Clinicians are less likely to recommend adjuvant chemotherapy for older adults based on their age alone. This study aimed to develop a mortality risk model to assist treatment decision making by identifying patients who are unlikely to live to benefit from chemotherapy. All lymph node-positive colon cancer patients ≥65 years who received surgery in New South Wales, Australia in 2007/2008 were identified using a linked population-based dataset (n=1550). A model predicting 1-year all-cause mortality was built using multilevel logistic regression. Risk scores derived from model factors were summed for each patient. One-year mortality was 11.5%. The risk model consisted of 14 factors, including comorbidities, hospital admission factors and other markers of frailty or health status. People with a total score of 0, 1 or 2 were considered at low risk (predicted 1-year mortality of 2.9%), those scoring 3 to 8 at medium risk (7.4% mortality) and those scoring 9 or above at high risk (24.7% mortality). The model had good discrimination (area under the receiver operating characteristic curve=0.788, 95% confidence interval: 0.752-0.825) and calibration (P=0.46). The risk model accurately predicts mortality for this cohort and could be useful in shifting the emphasis in chemotherapy decision making from chronological age to the identification of those of any age who will benefit.