BACKGROUND: inpatient falls are an important safety challenge, with around half causing physical injuries that compromise the recovery of older, frailer patients. Falls risk scores are in widespread use, but validation studies of their predictive values are few. OBJECTIVES: to assess the predictive values of the Morse falls score (MFS) in an acute general hospital. METHODS: age, admitting speciality, MFS, and any falls in the subsequent 7 days were collected in April 2011 through case note review and incident reporting systems. RESULTS: a total of 467 inpatients were included in the study; 51% were aged 75+ years; 56% had an MFS ≥25; 23% had an MFS ≥55; 28 fell. An MFS ≥25 was not significantly better than chance in the total sample or in any subgroups considered (YI: -0.01 to 0.15). An MFS ≥55 was significantly better than chance for the total sample (YI: 0.39), patients ≥75 years (YI: 0.31) and geriatrician-led wards (YI 0.37), although either sensitivity or specificity fell below 70% in each of these groups. Other subgroups did not demonstrate significantly better accuracy than chance, but may have been affected by type II error. CONCLUSIONS: using MFS ≥25 cannot be clinically justified, while using MFS ≥55 would be contingent on an effective intervention that was ethically acceptable to withhold from the patients with an MFS < 55, despite >40% of falls occurring in that group. Given similar limitations of alternative falls risk scores, hospitals should consider directly assessing and acting on individual patients' specific modifiable risk factors for falls.