaining 25% of responses, or the lowest quartile, ranged from 1 to 5. This class exceeds the minimum criteria for inclusion in a classification scheme of a rating of 3 or greater by at least 75% of the experts. Classes are listed in descending order of level of agreement in Figure 1. The range of responses for all classes was from 1 to 5. The diagnostic classes with the highest level of agreement were fracture, tumor, Scheurmann's disease, and spondylolisthesis, in which 75% of the responses for the diagnostic classes were 5. Fifteen additional classes met the minimum criteria for agreement of a rating of 3 or greater by at least 75% of experts. These classes, however, had larger ranges, between 3 and 5, for the upper three quartiles. For example, 50% of the responses for the diagnostic class of hypomobility dysfunction were 5, the next 25% of the responses were 4 and 5, and the lowest 25% of responses ranged from 1 to 4.

Six of the 25 diagnostic classes failed to meet the minimum criteria for agreement (Fig. 2). The following six additional diagnostic classes were suggested by more than one expert: vascular back pain (n=3), osteoporosis (n=2), coccydynia (n=2), spondylosis (n=2), spondylolysis (n=2), and diffuse idiopathic skeletal hyperostosis (n=2).

The range of responses for each expert was from 1 to 5. There was no difference in responses based on geographical location, for any diagnostic class or essential finding.

Essential Findings for Each Diagnostic Class

The rating of essentialness of clinical findings associated with agreed upon diagnostic classes was summarized using box plots (Fig. 3). Additional clinical findings considered to be essential to the class by more than one expert are indicated. Diagnostic classes that did not meet the minimum criteria for agreement are not included, as further work is required to define these classes.

Sacroiliac hypermobility is an example of a diagnostic class with clinical findings that met the criteria as essential to this diagnosis as well as findings that did not meet the criteria and additional findings suggested by more than one expert. The clinical findings that met the criteria to be considered essential were (1) unilateral buttock or posterior thigh pain, or both, and (2) sarcoiliac kinetic tests indicative of hypermobility. With respect to the first finding, 50% of the responses were 4 or 5, the next 25% of the responses were 3 or 4, and the lowest 25% of the responses are 1 to 3. Two clinical findings surveyed--asymmetry of pelvic bony landmarks (ie, positional findings) and the sacroiliac standing flexion test indicative of hypermobility--did not meet the minimum criteria as essential to the diagnosis of sacroiliac hypermobility. For these findings, the upper three quartiles, or 75% of the responses, ranged from 2 to 5. Three findings were considered to be essential to this diagnosis by more than one expert: (1) unilateral weight bearing aggravates pain (n=4), (2) accessory range of motion indicates hypermobility (n=7), and (3) positive sacroiliac stress tests (n=10).

Usefulness of the Survey

The mean rating of the overall usefulness of conducting the expert survey for determining the level of agreement on diagnostic classes of LBP and relevant clinical findings was 4.3 on a five-point scale. This indicated that most experts completing the survey felt that it was a useful exercise.

Discussion

There is substantial literature available related to the pathophysiology, epidemiology, and clinical findings of various diagnostic classes and syndromes of LBP. No quantitative analysis of the relative importance or acceptance of these classes among physical therapy experts has been reported. This survey provides information as to the agreement on terminology related to classification of LBP and the degree to which a group of physical therapy experts agree on signs and symptoms essential to each diagnostic class. The experts agreed on the appropriateness of 19 diagnostic categories, and an additional 6 categories were suggested by more than one expert. No classification system in the literature we reviewed included the full spectrum of classes that experts agree are appropriate. Although this study was an initial step toward documenting agreement on diagnostic classes and essential findings relevant to each class, there are several limitations to the survey.

The selection of experts was not representative of the physical therapy community at large. There is a lack of consensus on specific standards or criteria for defining "expert orthopedic physical therapists." Our criteria for selection of experts were designed to achieve a minimum standard of expertise in orthopedic physical therapy, but our definition and criteria could be challenged. In addition, all experts meeting these criteria were not surveyed.

The survey was more representative of physical therapy experts in Canada than any other country, because these experts contributed to 75% of the response pool. Several other countries were included in order to examine geographical differences, although no such differences were found on any survey item.

An attempt was made to be comprehensive in the diagnostic classes included in the survey; however, six classes suggested by more than one expert were not included. A lack of internationally accepted terminology and the choice of terminology for items in the survey could have negatively biased the results. Experts could, for example, be disagreeing on the terminology used, rather than on the existence of the clinical entity. In a survey such as ours, a bias may exist in which respondents agree or disagree in general with the content of the survey, rather than making judgments on individual items. The range of responses for all experts was from 1 to 5, indicating that such a bias was not occurring.

We found the highest levels of agreement on appropriateness for inclusion and essential clinical findings for classes that can be defined by specific diagnostic tests, such as blood tests or radiographs. These classes included fracture, tumor, Scheurmann's disease, and spondylolisthesis. They may be less relevant to physical therapy practice than many classes surveyed because of the lack of suitability of physical therapy intervention in these classes. Physical therapists, however, are responsible for the recognition of signs and symptoms of patients with diagnoses not amenable to physical therapy management, such as fracture and tumor. These diagnoses may subsequently be confirmed by investigations not within the scope of physical therapy practice.

In the literature reviewed, many categories of back pain that cannot be proven by diagnostic test procedures (eg, radiography) have been identified but poorly documented as to essential clinical features. The exception is McKenzie's extensive descriptions of the derangement, dysfunctional, and postural syndromes.[5] Using the concept of essential, or critical, clinical features of diagnostic classes will improve our diagnostic accuracy and reduce confusion when several classes coexist.

Three diagnostic classes particularly relevant to physical therapy practice had high levels of agreement as to their appropriateness as a class as well as in their respective constellations of clinical findings: hypomobility dysfunction, nerve root adhesion, and sacroiliac hypermobility. These three classes, therefore, were not only agreed upon as appropriate for inclusion in a diagnostic classification scheme for LBP, but physical therapy clinicians and researchers will be able to use agreed-upon guidelines for identifying LBP in these subgroups. Through continued critical evaluation of the clinical picture and response to treatment of patients with LBP, clinicians will assist in further defining these groups. Similar clinical course and response to intervention in a subgroup lends verification to the notion that the subgroup is a homogeneous diagnostic class. This process of confirming the existence of a group of similar patients through their response to treatment is termed prescriptive validity.[24] This concept will be a useful process for clinicians and researchers in confirming appropriate subgroups, particularly when specific diagnostic tests are not available.

Diagnostic classes that had poor agreement require further investigation. Possible reasons for lack of agreement on these classes include confusion regarding terminology and lack of criteria for distinguishing closely related diagnoses. It may also be that the class is not clinically distinguishable and should be subsumed under another category.

The clinical implications of the survey results for physical therapy are numerous. Common terminology for subgroups of patients with IBP will enhance communication between patients and physical therapists and among health care professionals. These survey results provide an initial step in establishing essential clinical features for each diagnostic class. Lists of clinical features that identify distinct subgroups of patients will continue to be modified by clinicians as further information and diagnostic techniques become available. Physical therapists should critically evaluate diagnostic categories not agreed upon by the experts, as these categories may be subsumed under another more appropriate category or may be difficult to establish because of lack of documented signs and symptoms. A standardized diagnostic scheme for LBP will allow physical therapists to develop management approaches specific to subgroups of patients with LBP and to critically evaluate the effect of intervention on different subgroups.

Conclusions

This survey has established the level of agreement among expert orthopedic physical therapists on diagnostic classes of LBP and associated clinical findings. Much of the current terminology for identifying subgroups of patients with LBP implies a known pathological or anatomical cause, such as posterolateral disk derangement. For most of these classes, however, the specific pathological or anatomical cause of LBP cannot be validated with current diagnostic procedures. Determining levels of agreement on patient presentation, or constellations of clinical findings, is the most important outcome of this survey and most critical to development of homogeneous subgroups on which patient management decisions and effectiveness trials may be based.

Current work is focused on collecting data on patients with LBP using the Toronto-Hamilton lumbar database to identify proportions of patients in each diagnostic class and clinical findings that describe each subgroup for comparison with the results of the expert survey. In addition, the reliability with which physical therapists use the classification system is being examined. Sensitivity and specificity of the classification system need to be determined. Validity will be sought using the concept of establishing prescriptive validity.[24] Effectiveness trials will be used with the objective of determining whether outcomes of physical therapy interventions differ across our diagnostic classes of patients with LBP.

Acknowledgments

We gratefully acknowledge the participation, expertise, and thoughtful suggestions of the following expert physical therapists: Wendy Aspinall, Bert Chesworth, Peter Edgelow, Brian Edwards, David Egan, Robert Elvey, Margo Fanaki, Maureen Hunt, Anne Kennedy, Mark Laslett, Dianne Lee, Therese Lord, Catherine Macleod, David Magee, Elaine Maheu, Geoffrey Maitland, Robin McKenzie, Laurie McLaughlin, Jim Meadows, John Oldham, Beverly Padfield, Stanley Paris, Erl Pettman, Lance Twomey, and Riki Yamada. The physical therapists at the Orthopaedic and Arthritic Hospital in Toronto and at the Chedoke Division of Chedoke-McMaster Hospitals and St Joseph's Hospital in Hamilton are acknowledged for their ongoing commitment to this project.

References

[1] Frymoyer JW. Magnitude of the problem. in: Weinstein JN, Wiesel SW, eds. The Lumbar Spine. Philadelphia, Pa: WB Saunders Co; 1990:32-38. [2] Chapman J, Saarinen H, Solomon P, et al. Back to backs: an approach to designing a curriculum. Presented at the World Confederation for Physical Therapy; London, England; July 1991. [3] Sahrmann SA. Diagnosis by the physical therapist--a prerequisite for treatment: a special communication. Phys Ther 1988;68:1703-1706. [4] Delitto A, Shulman AD, Rose SJ. On developing expert-based decision-support systems in physical therapy: the NIOSH low back atlas. Phys Ther. 1989;69:554-558. [5] McKenzie RA. The Lumbar Spine: Mechanical Diagnosis and Therapy. Waikanae, New Zealand: Spinal Publications; 1981. [6] Sypert GW. Low back pain disorders. Trans Assoc Life Insur Med Dir Am. 1988;71:174-197. [7] Spitzer WO, coordinator. Scientific approach to the assessment and management of activity-related spinal disorders: report on the Quebec Task Force on Spinal Disorders. Spine. 1987; 12:9-54. [8] Finch F, Binkley J. The Toronto-Hamilton Lumbar Database and Classification System for Low-Back Pain Patients. Toronto, Ontario, Canada: University of Toronto Press; 1990. [9] International Classification of Impairments, Disabilities and Handicaps Geneva, Switzerland: World Health Organization; 1980. [10] Mooney V. The syndromes of low back disease. Orthop Clin North Am. 1983;14:505-515. [11] Deyo RA. Early diagnostic evaluation of low back pain. J Gen Intern Med. 1986;1: 328-338. [12] Saunders HD. Classification of musculoskeletal spinal conditions. Journal of Orthopaedic and Sports Physical Therapy. 1979;3: 3-15. [13] Porter RW. Mechanical disorders of the lumbar spine. Ann Med. 1989;21:361-366. [14] Fast A. Low back disorders: conservative management. Arch Phys Med Rehabil. 1988;69: 880-891. [15] Bernard TN, Kirkaldy-Willis WH. Recognizing specific characteristics of nonspecific low back pain. Clin Orthop. 1987;217:266-280. [16] Levine A. A model for health projections using knowledgeable informants. World Health Stat Q. 1984;37:306-317. [17] Couper MR. The Delphi technique: characteristics and sequence model. Adv Nurs Sci. 1984;7:72-77. [18] Milholland AV, Wheeler SG, Heieck JJ. Medical assessment by a Delphi group opinion technic. N Engl J Med. 1973;288:1272-1275. [19] Chaney HS. Needs assessment: a Delphi approach. Journal of Nursing Staff Development. 1987;3:48-53. [20] Goodman CM. The Delphi technique: a critique. J Adv Nurs. 1987; 12:729-734. [21] Romm FJ, Hulka BS. Developing criteria for quality of care assessment: effect of the Delphi technique. Health Serv Res. 1979;14: 309-312. [22] Scheibe M, Skutsch B, Schofer J. Experiments in Delphi methodology. in: Linestone H, Turoff M, eds. The Delphi Method.. Techniques and Applications. London, England: Addison-Wesley Publishers Ltd; 1975. [23] McGill R, Tukey JW, Larsen WA. Variations of box plots. Amer Stat. 1978;32:12-16. [24] Task Force on Standards for Measurement in Physical Therapy. Standards for tests and measurements in physical therapy practice. Phys Ther. 1991;71:589-622.

Commentaries

Following are three commentaries on "Diagnostic Classification of Patients with Low Back Pain: Report on a Survey of Physical Therapy Experts"

Binkley et al have addressed several important issues relevant to future considerations for physical therapy management of patients who have low back pain (LBP). They are to be congratulated for furthering this important discussion, because it is apparent that sweeping changes loom in regard to management strategies addressing the epidemic of LBP.

The first basic premise stated by the authors is of critical importance. The development of a standardized diagnostic classification system is essential. They rightly point out that efficacy studies that simply compare one treatment technique with another without regard to specific subgroups of patients with LBP are illogical. Such studies make the assumption that all LBP has the same cause and characteristics. Clinicians would not tolerate studies that compared two different treatment techniques for knee pain" unless such studies attempted to clearly distribute patients with knee pain into similar subgroups. Binkley et al suggest that we utilize the same such standard for the low back and hence strive to pursue a diagnostic classification system. In a previous article,[1] we had suggested subgroups based on the pathomechanics gleaned through the evaluation. Regardless of one's bias, it is clear that the profession must develop some standardized classification system.

The literature review clearly identifies the dilemma for the clinician--multiple and loosely related diagnostic groupings. This fact has hampered the opportunity for clinicians to communicate among each other in regard to patients with LBP and limits the ability to test the effects of therapeutic intervention.

The authors determined that a rating of 3 or greater constituted agreement of diagnostic classifications. This appears generous, especially when the rating of 3, as viewed by its placement in Appendix 1, does not compel the rater to the side of appropriate or not appropriate, but allows a more neutral position to be taken on the question. This neutrality is then factored into "constituted agreement" and potentially obscures the actual meaning of the results.

Clearly, the agreement on diagnostic classifications was highest in those classes that featured specific sophisticated laboratory tests. Once these first six classes are removed from further consideration, we are left with the more difficult to define classes. It might also be argued that it is precisely these next classifications that are most often seen in the physical therapy clinic and hence should be the focus of future discussion. Although the intent of the survey

approach was to measure agreement on a diagnostic classification for LBP, I vas left with the impression that the survey was asking less about a diagnostic classification than a clinical finding. I believe that there is an important difference between the two. The authors' diagnostic classification of hypomobility dysfunction illustrates this difference. Although having consensus as to the essential clinical findings for such a diagnosis, the fundamental question remains unanswered regarding its correlation with the patient's problem This is akin to the recognized shortcomings of using the findings of radiographs to establish the diagnosis of degenerative joint disease of the spine and making the assumption that is the cause of the patient's low back complaint. Perhaps the additional finding noted as essential for the diagnostic classification of sacroiliac hypermobility illustrates why this correlation is essential to consider. Ten of the 24 respondents (without cue from the written survey) independently noted that positive sacroiliac stress tests were essential for this classification. I interpret this to mean that they are correlating the patient's complaint with the clinical finding, which should be the ideal goal with each classification and finding. This shortcoming aside, Binkley et al have made an important initial step in at least helping us determine whether agreement among clinicians is realistic in the first place.

It was revealing that chronic pain syndrome did not meet the agreement criteria for diagnostic classes, and I would have been interested in the essential diagnostic findings suggested by the authors. Perhaps this points out an important limitation of physical therapists that needs to be addressed. The physical therapist must become aware of the current medical strategy of identifying, as early as possible, the potential patient with chronic pain syndrome or those patients with a propensity toward disability. The skill to identify this diagnostic class is essential because a disproportionate percentage of patients (10%) account for 80% of the expenditures.[2,3] Although early identification appears to be a daunting task, the literature clearly points out the necessity of developing such clinical skills. It has been suggested, for example, that there may be four different dimensions of pain behavior: (1) marked distortion in ambulation or posture, (2) a negative affect, (3) facial or audible expressions of distress, and (4) the avoidance of activity.[4] Perhaps increased attention in the future needs to be placed on identification of this important diagnostic classification.

Future physical therapy research for LBP will most likely be focused primarily on efficacy of treatment and patient outcomes, because society will place increasing demands on accountability. For this to occur, however, a common terminology among clinicians must be used. Research on LBP is complicated by the fact that no two syndromes are exactly the same and the patient's subjective response to the pain is unique. It is essential, therefore, that clinicians at least develop a commonality of clinical language that allows for efficacy to be scientifically evaluated rather than relying wholly on anecdotal reports. In this regard, Binkley and colleagues' contribution to the literature is both timely and relevant.

Carl P Derosa, PT Associate Professor and Chairman Physical Therapy Program Northern Arizona University Box 15105 Flagstaff, AZ 86011 [1] Derosa CP, Porterfield JP. A physical therapy model for the treatment of low back pain. Phys Ther. 1992;72:261-272. [2] Cats-Baril WL, Frymoyer JW. Identifying patients at risk of becoming disabled due to low back pain: the Vermont Rehabilitation Engineering Predictive Model. Spine. 1991;16:605-607. [3] Bigos S, Spengler D, Martin N, et al. Back injuries in industry--a retrospective study, III: employee-related factors. Spine. 1986;11:252-256. [4] Turk DC, Wack JT, Kerns RD. An empirical examination of the "pain behavior" construct. J Behav Med. 1985;8:119-130.

Numerous physical therapy treatments have been advocated for the patient with low back pain (LBP). An important barrier to determining the effectiveness of these treatments is the lack of a valid classification system for patient's with LBP. Binkley et al have presented important and timely data that will assist clinicians and researchers toward overcoming this barrier. Several "expert" orthopedic therapists have agreed on categories by which to classify patients and on the essential findings that must be present for this classification. In their conclusions, the authors caution us that in many of the classes, the "specific pathological or anatomical cause of LBP cannot be validated with current diagnostic procedures." I agree with this statement and would like in my discussion to focus on the "clinical meaningfulness" of the classification scheme that was agreed on by the experts who were surveyed.

A fundamental goal of the treatment of patients with musculoskeletal disorders is to identify the structural abnormality causing the patient's impairment or disability. This knowledge allows the therapist to develop a mechanical basis for the patient's problem, to identify which tissues to treat, and also to predict outcomes. Relative to the extremity joints, this goal typically can be achieved with our current diagnostic procedures. The identification of structural problems that cause LBP, however, is problematic. The use of the pathological model as a means of predicting LBP may not be valid for several reasons.[1-3] The spine is a multisegmental system in which tissues work together, and presumably are injured together; thus, identifying one culpable tissue (eg, disk or muscle) may be inadequate.[3,4] Many radiographic and magnetic resonance imaging (MRI) findings from the lumbar spine correlate poorly with LBP.[5-8] In contrast to the extremity joints, most spinal structures are not palpable. This creates difficulty when attempting to identify painful tissues during a physical examination.[9] Additionally, when one examines the spine, there is no "good side" for comparison. Finally, the clinical presentation of a person with LBP is often greatly influenced by psychosocial factors.[2] Considering this, the classification scheme agreed on by the experts surveyed in this study generates two questions: (1) Within the limits of our diagnostic abilities, how reliably can clinicians and researchers classify patients with LBP? and (2) Presuming that the patients are reliably classified, does this identify a causal relationship for their LBP?

The first issue relates to the reliability of classifying individuals based on criteria that are considered to be essential by the experts. Figure 1 identifies those diagnostic classes that meet the agreement criteria. Those diagnostic classes that describe a specific pathology (eg, fractures, spinal tumors, Scheurmann's disease, spondylolisthesis, spinal stenosis, ankylosing spondylitis, scoliosis, visceral disorders, congenital anomalies, infections) can generally be achieved reliably by a combination of medical imaging procedures such as radiographs, bone scans, and MRI, as well as by hematological data. Other categories, however, require more qualitative-based decisions on the part of the examiner. For example, hypomobility dysfunction, segmental hypermobility, and sacroiliac hypermobility and hypomobility require that the accessory motion be either less than or greater than "normal." Unfortunately, "normal accessory motion is not operationally defined for the lumbar vertebral joints or the sacroiliac joints.[10] The movement that occurs at these joints varies greatly because of the high incidence of congenital anomalies, as well as because of age, gender, and body type. The degree of reliability by which practicing clinicians can determine the presence of abnormal vertebral accessory motion in patients with LBP has not been established. Thus, the decision to classify an individual as having abnormal accessory motion is subjective. Although the changes of the pain behavior with activity would appear to be easier to quantify, it is interesting that the experts weighted the results of accessory motion testing or sacroiliac kinetic tests as more essential.

The second critical issue relating to the classification scheme is the question: Does the presence of the diagnostic criteria infer a causal relationship with LBP? As one evaluates the diagnostic classes meeting the agreement criteria, several issues arise that relate to the ambiguity of the classes and to their causal relationship to LBP. For example, the presence of a spinal fracture can be highly related to a patient's problem such as a recent compression fracture of the vertebral body of L-1, or it may rarely be related to LBP such as pars defect without spondylolisthesis.[5,6,11] The radiographic finding of spondylolisthesis has a causal relationship with LBP in extreme cases only.[5,6,11] How should a person who has a radiographic finding of spondylolisthesis, but who otherwise meets the essential criteria for postural syndrome, be classified? Spinal stenosis presents many variations, depending on the location of the stenotic changes.[12] Nerve root irritation is a common result of spinal stenosis. How does one classify this patient? It is unclear what magnitude of a scoliotic curve would result in LBP. The experts agreed that on observable scoliosis with radiographic corroboration is essential, yet there was no mention of the degree of curvature. If a person has a 10-degree scoliosis as determined by Cobb angle measurements, is this the cause of his or her pain? The classification of disk herniation as it is described is problematic. The finding of "positive neurological signs" does not exclusively define the presence of a herniated disk. Mooney and Robertson[13] reported diminished deep tendon reflexes following the injection of lumbar facet joints in patients with LBP. Conversely, the presence of a herniated disk does necessarily result in symptoms.[7,8] Finally, spinal congenital anomalies are very common, and only in rare cases are they the source of pain.[3,5,6,11]

Each day clinicians and researchers are faced with the need to classify patients with LBP. The diagnostic classifications based on physical findings have uncertain reliability, and the diagnostic classifications that are based on radiographic findings have limited validity relative to identifying a causal relationship with a patient's LBP. Factors such as multiple or inconsistent symptoms add further confusion. Binkley et al have given provocative suggestions for further study, which have been in part generated by this study. The information from the Toronto-Hamilton database will be invaluable to our profession.

Paul Beattie, PhD, PT, OCS Physical Therapist Department of Rehabilitation Services University Hospital University of New Mexico 2211 Lomas NE Albuquerque, NM 87131

References

[1] Haldeman S. Presidential address, North American Spine Society: Failure of the pathology model to predict back pain. Spine. 1990; 15:718-724. [2] Waddell G. A new clinical model for the treatment of low back pain. In: The Lumbar Spine. Philadelphia, Pa: WB Saunders Co; 1990: 38-51. [3] Nachemson AL. Advances in low back pain. Clin Orthop. 1985;200-266-278. [4] DeRosa CP, Porterfield JA. A physical therapy model for the treatment of low back pain. Phys Ther. 1992;72:261-272. [5] Magora A, Schwartz A. Relation between low back pain and x-ray changes: lysis and olisthesis. Scand J Rehabil Med. 1980;12:47-52. [6] White AA, Panjabi MM. Clinical Biomechanics of the Spine. Philadelphia, Pa:JB Lippincott Co; 1978. [7] Paajanen H, Erkintalo M, Timo K, et al. Magnetic resonance study of disc degeneration in young low-back pain patients. Spine. 1989;14: 982-985. [8] Powell MC, Wilson M, Szypryt P, et al. Prevalence of lumbar disc degeneration observed by magnetic resonance in symptomless women. Lancet. December 1986:1366-1367. [9] McCombe PF. Reproducibility of physical signs in low-back pain. Spine. 1989;14:908-918. [10] Riddle DL. Measurement of accessory motion: critical issues and related concepts. Phys Ther. 1992;72:65-874. [11] Splithoff CA. Lumbosacral junction: roentgenographic comparison of patients with and without backaches. JAMA. 1953;152:1610-1613. [12] Anderson GBJ, McNeill TW. Definition and classification of spinal stenosis. In: Andersson GBJ, McNeill TW, eds. Lumbar Spinal Stenosis. St Louis, Mo: Mosby-Year Book Inc; 1992:9-16. [13] Mooney V, Robertson J. The facet syndrome. Clin Orthop. 1976;115:149-156.

The current lack of uniformity in diagnostic terminology for spinal disorders is both a major barrier and a key challenge to reversing the epidemic problem of low back pain (LBP). Often considered the first important step in the management of the patient with LBP, the task of making a diagnosis is currently characterized by uncertainty and variability according to the idiosyncracies of the practitioner. This initial source of error is only compounded further as cases proceed through the various steps of treatment and reassessment.

Binkley and co-workers have exhibited considerable foresight and an equal amount of courage in facing this challenge in their article. This report describes a first step toward a comprehensive strategy to the management of LBP by unifying the way in which spinal disorders are classified. The authors present an effective and cogent argument that such a system can benefit clinical decision making and the conduct of scientific research. We can further appreciate this work by keeping in mind that the authors are active in formulating a comprehensive evaluation system for patients with LBP based on standards of clinical presentation and functional outcome.

This article focuses on the development of the first dimension of classification (ie, the physical diagnosis). The authors have chosen to develop the diagnostic dimension by surveying a small group of orthopedic physical therapy "experts" to identify appropriate diagnostic terminology and the clinical signs and symptoms that are effective in selecting each diagnostic category. I found it helpful when reading this article to remind myself that the authors are only presenting the results of one initial step in formulating a classification system. As such, we cannot in fairness criticize the overall system until it is more fully developed. I do not believe we can even assume that the diagnostic categories or clinical findings that meet the minimal agreement criteria in this study will necessarily comprise the final diagnostic categories of their evaluation system. I base this on the authors' description of future work that will compare the results of their survey with a patients database. Other work is also planned to address such issues as the sensitivity and specificity of their classification system, its reliability, and its prescriptive validity. Definitive diagnostic entities, therefore, will ultimately be generated by these appropriate research efforts.

This study stimulates thought and raises questions in my mind regarding the nature of the LBP dilemma and how physical therapists can contribute to more effective management of this syndrome. A leading expert in this field and strong proponent for reform stated in a recent essay on LBP that "the ambiguities and difficulties of studying low back pain have made it an |orphan condition' both with regard to clinical care and research."[1] Physical therapists are in a unique position to "adopt" this "orphan" when one considers that as many as 40% of all out-patient treatments have been estimated to be for spinal pain disorders. We are further compelled to act by reports that have independently concluded on the basis of reviewing the scientific literature that physical therapy is of questionable long-term value for LBP. Clinical trials for LBP treatment are becoming more prevalent; however, many researchers, including myself, would agree that these efforts could be more effective if methodological research to develop accurate diagnostic and prognostic classifications were available. The same can be said for developing more innovative outcome assessments, particularly in terms of functional ability.

In the interim, before such definitive diagnostic categories are validated, researchers and clinicians should avail themselves of classification schemes that describe subjects according to simple clinical criteria encountered in practice, such as medical history and physical findings, rather than presupposing a pathological entity. The classification scheme developed by the Quebec Task Force on Spinal Disorders[2] and mentioned by Binkley and co-workers is an excellent example of a well thought out and useful procedure that includes simply derived categories based on medical history, clinical examination, and response to treatment. Information clearly related to treatment choice and outcome, such as the LBP duration and work status, are also included. Similar to the plan recommended by Binkley et al, assessment must include a functional index. This can be achieved by selecting from among several currently available functional disability scales that have been tested among the spinal pain population.[3]

Although Binkley et al limited their survey to physical therapists, it might be even more useful to not limit a classification scheme to a single profession, or even, as in this case, to a specific subpopulation of the profession (ie, mostly Canadian orthophedic "specialists" in physical therapy). If the goal is to have a unified terminology and classification scheme, then it should be possible to make that scheme include categories acceptable to other medical professions who have a role in assessing and treating patients who have LBP. This short-coming of their report is identified to some extent by the authors when they describe their attempts to included therapists outside of Canada. There simply were not enough numbers from these other countries, however, to make an argument for much diversity in this group. The results of the survey further point to a homogeneous group that is likely to have similar specialized training in manual therapy procedures, judging from the high number of categories included in the questionnaire that describe segmental and sacroiliac instabilities and the fact that they were usually sources of high agreement.

It would be interesting to know more about the backgrounds of these therapists and how representative their philosophies of treatment are compared with the physical therapist population in general. The need is for a universal classification system, rather than one reflecting a particular discipline or school of thought within the profession. We should at the same time not confuse the need for a more unified approach to this problem with the interpretation that the different schools of treatment should not have their own classification schemes to guide their specific therapies. This is perfectly acceptable if it helps the care provider implement the modalities. When reporting the results of one's intervention to the medical community, the health care provider, or the research community, however, a more universally accepted classification based on clinical and functional criteria is preferred.

Because the work of Binkley and others will likely be a relatively long-term commitment before definitive results can be made available to current practitioners, more surveys are needed among physical therapists as well as other care providers to identify how physical therapists are currently assessing LBP and to identify what forms of treatment are being used and in which combination. It would also be useful to know how well be accept the theoretical basis for these approaches and, similar to the survey conducted by Binkley et al, which clinical findings are most likely to guide these treatments. It would also be of immediate practical value to survey the profession to rate the acceptability or relevance of currently available comprehensive classification systems. One of the shortcomings of the survey conducted by Binkley et al is that the diagnostic terms selected for survey may have received different agreement scores if they had been presented in the context of a particular classification scheme rather than as isolated terms. For instance, the term "chronic pain behavior" might not have been rejected as a diagnostic term had it been presented as a necessary part of a particular classification system.

A final thought prompted by this article is how we, as a profession, can succeed in implementing standardization to assessment and measurement. The fact that so few people currently attempt to use scales that result from even the best of research efforts is cause for concern. Part of the solution must come from our professional educational system to be more innovative in teaching awareness of health care issues and from our support of professional organizations to formulate policy.

Serge H Roy, ScD, PT Research Assistant Professor NeuroMuscular Research Center Boston University 44 Cummington St Boston, MA 02215

References

[1] Deyo RA. Low back pain. In: Max M, Portenoy R, Laska E, eds. Advances in Pain Research and Therapy. New York, NY: Raven Press; 1992;18:291-303. [2] Spitzer WO, coordinator. Scientific approach to the assessment and management of activity-related spinal disorders; report on the Quebec Task Force on Spinal Disorders. Spine. 1987; 12:9-54. [3] Millard RW. A critical review of questionnaires for assessing pain-related disability. J Occup Rehabil. 1991;1:289-302.

Author Response

We thank Mr DeRosa, Dr Beattie, and Dr Roy for their insightful comments and suggestions, and we appreciate having the opportunity to discuss the issues raised in their commentaries.

The commentary by Mr DeRosa raises two critical issues. The first is that of the relationship of diagnosis to the patient's low back complaint. Patients will often have essential signs and symptoms suggestive of more than one diagnostic class. A classic example of this is the patient with a spondylolisthesis as revealed by radiography and a disk herniation. Determining the source of the patient's complaint is part of the clinical decision-making process. In this clinical case, the orthopedic physical therapist or orthopedic surgeon will have little difficulty in making a judgment using clinical test procedures together with knowledge of anatomy and pathology. In many cases, however, this judgment is not as straightforward. The decision on the most relevant diagnosis may be a factor that separates the novice and the experienced clinician. Our goal in the development of a diagnostic system was to provide standardized terminology from which the health care professional could select one or more diagnostic classes relevant to the patient. We have not yet dealt with the reliability among therapists in selecting diagnoses or identifying the one that is the cause of the patient's complaint.

The second issue raised by Mr DeRosa relates to the classification of chronic pain syndrome. The fact that it did not meet the agreement criteria may represent the difficulty that many of us have in objectively identifying and offering management strategies for patients with chronic low back pain (LBP). DeRosa points out several factors identified in the literature as important in determining chronic pain behavior. We agree that we must work to improve our methods of identifying these patients.

The issue raised by Dr Beattie of the reliability of classifying patients is critical. We see this as a two-step process. First, the reliability of individual diagnostic test procedures considered essential to a given diagnosis should be determined. Dr Beattie correctly points out that the reliability of determining the presence of abnormal vertebral accessory motion in patients has not been established. In conjunction with our ongoing LBP research project, we have completed two studies on the reliability of measurements of lumbar flexion and extension mobility and accessory motion testing, and our reports on those studies are currently in the review process. There is a substantial amount of work remaining to be done in establishing the reliability of test procedures used in the assessment of patients with LBP.

The next stage planned in determining the reliability of categorizing patients is to examine the process in a realistic clinical setting. The agreement among physical therapists on the diagnostic class(es) of patients must be determined. In addition, agreement on identification of the critical diagnosis related to a patient's complaint must be determined. We have not done work as yet in this area but would like to make a personal observation. Interrater agreement on a patient's diagnostic class may be better than on individual test procedures. In our experience, physical therapists tend to agree more on a overall clinical picture following a complete assessment that on an overall clinical picture following a complete assessment than on an isolated test result such as segmental mobility. This may be related to the multiple cues available and the similarity to clinical practice in the first situation. This informal observation remains to be tested.

Dr Beattie also raises the issue of the relationship of diagnostic class to the patient's complaint. We hope that this issue has been addressed adequately in our response to Mr DeRosa's commentary.

Dr Roy emphasizes the need for working to develop a standard classification system for patients with LBP. His reference to Deyo's description of LBP an an "orphan condition" because of its ambiguities is apt. We have worked as a clinical research team in the area of LBP for several years and often wonder why we adopted this problem child. We have found that dealing with the problem of LBP clinically and from a research standpoint is fraught with complexities. That it is an orphan is no wonder!

We can only reinforce Dr Roy's observation that the diagnostic system is in its infancy and several important steps are currently in progress and planned for the future. We are currently examining the frequency of diagnostic classes identified in a physical therapy outpatient setting. The relationship between the essential signs and symptoms agreed upon in the survey and those being used to identify diagnostic classes in a clinical setting is being evaluated using our Toronto-Hamilton database. We have also continued work on the measurement of function in patients with LBP. The results of some of our work on the responsiveness of several currently used functional scales will be published in Spine in the near future by Stratford and collegues.

Diagnostic classes for LBP should be based on pathological dysfunction or sets of signs and symptoms rather than related to treatment approaches. Various health care professionals may have different, but equally acceptable, approaches to treatment of a particular diagnostic class. We agree with Dr Roy that a diagnostic classification system for patients with LBP could be standard across professions involved with the assessment and treatment of patients with LBP. This is reflected in our inclusion of classes such as spinal stenosis as well as hypomobility dysfunction. Many classes will be "profession-specific" because of diagnostic procedures available to the examiner and interventions appropriate to the class. Communication and, hence, patient care will be enhanced significantly by terminology that is common to all health care professionals dealing with patients with LBP.

In selecting experts, approaches such as those described by Cyriax, Maitland, McKenzie, Kaltenbourne, and others are represented. We made a commitment to maintain the anonymity of specific expert respondents at the outset of our survey. This may have jeopardized our credibility in stating that all experts are well versed in one or more of these schools of thought. many of the Canadian experts have studied and taught internationally. We are confident that the experts surveyed represent a variety of well-known schools of thought in orthopedic physical therapy internationally.

We considered presenting a classification scheme rather than isolated diagnoses to the experts in our study. This was rejected because of our concern that experts might reject the diagnostic scheme outright but not necessarily the individual categories. This could result in a wasted opportunity to acquire data on individual classes and related essential signs and symptoms. We chose to organize the agreed-upon classes into a meaningful system following the survey.