The Impact of Obesity on Risk Factors for Adverse Outcomes in Patients Undergoing Elective Posterior Lumbar Spine Fusion

Deeptee Jain, MD; Wesley Durand, BS; Jeremy D. Shaw, MD; Shane Burch, MD; Vedat Deviren, MD; Sigurd Berven, MD

Disclosures

Spine. 2021;46(7):457-463. 

In This Article

Methods

This was a retrospective case–control study of patients undergoing elective posterior LSF. There were no sources of funding for this study.

Data Sample

Administrative claims from the State Inpatient Databases (SIDs) from 2006 to 2010 of New York, North Carolina, and Utah, and Nebraska from 2006 to 2011 of California, and from 2006 to 2014 of Florida were utilized. The SIDs are part of the Healthcare Cost and Utilization Project under the Agency for Healthcare Research and quality. They contain data on all inpatient hospital discharges using the International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM). SID uses unique patient identifiers that allows for longitudinal follow-up.

Inclusion/Exclusion Criteria

All patients were identified ICD-9 codes. Patients age >18 years who underwent primary posterior lumbar fusion (ICD-9 codes 81.07 and 81.08) were included. Before applying exclusion criteria, there were 316,070 patients identified. Patients undergoing revision surgery (ICD-9 codes 81.34–81.38) were excluded. Patients with ICD-9 diagnosis codes for congenital deformity, bone cancer/metastases, infection, and trauma were excluded (Table S1, http://links.lww.com/BRS/B671). To ensure a minimum 90-day follow-up, patients who underwent surgery the last quarter of the dataset were excluded. Patients with missing data for any outcome or independent variables were excluded.

Patient Characteristics and Outcomes

Data were queried for demographic data including age, sex, insurance status, and ethnic background, as well as surgical characteristics, including anterior fusion and levels fused. Patients were identified as obese using codes for obesity/morbid obesity or body mass index (BMI) >30 (V85.30-.45, V85.3, V85.4, 278.0, 278.00, 278.01).

Data were also queried for 17 medical comorbidities as previously described,[13] as well as mental health disease (codes 290–316). The other variables outside of the 17 cited medical comorbidities were selected based on a literature review of factors shown to be associated with the outcomes of interest after spine surgery. Smoking was specifically not included, as there has been prior literature demonstrating a paradoxical relationship between smoking and adverse outcomes specifically in administrative claims databases.[14]

Data were then queried for 90-day readmissions, as well as infection, reoperation (defined as any surgery on the lumbar spine), and major medical complications (Table S2, http://links.lww.com/BRS/B672). Major medical complications included pulmonary embolism, deep vein thrombosis, acute myocardial infarction, respiratory failure, and cerebrovascular accident.

As per SID guidelines, variables with absolute counts <10 were not included (any malignancy except skin, metastatic solid tumor, and moderate to severe liver disease).

Statistical Analysis

Descriptive statistics were generated. Bivariate analysis was performed to compare outcomes between obese and non-obese patients. All variables were included in the models. First, although not the main purpose of this study, multivariable logistic analyses including all variables were performed on the entire cohort for all outcome variables. Next multivariable logistic regression models were utilized to account for confounding variables. Interactions between obesity and each other independent variable were tested in separate multivariable models for each outcome. Odds ratios for the effect of each variable were calculated for obese and nonobese patients, and the corresponding P value comparing these odds ratios was generated. Given the large number of statistical tests performed, statistical significance was defined by controlling the false discovery rate with the Benjamini-Hochberg procedure.[15] Statistical analysis was conducted with SAS 9.4 (SAS Institute, Cary, NC) and Julia 1.1.1 (JuliaLang).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....