Keene D.J.1, Schlüssel M.M.1, Hagan D.1, Thompson J.1, Williams M.A.2, Byrne C.1, Gwilym S.E.1,3, Goodacre S.4, Cooke M.W.5, Hormbrey P.6, Bostock J.7, Collins G.S.1, Lamb S.E.1
1University of Oxford, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, Oxford, United Kingdom, 2Oxford Brookes University, Department of Sport and Health Sciences, Oxford, United Kingdom, 3Oxford University Hospitals NHS Foundation Trust, Trauma and Orthopaedic Surgery, Oxford, United Kingdom, 4University of Sheffield, School of Health and Related Research, Sheffield, United Kingdom, 5University of Warwick, Coventry, United Kingdom, 6Oxford University Hospitals NHS Foundation Trust, Emergency Department, Oxford, United Kingdom, 7Patient and Public Representative, n/a, United Kingdom
Background: Ankle sprains are one of the most common musculoskeletal injuries. Although recovery can occur within weeks, a third of patients still have problems with their ankle at one year. In the acute phase there is no reliable way of establishing which patients are at risk of having a poor outcome.
Purpose: To develop and externally validate a prognostic model for identifying patients at increased risk of poor outcome after an acute ankle sprain.
Methods: Stage 1: The model was developed using data from 584 ankle sprain patients participating in a randomised controlled trial, in 8 emergency departments in the UK. Twenty-three candidate predictors were examined: age, sex, sprain severity, pain, previous injury, ankle stability tests, weight bearing ability, and severity of presenting clinical signs and symptoms. Two composite outcomes at 9 months indicating poor recovery were investigated and defined as the presence (or a combination) of either pain, functional difficulty, lack of confidence in the ankle (outcome 1) or reoccurrence of ankle sprain
(outcome 2). Multiple imputation was used to handle missing data with imputed 10 datasets created. Multivariable logistic regression models using backwards elimination were used to identify the statistically significant predictors for each outcome. Predictive accuracy of the models was evaluated by assessing model discrimination (quantified by the c-statistic) and model calibration (calibration plot).
Stage 2: External validation of the prognostic model will be done using data from a prospective cohort study of 682 patients from 10 emergency departments across the UK. Data on the prognostic variables identified in stage 1 have been collected at baseline and 4 weeks with the outcome assessed at 4 and 9 months.
Results: Stage 1: At baseline, patients were aged 30 (SD 11; range: 16 to 72) years and all were within 7 days of injury. Discrimination (c-statistic) was 0.77 (95% CI: 0.75 to 0.79) for the first model (outcome 1) and 0.73 (95% CI: 0.70 to 0.75) for the second model (outcome 2). The predictors included in the final model for both outcomes were age, pain at rest, pain on weight bearing, days from injury, weight bearing ability, injury mechanism, employment status and previous sprain. Body Mass Index and ankle catching/locking were specific to outcome 1. Ability to dorsiflex and plantarflex the ankle were specific to outcome 2.
Stage 2: The external validation study recruitment was completed in March 2016 and follow-up will be completed in November 2016, results from this stage of the research will be presented.
Conclusion(s): Key patient characteristics and clinical findings that predict a poor outcome with a good model discrimination were identified in a large cohort of ankle sprain patients. The external validation cohort study will determine the predictive accuracy of the prognostic model in an independent data set.
Implications: This prognostic model has the potential to help identify patients at increased risk of a poor outcome and so enable targeted monitoring and rehabilitation.
Funding acknowledgements: UK National Institute for Health Research Health Technology Assessment (NIHR HTA) Programme (ref. 13/19/06).
Topic: Musculoskeletal: peripheral
Ethics approval: National Research Ethics Service Committee London - Chelsea (ref. 15/LO/0538).
Study registration: ISRCTN12726986
All authors, affiliations and abstracts have been published as submitted.
