Introduction:
Many tools have been used to try to risk stratify older patients, with a growing body of literature in head and neck surgery that demonstrates the importance of measuring frailty to aid in the identification of patients at risk of complications, increased length of stay (LOS) and discharge to alternative levels of care. There are two main conceptual models of frailty, one being a phenotypic model and the other being the deficit accumulation model. The deficit accumulation model incorporates more deficits than the phenotypic model and thus, may be a better tool to predict adverse outcomes following major head and neck surgery.

Objective:
Develop a deficit accumulation frailty score to measure frailty and evaluate whether there is an association between the deficit accumulation frailty score with post-operative complications and length of stay and secondarily to compare its ability to predict these outcomes with that of the previously evaluated phenotypic model of frailty.

Methods: A prospective cohort study was performed with patients undergoing major head and neck cancer procedures between December 2011 and April 2014 at a referral center. Patients frailty was assessed according to Fried's Frailty Score (-FFS- a phenotypic model) and a deficit accumulation model (DAM). The DAM was created selecting rigorous criteria such as; association with health status, prevalence increase with age, cover a range of systems. A total of 40 deficits were included and expressed as a ratio. The outcomes included LOS, complications within 30 days of surgery, and overall survival. Complications were assessed based on having any complication, as well as type (i.e. medical vs surgical complications) and grade of complication (Clavien-Dindo grading system).

Results: A total of 274 patients were enrolled. The mean age of the entire cohort was 67.8 years (range 50 to 88 years) with 96 patients aged 55 to 64 and 129 aged 65 and older. In univariate models, neither the FFS or DAM were significant predictors of overall complications. However, on multivariate analysis both indices were independent predictors for medical complications while controlling for tumor site, age and operative hours (p=0.046). None of them was significant for surgical complications. When adjusting for adjusting for tumor site, free flap and operative hours the DAM was an independent predictor of increased LOS (p=<0.001). Similarly, when controlling for age, cancer stage and sex, DAM was also significant (p=0.041).

Conclusion: We present a direct comparison of these models, constructing a cumulative model following specific criteria and a prospective referral center database. The cumulative model performed better predicting medical complications, LOS and survival compared with the phenotypical model. We hope that frailty research in head and neck cancer continues to evolve, aiming to endorse the importance of investigating frailty as a method and to incorporate it routine practice.