Introduction

Elderly patients can develop clinical frailty due to the composite presence of geriatric and medical conditions that multi-modally affect different domains of functions, dependence, ambulation, and medical/clinical performance [1]. Essentially, frailty combines conditions of malnutrition, physiological aberrations in metabolism, feeding difficulties, and fecal and urinary incontinence with deficiencies in the socioeconomic sphere, including low-income status, housing difficulties and social isolation [1, 2] to describe a clinical phenotype that is prone to increasingly adverse effects following surgeries and medical interventions [3]. The Johns Hopkins Adjusted Clinical Groups (ACG) definition of frailty is based on the consideration of 10 diagnostic clusters, which are listed in Supplementary Tables 1 and 4. This definition of frailty is particularly applicable in patients that require gastric surgery for gastric tumors [5, 6], as these patients are often elderly or have underlying feeding and nutritional difficulties that predispose them to frailty-development [7, 8]. However, frailty is not only restricted to elderly patients but can also be manifested in younger patients with malignancy, which induces conditions overlapping with frailty-defining features that range from nutritional deficiencies to social maladjustments. Prior NIS studies have also used Johns Hopkins ACG frailty criteria to define frailty in patients with malignancy undergoing resective procedures.

In surgical literature, it is currently well known that the underlying presence of frailty imposes a significant degree of risk in postoperative outcomes following various surgical procedures [4]. However, while the effects of frailty have been alluded to in multiple studies, its effects on post-gastrectomy remain undefined due to the sparsity of studies. Thus, in light of this, we plan to systematically evaluate the postoperative effects of frailty on gastrectomy outcomes in patients with gastric malignancy; to accomplish this, we use a national registry of hospital data and propensity-score matching to evaluate the respective risks associated with frailty on post-gastrectomy outcomes.

Methods

NIS database and variable selection

This study uses 2011–2017 data from the National Inpatient Sample (NIS), an annually updated database pertaining up to 40 ICD diagnoses and 25 ICD-10 PCS codes per patient, such as demographics, admission, and discharge characteristics [9]. It is derived from the State Inpatient Databases (SID), which compiles billing information from non-Federal US hospitals excluding centers for rehabilitation or long-term acute care. As part of the Healthcare Cost and Utilization Project (HCUP), NIS is funded by the Agency for Healthcare Research and Quality (AHRQ). NIS underwent several changes in 2011–2017. Previously, NIS derived all its data from 1000 hospitals; since 2014, it began gathering data from a fraction of discharges from all hospitals [9, 10]. It was further updated in 2015 to follow ICD-10. This study adheres to the updated NIS guidelines [11].

Variables of interest for this study were chosen using a search engine that browses for keywords in a database integrating the official CMS ICD-9 to ICD-10 conversion tables as well as diagnosis-related group (DRG) sets for each ICD system [12,13,14,15,16,17,18,19]. Cross-linking between the ICD codes allowed for variables to be selected with minimum heterogeneity.

Study cohort and study variables

The study cohort includes patients who underwent either total or partial gastrectomy (surgical removal of the stomach) for the treatment of gastric cancer; those younger than 18 years old and those who received multiple procedures were not excluded in the analysis. Frailty was used as the exposure variable to divide the cohort into frailty-present and frailty-absent cohorts. Like previous National Inpatient Studies, this study employs the Johns Hopkins ACG definition of frailty [20,21,22], a diagnostic index composed of criteria, such as dementia, housing needs and homelessness, difficulties with ambulation, weight loss and feeding difficulties, vision impairment, frequent falls, urinary and fecal incontinence, malnutrition, and pressure ulcerations. The primary endpoints investigated in this study are mortality, length of stay, hospitalization costs, and disposition at discharge. Secondary endpoints include postoperative bleeding, postoperative infection (i.e., surgical-site infections, infections of the postsurgical seroma, postsurgical abscesses, peritonitis/abdominal wall infections, internal abscesses, sepsis, and other deep and superficial surgery-related infections), postoperative wound complications (i.e., disruptions in the surgical wound and disruptions due to trauma, wound dehiscence, and fistulous malformations), and postoperative respiratory failure. Further analysis of subgroups divided by age (younger than 65 years vs 65 years and older) and by gastrectomy type (partial vs total) was conducted and recorded in the supplementary tables. All variables used along with their corresponding ICD-9 and 10 codes are documented in the supplementary tables.

Study design and statistical analysis

Propensity-score matching between cohorts was conducted using the following covariates: age, gender (female), race, diabetes, hyperlipidemia, hypertension, chronic obstructive pulmonary disease, coronary artery disease, chronic kidney disease, congestive heart failure, coagulopathy, alcohol use disorder, cigarette use, obesity, nature of procedure (emergent vs non-emergent in relation to time of admission), spread to lymph nodes, spread to non-gastrointestinal organs, spread to gastrointestinal organs, and chemotherapy history. These comorbidities were designated covariates based on their projected impact on mortality and clinical outcomes. Post matching, the frailty-present and frailty-absent cohorts underwent univariate and multivariate analyses. In univariate analysis, parametric/non-parametric tests in addition to Chi-squared or Fisher’s Exact tests were employed to compare mean values for demographics, medical comorbidities, socioeconomic status, hospital information, and other endpoints. The multivariate analysis generated primary endpoints (mortality, length of stay, hospitalization costs, and disposition at discharge) as dependent variables through either logistic or Poisson’s regression analysis while controlling for variables excluded from the matching process. All multivariate findings were evaluated for multicollinearity, fit, and accuracy using several analyses: Akaike information criterion (AIC) assessment, Bayesian information criterion (BIC) assessment, and variance inflation factor (VIF) analysis [23,24,25]. For every endpoint, adjusted and non-adjusted odd ratios were generated with confidence intervals, and statistical significance was assigned to p values less than 0.05.

The multivariate imputation by chained equation (MICE) method was used to fill in any missing NIS data. The MICE method is commonly employed by large database studies over singular imputation methods for its relative accuracy in imputing missing values, owed to its process of running multiple sequential imputations according to feedback [26,27,28].

All calculations involved in analysis were generated by Rstudio version 1.2.5042 with R code version 3.6.3. With all data originating from the HCUP-NIS database, this study was not required to undergo National Review Board or institutional review board approval.

Results

Patient selection

Figure 1 outlines the patient selection procedure of the study. Of 5,286 patients comprising the initial pre-match cohort, 1171 patients with and 1171 patients without frailty were included in the post-match cohort.

Fig. 1
figure 1

This figure demonstrates the patient selection procedure of the study

Full size image

Comparison of demographics and comorbidities

Table 1 demonstrates pre-match and post-match information on the demographics and comorbidities of gastrectomy patients with and without frailty. There were no differences in age, gender, race, or in the incidence of medical, gastric, and cancer-related comorbidities between the two cohorts following the matching process.

Table 1 Pre- and post-match comparisons of patients with and without frailty; patient demographics and comorbidities of those who underwent gastrectomy for gastric cancer
Full size table

Comparison of socioeconomic status and hospital characteristics

Table 2 demonstrates pre- and post-match socioeconomic and hospital data for frail and non-frail patients who underwent gastrectomy. After matching, patients with frailty were more likely to be in the lowest income quartile and less likely to be seen in hospitals in the Northeast. They were not significantly different in terms of hospital location (urban vs rural), hospital teaching status, or method of payment.

Table 2 Pre- and post-match comparisons of patients with and without frailty; patient socioeconomic and hospital characteristics of those who underwent gastrectomy for gastric cancer
Full size table

Comparison of hospital outcomes and complications

Table 3 demonstrates pre- and post-match hospital outcomes of patients undergoing gastrectomy. After matching, patients with frailty had higher mortality (6.83 vs 3.50% p < 0.001, OR 2.02 95%CI 1.37–2.97), longer length of stay (16.7 vs 12.0d; p < 0.001), and higher costs ($191,418 vs $131,367; p < 0.001) than patients without frailty. They were more likely to be discharged to non-routine care facilities or die during hospitalization. Regarding postoperative complications, frail patients had an increased incidence of postoperative wound complications (3.42 vs 0.94% p < 0.001, OR 3.73 95%CI 1.90–7.31), infection (5.98 vs 3.67% p = 0.012, OR 1.67 95%CI 1.13–2.46), and respiratory failure (6.32 vs 3.84% p = 0.0084, OR 1.69 95%CI 1.15–2.47), but no difference in bleeding complications. Multivariate analysis confirmed the univariate findings in showing frail patients to have increased mortality (p < 0.001, aOR 2.04 95%CI 1.38–3.01), length of stay (p < 0.001, aOR 1.40 95%CI 1.37–1.43), and hospitalization costs (p < 0.001, aOR 1.46 95%CI 1.46–1.46). Figure 2 demonstrates the effects of frailty on post-gastrectomy outcomes using mortality as the primary outcome variable. Furthermore, this combined model includes sub-stratified populations who underwent either partial or total gastrectomy (which is represented in the supplementary tables — see below).

Table 3 Pre- and post-match comparison of patients with and without frailty; patient clinical outcomes of those who underwent gastrectomy for gastric cancer
Full size table
Fig. 2
figure 2

This figure demonstrates the combine multivariate model using frailty as the primary exposure variable and mortality as the outcome; the included strata are those who underwent gastrectomy (as a composite) or surgery subtypes (partial and total gastrectomy)

Full size image

Supplementary tables

Additional post-match subgroup analyses are documented in Supplementary Tables 2 and 3. Supplementary Table 2 compares clinical outcomes of younger (< 65 years) and older patients (≥ 65 years); frailty increases primary outcomes for both cohorts except for mortality in the younger cohort. Supplementary Table 3 compares clinical outcomes for partial and total gastrectomy; all primary endpoints are increased except for mortality in patients receiving total gastrectomy. The cohort-specific stratifications as per the Johns Hopkins ACG frailty criteria are shown in Supplementary Fig. 4.

Discussion

Our post-match findings show that mortality, length of hospital stay, and total cost of care are significantly increased in individuals with frailty undergoing gastrectomy compared to those without frailty. Frailty is also shown to be an independent risk factor for post-procedural complications, such as infection, respiratory failure, and wound complications, and to increase the likelihood of discharge to additional care facilities. Our findings in the supplementary tables confirm the adverse effects of frailty on both age cohorts, though on a greater scale in the elderly.

A large number of cohort studies have attempted to investigate the effects of frailty on post-gastrectomy outcomes. One shortcoming, however, is their utilization of a variety of criteria and indices in classifying frailty, which presents an ongoing challenge in interpreting the collective data [1, 29]. Our study uses the Johns Hopkins ACG system, which defines frailty based on the combined presence of the following domains: malnutrition, impaired vision, dementia, decubitus ulcers, fecal and urinary incontinence, loss of weight, digestive difficulties, poverty, inadequate housing, difficulty in walking, and falls. The findings from our study corroborate several prior studies that demonstrate the positive association between the domains of frailty and mortality, length of stay, and post-operative complications in patients receiving major resective procedures, such as gastrectomy [30,31,32]. Our study strengthens these findings by assessing the effects of frailty across a cohort derived from a national database, using propensity-score matching to optimize the accuracy of analysis. Moreover, it evaluates these effects separately in elderly and younger cohorts, which allows us to further elucidate trends that may vary based on age.

Frailty negatively affects clinical outcomes through numerous physiological mechanisms, especially in patients undergoing high-risk procedures [33]. In general, the state of nutritional deficiency in a frail patient impedes the calorically taxing processes involved in post-surgical recovery [34,35,36]. Protein and energy insufficiency impairs proper wound healing, coagulation, and immunological defenses, which elevates the risk of infection and bleeding [37, 38]. Decreased energy stores also depress neural drive and muscular strength, both of which may contribute to deterioration in diaphragmatic function and ultimately to post-procedural respiratory complications. Furthermore, the ambulatory decline and overall weakness that accompany the diagnosis of frailty subject patients to a higher risk of falls or injury, prolong their stay in the hospital and raising cost of treatment [39]. With delayed discharge, patients are prone to subsequent complications, such as decubitus ulcers and acquired infections, contributing to further disruption from recovery [40].

Frailty’s propensity to induce adverse post-procedural outcomes, as demonstrated in this study, renders it essential for patients to be screened early for indications of frailty during their preoperative assessment. Once components of frailty are identified, clinicians should aim to mitigate those responsive to treatment, such as malnutrition or poor physical performance, through measures, such as dietary supplementation and guided physical training [41, 42]. After the procedure, it is advisable to place the patient under the supervision of a multidisciplinary team that can provide integrated care in multiple areas, such as nutrition, wound care, and respiratory support [43]].

One noteworthy finding in this study was that mortality was not significantly increased in frail patients receiving total gastrectomy, unlike for partial gastrectomy, where a significant increase was seen. While the precise mechanism for this cannot be exacted, part of the reason may be due to the fact that total gastrectomy is a much more invasive and comprehensive procedure with implicitly higher mortality and complication risks, thus diminishing the effect of frailty on its overall outcomes.

Limitations

The primary limitation of this paper pertains to the nature of the database used; the NIS database does not contain post discharge information, which limits the study results to the cross-sectional evaluation performed during admission and in-hospital postoperative course. While the immediate effects of frailty are well characterized in this study, a follow-up prospective study would be helpful in delineating the long-term effects of frailty on patient recovery.