Preoperative Risk Index Among Patients Undergoing Thyroid or Parathyroid Surgery

Key Points

Question  Which frailty-related factors are most associated with short-term postoperative adverse events in patients undergoing thyroid or parathyroid surgery?

Findings  Of the 154 895 patients analyzed in this cohort study, 2.1% had major postoperative adverse events within 30 days of surgery; older age, male sex, current smoking, dyspnea, anemia, leukocytosis, hypoalbuminemia, functional dependence, recent weight loss, obesity, anticoagulation use, inpatient status, and length of surgery were independently associated with major adverse events or death. The Cervical Endocrine Surgery Risk Index outperformed other models to determine major adverse events.

Meaning  The Cervical Endocrine Surgery Risk Index provides an acceptable measure of postoperative risk that surgeons may use to counsel patients awaiting thyroid or parathyroid surgery.

Importance  Frailty represents a multidimensional syndrome that is increasingly being used to stratify risk in surgical patients. Current frailty risk models are limited among those undergoing thyroid or parathyroid surgery.

Objective  To develop and compare preoperative risk indices to determine factors associated with short-term major postoperative adverse events in patients undergoing thyroid or parathyroid surgery.

Design, Setting, and Participants  This cohort study evaluated 154 895 patients in the American College of Surgeons National Surgical Quality Improvement Program who underwent thyroid or parathyroid surgery from January 1, 2007, to December 31, 2016.

Exposures  Preoperative frailty-related and surgical factors from a derivation cohort were evaluated using simple and multiple logistic regression. Variables potentially associated with postoperative adverse events were subsequently combined into a personalized preoperative Cervical Endocrine Surgery Risk Index (CESRI) and compared with existing risk models using the validation cohort.

Main Outcomes and Measures  Composite variable of any major postoperative adverse event, including death, within 30 days of surgery.

Results  Of the 154 895 operations reviewed, 3318 patients (2.1%; 2296 women and 1022 men; mean [SD] age, 56.1 [15.6] years) experienced a major postoperative adverse event, with 163 deaths (0.1%). Older age (age, ≥80 years: odds ratio [OR], 2.35; 95% CI, 1.74-3.13), inpatient status (OR, 3.55; 95% CI, 3.08-4.11), male sex (OR, 1.49; 95% CI, 1.29-1.71), current tobacco smoking (OR, 1.25; 95% CI, 1.05-1.48), dyspnea (OR, 1.58; 95% CI, 1.29-1.91), recent weight loss (OR, 1.88; 95% CI, 1.23-2.78), functional dependence (OR, 2.77; 95% CI, 2.05-3.69), obesity (OR, 1.33; 95% CI, 1.10-1.60), anemia (OR, 2.14; 95% CI, 1.82-2.52), leukocytosis (OR, 1.73; 95% CI, 1.38-2.14), hypoalbuminemia (OR, 1.87; 95% CI, 1.56-2.23), use of anticoagulation (OR, 2.16; 95% CI, 1.64-2.81), and length of surgery (>4 hours: OR, 2.92; 95% CI, 2.37-3.59) were independently associated with major adverse events or death on multiple regression analysis (C statistic, 0.77; 95% CI, 0.76-0.78). The area under the curve of the CESRI to determine major adverse events, including death, using the validation cohort was 0.63 (95% CI, 0.61-0.64), with a sensitivity of 0.66 (95% CI, 0.64-0.68) and specificity of 0.66 (95% CI, 0.65-0.66). The CESRI outperformed other risk models for determining adverse events (CESRI vs 5-Factor Modified Frailty Index: delta C index, 0.11; 95% CI, 0.09-0.13; CESRI vs American Society of Anesthesiologists Physical Status Classification System: delta C index, 0.05; 95% CI, 0.03-0.07; CESRI vs American College of Surgeons Risk Calculator: delta C index, 0.02; 95% CI, 0.01-0.03; and CESRI vs Head and Neck Surgery Risk Index: delta C index, 0.04; 95% CI, 0.03-0.06).

Conclusions and Relevance  This study suggests that the CESRI is able to determine major postoperative adverse events in patients undergoing thyroid or parathyroid surgery.

As the population ages, so does the number of frail elderly individuals.¹ The prevalence of frailty among seniors varies greatly from 4% to 59%.^1,2 This complex syndrome is marked by a reduction of physiological reserve, leading to vulnerability to stressors and disproportionate response to even minor procedures.^3-5 More important, frailty encompasses a multidimensional syndrome spanning psychosocial, cognitive, physical, and functional domains.^6,7 Growing interest exists in quantifying preoperative risk using frailty as the main indicator.^8,9 Various authors have developed risk indices to quantify cumulated physical deficits in patients undergoing elective surgery.^8-10 One widely accepted model is the Modified Frailty Index (mFI) developed using the National Surgical Quality Improvement Program (NSQIP) from the American College of Surgeons (ACS).⁹ Current evidence supports the use of well-validated frailty assessment models when evaluating risk for adverse postoperative outcomes in general and cardiovascular surgery.^4,8-10 Identifying frail surgical patients may help surgeons and perioperative medicine specialists guide preoperative patient counseling while targeting those in need of supplementary postoperative care.

Frailty in patients with head and neck cancer has been linked to a significant increase in postoperative morbidity and mortality.^11-14 Despite evidence from the mFI on patients undergoing thyroid or parathyroid surgery, the literature is less clear in this surgical cohort.^15,16 This study aims to determine which frailty-related factors are associated with short-term major postoperative adverse events and to develop a preoperative risk index for patients undergoing thyroid or parathyroid surgery.

The ACS NSQIP is an outcomes-based program designed to measure and improve the quality of surgical care. Preoperative patient factors as well as 30-day postoperative adverse events are accurately reported.¹⁷ In this study, a retrospective cohort analysis of patients undergoing thyroid or parathyroid surgery registered in the ACS NSQIP from January 1, 2007, to December 31, 2016, was performed. Current Procedural Terminology codes were used to identify patients (codes 60210, 60212, 60220, 60225, 60240, 60252, 60254, 60260, 60271, 60500, and 60502). Those undergoing emergency surgery, not receiving general anesthesia, and pediatric patients were excluded. Ethics approval was granted from the McGill University Health Centre research board (MP-37-2018-3568). Patient consent was waived as data from the NSQIP public user files were anonymous according to the ACS.

Only preoperative variables retained on the most current NSQIP public user files were included to reduce the amount of missing data. Preoperative variables were categorized into sociodemographic, clinical, and frailty-related risk factors. Sociodemographic data consisted of age, sex, and smoking status (smoking within 1 year of surgery). Clinical factors included procedure type, comorbid status (metabolic, cardiovascular, respiratory, renal, hepatic, or hematologic), and length of surgery. The American Society of Anesthesiologists Physical Status Classification System and admission status (inpatient or outpatient) were also recorded. Frailty-related factors consisted of functional status (partial or total dependence for activities of daily living), dyspnea, recent weight loss, and body mass index. In addition, relevant preoperative laboratory values (complete blood count, creatinine, and albumin) were analyzed. Continuous variables were categorized using clinically relevant cutoff values to facilitate model development.

The main outcome was a composite variable of any major postoperative adverse event or death within 30 days of surgery. Major postoperative adverse events in the ACS NSQIP were defined by several authors as deep neck space infection, pulmonary embolism, acute kidney injury, cerebrovascular accident, coma, myocardial infarction, cardiac arrest, sepsis, septic shock, failure to wean off ventilator, reintubation, blood transfusions, return to the operating room, and death^13,16,18,19; these adverse events were coded as binary variables. The secondary outcome was the ability of a risk model to determine any postoperative morbidity and mortality.

A total of 70% of patients meeting the study criteria were randomly assigned to the derivation cohort, while the remaining 30% were assigned to the validation cohort. Descriptive statistics were used to compare preoperative risk factors among patients with and without major postoperative adverse events within the derivation cohort. Mann-Whitney tests were used for continuous data and χ² tests were used for categorial data. Variables were then compared with major postoperative adverse events using logistic regression within the derivation cohort. Factors associated with adverse events were retained from univariate analysis if the lower bounds of the 95% CI was above 0.8, the upper bounds were below 1.2, or deemed clinically significant based on the current literature. The final model was derived using the lowest Bayesian Information Criterion and predictive value using the C statistic. Sensitivity analysis was used to compare type of surgery and postoperative diagnosis. The β coefficients from the final multiple logistic model were then used to derive the risk index based on a 10-fold multiple rounded to the nearest integer. Internal validity was tested using the validation cohort. The fit and accuracy of models was performed using the Hosmer-Lemeshow test and calibration curves from both derivation and validation cohorts. Subsequently, the developed risk index was compared with other preoperative risk models, including the ACS Risk Calculator, 5-Factor mFI (mFI-5), Head and Neck Surgery Risk Index, and American Society of Anesthesiologists Physical Status Classification using receiver operator characteristic curves. Data analysis was performed using R software, version 3.3 (R Foundation for Statistical Computing) and using libraries DescTools, survminer, survival, MASS, survAUC, Hmisc, generalhoslem, and pROC. All P values were from 2-sided tests, and results were deemed statistically significant at P < .05.

A total of 157 445 patients undergoing thyroid or parathyroid surgery were included in the study using Current Procedural Terminology codes. After excluding 348 patients undergoing emergency surgery and 2202 patients not undergoing general anesthesia, 154 895 patients were analyzed. A total of 116 214 patients (75.0%) underwent thyroid surgery (lobectomy, total thyroidectomy, or completion thyroidectomy) and 38 681 patients (25.0%) underwent parathyroid surgery (parathyroidectomy and/or 4-gland exploration). Overall, 3318 major postoperative adverse events occurred (2.1%), including 163 deaths (0.1%). The derivation cohort consisted of 108 489 patients; Table 1 describes the patient characteristics. All postoperative adverse events are in eTable 1 in the Supplement.

Using the derivation cohort, patients who experienced major postoperative adverse events tended to be older, male, and current smokers, with cardiorespiratory and renal comorbidities (eTable 2 in the Supplement). Patients scheduled for inpatient surgery were also more at risk of the primary outcome. In the multiple logistic regression analysis, the combination of sociodemographic, frailty, and clinical risk factors provided the best model, with a C statistic of 0.77 (95% CI, 0.76-0.78). Interactions among variables were evaluated; however, no combination of variables with a meaningful prevalence was identified. The most significant potentially reversible risk factors based on population attributable fractions were anemia (odds ratio [OR], 2.14; 95% CI, 1.82-2.52), current smoking (OR, 1.25; 95% CI, 1.05-1.48), obesity (OR, 1.33; 95% CI, 1.10-1.60), dyspnea (OR, 1.58; 95% CI, 1.29-1.91), and hypoalbuminemia (OR, 1.87; 95% CI, 1.56-2.23) (Table 2). Other factors independently associated with major adverse events or death included older age (age, ≥80 years: OR, 2.35; 95% CI, 1.74-3.13), inpatient status (OR, 3.55; 95% CI, 3.08-4.11), male sex (OR, 1.49; 95% CI, 1.29-1.71), recent weight loss (OR, 1.88; 95% CI, 1.23-2.78), functional dependence (OR, 2.77; 95% CI, 2.05-3.69), leukocytosis (OR, 1.73; 95% CI, 1.38-2.14), use of anticoagulation (OR, 2.16; 95% CI, 1.64-2.81), and length of surgery (>4 hours: OR, 2.92; 95% CI, 2.37-3.59). The Cervical Endocrine Surgery Risk Index (CESRI) was developed using β coefficients from the final regression model (Table 3 and Table 4). The sensitivity analysis did not show a significant difference in point estimates of coefficients in patients undergoing different types of surgery or with different postoperative diagnoses. The Hosmer-Lemeshow goodness of fit test showed no significant difference among cohorts.

The CESRI had the highest predictive index compared with other models with similar data (CESRI vs mFI-5: delta C index, 0.11; 95% CI, 0.09-0.13; CESRI vs American Society of Anesthesiologists Physical Status Classification System: delta C index, 0.05; 95% CI, 0.03-0.07; CESRI vs ACS Risk Calculator: delta C index, 0.02; 95% CI, 0.01-0.03; and CESRI vs Head and Neck Surgery Risk Index: delta C index, 0.04; 95% CI, 0.03-0.06) (Table 5). Overall, the CESRI provided the highest determination of morbidity within 30 days of surgery. The ACS Risk Calculator outperformed all other models for determining mortality. The area under the curve of the CESRI to determine major adverse events, including death, using the validation cohort was 0.63 (95% CI, 0.61-0.64). The sensitivity of the CESRI using the validation cohort to determine any major postoperative adverse event was 0.66 (95% CI, 0.64-0.68) and the specificity was 0.66 (95% CI, 0.65-0.66). This risk index determined death within 30 days of surgery with a sensitivity of 0.75 (95% CI, 0.65-0.83) and a specificity of 0.78 (95% CI, 0.78-0.79).

Major postoperative adverse events in thyroid or parathyroid surgery are uncommon, with 2.1% of patients experiencing a serious complication within 30 days of surgery. Compared with other risk models, the CESRI contains 14 parameters and is based on a multidimensional framework including sociodemographic, frailty, and surgical risk factors. This model provided an acceptable determination of postoperative outcomes, including death. All evaluated risk models in this study provided only satisfactory determination of morbidity. In addition, not all available preoperative risk indices could have been compared with the CESRI, such as the Physiology and Operative Severity Score for the enumeration of Mortality and Morbidity, Estimation of Physiologic Ability and Surgical Stress, and Charlson Comorbidity Index, given the NSQIP data set.

Frailty is increasingly being used as a risk factor for poor outcomes in patients undergoing surgery.^20-23 Beyond simply combining comorbidities, frailty encompasses psychosocial, physical, functional, emotional, and sarcopenic domains and allows for a more extensive preoperative assessment of patients.^6,7 Similar to the 11-Factor mFI, the mFI-5 provides a simple method to stratify risk; however, it is composed mostly of comorbidities, which limits the tool’s ability to be multifaceted.⁹ The CESRI was designed to include several domains of frailty, including sarcopenia, physical, and comorbid function in addition to surgical and sociodemographic factors to provide risk quantification, as accurate determination of risk can help surgeons with preoperative counseling of patients as well as potentially alter perioperative care.

The ability to determine outcomes can identify potentially modifiable factors that may be addressed in the perioperative period to mitigate risk. Within this surgical cohort, the population attributable fraction for smoking, dyspnea, anemia, hypoalbuminemia, and obesity amounted to 20% should these factors be causal. Previous studies have reported improvements in postoperative outcomes with prehabilitation programs in patients undergoing colorectal surgery.^24-26 Similarly, preoperative nutritional supplementation has shown benefit in patients undergoing ablative head and neck surgery.²⁷ It may be conceivable that modification of these identified factors may reduce short-term postoperative risk if it is initiated prior to surgery.

Study limitations include the design as a cohort analysis of surgical patients, as those who did not undergo surgery were not considered. Also, tumor-related factors, such as the TNM stage for thyroid cancers, was not recorded. Inpatient status was based on surgeon or preoperative medical consultation discretion. Clustering of cases based on high-volume sites could not be performed. Moreover, thyroid- and parathyroid-specific outcomes, including recurrent laryngeal nerve injury, hypocalcemia, and cervical hematoma, were not directly recorded in the NSQIP.

This study suggests that the CESRI accurately determines major postoperative adverse events for patients undergoing thyroid or parathyroid surgery. This risk model can help surgeons counsel patients preoperatively and may potentially allow for targeting of modifiable factors to mitigate perioperative risk.

Accepted for Publication: July 7, 2019.

Corresponding Author: Nancy Mayo, BSc, MSc, PhD, Department of Epidemiology, Biostatistics and Occupational Health, McGill University, 3654 prom Sir-William-Osler, Montreal, QC H3G 1Y5, Canada (nancy.mayo@mcgill.ca).

Published Online: September 5, 2019. doi:10.1001/jamaoto.2019.2413

Author Contributions: Dr Mascarella had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: Mascarella, Milad.

Drafting of the manuscript: Mascarella, Milad.

Critical revision of the manuscript for important intellectual content: Mascarella, Richardson, Mlynarek, Payne, Forest, Hier, Sadeghi, Mayo.

Statistical analysis: Mascarella, Milad, Mayo.

Obtained funding: Mascarella.

Administrative, technical, or material support: Mascarella, Milad.

Supervision: Richardson, Mlynarek, Payne, Forest, Hier, Sadeghi, Mayo.

Conflict of Interest Disclosures: Dr Mascarella reported receiving grants from Fonds de Recherche du Quebéc Santé during the conduct of the study. No other disclosures were reported.

Funding/Support: This study was funded by grant 255110 from the Fonds de Recherche du Quebéc–Santé.

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Information: Performing this study was part of the Master of Science Thesis in Epidemiology requirement in the Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Quebec, Canada (Dr Mascarella).