Machine learning for frailty assessment and outcome prediction in cardiovascular disease: a scoping review
Quach J, Theou O, Maxwell S, Abidi SSR, Song X, Rockwood K, Kehler DS. Ageing Res Rev. 2026;118:103122.
This scoping review found 31 studies (2016–2025) that used machine learning with frailty measures in people with cardiovascular disease, 71% of them in heart failure. The frailty phenotype, frailty index and Clinical Frailty Scale were the most used of 13 frailty measures. Reported AUCs were 0.80 to 0.96 for predicting frailty and 0.70 to 0.93 for predicting death. Only 3 studies tested their models in external data, and 4 reported calibration.
Cardiac rehabilitation and frailty: a systematic review and meta-analysis
MacEachern E, Quach J, Giacomantonio N, Theou O, Hillier T, Abel-Adegbite I, Gonzalez-Lara M, Kehler DS. Eur J Prev Cardiol. 2024;31(16):1960–1976.
This review pooled 34 studies (19,360 participants) of frailty in cardiac rehabilitation. Frailty at the start of rehabilitation was common: 46% by frailty index or Kihon Checklist (14 studies) and 40% by frailty phenotype (11 studies). Frailty improved after rehabilitation (6 studies), and higher frailty at the start was linked to higher mortality.
The association of frailty on cardiac rehabilitation goal achievement
MacEachern E, Quach J, Giacomantonio N, Theou O, Hillier T, Firth W, Kehler DS. Front Cardiovasc Med. 2024;11:1441336.
The study included 759 people referred to cardiac rehabilitation (2005–2015; mean age 59.5, 24% female), and 80% of them achieved a rehabilitation goal. People who were less frail at the start, on a 25-item frailty index, were more likely to reach a goal. Those with an index below 0.20 had about 4.7 times the odds of those above 0.40.
Comparing virtual and center-based cardiac rehabilitation on changes in frailty
MacEachern E, Giacomantonio N, Theou O, Quach J, Firth W, Abel-Adegbite I, Kehler DS. Int J Environ Res Public Health. 2023;20(2):1554.
The study included 132 people in a Halifax cardiac rehabilitation program (2021–2022). It compared virtual rehabilitation with a centre-based program that COVID-19 restrictions had shortened to 6 weeks. Frailty was measured on a 65-item frailty index. Overall, the type of program made no difference to the change in frailty. People who were frailer at the start improved more with virtual than with centre-based rehabilitation.
Association of admission frailty and frailty changes during cardiac rehabilitation with 5-year outcomes
Quach J, Kehler DS, Giacomantonio N, McArthur C, Blanchard C, Firth W, Rockwood K, Theou O. Eur J Prev Cardiol. 2023;30(9):807–819.
The study followed 3,371 people in a 12-week cardiac rehabilitation program in Halifax, Nova Scotia (2005–2015). Frailty was measured with a 25-item frailty index at the start and end. People who were frailer at the start had a higher risk of death, hospital admission and emergency visits over 5 years. Frailty improved by 0.07 on average. Improvement was linked to a longer time before any hospital admission, but not to the other outcomes.
The impact of cardiovascular health and frailty on mortality for males and females across the life course
Quach J, Theou O, Godin J, Rockwood K, Kehler DS. BMC Med. 2022;20(1):394.
The study included 35,207 US adults aged 20 to 85 (NHANES 1999–2015), followed for a median of 8.1 years. It combined a 33-item frailty index (cardiovascular items left out) with the American Heart Association's Life's Simple 7 score. In females, frailty and poorer cardiovascular health were together linked to a higher risk of death, and the combined risk was similar across ages. In males, the combined risk was higher at ages 60 and 70, but not at 30 to 50.
