Related Experiment Videos
A classification tree analysis of selection for discretionary treatment
J Feinglass1, P R Yarnold, W J McCarthy
1Division of General Internal Medicine, Northwestern University Medical School, Chicago, Illinois, USA.
Medical Care
|May 22, 1998
Summary
This study developed a classification tree model to identify factors influencing treatment decisions for intermittent claudication, revealing key clinical and patient-reported variables critical for observational research on treatment effectiveness.
Area of Science:
- Vascular Surgery
- Outcomes Research
- Health Services Research
Background:
- Observational studies of treatment effectiveness are susceptible to bias from unmeasured confounding factors.
- Understanding factors influencing treatment selection is crucial for accurate outcomes research.
- Intermittent claudication management involves complex clinical and patient-driven decisions.
Purpose of the Study:
- To develop a nonlinear classification tree model (CTA) to identify clinical and psychosocial factors influencing treatment selection for lower extremity bypass surgery or angioplasty.
- To assess treatment bias in observational outcomes research by modeling interventional management selection.
- To identify critical control variables for valid observational studies of treatment effectiveness in patients with intermittent claudication.
Main Methods:
- A prospective outcomes study enrolled 532 patients with mild to moderate lower extremity vascular disease.
- Classification tree analysis (CTA) was employed to model treatment selection based on baseline sociodemographic, clinical, and patient-reported health status data.
- Model stability was assessed using jackknife validity analysis, with experimentwise Type I error controlled at P < 0.05 by the Bonferroni method.
Main Results:
- The CTA model achieved an overall classification accuracy of 89.5% (67.6% sensitive, 92.9% specific).
- Ten patient attributes were identified as significant predictors of undergoing interventional procedures within six months.
- Eleven model prediction endpoints indicated a 33-fold difference in the odds of receiving lower extremity revascularization.
Conclusions:
- Initial ankle-brachial index, leg symptom status, self-reported walking distance, and prior willingness for procedures were key predictors of treatment selection.
- These identified attributes are critical control variables for ensuring the validity of observational studies on treatment effectiveness.
- The findings highlight the importance of incorporating both clinical and patient-reported factors to mitigate bias in outcomes research.