Generalisable prediction models for outcomes after lumbar spinal stenosis surgery: a model development and external...

Area of Science:

• Spine Surgery Outcomes Research
• Clinical Prediction Modeling
• Health Services Research

Background:

• Lumbar spinal stenosis is a common reason for spine surgery, yet patient outcomes vary significantly.
• Predicting surgical outcomes is crucial for informed shared decision-making between patients and surgeons.

Purpose of the Study:

• To develop and externally validate prediction models for 12-month disability and pain after lumbar spinal stenosis surgery.
• To provide tools that support shared decision-making in elective spine surgery for lumbar spinal stenosis.

Main Methods:

• A registry-based multicentre cohort study utilized data from Norwegian, Swedish, and Danish spine registries (n > 68,000).
• Models were developed using logistic regression, linear regression, and XGBoost, with internal-external cross-validation (IECV) and external validation.
• Primary outcomes included Oswestry Disability Index (ODI); secondary outcomes were Numeric Rating Scale (NRS) for back and leg pain at 12 months.

Main Results:

• Prediction models for disability demonstrated good calibration and generalizability across Scandinavian countries (Adjusted R² 0.26-0.33; C-statistics 0.75-0.78).
• Models for pain showed lower performance compared to disability models (MAE 2.2-2.6; C-statistics 0.64-0.73).
• Linear regression models achieved acceptable performance for predicting 12-month ODI, with pooled MAE of 12.4 after IECV.

Conclusions:

• Validated prediction models for lumbar spinal stenosis surgery outcomes are available and generalizable.
• These models offer a foundation for improving shared decision-making and potentially enhancing patient outcomes in clinical practice.
• Further prospective evaluation is recommended to confirm the impact of these models on clinical decision-making.