Predicting Outcome in Acute Low Back Pain Using Different Models of Patient Profiling

Abstract

Prospective observational study of prognostic indicators, using data from a randomized, controlled trial of physiotherapy care of acute low back pain (ALBP) with follow-up at 6 weeks, 3 months, and 6 months. To evaluate which patient profile offers the most useful guide to long-term outcome in ALBP. The evidence used to inform prognostic decision-making is derived largely from studies where baseline data are used to predict future status. Clinicians often see patients on multiple occasions so may profile patients in a variety of ways. It is worth considering if better prognostic decisions can be made from alternative profiles. Clinical, psychological, and demographic data were collected from a sample of 54 ALBP patients. Three clinical profiles were developed from information collected at baseline, information collected at 6 weeks, and the change in status between these 2 time points. A series of regression models were used to determine the independent and relative contributions of these profiles to the prediction of chronic pain and disability. The baseline profile predicted long-term pain only. The 6-week profile predicted both long-term pain and disability. The change profile only predicted long-term disability (P < 0.01). When predicting long-term pain, after the baseline profile had been added to the model, the 6-week profile did not add significantly when forced in at the second step (P > 0.05). A similar result was obtained when the order of entry was reversed. When predicting long-term disability, after the 6-week profile was entered at the first step, the change profile was not significant when forced in at the second step. However, when the change profile was entered at the first step and the 6-week clinical profile was forced in at the second step, a significant contribution of the 6-week profile was found. The profile derived from information collected at 6 weeks provided the best guide to long-term pain and disability. The baseline profile and change in status offered less predictive value.