Transition of care from the emergency department to skilled nursing facility: Retrospective case‐control study

Abstract

Objective

The primary objective of this study is to describe associations between emergency department (ED)‐to‐skilled nursing facility (SNF) transition and ED length‐of‐stay (LOS). The secondary objective is to explore how social determinants of health (SDOH) influence ED‐to‐SNF transition visit parameters. In 2020, The Centers for Medicare & Medicaid Services issued the “COVID‐19 Emergency Declaration Blanket Waivers for Health Care Providers” eliminating the requirement of a 3‐day qualifying hospital stay before SNF placement. The waiver allowed ED patients to be transitioned directly to an SNF from the ED.

Methods

We conducted a descriptive retrospective case‐control study of adult patients who sought care in the University of North Carolina Hospitals (UNCH) ED between March 1, 2020, and March 1, 2022, lived in a non‐SNF residence before their ED visit, and were transitioned directly to an SNF from the ED (n ₁ = 27), compared with a group seen in the ED and admitted to hospital for SNF placement (n ₂ = 54).

Results

The ED‐to‐SNF group experienced a significantly longer ED LOS compared to the ED‐to‐Inpatient‐to‐SNF group: 72.8 hours (95% confidence interval [CI], 59.2–86.4) compared to 14.5 hours (95% CI, 12.1–16.9). We found no significant differences in SDOH between the ED‐to‐SNF group and the ED‐to‐Inpatient‐to‐SNF group.

Conclusion

Patients who transition from the ED to an SNF experience long ED stays that may adversely affect health and well‐being. Transitioning directly from the ED to an SNF may contribute to ED boarding and overcrowding.

1. INTRODUCTION

1.1. Background

Crowding, which occurs when the identified need for emergency services exceeds available resources for patient care in the emergency department (ED), hospital, or both, has become an increasingly detrimental issue in emergency medicine. ¹ With the role of emergency medicine expanding to include more prolonged diagnostic and intensive care workups, many services that used to be conducted in the inpatient realm are now performed in EDs. In a 2019 systematic review, ED crowding was found to increase medical costs, medical errors, and hospital length‐of‐stay (LOS), while decreasing hospital revenues and patient satisfaction. ² Several factors have been identified that contribute to the issue of ED crowding including non‐urgent visits, patients who incur 4 or more ED visits per year, inadequate staffing, inpatient boarding, hospital bed shortages, and delays in ancillary services. ³ ED boarding was also found to be directly related to increases in 10‐ and 30‐day mortality rates. ² Furthermore, studies have shown that ED patients who experience long lengths of stay (LLOS) are more likely to belong to the oldest age bracket of 85–95 years old, reflecting a population likely to live in a skilled nursing facility (SNF) or assisted living facility (ALF). ⁴

1.2. Importance

There is ample literature regarding the outcomes of patients transferred from inpatient care to SNFs, which is far from an ideal process. Patients and family members report feeling rushed making continuing care decisions and feeling that they were not adequately prepared for SNF transfer and their post‐acute needs. These sentiments were echoed by hospital and SNF staff members. ⁵ Social determinants of health (SDOH) may also affect the process of SNF placement for many ED patients. The process of hospital‐to‐SNF transfers changed considerably during the COVID‐19 pandemic. The Centers for Medicare & Medicaid Services (CMS) issued a “COVID‐19 Emergency Declaration Blanket Waivers for Health Care Providers” that contained language specifically about SNF placement. Previously, patients were required to undergo a 3‐day “qualifying hospital stay” before SNF placement. Additionally, patients had to undergo a Preadmission Screen and Resident Review (PASARR), which screens potential nursing facility residents for serious mental illness (SMI) and/or intellectual disability to ensure that the facility had the proper accommodations. ⁶ Both of these requirements were waived by CMS's emergency declaration with a retroactive effective date of March 1, 2020. ⁷ This created a new pathway for patients to flow from the community through the ED to a SNF without an inpatient stay. It is unclear how this new pathway has affected the patients ED visit. Even before this waiver, the community‐to‐ED‐to‐Inpatient‐to‐SNF pathway was understudied.

1.3. Goals of this investigation

The primary objective of this study is to describe the associations between ED‐to‐SNF transition and ED LOS, patient visit parameters that may affect patient comfort, safety, ED crowding, and patient flow. The secondary objective of this study is to explore whether SDOH influences ED‐to‐SNF transition visit parameters within our study population.

The Bottom Line

Patients who are transitioned directly to a skilled nursing facility from the emergency department experienced longer length‐of‐stay, compared to those admitted to the hospital for placement. In addition, transitioning directly from the emergency department to a skilled nursing facility contributed to boarding and crowding.

2. MATERIALS AND METHODS

2.1. Study design, setting, and selection of participants

A descriptive retrospective case‐control study of patients who sought care in the University of North Carolina Hospitals (UNCH) ED between March 1, 2020, and March 1, 2022, was conducted. The study was approved by the University of North Carolina (UNC) Office of Human Research Ethics institutional review board. UNCH is one of the largest referral centers in North Carolina and is a level I trauma center, an American Burn Association burn center, and a Joint Commission‐certified comprehensive stroke center. Specifically, patients included in the study were 18–105 years old at the time of patient encounter and lived in a non‐SNF residence before their ED visit. The case group consisted of patients who were transitioned directly to an SNF from the ED. These patients did not meet the criteria for admission as determined by the attending emergency physician or hospitalist. An example of a case group subject would be an elderly individual who sustained a pelvic ramus fracture, non‐operative, and weight‐bearing as tolerated. This individual might not meet strict criteria for hospital admission, yet is unable to safely return home. Physical therapy and occupational therapy would typically be consulted and recommend SNF level of care. The control group consisted of randomly selected patients who previously lived in a non‐SNF residence and were admitted from the ED to the hospital floor and then transitioned to an SNF. An example of a patient in the control group would be an elderly individual who suffered a hip fracture requiring operative treatment. Patients with a primary psychiatric diagnosis, who were incarcerated, or who required immediate, intensive care unit‐level care were excluded from the study. Patients transferred from an ED at another facility were excluded if they were expedited through the UNCH ED for admission. In these instances, the expected transfer patient would have been discussed with the UNC ED attending with appropriate pre‐arrival consults. Additionally, patients admitted to the following services were excluded from the study: nephrology, infectious disease, cardiology, heart failure, medicine intensive, and cardiac intensive care. We excluded these specialties to keep our cases and controls as comparable as possible, as many of the case patients required less intensive medical care.

Informatics for Integrating Biology and the Bedside (I2B2) was used to identify a potential set of patients who were seen in the UNCH ED between March 1, 2020, and March 1, 2022, originated from a non‐SNF residence, and subsequently transferred to an SNF. This set of patients was then imported into the Electronic Medical Record Search Engine to further narrow the study sample by searching for patients with a “SNF referral form,” a document used by case management to place previously community‐dwelling patients into a SNF. To obtain more detailed information relating to the ED timeline, EPIC, an electronic medical record system, was studied by the data extractor.

Two separate queries were made from I2B2 to identify case patients and control patients based on ED‐to‐SNF transition types: (1) patients who were transferred directly from the ED to an SNF were considered cases (ED‐to‐SNF, n ₁ = 27), and (2) patients who were admitted to the hospital from the ED before being transferred to an SNF were considered controls (ED‐to‐Inpatient‐to‐SNF, n ₂ = 54). All patients with a non‐psychiatric complaint who were transitioned from home‐to‐ED‐to‐SNF during the 2‐year study period were included, except for incarcerated individuals. The original query list of 236 MRNs included multiple encounters for some patients; these patients were included once if they had one encounter that met our inclusion criteria. Encounters that were not for the UNC Main campuses, but included other campuses within the UNC system despite search criteria were excluded. Within the EHR, the reviewer manually checked each encounter for each MRN to find the qualifying encounter in which the patient transitioned from home‐to‐UNC Main Campus ED‐to‐SNF.  Meticulous manual chart review for each encounter was required to find accurate point of origin information, which was located in care management narrative notes. For the ED‐to‐SNF group, 236 patients were screened as described in detail above, and 27 were found to be eligible for inclusion in the study. The main reason for exclusion was that patients did not originate from non‐SNF residence despite being included in query results. Our query returned 1305 potential patients to be included in the ED‐to‐Inpatient‐to‐SNF group. This larger number was expected as inpatient admission has been the more frequent pathway to SNF placement. We sought to have 2 controls for each case. The 1305 potential subjects for the control group was narrowed to 54 patients via stratified random sampling using a random number generator. Stratified random sampling was used to avoid bias due to variable ED and hospital capacity during surge versus non‐surge periods. Patients were randomly sampled in a 2:1 ratio from the 2021–22 and 2020–21 arrival years, respectively, to better reflect the distribution of year of arrival among the ED‐to‐SNF group and keep the case and control groups as comparable as possible. In total, 81 patients were included in this study (N = 81).

2.2. Data collection and variables of interest

A comprehensive chart review was performed for each patient included in the study. The abstractor was a GCP‐trained graduate assistant. The PI independently reviewed 5% of the sample for accuracy. The chart review focused on ascertaining patient demographics, SDOH, ED visit elements and medical history. Each patient selected for chart review was given a de‐identified study ID, and all data collected from the chart review were entered into a secure REDCap database to ensure patient confidentiality and data security throughout the data collection and analysis process.

The main variable of interest was ED LOS, defined as the duration of time from the patients arrival time at the ED to their SNF transfer time/off‐the‐floor ED time. SDOH was gathered from the patients chart. “Other” races included patients of American Indian/Native Alaskan, Asian, Native Hawaiian/Pacific Islander, and all other racial groups. Illicit drugs were defined as marijuana, cocaine, and opioids. Percent of zip code in poverty was categorized based on the percent of all people that are living in poverty as of 2016–2020 within the patients zip code. ⁸ Family or social support for a patient was considered present if family or friends were mentioned anywhere in the chart notes as being able to help the patient or having been involved with helping the patient previously. Similarly, the presence or absence of financial resource strain, secure housing, and food insecurity was based on the evaluation of the patient by the case manager and the social history found in their chart notes. ED visit parameters were extracted from the visit timeline in EPIC; inpatient discharge time was extracted from the charts based on the discharge summary signed date. The discharge summary signed date and time were used for this data point because it was the most accessible and consistently documented. Waiting room time was defined as the duration of time from when the patient arrived at the ED to when they were initially placed in a bed. The patients ED primary diagnosis was also collected and then categorized based on the Centers for Medicare and Medicaid Services major diagnostic groups. ⁹ Lack of SNF bed availability as a barrier to SNF placement was defined as a general lack of SNF bed availability in the county, and delays due to the weekend were not included in this category.

2.3. Data analysis

Continuous demographic characteristics and ED visit parameters are described as the mean and standard deviation. Categorical demographic characteristics and ED visit parameters are described as frequency and percentage. All time variables were coded as DateTime values and converted to decimal hours. Any variables that were considered “unknown” were coded as missing data; patients had unknown data for education, financial resource strain, food insecurity, lack of transportation, and percent of zip code in poverty.

Statistical comparisons were performed using SAS version 9.4 and Stata version 16. ¹⁰ , ¹¹ A 2‐sample t test was used to determine whether the average ED LOS was different in ED‐to‐SNF patients and ED‐to‐Inpatient‐to‐SNF patients. The number of ED‐to‐SNF patients and ED‐to‐Inpatient‐to‐SNF patients was large enough to detect differences in average ED LOS, with a power of 80% and α of 5%. A 2‐sample t test was also used to determine whether the average age and hours spent in desirable versus undesirable bed types were different between the 2 groups. For a 2‐sample t test to be valid, either the population has to be normally distributed or if non‐normal, the sample size has to be approximately 25 or 30 in each comparison group. ¹² Some variables analyzed did not meet either of these requirements, so a Wilcoxon rank‐sum test was conducted instead.

A χ² test was used to determine whether there was an association between the categorical demographic and ED visit variables for those in the ED‐to‐SNF group compared to those in the ED‐to‐Inpatient‐to‐SNF group. For a χ² test to be valid, the expected number of observations in each category must be at least 5 for a 2 × 2 table; for a bigger table, the expected number had to be at least 1, and the average expected number had to be at least 5. However, some variables did not meet this requirement and were collapsed into fewer categories. The race of the patient was re‐coded as a binary variable (Caucasian/non‐Caucasian). The marital status of the patient was re‐coded as a categorical variable with 4 categories (single, married or long‐time partner, legally separated, and widow). The percent of zip code in poverty was re‐coded as a binary variable (13.70% or less and 13.71% or greater of the population below the federal poverty line). Substance use, lack of transportation, patient moved rooms, subspecialty consults, and barriers to SNF placement were analyzed as binary variables (Yes/No).

As a secondary analysis, we investigated if there was a difference in SDOH within the ED‐to‐SNF transition group based on their ED LOS. Using the median ED LOS as the cutoff point, the ED‐to‐SNF group was split into those patients who had an ED LOS less than the median 69.3 hours (n ₁ = 14) and those patients who had an ED LOS greater than or equal to 69.3 hours (n ₂ = 13).

In both the primary and secondary analyses, a Fisher exact test was used for categorical variables that did not meet the χ² requirement for the expected number of observations and were already binary or could not be collapsed in a logical way. A Fisher exact test was conducted on the following variables in the primary analysis: ethnicity, secure housing, food insecurity, family and social support, lack of transportation, insurance type, accountable care organization/Medicare shared savings program status, prior cerebrovascular incident as a pertinent past medical history, initial bed type, and hospital bed ordered in ED. In the secondary analysis, a Fisher exact test was conducted on all variables, with the exception of percent of zip code in poverty. Additionally, a Wilcoxon rank‐sum test was conducted on time spent in the waiting room because it did not meet the normality assumption required for a 2‐sample t test. Results were considered statistically significant when the p value was less than 0.05, and all statistics are 2‐sided.

3. RESULTS

3.1. Patient demographic characteristics

Demographic characteristics of the 81 patients who experienced a home‐to‐ED‐to‐SNF transition are presented in Table 1. Patients comprising the ED‐to‐SNF transition group did not differ significantly from the patients in the ED‐to‐Inpatient‐to‐SNF group in terms of demographics. The mean age of patients in the ED‐to‐SNF group and ED‐to‐Inpatient‐to‐SNF group was 76.6 years (95% CI, 72.7–80.6) and 75 years (95% CI, 72.2–77.8), respectively (P = 0.5). Among patients in the ED‐to‐SNF group, 70.4% of patients were Caucasian compared to 83.3% in the ED‐to‐Inpatient‐to‐SNF group (P = 0.2). In the ED‐to‐SNF group, 63% of patients reported substance use compared to 57.4% of patients in the ED‐to‐Inpatient‐to‐SNF group (P = 0.6). The majority of patients in both groups were insured through Medicare (92.6% % vs. 88.9%, P = 0.3).

TABLE 1.

Demographic characteristics of eligible patients who visited the UNC medical center emergency department from March 1, 2020, to March 1, 2022 (N = 81).

	ED‐to‐SNF (n 1 = 27), No. (%)	ED‐to‐Inpatient‐to‐SNF (n 2 = 54), No. ( %)	P value
Age (years)
Mean (SD)	76.6 (10.0)	75.0 (10.4)	0.5
Sex
Male	13 (48.2)	22 (40.7)	0.5
Female	14 (51.9)	32 (59.3)
Race a
Caucasian	19 (70.4)	45 (83.3)	0.2
Black/African American	8 (29.6)	8 (14.8)
Other		1 (1.9)
Ethnicity b
Hispanic/Latino		2 (3.7)	0.6
Non‐Hispanic/Latino	27 (100.0)	52 (96.3)
Substance use c
No	10 (37.0)	23 (42.6)	0.6
Yes	17 (63.0)	31 (57.4)
Cigarettes	15 (88.2)	27 (87.1)
Alcohol	1 (5.9)	3 (9.7)
Illicit drugs	1 (5.9)	1 (3.2)
Marital status d
Single	7 (25.9)	12 (22.2)	0.7
Long‐time partner	6 (22.2)	1 (1.9)
Married	7 (25.9)	17 (31.5)
Legally separated	7 (25.9)	10 (18.5)
Widow		14 (25.9)
Education
Less than high school degree	3 (21.4)	3 (8.3)	0.7
High school degree	2 (14.3)	4 (11.1)
Some college/Associate degree	2 (14.3)	5 (13.9)
Bachelor's degree	5 (35.7)	18 (50.0)
Advanced degree	2 (14.3)	6 (16.7)
Missing	13	18
Financial resource strain
None/low risk	17 (63.0)	41 (77.4)	0.4
Medium risk	7 (25.9)	8 (15.1)
High risk	3 (11.1)	4 (7.6)
Missing		1
Secure housing b
Yes	26 (96.3)	48 (88.9)	0.4
No	1 (3.7)	6 (11.1)
Food insecurity b
Yes	2 (7.7)	8 (15.1)	0.5
No	24 (92.3)	45 (84.9)
Missing	1	1
Family and social support b
Yes	24 (88.9)	50 (92.6)	0.7
No	3 (11.1)	4 (7.4)
Lack of transportation b , c
No	22 (88.0)	43 (84.3)	1.0
Yes e	3 (12.0)	8 (15.7)
Medical	3 (100.0)	8 (100.0)
Non‐medical	3 (100.0)	7 (87.5)
Missing	2	3
Percent of zip code in poverty f
5.47 or less	2 (7.4)	1 (1.9)	0.6
5.48–9.31	9 (33.3)	18 (34.0)
9.32–13.7	4 (14.8)	7 (13.2)
13.71–20.33	12 (44.4)	25 (47.2)
20.34 or greater		2 (3.8)
Missing		1
Insurance type b
Private	2 (7.4)	1 (1.9)	0.3
Medicare	25 (92.6)	48 (88.9)
Medicaid		2 (3.7)
Uninsured		3 (5.6)
ACO/MSSP status b
Yes	3 (11.1)	9 (16.7)	0.7
No	24 (88.9)	45 (83.3)

Abbreviations: ACO, accountable care organization; MSSP, Medicare Shared Savings Program.

^aRace was analyzed as a binary variable: Caucasian and non‐Caucasian.

^b Fisher exact test.

^c Analyzed as a binary variable (Yes/No).

^d Marital status was analyzed as a categorical variable with the following categories: single, married or long‐time partner, legally separated, and widow.

^e Patients can lack both medical and non‐medical transportation.

^f Percent of zip code in poverty was analyzed as a categorical variable with the following categories: 13.70% or less and 13.71% or greater of the population below the federal poverty line.

3.2. ED length of stay

ED LOS at the UNCH ED differed between patients with an ED‐to‐SNF transition and patients with an ED‐to‐Inpatient‐to‐SNF transition; the ED LOS was significantly longer among patients in the ED‐to‐SNF group than among patients in the ED‐to‐Inpatient‐to‐SNF group (P < 0.0001). Specifically, the mean ED LOS in the ED‐to‐SNF group was 72.8 hours (95% CI, 59.2–86.4) compared to a mean of 14.5 hours (95% CI, 12.1–16.9) in the ED‐to‐Inpatient‐to‐SNF group (Table 2).

TABLE 2.

ED LOS at the UNC Medical Center ED is significantly longer among patients transferred directly from the ED to an SNF than among patients transferred from the ED to inpatient to an SNF.

Characteristic	ED‐to‐SNF (n 1 = 27)	ED‐to‐Inpatient‐to‐SNF (n 2 = 54)	Difference	P value
ED LOS, mean hours (SD)	72.8 (34.5)	14.48 (8.7)	58.3 (95% CI, 44.5–72.1)	<0.001

3.3. ED visit parameters

ED visit parameters from the 81 patients who experienced a home‐to‐ED‐to‐SNF transition are presented in Table 3. The reason a patient was placed in an SNF significantly differed between the ED‐to‐SNF group and the ED‐to‐Inpatient‐to‐SNF group (P = 0.002). Specifically, a smaller proportion of patients in the ED‐to‐SNF group were placed in an SNF due to inability to complete activities of daily living or advanced medical needs than the proportion of patients in the ED‐to‐Inpatient‐to‐SNF group (70.4%  vs. 92.6% and 7.4% vs. 72.2%). Patients in the ED‐to‐SNF group were also more likely to experience barriers to SNF placement than patients in the ED‐to‐Inpatient‐to‐SNF group (81.5% vs. 40.7%, P = 0.001). The most common barrier to placement encountered was the lack of availability of SNF beds (59.9% of patients in the ED‐to‐SNF group encountered this barrier).

TABLE 3.

Parameters for eligible patients’ visits to the UNC medical center emergency department from March 1, 2020, to March 1, 2022 (N = 81).

	ED‐to‐SNF (n 1 = 27), No. (%)	ED‐to‐Inpatient‐to‐SNF (n 2 = 54), No. (%)	P value
Pertinent past medical history a
No
Yes	27 (100.0)	54 (100.0)
Dementia	6 (22.2)	9 (16.7)	0.5
Prior cerebrovascular accident b	4 (14.8)	9 (16.7)	1.0
Cancer	7 (25.9)	14 (25.9)	1.0
Frequent falls	6 (22.2)	18 (33.3)	0.3
Diabetes	10 (37.0)	24 (44.4)	0.5
Hypertension	18 (66.7)	39 (72.2)	0.6
Heart disease	15 (55.6)	28 (51.9)	0.8
Other	18 (66.7)	34 (63.0)	0.7
Reason for SNF placement a
Inability to complete ADLsc	19 (70.4)	50 (92.6)	0.002
Lack of adequate home care	9 (33.3)	14 (25.9)
Advanced medical needs	2 (7.4)	39 (72.2)
Other		2 (3.7)
ED arrival day
Sunday	6 (22.2)	6 (11.1)	0.2
Monday	3 (11.1)	7 (13.0)
Tuesday	5 (18.5)	6 (11.1)
Wednesday	5 (18.5)	7 (13.0)
Thursday	1 (3.7)	14 (26.0)
Friday	5 (18.5)	7 (13.0)
Saturday	2 (7.4)	7 (13.0)
Hours spent in ED by arrival day, mean (SD) d
Sunday	74.8 (26.4)	16.4 (8.3)
Monday	48.8 (10.0)	13.0 (8.5)
Tuesday	56.6 (14.9)	18.4 (13.0)
Wednesday	60.7 (49.8)	19.7 (11.7)
Thursday	117.7 (0.0)	14.5 (7.3)
Friday	109.0 (30.0)	7.7 (2.9)
Saturday	60.7 (11.4)	12.5 (4.8)
ED primary diagnosis e
Nervous system	1 (3.7)	3 (5.6)
Ear, nose, mouth, throat	1 (3.7)	1 (1.9)
Respiratory system		1 (1.9)
Circulatory system		1 (1.9)
Digestive system		3 (5.6)
Hepatobiliary system, pancreas		1 (1.9)
Musculoskeletal, connective tissue	20 (74.1)	27 (50.0)
Skin, subcutaneous tissue, breast	1 (3.7)	1 (1.9)
Endocrine, nutritional, metabolic		4 (7.4)
Kidney, urinary tract	3 (11.1)	5 (9.3)
Blood, immunologic		1 (1.9)
Infectious, parasitic		5 (9.3)
Alcohol use, drug use		1 (1.9)
Health status, health services	1 (3.7)
Initial bed type b
Private	9 (33.3)	20 (37.0)	0.8
Private with window	2 (7.4)	4 (7.4)
Curtain, low traffic	1 (3.7)
Trauma bay room	2 (7.4)	7 (13.0)
Curtain, high traffic	13 (48.2)	22 (40.7)
Hallway beds		1 (1.9)
Patient moved rooms f
No	6 (22.2)	38 (70.4)	<0.0001
Yes	21 (77.8)	16 (29.6)
Moved twice	8 (29.6)	3 (5.6)
Moved 3 times	4 (14.8)	0 (0.0)
Moved 4 times	1 (3.7)	0 (0.0)
Secondary bed type d
Private	21 (61.8)	12 (63.2)
Private with window	3 (8.8)
Curtain, low traffic
Trauma bay room	1 (2.9)	2 (10.5)
Curtain, high traffic	6 (17.6)	3 (15.8)
Hallway beds	3 (8.8)	2 (10.5)
Hours spent in each bed type, mean (SD)
Desirable
Private	36.71 (12.0)	10.99 (2.9)	<0.001
Private with window	0.54 (12.9)	9.99 (1.7)	0.0004
Curtain, low traffic h	48.44 (5.5)	18.98 (7.2)	0.0007
	7.30 (0.0)
Undesirable
Trauma bay room	24.31 (6.0)	8.82 (2.3)	<0.001
Curtain, high traffic	6.98 (0.0)	8.34 (3.4)	0.258
Hallway beds h	29.20 (6.9)	9.38 (2.0)	<0.001
	2.64 (0.0)	5.55 (0.0)
Hours spent in waiting room g
Mean (SD)	1.94 (2.7)	1.11 (1.8)	0.7
SNF transfer day
Sunday			0.9
Monday	4 (14.8)	6 (11.1)
Tuesday	5 (18.5)	12 (22.2)
Wednesday	6 (22.2)	10 (18.5)
Thursday	7 (25.9)	10 (18.5)
Friday	4 (14.8)	13 (24.1)
Saturday	1 (3.7)	3 (5.6)
Subspecialty consults a , f
No
Yes	10	18 (33.3)	0.7
Medical admissions officer	(37.0)17.7)	36 (66.7)
Psychiatry	1 (5.9)	1 (2.8)
Other	14 (82.4)	36 (100.0)
Barriers to SNF placement a , f
No	5 (18.5)	32 (59.3)	0.001
Yes	22 (81.5)	22 (40.7)
Complex care needs	2 (9.1)	5 (22.7)
Family/patient preferences	8 (36.4)	10 (45.5)
Insurance issues	4 (18.2)	4 (18.2)
Hospital factors	6 (27.3)	1 (4.6)
Lack of SNF bed availability	13 (59.1)	6 (27.3)
Transport delays	2 (9.1)	2 (9.1)
Patient COVID‐19–positive	2 (9.1)	2 (9.1)
Other	9 (40.9)	3 (13.6)
Home medication ordered in ED c
Yes	22 (81.5)	10 (18.5)	<0.0001
No	5 (18.5)	44 (81.5)
Hospital bed ordered in ED b
Yes	7 (25.9)	2 (3.7)	0.003
No	20 (74.1)	52 (96.3)
PT/OT consult in the ED
No	2 (7.4)	47 (87.0)	<0.001
Yes	25 (92.6)	7 (13.0)

Note: Variables that were significantly different between those transferred directly to a SNF from the ED and those transferred from the ED to inpatient to an SNF are in bold.

Abbreviations: ADL, activities of daily living; OT, occupational therapy; PT, physical therapy.

^a Patients can have multiple pertinent past medical histories, reasons for SNF placement, subspecialty consults, and barriers to SNF placement.

^b Fishers exact test.

^c Also includes decreased mobility.

^d Time spent in the ED by arrival day and secondary bed type were included for descriptive purposes only, thus no statistical test was done.

^e Greater than 50% of categories have no observations, thus statistical difference was not calculated.

^f Analyzed as a binary variable (Yes/No).

^g Wilcoxon‐rank sum test.

^h No observations and/or standard deviation in 1 of the 2 groups, thus statistical difference was not calculated.

Patients within the ED‐to‐SNF group were more likely to experience a room move whereas in the ED than patients in the ED‐to‐Inpatient‐to‐SNF group. A total of 77.8% of the ED‐to‐SNF group moved rooms, as opposed to only 29.6% of the ED‐to‐Inpatient‐to‐SNF group (P < 0.0001). Within the ED‐to‐SNF group, 8 of 27 patients (29.6%) spent time in a third bed, 4 of 27 patients (14.8%) spent time in a fourth bed, and 1 of 27 patients (3.7%) spent time in a fifth bed. Among those who moved rooms, the majority of patients in both the ED‐to‐SNF group and ED‐to‐Inpatient‐to‐SNF group had a private room as their secondary bed type (61.8% and 63.2%) followed by curtain high‐traffic beds (17.6% and 15.8%).

Patients in the ED‐to‐SNF group spent, on average, significantly longer times in private rooms with a window and curtain high‐traffic beds, compared to patients in the ED‐to‐Inpatient‐to‐SNF group (P = 0.0007 and P < 0.001). Whereas, patients in the ED‐to‐Inpatient‐to‐SNF group spent significantly longer time in private rooms (P = 0.0004). In general, patients in the ED‐to‐SNF group spent an average of 24.3 hours (95% CI, 21.8–26.8) in undesirable beds (ie, trauma bay beds, curtain high‐traffic beds, and hallway beds) compared to ED‐to‐Inpatient‐to‐SNF patients, who spent an average of 8.8 hours (95% CI, 8.0–9.6) in undesirable beds (P < 0.001). This difference was driven by the fact that patients in the ED‐to‐SNF group spent an average of 29.2 hours (95% CI, 25.9–32.5) in curtain high‐traffic beds compared to patients in the ED‐to‐Inpatient‐to‐SNF group who spent an average of 9.4 hours (95% CI, 8.6–10.2) in the same bed type.

3.4. Visit quality and social determinants of health for ED‐to‐SNF patients

Among patients with an ED‐to‐SNF transition type, the quality of their visit to the UNCH ED was investigated. A total of 93% of patients who were transferred directly to a SNF from the ED received a physical therapy and/or occupational therapy consultation while in the ED. A total of 81% and 26% of patients in the ED‐to‐SNF group had home medications and hospital beds ordered while in the ED. Additionally, several SDOH were investigated within the ED‐to‐SNF group: level of education, risk of financial resource strain, presence of secure housing, food insecurity, family and social support, and lack of transportation. There was not a significant difference between any of these determinants and those patients who had an ED LOS less than the median 69.3 hours compared to those who had an ED LOS greater than or equal to 69.3 hours (Tables 3 and  4).

TABLE 4.

Within the ED‐to‐SNF group, there was not a statistically significant difference between any of the social determinants of health and those patients who had an ED LOS <69.3 hours compared to those who had an ED LOS ≥69.3 hours.

Social determinant of health	ED LOS <69.32 hours (n 1 = 14), No. (%)	ED LOS >69.32 hours (n 2 = 13), No. (%)	P value
Education
Less than high school degree	2 (25.0)	1 (16.7)	1.0
High school degree	1 (12.5)	1 (16.7)
Some college/Associate degree	1 (12.5)	1 (16.7)
Bachelor's degree	3 (37.5)	2 (33.3)
Advanced degree	1 (7.1)	1 (16.7)
Missing	6	7
Financial resource strain
None/low risk	10 (71.4)	7 (53.9)	0.6
Medium risk	3 (21.4)	4 (30.8)
High risk	1 (7.1)	2 (15.4)
Secure housing
Yes	14 (100.0)	12 (92.3)	0.5
No	0	1 (7.7)
Food insecurity
Yes	0	2 (15.4)	0.5
No	13 (100.0)	11 (84.6)
Missing	1
Family and social support
Yes	13 (92.9)	11 (84.6)	0.6
No	1 (7.1)	2 (15.4)
Lack of transportation
No	11 (84.6)	11 (91.7)	1.0
Yes	2 (15.4)	1 (8.3)
Missing	1	1
Percent of zip code in poverty
13.70 or less	9 (64.3)	6 (46.2)	0.3
13.71 or greater	5 (35.7)	7 (53.9)

Note: All variables were analyzed with a Fisher exact test, with the exception of percent of zip code in poverty.

4. LIMITATIONS

There are multiple limitations to our study. Chart review is subject to selection bias by data extractors who were not blinded to the study purpose. Chance reporting of SDOH is inaccurate and there were many elements of SDOH that were lacking in the chart. This is due to inconsistent practices in collecting and documenting this information at the clinic level. Additionally, this is a small single‐center study and results may not be generalizable to other institutions or health systems. The sample sizes for the secondary analysis within the ED‐to‐SNF group were small, and therefore the analysis was more susceptible to random error; we may have seen significant results had the analysis been conducted on a larger sample size. Checking the normality of data is also hindered by a small sample size, so t tests for hours spent in each bed type, with the exception of private rooms, should be interpreted with caution.

5. DISCUSSION

ED crowding is the result of multiple factors, and the ED is the canary in the coal mine for new stressors on the US Healthcare system. ¹³ For many years, when a patient was unable to be discharged from the ED to a safe disposition, they would be admitted to an inpatient hospital bed for skilled nursing facility placement. This placement required a 3‐day qualifying stay in the hospital for Medicare recipients. This was problematic when the patient did not have a qualifying admitting diagnosis. This may be due to administrative pressure related to the anticipated lack of reimbursement for those without a qualifying diagnosis. After the CMS waiver of the 3‐day rule, patients without safe disposition and qualifying admission diagnoses must board in the ED, awaiting care management arrangements for transfer to an appropriate facility.

Although the 2020 CMS Emergency waiver of the 3‐day inpatient stay rule for SNF placement was intended to relieve pressure on hospital inpatient censuses, it paradoxically increased ED boarding by patients who did not meet strict admission criteria and were unsafe for disposition home. These patients are subject to the lengthy process of SNF placement directly from the ED as opposed to the relative comfort of inpatient hospital admission. Although the traditional meaning of the term “ED boarding” referred to patients who were admitted awaiting an inpatient bed, ED‐to‐SNF transition patients are a new category of ED boarders. ¹⁴

In our study, as expected, the ED‐to‐SNF group experienced a much longer ED LOS when compared to the ED‐to‐Inpatient‐to‐SNF group. We found no significant difference in SDOH between the ED‐to‐SNF group versus the ED‐to‐Inpatient‐to‐SNF group. In other words, patients boarding in the ED for an average of 72.8 hours for SNF transfer were not uninsured or marginalized individuals. There is sparse research related to this subject. Our study was limited by the incomplete availability of SDOH documentation in the chart. Future prospective studies of ED‐to‐SNF transition cases in which thorough SDOH data are prospectively collected may detect differences.

A 2018 pilot study abstract described cost savings when the 3‐day inpatient stay waiver was applied to 12 qualifying ACO patients who were transferred from ED directly to SNF who were participating in network. ¹⁵ However, the authors did not describe the costs of ED boarding, and indirect costs due to lack of bed turnover and increased wait times. The process is more difficult when the SNFs are not part of the hospital network or participating in a cooperative referral program. The median ED visit LOS at our institution during the study period was 5 h 25 min, thus, rooms housing individuals awaiting nursing home placement could have been “turned over” multiple times.  Because of staff shortages and adherence to safe nurse‐patient ratios, space within the ED is limited. ED boarding effectively reduces the size of the ED. ¹³ Long occupancy of acute care rooms means that others must wait in the waiting room and other sub‐optimal spaces such as stretcher triage and hallways.

A long stay in the ED affects a patients privacy, comfort, and health. ¹³ Negative impacts on the individual patient have been well documented. Patients who stay in the ED while awaiting SNF placement are moved from room to room, sometimes as many as four times. These moves may be initiated to improve privacy and comfort, may be required by the needs of other patients for specialized rooms, and some moves may have been necessary due to staffing as parts of the department close at various hours. We found that our patients spent significant amounts of time in beds with less‐than‐optimal conditions.

The patient experience is negatively affected when time is spent in beds with only curtain barriers or in hallways. Patients are subject to noise levels that typify the ED, including sounds of other patients in pain, alarms, staff conversation, and overhead pages and announcements. Those housed in the trauma bays are exposed to disturbing sounds of trauma and medical code resuscitation. Elderly and debilitated patients who have long stays on ED stretchers are at risk of pressure injury to the skin or exacerbation of existing decubiti. Exposure to 24/7 lighting may contribute to sleep deprivation, disorientation, and delirium. Our ED is actively working to improve the care of long‐stay patients by providing hospital beds in place of stretchers, home meds, and DVT prophylaxis. These mitigation efforts are important but do not address the core problem of long‐stay boarding.

Although our study of ED to SNF transition boarders involved a relatively small number, we describe a new group of boarders that exacerbate ED crowding due to very long LOS. It is unclear how hospitals will deal with this group of patients when the CMS waiver is lifted on May 11, 2023. ¹⁶ Many EDs across the country have been reporting all‐time high levels of ED boarding due to multi‐factorial issues such as staffing shortages and influxes of patients. ¹⁷ ED boarding has a profound effect on ED operations and ultimately the care of all ED patients. This phenomenon needs more study in a broader context to fully understand the problem, its impact, and potential solutions.

AUTHOR CONTRIBUTIONS

Alec P. Tolentino, Kelli S. Gaus, and Eugenia B. Quackenbush conceived the study, designed the trial and data collection methods, and submitted an institutional review board proposal. Alec P. Tolentino conducted chart review and collected the data, with assistance from Yingqiu Gao. Kelli S. Gaus designed the statistical methods and analyzed the data. Alec P. Tolentino, Kelli S. Gaus, and Eugenia B. Quackenbush drafted the manuscript, and all authors contributed to its revision. Jane Helen Brice provided oversight and advice throughout the duration of the study. Alec P. Tolentino takes responsibility for the article as a whole.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Biography

Alec P. Tolentino, BA, is a clinical trial assistant at the University of North Carolina at Chapel Hill, Chapel Hill, NC.

Tolentino AP, Gaus KS, Gao Y, Chronowski KJ, Brice JH, Quackenbush EB. Transition of care from the emergency department to skilled nursing facility: Retrospective case‐control study. JACEP Open. 2023;4:e13022. 10.1002/emp2.13022