The aim of this study was to derive a screening tool for predicting the risk of falls in community-dwelling older adults with knee osteoarthritis. This prospective cohort study was conducted in four orthopedic clinics. The main outcome measure was falls or nonfalls for 5 months, and the predictors were sex, age, BMI, Kellgren–Lawrence grade, laterality, number of comorbidities, pharmacotherapy, physical therapy period, use of a cane, household, previous history of falls, visual analog scale for pain, one-leg standing test (OLST), five times sit-to-stand test (FTSST), and Frenchay activities index. Ninety outpatients (74 females and 16 males) with a mean (SD) age of 73.1 (9.3) years completed a 5-month follow-up. According to the binomial logistic regression analysis, previous history of falls [odds ratio (OR), 6.85; P = 0.019], OLST (OR, 5.97; P = 0.048), and FTSST (OR, 12.93; P = 0.034) were identified as risk factors for falls, and the clinical prediction rule was derived from these variables. The pretest probability of fallers in this study was 21.1% (19 of 90 participants). When the total screening tool score was three points (the participant scored one point for each item: previous history of falls, yes; OLST, ≤6.84 s; FTSST, ≥8.77 s), the positive likelihood ratio was 16:19, and the posttest probability increased to 81.3%. Therefore, this simple screening tool possesses potential clinical utility for identifying patients with knee osteoarthritis at high risk of falls in the future because it demonstrated sufficient diagnostic test accuracy.