The Association of Frailty and Neighborhood Disadvantage with Emergency Department Visits and Hospitalizations in Older Adults

Abstract

Background

Risk stratification and population management strategies are critical for providing effective and equitable care for the growing population of older adults in the USA. Both frailty and neighborhood disadvantage are constructs that independently identify populations with higher healthcare utilization and risk of adverse outcomes.

Objective

To examine the joint association of these factors on acute healthcare utilization using two pragmatic measures based on structured data available in the electronic health record (EHR).

Design

In this retrospective observational study, we used EHR data to identify patients aged ≥ 65 years at Atrium Health Wake Forest Baptist on January 1, 2019, who were attributed to affiliated Accountable Care Organizations. Frailty was categorized through an EHR-derived electronic Frailty Index (eFI), while neighborhood disadvantage was quantified through linkage to the area deprivation index (ADI). We used a recurrent time-to-event model within a Cox proportional hazards framework to examine the joint association of eFI and ADI categories with healthcare utilization comprising emergency visits, observation stays, and inpatient hospitalizations over one year of follow-up.

Key Results

We identified a cohort of 47,566 older adults (median age = 73, 60% female, 12% Black). There was an interaction between frailty and area disadvantage (P = 0.023). Each factor was associated with utilization across categories of the other. The magnitude of frailty’s association was larger than living in a disadvantaged area. The highest-risk group comprised frail adults living in areas of high disadvantage (HR 3.23, 95% CI 2.99–3.49; P < 0.001). We observed additive effects between frailty and living in areas of mid- (RERI 0.29; 95% CI 0.13–0.45; P < 0.001) and high (RERI 0.62, 95% CI 0.41–0.83; P < 0.001) neighborhood disadvantage.

Conclusions

Considering both frailty and neighborhood disadvantage may assist healthcare organizations in effectively risk-stratifying vulnerable older adults and informing population management strategies. These constructs can be readily assessed at-scale using routinely collected structured EHR data.

Supplementary Information:

The online version contains supplementary material available at 10.1007/s11606-023-08503-x.

INTRODUCTION

As the population aged 65 years and older grows in the USA, healthcare organizations face challenges in meeting an increased need for chronic disease management, complex care, and healthcare resource consumption.^{1, 2} Older adults experience high rates of emergency department (ED) visits³ and 16.8% have at least one hospitalization per year,⁴ which may be avoidable in the setting of high quality ambulatory care or preventive services^5–7 and social support.⁸ The Behavioral Model of Health Service Use posits that contextual- and individual-level predisposing, enabling, and need factors are the primary processes that explain health behaviors, such as utilization, which consequently affect downstream health outcomes.⁹ Frailty, an individual-level evaluated need factor, and neighborhood disadvantage, a contextual community-level predisposing factor for hospitalizations, are two constructs that may help to identify vulnerable subsets of older adults with unmet or future medical, functional, and social needs.

Frailty is an age-related biological process resulting in a decrease in physiologic and functional reserve, which leads to increased vulnerability and elevated risk of adverse health outcomes,¹⁰ such as mortality, falls, and disability.^11–14 Even controlling for age and multi-morbidity, frailty further risk-stratifies a population of vulnerable older adults¹¹ and has emerged as an international target for screening in healthcare.^15–17 Frailty is also associated with higher levels of healthcare utilization, inclusive of the spectrum from primary care visits to inpatient hospitalizations.¹⁸ Frailty can increase the likelihood that individuals will seek or require acute care services due to diminished physical and functional health and increased susceptibility to illness and injuries.⁹ Importantly, when based on the theory of deficit accumulation, a measure of frailty can be operationalized using routine data captured in the electronic health record (EHR) as an electronic Frailty Index (eFI).¹³ The validity of such an approach has now been demonstrated repeatedly across several health systems and countries.^{13, 15–17, 19, 20}

Socioeconomic status (SES), a component of Social Determinants of Health (SDoH), can also influence health-related risks, outcomes, and healthcare access.^{21, 22} Outside of proxy measures such as insurance status (i.e., self-pay or qualifying for Medicaid), SES is not ascertained as part of routine medical care. This has led to considerable interest in geographic, area-based measures such as the area deprivation index (ADI), an indicator of area disadvantage. The ADI is a composite measure of 17 variables across SES domains of income, education, employment, and housing derived from U.S. Census Bureau data.^{23, 24} Living in an area of higher disadvantage, or lower SES, is associated with increased risk of mortality²⁵ and morbidity,^26–29 in addition to higher rates of healthcare consumption.^{5, 24, 30–32} Area disadvantage encompasses various socioeconomic and environmental factors that collectively predispose individuals living in that neighborhood to have more healthcare needs, contributing to higher utilization rates.⁹ The ADI is publicly available and has the potential to facilitate screening to identify older adults who may lack social or community-based resources.

Frailty is associated with lower SES levels among older adults.^33–36 SES has been examined as a moderator of frailty’s effect on mortality,^{37, 38} but most studies have examined the independent contribution of each factor to utilization,^{12, 13, 24, 30, 39–42} leaving a gap in the exploration of the intersection of frailty and socially vulnerable populations.⁴³ In addition, the myriad measures of frailty⁴⁴ and SES⁴⁵ further complicate our ability to translate findings into meaningful clinical action. While several studies have shown a positive association between frailty and neighborhood disadvantage with acute healthcare utilization,^{12, 13, 24, 30, 39–42} these studies have largely focused on re-admissions and cohorts with specific chronic diseases. Our goal was to examine the association between frailty and neighborhood disadvantage collectively, ascertained pragmatically from the EHR, with recurrent acute healthcare utilization in a primary care population of older adults. The goal is that this data may help identify vulnerable subsets of older adults with unmet or future medical, functional, and social needs who may be amenable to population health management initiatives.^{1, 5, 9}

METHODS

Setting and Population

This retrospective observational cohort study included patients identified using Atrium Health Wake Forest Baptist’s (AHWFB) electronic health record (EHR) (Epic, Verona, WI), who were aged 65 years and older and lived in North Carolina or lower Virginia as of January 1, 2019 (index date), and who were attributed to an affiliated Accountable Care Organization (ACO) registry. AHWFB provides primary, specialty, and hospital-based care across a 24-county region in western North Carolina. We required the presence of at least 2 ambulatory visits with blood pressure measurements taken in the 2 years preceding the index date to focus on older adults receiving some level of outpatient or primary care from AHWFB. We excluded individuals for whom we could not calculate the eFI or obtain an ADI value based on their residence. The study was approved by the Wake Forest University School of Medicine Institutional Review Board.

Key Variables

The composition and calculation of the eFI have been previously described, and it is currently calculated weekly for all patients aged 55 years and older at AHWFB.¹³ Frailty status was categorized as fit (eFI ≤ 0.10), pre-frail (0.10 < eFI ≤ 0.21), and frail (eFI > 0.21). For our cohort, we obtained block-level geographical identifiers (GEOID) from structured address data in the EHR as of the index date, which we linked to ADI data from 2018.²⁴ We used national ADI percentiles, which range from 1 to 100, with 100 representing the highest level of neighborhood disadvantage. Consistent with previous analyses of readmission risk^{24, 30} and guided by visual observation of unadjusted rates of utilization across ADI quantiles (see Fig. 1), we categorized neighborhood disadvantage into three groups: low deprivation (AHWFB’s bottom 50th percentile; ADI percentiles 3–63), mid-deprivation (AHWFB’s middle 35th percentile; ADI percentiles 64–82), and high deprivation (AHWFB’s top 15th percentile; ADI percentiles 83–100). We used structured data from AHWFB’s EHR and data from an admission, discharge, and transfer network (Bamboo Health), which aggregates utilization data across healthcare organizations participating in affiliated ACOs. We defined acute healthcare utilization as recurrent all-cause events, encompassing ED visits, observation stays, and inpatient hospitalizations. Each individual could experience multiple events on different days of follow-up, with events occurring on consecutive days treated as a single event. For instance, a hospitalization that followed an ED visit the next day was considered a single event, and the day of the ED visit was used as the time of the event.

Figure 1.

Line graph of unadjusted acute healthcare utilization rates across ADI quantiles. Note: National ADI values were categorized into 20 quantiles based on AHWFB’s distribution of national ADI values. Parentheses and brackets are used to indicate whether an endpoint value is not included or included in the quantile, respectively. ADI indicates area deprivation index and AHWFB indicates Atrium Health Wake Forest Baptist.

We examined ED visits, observation stays, and inpatient visits over one year of follow-up. We chose this shorter length of follow-up to circumvent potential bias introduced by the COVID-19 pandemic, which altered regular clinical activities in early 2020. Mortality information was supplemented by deterministic linkage (based on name, age, gender, date of birth, and race/ethnicity) to the North Carolina State Center for Health Statistics death index.

Statistical Analysis

To model the statistical interaction between frailty status and neighborhood disadvantage, we used a multivariate recurrent time-to-event model within a Cox proportional hazards framework, including random effects to account for within-individual correlations.⁴⁶ Observations were right-censored at the time of death or at the end of a 365-day follow-up window. Covariates were chosen based on theoretical and empirical association with outcomes from previous studies¹³ and availability within the EHR. Demographic variables included sex (male, female), race and ethnicity (White, Black, Hispanic or Latinx, other), and age (years). Number of outpatient encounters (≤ 1, 2 to 4, ≥ 5), and number of ED visits, observation stays, and inpatient encounters (0, 1, or ≥ 2) in the year prior to index date were included to account for informed presence bias⁴⁷ and were categorized according to each variable’s distribution. Multi-morbidity was not included as a specific covariate because it is included in the eFI and thus is at least partially collinear.⁴⁸ We added a two-way interaction term to evaluate whether the relative associations of frailty status and area disadvantage with utilization were dependent. We present associations (hazard ratios) of groups with varying combinations of eFI and ADI levels relative to a reference group of older adults who were fit and living in an area of low deprivation. To quantify statistical interactions, we report the relative excess risk index (RERI) for additive effects and the multiplicative interaction (INT_M) ratios for pre-frailty and frailty by mid- and high deprivation and vice versa.^{49, 50} A P-value of  < 0.05 was considered statistically significant. Analyses were performed in R (Version 4.2.2).⁵¹

RESULTS

Of 53,791 patients attributed to affiliated ACOs that met the inclusion criteria, 49,366 (91.8%) were geocoded and linked to an ADI percentile, and 51,327 (95.4%) had an eFI. A total of 47,566 individuals had both eFI and ADI values and were able to be included in our analysis. Visual analysis indicated the distribution of our population skewed towards higher disadvantage relative to national ADI percentiles (Supplemental Fig. 1). Those categorized as frail were more likely to be female and older, and with higher comorbidity scores in addition to higher utilization of acute and outpatient care in the prior year (Table 1). Those with frailty were more likely to live in areas of higher deprivation relative to those categorized as pre-frail or fit.

Table 1.

Characteristics of Population by eFI Category

	Overall	Fit	Pre-frail	Frail	P-value
N	47,566	11,600	25,943	10,023
Age, years, (median [IQR])	73.4 [69.3–79.3]	71.4 [68.2–75.9]	73.6 [69.43–79.2]	76.3 [71.1–83.0]	< 0.001
Age, no. (%)					< 0.001
65 to < 75	25,203 (53.0)	7659 (66.0)	13,546 (52.2)	3998 (39.9)
75 to < 85	17,460 (36.7)	3419 (29.5)	9884 (38.1)	4157 (41.5)
85 +	4903 (10.3)	522 (4.5)	2513 (9.7)	1868 (18.6)
Sex, no. (%)					< 0.001
Female	28,287 (59.5)	6243 (53.8)	15,330 (59.1)	6714 (67.0)
Male	19,279 (40.5)	5357 (46.2)	10,613 (40.9)	3309 (33.0)
Race/ethnicity, no. (%)					< 0.001
Black non-Hispanic or Latinx	5518 (11.6)	1084 (9.3)	3176 (12.2)	1258 (12.6)
Hispanic or Latinx	610 (1.3)	163 (1.4)	323 (1.2)	124 (1.2)
Other non-Hispanic or Latinx	1076 (2.3)	340 (2.9)	561 (2.2)	175 (1.7)
White non-Hispanic or Latinx	40,362 (84.9)	10,013 (86.3)	21,883 (84.4)	8466 (84.5)
Charlson comorbidity index category, no (%)					< 0.001
0 to < 2	27,343 (57.5)	9969 (85.9)	15,112 (58.3)	2262 (22.6)
2 to < 4	13,273 (27.9)	1472 (12.7)	8046 (31.0)	3755 (37.5)
4 +	6950 (14.6)	159 (1.4)	2785 (10.7)	4006 (40.0)
Area deprivation					< 0.001
Low (lowest 50%)	22,944 (48.2)	6365 (54.9)	12,312 (47.5)	4267 (42.6)
Mid (mid 35%)	17,208 (36.2)	3848 (33.2)	9491 (36.6)	3869 (38.6)
High (top 15%)	7414 (15.6)	1387 (12.0)	4140 (16.0)	1887 (18.8)
Outpatient visits in prior year, no. (%)					< 0.001
0–1	19,853 (41.7)	6267 (54.0)	10,724 (41.3)	2862 (28.6)
2–4	13,154 (27.7)	3432 (29.6)	7364 (28.4)	2358 (23.5)
5 +	14,559 (30.6)	1901 (16.4)	7855 (30.3)	4803 (47.9)
Emergency, observation, and inpatient visits in prior year, no. (%)					< 0.001
0	40,055 (84.2)	10,796 (93.1)	22,257 (85.8)	7002 (69.9)
1	5005 (10.5)	654 (5.6)	2665 (10.3)	1686 (16.8)
2 +	2506 (5.3)	150 (1.3)	1021 (3.9)	1335 (13.3)
Mortality during analytic period, no. (%)	376 (0.8)	24 (0.2)	176 (0.7)	176 (1.6)	< 0.001

Differences between groups were assessed using a chi-square test for categorical variables and Kruskal-Wallis for nonparametric numeric variables. Log-rank P-value reported for mortality

Those living in areas of high deprivation had the highest mean cumulative event count over time within eFI categories (Fig. 2). Unadjusted analyses indicated a higher hazard of acute care utilization for pre-frail and frail individuals relative to those who were fit (Table 2). Similarly, those who lived in areas of mid- and high deprivation experienced a higher hazard of acute healthcare utilization compared to those who lived in areas of low deprivation.

Figure 2.

Unadjusted cumulative count of acute healthcare utilization over time by deprivation and eFI. Note: eFI indicates electronic Frailty Index, and 95% confidence intervals were derived from bootstrapping with 1000 samples with replacement for each category.

Table 2.

Unadjusted Associations of eFI and Area Deprivation Categories with Acute Healthcare Utilization

Variable	Event count/rate per 100 person years	Unadjusted HR (95% CI)	P-value
eFI category
Fit	2208/19.1	Reference
Pre-frail	9688/37.5	1.97 (1.85–2.10)	< 0.001
Frail	8669/87.3	4.60 (4.30–4.92)	< 0.001
Area deprivation
Low (lowest 50%)	8662/37.9	Reference
Mid (mid 35%)	7400/43.2	1.14 (1.09–1.20)	< 0.001
High (top 15%)	4503/61.0	1.61 (1.52–1.71)	< 0.001

In an adjusted model, we observed a statistically significant interaction between frailty and disadvantage (P = 0.023), indicating dependent associations holding all other covariates constant. The hazard of acute healthcare utilization increased across ascending levels of eFI and ADI and was highest for those who were frail and living in high deprivation areas (Fig. 3). Frailty was associated with utilization across all categories of deprivation, as was deprivation across eFI categories (Table 3). The magnitude of associations for frailty and pre-frailty with utilization were notably larger than the associations for living in areas of mid- to high deprivation. Although we did not observe multiplicative effects across levels of frailty and neighborhood disadvantage (indicating the combined effects were not greater than the product of individual effects), we observed additive effects for frailty and mid- to high deprivation categories. The estimated joint effect of frailty and high deprivation were greater than the sum of the estimated individual effects of these variables. The interaction accounted for an additional 14.9% of the total association with utilization beyond the percent of effects attributed by frailty (56.2%) and high deprivation (28.9%) alone. Similarly, the estimated joint effects of pre-frailty and mid-deprivation were greater than the sum of the individual effects. Upon further exploration, we discovered that the effects of frailty relative to the pre-frailty group were greater for those living in mid- (INT_M = 1.11, 95% CI 1.04–1.08; P = 0.003) and high (INT_M = 1.08, 95% CI > 1.00–1.17; P = 0.046) deprivation areas compared to the effects within the low deprivation area.

Figure 3.

Associations of eFI and area deprivation categories with acute healthcare utilization. The reference category comprises those who are fit living in areas of low deprivation. Model adjusted for age, race, gender, outpatient and acute healthcare utilization in the prior year, the interaction between eFI and ADI, and random effects for within-subject correlation. All P-values corresponding to hazard ratios (HRs) are  < 0.001. PY indicates person-years, HR indicates hazard ratio, and CI indicates confidence interval.

Table 3.

Relative Associations of eFI and Area Deprivation Categories with Acute Healthcare Utilization Including Additive and Multiplicative Effects

	eFI contrast HR (95% CI); P
Area deprivation	Pre-frail: Fit	Frail: Fit
Low	1.54 (1.45–1.65); < 0.001	2.39 (2.232.56); < 0.001
Mid (mid 35%)	1.43 (1.32–1.55); < 0.001	2.44 (2.25–2.65); < 0.001
High (top 15%)	1.58 (1.41–1.78); < 0.001	2.64 (2.35–2.97); < 0.001
	Deprivation contrast HR (95% CI); P
eFI category	Mid: Low	High: Low
Fit	1.16 (1.06–1.28); 0.001	1.22 (1.08–1.38); 0.002
Pre-frail	1.08 (1.03–1.13); 0.001	1.25 (1.19–1.32); < 0.001
Frail	1.19 (1.13–1.25); < 0.001	1.35 (1.28–1.43); < 0.001
	Measure of interaction
eFI and ADI category	Additive effects (95% CI); P	Multiplicative effects (95% CI); P
Pre-frail and Mid	− 0.05 (− 0.17–0.08); 0.763	0.92 (0.83–1.02); 0.135
Pre-frail and High	0.17 (< 0.01–0.34); 0.028	1.02 (0.90–1.17); 0.720
Frail and Mid	0.29 (0.13–0.45); < 0.001	1.02 (0.92–1.13); 0.675
Frail and High	0.62 (0.41–0.83); < 0.001	1.11 (0.97–1.27); 0.142

Adjusted for age, race, gender, outpatient, and acute care utilization in the prior year, the interaction between eFI and ADI, and random effects for within-subject correlation. eFI indicates electronic frailty index, ADI indicates area deprivation index, HR indicates hazard ratio, and CI indicates confidence interval

DISCUSSION

Our findings indicated that high-cost, high-burden acute healthcare utilization was strongly associated with pragmatic measures of frailty and neighborhood disadvantage. While these measures can be readily available from the EHR, they are not, to date, typically used as part of resource planning and risk stratification in the USA.^{17, 32, 45} The associations between frailty, area deprivation, and utilization are consistent with the extant literature,^{5, 18, 30, 32, 42} and suggest value in integrating these passive digital markers of risk into healthcare organizations’ data infrastructure to use in scalable risk-stratification practices to identify vulnerable older adults and inform population health management efforts. A notable finding was the substantial magnitude of the correlation between frailty and utilization across all levels of neighborhood disadvantage, in accordance with trends reported for readmission rates.¹² We also observed that a larger proportion of older adults with frailty living in areas of mid- and high-deprivation relative to fit and pre-frail groups consistent with our knowledge that adverse SDoH conditions contribute to the deterioration of physical health and the development of frailty.^{9, 52, 53} It may, therefore, be prudent to emphasize preventive interventions targeting social needs and functional limitations not only for the highest-risk strata but also for older adults with pre-frailty and living in areas of higher deprivation to reduce current and future high-burden events through preventing or slowing individuals’ progression to frailty.

At the policy level, eFIs have gained traction in the USA^{13, 17, 54} but are not mandated as a tool for identifying vulnerable older adults in healthcare settings as is done within the UK.¹⁶ We have also recognized the need for incorporating and addressing social needs in the healthcare sector^{45, 55} through leveraging the widespread adoption of the EHR as outlined by the Centers for Medicare and Medicaid.⁵⁶ Healthcare organizations engaging in population health management may efficiently incorporate these scalable measures of vulnerability into automated, EHR-based risk-stratification mechanisms to quickly identify older adults with potential unmet medical, functional, and/or social needs and intervene accordingly. In a systematic review, Preston et al.⁵⁷ found that organization-level changes that initiated practices of identifying frail and high-risk older adults (risk-stratification) and/or providing specific care management interventions could reduce ED visits, reduce inpatient admissions, and improve discharge outcomes. Tools and strategies for addressing frailty (support services, therapy services, rehabilitation, etc.), however, may differ according to the resources available in various environments. For example, those living in areas of deprivation may not reside in a safe, walkable neighborhood, or have community-based exercise opportunities, or even reliable transportation or geographic proximity to physical therapy services to participate in interventions incorporating physical activity. Care management programs, patient navigators, and community health workers are interventions that can fill social gaps, extend health services, and reduce barriers to primary and preventive care to reduce disparities and high-burden utilization.^58–63 Providing these supports can increase the potential for addressing health-related issues before they become more critical, thereby reducing the necessity for acute healthcare utilization. Additionally, creating opportunities for accessible healthcare and social services offers alternative avenues for care beyond what individuals might perceive as their only option, potentially curbing unnecessary ED visits. Healthcare organizations may use risk stratification based on the eFI and ADI to more efficiently direct these types of interventions or even partner with high-risk communities to expand resources.

This study has several strengths. First, we were able to calculate an eFI and link an ADI percentile to nearly 90% of ACO attributed patients in our health system, which demonstrates that these tools are accessible, easily applied, and therefore scalable within a healthcare organization. Second, we used multiple data sources (EHR, Bamboo Health, and vital statistics) to create a better picture of our cohort’s health and utilization to compensate for the USA’s fragmented data environment across competitive healthcare systems. Lastly, we used a publicly available, objective measure of neighborhood disadvantage, which aggregates multiple attributes of the context in which a person lives. Although the ecological fallacy cautions us in ascribing community-level characteristics to individuals, area-level variables capture an objective risk unattainable through subjective patient inquiry and can serve as a preliminary screening mechanism to identify patients who might benefit from a more thorough investigation of social needs. Area-level data may also help healthcare organizations identify communities that would most benefit from partnerships or resource distribution.

Our study has a few limitations. First, generalizability may be limited, as this study includes ACO-attributed patients within a single healthcare system. Future research should examine trends across different populations and institutional settings, especially those with clinical contact patterns where patients are not embedded in a primary care network of a given healthcare system. Second, as with most EHR studies, there is potential for incomplete ascertainment of utilization and vital status, which may inflate estimates of follow-up time free from acute healthcare utilization. This may be why our observed all-cause mortality rates for those with frailty are slightly lower than rates reported for other US-based cohorts after one year of follow-up.^{64, 65} We also considered the impact of historical tracking of ACO enrollment in the EHR and disenrollment at end-of-life, which introduces potential selection bias for presumably healthier individuals remaining in ACO plans. Our acute healthcare utilization rates are in line with what we would expect given national rates for older adults^{3, 4} and considering a lack of comparable study populations and outcomes as well as the substantial heterogeneity in reported effect sizes for the relative associations between frailty and hospitalizations.⁶⁶ Lastly, we did not account for the possibility of a time-dependent nature of area disadvantage and frailty. However, we would not anticipate a sizeable number of individuals would experience frailty progression or relocate to an area with a substantially different level of disadvantage within the span of a year.

Future research should test whether linking the eFI and ADI to specific interventions will reduce high cost and high-burden utilization. AHWFB is currently piloting an intervention (eFRIEND, NCT05293730 clinicaltrials.gov) in which community health workers connect older adults with frailty to resources that address the functional and social needs of each patient. This pragmatic pilot trial will assess effects of the intervention on the number of ED visits and inpatient hospitalization as a primary outcome.

CONCLUSIONS

Frailty and neighborhood disadvantage, measured pragmatically using the eFI and ADI, are accessible measures of risk derived from routine data collected in EHRs that may assist healthcare organizations in more effectively risk-stratifying vulnerable older adults and implementing population health management strategies. Policy and targeted interventions have the potential to reduce costly and burdensome emergency and inpatient healthcare utilization by addressing patients’ unmet medical, functional, and social needs.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contribution:

KML wrote the first draft of the manuscript and performed all analyses. KEC conceptualized the overarching idea, and KML, NMP, EW, and DP participated in planning. RP provided input on analyses, and RP, KEC, and NMP provided a substantial first review and editing of several drafts. EW, DP, NMP, AH, JMH, JG, MD, BJW, and JH participated in the review and revision of the final manuscript.

Data Availability

The datasets generated during and/or analyzed during the current study are not publicly available due to the use of sensitive data from healthcare records. However, these datasets can be made available from the corresponding author upon reasonable request and subject to appropriate approvals or permissions.

Declarations:

Disclosure:

Wake Forest University Health Sciences has made its eFI a commercial product, so that other health systems may use it for their patients; royalties will go to support the team for future frailty research. The electronic Frailty Index was developed with support from the National Center for Advancing Translational Sciences, National Institutes of Health (NIH) (UL1TR001420), the Wake Forest Baptist Health President’s Office, and the Wake Forest Center for Healthcare Innovation. Additional support was provided by the James D. Duke Endowment and the National Institutes of Health through K76AG059986 (KEC), K23HL146902 (DP), K23AG070234 (JG), P30AG021332 (NMP), and U24AG059624 (NMP).

Disclaimer:

The NIH, the Wake Forest Center of Healthcare Innovation, and the Duke Endowment Foundation had no role in the design and conduct of the study, collection, management, analysis, interpretation of the data, preparation, review, approval of the manuscript, or decision to submit the manuscript for publication.

Ethics:

This study was approved by the Wake Forest University Health Sciences Institutional Review Board with a waiver of informed consent.

Conflict of Interest

We have reported disclosures to ensure transparency and have no conflicts to report. We want to emphasize that these disclosed interests did not have any direct influence on the design, execution, or interpretation of the research and that the findings are reported objectively.