Towards improving the identification of anterior cruciate ligament tears in primary point-of-care settings

Background Only a small proportion of anterior cruciate ligament (ACL) tears are diagnosed on initial healthcare consultation. Current clinical guidelines do not acknowledge that primary point-of-care practitioners rely more heavily on a clinical history than special clinical tests for diagnosis of an ACL tear. This research will assess the accuracy of combinations of patient-reported variables alone, and in combination with clinician-generated variables to identify an ACL tear as a preliminary step to designing a primary point-of-care clinical decision support tool . Methods Electronic medical records (EMRs) of individuals aged 15–45 years, with ICD-9 codes corresponding to a knee condition, and confirmed (ACL + ) or denied (ACL − ) first-time ACL tear seen at a University-based Clinic between 2014 and 2016 were eligible for inclusion. Demographics, relevant diagnostic indicators and ACL status based on orthopaedic surgeon assessment and/or MRI reports were manually extracted. Descriptive statistics calculated for all variables by ACL status. Univariate between group comparisons, clinician surveys ( n = 17), availability of data and univariable logistic regression (95%CI) were used to select variables for inclusion into multivariable logistic regression models that assessed the odds (95%CI) of an ACL-tear based on patient-reported variables alone (consistent with primary point-of-care practice), or in combination with clinician-generated variables. Model performance was assessed by accuracy, sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios (95%CI). Results Of 1512 potentially relevant EMRs, 725 were included. Participant median age was 26 years (range 15–45), 48% were female and 60% had an ACL tear. A combination of patient-reported (age, sport-related injury, immediate swelling, family history of ACL tear) and clinician-generated (Lachman test result) variables were superior for ACL tear diagnosis [accuracy; 0.95 (90,98), sensitivity; 0.97 (0.88,0.98), specificity; 0.95 (0.82,0.99)] compared to the patient-reported variables alone [accuracy; 84% (77,89), sensitivity; 0.60 (0.44,0.74), specificity; 0.95 (0.89,0.98)]. Conclusions A high proportion of individuals without an ACL tear can be accurately identified by considering patient-reported age, injury setting, immediate swelling and family history of ACL tear. These findings directly inform the development of a clinical decision support tool to facilitate timely and accurate ACL tear diagnosis in primary care settings.