The Interaction Between Geriatric & Neighborhood Vulnerability: Delineating Pre-Hospital Risk Among Older Adult Emergency General Surgery Patients

Abstract

Background:

When presenting for EGS care, older adults frequently experience increased risk of adverse outcomes owing to factors related to age (‘geriatric vulnerability’) and the social determinants of health unique to the places in which they live (‘neighborhood vulnerability’). Little is known about how such factors collectively influence adverse outcomes. We sought to explore how the interaction between geriatric and neighborhood vulnerability influences EGS outcomes among older adults.

Methods:

Older adults, ≥65 years, hospitalized with an AAST-defined EGS condition were identified in the 2016–2019, 2021 Florida State Inpatient Database. Latent variable models combined the influence of patient age, multimorbidity, and Hospital Frailty Risk Score into a single metric of “geriatric vulnerability.” Variations in geriatric vulnerability were then compared across differences in “neighborhood vulnerability” as measured by variations in Area Deprivation Index, Social Vulnerability Index, and their corresponding subthemes (e.g. access to transportation).

Results:

A total of 448,968 older adults were included. For patients living in the least vulnerable neighborhoods, increasing geriatric vulnerability resulted in up to six-times greater risk of death (30-day risk-adjusted HR[95%CI]: 6.32[4.49–8.89]). The effect was more than doubled among patients living in the most vulnerable neighborhoods, where increasing geriatric vulnerability resulted in up to fifteen-times greater risk of death (30-day risk-adjusted HR[95%CI]: 15.12[12.57–18.19]). When restricted to racial/ethnic minority patients, the multiplicative effect was four-times as high, resulting in corresponding 30-day hazard ratios for mortality of 11.53(4.51–29.44) versus 40.67(22.73–72.78). Similar patterns were seen for death within 365 days.

Conclusion:

Both geriatric and neighborhood vulnerability have been shown to affect pre-hospital risk among older patients. The results of this study build on that work, presenting the first in-depth look at the powerful multiplicative interaction between these two factors. The results show that where a patient resides can fundamentally alter expected outcomes for EGS care such that otherwise less vulnerable patients become functionally equivalent to those who are, at baseline, more aged, more frail, and more sick.

Level of evidence:

Prognostic and Epidemiological; Level III

Introduction

Accounting for upwards of 2.5 million hospitalizations¹ and 500,000 procedures² per year, emergency general surgery (EGS) has evolved into a thriving specialty caring for some of the most high-risk and critically-ill patients.^3,4 Among the high-risk patients that EGS providers treat, older adults (≥65 years) represent a unique population—one often faced with increased risk of adverse outcomes owing to factors related to age (e.g. frailty, preexisting multimorbidity) that put older EGS patients at increased risk for mortality during the perioperative period (e.g. death within 30 days) and over the subsequent course of prolonged recovery (e.g. death within 365 days). Research published over the last 10–20 years has shown consistently strong relationships between factors related to age,^5–9 frailty,^10–14 and a patient’s extent of multimorbidity^15–20 that affect both short- and long-term mortality among all EGS patients.^10,11

Emerging research in recent years has expanded these efforts to further explore how variations in where patients live can influence differences in baseline EGS mortality risk.^21–26 Among older adults, increases in neighborhood vulnerability as measured by differences in the United States (US) Center for Disease Control and Prevention/Agency for Toxic Substances and Disease Registry Social Vulnerability Index (SVI)²⁷ have been found to result in modest increases in mortality as far out as 30 days for both low- and high-risk procedures.²⁴ What remains less clear is how the two sets of pre-hospital risk-factors interact. Little is known about how variations in risk related to age (“geriatric vulnerability”) and risk related to the social determinants of health unique to the places in which EGS patients live (“neighborhood vulnerability”) combine to influence mortality among older adults.

Recognizing the challenge that such lack of knowledge poses for the meaningful development of pre-hospital risk-stratification and efforts to provide appropriate resources to older adults when presenting for EGS care, we sought to explore how the interaction between geriatric and neighborhood vulnerability influences differences in 30- and 365-day mortality among older adults. A Bayesian latent variable model was used to combine the influence of age, frailty, and multimorbidity into a single metric of geriatric vulnerability. Differences in neighborhood vulnerability were explored by assessing variations in SVI and an alternative metric known as the Area Deprivation Index (ADI).^28,29 As a secondary objective, differences in 30- and 365-day mortality were further evaluated among older adults identifying as non-Hispanic Black or Hispanic in an effort to account for potential differences in the interaction between geriatric and neighborhood vulnerability encountered among historically disadvantaged populations.

Methods

Data source and study population

Records of index admissions for all older adults (≥65 years) hospitalized with a primary diagnosis of an EGS condition were identified in the 2016–2019, 2021 Florida State Inpatient Database (SID). Inpatient admissions for 2020 were omitted owing to interruptions in usual practice patterns during the COVID-19 pandemic. To be included, older adults needed to present with a primary International Classification of Diseases, 10^th edition, Clinical Modification (ICD-10-CM) diagnosis code corresponding to 1 of the 16 common EGS conditions defined by the American Association for the Surgery of Trauma (AAST).³⁰

Provided by the US Agency for Healthcare Research and Quality (AHRQ) as a part of the Healthcare Cost and Utilization Project, the Florida SID provides longitudinal information on patients admitted to all hospitals within the state, regardless of age or primary insurance payer. It contains data on up to 25 ICD-10-PCS procedure and 40 ICD-10-CM diagnosis codes in addition to information on residential zip codes required to link inpatient records to regional information on neighborhood vulnerability. Data included in the Florida SID were collected from US hospitals by AHRQ in accordance with predetermined data standards.³¹

Defining neighborhood vulnerability

Two publicly-available indices were used to define neighborhood vulnerability: ADI designed by the Center for Health Disparities Research at the University of Wisconsin School of Medicine and Public Health^28,29 and SVI designed by the US Centers for Disease Control and Prevention.²⁷ Both indices have been previously used to look at differences in outcomes among EGS patients.^21–26 Intended to reflect differences in “income, education, employment, and housing quality,” ²⁹ ADI uses data from the American Community Survey to provide an overall assessment of a neighborhood’s extent of regional risk graded on a national scale from 1 (lowest) to 100 (highest). It does not include race as a risk-adjustment factor and can, for that reason, be used by federal programs to allocate resources. SVI, in contrast, was designed to respond to and promote preparedness for national disasters. It uses similar data from the American Community Survey to provide a different overall estimate of regional risk that is inclusive of race. SVI is graded on a national scale from 0.01 (lowest) to 1.00 (highest). It provides information on four constituent domains: socioeconomic status, household composition and disability, minority status and language, and housing type and transportation.

Defining geriatric vulnerability

Differences in patient age, frailty (measured by variations in Hospital Frailty Risk Score—a frailty index designed in 2018 in the United Kingdom based on the presence/absence of a select set of 80 three-digit ICD-10-CM codes, widely validated for use among older adults in acute care hospital settings), and extent of multimorbidity (measured by a patient’s total number of Elixhauser Comorbidity Index comorbidities) were combined into a single metric of geriatric vulnerability using a Bayesian latent variable model. A latent variable model is a statistical technique that relates a set of observable (known) variables to a set of latent (unknown) variables using a priori determined structural equations (i.e. anticipated relationships between the known and unknown variables). In this case, differences in known age, frailty, and extent of multimorbidity were used to predict unknown differences in geriatric vulnerability. Prior to inclusion in the latent variable model, all three known continuous variables were standardized (set to have mean=1, standard deviation=1), enabling them to be interpreted on the same scale and given model-determined appropriate weight.

Hospital Frailty Risk Score and Elixhauser Comorbidity Index values were calculated using ICD-10-CM primary/secondary diagnosis codes. Variations in resultant geriatric vulnerability were reported and analyzed as quintiles with “Q1” (quintile 1) representing the lowest geriatric vulnerability and “Q5” (quintile 5) representing the highest geriatric vulnerability.

Accounting for potential confounding

Potential confounders accounted for in risk-adjusted models included: gender (categorized as male-vs-female), EGS condition (16x AAST-defined EGS conditions),³⁰ EGS severity (simple-vs-complex),⁴ and presence/lack of 31 individual comorbidity indicators contained within the Elixhauser Comorbidity Index. EGS severity and Elixhauser Comorbidity Index values were calculated using ICD-10-CM primary/secondary diagnosis codes.

Statistical analysis

Associations between geriatric vulnerability, neighborhood vulnerability and differences in 30-, 365-day mortality (measured from the date of index admission to include in-hospital deaths) were first compared separately across quintiles of (1) geriatric vulnerability and (2) neighborhood vulnerability using multivariable hierarchical survival models that accounted for potential confounding and clustering of patient within hospitals, yielding hazard ratios (HR) and corresponding 95% confidence intervals (95%CI). Interactions between geriatric vulnerability and neighborhood vulnerability were then examined by repeating analyses for quintiles of geriatric vulnerability stratified among subsets of population living within: (1) neighborhoods in Florida with the highest (>80) versus the lowest (<20) national quintile of ADI, (2) neighborhoods in Florida with the highest (>0.80) versus the lowest (<0.20) national quintile of the SVI, and (3) neighborhoods in Florida with the highest (>0.80) versus the lowest (<0.20) national quintile of each of SVI’s four domains.

In an effort to examine whether potential interactions were more pronounced for historically disadvantaged populations, stratified analyses were repeated among older adults identified in the Florida SID as non-Hispanic Black or Hispanic. All data analyses were performed using Stata Statistical Software: Release 17.0 (College Station, TX). Two-sided p-values <0.05 were considered significant. The Yale Human Investigation Committee approved the study. The study was reported in accordance with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Guidelines.

Results

Association between geriatric vulnerability and mortality

A total of 448,968 older adults from 219 hospitals were included, of whom 2.2% died within 30 days (2.5% within 365 days)—numbers which were significantly influenced by baseline differences in age-related pre-hospital risk (Table 1). Compared to older adults falling within the lowest quintile of age (65–69 years), older adults within the highest quintile (86–108 years) were nearly three times as likely to die within 30 days (1.3% versus 3.8%; risk-adjusted HR[95%CI]: 2.88[2.71–3.07]) and 365 days (1.5% versus 4.2%; 2.86[2.69–3.04]). Similar associations were seen for differences in a patient’s extent of multimorbidity. Compared to older adults falling within the lowest quintile of multimorbidity (0–1 comorbidities), older adults within the highest quintile (5–11 comorbidities) were 14-times more likely to die (e.g. 30-day risk-adjusted HR[95%CI]: 14.01[12.81–15.31]). Differences in frailty (lowest quintile: 0–2 versus highest quintile: 9–29) were associated with 12-times the risk of perioperative death (e.g. 30-day risk-adjusted HR[95%CI]: 12.22[11.32–13.20]).

Table 1.

Association between (1) neighborhood vulnerability and (2) geriatric vulnerability and the risk of mortality among older adults within 30 and 365 days

Neighborhood Vulnerability
	All Older Adults
	30-day Mortality			365-day Mortality
	HR	95%CI		HR	95%CI
Area Deprivation Index
Q1 (lowest: < 20)	1.00	--		1.00	--
Q2 (20–40)	1.08	0.94	1.26	1.15	0.99	1.33
Q3 (40–60)	1.05	0.92	1.22	1.13	0.99	1.29
Q4 (60–80)	1.17	1.02	1.35	1.28	1.12	1.47
Q5 (highest: > 80)	1.24	1.07	1.44	1.32	1.15	1.52
Social Vulnerability Index
Q1 (lowest: < 0.20)	1.00	--		1.00	--
Q2 (0.20–0.40)	1.14	0.99	1.21	1.11	0.99	1.18
Q3 (0.40–0.60)	1.01	0.94	1.07	1.01	0.95	1.08
Q4 (0.60–0.80)	1.06	1.00	1.13	1.07	1.01	1.14
Q5 (highest: > 0.80)	1.08	1.01	1.57	1.09	1.03	1.16
Geriatric Vulnerability
	All Older Adults
	30-day Mortality			365-day Mortality
	HR	95%CI		HR	95%CI
Geriatric Vulnerability
Q1 (lowest)	1.00	--		1.00	--
Q2	1.36	1.19	1.54	1.49	1.33	1.66
Q3	2.80	2.50	3.13	2.65	2.39	2.93
Q4	5.46	4.91	6.07	4.90	4.45	5.38
Q5 (highest)	14.44	13.05	15.97	11.98	10.95	13.10
Contributing Factors
Hospital Frailty Risk Score	Contribution to LVM:			54.5%
Q1 (lowest)	1.00	--		1.00	--
Q2	2.02	1.80	2.27	1.76	1.59	1.96
Q3	2.41	2.20	2.64	2.12	1.95	2.30
Q4	5.02	4.62	5.45	4.06	3.78	4.35
Q5 (highest)	12.22	11.32	13.20	9.32	8.72	9.96
Number of Comorbidities	Contribution to LVM:			32.5%
Q1 (lowest)	1.00	--		1.00	--
Q2	2.04	1.83	2.26	2.01	1.84	2.21
Q3	3.95	3.58	4.34	3.46	3.17	3.77
Q4	7.00	6.36	7.69	6.15	5.65	6.69
Q5 (highest)	14.01	12.81	15.31	11.01	10.09	12.01
Age	Contribution to LVM:			14.1%
Q1 (lowest)	1.00	--		1.00	--
Q2	1.16	1.07	1.25	1.14	1.06	1.22
Q3	1.55	1.44	1.66	1.51	1.41	1.61
Q4	1.91	1.79	2.04	1.95	1.83	2.07
Q5 (highest)	2.88	2.71	3.07	2.86	2.69	3.04

Abbreviation: Q - quartile, HR - hazard ratio, 95%CI - 95% confidence interval, LVM - latent variable model

Results taken from multivariable (risk-adjusted) hierarchical survival models, accounting for clustering of patients within hospitals and risk-adjustment for known patient-level differences in gender, EGS condition, EGS severity, and presence of 31 individual comorbidities contained within the Elixhauser Comorbidity Index.

All three standardized risk-factors significantly contributed to the latent variable model (p<0.001 for each). Frailty accounted for 54.5% of the difference, extent of multimorbidity accounted for 32.5% of the difference, and age accounted for 14.1% of the difference. When implemented as a resultant single score (Table 1), differences in geriatric vulnerability yielded a larger increased risk of death than that reflected by any of the three contributing factors. Across increasing quintiles of geriatric vulnerability, patients’ risk of death within 30 days rose in a stepwise fashion from a risk-adjusted HR(95%CI) of 1.36(1.10–1.54) (Q2-vs-Q1) to 2.80(2.50–3.13) (Q3-vs-Q1) to 5.46(4.91–6.07) (Q4-vs-Q1) and 14.44(13.05–15.97) (Q5-vs-Q1). Similar results were seen for patients’ risk of death within 365 days.

Association between neighborhood vulnerability mortality

Differences in neighborhood vulnerability defined as variations in ADI and SVI yielded significant, albeit smaller, influences on older adult EGS patients’ risk of death (Table 1). Compared to older adults living within the least vulnerable neighborhoods (defined nationally as an ADI <20), older adults living in the most vulnerable neighborhoods (ADI >80) were 1.24 times as likely to die within 30 days (risk-adjusted HR[95%CI]:1.24[1.07–1.44])—a number which increased to 1.32 times within 365 days (1.09[1.03–1.16]). More modest associations were seen when comparing differences among older adults living in the least (defined nationally as an SVI <0.20) versus most (SVI >0.80) vulnerable neighborhoods based on regional variations in SVI (30-day risk-adjusted HR[95%CI]: 1.08[1.01–1.57]; 365-day: 1.09[1.03–1.16]). The geographic distribution of regional variations in ADI is presented in Figure 1.

Figure 1.

Distribution of national Area Deprivation Index (ADI) scores across Florida, 2021. Data utilized in this figure were taken from the Neighborhood Atlas run by the Center for Health Disparities Research at the University of Wisconsin School of Medicine and Public Health.

Interaction between geriatric and neighborhood vulnerability

Despite their comparatively small influence on older adults’ risk of death within 30 and 365 days, variations in neighborhood vulnerability strongly influenced the relationship between geriatric vulnerability and death. Stratified differences in patients’ risk-adjusted stepwise changes across quintiles of geriatric vulnerability are presented in Tables 2 (ADI) and 3 (SVI). Among patients living in the least vulnerable neighborhoods (ADI <20), increasing geriatric vulnerability resulted in six times greater risk of death within 30 days (risk-adjusted HR[95%CI]: 6.32[4.49–8.89]). Among patients living in the most vulnerable neighborhoods (ADI >80), the effect was more than doubled: increasing geriatric vulnerability resulted in 15-times greater risk of death within 30 days (risk-adjusted HR[95%CI]: 15.12[12.57–18.19]). By 365 days, the multiplicative interaction remained strong, yielding corresponding risk-adjusted hazard ratios of 5.61(4.04–7.79) versus 12.24(10.39–14.42).

Table 2.

Association between geriatric vulnerability and the risk of mortality among older adults within 30 and 365 days, stratified by older adults living in the least versus the most vulnerable neighborhoods (Area Deprivation Index)

Lowest Neighborhood Vulnerability (Area Deprivation Index <20)
	All Older Adults						Racial/Ethnic Minority Patients
	30-day Mortality			365-day Mortality			30-day Mortality			365-day Mortality
	HR	95%CI		HR	95%CI		HR	95%CI		HR	95%CI
Geriatric Vulnerability
Q1–2 (lowest)	1.00	--		1.00	--		1.00	--		1.00	--
Q3	0.88	0.51	1.51	0.78	0.46	1.33	2.41	0.77	7.53	2.41	0.78	7.54
Q4	1.58	1.00	2.49	1.68	1.10	2.58	7.23	2.68	19.51	7.24	2.68	19.51
Q5 (highest)	6.32	4.49	8.89	5.61	4.04	7.79	11.53	4.52	29.44	11.53	4.51	29.44
Highest Neighborhood Vulnerability (Area Deprivation Index >80)
	All Older Adults						Racial/Ethnic Minority Patients
	30-day Mortality			365-day Mortality			30-day Mortality			365-day Mortality
	HR	95%CI		HR	95%CI		HR	95%CI		HR	95%CI
Geriatric Vulnerability
Q1–2 (lowest)	1.00	--		1.00	--		1.00	--		1.00	--
Q3	3.52	2.83	4.37	3.27	2.70	3.96	8.72	4.62	16.44	7.29	4.29	12.37
Q4	5.26	4.29	6.46	4.62	3.85	5.54	12.20	6.58	22.62	10.42	6.24	17.37
Q5 (highest)	15.12	12.57	18.19	12.24	10.39	14.42	40.67	22.73	72.78	30.13	18.64	48.69

Abbreviation: Q - quartile, HR - hazard ratio, 95%CI - 95% confidence interval

Results taken from multivariable (risk-adjusted) hierarchical survival models, accounting for clustering of patients within hospitals and risk-adjustment for known patient-level differences in gender, EGS condition, EGS severity, and presence of 31 individual comorbidities contained within the Elixhauser Comorbidity Index.

Table 3.

Association between geriatric vulnerability and the risk of mortality among older adults within 30 and 365 days, stratified by older adults living in the least versus the most vulnerable neighborhoods (Social Vulnerability Index)

Lowest Neighborhood Vulnerability (Social Vulnerability Index <0.20)
	All Older Adults						Racial/Ethnic Minority Patients
	30-day Mortality			365-day Mortality			30-day Mortality			365-day Mortality
	HR	95%CI		HR	95%CI		HR	95%CI		HR	95%CI
Geriatric Vulnerability
Q1–2 (lowest)	1.00	--		1.00	--		1.00	--		1.00	--
Q3	1.27	1.01	1.59	1.09	0.90	1.33	1.55	0.76	3.22	1.55	0.74	3.22
Q4	4.54	3.83	5.38	3.60	3.11	4.18	6.17	3.54	10.76	6.00	3.68	11.69
Q5 (highest)	11.09	9.52	12.94	8.04	7.05	9.17	18.46	11.28	30.24	19.30	11.80	31.58
Highest Neighborhood Vulnerability (Social Vulnerability Index >0.80)
	All Older Adults						Racial/Ethnic Minority Patients
	30-day Mortality			365-day Mortality			30-day Mortality			365-day Mortality
	HR	95%CI		HR	95%CI		HR	95%CI		HR	95%CI
Geriatric Vulnerability
Q1–2 (lowest)	1.00	--		1.00	--		1.00	--		1.00	--
Q3	3.95	3.22	4.84	3.03	2.55	3.60	5.55	3.75	8.22	3.68	2.70	5.04
Q4	6.97	5.77	8.41	5.11	4.36	5.99	10.44	7.26	15.03	6.36	4.79	8.46
Q5 (highest)	18.08	15.19	21.52	12.81	11.09	14.80	32.16	22.88	45.18	19.64	15.18	25.41

Abbreviation: Q - quartile, HR - hazard ratio, 95%CI - 95% confidence interval

Results taken from multivariable (risk-adjusted) hierarchical survival models, accounting for clustering of patients within hospitals and risk-adjustment for known patient-level differences in gender, EGS condition, EGS severity, and presence of 31 individual comorbidities contained within the Elixhauser Comorbidity Index.

Similar, albeit less pronounced, results were seen for stratified differences in SVI, where 30-day hazard ratios for the least (SVI <0.20) versus the most (SVI >0.80) vulnerable neighborhoods were 11.09(9.52–12.94) and 18.08(15.15–21.52), respectively. When broken down by SVI domain (Figure 2), the largest contributors toward differences in 30-day hazard ratios were older adults’ tendency to live in predominately non-White, non-English-speaking neighborhoods (30-day risk-adjusted HR: 7.93-vs-16.44) and to live in low socioeconomic status neighborhoods (11.28-vs-18.19). Differences in neighborhood access to secure housing and transportation accounted for a difference of 11.24-vs-17.09, while differences in household composition were not significant.

Figure 2.

Contributions from Social Vulnerability Index (SVI) domains: Association between geriatric vulnerability (highest versus lowest quintile) and the risk of mortality among older adults within 30 days, stratified by older adults living in the least (black) versus the most (grey) vulnerable neighborhoods

Minority patients: Interaction between geriatric and neighborhood vulnerability

Repeated results restricted to older adults identified in the Florida SID as non-Hispanic Black or Hispanic are presented in Tables 2 (ADI) and 3 (SVI). Compared to a multiplicative interaction that doubled the risk of geriatric vulnerability among all older adults, when restricted to racial/ethnic minority patients, the multiplicative effect of neighborhood vulnerability on geriatric vulnerability was four-times as high. For racial/ethnic minority patients living in the least vulnerable neighborhoods (ADI <20), increasing geriatric vulnerability resulted in 11-times greater risk of death within 30 days (risk-adjusted HR[95%CI]: 11.53[4.52–29.44]). In contrast, among racial/ethnic minority patients living in the most vulnerable neighborhoods (ADI >80), increasing geriatric vulnerability resulted in 40-times greater risk of death within 30 days (risk-adjusted HR[95%CI]: 40.67[22.73–72.78]). By 365 days, the multiplicative interaction remained strong with corresponding risk-adjusted hazard ratios of 11.53(4.51–29.44) versus 30.13(18.64–48.69).

Analysis of SVI domains (Figure 2) suggested that the largest drivers of differences in racial/ethnic minority patients’ 30-day hazard ratios were older adults’ tendency to live in predominately non-White, non-English-speaking neighborhoods (risk-adjusted HR: 4.10-vs-19.46) followed by differences in neighborhood access to safe housing and transportation (12.91-vs-34.17) and neighborhood socioeconomic status (11.81-vs-25.77).

Discussion

Both geriatric and neighborhood vulnerability have been shown to affect pre-hospital risk among older EGS patients. The results of our study echo this work, revealing marked differences in 30- and 365-day mortality among older adults of more advanced age, with greater frailty, and with higher baseline extents of multimorbidity. The collective influence of these factors became more pronounced when combined into a single metric of geriatric vulnerability. Neighborhood vulnerability, in contrast, exhibited a more subtle influence on 30- and 365-day mortality. Its importance became pronounced when looking at neighborhood vulnerability’s indirect effects (i.e. its multiplicative interaction with the relationship between geriatric vulnerability and death). Stratifying by differences in neighborhood vulnerability: (1) doubled the risk of death due to geriatric vulnerability among all older EGS patients; (2) quadrupled the risk of death due to geriatric vulnerability among racial/ethnic minority EGS patients; and (3) effectively made otherwise less vulnerable older adults functionally equivalent to those who were, at baseline, more aged, more frail, and more sick.

By 2050, the number of US older adults is projected to rise to 89 million (representing 22.1% of the total US population),³² with it is expected to come a parallel increase in the number of older EGS patients. The results of our study suggest that in order to adequately account for their needs and appropriately risk-stratify older adults on EGS presentation, careful attention needs to be paid to the complex interplay of physiologic and sociodemographic factors affecting which patients present when and where for care. Subanalyses looking at differences among SVI domains indicate that the largest driver of the multiplicative effect is older adults living within historically non-White and non-English-speaking neighborhoods (additional influence was also found among older adults living in neighborhoods with lower socioeconomic status and neighborhoods with reduced access to secure housing and transportation). Such findings differ from results based on the influence of SVI alone, where average neighborhood household composition and disability followed by housing type and transportation and socioeconomic status were found to drive differences in mortality.²⁴ In their assessment, variations in minority status and language actually exhibited a small protective effect (30-day HR[95%CI]: 0.95[0.95–0.96]).²⁴ Such differences, combined with the sheer magnitude of the multiplicative interaction, underscore just how difficult it is to disentangle the array of factors underlying the existence of disparities in healthcare in the US.^33,34 Yet for all of the complexity, the results of the subanalyses make one thing painfully clear: historical practices that have resulted in long-standing neighborhood segregation (e.g. redlining in Florida with recently reported impacts on access to pharmacies³⁵ and walkability³⁶ among older adults, legalized segregation until 1967³⁷) and challenges of navigating access to care in a state with a large non-English-speaking immigrant population^38–40 continue to enact a devastating price on the lives of affected patients.

The challenge then becomes: where do we go from here? Recognizing that no perfect and few simple solutions exist, research suggests that impactful interventions aimed at addressing disparities optimally need to be implemented on a number of levels,⁴¹ including targeting individual patient-level factors (e.g. reducing barriers such as access to safe housing and food,⁴² delivering health programs that are culturally and linguistically tailored to meet the needs of specific individuals/groups⁴³) and health-system constraints (e.g. addressing issues related to variations in insurance, health-system resources and quality, and language services⁴⁴). A 2017 review by Thornton et al.⁴⁵ expands on these recommendations to specifically address interventions aimed at reducing disparities related to social determinants of health at the neighborhood level, noting the importance of: education and early childhood development, urban planning and community development, housing, income supplements, and employment interventions. A 2019 review by Marquez, Calman, and Crump⁴⁶ further contends that among US Latino populations, in particular, it will be important to accommodate heterogeneity (i.e. multiple Latino identities/ethnicities) by addressing key modifiable risk-factors, including variations in: language access, culturally-appropriate care, insurance, and social support services. Future research is warranted to see if/how interactions between geriatric and neighborhood vulnerability differ among other types of demographic factors (e.g. gender, additional racial/ethnic identities: Asian, Pacific Islander, Native Hawaiian; American Indian or Alaskan Native).

The study has limitations. The most important reflect its reliance on administrative claims, including the potential for absent or misreporting of events and limited access to clinical detail. Use of the Florida SID provided a large diverse sample of all older adult EGS patients, inclusive variations in rurality, insurance status, and neighborhood and individual patient-level social determinants of health. Few databases enable such assessment. However, in relying on the outcomes of a single state, the results might not be nationally representative. Additional research is encouraged to further assess how the observed interactions vary across geographic regions and among populations with theoretically equal access to care (e.g. patients enrolled nationally in Medicare or the Veterans Health Administration). Use of data in SID does not allow for tracking of deaths occurring across state lines or deaths that take place outside of an inpatient setting. Matching of zip codes to ADI and SVI requires aggregation up from imperfectly aligned census-block groups. Data in ADI (2015, 2020, 2021) and SVI (2016, 2018, 2020) are derived from multiple years of the American Community Survey and, as such, might not represent a perfect temporal match with the lived experience of older adults admitted in 2016–2019, 2021. Where possible, the closest year approximations were used.

The results of this study present the first in-depth look at the powerful multiplicative interaction between geriatric and neighborhood vulnerability among older EGS patients. They show that where a patient resides can fundamentally alter expected outcomes for EGS care such that otherwise less vulnerable patients become functionally equivalent to those who are, at baseline, more aged, more frail, and more sick. Ongoing efforts are needed to reduce disparities in neighborhood access to care, better risk-stratify older adults, and account for the complex interplay between physiologic and sociodemographic factors that affect which patients present when and where for care.