Emergency department use among assisted living residents after Hurricane Irma

Abstract

Objectives:

Nursing home residents are especially vulnerable to adverse outcomes after a hurricane. Prior research suggests that emergency department (ED) visits increase among community residing older adults after natural disasters. However, little is known about the impact of hurricanes on the large population of older adults residing in assisted living (AL) settings, particularly the influence of storms on the rates and causes of ED visits. We examined whether rates of ED use for injuries and other medical reasons increased after Hurricane Irma in 2017 among AL residents in Florida.

Design:

Retrospective cohort study

Setting and Participants:

Samples of 30,358 Medicare fee-for-service beneficiaries in 2016 and 28,922 beneficiaries in 2017 who resided in Florida AL communities.

Measures:

The number injury-related and other medical visits per 1,000 person-days within 30 and 90 days of September 1st in 2016 and 2017. We adjusted for age, race, sex, and chronic conditions using linear regression with AL fixed effects. We compared the top 10 primary diagnoses resulting in an ED visit between 2016 and 2017.

Results:

Adjusted rates of injury-related visits were 13% higher at 30 days but did not differ at 90 days. Other medical visits were 14% higher at 30 days in 2017 than in 2016 and 9% higher at 90 days. Heart failure was a leading cause of ED visits within 90 days of September 1st in 2017, unlike in 2016.

Conclusions and Implications:

Increased attention should be paid to AL communities in disaster preparedness and response efforts given the increased likelihood of ED visits following a hurricane.

Brief Summary:

We examined risk-adjusted rates of emergency department use among Florida assisted living residents after Hurricane Irma and described types of visits. Our study documented significant increases in utilization rates.

Impact of Hurricane Irma on ED Visits among Florida Assisted Living Residents

Assisted living (AL) communities provide housing and personal care services to over 800,000 older adults in the United States but do not provide continuous 24-hour skilled nursing services like nursing homes. Nonetheless, in 2015, nearly 42% of AL residents had dementia and 29% had difficulty transferring out of bed.¹ Older adults residing in long-term care facilities are especially vulnerable to disasters such as hurricanes.^2,3 However, nothing is published about emergency department (ED) use after disasters in AL communities. ED visits increase in the weeks after disasters among community residing individuals’^4-7 Because AL communities are state regulated and are not held to the more stringent federal emergency preparedness activities required of nursing homes, residents are potentially at risk.⁸

Additionally, the most common reasons for ED visits after disasters among AL residents are unknown. Injuries increase in the immediate weeks after disasters among adults residing in the community, often the result of falls due to hazardous conditions.^9,10 Disasters also cause disruptions in healthcare systems,¹¹ potentially contributing to ED visits because resources are unavailable in the community.⁵ For example, many ED visits following hurricane Katrina occurred because individuals ran out of medication.^5,7

Hurricane Irma made landfall on the western coast of Florida on September 10, 2017, causing an estimated $30 billion in damage.¹² This study evaluated the 30- and 90-day impact of Hurricane Irma on ED utilization among Florida AL residents by comparing adjusted rates of ED use against those of residents living in the same AL communities the year prior. We also examined the primary diagnoses resulting from ED visits 90 days after the storm, and compared them with the primary diagnoses during the same time period in 2016.

Methods

Data

Data came from the Medicare Master Beneficiary Summary File (MBSF), a Residential History File (RHF),¹³ a ZIP Code History file (ZHF), a list of licensed AL communities, the MedPar inpatient claims, and the Medicare outpatient claims. The MBSF provided information on demographic characteristics and chronic conditions. The Residential History File (RHF) provided information regarding hospital and nursing home stays using a combination of nursing home Minimum Data Set assessments, Medicare claims, and the Home Health Outcome and Assessment Information Set.¹³ The ZHF provided a person’s ZIP code each day within a calendar year. The ZHF was created by integrating the MBSF and the Enterprise Cross Reference File, which contained beneficiaries’ 9-digit ZIP codes. The addresses of the AL communities came from a national list of state licensed residential care settings. The inpatient and outpatient claims provided information on emergency department visits.

We included AL communities in the entire state of Florida because the large size and uncertain path of Hurricane Irma prompted widespread evacuations.¹⁴ Disruptions in care, evacuations, and stress that occur before a storm can contribute to morbidity among older adults.² We chose September 1st as the study date as opposed to the landfall date for this same reason. Consistent with previous research, we included individuals living in AL communities of 25+ beds that were licensed to serve an older adult population.^15,16 We used a validated¹⁵ methodology to identify Medicare beneficiaries residing in AL communities with at least 25 beds, which has been applied in other studies:^17,18using a combination of home health OASIS assessment data, Part B claims, the ZIP code history file, and our list of licensed AL communities, we created a finder file of unique 9-digit ZIP codes representing licensed AL communities. For additional explanation of the methodology to identify individuals residing in large ALs, please see the Appendix.

To identify ED visits, we used the 2016 and 2017 inpatient MedPAR and outpatient Medicare claims data. We excluded AL residents who were in any other healthcare setting (i.e., nursing home or hospital) on September 1st, capturing a total sample of 43,609 beneficiaries in 2016 and 41,449 in 2017. Because claims data were not available for Medicare Advantage (MA) beneficiaries, we excluded AL residents with any MA coverage within 180 days before or after September 1st in 2016 and 2017, resulting in a final analytic sample of 30,358 Medicare fee-for-service beneficiaries residing in large FL ALs in 2016 and 28,922 in 2017. Additional information about the data and methods used for these analyses can be found in the Brown University Digital Repository(doi: https://doi.org/10.26300/svpm-ey59).

Measures

We classified all ED visits experienced during each point in time as injury-related or “other medical” using a previously validated algorithm developed by New York University.^20,21 This algorithm provided the probability that an ED visit primary discharge diagnosis code fell into categories based on the severity of the diagnosis. Injuries were assigned a mutually exclusive category.^20,21 For additional information about the NYU algorithm, please see the Appendix.

We summed the number of visits in each 30-day study window and expressed them as the number of visits per 1,000 person-days alive. We included an indicator for whether the AL resident had been diagnosed with one of the following chronic conditions before September 1: anemia, atrial fibrillation, cancer, chronic kidney disease, chronic obstructive pulmonary disease, depression, diabetes, heart failure, hyperlipidemia, hypertension, ischemic heart disease, stroke, and Alzheimer’s disease and related dementias (ADRD). We categorized the number of these chronic conditions into three groups (<2, 2-3,4+) to account for comorbidity burden. We also categorized age into three groups (65-74, 75-84, 85+). We defined dual eligibility as beneficiaries who were enrolled in Medicare and Medicaid on September 1st.

Analysis

The rates of any ED visit in the 2017 cohort exposed to Irma were contrasted to the rates from the control group in 2016. We also compared the rates by type of ED visit each year. We adjusted rates for race, age, sex, dual eligibility, chronic conditions, and AL fixed effects using linear models. AL fixed effects were used to adjust for any omitted variables that may affect ED use and differ between ALs, such as the quality of care provided. We used independent t-tests to examine differences in adjusted outcomes across cohorts. Data were analyzed using SAS 9.4 and STATA 16. The institutional review boards at Brown University and the University of South Florida approved the research protocol.

Results

Table 1 displays the descriptive characteristics of the cohort exposed to Hurricane Irma in 2017 compared to the cohort that was not exposed in 2016. In general, residents across years were very similar in terms of their demographic and chronic conditions. Statistically significant differences are noted in the table. In 2017, 2,611 Florida AL residents in our sample (9%) experienced 6,567 ED visits within 30 days of September 1st and 6,030 residents (21%) experienced 8,641 ED visits within 90 days.

Table 1.

Characteristics of Medicare beneficiaries aged 65+ residing in Florida assisted living communities, by exposure to Hurricane Irma (2016 vs. 2017)

	2016 n=30,358	95% CIs	2017 n=28,922	95% CIs
Age group
65-74	15.7	[15.3, 16.1]	16.3	[15.9, 16.8]
75-84	28.6	[28.1, 29.1]	18.5	[28.0, 29.0]
85+	55.7	[55.1, 56.2]	55.2	[54.6, 55.7]
Sex (%)
Male	33.7	[33.2, 34.3]	33.8	[33.2, 34.3]
Female	66.3	[65.7, 66.8]	66.2	[65.7, 66.8]
Race (%)
White	93.5	[93.2, 93.7]	93.3	[93.0, 93.6]
Black	1.9	[1.8, 2.1]	2.0	[1.9, 2.2]
Hispanic	3.3	[3.1, 3.5]	3.3	[3.1, 3.5]
Other	1.3	[1.6, 1.4]	1.4	[1.3, 1.5]
Dually eligible for Medicare and Medicaid (%)†	11.1	[10.8, 11.5]	11.4	[11.1, 11.8]
Chronic Conditions (%)‡
Alzheimer’s disease and related dementias	32.6	[32,1, 33.1]	34.0	[33.5, 34.5]
Anemia	45.1	[44.5, 45.7]	43.7	[43.1, 44.3]
Cancer	11.2	[10.8, 11.5]	11.3	[10.9, 11.7]
Chronic kidney disease	32.6	[32.1, 33.1]	35.1	[34.5, 35.6]
Chronic obstructive pulmonary disease	18.9	[18.5, 19.4]	19.5	[19.1, 20.0]
Congestive heart failure	25.6	[25.1, 26.0]	25.4	[24.9, 25.9]
Depression	28.1	[27.6, 28.6]	29.1	[28.6, 29.6]
Diabetes	29.5	[29.0, 30.0]	29.3	[28.7, 29.8]
Hyperlipidemia	64.2	[63.7, 64.8]	65.8	[65.2, 66.3]
Hypertension	79.4	[78.9, 79.8]	79.4	[78.9, 79.8]
Ischemic heart disease	47.5	[46.9, 48.0]	47.0	[46.5, 47.6]
Stroke	7.0	[6.7, 7.2]	7.1	[6.8, 7.4]
Number of Chronic Conditions§
Fewer than 2 chronic conditions	13.5	[13.1, 13.9]	13.0	[12.6, 13.4]
2-3 chronic conditions	27.5	[27.0, 28.0]	27.2	[26.7, 27.7]
4+ chronic conditions	59.0	[58.4, 59.5]	59.8	[59.2, 60.3]

Notes: Irma landfall was September 10th, 2017. We used September 1st as the reference date to account for evacuations, stress, and medical care disruptions before the storm.

^†Dually eligible on September 1^st

^‡Chronic conditions measured as whether the beneficiary was diagnosed with the chronic condition before the September 1st date of each year

^§Of the 12 chronic conditions listed above

Figure 1 presents the adjusted rates of visits, by type, during each 30-day window. Injury-related visits were 13% higher at 30 days following Hurricane Irma when compared to 2016 (p<.001). The rate of injury-related visits did not differ significantly between years within 90 days of the storm. Rates of other medical visits were 14% higher at 30 days in 2017 than in 2016 and 9% higher at 90 days (p<.001). Unadjusted rates, displayed in Supplemental Figure 1 in the Appendix, were similar to the adjusted rates.

Figure 1. Adjusted rates of emergency department visits among assisted living residents age 65+ residing in Florida, by type of ED visit and year.

Notes: Exposure date was September 1st in each year of each year (2016 and 2017)

Data came from the 2016 & 2017 Medicare Master Beneficiary Summary File and Chronic Conditions, MedPAR file, and Medicare outpatient claims. Residents were enrolled in fee-for-service Medicare and resided in assisted living on September 1 of 2016 or 2017. All rates were standardized by the 1,000 person-days alive during each study window. Visits were classified as injuries by the New York University (NYU) algorithm based on ICD-10 codes.

Rates adjusted for race, age, dual eligibility, the presence of chronic conditions, and AL fixed effects

Finally, we examined the top 10 primary diagnosis codes from the ED visits in 2017 within the 90-day window and compared to the diagnosis codes in 2016. Results are shown in Table 2. Reasons for visits were relatively similar; individuals had high rates of urinary tract infections, sepsis, and head injuries during both years. One notable difference was that in 2017, two of the top 10 primary diagnoses were related to heart failure but not in 2016.

Table 2.

The top 10 ICD-10 codes for years 2016 and 2017, 90 days after September 1st

	2016		2017
Place	Diagnosis	Number of visits (Percentage of all visits, n=8,487)	Diagnosis	Number of visits (Percentage of all visits, n=8,641)
1.	Urinary tract infection	419 (4.9%)	Urinary tract infection	385 (4.5%)
2.	Unspecified injury to the head	221 (2.6%)	Sepsis, unspecified	245 (2.8%)
3.	Syncope and collapse	188 (2.2%)	Unspecified injury to the head	209 (2.4%)
4.	Sepsis, unspecified	176 (2.1%)	Syncope and collapse	181(2.1%)
5.	Chronic obstructive pulmonary disease with acute exacerbation	160 (1.9%)	Other chest pain	166 (1.9%)
6.	Other chest pain	150 (1.8%)	Acute kidney failure, unspecified	158 (1.8%)
7.	Acute kidney failure, unspecified	138(1.6%)	Chronic obstructive pulmonary disease with acute exacerbation	147 (1.7%)
8.	Chest pain, unspecified	124 (1.5%)	Hypertensive heart disease with heart failure	143 (1.7%)
9.	Pneumonia, unspecified organism	115 (1.4%)	Hypertensive heart and chronic kidney disease with heart failure and stage 1 through stage 4 chronic kidney disease, or unspecified chronic kidney disease	126 (1.5%)
10.	Essential primary hypertension	100 (1.2%)	Pneumonia, unspecified organism	125 (1.5%)

Notes: Exposure date was September 1st in each year of each year (2016 and 2017)

Data came from the 2016 & 2017 Medicare Master Beneficiary Summary File and Chronic Conditions, MedPAR file, and Medicare outpatient claims. Residents were enrolled in fee-for-service Medicare and residing in assisted living on September 1 of 2016 or 2017

Discussion

We observed significantly increased rates of ED visits among Florida AL residents associated with Hurricane Irma. These findings are particularly salient given the high levels of physical and cognitive impairment among the population residing in AL. Because AL disaster requirements are not as prescriptive as the federal requirements for nursing homes,⁸ AL communities may be less prepared for evacuations or sheltering in place during a disaster. In previous studies, administrators of AL communities reported not having access to a generator.^8,22 In one study, only 19% of evacuating AL communities in Florida conducted an assessment of risk for different types of disasters.²²

Florida has more robust standards than other states, as it requires that all AL communities develop disaster preparedness plans that include a facility risk assessment, resident and staff communications plan, operations and staffing procedures, evacuation plans, and plans for sheltering in other facilities.²⁴ After a recent mandate, Florida AL communities were required to have the ability to generate adequate power to maintain cool temperatures in their buildings.²⁴ AL residents in other states may be at an even higher risk than of other medical visits after a disaster than Florida residents due to less prescriptive standards. Consideration, however, should be given regarding how mandates may disproportionately affect AL communities with fewer resources.

Our work found that injury-related ED visits increased for AL residents in the 30 days after Hurricane Irma when compared to 2016 but returned to baseline levels within 90 days. These results were similar to those in one study that examined injury-related ED use after Hurricane Maria and found that rates returned to their baseline levels 6 weeks after the storm.¹⁰ Urgent efforts to re-establish normalcy in caregiving might be a method to return injury risks to baseline.

Our study found that other medical visits increased after the storm and remained elevated at 90 days, which could be due to a variety of reasons. Disruptions in care due to medical facility closures can exacerbate chronic conditions such as hypertension.²⁵ During Hurricane Sandy, many dialysis centers closed, which resulted in a surge of dialysis patients in EDs. Our study found a greater proportion of diagnoses related to chronic kidney disease in 2017 than in 2016. Previous studies found that a lack of access to respiratory devices and prescription refills were reasons for ED visits following a hurricane.^3,5

Additionally, we found that heart failure rates were higher in 2017 when compared to 2016. These results are consistent with a study that found heart disease was the most common cause of avoidable death after Hurricane Maria in Puerto Rico²⁶ and studies that found increased heart failure rates after the Great East Japan Earthquake.^27,28 Some scholars hypothesize that physical and mental stress may be the trigger for heart failure.²⁹ Increased activity of the sympathetic nervous system may lead to increased blood pressure and heart rate, leading to heart failure.²⁹

There were limitations to this analysis. Our methodology was dependent on identifying residents in larger AL residences (25+ beds) in Florida, which limited its generalizability to smaller settings and those in other states. Our study did not include those enrolled in MA, which may impact findings given documented differences in the population of Medicare beneficiaries enrolled in MA versus traditional Medicare.³⁰ Future work would benefit from comparing the outcomes of MA enrollees to individuals enrolled in traditional Medicare as well as across MA plans with different case management strategies and benefit packages. We cannot control for anxiety, which is a limitation given there are higher rates of ED use among individuals with anxiety disorders.³⁰

Lastly, our study utilized only one year of data (2016) in comparison to the year 2017. Due to the change from ICD-9 to ICD-10 that occurred in 2015, we limited our comparison to 2016. It is noteworthy that Florida was affected by one smaller hurricane (Hurricane Hermine) and skirted by another (Hurricane Mathew) during calendar year 2016. The financial impact of both storms on the state of Florida was considerably smaller than Hurricane Irma.¹² However, we did examine differences in the sample of AL residents between 2015 and 2016 in terms of demographics and utilization and found no significant differences, providing further evidence of the uniqueness of Hurricane Irma’s impact on the state of Florida.

Conclusion

Our study documented significant increases in rates of emergency department visits utilization among AL residents in FL after Hurricane Irma. Our findings suggest that more attention must be placed on disaster mitigation, preparedness, and response activities in the AL setting given the size and vulnerability of this population.