BackgroundFalls are common in hospitalized patients, especially in older adults. Currently, risk assessment tools lack specificity and sensitivity to be clinically useful. Recently it was discovered free text in nursing notes contains valuable information on fall risk factors. We used text mining techniques to search for any text characteristics (not only related to known risk factors) associated with falls. MethodsIn this retrospective case-control pilot study, hospitalized patients aged ≥70 years who experienced a fall were included using incident reports. Controls were matched for sex, age and length of stay. Data were collected from free text in nursing notes 72 h prior to the fall and for similar hours in controls. Number of words, frequencies of single words and word combinations were calculated and compared between both groups. Results19 fallers and 19 non-fallers were included, with a total of 362 nursing notes. More words were used in nursing notes in fallers in total (10,523 vs 7,510, p < 0.01) and per nursing note (median 47 vs 34.5, p < 0.01). More unique words were used in fallers (2,465 vs 1,887, p < 0.01). 21 words were associated with falling, including words describing fall prevention and delirium. 8 words and 6 combinations of words were associated with not falling. ConclusionsText mining in nursing notes can help to find words used more frequently in patients who experienced a fall and is thus a promising method to identify older adults at high risk for a fall.