Prospectively registered systematic review, PROSPERO 2018 CRD42018070241.
Data Sources and Searches
An electronic search was conducted in 3 databases (EMBASE, PubMed-Medline, and CINAHL) using search terms related to "back pain," "guideline recommendations," and "medical records" from inception to May 2019 (eTable 1, available as supplemental digital content at https://links.lww.com/PAIN/A913). Further potentially relevant studies were identified through consultation with experts and citation tracking on the included studies.
Studies were included if they (1) reported family practice or ED physicians' assessment and/or treatment of adult patients (aged >18 years) with LBP of any duration and (2) had a quantitative design assessing actual treatment records. Qualitative studies and studies that measured usual care through recall or hypothetical scenarios/vignettes were not included. Studies that analysed data from before 2000 were excluded as we were not interested in historical patterns of practice. Gray literature including non–peer-reviewed literature, theses, and letters to the editor was not included. Non–English-language studies were included and translated as necessary. A full list of the eligibility criteria appears in eTable 2 (available as supplemental digital content at https://links.lww.com/PAIN/A913). Two reviewers independently screened all titles and abstracts, and then potentially eligible full texts. Disagreements were resolved by discussion, arbitrated by a third reviewer as needed.
Two reviewers independently extracted study characteristics and outcomes data into an Excel spreadsheet. Study characteristics included the following: health care setting, LBP duration, period of data collection, data source, and sample size. Family practice was defined as primary-contact outpatient/ambulatory care that was not in ED. For each outcome, data were extracted on collection method, metric used to assess utilisation (eg, proportion or rate), and the denominator used (eg, episodes of care and number of patients). Studies most commonly reported the proportion of patients that received a particular aspect of care, which could have been at a single appointment or over a number of visits, we categorised this as "per patient." Fewer studies reported the proportion of visits that involved that aspect of care, we categorised this as "per episode" of care. One study reported the proportion of physicians that delivered that aspect of care, "per physician."
Reporting and Methodological Quality
The STROBE (STrength of Reporting in OBservational Studies in Epidemiology) Statement and its extension statement entitled RECORD (REporting of studies Conducted using Observational Routinely collected health Data) were used to assess the transparency of reporting. We used items from key domains for assessing susceptibility to bias in observational studies as recommended in Sanderson et al. This includes items in 4 domains: representativeness of the sample, potential misclassification of patients, potential misclassification of outcomes, and inconsistent data. Because we used methodological quality to prioritise interpretation of findings, we also considered precision as an indicator of study quality. Included studies were considered high quality if they met criteria for ≥4 of 5 items.
Findings from included studies were divided by health care setting (family practice or ED) and organised according to outcome category: assessments, treatment advice, imaging, medication, and referrals. Within each outcome category (eg, imaging), individual types are described separately (eg, X-ray, CT [computed tomography], MRI [magnetic resonance imaging], and any image). We planned meta-analysis of single proportions; however, clinical heterogeneity prevented meta-analysis for any outcome. We present all available estimates in the tables and focus our interpretation on the range of estimates that came from the high-quality studies. A narrow range of high-quality estimates from several studies provided greater confidence in the findings and vice versa.
Pain. 2020;161(4):694-702. © 2020 Lippincott Williams & Wilkins