Development and validation of prognostic models designed to predict wheelchair skills at discharge from spinal cord injury rehabilitation.

OBJECTIVE

To develop and validate a statistical model to predict wheelchair skills at discharge (t(2)) from personal and lesion characteristics and wheelchair skills at the start of spinal cord injury inpatient rehabilitation (t(1)).

DESIGN

Prospective cohort study.

SETTING

Eight Dutch rehabilitation centres.

SUBJECTS

One hundred and forty-two patients with a spinal cord injury.

MAIN MEASURES

Models were developed with the performance time and ability score at t(2) as dependent variables and t(1) scores of performance time and ability score, age, gender, body mass index, level and completeness of the lesion as independent variables. The statistical models were evaluated by comparing individual estimated scores with actual measured scores.

RESULTS

The main independent variables to predict wheelchair skills at discharge were the t(1) performance time and ability score, age, gender and lesion level. The intraclass correlation coefficient between the estimated and actual ability score was 0.79 and for the performance time 0.86. However, the 95% limits of agreement and their confidence intervals were relatively wide for both ability score (-2.3 to 3.4, range 0-8) and performance time (-12.5 to 8.2, range 11-40 seconds).

CONCLUSION

The prognostic models developed in this study to predict future wheelchair skills might help planning the course of rehabilitation. The models should be used with caution in daily clinical practice, but may add useful information to clinical expertise and knowledge of the individual patient.