Impact of frailty on inpatient outcomes in thyroid cancer surgery: 10-year results from the U.S. national inpatient sample

Background Frailty is linked to perioperative morbidity and mortality. We evaluated the impact of preoperative frailty on inpatient outcomes of patients undergoing surgery for thyroid malignancy. Methods This population-based, retrospective observational study extracted data of hospitalized patients who were 18 years and older with a primary diagnosis of thyroid cancer undergoing thyroidectomy from the US Nationwide Inpatient Sample (NIS) database (2005–2014). Participants were stratified into frail and non-frail using the Johns Hopkins (ACG) frailty-defining diagnosis indicator. Study endpoints were in-hospital mortality, incidence of surgical and medical complications and prolonged length of stay. Univariate and multivariate analysis were performed to determine associations between the endpoints and frailty. Results Data of 38,202 patients were included. After adjusting for possible confounders, frailty remained significantly associated with higher odds of in-hospital mortality (OR: 3.839, 95% CI: 1.738–8.480), prolonged length of stay (OR: 5.420, 95% CI: 3.799–7.733), surgical complications (OR: 3.144, 95% CI: 2.443–4.045) and medical complications (OR: 6.734, 95% CI: 5.099–8.893) compared with non-frailty. In patients > age 65 years, adjusted odds ratio for frailty was 4.099 (95% CI: 1.736–9.679) for in-hospital mortality, 6.164 (95% CI: 3.514–10.812) for prolonged length of stay, 3.736 (95% CI: 2.620–5.328) for surgical complications, and 5.970, 95% CI: 4.088–8.720 for medical complications, all with significance. Conclusion Frailty is associated with increased risk for adverse inpatient outcomes, including prolonged hospital stay, surgical and medical complications and mortality independent of age and comorbidities in thyroid cancer patients undergoing surgery. Study findings may provide valuable information for preoperative risk stratification.


Introduction
Frailty is highly prevalent among older adults worldwide and is associated with increased risk of falling and hospitalizations, high disability, morbidity and mortality [1,2]. Frailty is defined as a syndrome of lower energy reserve and increased vulnerability to stressors during aging along with age-associated declines in lean body mass, strength, endurance and activity levels [1]. Frailty has also been linked to perioperative morbidity and mortality among older adults in various surgical arenas, including vascular surgery [3], cardiac [4] and elective non-cardiac surgeries [5]. Frailty is shown to be more significant than age itself in predicting poor outcomes [2,6]. As such, frailty is a recognized factor in geriatric oncology, with increased risk of adverse outcomes in older adult cancer patients [6,7].
More than half of older adult cancer patients were found to have either pre-frailty or frailty, increasing risk of chemotherapy intolerance, postoperative complications and 30-day mortality [7].
Thyroid cancer is the most common malignancy in endocrine system organs and its incidence is increasing globally. In the United States, the annual percentage change was 2.4% between the 1980s to 1997 and increased to 6.6% between 1997 and 2009 [8]. Many researchers attribute this increase to enhanced detection through more sensitive screening and diagnostic procedures [9,10]. In fact, the increased incidence in the U.S. is considered to be of epidemic proportions, especially in women, who have higher thyroid cancer rates but lower prevalence at autopsy than men-a trend suggesting that the epidemic is one of higher detection rates than actual disease rates [10]. Although thyroid cancer is more commonly diagnosed at younger ages, and is associated with a good overall prognosis, older adults with thyroid cancer are challenged by a more aggressive disease that can lead to significant morbidity as well as to a significant economic burden on healthcare systems [11][12][13]. In particular, older adults undergoing thyroidectomy for thyroid cancer are prone to more complications, greater risk of recurrence and distant metastases and reduced overall and disease-free survival [11].
Prior studies have investigated the effects of frailty on patients undergoing surgery for a broad range of thyroid disease [14][15][16]. Frailty was more predictive of postthyroidectomy systemic complications than age in older adults with multimodal goiter [14]. In common ambulatory general surgery operations including thyroid surgeries, frailty was associated with increased perioperative morbidity independent of age [15]. A preoperative risk index based on multidimensional framework also incorporated frailty as a major component and was found acceptable to determine major adverse events after thyroid or parathyroid surgery, whereas not for malignancy specifically [16].
However, despite the evidence gained from these studies, the impact of frailty on the outcomes of older adults undergoing surgery for thyroid malignancy has yet to be investigated directly. Therefore, the present study aimed to evaluate the impact of preoperative frailty on inpatient outcomes in older adults undergoing surgery for thyroid malignancy.

Study design and data source
This population-based, retrospective observational study extracted all data from the US Nationwide Inpatient Sample (NIS) database, which is the largest all-payer, continuous inpatient care database in the United States, including about 8 million hospital stays each year [17]. It is administered by the Healthcare Cost and Utilization Project (HCUP) of the National Institutes of Health (NIH). Patient data include primary and secondary diagnoses, primary and secondary procedures, admission and discharge status, patient demographics, expected payment source, duration of hospital stay, and hospital characteristics (i.e., bed size/location/teaching status/hospital region). All patients are initially considered for inclusion. The NIS database derives patient data from about 1050 hospitals from 44 States in the US, sampled to represent a 20% stratified sample of US community hospitals as defined by the American Hospital Association.

Ethics statement
All data were obtained through request to the Online Healthcare Cost and Utilization Project (HCUP) Central Distributor (available at: https://www.distributor.hcupus.ahrq.gov/), which administers the database (certificate # HCUP-4R08J60JV). This study conforms to the datause agreement of the NIS from HCUP. As this study was an analysis of secondary data from the NIS database, patients and the public were not involved directly. The study protocol was submitted to the Institutional Review Board (IRB) of Renji Hospital, School of Medicine, Shanghai Jiao Tong University, which exempted the study from IRB approval. Since all data in the NIS database are de-identified, the requirement for informed consent was also waived.

Study population
Adults ≥18 years old admitted to U.S. hospitals between 2005 and 2014 with a primary diagnosis of thyroid cancer undergoing thyroidectomy were identified in the NIS database through the International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes (code 193) and procedure codes (code 06.2, 06.3, 06.4, 06.5). Participants with no information on mortality status or length of stay were excluded.
The participants were further stratified into frail and non-frail groups, based on the 10 clusters of frailtydefining diagnoses that comprise the Johns Hopkins Adjusted Clinical Groups (ACG)frailty-defining diagnosis indicator, a binary variable, using ICD-9 codes assigned during admission, as described previously [18,19]. Details of relevant codes for frailty are shown in Table 1.

Study variables and outcome measures
Study endpoints were in-hospital mortality, incidence of any surgical complications, any medical complications and prolonged length of stay. Surgical complications included tracheostomy, hoarseness, hemorrhage, hematoma or seroma, cystitis, hypocalcemia, vocal cord paresis or paralysis, wound and technical complications. Medical  Table 1. Prolonged length of stay was defined as length of stay ≥75th percentile of the study cohort.

Covariates
Patients' characteristics included age (grouped into: < 45, 45-64, and ≥ 65 years), gender, race/ethnicity, household income level, insurance status (primary payer), and admission type (elective or emergent). Procedure types were divided into: unilateral / partial thyroidectomy, total (complete) thyroidectomy, and substernal thyroidectomy. Patients requiring cervical lymph node dissection, or with tumor metastasis to the lung or bone, smoking status, obesity or overweight status were also identified through ICD-9 diagnostic or procedure codes. Comorbidities were graded using the Romano adaptation of the Charlson comorbidity index [20,21], excluding codes for the index cancer diagnosis from the solid tumor category, in accordance with previous study [22]. Hospital-related characteristics (bed size/location/teaching status/hospital region) and hospital volume of thyroidectomy cases were also extracted from the database as part of the comprehensive data available for all participants.

Statistical analysis
All categorical variables are expressed as counts (percentages). Comparisons of proportions between groups for categorical variables were performed using Pearson's chi-square test or Fisher's exact test. Univariate and multivariate analysis were performed to determine the associations between frailty and in-hospital mortality, prolonged length of stay (of survivors), incidence of surgical complications, and incidence of medical complications. Additional stratified analyses were performed according to age group and number of comorbidities. All statistical analyses were performed using SAS statistical software version 9.4 (SAS, Cary, NC, USA). Twosided p-values less than 0.05 were considered statistically significant. were determined to be frail, of which 50.99% were ≥ 65 years old, 41.39% male and 58.61% female. Significant differences were found between the frailty and nonfrailty groups in age, gender, race, household income, insurance status, admission type, procedure type, whether or not required cervical lymph node dissection, metastasis to the lung, metastasis to the bone, hospital region and hospital volume (all P < 0.001) ( Table 2). The inpatient outcomes of the patients after thyroidectomy are shown in Table 3. Significant differences were found between frailty and non-frailty groups in all four outcomes of interest: in-hospital mortality, prolonged length of stay, any surgical complications and medical complications, with a greater proportion of these outcomes observed in frail patients (all P < 0.001) ( Table 3).

Associations between inpatient outcomes and frailty
The results of univariate and multivariate regression analysis on associations between frailty and inpatient outcomes are summarized in Associations between outcomes and frailty according to age and number of comorbidities Stratified analysis between frailty and the outcomes of interest according to age and number of comorbidities are summarized in Table 5. Frailty was significantly associated with increased odds of all inpatient outcomes among both older (> = 65 years old) and younger (< 65 years old) subgroups except for in-hospital death. Among the older subgroup, the adjusted odds ratio for frailty was 4.099 (95% CI: 1.736-9.679) for in-hospital mortality, 6.164 (95% CI: 3.514-10.812) for prolonged length of stay, 3.736 (95% CI: 2.620-5.328) for surgical

Discussion
Results of the present study have shown that, after adjusting for all relevant comorbidities and clinical characteristics, hospitalized adult patients with preoperative frailty who underwent surgery for thyroid cancer are at over 3 times risk of in-hospital death and having surgical complications, and at over 5 times risk of having medical complications and prolonged length of stay compared to patients without frailty. The adverse predictive role of on the inpatient outcomes remained significant across subgroups, including younger or older age and number of comorbidities.
The effects of frailty in patients receiving thyroidectomy have been investigated by other authors, reporting surgeries performed for benign thyroid conditions mostly [14][15][16]. Among patients with multimodal goiter who underwent thyroidectomy, the frailty index provided more reliable risk assessment than age for complications associated with thyroidectomy [14]. In the present study, frailty was independently associated with increased risk of having medical complications, surgical complications and prolonged length of stay among both older (> = 65 years old) and younger (< 65 years old) patients. In previous studies focused on head and neck cancer patients, frailty was shown to be a stronger predictor than age for prolonged length of stay, surgical complications or medical complications [6,19]. It was suggested that preoperative frailty assessment can provide useful information about health status and predictive information about outcomes; in cancer patients, in particular, tolerance to chemotherapy and radiotherapy can be predicted independent of age [6]. Nieman et al. [19] found frailty to be an independent predictor of postoperative morbidity and mortality, length of hospital stays and related costs in patients undergoing surgery for head and neck cancer. In that study, interactions with comorbidities also had a greater impact on complications and length of stay when accompanied by frailty. The findings of these studies and ours together contribute to the expanding literature highlighting the relevance of frailty rather than that of chronological age in preoperative decision making and perioperative patient care.
A previous study documented that rehospitalization among elderly patients in Medicare beneficiaries with thyroid cancer after thyroidectomy is both prevalent and costly, thus further predictors should be studied to enhance preoperative risk stratification, improve discharge planning, and increase outpatient support [13]. The present study did not evaluate readmission because the data were not available in the NIS database, which collects admission data separately for prior or subsequent admissions. While the present study focused only on inpatient outcomes provided by the NIS database, it was not the main focus of the investigation. Evidence indicates that frailty assessment using riskstratification tools are of particular help in understanding risks associated with individual older adult patients undergoing emergent general surgery as well as to assist with postoperative management and improve geriatric-  The significance value is shown in bold a Multivariate analysis was adjusted for age, gender, race, household income, insurance status/primary payer, admission type, procedure type, required cervical LN dissection, metastasis to the lungs, metastasis to the bone, smoking, overweight and obesity, CCI and hospital characteristics centered outcomes [23]. Preoperative risk indices for frailty are being used more and more in patient undergoing thyroid and parathyroid surgery to predict major adverse events, including death with 30 days of surgery [16]. The present study defined frail and non-frail groups based on 10 clusters of frailty-defining diagnoses that comprise the Johns Hopkins Adjusted Clinical Groups (ACG) frailty-defining diagnosis indicator, a binary variable that uses ICD-9 codes assigned during admission [18,19]. Among multiple measures of frailty in common use, the Johns Hopkins ACG frailty-defining diagnoses indicator was developed and validated recently to be used specifically with health administrative data, and is not intended to distinguish between degrees of frailty. Briefly, alternatives include the frailty phenotype reported by Fried et al. [1], which is based on five criteria: unintentional weight loss, exhaustion, decreased grip strength, decreased walking speed and low physical activity. The Rockwood index calculates subjective deficits that provide useful predictive information about frailty [24]. The ACG indicator used in the present study has been increasingly applied in studies based on administrative databases, but this new tool is not yet as widely used as other established indices.

Strengths and limitations
The main strength of the present study and its findings is the use of the NIS database, a large and comprehensive database that closely represents the population of the United States and allows results to be generalized to a national population. The first analysis focused on the effects of frailty on the postoperative outcomes of older adults undergoing surgery for thyroid malignancy. Important confounding variables such as patients' comorbidities and hospital characteristics, including hospital volume, were considered and adjusted in the analyses. Nevertheless, this study has a few limitations that may interfere with the analysis and interpretation of result. The ICD-9 coding system was used to identify comorbidities in the included patients. Although comorbidities were graded based on the Charlson Index, the severity of individual comorbidities were unknown, which has the possibility of skewing results. Also, Johns Hopkins ACG frailty-defining index relies on ICD-9 codes. Frailty may The significance value is shown in bold CCI Charlson Comorbidity Index a Multivariate analysis was adjusted for gender, race, household income, insurance status/primary payer, admission type, procedure type, required cervical LN dissection, metastasis to the lung, metastasis to the bong, smoking, overweight and obesity, CCI and hospital characteristics b Multivariate analysis was adjusted for age, gender, race, household income, insurance status/primary payer, admission type, procedure type, required cervical LN dissection, metastasis to the lungs, metastasis to the bone, smoking, overweight and obesity and hospital characteristics be underestimated due to under-coding of the frailtydefining diagnoses. In addition, the index does not allow for defining degrees of frailty. As noted in previous work with the NIS database, the database lacks information on adjuvant therapy, subtype, grade, and stage of thyroid cancer, which may possibly bias the results. The NIS database lacks patients' follow-up data after discharge, precluding the evaluation of later morbidity and mortality. Certain other confounding variables not collected by the NIS such as operation time may complicate analysis and limit the interpretation of results. Further, other important outcomes such as readmission or quality of life are not available in the database, limiting our evaluation and comparison of patient data.