Adding 3-month patient data improves prognostic models of 12-month disability, pain, and satisfaction after specific lumbar spine surgical procedures: development and validation of a prediction model

Background context: Prognostic models including early postoperative variables may provide optimal estimates of long-term outcomes and help direct postoperative care.

Purpose: To develop and validate prognostic models for 12-month disability, back pain, leg pain, and satisfaction among patients undergoing microdiscectomy, laminectomy, and laminectomy with fusion for degenerative lumbar conditions.

Study design/setting: Retrospective cohort study using the Quality Outcomes Database.

Patient sample: Patients receiving elective lumbar spine surgery due to degenerative spine conditions.

Outcome measures: Oswestry Disability Index, pain numerical rating scale, and NASS Patient Satisfaction Index.

Methods: Prognostic models were developed using proportional odds ordinal logistic regression using patient characteristics and baseline and 3-month patient-reported outcome scores. Models were fit for each outcome stratified by type of surgical procedure. Adjusted odds ratio and 95% confidence intervals were reported for all predictors by procedure. Models were internally validated using bootstrap resampling. Discrimination was reported as the c-index and calibration was presented using the calibration slope. We compared the performance of models with and without 3-month patient-reported variables. This research was supported by the Foundation for Physical Therapy's Center of Excellence in Physical Therapy Health Services, and Health Policy Research and Training grant.

Results: The sample consisted of 5,840 patients receiving a microdiscectomy (n=2,085), laminectomy (n=1,837), or laminectomy with fusion (n=1,918). The 3-month Oswestry score was the strongest and most consistent predictor associated with 12-month outcomes. All prognostic models performed well with overfitting-corrected c-index values ranging from 0.718 to 0.795 and all optimism corrected calibration slopes over 0.92. The increase in c-index values ranged from 0.09 to 0.21 when adding 3-month patient-reported outcome scores.

Conclusions: Models had good discrimination and were well calibrated for estimating 12-month disability, back pain, leg pain, and satisfaction. Patient-reported outcomes at 3 months after surgery, especially 3-month Oswestry scores, improved the 12-month performance of all prognostic models beyond using only baseline variables.