Complication Trends and Costs of Surgical Management in 11,086 Osteoporotic Patients Receiving Lumbar Fusion

Shane Shahrestani, MS; Xiao T. Chen, BA; Alexander M. Ballatori, BA; Andy Ton, BS; Joshua Bakhsheshian, MD; Raymond J. Hah, MD; Jeffrey C. Wang, MD; Zorica Buser, PhD


Spine. 2021;46(21):1478-1484. 

In This Article


Data Source

We utilized the National Readmission Database (NRD), a yearly database provided by the Healthcare Cost and Utilization Project (HCUP) and the agency for Healthcare Research and Quality (AHRQ), for our analysis of osteoporotic patients receiving lumbar fusion. The NRD contains >35 million patient encounters, including both primary admissions and readmissions, and discharges in 2016 and 2017 with corresponding data regarding patient demographics, complications, comorbidities, and readmissions. All data regarding patient diagnoses and procedures were queried using International Classification of Diseases, Tenth Revision (ICD-10) codes in all patient admissions and readmissions. Patients were deidentified and represented as unique patient linkages to allow for accurate patient tracking throughout the calendar year. Charges described in this study reflect total hospital charges not including professional fees and noncovered charges. Institutional Review Board approval and informed consent were not required as we used a de-identified publicly available database.

Patient Selection Criteria

All primary patient encounters involving lumbar fusion surgery were queried from the database and stratified into either an osteoporotic or non-osteoporotic group based on ICD-10 codes shown in Supplementary Table 1, Non-osteoporotic patients were excluded from analysis. Osteoporotic patients were grouped into single and multi-level fusion. Within these groups, complications were queried using ICD-10 coding and compared through multivariable statistics. Additionally, multivariable models were developed to compare complications following single and multi-level fusion groups, including readmission, inflection, length of stay (LOS), and total hospital charges and whether the anterior or posterior column was fused. Demographics, including age, patient sex, and comorbidity status, and hospital information were collected for each elective and nonelective inpatient stay and readmission was analyzed at 30 days, 90 days, and 180 days. Patients who received fusion with both autologous and nonautologous graft materials and those who received fusion of both anterior and posterior columns were excluded.

Outcomes of Interest

Outcomes of interest in this study included both medical and surgical complications encountered during readmission following surgical intervention. These included infection, postoperative pain, lumbar pathology, wound dehiscence, hardware failure, and lumbar vertebral fractures. In addition, readmission rates, inpatient LOS, and hospital charges were analyzed as additional outcomes.


All statistics included in this analysis were conducted in RStudio (Version 1.2.5001). All statistical tests were two-sided and used an α = 0.05 level of significance. Patients receiving synthetic biologic grafts, including calcium phosphates, ceramics, and other synthetic materials with similar bone biomechanics, were excluded due to considerable heterogeneity within that biologic category and the inability of ICD-10 coding to differentiate the different synthetic compositions. Autologous graft material included both iliac and local bone grafts, whereas nonautologous included allogenic cadaveric bone from a tissue bank.

Univariable Analysis. Univariable analysis used Welch two-sample t tests to compare osteoporotic patients who received autologous versus nonautologous biologic graft materials. Preliminary data showed no differences in age, sex, or Charlson Comorbidity Index (CCI)[13] at any time points within the two biologic groups, and as such, univariable rather than multivariable analysis was utilized for analysis of biologic graft materials. All CCI scores were obtained using ICD-10 coding.

Multivariable Analysis. Binomial multivariable logistic regression was used to compare osteoporotic patients who received anterior column versus posterior column fusion because preliminary data analysis showed significant differences in age and CCI between fusion groups. Dependent variables included: 30-day readmission rate, 90-day readmission rate, 180-day readmission rate, and 30-day infection rates. Gaussian-fitted generalized linear regression modeling was used to compare the continuous dependent variables, including LOS and log-transformed total hospital charge. Log-transformed total hospital charge was used because it provided better model fitting in our analyses. Independent covariates for all analyses were age, sex, CCI, and column fused (anterior vs. posterior). Wald testing was performed to evaluate the effect of the weighted distance between the estimated value and the hypothesized true value under the null hypothesis on statistical parameters within each model.