We aimed to establish an algorithm to identify older adults at risk of falling using Classification Regression Trees (with CHAID). We used longitudinal data (2012–2014) with 1662 participants completed the second assessment. The following risk factors for falls were entered in the CHAID: age, sex, BMI, multimorbidity, cognitive deficit, depression, number of falls in last year, fear of Falling (FOF), chair-rise speed, balance, and gait speed. The CHAID identified six end nodes with three levels of partitions and four partitioning variables: number of falls, FOF, chair-rise score, and age group. CHAID identified three subgroups based on number of falls (none, one, ≥ two). Then, the ‘no falls’ subgroup was split using FOF into high and moderate FOF groups. Those with multiple falls were also split by their chair-rise speed into ‘slow or fast’ and intermediate speed groups. Then the ‘slow or fast’ group was further split into two age groups (64–69 and 70–75). Overall, those with more than two falls last year, slow or fast chair-rise speed, and aged 64–69 years had the greatest odds of falls (OR: 5.25, 95%CI: 3.16–8.88). While those who had no falls last year and low or medium FOF had the lowest odds of falls (0.42, 95%CI: 0.33–0.52).