Homecare after cancer surgery for older adults: derivation and validation of a predictive tool

Background — Need for home support after surgery is a patient-centred outcome and marker of functional recovery for older adults. We developed a risk prediction model, HOMECARE, to estimate the risk of using homecare for older adults after cancer surgery.

Methods — We conducted a population-based retrospective cohort study of adults ≥ 70 years having surgery for solid malignancy (2007–2019). Candidate predictors were preoperative sociodemographic and clinical factors. Receipt of immediate (within 1 month) and chronic (at 7–12 months) homecare was predicted. Internal validation used bootstraps with 500 samples with replacement. Logistic regression models were used. The predictive model included age, sex, rural residence, previous cancer diagnosis, frailty, prior homecare use, cancer site, cancer stage, and type of surgery. We performed bootstrap validation by using 500 samples with replacement.

Results — Of 93,883 patients included, 39,169 (41.7%) required immediate homecare; of the 88,252 alive after month 6 postoperatively, 22,031 (25%) required chronic homecare. For immediate homecare, the area-under-the-curve was 0.77 and the deviation of predicted from observed probability was − 0.002% (95% CI 0.004 to − 0.009). For chronic homecare, the area-under-the-curve was 0.76, and the deviation of predicted from observed probabilities was − 0.004% (95% CI 0.002 to − 0.009). Deviation between predicted and observed probabilities ranged from − 0.04 to 0.03% across risk deciles for immediate homecare and − 0.05 to 0.04% for chronic homecare.

Conclusions — The HOMECARE tool presents good discrimination and is well calibrated. Implemented as an online calculator, individualized risk estimates from this tool could support risk communication with older adults selected for cancer surgery.