Abstract
Background Early prediction of outcome after stroke is becoming increasingly important, as most patients are discharged from hospital stroke units within several days after stroke.
Objectives The primary purposes of this study were: (1) to determine the accuracy of physical therapists' predictions at hospital stroke units regarding upper-limb (UL) function, (2) to develop a computational prediction model (CPM), and (3) to compare the accuracy of physical therapists' and the CPM's predictions. Secondary objectives were to explore the impact of timing on the accuracy of the physical therapists' and CPM's predictions and to investigate the direction of the difference between predicted and observed outcomes. Finally, this study investigated whether the accuracy of physical therapists' predictions was affected by their experience in stroke rehabilitation.
Design A prospective cohort study was conducted.
Methods Physical therapists made predictions at 2 time points—within 72 hours after stroke onset (T72h) and at discharge from the hospital stroke unit (Tdischarge)—about UL function after 6 months in 3 categories, derived from the Action Research Arm Test. At the same time, clinical variables were measured to derive a CPM. The accuracy of the physical therapists' and CPM's predictions was evaluated by calculating Spearman rank correlation coefficients (rs) between predicted and observed outcomes.
Results One hundred thirty-one patients and 20 physical therapists participated in the study. For the T72h assessment, the rs value between predicted and observed outcomes was .63 for the physical therapists' predictions and .75 for the CPM's predictions. For the Tdischarge assessment, the rs value for the physical therapists' predictions improved to .75, and the rs value for the CPM's predictions improved slightly to .76.
Limitations Physical therapists administered a test battery every 3 days, which may have enhanced the accuracy of prediction.
Conclusions The accuracy of the physical therapists' predictions at T72h was lower than that of the CPM's predictions. At Tdischarge, the physical therapists' and CPM's predictions are about equally accurate.
Footnotes
Ms Nijland, Dr van Wegen, and Dr Kwakkel provided concept/idea/research design and writing. Ms Harmeling-van der Wel provided data collection. Dr van Wegen, Ms Harmeling-van der Wel, and Dr Kwakkel provided project management. Dr Kwakkel provided fund procurement, facilities/equipment, and institutional liaisons. Dr van Wegen and Dr Kwakkel provided consultation (including review of manuscript before submission).
The authors acknowledge the EPOS Investigators: Executive committee: Professor Gert Kwakkel, PhD, principal investigator, Department of Rehabilitation Medicine, Research Institute MOVE, VU University Medical Center, Amsterdam, the Netherlands; and Barbara C. Harmeling-van der Wel, co-principal investigator, Department of Rehabilitation Medicine and Physical Therapy, Erasmus MC University Hospital, Rotterdam, the Netherlands. Steering committee and data management: Janne M. Veerbeek and Rinske H.M Nijland, MSc, Department of Rehabilitation Medicine, Research Institute MOVE, VU University Medical Centre. Revision of the manuscript: Erwin E.H. van Wegen, PhD, Department of Rehabilitation Medicine, Research Institute MOVE, VU University Medical Centre. Monitoring board: Marijke A. van der Beek, UMC Utrecht, Utrecht, the Netherlands; Wam A.M. Cornelissen, AMC Amsterdam, Amsterdam, the Netherlands; Jos A.G. Goos, Franciscus Hospital, Roosendaal, the Netherlands; Claudia S. Steeg, UMC Sint Radboud, Nijmegen, the Netherlands; Joke M. Timmermans, Leiden University Medical Center, Leiden, the Netherlands; and Richard Tichelaar, Amphia Hospital, Breda the Netherlands.
This study was part of the EPOS research project funded by the Wetenschappelijk College Fysiotherapie (WCF: no. 33368) of the Royal Dutch Society for Physical Therapy (KNGF), the Netherlands. This part of the EPOS project is co-financed by ZonMw (grant no. 89000001) as a part of the EXPLICIT-stroke program (www.explicit-stroke.nl).
- Received March 19, 2012.
- Accepted November 2, 2012.
