Purpose: The relatively large number of participants lost to follow-up (attrition) in spinal registers calls for studies that investigate the features of these individuals and their possible outcome. The aim was to explore the effect of attrition on patient-reported outcome in patients undergoing degenerative lumbar spine surgery. Three groups were studied: spinal stenosis (LSS), disc herniation (LDH) and degenerative disc disorder (DDD).
Methods: Patients who underwent surgery for degenerative lumbar spine conditions during 2008-2012 according to registration in the Swespine national register were eligible for the study. Non-respondents were registered in Swespine prior to surgery, but not at follow-up. Swespine data were merged with hospital data from seven Swedish regions (65% of the population), Statistics Sweden, the National Patient Register and the Social Insurance Agency. Baseline characteristics of non-respondents were described and compared to those of the respondents. Coefficients from regression analyses on PROM values for respondents were used to estimate the levels of PROM values for non-respondents, assuming the same effects of baseline characteristics for the two subgroups. Regression analyses were then conducted to identify variables associated with non-response. The results from the regression analyses were used to predict outcomes for patients with the characteristics of a non-respondent. Primary outcome variable in LSS and LDH was Global Assessment for leg pain, and in DDD, Global Assessment for back pain.
Results: Age, sex, educational level, smoking, living alone, being born outside the EU, previous spine surgery and unexpected events before follow-up were factors that were significantly associated with non-response. Being born inside, the EU was important in all of the studied groups (LSS: OR 0.61 p = < 0.000; LDH: OR 0.68 p = 0.001; DDD: OR 0.58 p = 0.04). For spinal stenosis patients, an unexpected event appeared particularly important (OR 3.40, p = 0.000). The predicted outcome of non-respondents was significantly worse than for respondents (LSS: 75.4% successful outcome vs. 78.7%; LDH: 53.9% vs. 58.2%; DDD: 62.7% vs. 67.5%. P-value in all groups = < 0.000).
Conclusion: Attrition in Swespine cannot be ignored, as non-respondents were predicted to have worse outcome. The effect of attrition bias should always be considered when contemplating outcome recorded in a quality register with patients lost to follow-up.
Keywords: Attrition bias; Non-response to follow-up; Quality register; Spine surgery.