Skip to main content
  • Development of a Validated Computer Based Preoperative Predictive Model for Reaching ODI MCID for Adult Spinal Deformity (ASD) Patients

    Final Number:

    Justin K Scheer BS; Justin S. Smith MD, PhD; Frank Schwab MD, PhD; Robert Hart MD; Richard A. Hostin MD; Virginie Lafage PhD; Amit Jain BS; Douglas C. Burton MD; Shay Bess MD; Tamir T. Ailon MD, MPH; Themistocles Protopsaltis MD; Eric Klineberg MD; Christopher I. Shaffrey MD, FACS; Christopher P. Ames MD; International Spine Study Group

    Study Design:

    Subject Category:

    Meeting: Congress of Neurological Surgeons 2015 Annual Meeting

    Introduction: Surgical correction of ASD results in significant improvement in patients' disability as measured by ODI with the goal of reaching at least 1 MCID. However, it remains unknown what are the specific predictors of reaching MCID. This study attempts to develop a preoperative predictive model to identify patients likely to reach ODI MCID.

    Methods: Inclusion criteria: age =18, ASD, baseline ODI =30. 43 variables were included in the initial training of the model and included demographic data, comorbidities, modifiable surgical variables, baseline HRQOL, and coronal and sagittal radiographic parameters. Patients were grouped by reaching at least 1 ODI MCID or not at 2-year follow- up. An ensemble of decision trees was constructed using the C5.0 algorithm with 5 different bootstrapped models. Internal validation was accomplished via a 70:30 data split for training and testing each model, respectively. Final predictions from the models were chosen by voting with random selection for tied votes. Overall accuracy, and the area under a receiver operator characteristic curve (AUC) were calculated.

    Results: 198 patients were included, (MCID:109, NotMCID:89). The overall model accuracy was 86.0% correct with an AUC of 0.94 indicating a very good model fit. The top 11 predictors (importance =0.90, Figure) of reaching MCID, in decreasing order of importance were: gender, SRS activity, back pain, SVA, PI-LL, primary vs revision, T1SPI, ASA grade, T1PA, SRS pain, SRS total.

    Conclusions: A successful model (86% accuracy, 0.94 AUC) was built predicting reaching ODI MCID. Most of the important predictors were not modifiable surgically indicating that the baseline clinical and radiographic status of the patient is a critical factor for reaching ODI MCID. This model can set the groundwork for preop point of care decision making and improved patient counseling regarding expected outcomes of ASD surgery.

    Patient Care: This research study will improve patient care by aiding surgeons with preoperative decision making and patient counseling.

    Learning Objectives: By the conclusion of this session, participants should be able to: 1) comprehend the accuracy of the predictive model which uses 43 baseline demographic, radiographic and surgical variables, 2) recognize the top 11 predictors of MCID outcome as provided, and 3) apply the concept of this model to predict reaching ODI MCID.


We use cookies to improve the performance of our site, to analyze the traffic to our site, and to personalize your experience of the site. You can control cookies through your browser settings. Please find more information on the cookies used on our site. Privacy Policy