Family Practice Advance Access published online on June 13, 2008
Family Practice, doi:10.1093/fampra/cmn031
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Predictive value of self-reported patient information for the identification of lumbar spinal stenosis
a Department of Epidemiology and Health Care Research, Graduate School of Medicine and Public Health, Kyoto University, Kyoto, Japan
b Department of Orthopedic Surgery, Fukushima Medical University, Fukushima, Japan
c Institute for Health Outcomes and Process Evaluation Research, Kyoto, Japan
Correspondence to Yasuaki Hayashino, Department of Epidemiology and Health Care Research, Graduate School of Medicine and Public Health, Kyoto University, Yoshida-Konoe-cho, Sakyo-ku, Kyoto 606-8501, Japan; Email: hayasino-y{at}umin.net
Received 10 October 2007; Accepted 18 May 2008.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionDeclarationAppendix. Questionnaire used to...References |
|---|
Background. To our knowledge, no objective criterion has been identified for the diagnosis of lumbar spinal stenosis (LSS) and no study has evaluated the predictive value of self-reported patient information for the identification of LSS.
Objective. To develop and validate a prediction rule for the identification of LSS based on self-reported patient information alone.
Methods. Prospective derivation study using a coefficient-based multivariable logistic regression scoring method with internal validation with primary care clinics and orthopaedic departments of medical centres, as well as university and other hospitals. Participants were consecutive patients with primary symptoms of pain or numbness in the lower extremities. Physician-diagnosed LSS was the main outcome measure.
Results. Of 468 patients included in the analysis, 47.3% were diagnosed with LSS and divided into derivation and validation sets. The following items were retained at the conclusion of the derivation process: age (<60, 60–70 and >70), duration of symptoms over 6 months, symptom improvement when bending forward, symptom improvement when bending backward, symptom exacerbation while standing up, intermittent claudication and urinary incontinence. To derive a risk score for each patient, integer-based scores were assigned and summed. In the validation data sets, prevalence of LSS in patients from the first to fourth risk score quartile were 13.3%, 47.6%, 55.2% and 65.5%, respectively. Further, the likelihood ratio in the low-risk category was 0.154.
Conclusions. We developed a prediction rule for the identification of LSS based on self-reported patient information alone. Further, the likelihood ratio in the low-risk category was sufficiently low. This rule may be used for screening of LSS.
Keywords. Lumbar spinal stenosis, prediction rule, self-reported patient information.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationAppendix. Questionnaire used to...References |
|---|
Lumbar spinal stenosis (LSS) is a condition caused by compression of the cauda equina or spinal nerve roots.1 LSS may lead to substantial functional disability, intermittent claudication or vesico-rectal disturbance and an associated decrease in quality of life and increase in social burden.2,3 Although the effectiveness of surgical treatment remains unclear due to limited scientific evidence, early diagnosis and identification of patients needing treatment is reported to improve the outcome of LSS.4–6 However, the most common initial symptoms of LSS, namely leg symptoms, are frequently misdiagnosed.3,7 Data on the population-based epidemiology of LSS is relatively limited possibly due in part to the difficulty of diagnosing LSS. In particular, because computed tomography and magnetic resonance imaging (MRI) are often non-specific8–10 and therefore not sufficiently reliable for diagnosis, no objective criteria for diagnosis are available. Although self-reported patient information, including medical history, symptoms and signs, is a necessary component of the clinical diagnosis of LSS, its predictive value has not been extensively evaluated.
Given these conditions, the clinical prediction rule (CPR), which consists in calculating integer-based scores from information on patient symptoms and physical examination results, is a useful tool for predicting the probability of a disease.11 Evaluation of the diagnostic performance of CPR enables the estimation of the predictive value of patient information as a group of items. Further, the predictive value of self-reported patient information in the diagnosis of LSS may potentially facilitate the conduct of community-based screening, as well as the estimation of LSS prevalence or incidence.
Here, the predictive value of self-reported patient information was evaluated for the identification of LSS in patients with leg symptoms by developing and validating the test performance of a CPR composed of self-reported patient information alone.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationAppendix. Questionnaire used to...References |
|---|
Subjects
The target population was identified as patients visiting primary care clinics, which were recruited for the study. Selection was limited to consecutive patients showing primary symptoms of pain or numbness in the lower extremities, including the buttocks, thighs and lower legs, aged more than 20 years, and able to visit the clinic alone without assistance. Patients having visited other hospitals or clinics for the above symptoms within the year before study participation, as well as those showing severe psychiatric disorders, namely dementia, were excluded. If necessary, the study was approved by the institutional review board of each study institution, and written informed consent was obtained from all patients.
Data collection
Data collection from patients was planned in two steps. First, eligible patients completed a self-administered questionnaire at the primary care clinics before consulting the attending physician to avoid incorporation bias. Second, patients were referred to orthopaedic departments in hospitals and medical history, physical examination, lumbar X-ray and MRI results were recorded by the orthopaedic staff physicians.
Questionnaire
The questionnaire queried patients on the association between their symptoms and posture (lying in bed, sitting on a chair, standing upright, bending forward or backward), as well as activities (walking and riding a bicycle). An additional question enquired about intermittent claudication: ‘If your symptom occurs while walking, does it improve by resting?’. Other items included age, gender, time of symptom occurrence, symptom history, treatment history, sensory or motor disturbances in the legs, cauda equina syndrome (urinary or faecal incontinence, dysuria, urinary retention, nocturia and penile erection while walking), smoking habit, alcohol drinking and comorbidities. Items from the Japanese version of the modified Roland–Morris Disability Questionnaire12 were also included by altering the expression ‘low back pain’ to ‘pain or numbness in the lower extremities’.
LSS diagnosis process
The association between questionnaire responses and LSS diagnosis was prospectively evaluated. For all patients, orthopaedic staff physicians in each institution recorded medical history and performed a physical examination, as well as lumbar X-ray and MRI in accordance with a standardized protocol. The medical history included the type and distribution of symptoms (e.g. leg and lumbar pain), postures attenuating or exacerbating symptoms, as well as comorbidities, including diabetes and peripheral artery disease. Physical examination included the ankle pressure index and various tests designed to elicit dysfunction in the lumbopelvic region.13 To avoid verification bias, all patients then underwent lumbar X-ray and MRI. Clinical and diagnostic test information was recorded by the attending physician in a standardized form, which was then sent for diagnosis and information verification to the study coordinator, who is an experienced orthopaedic surgeon. In the absence of a universally accepted reference standard for LSS, studies show that the opinion of expert clinicians provides a reasonable method for establishing clinical diagnosis.14 In addition, this approach has been used in the development of classification criteria for rheumatic diseases, which similarly to LSS cannot be defined by single laboratory measurements.15,16
The flow of final patient diagnosis is detailed in Figure 1. First, the attending orthopaedic physician reached a clinical diagnosis based on history taking, examination and radiographic findings. The study coordinator then verified diagnosis using copies of the clinical information and radiographic images. Interobserver agreement between the attending orthopaedic physician and study coordinator was assessed by calculating agreement ratios and kappa values.17
|
Substantial discrepancy between diagnoses [interobserver agreement rate on LSS diagnosis: 60.8%, kappa value: 0.261 (95% confidence interval, CI, 0.185–0.336)] were resolved in a second diagnostic step using a consensus panel, composed of 10 orthopaedic physicians with specialist clinical experience in LSS, in which all members were either professors or associate professors at university hospitals or the head of the orthopaedic department at teaching hospitals in Japan. For each discrepant case, panel members arrived at a four-scale score on the probability of LSS (lowest = 1 and highest = 4) based on clinical information and imaging studies. Scoring was performed without consultation with other panel members. Mean scores for each patient were then calculated, with a mean score of 3 or more confirming LSS, whereas a mean score of 2 or less was not considered as LSS. For cases with mean scores between 2 and 3, the consensus panel members carefully discussed the discrepancy and achieved final diagnosis. Cases not having reached consensus were removed from analysis.
Data analysis
Derivation.
Before beginning data analysis, all observations were randomly assigned to derivation and validation sets at a 4 to 1 ratio. The univariate relationships between LSS and each item in the questionnaire from the derivation set were evaluated by univariate logistic regression analysis, and an odds ratio was generated. All items with a P-value of less than 0.05 in the univariate analysis were entered into a multiple logistic regression model using stepwise model selection. In addition, two items with a P-value of more than 0.05 but considered clinically important were also entered: ‘diabetes as comorbidity’, due to diabetic neuropathy being an important differential diagnosis of LSS, and ‘symptom improves when bending backward’, considered critical in LSS diagnosis. Only variables with a P-value of less than 0.05 were kept in the final model. Model calibration was evaluated using the Hosmer–Lemeshow chi-square statistic.18
Development of CPR. A score-based prediction rule for the final diagnosis of LSS was developed for each step based on results from multivariable logistic regression equations using a regression coefficient-based scoring method. To generate a simple integer-based point score for each predictor variable, scores were calculated by dividing the β-coefficient by the half-sum of the two smallest coefficients in the model, followed by rounding up to the nearest integer. To achieve a value of 1 for the variable with smallest coefficient score, the denominator was selected. The overall risk score for each patient was calculated by summing up the scores for each variable.19 After the total risk score was calculated, patients were further categorized into total risk score quartiles, and their observed probabilities of LSS compared. Discriminatory performance of the rule was assessed by calculating the area under the receiver operating characteristic (ROC) curve.20 Sensitivity and specificity were calculated at the cut-off score point of 5, which divides the second and third quartiles.
Validation of CPR. The prediction rule in the validation set was validated, and patients were categorized in risk score quartiles using the same derivation process. The observed probabilities of LSS in patients were compared and likelihood ratios calculated. Discriminatory performance of the rule was assessed by calculating the area under the ROC curve. Sensitivity and specificity were calculated at the cut-off score point of 5. All statistical analyses were performed using Stata, version 9.2 (Stata Corp., College Station, TX).
Sample size. Sample size was determined to enable the detection of a significant difference in LSS prevalence between patients clinically showing an improvement in symptoms when bending forward and those showing no improvement. LSS prevalence in groups showing ‘symptom improvement’ and ‘no symptom improvement’ was 60% and 41%, respectively, which revealed that a sample size of at least 308 patients was needed to reach a statistical power of 0.90 at a 0.05 significance level (two-sided test).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationAppendix. Questionnaire used to...References |
|---|
Patient characteristics
From December 2002 to December 2004, a total of 469 patients were evaluated by 104 orthopaedic surgeons in 22 clinics and 50 hospitals in various sites around Japan. Ages ranged from 20 to 96 years, with a mean age of 65.2 years, and 54.2% of patients were male. Although recruitment of all patients was planned to be carried out at primary care settings, protocol violation occurred in some institutions, in which patients were recruited from the orthopaedic department of the hospital. However, recruitment of at least one-third of patients at primary care clinics was clearly confirmed. Of the 469 participants, diagnosis from the two observers was consistent in 226, of whom 126 were diagnosed with LSS. Of the 243 discrepant cases, only nine cases' mean scores were between 2 and 3, and thus were discussed by the consensus panel, of which agreement was reached in 8. In total, the consensus panel discussion diagnosed 96 cases with LSS. The interobserver agreement rate between attending physicians and expert panel members was 85.7%, with a kappa value of 0.71 (95% CI, 0.621–0.799). In contrast, the agreement rate between the study coordinator and expert panel members was 67.5%, with a kappa value of 0.36 (95% CI, 0.272–0.445).
The consensus panel failed to reach agreement in one case only, which was removed from analysis, leaving 468 cases. Results showed a 47.3% prevalence of LSS in the patient sample group. Other diagnoses included lumbar disc herniation (17.7%), diabetic neuropathy (2.8%) and peripheral artery disease (8.3%). In the remaining patients (23.7%), no specific diagnosis other than ‘not LSS’ was determined (Table 1).
|
Derivation. From the total patient sample, 374 and 94 patients were selected as the derivation and validation set, respectively. First, 26 variables with a P-value less than 0.05 were identified (Table 2). Second, the following variables with P-values less than 0.05 were retained in the multivariable model as independent predictors: age more than 60 years (with a higher risk for more than 70 years), symptom present for more than 6 months, symptom improvement when bending forward, symptom improvement when bending backward, symptom exacerbation while standing up, intermittent claudication and urinary incontinence (Table 3). For the final model, results show a Hosmer–Lemeshow statistic of 2.47 (P = 0.9632), which indicates the predicted probabilities of LSS statistically agreed with the observed frequencies.18,21
|
|
Development of CPR. An integer-based score, derived in the second step from the β-coefficient, is shown in Table 3 and Appendix, from which a mean overall risk score of 5.05 (range: –2 to 10) was obtained for each patient in the derivation set. These patients were categorized into risk score quartiles: the first (Q1), second (Q2), third (Q3) and fourth (Q4) quartiles were defined by risk scores of 2 or less, 3–4, 5–6 and 7 or more, respectively. Of the 374 patients in the derivation set, Q1 showed a 17.7% (9/51) probability of LSS, whereas Q2, Q3 and Q4 showed 25.3% (25/99), 50.8% (62/122), and 77.5% (79/102), respectively. Table 4 shows the model performance indices. For the derivation set, calculations show an area under the ROC curve of 0.77 (Fig. 2). Sensitivity and specificity at the cut-off score point of 5 were 0.81 and 0.58, respectively. Further, the likelihood ratio in Q1 of the derivation set was 0.24.
|
|
Validation of CPR. In the validation set, results show a mean overall risk score for each patient of 5.21 (range: 0–10). Of the 94 patients in the validation set, Q1 showed a 13.3% (2/15) probability of LSS, whereas Q2, Q3 and Q4 showed 47.6% (10/21), 55.2% (16/29) and 65.5% (19/29), respectively (Fig. 3). In addition, calculations show an area under the ROC curve of 0.67 (Fig. 2), indicating a slight decrease in performance compared with the derivation set. However, no statistical significance was observed between the area under the ROC curves of the derivation and validation sets (P = 0.125). Sensitivity and specificity at the cut-off score point of 5 were 0.75 and 0.51, respectively. Further, the likelihood ratio in Q1 of the validation set was 0.15.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationAppendix. Questionnaire used to...References |
|---|
Summary of main findings
To our knowledge, this study is the first to evaluate the predictive value of self-reported patient information for the identification of LSS in patients with leg symptoms. Results show that the likelihood ratio in the low-risk category was sufficiently low to allow patients with a low-risk score to be excluded from diagnosis (Table 4). Although based on self-reported information, items in the present study are closely consistent with those previously reported as associated with LSS diagnosis.1,22
Evaluation of predictive value of patient histories
CPRs usually consist of items related to medical history, physical information and diagnostic tests. To diagnose physical conditions, rules which use only self-reported information from patients are rare. Although evaluated as a group of items in the present study, medical history as a diagnostic tool is often assessed as an independent item. For example, the Journal of the American Medical Association publishes a periodical feature entitled the ‘Rational Clinical Examination Series’,23 in which the predictive value of clinical information is evaluated, such as that of symptoms for the diagnosis of heart failure or influenza.24,25 A limitation of the individual evaluation of every piece of clinical information is the overpowering sensitivity or specificity for single items due to the dependence of final diagnosis on information from a number of diverse sources.24 In the present study, the value of self-reported information from patients was evaluated as a group of items, which may potentially improve the value of clinical information in use.
Cauda equina syndrome
An additional item, urinary incontinence, is also contained in the present prediction rule. A great number of patients presenting with an advanced form of LSS manifest varying degrees of bladder dysfunction, including urinary incontinence,26 with more severe clinical signs of cauda equina syndrome in patients with neuropathic bladder than in those without.27 Early diagnosis and decompression was reported to improve the outcome in patients with cauda equina syndrome, which includes bladder dysfunction28. Given this finding, the careful history and physical examination of patients with suspected cauda equina syndrome is recommended.
Application for primary care practice
A CPR composed only of self-reported patient information may be useful for a variety of purposes. First, with the recent ageing population, the lack of epidemiological data on LSS may hamper appreciation of its importance in policymakers. Given the ease of use and distribution of self-reported information in the community, our tool may be used to estimate the incidence or prevalence of LSS in the general population. Second, this method may be used by patients with leg symptoms for self-screening of LSS. An ecology of medical care study revealed that only a small portion of patients reporting symptoms visit a hospital.29 Given this observation, our tool may facilitate the decision to seek medical care for patients. Third, because of the simplicity of our tool, assessing the need to refer patients to an orthopaedic specialist may be facilitated for non-orthopaedic specialists. Wider application of our tool is predicated on its validation in a variety of settings.
Limitations
Although our method shows potential in its ease of use, several limitations should be noted. First, no consensus has been reached on an objective standard for LSS diagnosis, which essentially remains clinical. In the absence of explicit criteria, however, expert opinion is a reasonable strategy for the diagnosis of clinical syndromes14 and has been previously used in a number of disorders.15,16 In the present study, inconsistent diagnosis from two observers were resolved by a consensus panel composed of 10 orthopaedic specialists, who may in turn qualify the reference standard used to diagnose LSS.
Second, because a number of patients were recruited from hospitals, LSS prevalence in the present study appears higher than that using patients recruited from primary care clinics. Unlike in the UK, no official gatekeeping system by GPs is available in Japan, with patients free to choose hospitals or clinics when seeking medical care.30 A study showed that among 1000 Japanese residents, 232 visited a primary care physician and 88 a hospital-based outpatient clinic as a first visit in a month.31 These findings show that in some situations, hospitals play a role as primary care providers. Despite this situation, further investigation is necessary to validate our rule in other primary care populations.
Third, the likelihood ratio in the high-risk category was low, which suggests that this rule is possibly weak for a definite diagnosis of LSS. However, especially in primary care settings, a screening tool is crucial to eliminate the possibility of specific diseases in diagnosis. Further, the likelihood ratio in the low-risk category was sufficiently low, which implies that this rule useful in safely excluding LSS patients.
Conclusion
The present study is the first to report a prediction rule for LSS identification based only on self-reported patient information. Results show that the likelihood ratio in the low-risk category was sufficiently low. Our rule may be used for LSS screening and is suggested to improve the quality of LSS diagnostic practice in primary care.
|
Declaration |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationAppendix. Questionnaire used to...References |
|---|
Funding: Institute for Health Outcomes and Process Evaluation Research and the Japanese Society for Spine Surgery and Related Research.
Ethical approval: This study was approved by the institutional review board of each study institution.
Conflicts of interest: None.
|
Appendix. Questionnaire used to identify patients with possible LSS among those with pain or numbness in the lower extremities (including thigh, buttock and lower legs) |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationAppendix. Questionnaire used to...References |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||
|
Notes |
|---|
Sugioka T, Hayashino Y, Konno S, Kikuchi S and Fukuhara S. Predictive value of self-reported patient information for the identification of lumbar spinal stenosis. Family Practice 2008; Pages 1–8 of 8.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationAppendix. Questionnaire used to...References |
|---|
1 Katz JN, Dalgas M, Stucki G, et al. Degenerative lumbar spinal stenosis. Diagnostic value of the history and physical examination. Arthritis Rheum (1995) 38:1236–1241.[CrossRef][Web of Science][Medline]
2 Stucki G, Liang MH, Lipson SJ, Fossel AH, Katz JN. Contribution of neuromuscular impairment to physical functional status in patients with lumbar spinal stenosis. J Rheumatol (1994) 21:1338–1343.[Web of Science][Medline]
3 Goldman SM, Funk JD, Christensen VM. Spinal stenosis. A common cause of podiatric symptoms. J Am Podiatr Med Assoc (1997) 87:117–124.[Abstract]
4 Gibson JN, Wadell G. Surgical interventions for lumbar disc prolapse. Cochrane Database of Systematic Reviews (2007) issue 2. Art. No.: CD001350. [Review].
5 Malmivaara A, Slatis P, Heliovaara M, et al. Surgical or nonoperative treatment for lumbar spinal stenosis? A randomized controlled trial. Spine (2007) 32:1–8.[CrossRef][Web of Science][Medline]
6 Weunstein JN, Lurie JD, Tosteson TD, et al. Surgical versus nonsurgical treatment for lumbar degenerative spondylolisthesis. N Engl J Med (2007) 356:2257–2270.
7 Goh KJ, Khalifa W, Anslow P, Cadoux-Hudson T, Donaghy M. The clinical syndrome associated with lumbar spinal stenosis. Eur Neurol (2004) 52:242–249.[CrossRef][Web of Science][Medline]
8 Wiesel SW, Tsourmas N, Feffer HL, Citrin CM, Patronas N. A study of computer-assisted tomography. I. The incidence of positive CAT scans in an asymptomatic group of patients. Spine (1984) 9:549–551.[CrossRef][Web of Science][Medline]
9 Boden SD, Davis DO, Dina TS, Patronas NJ, Wiesel SW. Abnormal magnetic-resonance scans of the lumbar spine in asymptomatic subjects. A prospective investigation. J Bone Joint Surg Am (1990) 72:403–408.
10 Powell MC, Wilson M, Szypryt P, Symonds EM, Worthington BS. Prevalence of lumbar disc degeneration observed by magnetic resonance in symptomless women. Lancet (1986) 2:1366–1367.[Web of Science][Medline]
11 Tu JV, Naylor CD. Clinical prediction rules. J Clin Epidemiol (1997) 50:743–744.[CrossRef][Web of Science][Medline]
12 Suzukamo Y, Fukuhara S, Kikuchi S, et al. Validation of the Japanese version of the Roland-Morris Disability Questionnaire. J Orthop Sci (2003) 8:543–548.[CrossRef][Medline]
13 Waddell G, Somerville D, Henderson I, Newton M. Objective clinical evaluation of physical impairment in chronic low back pain. Spine (1992) 17:617–628.[Web of Science][Medline]
14 Feinstein AR. Clinical Epidemiology: The Architecture of Clinical Research (1985) Philadelphia, PA: WB Saunders.
15 Katz JN, Liang MH. Classification criteria revisited. Arthritis Rheum (1991) 34:1228–1230.[Web of Science][Medline]
16 Arnett FC, Edworthy SM, Bloch DA, et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum (1988) 31:315–324.[Web of Science][Medline]
17 Chalmers I. Underreporting research is scientific misconduct. JAMA (1990) 263:1405–1408.
18 Hosmer D, Lemeshow S. Assessing the fit of the model. In: Applied Logistic Regression—Wiley J, ed. (1999) New York. 135–175.
19 Leung GM, Rainer TH, Lau FL, et al. A clinical prediction rule for diagnosing severe acute respiratory syndrome in the emergency department. Ann Intern Med (2004) 141:333–342.
20 Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology (1982) 143:29–36.
21 Amar D, Shi W, Hogue CM Jr, et al. Clinical prediction rule for atrial fibrillation after coronary artery bypass grafting. J Am Coll Cardiol (2004) 44:1248–1253.
22 Szpalski M, Gunzburg R. Lumbar spinal stenosis in the elderly: an overview. Eur Spine J (2003) 12:S170–S175.[CrossRef][Web of Science][Medline]
23 Williams JW, Simel DL. The rational clinical examination. Does this patient have ascites? How to divine fluid in the abdomen. JAMA (1992) 267:2645–2648.
24 Wang CS, FitzGerald JM, Schulzer M, Mak E, Ayas NT. Does this dyspneic patient in the emergency department have congestive heart failure? JAMA (2005) 294:1944–1956.
25 Call SA, Vollenweider MA, Hornung CA, Simel DL, McKinney WP. Does this patient have influenza? JAMA (2005) 293:987–997.
26 Deen HG Jr, Zimmerman RS, Swanson SK, Larson TR. Assessment of bladder function after lumbar decompressive laminectomy for spinal stenosis: a prospective study. J Neurosurg (1994) 80:971–974.[Web of Science][Medline]
27 Kawaguchi Y, Kanamori M, Ishihara H, et al. Clinical symptoms and surgical outcome in lumbar spinal stenosis patients with neuropathic bladder. J Spinal Disord (2001) 14:404–410.[CrossRef][Web of Science][Medline]
28 Kennedy JG, Soffe KE, McGrath A, Stephens MM, Walsh MG, McManus F. Predictors of outcome in cauda equina syndrome. Eur Spine J (1999) 8:317–322.[CrossRef][Web of Science][Medline]
29 White KL, Williams TF, Greenberg BG. The ecology of medical care. N Engl J Med (1961) 265:885–892.[Web of Science][Medline]
30 Guo Y, Kuroki T, Yamashiro S, Koizumi S. Illness behaviour and patient satisfaction as correlates of self-referral in Japan. Fam Pract (2002) 19:326–332.
31 Fukui T. The ecology of medical care in Japan. JMAJ (2005) 48:163–167.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||