Long-Term Mortality After Hurricane-Related Flooding Among Skilled Nursing Facility Residents With Dementia.

Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Disaster Med Public Health Prep . Author manuscript; available in PMC: 2026 Apr 12. Published in final edited form as: Disaster Med Public Health Prep. 2025 Dec 23;19:e363. doi: 10.1017/dmp.2025.10280 Search in PMC Search in PubMed View in NLM Catalog Add to search Long-term mortality after hurricane-related flooding among skilled nursing facility residents with dementia Nathan Ashe Nathan Ashe , PhD 1. Weill Cornell Medical College, 1300 York Avenue, NY, NY, USA 10065 Find articles by Nathan Ashe 1 , Orysya Soroka Orysya Soroka , MS 2. Department of Medicine, Weill Cornell Medicine, 525 E 68th St, NY, NY, USA 10065 Find articles by Orysya Soroka 2 , Arnab K Ghosh Arnab K Ghosh , MD, MSc, MA 1. Weill Cornell Medical College, 1300 York Avenue, NY, NY, USA 10065 2. Department of Medicine, Weill Cornell Medicine, 525 E 68th St, NY, NY, USA 10065 Find articles by Arnab K Ghosh 1, 2 Author information Article notes Copyright and License information 1. Weill Cornell Medical College, 1300 York Avenue, NY, NY, USA 10065 2. Department of Medicine, Weill Cornell Medicine, 525 E 68th St, NY, NY, USA 10065 ✉ Corresponding Authors Arnab K. Ghosh MD, MSc, MA, [email protected] ; Nathan Ashe PhD, [email protected] Author Contributions NA contributed to study conception and drafted the manuscript. OS contributed to study conception, performed the statistical analysis, and assisted with manuscript writing. AK conceived the research topic and assisted with manuscript writing. Collection date 2025 Dec 23. PMC Copyright notice PMCID: PMC13070004  NIHMSID: NIHMS2158276  PMID: 41431938 The publisher's version of this article is available at Disaster Med Public Health Prep Abstract Objectives: Examine the association between dementia and all-cause five-year mortality among skilled nursing facility (SNF) residents exposed to Hurricane Sandy flooding. Methods: This study analyzed Medicare fee-for-service (FFS) beneficiaries aged ≥ 65 receiving care in SNFs located in flooded ZIP codes in New York, New Jersey, and Connecticut (October 2012). A 20% Medicare FFS sample was linked to Minimum Data Set assessments, LTCFocus, Care Compare and American Community Survey data. Flooding exposure was defined using 2012 U.S. Geological Survey flood shapefiles. Follow-up extended five years. Analysis included Kaplan-Meier curves, multivariable Cox models, and propensity-score matching. Results: Of 1,627 SNF residents, 767 (47%) had dementia. Compared with those without dementia, they were older (≥85y: 52% vs 38%; P < 0.001), less often non-Hispanic White (67% vs 75%; P < 0.001), and more frequently dually eligible for Medicare/Medicaid (63% vs 40%; P < 0.001); Charlson comorbidity burden was similar (mean 4.9; p=0.95.). Dementia was associated with higher five-year mortality after full adjustment (HR 1.20, 95% CI 1.05-1.37), and propensity-score matching (HR 1.24, 95% CI 1.08-1.44). Median survival was 1.68 years versus 2.61 years. Conclusions: SNF residents with dementia in flooded areas had higher 5-year mortality, underscoring the need for dementia-specific disaster plans. Keywords: Flooding, hurricane, skilled nursing facility, dementia INTRODUCTION: Older adults disproportionately bear the health impacts of natural disasters. Frailty, chronic illness, and limited mobility place many older individuals at heightened risk of adverse health outcomes—including cardiovascular morbidity, hospitalization—and all-cause mortality after extreme weather events (EWEs) like hurricanes; these risks can remain elevated years after the disaster. 1 , 2 , 3 Approximately three-quarters of those who died in Hurricane Katrina, for example, were over the age of 60. 4 Dementia introduces additional layers of vulnerability in disaster settings. Cognitive impairment and memory loss can prevent those with dementia from recognizing danger, following emergency instructions, or finding safety without assistance. Older adults with Alzheimer’s or related dementias, therefore, are fundamentally dependent on caregivers during a crisis. 5 Disaster conditions often disrupt routines and support systems persons with dementia rely on, as access to medications and caregiving support are interrupted, and familiar environments are upended. These disruptions increase risk of mortality, as demonstrated in a recent cohort study of older adults with dementia exposed to hurricanes, which showed an increase in mortality following hurricane landfall that peaked after 3-6 months. 5 These findings align with reports that even planned evacuations can be deadly for the cognitively impaired elderly: one analysis of nursing home evacuations during Hurricane Gustav documented a 218% increase in 30-day mortality among residents with severe dementia who were relocated compared to non-evacuated residents. 6 This convergence of cognitive, behavioral, and care-dependence factors makes dementia a potent risk multiplier in disasters. Older adults residing in institutional settings—a group which overlaps with dementia patients—constitute another especially vulnerable group in EWEs due to their institutional care needs. Skilled nursing facilities (SNF) house older adults, many with significant disability or advanced illness, who depend on complex systems of support and on-site caregivers for survival. Infrastructure failure following EWEs can have devastating consequences in these facilities, including power outages, flood damage, supply disruptions, and staff shortages. SNF residents have been shown to experience significantly higher morbidity and mortality following major hurricanes. 7 A recent review of the impact of EWEs on health outcomes of nursing home residents by Gad et al. found increased risk of mortality or hospitalization across 10 studies. 8 In one analysis of Florida facilities, exposure to Hurricane Irma was associated with 433 excess nursing home resident deaths in the 90 days after the storm. 7 Critically, however, most disaster studies in long-term care have examined all nursing home residents as a group, without specifically examining dementia as a modifying factor. 5 , 8 Hurricane Sandy’s landfall in October 2012 provides a case study of disaster-induced infrastructure collapse in a densely populated region. Sandy struck the Tri-State area (New York, New Jersey, and Connecticut) with record storm surge flooding and prolonged and extensive power outages, causing widespread devastation. Older adults bore the impact: more than half of the fatalities attributed to Hurricane Sandy were adults aged 60 and above. 9 In New York City, several nursing homes located in flood zones were inundated by seawater, which knocked out electrical systems and back-up generators. 10 Hurricane Sandy thus represents a severe test of the resilience of SNFs and caregiving systems in the face of post-hurricane infrastructure failure. 11 Given these intersection vulnerabilities—age, dementia, and residence in a SNF—it is critical to determine whether cognitive impairment compounds long-term mortality risk for SNF residents during disasters. However, there has been little research examining dementia as a modifier of disaster outcomes in institutional care environments. As a result, it remains unclear to what extent residents with dementia suffer worse post-disaster outcomes in SNFs, where baseline frailty is high but on-site support is available. We hypothesized that, following hurricane-related flooding, SNF residents with dementia would experience higher long-term mortality than residents without dementia, reflecting their greater dependence on the stability of facility-based infrastructure, such as staffing and specialized Alzheimer’s care units, and on the broader community context in which those facilities operate. This study addresses that gap by evaluating whether dementia diagnosis was associated with increased long-term mortality among short-stay SNF residents in areas heavily flooded by Hurricane Sandy. Its objective is to clarify the role of dementia as a compounding vulnerability in disasters and to inform targeted strategies to protect this high-risk subgroup in future emergencies. METHODS: Study Population, Cohort Definition, Data Sources The cohort study population included individuals aged 65 years or older who were short-stay SNF residents that experienced flooding due to Hurricane Sandy. This study was approved by the institutional review board of the authors’ home institution. The Centers for Medicare & Medicaid Services Privacy Board approved a waiver of informed consent, as research using administrative datasets cannot practicably be carried out without such a waiver. This study used Medicare fee-for-service claims for beneficiaries who met inclusion criteria. These included Medicare claims for inpatient and outpatient services, as well as the Master Beneficiary Summary File. These records were linked to Minimum Data Set (MDS) assessments, facility level-data from the Long-Term Care: Facts on Care in the US (LTCFocus) database and Medicare’s Care Compare for SNFs. SNF resident clinical characteristics were derived from MDS assessments, which are federally mandated for all SNFs. These assessments include activities of daily living (ADL) score and Cognitive Function Scale (CFS), as well as information on demographic characteristics, diagnoses, and measures of both physical and cognitive functional status. The US Census provided census tract-level covariates of general and socioeconomic status and demographic characteristics. Geographic Information System (GIS) data from the US Geological Survey (USGS) Inundation Shapefile, based on USGS field-verified High-Water Marks and Storm Surge Sensor Data, was used to determine flooded ZIP-codes. Exposure: Dementia This study defined dementia as the presence of a subtype of dementia—Alzheimer’s, vascular, Lewy body, frontotemporal, other, or mixed—in at least one inpatient or two outpatient claims, occurring at least 7 days apart during the year prior to Hurricane Sandy’s landfall, as indicated by the Medicare FFS dataset. Outcomes: All-Cause Mortality All-cause mortality was defined using the Medicare Beneficiary Summary File as any deaths up to an approximate 5-year timeframe since hurricane exposure Other Independent Variables Individual demographic factors included age, sex, race and ethnicity, comorbidities, Medicare/Medicaid dual eligibility, total days spent in a SNF during the year prior to Hurricane Sandy, and number of hospitalizations during the year prior to Hurricane Sandy. Facility level characteristics included total number of beds, indicators of for-profit status, indicators of the facility being part of a chain, indicators of whether or not the facility is hospital-based, indicators of the presence of an Alzheimer’s disease Special Care Unit (SCU), proportion of residents present on the 1 st Thursday in April with a Cognitive Function Scale (CFS) greater than or equal to 5, the average Activities of Daily Living (ADL) score for all residents present on the 1 st Thursday in April, proportion of facility residents whose primary support is Medicare, number of occupied beds divided by the total number of beds, direct-care staff hours per resident day, and proportion of residents admitted during the calendar year who were White. ZIP-code level characteristics included percent of residents with a college education or higher, and percent of residents living below the federal poverty line. Statistical Analyses Unadjusted comparisons based on whether a subject had or did not have dementia were made using t test or Wilcoxon rank sum test for continuous variables and the χ 2 test for categorical variables. This study used Kaplan-Meier plots to visually inspect the cumulative risk of death for each group. Risk of mortality was calculated using a Cox proportional hazard model. This study adopted a stepwise strategy of adding variables to this model. With the Variance Inflation Factor (VIF), this study assessed if the covariates were collinear using a cut-off of 3. Participants were censored at the end of their Medicare Parts A, B coverage or Dec. 31, 2017 (whichever came first). This study chose 5 years after the hurricane’s landfall to examine 5-year mortality risk because existing long-term mortality studies demonstrate impacts up to 2 years after hurricane landfall, and have not measured risk beyond this range. Finally, this study tested the assumption of proportionality of all Cox models using the scaled Schoenfeld residuals approach. To address potential unobserved bias in modeling strategy, this study used propensity scoring, a method that reduces confounding by matching subjects on a number of variables relevant to both exposure and outcome. Informed by an integrative framework that links hurricane-related flooding to long-term health via individual demographics, clinical status, pre-event healthcare utilization, facility resources, and neighborhood socioeconomic context—adapted from Waddell et al .’s hurricane-health conceptual model and further shaped by post-acute dementia-care utilization findings and nursing-home disaster studies 7 , 11 , 12 , 13 —this study included the following covariates in its propensity scoring: age, sex, race/ethnicity, dual eligibility, Charlson Comorbidity Index score, total days spent in a SNF during the year prior to Hurricane Sandy, number of hospitalizations during the year prior to Hurricane Sandy, SNF ownership type, facility size (total beds), direct-care staff hours per resident day, and the median income, percent of residents with bachelor’s degree or higher, and percent of residents below the Federal poverty level of the ZIP code in which the subject resided. This study matched one-to-one, accounting for clustering within matched pairs, at a caliper of 0.5 and a caliper of 0.2. After the matching process, this study then derived both a crude and fully adjusted hazard ratio for each caliper. This study also did a subgroup analysis for SNFs with Alzheimer’s dementia SCUs to test the hypothesis that SNFs with SCUs may provide better care in the post hurricane period compared to non-SCU SNFs. This study performed all analyses using SAS statistical software, version 9.4 (SAS Institute, Cary, NC) and Stata, version 18.0 (StataCorp, College Station, TX). RESULTS: Unadjusted Comparisons Table 1 presents descriptive statistics for SNF residents and SNFs in flooded ZIP-codes. The study population included 1627 individuals (mean [SD] age, 82.35 [8.00]). Compared to patients without dementia, patients with dementia were older (proportion aged ≥ 85 years: 399 [52%] vs 324 [37.7%]; P < 0.001), were less likely to identify as non-Hispanic white (513 [66.9%] vs 647 [75.2%]; P < 0.001), and were more likely to be dually eligible for Medicare and Medicaid (479 [62.5%] vs 347 [40.3%]; P < 0.001). There was no significant difference in the mean or median Charlson comorbidity values between the two groups (mean [SD]: 4.9 [3.0] vs 4.9 [2.6]; P = 0.95; median [IQR]: 5.0 [3.0, 6.0] vs 5.0 [3.0, 7.0]; P = 0.59). While the median number of hospital stays during the baseline year was not significant, the median number of days at an SNF during the baseline year was (median [IQR]: 35.0 [5.0, 81.0] vs 13.0 [0.0, 43.0]). Table 1: Characteristics of the Participants Factor Level Number Dementia-Free at Hurricane Landfall Dementia at Hurricane Landfall p-value Individual Characteristics Total 1627 860 767 Male 582 (35.8%) 323 (37.6%) 259 (33.8%) 0.11 Age 65-69 126 (7.7%) 98 (11.4%) 28 (3.7%) <0.001 70-74 185 (11.4%) 121 (14.1%) 64 (8.3%) 75-79 253 (15.6%) 141 (16.4%) 112 (14.6%) 80-84 340 (20.9%) 176 (20.5%) 164 (21.4%) 85+ 723 (44.4%) 324 (37.7%) 399 (52.0%) Research Triangle Institute (RTI) Race Code Unknown 4 (0.2%) 3 (0.3%) 1 (0.1%) <0.001 Non-Hispanic White 1160 (71.3%) 647 (75.2%) 513 (66.9%) Black 260 (16.0%) 130 (15.1%) 130 (16.9%) Other 9 (0.6%) 5 (0.6%) 4 (0.5%) Asian or Pacific Islander 39 (2.4%) 22 (2.6%) 17 (2.2%) Hispanic 154 (9.5%) 52 (6.0%) 102 (13.3%) American Indian or Alaska Native 1 (0.1%) 1 (0.1%) 0 (0.0%) Non-Hispanic White 1160 (71.3%) 647 (75.2%) 513 (66.9%) <0.001 Medicaid Dual Eligibility 826 (50.8%) 347 (40.3%) 479 (62.5%) <0.001 Comorbidities Charlson-Deyo score, mean (SD) 4.9 (2.9) 4.9 (3.0) 4.9 (2.6) 0.95 Charlson-Deyo score, median (IQR) 5.0 (3.0, 7.0) 5.0 (3.0, 7.0) 5.0 (3.0, 6.0) 0.59 Congestive heart failure 859 (52.8%) 453 (52.7%) 406 (52.9%) 0.92 Myocardial infarction 298 (18.3%) 174 (20.2%) 124 (16.2%) 0.034 Peripheral vascular disease 482 (29.6%) 272 (31.6%) 210 (27.4%) 0.061 AIDS 8 (0.5%) 4 (0.5%) 4 (0.5%) 0.87 Any malignancy 318 (19.5%) 177 (20.6%) 141 (18.4%) 0.26 Prior cerebrovascular event 796 (48.9%) 388 (45.1%) 408 (53.2%) 0.001 Chronic pulmonary disease 728 (44.7%) 404 (47.0%) 324 (42.2%) 0.055 Dementia 638 (39.2%) 63 (7.3%) 575 (75.0%) <0.001 Diabetes 711 (43.7%) 361 (42.0%) 350 (45.6%) 0.14 Diabetes with complications 189 (11.6%) 116 (13.5%) 73 (9.5%) 0.013 Hemi- or paraplegia 50 (3.1%) 25 (2.9%) 25 (3.3%) 0.68 Metastatic solid tumor 108 (6.6%) 82 (9.5%) 26 (3.4%) <0.001 Liver disease, mild 14 (0.9%) 8 (0.9%) 6 (0.8%) 0.75 Liver disease, moderate or severe 34 (2.1%) 20 (2.3%) 14 (1.8%) 0.48 Peptic ulcer disease 94 (5.8%) 54 (6.3%) 40 (5.2%) 0.36 Renal disease 611 (37.6%) 348 (40.5%) 263 (34.3%) 0.010 Rheumatologic disease 162 (10.0%) 103 (12.0%) 59 (7.7%) 0.004 Number of hospital stays during baseline year, mean (SD) 2.7 (2.5) 2.8 (2.6) 2.7 (2.4) 0.71 Number of hospital stays during baseline year, median (IQR) 2.0 (1.0, 4.0) 2.0 (1.0, 4.0) 2.0 (1.0, 4.0) 0.81 Number of days at SNF during baseline year, mean (SD) 49.9 (78.7) 35.2 (63.0) 66.4 (90.5) <0.001 Number of days at SNF during baseline year, median (IQR) 20.0 (0.0, 61.0) 13.0 (0.0, 43.0) 35.0 (5.0, 81.0) <0.001 Facility Characteristics Total number of beds, mean (SD) 209.4 (126.7) 206.2 (129.8) 213.0 (123.1) 0.28 Total number of beds, median (IQR) 180 (124, 246) 180 (120, 246) 181 (128, 242.5) 0.054 Facility for-profit Yes 1135 (70.1%) 596 (69.6%) 539 (70.5%) 0.69 Proportion of residents present on the 1st Thursday in April with a Cognitive Function Scale (CFS) score of 4 (severe cognitive impairment), mean (SD) 18.0 (8.3) 17.9 (8.6) 18.1 (7.9) 0.68 Proportion of residents present on the 1st Thursday in April with a Cognitive Function Scale (CFS) score of 4 (severe cognitive impairment), median (IQR) 17.5 (12.5, 22.6) 17.4 (12.5, 22.9) 17.5 (12.7, 22.6) 0.82 Proportion of facility residents whose primary support is Medicare, mean (SD) 20.8 (16.5) 23.1 (18.2) 18.2 (14.0) <0.001 Proportion of facility residents whose primary support is Medicare, median (IQR) 15.7 (10.8, 23.7) 16.8 (11.5, 27.1) 14.3 (10.1, 20.9) <0.001 Number of occupied beds in facility divided by the total number of beds, mean (SD) 90.0 (9.2) 89.1 (10.0) 90.9 (8.0) <0.001 Number of occupied beds in facility divided by the total number of beds, median (IQR) 92.2 (86.7, 96.3) 91.4 (85.5, 96.1) 92.5 (88.3, 96.5) <0.001 Facility part of a chain Yes 462 (28.5%) 250 (29.2%) 212 (27.7%) 0.52 Facility hospital-based Yes 42 (2.6%) 28 (3.3%) 14 (1.8%) 0.069 Direct-care staff hours per resident day, mean (SD) 3.6 (0.8) 3.7 (0.8) 3.5 (0.7) <0.001 Direct-care staff hours per resident day, median (IQR) 3.5 (3.2, 3.9) 3.5 (3.3, 4.0) 3.4 (3.2, 3.8) <0.001 The average Activities of Daily Living (ADL) score for all residents present on the 1st Thursday in April, mean (SD) 17.9 (2.2) 18.1 (2.1) 17.8 (2.2) 0.005 The average Activities of Daily Living (ADL) score for all residents present on the 1st Thursday in April, median (IQR) 18.0 (16.6, 19.5) 18.3 (16.8, 19.6) 17.8 (16.6, 19.3) 0.008 Facility has an Alzheimer's disease Special Care Unit (SCU) No 1472 (90.9%) 786 (91.8%) 686 (89.8%) 0.16 Yes 148 (9.1%) 70 (8.2%) 78 (10.2%) Proportion of residents admitted during the calendar year who were White, mean (SD) 70.1 (26.6) 72.0 (25.6) 68.0 (27.5) 0.002 Proportion of residents admitted during the calendar year who were White, median (IQR) 79.4 (56.3, 92.0) 81.5 (59.3, 92.2) 77.4 (49.5, 91.3) 0.010 ZIP-Code Characteristics Median income, ACS-2012-5yrs, mean (SD) $65,164.6 ($24,823.1) $65,960.1 ($25,674.3) $64,272.6 ($23,817.7) 0.17 Median income, ACS-2012-5yrs, median (IQR) $62,191 ($46,325, $77,853) $63,352 ($48,236, $79,238) $61,213 ($46,001, $77,506) 0.28 Percent bachelor's degree or higher, mean (SD) 33.9 (17.3) 34.2 (17.4) 33.5 (17.2) 0.40 Percent bachelor's degree or higher, median (IQR) 28.5 (21.8, 42.5) 29.2 (22.0, 42.6) 27.9 (21.6, 41.2) 0.24 % under Federal Poverty Line, ACS-2012-5yrs, mean (SD) 13.3 (8.9) 13.0 (8.9) 13.6 (8.8) 0.24 % under Federal Poverty Line, ACS-2012-5yrs, median (IQR) 10.7 (6.4, 18.1) 10.4 (6.3, 17.2) 11.6 (6.6, 18.4) 0.095 Open in a new tab Residents with dementia were more likely to be in facilities with a lower proportion of residents whose primary support was Medicare (mean [SD]: 18.2 [14.0] vs 23.1 [18.2]; P < 0.001; median [IQR]: 14.3 [10.1, 20.9] vs 16.8 [11.5, 27.1]), a lower ratio of direct-care staff hours per resident day (mean [SD]: 3.5 [0.7] vs 3.7 [0.8]; P <0.001; median [IQR]: 3.4 [3.2, 3.8] vs 3.5 [3.2, 3.9]), and a lower proportion of residents admitted during the calendar year who were white, though only the mean was significant (mean [SD]: 68.0 [27.5] vs 72.0 [25.6]; P = 0.002; median [IQR]: 77.4 [49.5, 91.3] vs 81.5 [59.3, 92.2]; P = 0.010). Adjusted Comparisons Table 2 presents hazard ratios for individuals with and without dementia who resided in a SNF in a flooded ZIP-code during Hurricane Sandy, where the outcome is all-cause mortality. The crude (model 0) hazard ratio was 1.39 (95% CI: 1.24-1.56). Model 1 show adjustments for age and sex (HR [95% CI]: 1.31 [1.17, 1.47]), model 2 adds comorbidity as measured by the Charlson comorbidity index, race and ethnicity, and dual Medicaid status (HR [95% CI]: 1.27 [1.13, 1.43]), model 3 adds facility-level characteristics (HR [95% CI]: 1.20 [1.06, 1.37]), model 4 adds ZIP-code level characteristics (HR [95% CI]: 1.20 [1.06, 1.37]), and model 5 adds total days spent in an SNF during the year prior to Hurricane Sandy and the number of hospitalizations during the year prior to the hurricane (HR [95% CI]: 1.20 [1.05, 1.37]). Table 2: Hazard ratios for all-cause mortality among individuals with and without dementia who resided in an SNF in a flooded ZIP-code during Hurricane Sandy HR 95% CI Model 0: Crude (n=1627) 1.39 1.24 1.56 Model 1: Age-sex adjusted (n=1627) 1.31 1.17 1.47 Model 2: Model 1 + Charlson, Medicaid, race (n=1627) 1.27 1.13 1.43 Model 3: Model 2 + hospital variables (n=1304) 1.20 1.06 1.37 Model 4: Model 3 + percent bachelor or higher, percent below the poverty line (n=1304) 1.20 1.06 1.37 Open in a new tab Table 3 shows hazard ratios after one-to-one propensity score matching. At a caliper of 0.5 (n=1382) a crude hazard ratio without covariates was calculated (HR [95% CI]: 1.28 [1.14, 1.45]) as well as a model adjusted with the covariates included in model 5 above (HR [95% CI]: 1.21 [1.05, 1.38]). At a caliper of 0.2 (n=1210), a crude hazard ratio (HR [95% CI]: 1.26 [1.11, 1.45]) and an adjusted hazard ratio were again calculated (HR [95% CI]: 1.24 [1.07, 1.44]). Table 3: Hazard ratios after 1-to-1 greedy propensity score matching, accounting for clustering within matched pairs HR 95% CI HR Caliper 0.5 (n=1382) Crude (n=1382) 1.28 1.14 1.45 Fully adjusted (n=1131) 1.20 1.05 1.38 Caliper 0.2 (n=1210) Crude (n=1210) 1.26 1.11 1.44 Fully adjusted (n=987) 1.24 1.08 1.44 Open in a new tab Subgroup Analysis A subgroup analysis of patients with dementia diagnosis at Hurricane Sandy landfall (n=767) showed Alzheimer’s SCU residents (n=78) had a lower unadjusted one-year survival (HR [95% CI]: 1.19 [0.93, 1.54]), however this finding was not statistically significant (p=0.11) (see analysis in Supplement Figure S2 ). LIMITATIONS: This study has five important limitations. First, because this study restricted its analysis to SNFs located in ZIP codes that experienced flooding after Hurricane Sandy made landfall, comparing dementia patients to non-dementia patients, this study did not include a contemporaneous group of residents in unflooded facilities. Thus, the effect size noted (a ~20% increased risk of adjusted mortality) likely is both a combination of mortality risk jointly from flooding exposure and the dementia diagnosis. Nonetheless, despite this, we argue the findings are significant because they highlight the unique vulnerabilities of individuals with dementia in post-hurricane settings. Second, Medicare claims provide only ZIP-code level data, so this study could not verify flooding at the level of individual facility or resident. This spatial imprecision likely biases effect estimates toward the null but cannot be ruled out as a source of misclassification. Third, this study was unable to stratify by dementia subtype (e.g., Alzheimer’s disease, vascular dementia, Lewy body dementia). Prior studies show that survival trajectories differ across subtypes, so important heterogeneity may be masked. 14 Because the MDS includes instruments such as the CFS, future work could stratify by dementia severity to assess whether mortality risk is concentrated in those with advanced impairment. Fourth, evacuation status was not measured, and differences in decisions to evacuate or shelter in place—as well as the success of evacuation efforts—may have differentially affected residents with dementia and contributed to the observed mortality gap. We recognize this as an important area of future research given ambivalence in the literature as to the effects of sheltering-in-place vs evacuation and the long-term health outcomes associated with these strategies for nursing home residents. 13 , 15 Finally, as a retrospective observational study, residual confounding from unmeasured or incompletely measured variables may persist despite multivariable adjustment and propensity-score matching. DISCUSSION: In this retrospective cohort of 1627 SNF residents whose facilities flooded during Hurricane Sandy, dementia was associated with a 20% increase in the hazard of all-cause mortality over five years (fully-adjusted HR 1.20, 95% CI 1.05-1.37). Kaplan-Meier curves ( Figure 1 ) showed that survival in the dementia group diverged early and remained lower throughout follow-up, yielding a median survival of 1.68 years among residents with dementia versus 2.61 years among residents without dementia. Applying the fully-adjusted HR of 1.20 and the 47% prevalence of dementia, this study estimates that ~8.6% of all deaths (≈104 of 1205) were attributable to dementia beyond the baseline risk conferred by age, comorbidity, and facility and neighborhood factors. These data indicate that the excess risk attributable to dementia is not confined to the acute post-disaster period, but persists for years. Figure 1: Open in a new tab Kaplan-Meier curve comparing SNF residents with and without dementia within 5 years of Hurricane Sandy landfall This study’s five-year adjusted hazard ratio is directionally consistent with two prior disaster studies that used different effect measures and comparison groups. Bell et al. examined >346,000 Medicare beneficiaries with and without pre-existing dementia who resided in FEMA-declared disaster counties during Hurricanes Harvey, Irma, or Florence, tracing monthly all-cause mortality 12 months before landfall through 12 months after, and calculating attributable risks, attributable fraction, and relative risks (RRs). 5 They found an overall 12-month RR of 1.08 (95% CI 1.07-1.09); hurricane specific RRs were 1.12 for Harvey, 1.07 for Irma, and 1.09 for Florence. Dosa et al. focused on all Florida SNF residents (irrespective of dementia diagnosis) and compared post-Irma mortality with outcomes for residents of the same facilities over the same time period two years earlier. 7 They found 30-day and 90-day risk ratios of 1.12 and 1.07, respectively—i.e., a 12% and 7% increase in short-term death risk attributable to hurricane exposure when the baseline is a pre-event year in identical facilities. SNF preparedness offers several concrete pathways by which dementia could translate into sustained excess mortality after a post-hurricane flood. First, residents with dementia rely on stable, round-the-clock caregiving routines; when a SNF loses power or staff during a storm, behavioral cues, medication timing, and fall-prevention protocols break down, leading to delirium, injury, or accelerated functional decline. 6 Second, environmental hazards that arise inside an under-prepared facility can act as chronic stressors: failed heating, ventilation, and air conditioning (HVAC) systems elevate indoor temperatures, damp walls foster mold growth, and disrupted food supply chains alter diet quality—each of which can worsen cardiopulmonary disease and potentially accelerate neurodegeneration in cognitively impaired residents. 16 , 17 , 18 , 19 , 20 , 21 Third, observational data show that injury-related hospitalizations carry higher case-fatality among dementia patients. 22 Finally, post-disaster service gaps persist well beyond the initial emergency; for example, NYU Langone remained partially closed for 59 days after Hurricane Sandy landfall, 23 and some outpatient services in flooded ZIP codes were disrupted for months, prolonging barriers to medical follow-up. 11 In short, the degree to which a SNF’s physical infrastructure, staffing, and supply networks are disaster-ready may modify the impact and duration of dementia as a risk factor for increased all-cause mortality. Strengthening these facility-level preparedness domains is therefore essential to mitigating long-term disaster risk in this high-vulnerability population. Although this study adjusted for direct-care staff hours per resident day, the evidence linking staffing levels or staff training to mortality outcomes in disaster settings remains limited. Most disaster studies in skilled nursing care focus on evacuation rather than personnel factors. 24 Nonetheless, consistent staffing and dementia-specific training are likely critical for maintaining routines, preventing delirium, and ensuring timely care during crises—mechanisms that warrant further investigation in future research. This study hypothesized that a dedicated SCU could buffer these risks by concentrating trained staff and dementia-friendly design features. Its analysis showed a lower unadjusted one-year survival for SCU residents, however this result was not significant; larger datasets are needed to clarify whether specialized dementia units mitigate or merely select for higher risk. Key strengths of this study include linkage of Medicare claims, MDS assessments, facility-level characteristics, and GIS-based flood exposure; a five-year horizon, longer than most disaster-mortality studies; and rigorous adjustment plus propensity-score matching that yielded consistent estimates. CONCLUSION: This study’s five-year follow-up of 1,627 storm-surge-flooded SNF residents shows that dementia is not just an acute-phase vulnerability but a persistent amplifier of mortality: residents with dementia faced a 20% higher death hazard and survived nearly a year less than their cognitively intact peers. This long-tail risk underscores the need to move toward research-informed solutions. Future work should do four things. First, disentangle absolute flood effects by pairing flooded and non-flooded facilities within the same event and tracking recovery metrics such as power, water, and staffing restoration. Second, test facility-level interventions—backup generators sized for HVAC, dementia-trained surge staff, and resilient unit layouts—through prospective studies, as well as study post-disaster strategies to mitigate harm to individuals with dementia. Third, stratify analyses by dementia subtype and severity to pinpoint residents who would benefit most from targeted protections. And fourth, integrate healthcare-utilization variables and neighborhood social-vulnerability indices into disaster-risk models to refine triage and resource allocation. Rigorous work along these lines may lay the groundwork for interventions that more effectively safeguard cognitively impaired residents during flood events. Supplementary Material Supplement NIHMS2158276-supplement-Supplement.docx (68.5KB, docx) Funding sources This work was supported with funding from National Heart, Lung, and Blood Institute (K08HL163329 Ghosh), and National Center for Advancing Clinical Translational Science (R03TR004976 Ghosh). Abbreviations ADL Activities of Daily Living CFS Cognitive Function Scale EWE Extreme Weather Events FFS Medicare Fee-For-Service GIS Geographic Information System HVAC Heating, Ventilation, and Air Conditioning LTCFocus Long-Term Care: Facts on Care in the US MDS Minimum Data Set SCU Special Care Unit SNF Skilled Nursing Facility USGS US Geological Survey VIF Variance Inflation Factor Footnotes CONFLICTS OF INTEREST: The authors have no conflicts of interest. REFERENCES: 1. Adams V, Kaufman Sharon R., van Hattum Taslim, and Moody S. Aging Disaster: Mortality, Vulnerability, and Long-Term Recovery among Katrina Survivors. Med Anthropol. 2011;30(3):247–270. doi: 10.1080/01459740.2011.560777 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Malik S, Lee DC, Doran KM, et al. Vulnerability of Older Adults in Disasters: Emergency Department Utilization by Geriatric Patients After Hurricane Sandy. Disaster Med Public Health Prep. 2018;12(2):184–193. doi: 10.1017/dmp.2017.44 [ DOI ] [ PubMed ] [ Google Scholar ] 3. Gautam S, Menachem J, Srivastav SK, Delafontaine P, Irimpen A. Effect of Hurricane Katrina on the Incidence of Acute Coronary Syndrome at a Primary Angioplasty Center in New Orleans. Disaster Med Public Health Prep. 2009;3(3):144–150. doi: 10.1097/DMP.0b013e3181b9db91 [ DOI ] [ PubMed ] [ Google Scholar ] 4. Wilson N. Hurricane Katrina: Unequal Opportunity Disaster. Public Policy Aging Rep. 2006;16(2):8–13. doi: 10.1093/ppar/16.2.8 [ DOI ] [ Google Scholar ] 5. Bell SA, Miranda ML, Bynum JPW, Davis MA. Mortality After Exposure to a Hurricane Among Older Adults Living With Dementia. JAMA Netw Open. 2023;6(3):e232043. doi: 10.1001/jamanetworkopen.2023.2043 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Brown LM, Dosa DM, Thomas K, Hyer K, Feng Z, Mor V. The Effects of Evacuation on Nursing Home Residents With Dementia. Am J Alzheimers Dis Dementias ® . 2012;27(6):406–412. doi: 10.1177/1533317512454709 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Dosa DM, Skarha J, Peterson LJ, et al. Association Between Exposure to Hurricane Irma and Mortality and Hospitalization in Florida Nursing Home Residents. JAMA Netw Open. 2020;3(10):e2019460. doi: 10.1001/jamanetworkopen.2020.19460 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Gad L, Keenan OJ, Ancker JS, et al. Impact of Extreme Weather Events on Health Outcomes of Nursing Home Residents Receiving Post-Acute Care and Long-Term Care: A Scoping Review. J Am Med Dir Assoc. 2024;25(11):105230. doi: 10.1016/j.jamda.2024.105230 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Kleier JA, Krause D, Ogilby T. Hurricane preparedness among elderly residents in South Florida. Public Health Nurs. 2018;35(1):3–9. doi: 10.1111/phn.12344 [ DOI ] [ PubMed ] [ Google Scholar ] 10. Powell M, Fink S. Nursing Home Is Faulted Over Care After Storm. The New York Times. https://www.nytimes.com/2012/11/10/nyregion/queens-nursing-home-is-faulted-over-care-after-storm.html . November 10, 2012. Accessed April 20, 2025. [ Google Scholar ] 11. Lee S, Jayaweera DT, Mirsaeidi M, Beier JC, Kumar N. Perspectives on the Health Effects of Hurricanes: A Review and Challenges. Int J Environ Res Public Health. 2021;18(5):2756. doi: 10.3390/ijerph18052756 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Burke RE, Xu Y, Ritter AZ. Outcomes of post-acute care in skilled nursing facilities in Medicare beneficiaries with and without a diagnosis of dementia. J Am Geriatr Soc. 2021;69(10):2899–2907. doi: 10.1111/jgs.17321 [ DOI ] [ PubMed ] [ Google Scholar ] 13. Hua CL, Patel S, Thomas KS, et al. Evacuation and Health Care Outcomes Among Assisted Living Residents After Hurricane Irma. JAMA Netw Open. 2024;7(4):e248572. doi: 10.1001/jamanetworkopen.2024.8572 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Ono R, Sakurai T, Sugimoto T, et al. Mortality Risks and Causes of Death by Dementia Types in a Japanese Cohort with Dementia: NCGG-STORIES. J Alzheimer’s Dis. 2023;92(2):487–498. doi: 10.3233/JAD-221290 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Dosa D, Hyer K, Thomas K, et al. To Evacuate or Shelter in Place: Implications of Universal Hurricane Evacuation Policies on Nursing Home Residents. J Am Med Dir Assoc. 2012;13(2):190.e1–190.e7. doi: 10.1016/j.jamda.2011.07.011 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Hasegawa K, Yoshino H, Yanagi U, et al. Indoor environmental problems and health status in water-damaged homes due to tsunami disaster in Japan. Build Environ. 2015;93:24–34. doi: 10.1016/j.buildenv.2015.02.040 [ DOI ] [ Google Scholar ] 17. Ban J, Sutton C, Ma Y, Lin C, Chen K. Association of flooding exposure with cause-specific mortality in North Carolina, United States. Nat Water. 2023;1(12):1027–1034. doi: 10.1038/s44221-023-00167-5 [ DOI ] [ Google Scholar ] 18. Mendell MJ, Mirer AG, Cheung K, Tong M, Douwes J. Respiratory and Allergic Health Effects of Dampness, Mold, and Dampness-Related Agents: A Review of the Epidemiologic Evidence. Environ Health Perspect. 2011;119(6):748–756. doi: 10.1289/ehp.1002410 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Wang K, Liu Shoujun, Zhang Xufeng, and Sun D. Toxic Effect of Mycotoxins on Cardiovascular System: A Topic Worthy of Further Study. Food Rev Int. 2023;39(4):2203–2211. doi: 10.1080/87559129.2021.1950172 [ DOI ] [ Google Scholar ] 20. Vasefi M, Ghaboolian-Zare E, Abedelwahab H, Osu A. Environmental toxins and Alzheimer’s disease progression. Neurochem Int. 2020;141:104852. doi: 10.1016/j.neuint.2020.104852 [ DOI ] [ PubMed ] [ Google Scholar ] 21. Chu L, Warren JL, Spatz ES, et al. Floods and cause-specific mortality in the United States applying a triply robust approach. Nat Commun. 2025;16(1):2853. doi: 10.1038/s41467-025-58236-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Harvey L, Mitchell R, Brodaty H, Draper B, Close J. The influence of dementia on injury-related hospitalisations and outcomes in older adults. Injury. 2016;47(1):226–234. doi: 10.1016/j.injury.2015.09.021 [ DOI ] [ PubMed ] [ Google Scholar ] 23. •. NYU Langone Medical Center Reopens 2 Months After Sandy. NBC New York. December 27, 2012. Accessed May 15, 2025. https://www.nbcnewyork.com/news/local/nyu-langone-medical-center-sandy-evacuated-patients-federal-aid-power/2095324/ [ Google Scholar ] 24. Gad L, Keenan OJ, Ancker JS, et al. Impact of Extreme Weather Events on Health Outcomes of Nursing Home Residents Receiving Post-Acute Care and Long-Term Care: A Scoping Review. J Am Med Dir Assoc. 2024;25(11):105230. doi: 10.1016/j.jamda.2024.105230 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplement NIHMS2158276-supplement-Supplement.docx (68.5KB, docx) ACTIONS View on publisher site PDF (460.9 KB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top