Low specificity and high false-positive rates limit the usefulness of the STRATIFY tool and clinical judgement in predicting falls in older patients in an acute hospital setting
- Correspondence to Frances Healey
National Patient Safety Agency, 4–8 Maple Street, London W1T 5HD, UK;
This paper adds to the increasing evidence that even the best of risk scores may not predict falls accurately enough to be clinically useful. As the authors describe, falls in hospitals are a serious patient safety issue, with more than 1000 falls reported annually from larger hospitals1 and between 1% and 4% of falls resulting in serious injury.
Comparing falls screening methods
The study took place in an Australian acute hospital and compared STRATIFY scores calculated by research nurses with responses by ward nurses to the question “Do you think this patient is at risk of falling?” Overall predictive value was higher with the STRATIFY tool than nurse judgement (p=0.027), although nurse judgement led to fewer false negatives. The authors conclude that the high proportion of false positives means that neither method is clinically useful.
Can the findings be generalised to other hospitals?
The trial was generally well designed, and those seeking further information on the statistical methodology can find it in a related publication.2 However, ward nurses were asked to judge whether the patient was at risk of falling after answering questions on whether the patient had a history of falls or was agitated, for example (to enable the research nurses to calculate the STRATIFY score). This may have affected their judgement and therefore the validity of the study. A further issue is that, although all falls throughout the inpatient stay were analysed, STRATIFY scores were calculated and nurses were questioned within 48 h of admission; STRATIFY scores and nurses' judgement may change as the patient recovers or deteriorates.
Can falls in hospital be predicted?
The findings of this study are in line with a recent systematic review3 that found that nursing judgement, the Morse falls score and STRATIFY had similar, moderate levels of predictive accuracy and that all three methods had high levels of false positives, making them unlikely to be of clinical value in most settings. However, it is important to remember that we know the limitations of the Morse and STRATIFY scores because they are the only scores that have undergone systematic validation, and the accuracy of all other published tools is unknown.3 Least accurate of all are ‘home-made’ falls prediction tools.1
What does this evidence mean for nursing practice?
Although it is common for researchers to critique each other's work, it is rare for them to critique their own; however, Professor David Oliver, the author of STRATIFY, discourages staff from using it—or indeed from using any falls prediction tool.4 This relates not just to the predictive value discussed above but to the clinical practicalities of falls prevention. Vastly more research effort has been directed towards the development of falls risk prediction tools to identify high-risk patients so that ‘something can be done’ to prevent falls than has gone into investigating what the ‘something’ might be. But we do know that multiple actions aimed at those risk factors that can be changed—such as unnecessary sedative medication, incontinence, delirium, or poor vision—can reduce falls.5 These factors should be identified and acted on as part of good nursing assessment and medical review rather than considered only for patients at high risk. Is it, for example, acceptable to ignore unsafe footwear simply because the patient has a low falls risk score?
Understand your local score
If your hospital uses a falls score—whether this is Morse, STRATIFY or a local tool—you need to understand how well it works for your inpatient population, because every false negative represents a missed opportunity for falls prevention, and too many false positives can divert resources from the patients most at risk. An automatic calculator6 can be found at http://www.patientsafetyfirst.nhs.uk. Whether you are using a risk score or clinical judgement, prediction will be of no value unless it is followed by prevention, and assessment will be of no value unless it is followed by effective intervention.