Fourteen studies (3,980 participants, median 117, range 41 to 1,425) were included in the review. For the patient selection and flow and timing domains, risk of bias was generally considered low and there were few concerns about the applicability of the study populations. Details of blinding were often poorly reported so risk of bias for the index test and reference standard domains was frequently unclear. Variation in content of comprehensive geriatric assessment led to concerns about applicability in the reference standard domain.
Prevalence of frailty varied widely between study populations (median 68%, range 28% to 94%).
For the Vulnerable Elders Survey-13 (nine studies), median sensitivity and specificity values were 68% (range 39 to 88%) and 78% (range 62% to 100%).
For Geriatric 8 (six studies), the median sensitivity and specificity values were 87% (range 77% to 92%) and 61% (39% to 75%).
The Groningen Frailty Index was assessed by three studies. Reported sensitivities ranged from 39% to 62% and specificities ranged from 69% to 87%.
The Triage Risk Screening Tool was assessed by two studies (three data sets) that reporting sensitivities of 64% to 92% and specificities in the range 43% to 100%.
Two studies that assessed the Fried frailty criteria reported sensitivities of 25% and 37% and specificities of 96% and 86%.
The Comprehensive Geriatric Assessment was assessed by one study that found sensitivity to be 51% and specificity to be 97%.
Barber was assessed by one study that found sensitivity to be 59% and specificity to be 79%
Although there was a high prevalence of frailty in most study populations, even those screening methods with the highest sensitivity yielded negative predictive values of around 60%. Full details of positive and negative predictive values were reported in the article.