This permitted a comparison between two groups: participants with (n = 33) or without shoulder pain (n = 61). Several factors were observed to differ between those with or without pain (Table 1). Those with pain tended to be younger, took longer to be admitted to rehabilitation after their

stroke, and had lower Motor Assessment Scale (Carr et al 1985) scores for the arm. They also tended to have limited passive range of shoulder motion, shoulder subluxation, impaired sensation, and altered muscle tone. For this study, altered muscle tone included both hypotonia and hypertonia (Carr and Shepherd 1998). In contrast, no differences were observed for several variables including the presence of inattention, communication impairment, or area and side of stroke (Table 1). The selleck compound four predictors selected for inclusion in logistic regression were Motor Assessment Scale Upper Arm item, passive range of shoulder flexion, subluxation, and altered sensation. These were selected from the 10 variables that differentiated between people with and without pain (Table 1) for several reasons. The predictors focused on primary and secondary impairments

following the stroke rather than those relating to hospital processes (eg, days between onset and admission to rehabilitation). When two similar variables were moderately related, only one variable was selected. For instance, the Motor Assessment Scale Upper Arm item was selected over the Hand item as it was considered more relevant to the shoulder. Passive range of shoulder flexion was chosen over external rotation as it was ATM Kinase Inhibitor concentration considered easier to measure clinically given the reliance upon retrospective data. Although Nicks and colleagues (2007) suggested that less than 160 degrees shoulder flexion was a predictor for post-stroke shoulder pain, we used ≤ 150 degrees as a predictor due to the distribution

of shoulder ranges observed. Altered tone was not selected as a predictor as it related to several variables including Motor Assessment Scale scores, subluxation and shoulder range of motion. Logistic regression using the four predictors identified shoulder pain as reliably associated with two predictors: Motor Assessment Scale Upper Arm item and passive range of shoulder flexion (Box 1). These findings indicate old that the odds of experiencing shoulder pain are, on average, 14% greater for people with ≥150 degrees passive shoulder flexion relative to those with ≥ 150 degrees. The average odds of shoulder pain increase by 64% for each unit lower on the Motor Assessment Scale Upper Arm item (ie, a score of 5 has a 64% greater chance of shoulder pain than a score of 6). Based on the prediction equation, the mean odds and probabilities for experiencing shoulder pain are estimated for the range of people with stroke admitted to rehabilitation (Table 2). Regression coefficients of predictors Constant = 3.73 PROM shoulder flexion = −1.