Impact of Extreme Weather Events on Health Outcomes of Nursing Home Residents Receiving Post-Acute Care and Long-Term Care: A Scoping Review

Abstract

Objectives

To systematically examine the evidence of the association between extreme weather events (EWEs) and adverse health outcomes among short-stay patients undergoing post-acute care (PAC) and long-stay residents in nursing homes (NH).

Design

This is a scoping review. The findings were reported using the Preferred Reporting Items for Systematic Review and Meta-Analysis Extension for Scoping Reviews checklist.

Settings and Participants

Studies published on short-stay PAC and long-stay residents in NHs.

Methods

A literature search was performed in 6 databases. Studies retrieved were screened for eligibility against predefined inclusion/exclusion criteria. Studies were qualitatively synthesized based on the EWE, health outcomes, and special populations studied.

Results

Of the 5044 studies reviewed, ten met our inclusion criteria. All were retrospective cohort studies. Nine studies examined the association between hurricane exposure, defined inconsistently across studies, and PAC patients and long-stay residents in the NH setting in the Southern United States; the other study focused on post-flood risk among North Dakota NH residents. Nine studies focused on long-stay NH residents receiving custodial care, and one focused on patients receiving PAC. Outcomes examined were unplanned hospitalization rates and mortality rates within 30 and 90 days and changes in cognitive impairment. Nine studies consistently found an association between hurricane exposure and increased risk of 30- and 90-day mortality compared to unexposed residents.

Conclusions/Implications

Of the EWEs examined, hurricanes are associated with an increased risk of mortality among long-stay NH residents and those admitted to hospice, and with increased risk of hospitalization for short-stay PAC patients. As the threat of climate-amplified EWEs increases, future studies of NH residents should evaluate the impact of all types of EWEs, and not solely hurricanes, across wider geographic regions, and include longer term health outcomes, associated costs, and analyses of potential disparities associated with vulnerable populations in NHs.

Summary:

Extreme Weather Events are a risk to human health, with nursing home (NH) residents at higher risk. EWEs, like hurricanes, are associated with increased risk of mortality in long-term NH residents.

Introduction

Extreme weather events (EWEs), defined as hurricanes, floods, tornadoes, and wildfires,¹ have been associated with increased risk of adverse health outcomes.² NH residents include those receiving post-acute care (PAC) with the goal of returning to the community and long-stay residents receiving custodial care. Residents of these facilities are particularly vulnerable to EWEs.^3–5 They disproportionately have higher rates of functional and cognitive impairment, and chronic comorbidities.⁶ During EWEs, they often rely on the facility’s preparedness measures and evacuation protocols to protect against harm during EWEs; ^7,8 in the United States, the Centers of Medicare and Medicaid Services require NHs to have disaster preparedness plans. Moreover, from the immediate aftermath into the longer term, the disruptions to residents’ care and place of residence are likely to increase the risk of adverse health outcomes, such as unplanned hospitalizations and mortality, for prolonged periods.²

There is a growing literature that describes the impacts of EWEs on NH residents.⁹ In the United States, this literature has historically focused on hurricanes, and has also systemically evaluated mortality risk after emergency evacuations from NH.^10,11 However, EWEs of various types have increased in both their geographic reach and their frequency, as evidenced by the latest figures from the National Oceanic and Atmospheric Agency for 2023.¹² As a result, adverse health outcomes in the NH population may vary based on the type of EWE. Therefore, our objective was to systematically examine the evidence of the association between EWEs and adverse health outcomes among NH residents to date.

Methods

This study was performed following the PRISMA Extension for Scoping Reviews (PRISMA-ScR).¹³ A protocol was registered and made publicly available with Open Science Framework (OSF: osf.io/7gsj2).¹⁴

A medical librarian (MRD) performed searches to identify studies that addressed evacuation from EWE threats and health outcomes in the NH population.

Searches were run on March 1, 2024, in the following databases: Ovid MEDLINE (ALL - 1946 to Present); Ovid EMBASE (1974 to present); CINAHL (EBSCO); The Cochrane Library (Wiley); Web of Science – Core Collection (Clarivate); and Scopus (Elsevier). The search strategy included all appropriate controlled vocabulary and keywords for the concepts of “nursing homes” and “extreme weather.” The full search strategies for all databases are available in the Supplement. There were no language, publication date, or article type restrictions on the search strategy.

Retrieved studies were screened for inclusion using Covidence systematic review software.¹⁵ Titles and abstracts were reviewed against predefined inclusion/exclusion criteria by three independent reviewers (LG, OJK, AKG). Discrepancies were resolved by consensus. For final inclusion, full text was then retrieved and screened by the same three reviewers. Articles considered for inclusion were the following: population must reside in NHs (under 24-hour supervision), empirical data studies, published in the last 20 years (January 1, 2004- March 1), and focused on dates of post-EWE hospitalizations, health outcomes, and mortality rates. Studies were excluded due to the following: the study populations were living in assisted living centers where residents are not under 24-hour supervision, non-English manuscripts, articles being reviews or perspective pieces, being based on a single case study, timeframe following EWE for hospitalization and mortality not defined, or did not examine a health outcome. For articles selected for inclusion in this study, reference lists and citing articles were pulled from Scopus (Elsevier) and screened. The Figure describes the full PRISMA flow diagram outlining the study selection process.

Figure:

PRISMA 12020 flow diagram for new systematic reviews which included searches of databases and registers only

Data extraction was performed independently in duplicate with predefined, standardized templates. Data points defined for extraction were as follows: authors, year, study objective, study design, event, exposure, location, population, health outcomes studied, and results.

Results

We identified 5044 records through our database searches, selecting 36 studies for full text review. Of these 36 studies, ten were retained for data extraction (Table). All were published between 2010 and 2023.

Table 1:

Extraction Criteria

Authors, Year	Study Objective	Study Design	Event, Location, Year	Exposure definition	Comparison group	Population	Health Outcomes Studied	Results
Brown et. al., 201220	Examine the effects of evacuation during Hurricane Gustav on longterm care residents diagnosed with cognitive impairment	Retrospective cohort study comparing outcomes from before to after evacuation using DiD5 framework	Hurricane Gustav, LA17, 2008	Residents from NH12 that completely evacuated prior to Gustav	All other NH residents over 3-year period in counties that were in the path of Gustav (2006-2008)	N = 21,255 residents in 119 at-risk NHs12 over 3-year study period (2006-2008); mean age 82.96 years, 19% Black, 18% with severe cognitive impairment (CPS8 ≥ 5), 35% with DM6, 11% with CHF7 5,036 residents evacuated (82% of NHs12) vs 16,219 in place	30- and 90-day mortality rate	Among residents with high CPS scores, 30- and 90-day mortality rates higher in evacuated group compared to preceding years. In adjusted analysis, evacuated residents with high CPS8 had a 2.8% increased risk of 30-day mortality (95% CI4: 0.8 to 4.8%), and 3.9% increased risk of 90-day mortality (95% CI4: 1.1 to 6.7%) compared to residents who stayed in place, from before to after hurricane's impact.
Castle et. al., 201118	Examine postrelocation mortality and functional outcomes in nursing home residents after Hurricane Katrina	Retrospective study design using propensity score matching	Hurricane Katrina, New Orleans, 2005	NH12 residents from facilities that were closed due to the hurricane	Residents from NH urban facilities located in the southern US16 but outside MS17 and LA17 as these states were impacted by the hurricane	N = 46,719 (684 relocated vs 46,035 in comparison facilities). Among relocated residents, mean age was 79.4, 28% male, 56% Black, and mean CPS8 score was 2.5	Death, days of survival, and several psychological (cognitive performance, depression, and behavior issues), ADL1 status score	Compared to matched samples, relocated patients after Hurricane Katrina had a significantly increased risk of death (aOR3 2.43, SE 0.29, p < 0.001), and significant increase in ADL1 status score (3.08 vs. 2.98, p value < .01).
Dobbs et. al., 202223	Evaluate association between hurricane-related exposure and mortality in hospice, and hospice-related utilization for short-stay and long-stay NH12 residents	Retrospective cohort design	Hurricane Irma, FL, 2017	All NH12 residents residing in state of FL17 3 days before Hurricane Irma's landfall	All NH 12 residents in same NHs12 in 2015	N = 93,572 (45,882 in exposure group, of whom 7.8% in hospice, and 47,690 in control groups, of whom 7.4% in hospice)	30- and 90-day mortality odds for hospice and non-hospice patients	Compared to NH residents in previous years, there was a 12% (95% CI4: 1.00-1.26) increased odds of 30-day mortality for hurricane exposed NH residents in hospice, and 9% (95% CI4: 1.00-1.18) increased odds for non-hospice residents. At 90 days, the odds of mortality were 2% higher (95% CI4: 0.95-1.10) for hospice residents, and 6% higher for non-hospice residents (9%% CI4: 1.01-1.11).
Dosa et. al., 201017	Evaluate multi-state hurricane-related mortality and morbidity from Hurricane Katrina among NH12 residents	Retrospective cohort design	Hurricane Katrina, LA, and MS, 2005	Residents in NHs in parishes/counties defined by the National Weather Service's hurricane watch and all residents from NFs located in counties and parishes where are least 1 NF evacuated	Residents in the same NHs as those exposed but from June - June 2004	N = 28,540 (9260 exposed [141 NHs12] in 2005 cohort vs 19,280 in 2003/2204 cohort); in exposed cohort, 75.4% female, 22.6% Black, and 20.9% with severe cognitive impairment (CPS8 ≥ 5)	Functional decline defined as 4-point drop in 28-point ADL1 scale); 30- and 90-day mortality; 30- and 90-day hospitalization rates	Compared to preceding years, 6.77% of NH12 residents exposed to Hurricane Katrina had a significant ADL1 decline compared to 5.81% (2033) and 5.1% (2004), p < 0.001. 30-day and 90-day mortality rates significance differed between hurricane exposed NH12 residents and preceding years 2003 and 2004 (30 day - 3.88% vs 2.1% and 2.28%, p < 0.001; and 90-day - 9.27% vs 6.71% and 6.31%, p < 0.001). Similarly, 30-day and 90- day hospitalization rates differed (30 day - 9.87% vs 7.21% and 7.53%, p < 0.001; and 90-day -20.39% vs 18.61% and 17.82%, p < 0.001).
Dosa et. al., 201216	To examine the differential morbidity/mortality associated with evacuation versus sheltering in place for NH12 residents exposed to the 4 most recent Gulf-hurricanes	Retrospective cohort design	Hurricane Katrina August 2005, Hurricane Rita September 2005, Hurricane Gustav September 2008, and Hurricane Ike September 2008	Residents from NH facilities that completely evacuated prior to the date of landfall of each storm	Residents from NH facilities from the 2 years	Experienced Hurricane: 36,389 residents (one of four storms), Didn’t Experience Hurricane: 42,500; White (around 75% for exposed) and female majority (around 75% for exposed) NH12 residents long (>90 days) term NH12 with moderate CPS8 scores (3-4) (around 43% for both exposed and unexposed)	30-day, and 90- day mortality and hospitalization rates	Significant increase across all hurricanes in death at 90 days post evacuation (range from 2.7% in Gustav to 5.3% in Katrina) Significance in all but Ike in hospitalization at 90 days post evacuation
Dosa et. al., 202022	To examine the relationship between exposure to Hurricane Irma and hospitalization and mortality rates for short- and long-term stay NH12 residents in FL17	Retrospective cohort design	Hurricane Irma September, 2017	All NH12 residents (short- day and long- stay) in FL at time of hurricane landfall	All NH12 residents (short-day and long-stay) in FL in 2015	N = 61,564 from 640 NHs12 (exposed population in 2017) vs 68,921 (control from 2015); 68% vs 67% female, 79% vs 78% White	30-and 90-day mortality; 30-day and 90-day hospitalization	For long-stay residents, odds of mortality at 30 days increased by 18% (95% CI4: 1.18-1.29) and at 90 days increased 9% (95% CI4: 1.05-1.14), whereas for short-stay patients, odds of mortality at 30 days were not significant (aOR3 1.06, 95% CI4: 0.98-1.14) or 90 days (aOR3: 1.05, 95% CI4 1.00-1.10). At 30 days, odds of hospitalization for long term residents increased by 11% (95% CI4: 1.041.18). However, odds of hospitalization at 30 days (aOR3 1.08, 95% CI4: 1.03-1.13) and 90 days (aOR3 1.04, 95% CI4: 1.01-1.07) was significantly increased for short-stay patients.
Hua et. al., 20 2324	Examine NH12 excess all-cause 30-day and 90- day mortality associated with Hurricane Harvey by short- vs longterm stay residents	Retrospective cohort study	Hurricane Harvey, August, 2017	NH12 residents residing in TX17 county exposed to hurricane determined by FEMA9 risk determination, in the wind swath of storm, or NH12 in county evacuated	NH12 residents residing in same TX17 county from the 2 preceding years (2015-2016)	N = 57,716 over study period with exposed: 18,479 (18% short stay vs 82% long-stay) In exposed group, 67% female, 15% Black, 67% with ADRD2 diagnosis	30- and 90-day mortality	In unadjusted analysis, 90-day mortality rates after hurricane impact (2017) were not significantly different (2016, 2015) than 2 preceding years for short- stay residents (15.2% vs 13.5% [2016] and 14.1% [2015], p = 0.11) but was significant for long-term stay residents (7.5% vs 6.9% and 6.6%, p =0.004). In adjusted analysis, 90-day mortality rates were significant higher compared to 2015 and 2016 (7.5% vs 6.9% [2016] and 6.6% [2015], p =0.004). There was no significant difference in adjusted mortality rates at 30 and 90 days for short-stay residents.
Miller, 201425	Examine association between cognitive decline, and changes in mood and functional status for SNF15 residents impacted by evacuations during the 2009 Red River Flooding in Fargo, ND17	Retrospective study design	North Dakota Red River Floods, 2019	All SNF15 residents evacuated from Fargo	All SNF15 residents in Bismarck, Grand Forks, and Minot, ND	N = 1603 (exposed 530, controls from other 4 SNFs15 1,073); 63.5% were older than 80 years, 98.5% were white, and 71.83 % were female	Change in cognitive function, change in mood, and change in ADL1 score recoded using the MDS11 score	There were no significant differences between evacuated and non- evacuated SNF15 residents in the summary measures of cognitive function, change in mood, and ADL1 score.
Skarha et. al., 202121	Estimate associations between hurricane-related power loss and short- and longterm NH12 residents hospitalization and mortality among NH12 residents after Hurricane Irma.	Retrospective cohort study	Hurricane Irma, FL, September 2017	Residents residing in NHs with post-storm power outage determined by required the FLHealthSTAT voluntary Hurricane Irma Facility Impact (HIFI10) survey	Residents residing in NHs without post-storm power outage	N = 54,095 residents from 591 NHs12 (27,892 with power loss [299 NHs12] vs, 26,203 without [292 NHs12]), 67% women, 78% white, and 70% long-term residents 27,892 experienced power loss, 26,203 didn’t experience power loss; White (78%), Women (67%) NH12 (short (<90 days) and long (>90 days) term) residents with dementia in Fl17	First hospitalization within 30 days, 7- and 30-day mortality	In unadjusted and adjusted analysis, risk of first hospitalization was not significantly increased for residents experiencing power loss vs no power loss. However, risk of 7-day mortality significantly increased in unadjusted analysis (7-day mortality rate – OR13 1.25, 95% CI4: 1.05-1.48) Both 7-and 30-day mortality significantly increased in adjusted analysis (7-day mortality rate aOR3 1.25, 95% CI4: 1.05-1.48; 30- day mortality rate aOR3 1.12, % CI4: 1.02-1.23) Effect modification was noted by age category (interaction P =0.01) but not short-vs long-stay type.
Thomas et. Al., 2012 19	To examine the hospitalization and mortality rates associated with forced mass transfer of NH12 residents with the highest levels of functional impairment defined as an ADL1 greater than 23.	Retrospective cohort study	Hurricane Gustav, TX and FL, September, 2008	Residents transferred from NHs that were completely evacuated before the hurricane's landfall	Residents transferred from NHs that sheltered in place or did not completely evacuated before the hurricane's landfall	Evacuated: 1,295 (25% black, 24% male), Sheltered in Place: 367 (20% black, 21% male) exposed to Hurricane Gustav; long-stay (>90 days) of most functionally impaired NH12 residents who had resided in an at-risk facility with CHF7, DM6, and dementia	30- and 90-day mortality (rates were determined using the prehurricane period (3 months prior to landfall to 4 days prior to landfall), and cognitive status (CPS8 score severity: medium (score 3 –4) and high (score 5–6))	12% (n=155) of those evacuated were hospitalized within 30 days after the storm, 21% (n=273) were hospitalized within 90 days; 6.2% (n = 80) died within 30 days and 15.2% (n = 197) within 90 days 8% more hospitalizations by 30 (SE14: 2.9%) and 90 days (SE14: 3.8%) but there is not a statistically significant greater risk of death at 30 and 90 days

Study Characteristics

All studies were retrospective cohort studies. Of the ten studies that met the inclusion criteria, nine examined the impact of hurricanes that occurred in the Southern United States (Florida, Louisiana, Mississippi, and Texas) on individuals receiving care in NHs. The hurricanes studied were Hurricane Katrina (2005)^16–18, Hurricane Gustav (2008)^16,19,20, Hurricane Irma (2017)^21–23, and Hurricane Harvey (2017)²⁴. The one non-hurricane study that met our inclusion criteria was a doctoral thesis that examined the association between the Red River floods in Fargo, North Dakota (2009) and changes in cognitive and physical functioning on residents in skilled nursing facilities (SNFs).²⁵

Of the ten studies, eight were written by the same authorship group.^{16,17,19–24}

Exposure Definitions

The definition of hurricane exposure varied between studies, with exposure definitions based on either geographical location, facility level, hurricane-related evacuation, or hurricane-related power outage. For example, the three studies that examined the effects of Hurricane Irma defined the exposed areas as the entire state of Florida given its wide-ranging impacts.^21–23 In comparison, the exposure to Hurricane Gustav was defined by complete NH evacuation^16,19,20. Castle and Enberg (2011)¹⁸ defined exposure to Hurricane Katrina based on NHs that closed after landfall. Differently, Hua et. al. (2023)²⁴ defined the resident exposure to Hurricane Harvey by either counties that were: 1) In the ‘wind swath of the storm’ determined by the National Hurricane Center,²⁶ 2) Included in a state-provided list where at least one NH evacuated, or 3) On the Federal Emergency Management Agency’s at-risk list.²⁷ There were four studies that incorporated hurricane-related NH evacuations as the exposure.^16,18–20 These defined evacuations as the complete relocation of residents prior to the date of hurricane landfall. One study incorporated flooding-related NH evacuations as the exposure, defining evacuation as the voluntary or involuntary relocation of residents to a safer place.²⁵

By contrast, Skarha et. al (2021) defined hurricane-related exposures as power outages associated with these events.²¹ Power outages were identified using data reporting systems and impact surveys from FLHealthSTAT, an emergency reporting system used at the time in Florida and since replaced by an updated system, and the voluntary Hurricane Irma Facility Impact (HIFI) survey.²¹

Population Characteristics

Eight of the included studies defined post-acute and long-stay residents as those residing in NHs for up to or greater than 90 days, respectively.^{16,17,19–24} Skarha et. al. (2021) and Miller (2014) did not specify how post-acute and long-stay residents were defined.^21,25 On the other hand, Castle and Enberg (2011)¹⁸ defined post-acute residents as those receiving “post-acute care measures” but did not further specify the definition.

All studies extracted included information about the composition of their study populations by race and sex. For nine of the included studies, the majority the NH’s residents were White and the percentage of residents who were Black ranged from 15% to 23%. One study examined a NH population where 56% of the residents were Black and 28% were male.¹⁸ Several studies described the comorbidities associated with the study populations, which included diabetes mellitus (DM) and chronic heart failure (CHF).^19,20 In these studies, the populations included 35–36% of residents diagnosed with DM and 11–13% with CHF. Two studies focused on NH residents with severe cognitive impairment defined as Cognitive Performance Scale (CPS) scores of ≥ 5 (scale 0 to 6).^19,20

Data sets used to gather resident demographics and associated NH data included the Minimum Data Set (MDS),^16–25 the On-line Survey Certification and Recording (now named Certification and Survey Provider Enhanced Reports),^16,18,19 Centers for Medicare Services (CMS) Medicare claims,^{16,17,19,22,24} CMS Standard Analytical Files,²³ and Nursing Home Compare.²⁴

Outcomes

Outcomes analyzed in the included studies were mortality within 7, 30, and 90 days following the EWE, 30- and 90-day hospitalization, 30- and-90-day hospice enrollment, and worsening of functional status.

The definition of change in functional status varied by study. Dependence was scored 1 or greater, while independence was given a score of 0. Dosa et. al. (2010) used a four point drop in the ADL (Activities of Daily Living) scale,¹⁷ whereas Castle and Enberg (2011)¹⁸ created their own ADL status score and examined any change in this score. This ADL score was created using five MDS items (transfers, locomotion, dressing, eating, and toilet use), and scored on a six-item scale from 0 to 5.¹⁸ Conversely, Miller (2014) examined indicators for delirium using data from the North Dakota Department of Human Services. These delirium indicators reflected the presence or absence of various delirium signs/symptoms including easily distracted, periods of alerted perception, and episodes of disorganized speech.²⁵

Qualitative Synthesis of Findings

1. Mortality outcomes

Among the studies examining mortality,^16–24 all reported significant associations between hurricane exposure and adjusted 90-day mortality compared to non-exposed residents. Of note, the likelihood of 90-day mortality varied by the degree of cognitive impairment.^19,20 For example, Brown et. al. (2012) found that long-stay residents with high CPS scores (≥ 5) who were evacuated post-Hurricane Gustav experienced a significant 3.9% (CI: 1.1%-6.7%) increase in deaths within 90 days compared to those who stayed in place.²⁰

Studies that compared the mortality of long-stay and post-acute residents found differences between the two populations. In the Hua et. al. (2017) study, the year of the hurricane (Hurricane Harvey in 2017) was compared to two non-hurricane years (2015 and 2016). The study found no significant differences in 90-day adjusted mortality rates of hurricane-exposed and unexposed post-acute residents (15.2% vs 13.5% [2016] and 14.1% [2015], p = 0.11). Similar to Brown et. al. (2012)²⁰, they found significant differences in 90-day mortality rates for long-stay residents (7.5% vs 6.9% [2016] and 6.6% [2015], p =0.004),²⁴ but there was not a significant association between the hurricane and 30-day mortality rates for post-acute nor long-stay residents.

Four studies conducted subgroup analyses based on comorbidities, which included DM, chronic heart failure, and chronic obstructive pulmonary disease (COPD).^16,19,20,24 Hua et. al. (2023) found that, among long-stay NH residents with COPD exposed to Hurricane Harvey in 2017, there was a significant difference in adjusted 90-day mortality rates compared to non-exposed residents in 2015 (9.2% vs. 7.2%, p < 0.01).²⁴ In Thomas et. al. (2012), the authors found no significant differences in 30- and 90-day mortality and hospitalization rates between residents with DM and CHF exposed and not exposed to Hurricane Gustav.¹⁹

Only one paper studied the association between hurricane-related impacts, such as power outages, and health outcomes in post-acute and long-term NH residents. In this study, Skarha et. al. (2021) found that power outages due to Hurricane Irma resulted in significantly increased odds of 7-day (odds ratio [OR] 1.25; 95% CI, 1.05–1.47) and 30-day (OR 1.12; 95% CI, 1.02–1.23) mortality.²¹

One study, Dosa et. al. (2020), found an 18% (OR 1.18, 95% CI: 1.08–1.29) increased risk of 30-day mortality and a 9% (OR: 1.09, 95% CI: 1.05–1.14) increased risk of 90-day mortality for long-stay NH residents associated with Hurricane Irma. However, the same level of statistical significance was not observed for PAC residents (30-day OR 1.06, 95% CI: 0.98–1.14; 90-day OR: 1.05, 95% CI 1.00–1.10).²²

2. Hospitalization outcomes

The impact of hurricanes on hospitalization varied between post-acute and long-stay residents. In Dosa et. al. (2020), odds of hospitalization after Hurricane Irma were significantly increased for PAC residents at 30 days (OR 1.08, 95% CI: 1.03–1.13) and 90 days (OR 1.04, 95% CI: 1.01–1.07).²² Thomas et. al (2012) found that evacuation related to Hurricane Gustav was associated with 8% more hospitalizations at both 30 days (standard error [SE] 2.9%) and 90 days (SE 3.8%) for the most functionally impaired long-stay residents (ADL score ≥ 23).¹⁹

3. Other outcomes

Hurricane-associated hospice enrollment and mortality was analyzed in only one study.²³ Hurricane-associated hospice enrollment rates were calculated using the first hospice stay within 30 and 90 days after exposure to the hurricane. At 30 days post-Hurricane Irma, there were significant differences in rates of first hospice stay between exposed (2017) and unexposed (2015) long-stay NH residents (OR 1.15, 95% CI: 1.02–1.23), but not for post-acute residents (OR 1.02, 95% CI: 0.91–1.15). Similarly, 90-day hospice enrollment rates were significantly higher for long-stay residents (OR 1.12, 95% CI: 1.05–1.20).²³ Exposed NH residents in hospice experienced a 12% increase in the odds of mortality within 30 days (OR 1.12, 95% CI: 1.00–1.26), but the difference in 90-day mortality rates was not significant (OR 1.02, 95% CI: 0.95–1.10).²³

Changes in functional and cognitive status varied by study and exposure. Dosa et. al. (2010) found long-stay NH residents exposed to Hurricane Katrina experienced a significant decline in Activities of Daily Living (ADL) score compared with those in non-hurricane years (5.81% vs 5.1%, P < 0.001).¹⁷. Similarly, Castle and Enberg (2011) found that, in a propensity score matched analysis, Hurricane Katrina-related relocated residents experienced a significant increase in need with ADL assistance compared to those who did not relocate (3.08 vs. 2.98, p value < 0.01).¹⁸ Miller (2014) found that flood-related relocation had no significant impact on functional status, reporting that no significance was found between evacuated and non-evacuated SNF residents in measures of cognitive function, indicators of delirium, change in mood, and ADL scores.²⁵

Discussion

Our scoping review found ten articles that examined the health outcomes for post-acute and long-stay NH residents associated with EWEs. Of these articles, nine focused on hurricane events that occurred in the Southern United States, and one focused on flooding events in North Dakota²⁵. Consistently across all hurricane-related studies, a significant relationship was found between hurricane exposure and 90-day mortality rates among long-stay NH residents.

Our findings demonstrate a focus on hurricane-related exposure in the Southern United States. Several reasons likely underscore this finding. Over the past ten years, the number of hurricanes has increased in the United States, and these hurricanes have become more intense compared to earlier years.²⁸ Moreover, approximately 50% of the United States population resides along coastlines and hurricanes often strike coastal regions, frequently where NHs are located. This is often the case for hurricanes in the Southern United States, particularly Florida, where 21.6% of residents are aged greater than 65 years— the fourth highest percentage in the country.²⁹

A second notable finding is the various measures of exposures to EWEs used. Although nearly all of the manuscripts focused on hurricane-related exposures, how exposures were defined varied greatly, ranging from large geographic regions impacted by the hurricane (e.g., Hurricane Irma’s impact on Florida),²³ to more spatiotemporally refined measures that accounted for the hurricane’s pathway and/or trajectory,¹⁷ and even direct sequelae of the storm (e.g., post storm power outages).²¹ This highlights a notable challenge in studying the impacts of EWEs on NH residents, and may in part explain the heterogeneity of the effect sizes in mortality and hospitalization rates noted between studies.

Nonetheless, our analysis reveals several important gaps in the literature that warrant attention, particularly as EWEs become more frequent, severe, and intense. First, while our scoping review reveals a focus on the Southern United States, EWEs other than hurricanes extend beyond this region, affecting large swathes of the nation. Severe storm-related high-tide flooding has doubled in coastal regions over the past 20 years³⁰ in the United States. Likewise, flooding has increased in inland regions, particularly due to river-based flooding.³¹ Similarly, since the 1980s, the United States has seen a significant increase in not only the incidence of wildfires, but also the size of areas directly and indirectly affected by them and associated smoke. From 2002–2020, an annual average of 1.7 million acres has burned in July in the United States,³² with the highest increase in burned acreage occurring in states with high 65 years and older populations (19%) like Arizona and Oregon.²⁹ The incidence of other EWEs has also increased. Across the Central United States, there are about 1,000 reported tornadoes yearly, causing billions of dollars in damage and thousands of injuries.¹² These trends indicate a critical need to understand how health impacts may differ between NH residents receiving PAC versus long-term care, and whether those with multiple medical comorbidities,^24,33 greater ADL limitations, and cognitive impairment³⁴ are more vulnerable than other residents to improve disaster preparedness and management for NHs.

Second, our scoping review revealed that there is limited evidence of the longer-term health impacts of EWEs on NH residents (both PAC and long-stay). Of the studies that met our inclusion criteria, the longest period that any health outcome was observed was 90 days. However, there is growing evidence from studies of the general population that EWEs increase the risk of adverse long-term health outcomes, up to 10 years after these events.³⁵

Several mechanisms may contribute to the relationship between EWEs and longer term health outcomes, particularly in short-stay residents undergoing PAC with the expectation of returning to the community setting. Abramson et al. (2022) found among a cohort of persons experiencing housing instability after being displaced by Hurricane Katrina, the median time to achieve housing stability was 3 years.³⁶ In other studies, EWEs caused widespread disruption to the socioeconomic structure of communities, such as changes in economic opportunities and demographics, which highlights underlying pressures to existing infrastructure, including gentrification and the displacement of low-income and minoritized communities.^37,38 Therefore, compared to long-stay NH residents who have an increased risk of mortality after EWE exposure, short-stay residents undergoing PAC are more likely to experience these disruptions caused by EWEs into the longer term, and have greater risk of rehospitalization, which may explain the findings of studies that compared long- and short-stay residents.^22,24 These stressors have the potential to increase the likelihood of long-term adverse health outcomes, warranting further study to identify the resident and NH characteristics associated with higher risk of them, as well as the underlying mechanisms driving variation by region and EWE type.

Third, our findings emphasize that the costs and patterns of healthcare utilization for NH residents following EWEs need to be considered. Neither costs nor utilization was examined for NH residents in the ten studies analyzed for this review. There may also be variation based on resident characteristics. For example, many NH residents have Alzheimer’s disease and related dementias (ADRD), which previous research has associated with more healthcare use and higher costs compared to residents without ADRD.³⁹ Additionally, over the past two decades, more older adults across the country have been discharged from the hospital to PAC in NHs, resulting in a 25% increase in use of these services.⁴⁰ These services are the largest influence on geographic variation in Medicare spending.⁴¹ Thus, there is a need to understand how different EWEs impact healthcare spending for NH residents in different regions of the United States.

Fourth, our findings underscore the need to examine and quantify impacts of EWEs on vulnerable populations in NHs through the lens of environmental and climate justice -- the concept that individuals belonging to certain races and ethnicities, income brackets, and origins are disproportionately more exposed to environmentally hazards. Few of the studies have examined disparities in health outcomes by race and ethnicity through stratified analyses,¹⁸ or other climate vulnerabilities, including climate-sensitive comorbidities such as CHF and COPD.^16,19,20,24 However, there is widespread literature showing the disproportionate impact of EWE exposure on both people of color and those who are functionally and cognitively impaired, among others.^42,43 These considerations of environmental injustice are important to consider and elucidate not only because EWEs are likely to increase due to climate change, but also because by characterizing and quantifying the disparities created by EWEs, NH administrators and policy makers can consider new and unique ways to address these disparities in disaster and climate policy. Moreover, there are several forces that shape NH populations that may inform differences in health outcomes related to EWE exposure. Racial/ethnic make-up of NH residents has been associated with NH care quality,⁴⁴ payor mix (e.g., proportion of dually eligible Medicare-Medicaid beneficiaries),^45,46 and policies related to reimbursement.⁴⁶

In the manuscripts reviewed, all but one study included a NH population consisting of a White-female majority.^{16,17,19–25} Examinations of environmental justice are needed that consider disparities at the NH-level, examine racial and ethnic disparities in health outcomes, as well as differences in outcomes within NHs that have disproportionate shares of dually-eligible residents participating in Medicare and Medicaid and racial and ethnic minorities.

Fifth, our findings reveal limitations to the methodologies employed in existing studies. Several cohort studies considered multiple hurricanes across various time periods using difference-in-differences study designs. Other studies directly compared non-exposed residents in previous years to residents in years when hurricanes struck NHs, thus assuming no time-varying differences in the NHs. Although one study utilized matching techniques to account for observable differences between exposed and unexposed cohorts, many did not. Furthermore, when considering events occurring in multiple time periods, newer techniques, such as the approach developed by Sant’Anna and Calloway,⁴⁷ consider group-time average treatment effects of exposures when estimating associations, mitigating this source of bias.

While we employed systematic methods that accounted for potential biases in this scoping review, we note two limitations. First, we recognize that EWEs affect NH residents outside the United States, but we excluded non-English language papers as professional translation was cost-prohibitive for this unfunded study. However, only one non-English language paper was potentially eligible; it studied the impact of flooding-induced evacuation on the mortality of older adults in France.⁴⁸ Although it is possible that our search strategy missed other non-English manuscripts, it is unlikely given that we did not apply this exclusion criteria until the full-text review. Second, we limited our search to peer-reviewed manuscripts in the past 20 years. There may be manuscripts pre-dating our study period that analyzed the impact of EWEs on outcomes of NH residents. However, our interest was more contemporaneous given the climate amplification of these events.

In summary, our scoping review characterized the existing literature on the impacts of EWEs on NH residents, highlighting the focus on hurricanes in the Southern United States. We note several opportunities to expand research to other EWEs including wildfires, tornadoes, and storm-related flooding, in addition to studies of long-term health outcomes of NH residents, use of rigorous econometric/statistical and epidemiological methods, and greater consideration of environmental justice, particularly for vulnerable groups receiving care in NHs. As EWEs increasingly threaten the well-being of residents of these facilities, our proposed research agenda will provide critically needed evidence of the impacts of these events to design and implement better disaster preparedness and management strategies across the United States and globally.