Missing Data in the National Surgical Quality Improvement Program Database: How Does It Affect the Identification of Risk Factors for Shoulder Surgery Complications?

Abstract

The main purpose of this study was to establish whether different approaches to handling missingness affect the determination of risk factors associated with 30-day postoperative major and minor complications. A secondary purpose was to determine the frequency of missingness in the National Surgical Quality Improvement Program (NSQIP) records of patients who underwent shoulder surgery. We queried the American College of Surgeons NSQIP database using Current Procedural Terminology codes to identify patients who underwent shoulder surgery from 2011 to 2016 (n= 61,963). Data on major and minor postoperative complications were extracted. We also extracted data on patient characteristics, comorbidities, American Society of Anesthesiologists classifications, and preoperative laboratory values. We calculated the percentages of missingness for each variable. Each variable was then evaluated for associations with major and minor complications by using multivariable regression and 4 methods of handling missingness (involving imputation or exclusion, depending on the completeness of the data set). For 10 variables, the method using no exclusion or imputation produced higher odds of major complications compared with imputation. For 5 variables, the method using no exclusion or imputation produced higher odds of minor complications compared with imputation. Only 6.5% of all patients had no missing data (n= 4,042), whereas 44% had <10% missingness (n= 27,165). Fewer variables were associated with both major and minor complications after shoulder surgery when patient records with missing data were excluded from analysis. Different methods of handling missingness produced different odds ratios for some variables when determining risk factors for complications after shoulder surgery. III, Case control study.