Background Falls are frequent in Parkinson's disease (PD), but there is lack of information about predictors of injurious falls. Objectives To determine predictors of falls with injuries in people with PD; to compare circumstances and consequences of falls in single and recurrent fallers. Methods Participants (n = 225) were assessed by disease‐specific, self‐report, and balance measures, and followed‐up for 12 months with a diary to record falls, their circumstances, and injuries. Univariate and multivariate analyses were performed. Circumstances and consequences of falls presented by single and recurrent fallers were compared. Results A total of 805 falls were analyzed, 107 (13%) were falls with injuries. Multivariate logistic regression model revealed that greater PD duration and higher balance confidence were protective factors; better balance during gait, outdoor falls, and falls related to extrinsic factors were risk factors for falls with injuries, when compared to falls with no injuries. Multivariate multinomial regression model revealed that, when compared to zero fall, past falls and daily levodopa equivalent dose were predictors of falls with injuries; these predictors together with disability were predictors of falls with no injuries. Single falls (n = 27; 3%) were more common outdoors because of extrinsic factors, whereas recurrent falls (n = 778; 97%) were more common indoors because of intrinsic factors. Single falls led to more injuries than recurrent falls (P < 0.05). Conclusions Different predictors of falls with injuries were obtained when different outcomes were compared. It should be noted that falls with injuries might be influenced by fall‐related activities and environmental factors. Single and recurrent falls differed on circumstances and consequences.