At 4, 6, 8 and 12 months after discharge from rehabilitation 15 (11%), BLZ945 order 15 (11%), 20 (15%) and 25 (19%) of participants, respectively, were non-users. As the number of prosthetic non-users and variables were identical for

4 and 6 months, these data were analysed as one time frame. Of the 40 potential variables investigated for the univariate analysis (Box 1), a total of 16 variables were identified as being significant (p < 0.10) for prosthetic non-use at the 4-, 6- and 8-month timeframes, and 15 variables were significant at 12 months after discharge (Table 4, which is available in the eAddenda). The predictor variables significant (95% CI) for prosthetic non-use after being entered into the backwards-stepwise logistic regression model are reported below. Full details, including associated accuracy statistics, are presented in Table 5. At 4 (and 6) months, the five variables that were predictive of prosthetic learn more non-use included: amputation level above transtibial level, mobility aid use, dependence walking outdoors on concrete, very high number of comorbidities, and not having a diagnosis of type II diabetes. At 8 months, the three variables that were predictive of prosthetic non-use included: amputation level above transtibial level, mobility aid use, and dependence walking outdoors on concrete. At 12 months, the three variables that were predictive of prosthetic non-use included: amputation

level above transtibial level, mobility aid use, and delay to prosthesis. The multifactorial causes of delay to prosthesis included: wound complications (n = 8), comorbidities (n = 3), orthopaedic injuries (n = 2) and deconditioning (n = 1). From March 2011 until December 2012, 66 participants were interviewed, of whom 55 remained prosthetic users. There were eight non-users at 4 and 6 months after discharge from rehabilitation, which increased to ten at 8 months and eleven at 12 months. Similar to the retrospective cohort, prosthetic non-users and variables were identical for the 4-month and 6-month timeframes in the prospective cohort. second Survival curves (Figure 2) demonstrated a high level of concordance between

the retrospective and prospective cohorts. From discharge there was rapid progression to prosthetic non-use, followed by linear decline after 1 month. Associated accuracy statistics for having a combination of prosthetic non-use predictors (95% CI) for the clinical prediction rules time frames in the prospective cohort are reported below. Full details, including associated accuracy statistics, are presented in Table 6. If four out of five predictors were present (LR+ = 43.9, 95% CI 2.73 to 999+), the probability of non-use increased from 12 to 86% (p < 0.001). If all three predictors were present (LR+ = 33.9, 95% CI 2.1 to 999+), the probability of non-use increased from 15 to 86% (p < 0.001). If two out of three predictors were present (LR+ = 2.8, 95% CI 0.9 to 6.