Passive range of shoulder movement was measured using either a goniometer

or visual observation. Sensation was measured using a range of clinical assessments including light touch, proprioception, two-point and temperature discrimination. Subluxation was measured by palpation or calipers when the arm was unsupported in sitting. Shoulder pain BLU9931 was deemed present if documented in the weekly therapy reports, ward round, or case conference notes (eg, shoulder pain interfered with dressing or sleeping, therapeutic exercises, or task-related practice, or required analgesia). When possible, information about events (eg, a fall, change in mobility, or use of arm supports) preceding the onset of shoulder pain was collated. Data were summarised for the sample, and subsamples with and AZD2281 research buy without pain. Data were then analysed using Mann-Whitney (ordinal and interval data that was not normally distributed) and Chi-Square (categoric data) tests to determine how people with pain differed from those

without pain. To assist in interpreting the observed differences, odds ratios and mean group differences (with 95% CIs) for all variables were also calculated. Factors that differentiated the group with pain from those without pain were then explored in order to select predictors, and to reduce the likelihood of muticollinearity and overfitting within the multivariate model (Tabachnick and Fiddell 2001). Given the sample size, the multivariate analysis was restricted to a maximum of five predictors. Logistic regression was then conducted to explore factors associated with shoulder pain. The fit of the model was further explored by entering various combinations of predictors into the model. Level of statistical significance

was 0.05 for all analyses. The participants’ characteristics are summarised in Table 1. Of the 94 participants, 22 (23%) had shoulder pain when admitted to rehabilitation. A further 11 participants developed pain during rehabilitation, secondly leading to a total of 33 (35%) who experienced shoulder pain whilst hospitalised. Pain was reported at various frequencies for the 33 participants with pain (ie, median 33%, range 4% to 100%, of entries per participant). For the 11 participants not admitted with shoulder pain, the first report of pain was at a median of 4 (range 1 to 14) weeks after admission. Several events were noted that might have contributed to the onset of pain in these 11 participants. These included events or poor postures that may have traumatised the shoulder (eg, whilst having investigations such as radiology), altered use of arm supports, change in pattern of motor recruitment for the arm, and a fall.