Sensitivity and specificity of patient-entered red flags for lower back pain

Abstract Background Context Red flags are questions typically ascertained by providers to screen for serious underlying spinal pathologies. The utility of patient-reported red flags in guiding clinical decision-making for spine care, however, has not been studied. Purpose The aim of this study was to quantify the sensitivity and specificity of patient-reported red flags in predicting the presence of serious spinal pathologies. Study Design This was a retrospective nested case-control study. Patient Sample This study consisted of 120 patients with International Classification of Diseases, Ninth Revision, Clinical Modification codes for spinal pathologies and 380 randomly selected patients, from a population of 4,313 patients seen at a large tertiary care spine clinic between October 9, 2013 and June 30, 2014. Outcome Measures The presence of patient-reported red flags and red flags obtained from medical records was verified for chart review. The spinal pathology (ie, malignancy, fractures, infections, or cauda equina syndrome) was noted for each patient. Methods The sensitivity and specificity of patient-reported red flags for detecting serious spinal pathologies were calculated from data obtained from the 500 patients. Youden's J was used to rank performance. Agreement between patient-reported red flags and those obtained from medical record review was assessed via Cohen's kappa statistic. Results “History of cancer” was the best performing patient-reported red flag to identify malignancy (sensitivity=0.75 [95% confidence intervals, CI 0.53–0.90], specificity=0.79 [95% CI 0.75–0.82]). The best performing patient-reported red flag for fractures was the presence of at least one of the following: “Osteoporosis,” “Steroid use,” and “Trauma” (sensitivity=0.59 [95% CI 0.44–0.72], specificity=0.65 [95% CI 0.60–0.69]). The prevalence of infection and cauda equina diagnoses was insufficient to gauge sensitivity and specificity. Red flags from medical records had better performance than patient-reported red flags. There was poor agreement between patient red flags and those obtained from medical record review. Conclusions Patient-reported red flags had low sensitivity and specificity for identification of serious pathologies. They should not be used in insolation to make treatment decisions, although they may be useful to prompt further probing to determine if additional investigation is warranted.