 |
 |
 |
|
| home | CME | help | feedback | subscriptions | archive |
|
© 1999 The Journal of Bone and Joint Surgery, Inc. Relationship Between Rates and Outcomes of Operative Treatment for Lumbar Disc Herniation and Spinal Stenosis   Investigation performed at Maine Medical Assessment Foundation, Manchester
Background: Population-based variations in rates of operations for the treatment of lumbar disc herniation and spinal stenosis are well known. This variability may occur in part because of differences in the threshold at which physicians recommend an operation, reflecting uncertainty about the optimum use of an operative procedure. To the best of our knowledge, no previous reports have indicated whether differences in population-based rates of operative treatment are associated with patient outcomes. Methods: The Maine Lumbar Spine Study is an ongoing prospective study of 655 patients who had a herniated lumbar disc or spinal stenosis. The patients were enrolled by their physicians, who provided baseline demographic and treatment-related data. The patients completed baseline and follow-up questionnaires that focused on symptoms, function, satisfaction, and quality of life. Small-area variation analysis was used to develop three distinct so-called spine service areas in Maine. The outcomes (usually at four years; minimum, two years) were compared among these areas, in which a total of 250 patients had been managed operatively and had answered questionnaires. Results: Population-based rates of operative treatment derived from statewide data that had been collected over five years in the state of Maine ranged from 38 percent below to 72 percent above the average rate in the state (a greater than fourfold difference). The outcomes for the patients who had been managed by surgeons in the lowest-rate area were superior to those for the patients in the two higher-rate areas. Seventy-nine percent (fifty-seven) of seventy-two patients in the lowest-rate area had marked or complete relief of pain in the lower extremity compared with 60 percent (eighteen) of thirty patients in the highest-rate area. The improvements in the Roland disability score (p  0.01), quality of life (p 0.01), and satisfaction (p 0.05) were significantly greater among the patients in the lowest-rate area. The patients in the higher-rate areas generally had less severe symptoms and findings at baseline than those in the lowest-rate area did.
Conclusions: Higher population-based rates of elective spinal operations may be associated with inferior outcomes. This variability is possibly related to differences in physicians' preferences with regard to recommending an operation and in their criteria for the selection of patients. Physicians cannot assume that their outcomes will be the same as those of others, and therefore they need to evaluate their own results.
Since publication of the original paper describing small-area variations by Wennberg and Gittelsohn15, many authors have confirmed their seminal findings that population-based rates of hospitalization and procedures vary widely by geographic region and cannot be fully explained by differences in demographic features6,13. Despite the large number of reports on this topic in the literature, variations in practice patterns remain a major feature of medical care in the United States16. Simple counts of hospital admissions or procedures, or both, do not provide information about the rates of services among discrete populations. Small-area analysis provides a method with which to calculate population-based utilization rates. Variations in utilization of services may be quite appropriate. For example, if there are differences in the incidence of the disease or in environmental or genetic factors, requirements for medical services may differ from one geographic area to another. However, among similar populations, large-scale differences in the rates of services do not appear logical. Research has indicated that the practice styles and treatment preferences of individual physicians are major factors in these variations6. Although there appears to be broad agreement that variations in utilization among similar populations are inappropriate, there is no empirical evidence to indicate which of the many different rates is the so-called right rate. Analysis of variations demonstrates the inconsistency of many aspects of medical practice but provides little evidence to assist physicians in determining the right rate. A solid and consistent body of scientific literature could guide doctors and patients in the medical decision-making process. However, the clinical literature is replete with publications advocating a wide range of treatments and procedures, leaving clinicians free to choose the ones that they prefer. In addition, scientific analyses of the literature (meta-analyses) indicate weak scientific evidence for many conclusions and recommendations5,12. Driven in part by variations in practice patterns, patient-oriented clinical outcomes research has gained importance in recent years. The goal is to conduct high-quality, methodologically rigorous, prospective clinical studies that focus on outcomes of care that matter to patients7. It is reasoned that, if valid information can be provided to physicians and their patients, more consistent treatment choices will be made, variations will diminish, and outcomes will improve. To our knowledge, analyses of the relationship of the rates of operative procedures with the outcomes of those procedures have not been previously published. The goal of this study was to integrate the results of small-area utilization analyses in Maine with data from a prospective effectiveness study of the outcomes of treatment for herniated lumbar intervertebral discs and spinal stenosis. A number of studies have indicated a positive association between the volume of procedures performed by surgeons or within hospitals and various outcomes. Such an association between volume and outcome has been demonstrated in studies of cardiac operations, percutaneous coronary angioplasty, and total knee arthroplasty4,9,10. These analyses focused on the concept of practice makes perfect—that is, when complex procedures are performed more often by surgeons, the outcomes may be better. Our analysis focused on the rate of lumbar discectomy—that is, the number of procedures performed on the population of a defined geographic region in a given time-period. Within a given region, the volume of procedures performed by individual physicians or hospitals is independent of the population-based rate.
Small-Area Analysis This analysis resulted in spine service areas incorporating one to seventeen normal hospital service areas. One spine service area is made up of only its own hospital service area, whereas another area consists of sixteen hospital service areas in addition to the original hospital service area (Fig. 1). It is important to note that all procedures were allocated to the spine service area in which the patient resided, regardless of where the operation was performed. This process is necessary in order to create population-based rates of operations. In this study, all of the operations performed on residents of the spine service area were performed in institutions within that area.
Rates of lumbar discectomy and decompression laminectomy were calculated for each of the spine service areas. These were adjusted for age and gender with an indirect method. Three of the five spine service areas had a sufficient number of patients who were participants in the Maine Lumbar Spine Study to permit an outcome analysis (Fig. 2).
Outcomes Study Patients were recruited as part of normal practice, and they were followed prospectively with standardized questionnaires. Baseline information that was supplied by physicians included demographic data, details of physical examination and imaging findings, and clinical impressions. Details of the operative procedure and any complications have been reported separately3,8. When possible, imaging studies were obtained and subjected to blinded interpretation by a neuroradiologist8. Initial interviews with the patients were conducted at the patients' homes by trained interviewers. Because the patients who were enrolled in the study represented only a portion of the patients who were managed by the participating physicians, a random sample of patients who were not enrolled were sent a follow-up questionnaire in order to compare their baseline characteristics and outcomes with those of the enrolled patients8. Follow-up questionnaires were mailed to the patients at three, six, and twelve months and annually thereafter. Of the 279 operatively managed patients who were assigned to one of the three spine service areas, 250 (90 percent) completed at least one follow-up questionnaire between two and four years, and the data from those questionnaires are included in this analysis. When more than one follow-up questionnaire had been completed, the data from the most recent survey were used. The baseline characteristics and features of the patients included demographic information, work and disability status, previous episodes of back pain, other comorbid conditions, symptoms, and functional status2. The frequency of pain in the low back and the lower extremity (with a score of 0 points indicating "not at all" and a score of 6 points indicating "always") and the extent to which these symptoms were bothersome (with a score of 0 points indicating "not bothersome" and a score of 6 points indicating "extremely bothersome") were assessed at baseline. The pain in the low back and the lower extremity at baseline was categorized as severe if the patient reported that it was "usually" to "always" present and was "extremely bothersome" (a score of 5 or 6 points). A symptom frequency index ranging from 0 to 24 points was created by summing the responses to four questions regarding the lower extremity: pain in the lower extremity; numbness or tingling in the lower extremity, foot, or groin; weakness in the lower extremity or foot; and pain in the back or lower extremity while sitting (or while walking, for patients who had stenosis)11. For all scales, higher scores indicated more severe symptoms. Baseline measures of functional status included the back-specific modified Roland disability scale11 and the generic Medical Outcomes Study (MOS) 36-Item Short-Form Health Survey (SF-36)14. The modified Roland scores range from 0 to 23 points, reflecting a simple sum of items checked as present2. Four of eight domains of health included in the SF-36 are reported: physical function, bodily pain, and mental health and general health perceptions14. Each domain is scored from 0 (poor health) to 100 points (optimum health). Higher SF-36 scores reflect better function, but higher Roland scores reflect worse function. Outcome measures emphasized a spectrum of patient-reported health-related quality of life, including symptoms (pain), functional status (physical, general health, employment, and disability), and satisfaction2,11. These outcome measures have been previously validated11, and many have been incorporated into the current MODEMS (Musculoskeletal Outcomes Data Evaluation and Management Systems) questionnaire1 of the American Academy of Orthopaedic Surgeons, the North American Spine Society, and the Council of Musculoskeletal Specialties. Pain in the low back and the lower extremity was considered to be decreased if the patient reported that the symptom was "better" to "completely gone" compared with the baseline status. Changes in the symptom frequency index and the modified Roland scores were calculated by subtracting the baseline score from the score at the time of follow-up. General health status was assessed with use of the SF-36. Work and disability status was ascertained at each follow-up evaluation. In this analysis, patients who had operative management for a lumbar disc herniation or spinal stenosis were allocated to the appropriate spine service area. Our hypothesis was that outcomes would vary according to the population-based rate of operative treatment in the spine service area and that higher rates of operative treatment would result in less optimum outcomes. To assess for differences in baseline and outcome variables among the spine service areas—which were ranked as low-rate (area A), middle-rate (area B), and high-rate (area C)—the Mantel-Haenszel chi-square test of trend was used for dichotomous variables, with linear regression models for continuous variables. Additional analyses with the Cochran-Mantel-Haenszel test of trend were performed to examine whether outcomes across the spine service areas were different among specific subgroups. The subgroups were based on diagnosis (sciatica or stenosis), baseline disability compensation status (receiving or not receiving compensation), and duration of follow-up (four years compared with two or three years). The patients who were receiving disability compensation at baseline had worse outcomes regardless of the spine service area (data not shown). However, the relationships among outcome variables and spine service area were not different for any subgroup that was examined and, thus, only the overall results for each spine service area are presented. Independent predictors of the patients' satisfaction with their current state at the time of follow-up were examined with use of multiple logistic regression analysis. The patients who had a score of 1 point ("delighted"), 2 points ("pleased"), or 3 points ("mostly satisfied") on a 7-point scale, with "delighted" and "terrible" as the two end points, were classified as satisfied. The models included spine service area and other baseline characteristics that were potentially associated with the satisfaction of the patient (at p values of 0.20)—that is, education, disability compensation, smoking status, comorbid illnesses, duration of the current episode, severity of pain in the low back and lower extremity, and back-specific and generic measures of functional status. To account for associations among patients within each physician cluster, we used the generalized estimating equations methodology to estimate the covariance structure of the models17. All statistical analyses were performed with use of a commercial software package (Statistical Analysis System, SAS Institute, Cary, North Carolina) on a UNIX workstation (Sun Microsystems, Mountain View, California).
Outcomes of the Maine Lumbar Spine Study
Small-Area Analysis
Relationship Between Patient Outcomes and Operative Rates
The patients enrolled in the study represented only a portion of all of the patients who were managed by the participating physicians. We previously reported that 22 percent (655) of 2977 eligible patients who had an operation were referred to the study8. The rates of referral in the three spine service areas were comparable with the overall referral rate. Referred and nonreferred patients had similar demographic and clinical characteristics.
The physicians were generally optimistic about the expected outcome of the procedure, although the physicians in area A were less likely than the physicians in the other areas to believe that the patients' quality of life would be greatly improved (p Because different population-based rates of operative treatment imply that some patients who were managed operatively in one area may have been managed nonoperatively in another, we examined whether there were differences in baseline characteristics or outcomes among those managed nonoperatively in the different spine service areas. One hundred and twenty-four patients (thirty-six in area A, seventy in area B, and eighteen in area C) were managed nonoperatively. Regardless of the spine service area, the nonoperatively managed patients had less severe baseline symptoms and functional impairment than did the operatively managed patients, and the outcomes of the nonoperative treatment did not differ among the spine service areas (data not shown).
Patient Characteristics and Clinical Features at Baseline Evaluation
There were a number of differences among the spine service areas with regard to the characteristics of the lumbar conditions at baseline (Table III). Among the patients who had sciatica, those in area C were somewhat less likely to have a positive straight-leg-raising test (p = 0.06). However, among all patients, those in area C were somewhat more likely to have other abnormal physical findings. The physicians in area C reported more abnormal motor, sensory, and reflex findings, with area B having the least and area A having an intermediate amount of such reported findings. In contrast to the findings on physical examination, the severity of the abnormalities as assessed by the independent examination of the advanced imaging studies was the least for the patients in area C and the greatest for the patients in area A (p 0.01).
There were some differences in patient-reported symptoms across the three service areas. The percentage of patients who reported current symptoms that had lasted for less than six months decreased from 72 percent (fifty-seven of seventy-nine) in area A to 61 percent (eighty-two of 134) in area B to 49 percent (seventeen of thirty-five) in area C (p 0.05). The percentage of patients who had severe low-back pain was highest in area C and lowest in area A (p 0.1). In contrast, the percentages of those who had severe pain in the lower extremity and those who had unilateral symptoms in the lower extremity were highest in area A and lowest in area C (p 0.1). The percentages of patients who reported the symptoms in area B were generally intermediate between the percentages in the other two areas. With the numbers available, we were unable to detect significant differences in the symptom frequency index or in other measures of functional status, including the back-specific modified Roland disability scale and the generic SF-36 domains of physical function, bodily pain, and mental and general health perceptions (p > 0.1 for all except mental health perception, for which p = 0.08).
Patient-Reported Outcomes at the Time of Follow-up
There was a significantly greater decrease in the frequency of symptoms in the lower extremity (as measured by the symptom frequency index) between the baseline and follow-up evaluations in area A, the area with the lowest utilization rate (p 0.05). Although relief of pain in the low back and the lower extremity and relief of the predominant symptom (pain in either the low back or the lower extremity) were not significantly different among the spine service areas, the trends for these variables also favored the area with the lowest utilization rate. For example, pain in the lower extremity was rated as much better by 79 percent (fifty-seven) of the seventy-two patients in area A, 74 percent (ninety-five) of the 128 in area B, and 60 percent (eighteen) of the thirty in area C (p = 0.06). In addition, the back-specific functional status as assessed with the Roland scale improved significantly more in area A compared with area C (p 0.01). The general health status as assessed with the SF-36 was similar across the three spine service areas.
Patient-reported satisfaction with outcome was assessed with three questions regarding improvement in quality of life, the patients' satisfaction with their current state, and whether they would choose the operation again. Regardless of the specific question, the patients in the area with the lowest utilization rate (area A) were most satisfied with their care, the patients in the highest-rate area (area C) were least satisfied, and the patients in area B had a rate of satisfaction that was between the rates in the two other areas. Overall, 72 percent (fifty-seven) of the seventy-nine patients in area A were satisfied with their current state compared with 63 percent (eighty-five) of the 136 patients in area B and 49 percent (seventeen) of the thirty-five patients in area C (p As noted previously, the patients who were referred to the study represented only a portion of the potentially eligible patients and a survey of nonreferred patients was performed. Overall, with the numbers available, no significant differences in outcomes could be detected between the referred and nonreferred patients or among the nonreferred patients in the three spine service areas (data not shown)2.
Independent Baseline Predictors of Patient Satisfaction at the Time of Follow-up
To our knowledge, the present analysis represents the first attempt to associate differences in population-based rates of operative treatment with patient-reported outcomes of that treatment. We observed that better operative outcomes were associated with lower population-based rates of operative treatment. It is noteworthy that, when the Maine Lumbar Spine Study is considered in its entirety, the outcomes of operations for lumbar disc herniation and lumbar spinal stenosis are very favorable. The patients who had an operation had had more severe symptoms and findings at the time of enrollment than those who were managed nonoperatively, and at the time of follow-up they had significantly better outcomes2,3. For example, compared with baseline values, pain in the lower extremity at one year had decreased in 81 percent (170) of 209 patients who had had a disc excision compared with 56 percent (ninety-two) of 165 patients who had been managed nonoperatively for lumbar disc herniation. The predominant symptom was much better or gone in 71 percent (148) of the 209 patients who had had a disc excision compared with 43 percent (seventy-two) of the 167 nonoperatively managed patients. These findings remained essentially the same at four years (unpublished data). The results similarly favored operative management for patients who had spinal stenosis. Thus, with the results viewed globally, an operation could be seen as the preferred method of management for patients who have severe sciatica due to herniation of a lumbar disc or those who have spinal stenosis. However, not all of the patients who were managed operatively had the same outcome. Indeed, the outcomes of operative treatment were associated with the population-based rates of operative treatment. The patients from the area with the lowest rate of operative treatment (area A) had outcomes that were even better than those for the group as a whole, whereas patients from the highest-rate area (area C) had outcomes that were inferior to those of the entire group. The results in area B fell between those for the other two areas. These findings may be partly explained by differences in the baseline findings among the three groups. Patients in the lowest-rate area reported symptoms that were of a shorter duration, were more likely to be unilateral, and were generally more severe in the lower extremity and less severe in the back; they also had more severe findings on independently evaluated imaging studies. In the highest-rate area, surgeons recorded that the straight-leg-raising test was positive in 59 percent (seventeen) of twenty-nine patients compared with 79 percent (131) of 166 patients in the other two areas. Thus, although all patients met the same criteria for study eligibility, differences in baseline symptoms among the patients managed in the three spine service areas suggest that the superior outcomes in the lowest-rate area may be related to more stringent criteria for selection of patients for a procedure on the lumbar spine. We have previously shown that an operation has the least benefit in patients who have the lowest baseline severity of disease2. We have also demonstrated that patients who are receiving Workers' Compensation at baseline have inferior outcomes and are less likely to return to work2. We believe that the differences in the baseline characteristics of the patients in the three areas relate more to the decision-making patterns of the physicians in each of the areas than to differences in the socioeconomic status or employment patterns of the local populations. There is no evidence that surgeons in any of the areas were operating on patients who did not have evidence of disc herniation or spinal stenosis. Rather, surgeons in the two higher-rate areas may have had a lower threshold for recommending an operation to their patients than did surgeons in the lowest-rate area. The stability of the area-based operative rates over time (Fig. 2) indicates that these local practice styles are entrenched and that physicians remain consistent with regard to how they make recommendations and manage patients. Surgeons who perform a high rate of operations do not believe that they are practicing inappropriately or differently from others, and they have no way of knowing that their practice patterns are quite different from those of colleagues in the same state. Our results highlight several generic deficiencies in the manner in which physicians evaluate the outcomes of the patients whom they manage. First, physicians and surgeons are generally unaware of the rate of utilization in the area in which they practice because small-area variation analyses are not routinely available in most states and regions. Statewide hospital-discharge databases are required for this purpose. These are currently available in more than thirty states, but small-area analyses are infrequently performed. This problem has recently become exacerbated with the shift of many lumbar-disc procedures from the inpatient to the ambulatory setting, where utilization information is even less frequently collected and analyzed. Second, physicians do not routinely receive information about the patient-oriented outcomes of the care that they provide. The analysis that we have presented indicates that the outcomes of operations on the lumbar spine vary significantly among patients residing in different spine service areas in Maine. Before presentation of the results of the current study, surgeons in each of the three areas believed that they operated on patients with similar indications and that they produced similar outcomes. All surgeons tend to function under these beliefs. They are similarly trained, are usually board-certified, and have strong beliefs in their own judgment and skills. In the absence of other information, it is not unreasonable that they would be confident that their diagnostic and operative skills result in a level of utilization and quality of care that are equal to those of their peers. Our research indicates that individual physicians cannot assume that their treatment results will be the same as those of their peers or the same as the findings or recommendations published in the literature. Even the findings of a study in which they participated cannot be applied equally to all of the surgeons who enrolled patients. Patients are not likely to be aware of or to appreciate the implication of small-area analysis and variation in population-based rates of operative treatment. The probability of having an operation is highly dependent on the practice style of the surgeons in the area in which the patient lives. This is because patients tend to follow the advice of their physician—a phenomenon described as delegated decision-making. In areas where surgeons are predisposed to recommend an operation more frequently, more patients will have an operation. Outcomes research has the potential to provide information that will enable each patient to better understand the outcomes, risks, and benefits of an operation and other treatment. An operation for the treatment of lumbar disc herniation or spinal stenosis is almost always elective, as it is performed to decrease symptoms and improve function. Patients need to understand the risks and benefits of such an operation so that they can make better decisions about what works best for them. Many of them will elect to proceed with an operation and will do well, whereas others may choose watchful waiting and may have an equally good outcome. The response of Maine's spine surgeons to the analyses that we have presented is illuminating. Rather than dispute the findings and implications of these data, the surgeons asked what they could do to rectify the problem. The Maine Medical Assessment Foundation has responded by continuing to conduct feedback of the information from small-area variation analyses. Additionally, most of the practices from all five spine service areas are participating in a voluntary outcomes data management program, directed by the Maine Medical Assessment Foundation. Patients who are evaluated because of back problems (lumbar and cervical) are enrolled prospectively. The MODEMS spine questionnaires are used in this project1. Individual doctors will be able to compare their outcomes with those of their colleagues and peers. As a result of these efforts, participating physicians have involved themselves in a continuous quality-improvement effort in which valid and useful data are collected, analyzed, and fed back to the involved doctors on a continuous basis.
Limitations Our research did not lead us to conclude that high rates of lumbar disc excision for sciatica or laminectomy for spinal stenosis always produce outcomes that are inferior to those associated with lower rates. Although this was the case in the present study, additional research is needed to determine whether there is a generally optimum operative rate and whether this rate differs in various regions. In conclusion, this community-based effectiveness outcome study demonstrated that higher population-based rates of lumbar discectomy and laminectomy were associated with inferior patient-oriented outcomes. Based on these findings, we make two recommendations. First, it is important that physicians be provided with accurate and timely comparative data on the population-based rates of the operative and medical treatments that they are using. Although small-area analysis does not establish the so-called right rate of medical care, it does provide useful information about variations in practice patterns. These studies can provide information about potential overutilization and underutilization and can provide a stimulus for further evaluation and analysis. Second, surgeons and other physicians need to begin to collect information about the patient-oriented outcomes of the care that they provide. Only this level of rigorous self-analysis can provide data that will enable continuous improvement in quality. With high-quality outcomes data, doctors and their patients will be able to make better decisions about the best options for care. Outcomes information will allow surgeons to understand how their results compare with those of their colleagues and will give them the tools with which to improve when that is necessary.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. Funds were received in total or partial support of the research or clinical study presented in this article. The funding sources were Grants HS 06344 and HS 08194 from the Back Pain Outcome Assessment Team of the Agency for Health Care Policy and Research.
#Department of Medicine and Health Services, University of Washington, Seattle, Washington 98195.
  CiteULike  Connotea  Del.icio.us  Facebook  Technorati  Twitter What's this?
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Abstract
Introduction
| home | CME | help | feedback | subscriptions | archive | site index |
|
Stanford University Libraries' HighWire Press (TM) assists in the publication of JBJS Online. |