A "people-like-me" approach to predict individual recovery following lumbar microdiscectomy and physical therapy for lumbar radiculopathy.

Abstract

Background context: Lumbar microdiscectomy is an effective treatment for short-term pain relief and improvements in disability in patients with lumbar radiculopathy, however, many patients experience residual pain and long-term disability. The 'people like me' approach seeks to enhance personalized prognosis and treatment effectiveness, utilizing historical data from similar patients to forecast individual outcomes.

Purpose: The primary objective was to develop and test the 'people-like-me' approach for leg pain intensity and disability at 12-month follow-up after lumbar microdiscectomy and postoperative physical therapy. The secondary objective was to verify the clinical utility of the prediction tool via case vignettes.

Study design/setting: A 12-month prospective cohort study.

Patient sample: Patients (N=618, mean age: 44.7) with lumbar radiculopathy who undergo a lumbar microdiscectomy and postoperative physical therapy.

Outcome measures: Leg pain intensity (Visual Analogue Scale) and disability (Roland-Morris Disability Questionnaire) were measured at 12-months following surgery.

Methods: Predictors were selected from data collected in routine practice before and 3-months after lumbar microdiscectomy. Predictive mean matching was used to select matches. Predictions were developed using preoperative data alone or combined with 3-month postoperative data. The prediction performance was evaluated for bias (difference between predicted and actual outcomes), coverage (proportion of actual outcomes within prediction intervals), and precision (accuracy of predictions) using leave-one-out cross-validation.

Results: Overall, the 'people-like-me' approach using preoperative data showed accurate coverage and minimal average bias. However, precision based on preoperative data alone was limited. Incorporating 3-month postoperative data alongside preoperative predictors significantly enhanced prognostic precision for both leg pain and disability. Including postoperative data, leg pain prediction accuracy improved by 43% and disability by 23% compared to the sample mean. Adjusted R² values improved from 0.04 to 0.21 for leg pain, and from 0.07 to 0.34 for disability, enhancing model precision. The effectiveness of this method was highlighted in two case vignettes, illustrating its application in similar patient scenarios.

Conclusion: The 'people-like-me' approach generated an accurate prognosis of 12-months outcomes following lumbar discectomy and physical therapy. Scheduling a 3-month postoperative follow-up to evaluate the course, and refine therapy plans and expectations for patients undergoing lumbar microdiscectomy would be recommended to assist clinicians and patients in more personalized healthcare planning and expectation setting.