Discover Portal

NIHR Signal A new tool helps predict recovery from ankle sprain

Published on 19 February 2019

doi: 10.3310/signal-000734

The SPRAINED model may improve prediction of people who are at risk of delayed recovery from ankle sprain. This model was developed in the UK using clinical information from 584 adults with ankle injuries.

The model was validated using observational data from 682 people with ankle sprains across 10 different UK emergency departments. Delayed recovery from ankle injury was more likely to be detected when using the SPRAINED model than by clinicians using judgment alone.

Re-assessing pain levels when weight bearing four weeks after the injury improved the strength of the model. The practicalities of using the model and its effect on referrals could now be tested.

  •   Emergency and urgent care, Health management, Musculo-skeletal disorders, Orthopaedics, Trauma, Acute and general medicine
A new tool helps predict recovery from ankle sprain

Why was this study needed?

Ankle sprains are one of the most common musculoskeletal injuries, representing up to 5% of all emergency department attendances. It is estimated that around 30% of people have persistent problems one year after initial ankle sprain.

Physical examination of the ankle is often difficult at first because of swelling and pain. This impedes judgment and can make estimating prognosis uncertain.

This predictive model was developed to include a range of clinical factors that might help decide whether a patient is at risk of having problems nine months after their ankle sprain. The model needed to be tested to see whether it could reliably pick out those who, at nine months, did have on-going problems.

What did this study do?

This modelling study aimed to develop and externally validate a tool to predict poor recovery after ankle sprain, using two different datasets. Both datasets include people with ankle sprains. The first developmental model included 584 people from a UK multicentre randomised controlled trial (CAST study) between April 2003 and July 2005. The second test data set included 682 people attending 10 different UK emergency departments for ankle injuries between July 2015 and March 2016 (SPRAINED cohort).

Clinical indicators with the strongest relationship to poor recovery were selected from those reported in the CAST study. The ability of these indicators to detect people at risk of poor recovery was tested using statistical analysis and data from the SPRAINED cohort.

What did it find?

  • The number of ankle injuries was lower in the CAST study (6.7%) than in the SPRAINED cohort (19.9%)
  • Age, body mass index, pain when resting, pain when weight bearing, the ability to bear weight, days from injury until assessment and injury reoccurrence were the selected clinical indicators used as predictors of poor recovery.
  • The SPRAINED model performed well at predicting poor recovery nine months after initial injury (C-statistic 0.72; 95% confidence interval [CI] 0.66 to 0.79; where 0.5 signifies no predictive or discriminatory ability and 1.00 would be perfect). 
  • When adding the ability to bear weight at four weeks after injury, the predictive value of the model improved (C-statistic 0.78; 95% CI 0.72 to 0.84).

What does current guidance say on this issue?

The NICE guideline for sprains and strains published in March 2016, includes a section on the management, including when to refer, choice of analgesia, how to follow up the person, and how to prevent further injury. These include pain relief, injury management advice and how to avoid harm in the first 72 hours after the injury. This includes considering the need to immobilize the joint and review the strain after a few days to determine the need for referral to an orthopaedic specialist or routine referral for physiotherapy. 

What are the implications?

The SPRAINED model can be used to help clinicians in emergency departments advise patients of their likelihood of poor recovery from ankle sprain. This may help target rehabilitation or physiotherapy to those with the greatest need. The clinical indicators used as predictors are also easy for clinicians to assess in emergency departments. 

It would be useful to test the model in larger numbers of people with ankle sprains and to follow the consequences of its use in terms of important outcomes like return to sports and to work. Use of the model may lead to changes in referrals to other services such as physiotherapy.  

Citation and Funding

Schlussel MM, Keene DJ, Collins GS et al. Development and prospective external validation of a tool to predict poor recovery at 9 months after acute ankle sprain in the UK emergency departments: the SPRAINED prognostic model. BMJ Open. 2018;8(11):e022802.

The study was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme (project number 13/19/06).

Bibliography

NHS website. Sprains and strains. London: Department of Health and Social Care; 2018.

Expert commentary

In the emergency department, predicting poor recovery following an acute ankle sprain can be difficult. This study highlights a number of factors taken from routine examination at baseline and one taken four-weeks later that may help predict those patients at high-risk of poor recovery at nine months.

In the future, the tool may assist clinicians to decide which patients will benefit the most from treatment and in some cases, those who may require little or no treatment.

Further research is needed to determine the clinical and cost effectiveness of using such a tool to assist decision making in this patient population.

Sarah Harrisson, Research Associate, Arthritis Research UK Primary Care Centre, Keele University

The commentator works in the same research institute as, but not on any collaborations with Professor Richard Riley who was on the SPRAINED study steering group

Definitions

External validation refers to evaluating the models performance using a separate dataset to the one the model was developed on. External validation was tested using C-statistics, which is a measure of goodness of fit. C-statistics give the probability a randomly selected patient who experienced an event, had a higher score than a patient who had not experienced the event. A value over 0.7 indicates a good model.