Using few and scattered time points for analysis of a variable course of pain can be misleading: an example using weekly text message data

Spine J. 2014 Aug 1;14(8):1454-9. doi: 10.1016/j.spinee.2013.08.035. Epub 2013 Nov 5.


Background context: Because low back pain (LBP) is a fluctuating condition, the diversity in the prediction literature may be due to when the outcome is measured.

Purpose: The objective of this study was to investigate the prediction of LBP using an outcome measured at several time points.

Study design/setting: A multicenter clinical observational study in Sweden.

Patient sample: Data were collected on 244 subjects with nonspecific LBP. The mean age of the subjects was 44 years, the mean pain score at inclusion was 4.4/10, and 51% of the sample had experienced LBP for more than 30 days the previous year.

Outcome measures: The outcome used in this study was the "number of days with bothersome pain" collected with weekly text messages for 6 months.

Methods: In subjects with nonspecific LBP, weekly data were available for secondary analyses. A few baseline variables were chosen to investigate prediction at different time points: pain intensity, the presence of leg pain, duration of LBP the previous year, and self-rated health at baseline. Age and gender acted as additional covariates.

Results: In the multilevel models, the predictive variables interacted with time. Thus, the risk of experiencing a day with bothersome LBP varied over time. In the logistic regression analyses, the predictive variable's previous duration showed a consistent predictive ability for all the time points. However, the variables pain intensity, leg pain, and self-rated health showed inconsistent predictive patterns.

Conclusions: An outcome based on frequently measured data described the variability in the prediction of future LBP over time. Prediction depended on when the outcome was measured. These results may explain the diversity of the results of the predictor studies in the literature.

Keywords: Bothersomeness; Frequent data collection; Low back pain; Prediction; Retrospective Study; Text messages.

Publication types

  • Observational Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adolescent
  • Adult
  • Aged
  • Data Collection / methods*
  • Female
  • Humans
  • Leg
  • Logistic Models
  • Low Back Pain / diagnosis*
  • Male
  • Middle Aged
  • Multilevel Analysis
  • Outcome Assessment, Health Care
  • Pain / diagnosis
  • Pain Measurement / methods*
  • Research Design
  • Sweden
  • Text Messaging*
  • Time Factors
  • Young Adult