Underweight Patients Are the Greatest Risk Body Mass Index Group for 30-Day Perioperative Adverse Events After Total Shoulder Arthroplasty

Taylor D. Ottesen, BS; Walter R. Hsiang, BS; Rohil Malpani, BS; Allen D. Nicholson, MD; Arya G. Varthi, MD; Lee E. Rubin, MD; Jonathan N. Grauer, MD


J Am Acad Orthop Surg. 2021;29(3):e132-e142. 

In This Article


Patient Cohort

The NSQIP database records and aggregates more than 150 variables on individual surgeries from over 500 participating institutions. These variables, which include demographics, perioperative variables, and 30-day postoperative morbidity and mortality,[20] are collected by trained reviewers from medical records, operative reports, and patient interviews.[21] Inter-rater reliability disagreement rates are reported to be less than 2%.[20]

TSA cases were extracted from the 2005 to 2016 NSQIP data sets using the Current Procedural Terminology (CPT) code 23472. Because this code refers to both anatomic and reverse TSA, these could not be separated and studied independently. Only patients undergoing elective surgery were identified based on NSQIP coding and were included in the study cohort. Revision, fracture, trauma, infection, and tumor cases were excluded using both ICD-9 and ICD-10 codes.

Patient Demographics/Characteristics

Height (meters) and weight (kilograms) were directly abstracted from the NSQIP data set to calculate BMI. Each patient BMI was calculated as kilograms (kg)/meters (m)2 directly from the height and weight variables provided in the data set. Outputs were verified for a subset of the patients to ensure correct calculation of BMI. Then, BMI was broken down into categories as defined by the World Health Organization.[22] These categories were the following: underweight (BMI <18.5 kg/m2), normal weight (BMI, 18.5 to 24.9 kg/m2), overweight (BMI, 25.0 to 29.9 kg/m2), obese (BMI, 30.0 to 39.9 kg/m2), morbidly obese (BMI, 40.0 to 49.9 kg/m2), and super morbidly obese (BMI >50.0 kg/m2).

Age, sex, functional status before surgery, American Society of Anesthesiologists (ASA) classification, smoking status (current smoker within 1 year), and presence of noninsulin-dependent or insulin-dependent diabetes were also directly extracted from NSQIP. The ASA score was used to approximate the overall health of patients because past literature has substantiated using this as a marker of overall health.[23–26]

Perioperative Adverse Events and Readmission

NSQIP captures the occurrence of individual postoperative complications through the postoperative day 30, regardless of discharge status. Adverse event occurrences were extracted from NSQIP and investigated individually and in aggregated groups of any adverse, serious adverse, and minor events.

A serious adverse event (SAE) was defined as the occurrence of any of the following: deep infection, sepsis/septic shock/systemic inflammatory response syndrome, failure to wean, unplanned reintubation, postoperative renal failure, thromboembolic event, cardiac arrest, myocardial infarction, and stroke/cerebrovascular event. A minor adverse event (MAE) was defined as the occurrence of any of the following: superficial surgical site infection, wound dehiscence, pneumonia, urinary tract infection, and postoperative renal insufficiency. Any adverse event (AAE) was defined as the occurrence of a MAE or SAE.

In addition to being in the aggregated adverse event categories, the occurrences of postoperative infections, readmission, and deaths were considered separately. Postoperative infections were noted after the occurrence of superficial infection, deep infection, urinary tract infection, or sepsis. Occurrence of death and readmission within 30 days of operation was also abstracted (noted that the readmissions data element was only available for cases that occurred in 2011 to 2016). Furthermore, operative time (minutes from incision to closure) and length of stay (days from operation to hospital discharge, with a maximum of 30-day postoperation) were collected.

Data Analysis

Patients were striated into six BMI categories based on the WHO classification. All demographic, comorbidity, and perioperative outcome variables were delineated for each BMI category.

A histogram of the frequency of BMIs in the cohort was generated. This histogram was then overlaid with the relative risk of an adverse events or binomial outcome as a function BMI. Subsequently, univariate logistic regression models of BMI groups were used to calculated the odds of given perioperative adverse events relative to each of the defined BMI categories.

Subsequently, robust multivariate logistic regressions of BMI groups were fitted on major surgical adverse events (AAEs, SAEs, MAEs, postoperative infections, readmissions, and mortality) that occurred during the 30-day postoperative period. Analyses controlled for all patient demographics including age, sex, functional status, smoking status, and health status as measured by ASA class as has been modeled in previous publications.[23,26–31]

All statistical analyses were performed using STATA version 13 (StataCorp LP, College Station, TX). Our institutional review board granted exemption for studies using the NSQIP data set.