144 - What Factors Predict the Best Outcomes for Elderly Patients Operated for Grade 1 Degenerative Lumbar Spondylolisthesis? A Machine Learning Analysis from the Quality Outcomes Database

Saturday, June 3, 2023

12:06 PM – 12:08 PM

Location: Salons 13-15

Co-Author(s)

EY

Eunice Yang, n/a

Medical Student
Department of Neurological Surgery, Columbia University, The Och Spine Hospital at NewYork-Presbyterian, New York, New York
New York, New York, United States

Co-Author(s)

PM

Praveen V. Mummaneni, MD, MBA

Professor
University of California San Francisco
san francisco, California, United States
DC

Dean Chou, MD

Columbia University, United States
MB

Mohamad Bydon, MD

Rochester, Minnesota, United States
EB

Erica F. Bisson, MD, MPH

Utah, United States
CS

Christopher I. Shaffrey, Sr., MD

Duke University Medical Center
Durham, North Carolina, United States
Steven D. Glassman, MD

Medical Director
Norton Leatherman Spine Center
Louisville, Kentucky, United States
KF

Kevin T. Foley, MD

United States
EP

Eric A. Potts, MD

United States
MS

Mark E. Shaffrey, MD

United States
DC

Domagoj Coric, MD

Neurosurgeon
Carolina Neurosurgery and Spine Associates
Charlotte, North Carolina, United States
JK

John J. Knightly, MD

New Jersey, United States
PP

Paul Park, MD

United States
MW

Michael Y. Wang, MD

Professor
University of Miami Department of Neurosurgery
Miami, Florida, United States
KF

Kai-Ming Fu, M.D., Ph.D.

Weil Cornell Medical Center, United States
JS

Jonathan R. Slotkin, MD

United States
AA

Anthony L. Asher, MD

Carolina Neurosurgery and Spine Associates, United States
MV

Michael S. Virk, MD, PhD

United States
GM

Giorgos D. Michalopoulos, MD

Minnesota, United States
JG

Jian Guan, MD

United States
NA

Nitin Agarwal, M.D.

University of Pittsburgh Medical Center, United States
RH

Regis W. Haid, Jr., n/a

Atlanta Brain and Spine Care
Atlanta, Georgia, United States
AC

Andrew K. Chan, MD

Assistant Professor of Neurological Surgery (in Orthopaedic Surgery)
Columbia University
New York, New York, United States

Submitter(s)

EY

Eunice Yang, n/a

Medical Student
Department of Neurological Surgery, Columbia University, The Och Spine Hospital at NewYork-Presbyterian, New York, New York
New York, New York, United States

Presenting Author (Person doing the Podium Present(s)

VL

Vivian P. Le, M.P.H.

Columbia University Irving Medical Center, United States

Introduction: The prevalence of degenerative lumbar spondylolisthesis in the elderly population is increasing, yet the factors influencing surgical outcomes remain incompletely defined.
AIMS/

Objectives: Our study uses a machine learning approach to identify unique outcome clusters among elderly patients operated for grade 1 lumbar spondylolisthesis.

Methods: Data was obtained from the prospective Quality Outcomes Database registry spondylolisthesis cohort, including patients with grade 1 degenerative lumbar spondylolisthesis. We included patients with age ≥ 65. We used principal components analysis to generate a composite outcome score based on five patient-reported outcomes (PROs) collected 24 months postoperatively. Subsequently, a k-means clustering approach was used to differentiate elderly patients by composite operative outcome, with silhouette scoring used to identify the optimal number of clusters. Logistic regression was used to assess variable importance.

Results: Of 608 patients, a total of 233 patients over 65 were included with 24-month follow-up. Our k-means clustering approach stratified patients into two distinct clusters, with cluster 1 corresponding to patients with optimal outcomes (less back and leg pain, increased patient satisfaction, lower levels of disability, and higher quality of life) and cluster 2 corresponding to patients with suboptimal outcomes. The optimal-outcomes cluster exhibited significantly greater improvements for all PROs (p < 0.001) at 24 months. The optimal-outcomes cluster had 49.3% of patients achieve minimal clinically important difference (MCID) across both EQ5D and ODI as compared to 0% of patients in the suboptimal-outcomes cluster (p < 0.001). Clusters did not differ significantly by age (cluster 1: 71.3, cluster 2: 73.0, p=0.18), though patients with suboptimal outcomes did report higher baseline ODI and levels of leg pain. Notably, patients in the optimal-outcomes cluster (70.5%) were significantly more likely to have received a fusion procedure than were patients in the suboptimal cluster (51.7%) (p=0.01). In a logistic regression predictor analysis, performance of a fusion procedure was the only significant independent predictor of optimal outcomes. Patients receiving fusion were more likely to have optimal surgical outcomes (OR = 1.57; 95% CI 1.12-2.19; p=0.01).

Conclusion: Our analysis of elderly patients undergoing surgery for grade 1 degenerative lumbar spondylolisthesis reveals two distinct clusters: one with optimal and another with suboptimal outcomes. Among various patient and surgical characteristics, the addition of fusion to a procedure was associated with superior outcomes, with patients receiving a fusion having nearly 1.5 times the odds of reaching an optimal outcome. There was no evidence that age was significantly different between clusters—failing to support an age cutoff for surgery. These findings may be especially informative to surgeons choosing between decompression versus decompression and fusion approaches for elderly patients.