Development and validation of models to predict respiratory function in persons with long-term spinal cord injury.

Abstract

Multicenter, cross-sectional study. To validate previously developed respiratory function prediction models for persons with long-term spinal cord injury (SCI) and if necessary develop and validate new models. Ten SCI rehabilitation centers. Five respiratory function parameters were measured in adults with chronic, traumatic, motor complete SCI (C4-T12). First, the models published in 2012 were validated using Bland-Altman plots. Then, new models were calculated using 80% of the dataset by multiple regression analysis with the candidate predictors gender, age, height, weight, time post injury (TPI), lesion level, and smoking. In a third step, the new models were validated using the other 20% of the dataset by Bland-Altman plots. In total 613 participants were included. For persons with long-term SCI, the 2012 models were poorly predictive, especially for respiratory muscle strength (R2 = 0.4). Significant predictors for all respiratory function parameters in the new models (R2 = 0.7-0.8) were lesion level, gender and weight. Small effects on single outcome parameters were observed for TPI and age whereas smoking had no effect. For the new models the mean differences between measured and predicted values for respiratory muscle strength were 4.0 ± 36.0 cm H2O and for lung function parameters -0.5 ± 1.2 L (FVC), -0.3 ± 0.9 L (FEV1) and -0.5 ± 2.0 L/s (PEF). We did not find better models for lung function in long-term SCI but those for respiratory muscle strength showed better accuracy. The content of this publication was developed under grant from Wings for Life, grant number WFL-CH-017/14.