Development and initial validation of the Northwick Park Therapy Dependency Assessment

Abstract

Objectives: To describe the development and initial validation of the Northwick Park Therapy Dependency Assessment (NPTDA) as a measure of therapy interventions in neurorehabilitation. Design: An iterative development process, followed by comparison with systemic prospective activity analysis, and parallel application of prospective and retrospective scores Setting: A tertiary specialist inpatient neurorehabilitation service Participants: A total of 37 patients (M:F 21:16, mean age 41.8 (SD 14.7) years) with complex neurological disability in two consecutive cross-sectional cohorts. Methods: The NPTDA was developed and refined over 18 months, together with an algorithm that converts ordinal scores to estimated therapy hours/week. NPTDA-estimated hours were compared with ‘actual’ therapy hours/week, identified from activity analysis. In a subsequent cohort analysis, prospectively rated NPTDA scores (reflecting intended levels of intervention) were compared with retrospective NPTDA scores (actual interventions). Results: NPTDA-estimated therapy hours/week were strongly correlated with those identified from activity analysis, for total scores (Spearman rho 0.77, P<0.0001), and also for all five subdomains for direct (hands-on) intervention (rho 0.70—0.93, P<0.0001). The initial test algorithm overestimated therapy hours (Wilcoxon z =3.9, P<0.001). After adjustment, reanalysis using a revised algorithm showed this bias to be removed (Wilcoxon z =1.4 P =0.15). Prospective and retrospectively applied total NPTDA scores were strongly correlated (rho 0.61, P<0.0001). Although intended levels of intervention were higher than those actually delivered (Wilcoxon z =3.30, P<0.001), the differences corresponded to real deviations from intended practice. Conclusion: In this initial evaluation, after revision of the algorithm, the NPTDA provided acceptable estimate of therapy interventions. Further evaluation is now required in other populations and settings.