Identifying frailty in primary care

Frailty is a common accompaniment to aging, bringing reduced resilience to acute problems compared with healthier people.1 Recovery takes longer and is sometimes incomplete. Falls are frequent. Polypharmacy—much of it futile—is common.2 In theory, if frail people could be identified, some of these problems could be averted. This is the rationale behind the requirement for general practitioners in England to identify severely frail people on their lists, to review them for risk of falls, and to ensure their treatment is suitable.3 Frailty is not the same as mortality, although it shares several predictors. However, mortality is more amenable to prediction, so it may be worthy of consideration as a proxy for frailty.

In this issue, Hippisley-Cox and Coupland (doi:10.1136/bmj.j4208) have adapted their well oiled methods of examining a large English primary care database to produce a predictive algorithm for short term mortality (QMortality).4 They also develop a new classification of frailty, combining risk of death with risk of hospital admission (QFrailty categories), building on an existing electronic frailty index (EFI) from a similar English database, and a smaller Dutch study.56

Both English studies are large, are well conducted, and crucially use primary care data, unlike previous instruments.7 The EFI predicts one of three outcomes (death, unplanned hospital admission, or nursing home admission) and so is probably closer to a true definition of frailty than QMortality/QAdmissions, which omit nursing home admission. QMortality is better at predicting death than the EFI, probably reflecting the different weightings for each variable, whereas the EFI just scores each variable as present or absent.

Both algorithms only use variables typically stored in primary care records; none the less, missing data will weaken their performance in the real world.7 Crucially, when either QMortality or the EFI is used to select the “worst 2%” of patients in a general practice, many patients are misclassified as being false positives or false negatives. It is also likely that a substantial proportion of those identified by the EFI or QMortality would already be under regular review for chronic disease. So, we have helpful but imperfect tests for possible use, although no primary care algorithm measures frailty directly.

If the QMortality and the QFrailty instruments were implemented, are doctors honest enough to say to a patient, “You’re in the worst 2% for risk of death or unplanned hospital admission—let’s discuss drugs and falls”? From a patient perspective, being placed among the worst 2% is unlikely to be welcomed, especially if there is an even chance that the categorisation is wrong. Patients are very receptive to early diagnosis if the benefit is clear, either for ameliorating symptoms or averting death.8 When it comes to frailty, most patients are aware of their problems, and many are content to live within their limitations. Being told you are in the “mortality relegation zone” when there are few, if any, personal benefits available from intervention is much less attractive—arguably insulting.

Even if such terminology problems could be surmounted, how big are the benefits from interventions aimed at reducing polypharmacy and falls? A Dutch trial of a primary care programme to preserve daily function, using their frailty index, found small differences in favour of intervention, but these were of doubtful clinical value.9 A recent systematic review and meta-analysis of trials of de-prescribing in elderly patients reported no reduction in mortality, falls, or adverse events, nor did the review report any improvements in quality of life (other than in one trial), despite a decrease in the numbers of drugs used.10 The optimum methods for de-prescribing in primary care are uncertain, as is the cost effectiveness.

In contrast, good evidence supports intervention to reduce falls: a systematic review of exercise programmes reported a pooled rate ratio for falls, leading to an injury in the exercise group, of 0.63 (95% confidence interval 0.51 to 0.77), although no number needed to treat was calculated.11 Other interventions to reduce falls—or mitigate their impact—such as vitamin D supplementation and hip protectors may also be cost effective.12 Whether these interventions are best targeted at the worst 2% (however identified) is unknown. It is possible they would be better targeted at less frail patients.

This all adds up to a problem for primary care, and particularly English primary care, with the new contractual obligation for general practitioners to identify all those on their list older than 65 years with severe frailty. Is it appropriate to screen practice lists using a proxy outcome for frailty, such as risk of death or hospital admission? Such a policy arguably fails on several of the classic Wilson and Jungner criteria for screening, including those relating to patient acceptability and cost effectiveness, let alone the suitability of the screening test.13 The timing is poor too. The number of general practitioners in the UK is stable or falling, yet workload increased by 16% between 2007-8 and 2013-14, and probably continues to rise.4 This is not to downplay the importance of managing polypharmacy or falls. Even so, the existence of a problem such as frailty does not presuppose the existence of an effective solution—or even a flawed one.