Predictors of Nonelective Surgery for Spinal Metastases

Insights From a National Database

Hammad A. Khan, BS; Nicholas M. Rabah, BS; Vikram Chakravarthy, MD; Raghav Tripathi, MPH; Ajit A. Krishnaney, MD


Spine. 2021;46(24):E1334-E1342. 

In This Article


Data Source and Patient Selection

The present study retrospectively analyzed inpatient discharges from the Healthcare Cost and Utilization Project (HCUP) National Inpatient Sample (NIS) between January 1st, 2012 and September 30th, 2015.[16] The NIS is a stratified and survey-weighted sample of all discharges from non-federal, short-term community hospitals in the United States. The nation's largest all-payer inpatient database, it consists of records from >7 million hospitalization per year and an estimated 35 million hospitalizations per year once survey weights are applied. Using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) system, NIS captures up to 30 diagnoses and 15 procedures associated with each hospital encounter. Patients were included if their discharge record contained a diagnosis code for secondary malignant neoplasm of the spinal cord/brain, meninges, or bone (198.3, 198.4, 198.5) and procedure code for spine surgery (03.0, 03.4, 03.09, 81.0, 81.00–81.08) within the first three coded diagnoses and procedures, as has been performed previously.[17–19] Patients were then subclassified based on admission status, which is coded in the NIS as "elective" and "nonelective."

Data Collection

The primary objective of this study was to determine patient- and hospital-level factors associated with nonelective versus elective surgery for spinal metastases. Sociodemographic characteristics compared between groups included sex, age, race, income level, insurance status, residency status, and weekday admission. The presence of myelopathy was identified using the ICD-9-CM codes 336.3, 336.8, and 336.9. The All Patient Refined Diagnosis Related Group (APR-DRG) Severity Subclass was used as a surrogate for disease severity. Primary cancer site was identified using the NIS clinical classification system. For the purposes of multivariable analysis, primary cancer was reclassified as tumors with slow growth (breast, prostate), moderate growth (kidney), fast growth (lung, colon), and other/unknown, in accordance with the Tomita scoring system for spinal metastases.[20] Hospital characteristics analyzed included hospital bed size, type, and region. Comorbidity status was assessed using the Elixhauser comorbidity measure, which has been validated for use in administrative datasets.[21] Select Elixhauser comorbidities were also individually compared between groups.

Perioperative Outcomes

The secondary objective of this study was to compare perioperative outcomes between patients receiving elective versus nonelective surgery for spinal metastases. These included in-hospital complications, in-hospital mortality, discharge disposition (routine vs. nonroutine), LOS, and total hospital costs. Nonroutine discharge was defined as discharge home with health care services, short-term inpatient hospital, skilled nursing facility, intermediate care facility, or against medical advice. In-hospital complications were identified using ICD-9-CM diagnosis codes. LOS was classified as normal or extended (LOS >75th percentile for cohort, 11.37 days). All-payer cost-to-charge ratios were used to compute total hospital costs from total hospital charges, and costs were log-transformed to account for the inherent right-skewedness of this data. Total hospital costs were classified as normal or high-cost (cost >75th percentile for cohort, $56,680).

Statistical Analysis

Statistical analysis was conducted in SAS Version 9.4 (SAS Institute, Cary, NC) using proc survey procedures to account for NIS strata, clusters, and weighting. Weighted descriptive statistics summarizing sociodemographic characteristics, comorbidities, and hospital characteristics for patients receiving elective and nonelective surgery were generated. Weighted frequencies were compared between groups using the Rao-Scott χ 2 test. Cost and LOS were compared between groups using the Mann–Whitney U test and reported as medians with interquartile range (IQR), as these variables were not normally distributed. For our primary objective, a multivariable logistic regression model was constructed to identify factors independently associated with receiving nonelective surgery for spinal metastases, after adjusting for baseline covariates (sex, age, race, income, insurance status, Elixhauser comorbidity score, myelopathy, disease severity, primary cancer, patient residence, weekend admission, hospital bed size, hospital type, and hospital region). Additional multivariable logistic regression models were constructed to evaluate the association between nonelective surgery and the presence of any perioperative complication, in-hospital mortality, nonrou-tine discharge, extended LOS, and high cost. Statistical significance was defined at the P < 0.05 level.