Spine Infection Risk Calculation Heuristic Data Set I

Published: 23 January 2018| Version 4 | DOI: 10.17632/nnvxsg2ffz.4
Contributor:
Steven Shroyer

Description

Data: This is a database of 41 clinical variables (mostly dichotomous) collected non-consecutively and prospectively on 180 patients who presented to the emergency department with back/neck pain or radicular equivalent over a 6 year period in whom the physician was suspicious enough of spinal infection to order a CRP and/or MRI (or other advanced imaging) to definitively rule-in or rule-out infection. 180 patients were enrolled, 8 excluded, 43 spinal infections and 129 non-infected patients were identified. From this data a univariate analysis can be performed to identify history/physical exam/lab findings that differ significantly between patient groups having infection and No-infection. Definitions of historical variables and derivation of cut-off values for fever, CRP are provided in accompanying files as well as a univariate analysis of this data.

Files

Steps to reproduce

The sample of patients we selected were patients presenting to an emergency department with back or neck pain in whom there is enough suspicion of spine infection that a CRP and/or MRI (or other advanced imaging) was ordered to definitively rule-in or rule-out spine infection. Suspicion of infection generally meant the patient had back/neck pain and any of the following: fever, history of fever, chills, sweats, purported historical risk factors, any progressive neurological deficit including overflow incontinence, any recent spinal surgical procedure, multiple recent visits for back or neck pain and lastly, unusually severe back pain. Once ordered, the completion of MRI could not be ethically be performed if the CRP result returned low and if the clinician's suspicion of infection now also low. These patients were followed clinically without MRI at 2 months and possibly up to 4 months to verify no infection was present. A report of follow up type and the number lost to follow up must be maintained. 17 of the 41 variables we evaluated should be collected before the MRI is completed on each patient with careful attention to avoid any missing data and then perform a univariate analysis and subsequent logistic regression to identify independent variables that predict spine infection. Using the four resulting variables a prediction tool can be derived to decide which patients must have an MRI and which are so low risk that no lab tests or MRI are necessary. Our scoring system assigned weighted values to each of the four variables (1. presence of any historical risk factor=value of one; 2. fever/history of fever=value of two; 3. progressive neurological deficit=value of one; and 4. CRP=value of two). A summed score equal to or greater than 3 (called SIRCH ≥ 3) mandates an MRI; and patients having the triad of NO historical risk factors, NO fever or history of fever and NO progressive neurologic deficit (called SIRCH-negative) were deemed so low risk that there was no need any laboratory testing (including CRP) or advanced imaging (eg MRI). A validation study should consider also collecting ESR and procalcitonin on each patients as these may have superior diagnostic characteristics compared to CRP. The validation study should collect the 17 variables in each suspected patient and compare the diagnostic characteristics of the two derived tools (SIRCH ≥ 3 and SIRCH-negative) with the composite gold standard (MRI (or other advanced imaging) and clinical follow up of lower risk patients). Future validation would also compare diagnostic characteristics with our derivation. Our study goal was to create a tool with a high clinical bedside sensitivity (miss few spinal infections), without over-utilizing MRI. Any sensitivity greater than the historical 50% would be an improvement; and less MRI orders prompted by the tool than actually ordered in this cohort was considered an improvement in MRI utilization over "a clinician's high index of suspicion."