Economic Modeling on Cost‐effectiveness of Hip Protectors in Institutionalized Older People Based on Contentious Interpretation of Original Effectiveness Data

To the Editor: Honkanen et al.1 describe a cost-benefit analysis, based on an economic model, of the use of external hip protectors in institutionalized older adults. We believe that the model is based on selective assumptions that are potentially misleading. The analysis, as published, would lead the reader to believe that the mass purchase of hip protectors for routine use in care homes would lead to a significant reduction in hip fractures with attendant financial savings. The authors made three basic assumptions as a basis for their model. The relative risk of sustaining a hip fracture by the use of hip protectors is 0.43. The likely mean adherence to hip protector use is 48% (on a day-to-day basis rather than as a percentage of patients still wearing the devices over time) based on one cited study. Hip fractures are rare in patients who are wearing the devices at the time of a fall. We believe that the first two of these assumptions are flawed for the following reasons. The systematic review cited by the authors in defense of their choice of effect size comes from cluster randomized trials2 and fails to incorporate studies published since 2001. Yet the updated Cochrane Systematic Review in 2004 that the authors cite3 concluded that there was insufficient evidence to recommend the use of hip protectors, although they may have a role in selected high-risk patients. This view has been endorsed in the current falls guidelines from the UK National Institute for Health and Clinical Excellence.4 A systematic review, which is about to be published, uses different techniques for aggregation of data.5 Four trials involving institutionalized older people were included, with a pooled risk difference of −4.0% for those allocated to hip protector use. Although the relative risk of 0.53 in this meta-analysis seems impressive, the numbers needed to treat to prevent one fracture was 25—with wide confidence intervals (13–200). (Again, this is somewhat irrelevant. The public health payoff would still be there; what matters is whether the point effect of 0.53 is in the range of their sensitivity analysis.) In the Cochrane analysis, of the nine trials reported from care homes, only those that were cluster randomized showed a significant pooled effect on hip fracture rates, as opposed to those randomized at the individual level, which did not. Pooling of the five individually randomized trials showed no significant reduction in fracture incidence. The CONSORT guidelines on the design and interpretation of cluster randomized trials6 point to the main concern, where lack of a design effect is incorporated, with the result that reported standard errors are smaller than they should be. Moreover, a key paper from Northern Ireland7 published shortly before Honkanen's article describes the largest recruitment and longest intervention period in any cluster randomized trial, with 43 out of 120 care homes being randomized to hip protector use over 2 years. Using appropriate statistical analysis for a cluster randomized sample, the effect of external hip protectors on hip fracture rates was null. Adherence was also considerably lower (27%) than that described by Honkanen et al. Indeed, a systematic review of adherence to hip protectors8 has acknowledged wide variation in rates of initial acceptance of hip protectors and equally wide variations in ongoing rates of adherence and in how adherence is defined. Consequently, contrary to Honkanen et al.'s original assumption (unless adherence means 24 hours a day, 7 days a week—or at least at those times when patients are ambulant), the real adherence rates may be considerably lower than they assumed. Finally, it can be postulated that those very residents at highest risk of fracture (i.e., those with risk factors such as agitation, restlessness, gait instability, and incontinence) are those least likely to wear the devices, which may be responsible for the apparent lack of effect in individually randomized controlled trials and in the null results of the Northern Ireland study. It would be a mistake for commissioners of health care from care homes to be influenced by this article that hip protectors are a sound way to invest money on a population basis. There is better evidence for the use of fall-prevention strategies or calcium and vitamin D supplementation and certainly for different flooring types in this setting, although even in these instances, the case is not yet compelling. Financial Disclosure: None. Author Contributions: David Oliver, Theresa Allain, and James Connelly: interpretation of data and preparation of manuscript. Sponsor's Role: None.