- Email: [email protected]
Risk Factors for Serious Underlying Pathology in Adult Emergency Department Nontraumatic Low Back Pain Patients
The Journal of Emergency Medicine, Vol. -, No. -, pp. 1–11, 2014 Copyright Ó 2014 Elsevier Inc. Printed in the USA. All rights reserved 0736-4679/$ - see front matter
http://dx.doi.org/10.1016/j.jemermed.2013.08.140
Original Contribution
RISK FACTORS FOR SERIOUS UNDERLYING PATHOLOGY IN ADULT EMERGENCY DEPARTMENT NONTRAUMATIC LOW BACK PAIN PATIENTS Venkatesh Thiruganasambandamoorthy, MBBS, MSC,*†‡ Ekaterina Turko, MD,* Dominique Ansell, MSC,§ Aparna Vaidyanathan, MBBS,† George A. Wells, PHD,‡ and Ian G. Stiell, MD, MSC*†‡ *Department of Emergency Medicine, University of Ottawa, Ottawa, Ontario, Canada, †Ottawa Hospital Research Institute, The Ottawa Hospital, Ottawa, Ontario, Canada, ‡Department of Epidemiology and Community Medicine, University of Ottawa, Ottawa, Ontario, Canada, and §Faculty of Medicine, University of Ottawa, Ottawa, Ontario, Canada Reprint Address: Venkatesh Thiruganasambandamoorthy, MBBS, MSC, Ottawa Hospital Research Institute, Clinical Epidemiology Unit, The Ottawa Hospital, Civic Campus, 1053 Carling Avenue, 6th Floor, Rm F650, Ottawa, ON K1Y 4E9, Canada
, Abstract—Background: Nontraumatic low back pain (LBP) is a common emergency department (ED) complaint and can be caused by serious pathologies that require immediate intervention or that lead to death. Objective: The primary goal of this study is to identify risk factors associated with serious pathology in adult nontraumatic ED LBP patients. Methods: We conducted a health records review and included patients aged $ 16 years with nontraumatic LBP presenting to an academic ED from November 2009 to January 2010. We excluded those with previously confirmed nephrolithiasis and typical renal colic presentation. We collected 56 predictor variables and outcomes within 30 days. Outcomes were determined by tracking computerized patient records and performance of univariate analysis and recursive partitioning. Results: There were 329 patients included, with a mean age of 49.3 years; 50.8% were women. A total of 22 (6.7%) patients suffered outcomes, including one death, five compression fractures, four malignancies, four disc prolapses requiring surgery, two retroperitoneal bleeds, two osteomyelitis, and one each of epidural abscess, cauda equina, and leaking abdominal aortic aneurysm graft. Risk factors identified for outcomes were: anticoagulant use (odds ratio [OR] 15.6; 95% confidence interval [CI] 4.2–58.5), decreased sensation on physical examination (OR 6.9; CI 2.2–21.2), pain that is worse at night (OR 4.3; CI 0.9–20.1), and pain that persists despite appropriate treatment (OR 2.2; CI 0.8–5.6). These four predictors identified serious pathology with 91% sensitivity
(95% CI 70–98%) and 55% specificity (95% CI 54–56%). Conclusion: We successfully identified risk factors associated with serious pathology among ED LBP patients. Future prospective studies are required to derive a robust clinical decision rule. Ó 2014 Elsevier Inc. , Keywords—low back pain; emergency department; nontraumatic; outcomes; risk stratification
INTRODUCTION Low back pain is a common presenting complaint in the emergency department (ED). Literature shows that approximately 70–90% of adults will experience at least one episode of back pain during their lifetime (1,2). In the United States, low back pain is the fourth leading cause of ED visits, with 2.9 million visits in 2004 (3). The Saskatchewan Health and Back Pain Survey found that, at the time of the survey, 28.4% of adults were experiencing low back pain (4). In the majority of cases, the low back pain is selflimiting and does not require emergent intervention. Among the patients presenting with back pain in a primary care setting, 5–10% will have serious underlying spinal or
RECEIVED: 17 May 2013; FINAL SUBMISSION RECEIVED: 27 August 2013; ACCEPTED: 31 August 2013 1
2
V. Thiruganasambandamoorthy et al.
paraspinal pathology (fractures, abscesses, osteomyelitis, malignancies, cauda equina syndrome, and paraspinal visceral diseases such as abdominal aortic aneurysm), and many of them require immediate intervention (5). The prevalence of serious pathology in ED low back pain patients remains unknown. Few studies have identified predictors of such serious pathology (6–8). Deyo et al. identified predictors of cancer in low back pain patients in a primary care setting, Waldvogel et al. summarized 18 studies with vertebral osteomyelitis, and Domen et al. identified predictors of cauda equina in 58 ED patients who underwent magnetic resonance imaging (MRI) (6–8). There are no ED studies that aid emergency physicians in identification of patients with serious underlying pathology. The frequency of ED visits for nontraumatic low back pain, their outcomes, and the risk factors associated with serious underlying causes are also not known. The primary goal of this study is to identify risk factors associated with serious pathology in adult nontraumatic ED low back pain patients. METHODS Design and Setting This study was a health records review of consecutive nontraumatic low back pain patients presenting to the two tertiary care EDs of the Ottawa Hospital, with over 130,000 ED visits annually. We conducted the study over 3 months, from November 2009 to January 2010. The hospital research ethics board approved the protocol without the need for informed consent. Study Population We identified potentially eligible patients by searching The Ottawa Hospital health records database. This database uses the Canadian National Ambulatory Care Reporting System (NACRS) to capture data on patients visiting Canadian EDs. We searched for the following terms in the presenting complaint, and the primary and secondary discharge diagnoses fields of the database: ‘‘low back pain,’’ ‘‘back pain,’’ ‘‘radiculopathy,’’ ‘‘sciatica,’’ ‘‘dorsalgia,’’ ‘‘back ache,’’ ‘‘pain in the lower limb,’’ and ‘‘lumbago.’’ We also screened patients with the International Classification of Diseases codes M54.1 (radiculopathy), M54.2 (cervicalgia), M54.3 (sciatica), M54.4 (lumbago with sciatica), M54.5 (low back pain), M54.8 and M54.9 (dorsalgia), M51.1 (lumbago with sciatica due to intervertebral disc disorder), G57.0 (lesion of the sciatic nerve), and M79.61 (lower extremity pain). We included patients who were $ 16 years old, who had a local residential address, had a chief complaint of non-
traumatic low back pain (defined as back pain below the costal margins and above the buttocks), and who were assessed by an emergency physician. We excluded patients who had a history of nephrolithiasis confirmed by imaging and who presented with typical signs and symptoms consistent with renal colic. However, patients with flank pain with no prior documented nephrolithiasis were included. We excluded patients with history of back trauma immediately preceding the onset of the symptoms. We included patients regardless of whether they were admitted to the hospital or discharged home. Study Protocol and Data Abstraction The initial visit was defined as the first visit during the study period and the return visit as any visit within 30 days of the initial visit. We included each patient only once during the study period to avoid double counting. We selected variables for data abstraction based on published literature and recommendations by five experienced emergency physicians (6–13). We photocopied the ED records of treatment (physician, nurse, consultant, and paramedic notes), and one of two investigators (E.T. or D.A.) used these copies to assess the patient’s eligibility for the study. The final diagnosis, imaging results, and other portions of the chart indicating any outcome were blinded prior to data abstraction by a second investigator (E.T. or D.A.), different from the one assessing eligibility. The first 10% of the charts and, subsequently, another 10% of a random sample of the charts were reviewed by the principal investigator to standardize data abstraction and ensure data abstraction accuracy, respectively. We designed and piloted standardized data abstraction forms prior to conducting the study. The data abstraction form had a provision to document if data were missing for any variable. In total, we abstracted 56 predictor variables (two demographic, 28 historical including medications, 15 clinical examination, eight investigation, and three disposition variables). Among the variables abstracted, 47 were categorical and nine were continuous variables. The list of variables collected is detailed in Appendix I. We designed SAS (Statistical Analysis System; SAS Institute Inc., Cary, NC)-based data entry screens with built-in range and logic checks. We assessed the accuracy of data collection and data entry by regular frequency reports. Outcomes We defined a serious outcome as the identification of any one of the following underlying pathologies within 30 days of the index visit: compression fracture,
Risk Stratification of ED Nontraumatic Low Back Pain
osteomyelitis, spinal abscess, malignancy, cauda equina syndrome, severe disc prolapse requiring surgery, any condition requiring immediate intervention (such as abdominal aortic aneurysm that is leaking; retroperitoneal tumor, bleeding or infection that required treatment, spinal stenosis requiring surgery), or death due to any of the above serious conditions. The list of serious conditions that constitute an outcome in this study and their definition are detailed in Appendix II. We assessed for outcomes by reviewing all documents available through the computerized patient tracking system (ED records for the initial and return visits; hospital health records for inpatient, follow-up clinic or investigation, operation room documents, and death records). Our hospital is the only regional center with neurosurgery, spine, and vascular intervention capabilities. All outcomes were confirmed by an independent blinded reviewer, and disagreements were resolved by consensus. We collected information on the place of outcome detection–in the ED, as an inpatient, on return visit, or outside the hospital. We also collected data on follow-up arrangements made in patients with serious outcomes. Data Analysis We calculated descriptive statistics and report means with ranges or SDs as appropriate for the data distribution. We report percentages for categorical or dichotomous variables. We performed univariate analysis to determine the strength of association between the predictor variables and outcomes. For dichotomous variables, we used the chi-squared test with continuity correction or the Fisher exact test; and for continuous variables, we used an unpaired two-tailed t-test using separate or pooled variance estimates as appropriate. We performed univariate analysis with the assumption that the missing variables were both normal and abnormal. We used SAS (Version 9.2) software for descriptive statistics and univariate analysis. Variables with a p-value < 0.2 on univariate analysis were selected for chi-squared recursive partitioning analysis (using Knowledge SEEKER V5 Software; Angoss Software, Toronto, ON, Canada) to develop a final model. We aimed for a model that is highly sensitive, adequately specific, clinically sensible and acceptable, and one that would not require any computation or statistical aids. We calculated odds ratios for the component predictors in the final model using logistic regression (SAS). We assessed for lack of fit in the model using the Hosmer-Lemeshow goodness-of-fit test (14). We calculated test characteristics for the final model using publicly available software and report sensitivity, specificity, and predictive values (15). We determined the sample size based on the precision of the sensitivity of the model to be derived and the bound
3 Table 1. Patient Characteristics, Management, and Disposition of 329 Adult Emergency Department Nontraumatic Low Back Pain Patients Characteristics Demographics Age in years – Mean (range) Female Arrival by ambulance History of present illness* Insidious onset of pain (n = 266) Pain that is worst at night (n = 30) Motor symptoms [weakness/paralysis] (n = 213) Urinary retention or incontinence (n = 274) Fecal incontinence (n = 275) Medical history* Previous cancer (n = 326) Bacteremia procedure within previous 8 weeks (n = 324) Osteoporosis (n = 328) Immunosuppression† (n = 329) Anticoagulants (n = 328) Examination Midline spine tenderness (n = 225) Motor weakness (n = 250) Decreased sensation (n = 238) Abnormal deep tendon reflexes (n = 176) Saddle anesthesia (n = 100) Fullness bladder/suprapubic (n = 33) Investigations and imaging Imaging or blood work done Imaging done in ED X-ray lumbar spine CT lumbar spine MRI lumbar spine Other Imaging related to presentation done within last month Disposition Admitted to hospital Seen by specialist in ED Outpatient specialist follow-up arranged Outcomes Outcome in ED Outcome outside of ED
Patients, N = 329 n (%) 49.3 (17–93) 167 (50.8) 84 (25.5) 131 (39.8) 16 (4.9) 50 (15.2) 15 (4.6) 4 (1.2) 20 (6.1) 10 (3.0) 6 (1.8) 8 (2.4) 13 (4.0) 69 (21.0) 52 (15.8) 27 (8.2) 30 (9.1) 6 (1.8) 1 (0.3) 239 (72.6) 112 (34.0) 57 (17.3) 12 (3.7) 13 (4.0) 45 (13.7) 40 (12.6) 10 (3.0) 15 (4.6) 43 (13.1) 22 (6.7) 8 (2.4) 14 (4.3)
CT = computed tomography; MRI = magnetic resonance imaging; ED = emergency department. * The numbers in brackets indicate the number of patients who had the variable available in their charts. † Immunodeficiency or on immunosuppressive medications.
on the error of the precision estimate (the width of the 95% confidence interval), which is the standard technique used in sample size calculation for decision rule studies (16). We determined a priori that a sample size of 308 patients was needed to assess for risk factors. RESULTS There were a total of 31,705 ED visits during the study period. Of the potential 698 patients who met the
4
V. Thiruganasambandamoorthy et al.
screening criteria, 330 patients (1% of total) met the inclusion criteria and had no exclusion criteria. One patient’s ED documents were incomplete and missing several initial ED visit documents, and therefore that patient was excluded. As a result, 329 patients were included in the study. The patients’ demographics, back pain characteristics, medical history, medications taken, investigations conducted, and disposition are detailed in Table 1. The majority of patients (72.6%) had blood work, imaging, or both done during their visit. X-ray study of the lumbar spine was the most frequently requested radiological investigation, ordered in 17.3% of cases (57 patients); outcome was present in 7 (yield 12.3%; six compression fractures and one malignancy). Computed tomography scan of the lumbar spine was performed in 3.7% (12 patients) and was able to detect two outcomes (yield 16.7%; one case of osteomyelitis and one case of severe disc prolapse requiring surgery). MRI of the lumbar spine was ordered in 4% of cases (13 patients) and was helpful in detecting four outcomes (yield 30.1%; two cases of malignancy, one case of spinal stenosis requiring surgery, and one case of retroperitoneal bleeding). Thirty-nine patients (12%) had a related return ED visit within 30 days, with 3 patients (0.9%) returning more than once. The list of outcomes in the study patients, the place of outcome detection, and follow-up arrangements made for those who suffered outcomes are listed in Table 2. A total of 22 patients had serious underlying pathology causing their back pain, including one patient in whom the back pain was due to upper urinary tract infection that led to urosepsis and death. Four patients with fractures had other concomitant pathology (3 patients had underlying spinal malignancy and one had osteomyelitis with epidural abscess). In 2 patients, retroperitoneal bleeding was the cause of the back pain (one required emergent embolization and another emergent reversal of anticoagulation), and in one patient the back pain
was due to an infected and leaking abdominal aortic aneurysm graft. The majority (14 patients, 4.3%) of the outcomes were detected after the initial ED visit, and a small but significant number of patients (8 patients, 2.4%) was discharged home without appropriate follow-up arrangement. The results of the univariate analysis of the categorical and continuous variables are shown in Tables 3 and 4, respectively. We performed univariate analysis between the categorical predictor variables and serious outcomes with the assumption that the missing variables were normal (Table 3). We conducted sensitivity analysis with the assumption that the missing variables were abnormal. We found 17 categorical variables to be significant (p < 0.2), 13 variables with the assumption that missing variables were normal, and an additional four variables (history of insidious onset of pain, urinary incontinence, fecal incontinence, and decreased anal sphincter tone) with the assumption that the missing ones were abnormal. Six categorical variables (history of sensory symptoms, urinary retention, recent bacteremia procedure, history of cancer, anticoagulant use, and decreased sensation on physical examination) were significant under both assumptions. Table 4 shows the results of univariate analysis of continuous predictor variables. Two continuous variables, erythrocyte sedimentation rate and C-reactive protein, were not analyzed, as only 3 study patients had these tests performed. Of the seven variables analyzed, four (age, triage diastolic blood pressure, hemoglobin, and international normalized ratio [INR] for prothrombin time) were significant. These four variables were further analyzed by using various cut points, and three (age, hemoglobin, and INR) were found to be significant at different cut points with the assumption that missing variables were normal. Only the variable hemoglobin < 100 g/L was found to be significant under the assumption that missing variables were abnormal.
Table 2. Frequency of Serious Outcomes, Location of Outcome Detection Among 329 Adult Emergency Department Nontraumatic Low Back Pain Patients, and Follow-Up of Patients With Serious Outcomes Serious Outcomes
Total
In ED
Outside ED
No Appropriate Follow-Up Arranged
n Death Compression fracture Osteomyelitis Epidural abscess Malignancy Cauda equina Severe disc prolapse requiring surgery Retroperitoneal bleeding Spinal stenosis requiring surgery Abdominal aortic aneurysm graft leak
22 1 5 2 1 4 1 4 2 1 1
8 0 2 0 0 2 0 2 2 0 0
14 1 3 2 1 2 1 2 0 1 1
8 1 3 1 1 0 1 1 0 0 0
ED = emergency department.
Risk Stratification of ED Nontraumatic Low Back Pain Table 3. Univariate Analysis of Predictor Variables and Serious Outcomes Among the 329 Adult Emergency Department Nontraumatic Low Back Pain Patients
Variable History of present illness Duration < 1 day Duration of pain > 365 days History of insidious onset Pain persistent despite appropriate analgesia Pain that awakens patient at night Pain that is worse at night Pain that is worse at rest Progressive/worsening symptoms Pain in the middle of lumbar vertebrae Any radiation of pain to the leg Radiation to the leg below the knee Motor symptoms (weakness/ paralysis) Sensory symptoms Fever Urinary retention Urinary incontinence Fecal incontinence Medical history Recent bacteremia procedures History of osteoporosis History of cancer History of HIV On anticoagulants Prolonged steroids use On immunosuppressive medications Physical examination Midline spine tenderness Inability to walk Motor weakness Fever ($38 C) Any decreased sensation Saddle anesthesia Decreased anal sphincter tone Bladder/suprapubic fullness Abnormal deep tendon reflexes
Outcome Outcome Present Absent (n = 22) (n = 307)
p Value
3 2 11
83 9 120
0.21 0.16* 0.31
13
114
0.04*
3
19
0.17*
3 1 7
13 9 123
0.08* 0.50 0.44
0
11
1.00
8 6
145 90
0.32 0.84
6
44
0.10*
8 0 3 0 1
55 8 6 6 3
0.03* 1.00 0.02* 1.00 0.24
2
8
0.14*
1 3 0 5 0 0
5 17 2 8 3 6
0.34 0.14* 1.00 < 0.0001* 1.00 1.00
3 2 3 1 6 2 0 1 8
66 29 49 9 21 4 3 0 138
0.59 1.00 1.00 0.50 < 0.0001* 0.05* 1.00 0.07* 0.43
HIV = human immunodeficiency virus. * Variables used in multivariable analysis.
We found 15 predictor variables to be significant on univariate analysis with the assumption that missing variables were normal, and 11 with the assumption that the missing variables were abnormal. Figure 1 shows the four high-risk predictor variables derived by recursive partitioning that identify serious underlying pathology in nontraumatic ED low back pain patients: pain that persists despite appropriate treatment, use of anticoagulant medications, decreased sensation on physical examina-
5
tion, and pain that is worse at night. Figure 2 shows the receiver operator characteristic curve for the rule and Figure 3 shows the test characteristics of the model with the four high-risk variables. DISCUSSION To the best of our knowledge, our study is the first to report outcomes and to identify risk factors associated with serious underlying pathology in adult nontraumatic ED low back pain patients. We conducted the study as per the methodological standards for chart review to decrease the bias in patient selection and data abstraction (17–19). The proportion of ED visits to our study site that was due to low back pain (1%) is lower than previously reported (3). This could possibly be related to exclusion of traumatic low back pain patients. Our study shows that a small but important number of patients have underlying serious pathology that was not detected during the initial ED visit. This highlights the need for a robust risk-stratification strategy to detect those with serious underlying conditions. We identified four important predictor variables associated with serious outcomes in ED low back pain patients: pain that persists despite appropriate treatment, use of anticoagulant medications, decreased sensation on physical examination, and pain that is worse at night. Most published studies that report risk factors for serious pathology in low back pain patients include traumatic back pain patients and identify significant trauma as a risk factor for fracture (9,20,21). Constant and progressive pain was found to be a risk factor for underlying cancer and spinal infection, and altered sensation was a risk factor for underlying cancer, infection, and fracture in a large primary care study of acute low back pain patients (21). Deyo et al. reported pain that is worse at night and pain that does not respond to treatment as risk factors for underlying cancer (5,6). No studies in the literature report the use of anticoagulant medications as a risk factor. This could be partly explained by the fact that none of the studies has reported acute bleeding conditions such as a retroperitoneal hemorrhage as an outcome. In our study, 2 patients had retroperitoneal bleeding and both required acute treatment. The signs and symptoms associated with cauda equina, urinary retention, urinary or fecal incontinence, and saddle anesthesia were not part of the final model, as there was only one patient in our study who had cauda equina as the underlying cause of low back pain. Our study identifies four high-risk factors for serious underlying conditions in adult ED nontraumatic low back pain patients. Use of these four high-risk factors
6
V. Thiruganasambandamoorthy et al.
Table 4. Univariate Analysis of Continuous Predictor Variables With Discriminative Cut Points and Serious Outcomes Among the 329 Adult Emergency Department Nontraumatic Low Back Pain Patients Variable
Outcome Absent (n = 307)
p Value
65.2 (20.6) 21.5 (39.0) 36.2 (0.9) 136.2 (26.5)
48.1 (17.6) 18.0 (33.2) 37.2 (1.0) 138.3 (21.7)
< 0.0001 0.68 0.35 0.67
75.3 (12.7)
79.1 (12.5)
0.17
112.7 (22.2) 0.58 (0.51)
135.2 (20.7) 0.20 (0.40)
0.001 0.004
12 9 2 4 7
52 34 2 5 16
< 0.0001 < 0.0001 0.02 0.002 < 0.0001
Outcome Present (n = 22)
Continuous predictor variables (Mean, SD) Age Days of low back pain (n = 17,201) Temperature (n = 22,293) Mean triage systolic blood pressure (n = 22,303) Mean triage diastolic blood pressure (n = 22,303) Hemoglobin (n = 1176) INR (n = 1279) Continuous variables with discriminative cut points (n) Age > 65 years Age > 75 years Hemoglobin < 100 g/L Hemoglobin < 110 g/L INR $ 1
SD = standard deviation; INR = international normalized ratio.
would still have missed 2 patients with serious pathology if reapplied to the study population, one with severe disc prolapse and the second with retroperitoneal hematoma. Both patients had their pathology identified during the initial ED visit. Hence, the reported risk factors would
identify all patients with serious pathology not identified during their ED visit. These factors will be helpful to the practicing emergency physician to rule in serious conditions, but the model is not 100% sensitive and hence, should not be used to rule out serious underlying
Figure 1. Identification of predictors of serious outcomes in adult emergency department nontraumatic low back pain patients by recursive partitioning.
Risk Stratification of ED Nontraumatic Low Back Pain
7
had some of the predictor variables missing. We reduced the impact of missing variables by conducting analysis under the assumptions that they were either normal or abnormal. CONCLUSIONS
Figure 2. Receiver operating characteristic curve for the independent predictors of serious outcomes in adult emergency department nontraumatic low back pain patients.
pathology. Future prospective studies are required to derive a robust clinical decision rule. Limitations The limitations of our study are inherent to all medical record review studies. It is possible that our screening process may have missed some nontraumatic low back pain patients. We decreased this possibility by using broad screening criteria and by using all possible fields in the database (chief complaint, primary and secondary discharge diagnoses) to identify patients for screening. The national ambulatory database (NACRS) is a wellestablished database known to capture good quality data on ED visits (22). We excluded patients with previously confirmed nephrolithiasis presenting with typical pain to focus the study on true ED nontraumatic patients who present a dilemma to the practicing emergency physician. While we reviewed medical records within our hospital network, it is possible that patients suffered serious outcomes out of the local area or while in other parts of the country. A proportion of our study patients
Figure 3. Classification performance of the final model for detecting serious underlying pathology in adult emergency department nontraumatic low back pain patients. CI = confidence interval.
Nontraumatic low back pain is a common ED problem. A small but important number of patients have serious underlying pathology that is not detected during their ED visit. Our study has identified high-risk factors associated with serious outcomes. Future prospective studies are required to develop a robust clinical decision rule to identify all patients with serious underlying conditions to improve patient care and patient safety. Acknowledgments—This study was funded by grants from the Canadian Association of Emergency Physicians, and the Department of Emergency Medicine, University of Ottawa. Dr. Thiruganasambandamoorthy receives salary support from the Heart and Stroke Foundation of Canada through the Jump Start Resuscitation Scholarship. We gratefully acknowledge My-Linh Tran, Sheryl Domingo, Dr. Magdalena Kisilewicz, Rina Marcantonio, Lilliana Polesello, Dr. Erin O’Connor, Cathy Clement, RN, and Angela Marcantonio for their help with this study.
REFERENCES 1. Kelsey JL, White AA 3rd. Epidemiology and impact of low-back pain. Spine (Phila Pa 1976) 1980;5:133–42. 2. Mazanec DJ. Back pain: medical evaluation and therapy. Cleve Clin J Med 1995;62:163–8. 3. McCaig LF, Nawar EW. National Hospital Ambulatory Medical Care Survey: 2004 emergency department summary. Adv Data 2006;372:1–29. 4. Cassidy JD, Carroll LJ, Cote P. The Saskatchewan health and back pain survey. The prevalence of low back pain and related disability in Saskatchewan adults. Spine (Phila Pa 1976) 1998;23:1860–6. 5. Deyo RA, Weinstein JN. Low back pain. N Engl J Med 2001;344: 363–70. 6. Deyo RA, Diehl AK. Cancer as a cause of back pain: frequency, clinical presentation, and diagnostic strategies. J Gen Intern Med 1988;3:230–8. 7. Domen PM, Hofman PA, van Santbrink H, Weber WE. Predictive value of clinical characteristics in patients with suspected cauda equina syndrome. Eur J Neurol 2009;16:416–9. 8. Waldvogel FA, Papageorgiou PS. Osteomyelitis: the past decade. N Engl J Med 1980;303:360–70. 9. Davis PC, Wippold FJ 2nd, Brunberg JA, et al. ACR Appropriateness Criteria on low back pain. J Am Coll Radiol 2009;6:401–7. 10. Della-Giustina DA. Emergency department evaluation and treatment of back pain. Emerg Med Clin North Am 1999;17:877–93. 11. Deyo RA, Diehl AK. Lumbar spine films in primary care: current use and effects of selective ordering criteria. J Gen Intern Med 1986;1:20–5. 12. Deyo RA, Rainville J, Kent DL. What can the history and physical examination tell us about low back pain? JAMA 1992;268:760–5. 13. Winters ME, Kluetz P, Zilberstein J. Back pain emergencies. Med Clin North Am 2006;90:505–23.
8 14. Hosmer DW, Lemeshow S. Applied logistic regression. Hoboken, NJ: Wiley; 2000. 15. Pezzullo JC. Interactive statistics page. 2-Way contingency table analysis. Available at: http://www.statpages.org/ctab2x2.html. Accessed March 5, 2014. 16. Scheaffer RL, Mendenhall W, Ott L. Elementary survey sampling. Pacific Grove, CA: Duxbury Press; 1979. 17. Badcock D, Kelly AM, Kerr D, et al. The quality of medical record review studies in the international emergency medicine literature. Ann Emerg Med 2005;45:444–7. 18. Gilbert EH, Lowenstein SR, Koziol-McLain J, et al. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med 1996;27:305–8.
V. Thiruganasambandamoorthy et al. 19. Worster A, Bledsoe RD, Cleve P, et al. Reassessing the methods of medical record review studies in emergency medicine research. Ann Emerg Med 2005;45:448–51. 20. Henschke N, Maher CG, Refshauge KM. A systematic review identifies five ‘‘red flags’’ to screen for vertebral fracture in patients with low back pain. J Clin Epidemiol 2008;61:110–8. 21. Henschke N, Maher CG, Refshauge KM, et al. Prevalence of and screening for serious spinal pathology in patients presenting to primary care settings with acute low back pain. Arthritis Rheum 2009; 60:3072–80. 22. Canadian Institute of Health Information (CIHI). CIHI data quality study of Ontario emergency department visits for fiscal year 2004– 2005—executive summary. Ottawa, ON: CIHI; 2008.
Risk Stratification of ED Nontraumatic Low Back Pain
APPENDIX I: LIST OF PREDICTOR VARIABLES ABSTRACTED FOR STUDY PATIENTS
9
10
V. Thiruganasambandamoorthy et al.
APPENDIX II: LIST OF SERIOUS CONDITIONS DEFINED AS OUTCOMES IN THE STUDY
Risk Stratification of ED Nontraumatic Low Back Pain
ARTICLE SUMMARY 1. Why is this topic important? Low back pain (LBP) is a common Emergency Department (ED) presenting complaint. The frequency of ED visits due to LBP, their outcome and risk factors associated with serious underlying pathology are not known. 2. What does this study attempt to show? Through comprehensive data abstraction and detailed analysis the study identifies risk factors associated with serious underlying conditions in adult nontraumatic LBP patients. 3. What are the key findings? We studied 329 nontraumatic LBP patients. There were 22 patients (6.7%) who had serious underlying conditions causing the back pain. One patient died secondary to upper urinary tract infection, five had compression fractures, four had spinal malignancies, four had disc prolapses requiring surgery, two had retroperitoneal bleed, two had spinal osteomyelitis and one patient with each of epidural abscess, cauda equina and leaking abdominal aneurysm graft. Risk factors for these serious conditions include: anticoagulant use, decreased sensation on physical exam, pain that is worse at night and pain that persists despite appropriate treatment. 4. How is patient care impacted? An important number of nontraumatic LBP patients have a serious underlying cause, if undetected will lead to severe morbidity and even mortality. Future prospective studies are needed to derive a robust risk-tool to identify those at risk.
11
http://dx.doi.org/10.1016/j.jemermed.2013.08.140
Original Contribution
RISK FACTORS FOR SERIOUS UNDERLYING PATHOLOGY IN ADULT EMERGENCY DEPARTMENT NONTRAUMATIC LOW BACK PAIN PATIENTS Venkatesh Thiruganasambandamoorthy, MBBS, MSC,*†‡ Ekaterina Turko, MD,* Dominique Ansell, MSC,§ Aparna Vaidyanathan, MBBS,† George A. Wells, PHD,‡ and Ian G. Stiell, MD, MSC*†‡ *Department of Emergency Medicine, University of Ottawa, Ottawa, Ontario, Canada, †Ottawa Hospital Research Institute, The Ottawa Hospital, Ottawa, Ontario, Canada, ‡Department of Epidemiology and Community Medicine, University of Ottawa, Ottawa, Ontario, Canada, and §Faculty of Medicine, University of Ottawa, Ottawa, Ontario, Canada Reprint Address: Venkatesh Thiruganasambandamoorthy, MBBS, MSC, Ottawa Hospital Research Institute, Clinical Epidemiology Unit, The Ottawa Hospital, Civic Campus, 1053 Carling Avenue, 6th Floor, Rm F650, Ottawa, ON K1Y 4E9, Canada
, Abstract—Background: Nontraumatic low back pain (LBP) is a common emergency department (ED) complaint and can be caused by serious pathologies that require immediate intervention or that lead to death. Objective: The primary goal of this study is to identify risk factors associated with serious pathology in adult nontraumatic ED LBP patients. Methods: We conducted a health records review and included patients aged $ 16 years with nontraumatic LBP presenting to an academic ED from November 2009 to January 2010. We excluded those with previously confirmed nephrolithiasis and typical renal colic presentation. We collected 56 predictor variables and outcomes within 30 days. Outcomes were determined by tracking computerized patient records and performance of univariate analysis and recursive partitioning. Results: There were 329 patients included, with a mean age of 49.3 years; 50.8% were women. A total of 22 (6.7%) patients suffered outcomes, including one death, five compression fractures, four malignancies, four disc prolapses requiring surgery, two retroperitoneal bleeds, two osteomyelitis, and one each of epidural abscess, cauda equina, and leaking abdominal aortic aneurysm graft. Risk factors identified for outcomes were: anticoagulant use (odds ratio [OR] 15.6; 95% confidence interval [CI] 4.2–58.5), decreased sensation on physical examination (OR 6.9; CI 2.2–21.2), pain that is worse at night (OR 4.3; CI 0.9–20.1), and pain that persists despite appropriate treatment (OR 2.2; CI 0.8–5.6). These four predictors identified serious pathology with 91% sensitivity
(95% CI 70–98%) and 55% specificity (95% CI 54–56%). Conclusion: We successfully identified risk factors associated with serious pathology among ED LBP patients. Future prospective studies are required to derive a robust clinical decision rule. Ó 2014 Elsevier Inc. , Keywords—low back pain; emergency department; nontraumatic; outcomes; risk stratification
INTRODUCTION Low back pain is a common presenting complaint in the emergency department (ED). Literature shows that approximately 70–90% of adults will experience at least one episode of back pain during their lifetime (1,2). In the United States, low back pain is the fourth leading cause of ED visits, with 2.9 million visits in 2004 (3). The Saskatchewan Health and Back Pain Survey found that, at the time of the survey, 28.4% of adults were experiencing low back pain (4). In the majority of cases, the low back pain is selflimiting and does not require emergent intervention. Among the patients presenting with back pain in a primary care setting, 5–10% will have serious underlying spinal or
RECEIVED: 17 May 2013; FINAL SUBMISSION RECEIVED: 27 August 2013; ACCEPTED: 31 August 2013 1
2
V. Thiruganasambandamoorthy et al.
paraspinal pathology (fractures, abscesses, osteomyelitis, malignancies, cauda equina syndrome, and paraspinal visceral diseases such as abdominal aortic aneurysm), and many of them require immediate intervention (5). The prevalence of serious pathology in ED low back pain patients remains unknown. Few studies have identified predictors of such serious pathology (6–8). Deyo et al. identified predictors of cancer in low back pain patients in a primary care setting, Waldvogel et al. summarized 18 studies with vertebral osteomyelitis, and Domen et al. identified predictors of cauda equina in 58 ED patients who underwent magnetic resonance imaging (MRI) (6–8). There are no ED studies that aid emergency physicians in identification of patients with serious underlying pathology. The frequency of ED visits for nontraumatic low back pain, their outcomes, and the risk factors associated with serious underlying causes are also not known. The primary goal of this study is to identify risk factors associated with serious pathology in adult nontraumatic ED low back pain patients. METHODS Design and Setting This study was a health records review of consecutive nontraumatic low back pain patients presenting to the two tertiary care EDs of the Ottawa Hospital, with over 130,000 ED visits annually. We conducted the study over 3 months, from November 2009 to January 2010. The hospital research ethics board approved the protocol without the need for informed consent. Study Population We identified potentially eligible patients by searching The Ottawa Hospital health records database. This database uses the Canadian National Ambulatory Care Reporting System (NACRS) to capture data on patients visiting Canadian EDs. We searched for the following terms in the presenting complaint, and the primary and secondary discharge diagnoses fields of the database: ‘‘low back pain,’’ ‘‘back pain,’’ ‘‘radiculopathy,’’ ‘‘sciatica,’’ ‘‘dorsalgia,’’ ‘‘back ache,’’ ‘‘pain in the lower limb,’’ and ‘‘lumbago.’’ We also screened patients with the International Classification of Diseases codes M54.1 (radiculopathy), M54.2 (cervicalgia), M54.3 (sciatica), M54.4 (lumbago with sciatica), M54.5 (low back pain), M54.8 and M54.9 (dorsalgia), M51.1 (lumbago with sciatica due to intervertebral disc disorder), G57.0 (lesion of the sciatic nerve), and M79.61 (lower extremity pain). We included patients who were $ 16 years old, who had a local residential address, had a chief complaint of non-
traumatic low back pain (defined as back pain below the costal margins and above the buttocks), and who were assessed by an emergency physician. We excluded patients who had a history of nephrolithiasis confirmed by imaging and who presented with typical signs and symptoms consistent with renal colic. However, patients with flank pain with no prior documented nephrolithiasis were included. We excluded patients with history of back trauma immediately preceding the onset of the symptoms. We included patients regardless of whether they were admitted to the hospital or discharged home. Study Protocol and Data Abstraction The initial visit was defined as the first visit during the study period and the return visit as any visit within 30 days of the initial visit. We included each patient only once during the study period to avoid double counting. We selected variables for data abstraction based on published literature and recommendations by five experienced emergency physicians (6–13). We photocopied the ED records of treatment (physician, nurse, consultant, and paramedic notes), and one of two investigators (E.T. or D.A.) used these copies to assess the patient’s eligibility for the study. The final diagnosis, imaging results, and other portions of the chart indicating any outcome were blinded prior to data abstraction by a second investigator (E.T. or D.A.), different from the one assessing eligibility. The first 10% of the charts and, subsequently, another 10% of a random sample of the charts were reviewed by the principal investigator to standardize data abstraction and ensure data abstraction accuracy, respectively. We designed and piloted standardized data abstraction forms prior to conducting the study. The data abstraction form had a provision to document if data were missing for any variable. In total, we abstracted 56 predictor variables (two demographic, 28 historical including medications, 15 clinical examination, eight investigation, and three disposition variables). Among the variables abstracted, 47 were categorical and nine were continuous variables. The list of variables collected is detailed in Appendix I. We designed SAS (Statistical Analysis System; SAS Institute Inc., Cary, NC)-based data entry screens with built-in range and logic checks. We assessed the accuracy of data collection and data entry by regular frequency reports. Outcomes We defined a serious outcome as the identification of any one of the following underlying pathologies within 30 days of the index visit: compression fracture,
Risk Stratification of ED Nontraumatic Low Back Pain
osteomyelitis, spinal abscess, malignancy, cauda equina syndrome, severe disc prolapse requiring surgery, any condition requiring immediate intervention (such as abdominal aortic aneurysm that is leaking; retroperitoneal tumor, bleeding or infection that required treatment, spinal stenosis requiring surgery), or death due to any of the above serious conditions. The list of serious conditions that constitute an outcome in this study and their definition are detailed in Appendix II. We assessed for outcomes by reviewing all documents available through the computerized patient tracking system (ED records for the initial and return visits; hospital health records for inpatient, follow-up clinic or investigation, operation room documents, and death records). Our hospital is the only regional center with neurosurgery, spine, and vascular intervention capabilities. All outcomes were confirmed by an independent blinded reviewer, and disagreements were resolved by consensus. We collected information on the place of outcome detection–in the ED, as an inpatient, on return visit, or outside the hospital. We also collected data on follow-up arrangements made in patients with serious outcomes. Data Analysis We calculated descriptive statistics and report means with ranges or SDs as appropriate for the data distribution. We report percentages for categorical or dichotomous variables. We performed univariate analysis to determine the strength of association between the predictor variables and outcomes. For dichotomous variables, we used the chi-squared test with continuity correction or the Fisher exact test; and for continuous variables, we used an unpaired two-tailed t-test using separate or pooled variance estimates as appropriate. We performed univariate analysis with the assumption that the missing variables were both normal and abnormal. We used SAS (Version 9.2) software for descriptive statistics and univariate analysis. Variables with a p-value < 0.2 on univariate analysis were selected for chi-squared recursive partitioning analysis (using Knowledge SEEKER V5 Software; Angoss Software, Toronto, ON, Canada) to develop a final model. We aimed for a model that is highly sensitive, adequately specific, clinically sensible and acceptable, and one that would not require any computation or statistical aids. We calculated odds ratios for the component predictors in the final model using logistic regression (SAS). We assessed for lack of fit in the model using the Hosmer-Lemeshow goodness-of-fit test (14). We calculated test characteristics for the final model using publicly available software and report sensitivity, specificity, and predictive values (15). We determined the sample size based on the precision of the sensitivity of the model to be derived and the bound
3 Table 1. Patient Characteristics, Management, and Disposition of 329 Adult Emergency Department Nontraumatic Low Back Pain Patients Characteristics Demographics Age in years – Mean (range) Female Arrival by ambulance History of present illness* Insidious onset of pain (n = 266) Pain that is worst at night (n = 30) Motor symptoms [weakness/paralysis] (n = 213) Urinary retention or incontinence (n = 274) Fecal incontinence (n = 275) Medical history* Previous cancer (n = 326) Bacteremia procedure within previous 8 weeks (n = 324) Osteoporosis (n = 328) Immunosuppression† (n = 329) Anticoagulants (n = 328) Examination Midline spine tenderness (n = 225) Motor weakness (n = 250) Decreased sensation (n = 238) Abnormal deep tendon reflexes (n = 176) Saddle anesthesia (n = 100) Fullness bladder/suprapubic (n = 33) Investigations and imaging Imaging or blood work done Imaging done in ED X-ray lumbar spine CT lumbar spine MRI lumbar spine Other Imaging related to presentation done within last month Disposition Admitted to hospital Seen by specialist in ED Outpatient specialist follow-up arranged Outcomes Outcome in ED Outcome outside of ED
Patients, N = 329 n (%) 49.3 (17–93) 167 (50.8) 84 (25.5) 131 (39.8) 16 (4.9) 50 (15.2) 15 (4.6) 4 (1.2) 20 (6.1) 10 (3.0) 6 (1.8) 8 (2.4) 13 (4.0) 69 (21.0) 52 (15.8) 27 (8.2) 30 (9.1) 6 (1.8) 1 (0.3) 239 (72.6) 112 (34.0) 57 (17.3) 12 (3.7) 13 (4.0) 45 (13.7) 40 (12.6) 10 (3.0) 15 (4.6) 43 (13.1) 22 (6.7) 8 (2.4) 14 (4.3)
CT = computed tomography; MRI = magnetic resonance imaging; ED = emergency department. * The numbers in brackets indicate the number of patients who had the variable available in their charts. † Immunodeficiency or on immunosuppressive medications.
on the error of the precision estimate (the width of the 95% confidence interval), which is the standard technique used in sample size calculation for decision rule studies (16). We determined a priori that a sample size of 308 patients was needed to assess for risk factors. RESULTS There were a total of 31,705 ED visits during the study period. Of the potential 698 patients who met the
4
V. Thiruganasambandamoorthy et al.
screening criteria, 330 patients (1% of total) met the inclusion criteria and had no exclusion criteria. One patient’s ED documents were incomplete and missing several initial ED visit documents, and therefore that patient was excluded. As a result, 329 patients were included in the study. The patients’ demographics, back pain characteristics, medical history, medications taken, investigations conducted, and disposition are detailed in Table 1. The majority of patients (72.6%) had blood work, imaging, or both done during their visit. X-ray study of the lumbar spine was the most frequently requested radiological investigation, ordered in 17.3% of cases (57 patients); outcome was present in 7 (yield 12.3%; six compression fractures and one malignancy). Computed tomography scan of the lumbar spine was performed in 3.7% (12 patients) and was able to detect two outcomes (yield 16.7%; one case of osteomyelitis and one case of severe disc prolapse requiring surgery). MRI of the lumbar spine was ordered in 4% of cases (13 patients) and was helpful in detecting four outcomes (yield 30.1%; two cases of malignancy, one case of spinal stenosis requiring surgery, and one case of retroperitoneal bleeding). Thirty-nine patients (12%) had a related return ED visit within 30 days, with 3 patients (0.9%) returning more than once. The list of outcomes in the study patients, the place of outcome detection, and follow-up arrangements made for those who suffered outcomes are listed in Table 2. A total of 22 patients had serious underlying pathology causing their back pain, including one patient in whom the back pain was due to upper urinary tract infection that led to urosepsis and death. Four patients with fractures had other concomitant pathology (3 patients had underlying spinal malignancy and one had osteomyelitis with epidural abscess). In 2 patients, retroperitoneal bleeding was the cause of the back pain (one required emergent embolization and another emergent reversal of anticoagulation), and in one patient the back pain
was due to an infected and leaking abdominal aortic aneurysm graft. The majority (14 patients, 4.3%) of the outcomes were detected after the initial ED visit, and a small but significant number of patients (8 patients, 2.4%) was discharged home without appropriate follow-up arrangement. The results of the univariate analysis of the categorical and continuous variables are shown in Tables 3 and 4, respectively. We performed univariate analysis between the categorical predictor variables and serious outcomes with the assumption that the missing variables were normal (Table 3). We conducted sensitivity analysis with the assumption that the missing variables were abnormal. We found 17 categorical variables to be significant (p < 0.2), 13 variables with the assumption that missing variables were normal, and an additional four variables (history of insidious onset of pain, urinary incontinence, fecal incontinence, and decreased anal sphincter tone) with the assumption that the missing ones were abnormal. Six categorical variables (history of sensory symptoms, urinary retention, recent bacteremia procedure, history of cancer, anticoagulant use, and decreased sensation on physical examination) were significant under both assumptions. Table 4 shows the results of univariate analysis of continuous predictor variables. Two continuous variables, erythrocyte sedimentation rate and C-reactive protein, were not analyzed, as only 3 study patients had these tests performed. Of the seven variables analyzed, four (age, triage diastolic blood pressure, hemoglobin, and international normalized ratio [INR] for prothrombin time) were significant. These four variables were further analyzed by using various cut points, and three (age, hemoglobin, and INR) were found to be significant at different cut points with the assumption that missing variables were normal. Only the variable hemoglobin < 100 g/L was found to be significant under the assumption that missing variables were abnormal.
Table 2. Frequency of Serious Outcomes, Location of Outcome Detection Among 329 Adult Emergency Department Nontraumatic Low Back Pain Patients, and Follow-Up of Patients With Serious Outcomes Serious Outcomes
Total
In ED
Outside ED
No Appropriate Follow-Up Arranged
n Death Compression fracture Osteomyelitis Epidural abscess Malignancy Cauda equina Severe disc prolapse requiring surgery Retroperitoneal bleeding Spinal stenosis requiring surgery Abdominal aortic aneurysm graft leak
22 1 5 2 1 4 1 4 2 1 1
8 0 2 0 0 2 0 2 2 0 0
14 1 3 2 1 2 1 2 0 1 1
8 1 3 1 1 0 1 1 0 0 0
ED = emergency department.
Risk Stratification of ED Nontraumatic Low Back Pain Table 3. Univariate Analysis of Predictor Variables and Serious Outcomes Among the 329 Adult Emergency Department Nontraumatic Low Back Pain Patients
Variable History of present illness Duration < 1 day Duration of pain > 365 days History of insidious onset Pain persistent despite appropriate analgesia Pain that awakens patient at night Pain that is worse at night Pain that is worse at rest Progressive/worsening symptoms Pain in the middle of lumbar vertebrae Any radiation of pain to the leg Radiation to the leg below the knee Motor symptoms (weakness/ paralysis) Sensory symptoms Fever Urinary retention Urinary incontinence Fecal incontinence Medical history Recent bacteremia procedures History of osteoporosis History of cancer History of HIV On anticoagulants Prolonged steroids use On immunosuppressive medications Physical examination Midline spine tenderness Inability to walk Motor weakness Fever ($38 C) Any decreased sensation Saddle anesthesia Decreased anal sphincter tone Bladder/suprapubic fullness Abnormal deep tendon reflexes
Outcome Outcome Present Absent (n = 22) (n = 307)
p Value
3 2 11
83 9 120
0.21 0.16* 0.31
13
114
0.04*
3
19
0.17*
3 1 7
13 9 123
0.08* 0.50 0.44
0
11
1.00
8 6
145 90
0.32 0.84
6
44
0.10*
8 0 3 0 1
55 8 6 6 3
0.03* 1.00 0.02* 1.00 0.24
2
8
0.14*
1 3 0 5 0 0
5 17 2 8 3 6
0.34 0.14* 1.00 < 0.0001* 1.00 1.00
3 2 3 1 6 2 0 1 8
66 29 49 9 21 4 3 0 138
0.59 1.00 1.00 0.50 < 0.0001* 0.05* 1.00 0.07* 0.43
HIV = human immunodeficiency virus. * Variables used in multivariable analysis.
We found 15 predictor variables to be significant on univariate analysis with the assumption that missing variables were normal, and 11 with the assumption that the missing variables were abnormal. Figure 1 shows the four high-risk predictor variables derived by recursive partitioning that identify serious underlying pathology in nontraumatic ED low back pain patients: pain that persists despite appropriate treatment, use of anticoagulant medications, decreased sensation on physical examina-
5
tion, and pain that is worse at night. Figure 2 shows the receiver operator characteristic curve for the rule and Figure 3 shows the test characteristics of the model with the four high-risk variables. DISCUSSION To the best of our knowledge, our study is the first to report outcomes and to identify risk factors associated with serious underlying pathology in adult nontraumatic ED low back pain patients. We conducted the study as per the methodological standards for chart review to decrease the bias in patient selection and data abstraction (17–19). The proportion of ED visits to our study site that was due to low back pain (1%) is lower than previously reported (3). This could possibly be related to exclusion of traumatic low back pain patients. Our study shows that a small but important number of patients have underlying serious pathology that was not detected during the initial ED visit. This highlights the need for a robust risk-stratification strategy to detect those with serious underlying conditions. We identified four important predictor variables associated with serious outcomes in ED low back pain patients: pain that persists despite appropriate treatment, use of anticoagulant medications, decreased sensation on physical examination, and pain that is worse at night. Most published studies that report risk factors for serious pathology in low back pain patients include traumatic back pain patients and identify significant trauma as a risk factor for fracture (9,20,21). Constant and progressive pain was found to be a risk factor for underlying cancer and spinal infection, and altered sensation was a risk factor for underlying cancer, infection, and fracture in a large primary care study of acute low back pain patients (21). Deyo et al. reported pain that is worse at night and pain that does not respond to treatment as risk factors for underlying cancer (5,6). No studies in the literature report the use of anticoagulant medications as a risk factor. This could be partly explained by the fact that none of the studies has reported acute bleeding conditions such as a retroperitoneal hemorrhage as an outcome. In our study, 2 patients had retroperitoneal bleeding and both required acute treatment. The signs and symptoms associated with cauda equina, urinary retention, urinary or fecal incontinence, and saddle anesthesia were not part of the final model, as there was only one patient in our study who had cauda equina as the underlying cause of low back pain. Our study identifies four high-risk factors for serious underlying conditions in adult ED nontraumatic low back pain patients. Use of these four high-risk factors
6
V. Thiruganasambandamoorthy et al.
Table 4. Univariate Analysis of Continuous Predictor Variables With Discriminative Cut Points and Serious Outcomes Among the 329 Adult Emergency Department Nontraumatic Low Back Pain Patients Variable
Outcome Absent (n = 307)
p Value
65.2 (20.6) 21.5 (39.0) 36.2 (0.9) 136.2 (26.5)
48.1 (17.6) 18.0 (33.2) 37.2 (1.0) 138.3 (21.7)
< 0.0001 0.68 0.35 0.67
75.3 (12.7)
79.1 (12.5)
0.17
112.7 (22.2) 0.58 (0.51)
135.2 (20.7) 0.20 (0.40)
0.001 0.004
12 9 2 4 7
52 34 2 5 16
< 0.0001 < 0.0001 0.02 0.002 < 0.0001
Outcome Present (n = 22)
Continuous predictor variables (Mean, SD) Age Days of low back pain (n = 17,201) Temperature (n = 22,293) Mean triage systolic blood pressure (n = 22,303) Mean triage diastolic blood pressure (n = 22,303) Hemoglobin (n = 1176) INR (n = 1279) Continuous variables with discriminative cut points (n) Age > 65 years Age > 75 years Hemoglobin < 100 g/L Hemoglobin < 110 g/L INR $ 1
SD = standard deviation; INR = international normalized ratio.
would still have missed 2 patients with serious pathology if reapplied to the study population, one with severe disc prolapse and the second with retroperitoneal hematoma. Both patients had their pathology identified during the initial ED visit. Hence, the reported risk factors would
identify all patients with serious pathology not identified during their ED visit. These factors will be helpful to the practicing emergency physician to rule in serious conditions, but the model is not 100% sensitive and hence, should not be used to rule out serious underlying
Figure 1. Identification of predictors of serious outcomes in adult emergency department nontraumatic low back pain patients by recursive partitioning.
Risk Stratification of ED Nontraumatic Low Back Pain
7
had some of the predictor variables missing. We reduced the impact of missing variables by conducting analysis under the assumptions that they were either normal or abnormal. CONCLUSIONS
Figure 2. Receiver operating characteristic curve for the independent predictors of serious outcomes in adult emergency department nontraumatic low back pain patients.
pathology. Future prospective studies are required to derive a robust clinical decision rule. Limitations The limitations of our study are inherent to all medical record review studies. It is possible that our screening process may have missed some nontraumatic low back pain patients. We decreased this possibility by using broad screening criteria and by using all possible fields in the database (chief complaint, primary and secondary discharge diagnoses) to identify patients for screening. The national ambulatory database (NACRS) is a wellestablished database known to capture good quality data on ED visits (22). We excluded patients with previously confirmed nephrolithiasis presenting with typical pain to focus the study on true ED nontraumatic patients who present a dilemma to the practicing emergency physician. While we reviewed medical records within our hospital network, it is possible that patients suffered serious outcomes out of the local area or while in other parts of the country. A proportion of our study patients
Figure 3. Classification performance of the final model for detecting serious underlying pathology in adult emergency department nontraumatic low back pain patients. CI = confidence interval.
Nontraumatic low back pain is a common ED problem. A small but important number of patients have serious underlying pathology that is not detected during their ED visit. Our study has identified high-risk factors associated with serious outcomes. Future prospective studies are required to develop a robust clinical decision rule to identify all patients with serious underlying conditions to improve patient care and patient safety. Acknowledgments—This study was funded by grants from the Canadian Association of Emergency Physicians, and the Department of Emergency Medicine, University of Ottawa. Dr. Thiruganasambandamoorthy receives salary support from the Heart and Stroke Foundation of Canada through the Jump Start Resuscitation Scholarship. We gratefully acknowledge My-Linh Tran, Sheryl Domingo, Dr. Magdalena Kisilewicz, Rina Marcantonio, Lilliana Polesello, Dr. Erin O’Connor, Cathy Clement, RN, and Angela Marcantonio for their help with this study.
REFERENCES 1. Kelsey JL, White AA 3rd. Epidemiology and impact of low-back pain. Spine (Phila Pa 1976) 1980;5:133–42. 2. Mazanec DJ. Back pain: medical evaluation and therapy. Cleve Clin J Med 1995;62:163–8. 3. McCaig LF, Nawar EW. National Hospital Ambulatory Medical Care Survey: 2004 emergency department summary. Adv Data 2006;372:1–29. 4. Cassidy JD, Carroll LJ, Cote P. The Saskatchewan health and back pain survey. The prevalence of low back pain and related disability in Saskatchewan adults. Spine (Phila Pa 1976) 1998;23:1860–6. 5. Deyo RA, Weinstein JN. Low back pain. N Engl J Med 2001;344: 363–70. 6. Deyo RA, Diehl AK. Cancer as a cause of back pain: frequency, clinical presentation, and diagnostic strategies. J Gen Intern Med 1988;3:230–8. 7. Domen PM, Hofman PA, van Santbrink H, Weber WE. Predictive value of clinical characteristics in patients with suspected cauda equina syndrome. Eur J Neurol 2009;16:416–9. 8. Waldvogel FA, Papageorgiou PS. Osteomyelitis: the past decade. N Engl J Med 1980;303:360–70. 9. Davis PC, Wippold FJ 2nd, Brunberg JA, et al. ACR Appropriateness Criteria on low back pain. J Am Coll Radiol 2009;6:401–7. 10. Della-Giustina DA. Emergency department evaluation and treatment of back pain. Emerg Med Clin North Am 1999;17:877–93. 11. Deyo RA, Diehl AK. Lumbar spine films in primary care: current use and effects of selective ordering criteria. J Gen Intern Med 1986;1:20–5. 12. Deyo RA, Rainville J, Kent DL. What can the history and physical examination tell us about low back pain? JAMA 1992;268:760–5. 13. Winters ME, Kluetz P, Zilberstein J. Back pain emergencies. Med Clin North Am 2006;90:505–23.
8 14. Hosmer DW, Lemeshow S. Applied logistic regression. Hoboken, NJ: Wiley; 2000. 15. Pezzullo JC. Interactive statistics page. 2-Way contingency table analysis. Available at: http://www.statpages.org/ctab2x2.html. Accessed March 5, 2014. 16. Scheaffer RL, Mendenhall W, Ott L. Elementary survey sampling. Pacific Grove, CA: Duxbury Press; 1979. 17. Badcock D, Kelly AM, Kerr D, et al. The quality of medical record review studies in the international emergency medicine literature. Ann Emerg Med 2005;45:444–7. 18. Gilbert EH, Lowenstein SR, Koziol-McLain J, et al. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med 1996;27:305–8.
V. Thiruganasambandamoorthy et al. 19. Worster A, Bledsoe RD, Cleve P, et al. Reassessing the methods of medical record review studies in emergency medicine research. Ann Emerg Med 2005;45:448–51. 20. Henschke N, Maher CG, Refshauge KM. A systematic review identifies five ‘‘red flags’’ to screen for vertebral fracture in patients with low back pain. J Clin Epidemiol 2008;61:110–8. 21. Henschke N, Maher CG, Refshauge KM, et al. Prevalence of and screening for serious spinal pathology in patients presenting to primary care settings with acute low back pain. Arthritis Rheum 2009; 60:3072–80. 22. Canadian Institute of Health Information (CIHI). CIHI data quality study of Ontario emergency department visits for fiscal year 2004– 2005—executive summary. Ottawa, ON: CIHI; 2008.
Risk Stratification of ED Nontraumatic Low Back Pain
APPENDIX I: LIST OF PREDICTOR VARIABLES ABSTRACTED FOR STUDY PATIENTS
9
10
V. Thiruganasambandamoorthy et al.
APPENDIX II: LIST OF SERIOUS CONDITIONS DEFINED AS OUTCOMES IN THE STUDY
Risk Stratification of ED Nontraumatic Low Back Pain
ARTICLE SUMMARY 1. Why is this topic important? Low back pain (LBP) is a common Emergency Department (ED) presenting complaint. The frequency of ED visits due to LBP, their outcome and risk factors associated with serious underlying pathology are not known. 2. What does this study attempt to show? Through comprehensive data abstraction and detailed analysis the study identifies risk factors associated with serious underlying conditions in adult nontraumatic LBP patients. 3. What are the key findings? We studied 329 nontraumatic LBP patients. There were 22 patients (6.7%) who had serious underlying conditions causing the back pain. One patient died secondary to upper urinary tract infection, five had compression fractures, four had spinal malignancies, four had disc prolapses requiring surgery, two had retroperitoneal bleed, two had spinal osteomyelitis and one patient with each of epidural abscess, cauda equina and leaking abdominal aneurysm graft. Risk factors for these serious conditions include: anticoagulant use, decreased sensation on physical exam, pain that is worse at night and pain that persists despite appropriate treatment. 4. How is patient care impacted? An important number of nontraumatic LBP patients have a serious underlying cause, if undetected will lead to severe morbidity and even mortality. Future prospective studies are needed to derive a robust risk-tool to identify those at risk.
11