Emergency Department Utilization Among Deaf and Hard-of-Hearing Patients: A Retrospective Chart Review

Abstract

Background.

Deaf and hard-of-hearing (DHH) patients are an underserved priority population. Existing, although contextually limited, findings indicate that DHH patients are more likely to use the emergency department (ED) than non-DHH patients. However, little attention has been given to the differences in ED utilization by patients’ language modality.

Objective/Hypothesis:

We hypothesized that DHH ASL-users and DHH English-speakers would have higher rates of ED utilization in the past 36 months, than non-DHH English-speakers.

Methods.

We used a retrospective chart review design using data from a large academic medical center in the southeastern United States. In total, 277 DHH ASL-users, 1,000 DHH English-speakers, and 1,000 non-DHH English-speakers were included. We used logistic regression and zero-inflated modeling to assess relations between patient segment and ED utilization in the past 12- and 36-months. We describe primary ED visit diagnosis codes using AHRQ Clinical Classifications Software.

Results.

DHH ASL-users and DHH English-speakers had higher adjusted odds ratios of using the ED in the past 36-months than non-DHH English-speakers (aORs = 1.790 and 1.644, respectively). Both DHH ASL-users and DHH English-speakers had higher frequency of ED visits, among patients who used the ED in the past 36-months (61.0% and 70.1%, respectively). The most common principal diagnosis code was for abdominal pain, with DHH English-speakers making up over half of all abdominal pain encounters.

Conclusions.

DHH ASL-users and DHH English-speakers are at higher risk of using the ED compared to non-DHH English-speakers. We call for additional attention on DHH patients in health services and ED utilization research.

Introduction

People who are deaf or hard-of-hearing (DHH) are members of a priority population for health services research,¹ representing approximately 15–17% (or between 49 and 55 million) of people in the United States.^2,3 The DHH population can be segmented by a variety of characteristics such as the etiology and age of onset of hearing loss, cultural affiliation, or language modality.^3,4 In recognizing that the DHH population is heterogenous, it is necessary to consider this heterogeneity when assessing health outcomes.

Language modality is an important consideration for socioeconomic position, and associated health outcomes, within the DHH population. In the United States, DHH individuals are predominantly English-speakers or American Sign Language (ASL) users. A majority of DHH English-speakers are people with age-related hearing loss.⁵ Due to the age of onset of hearing loss, these individuals likely have stronger English-proficiency and, therefore, better access to social resources and skills required to navigate the English-dominated healthcare environment. DHH people who use ASL to communicate typically have younger ages of onset and, like all DHH children with young age of onset, are at high risk of early childhood experiences detrimental to health promoting behavior and healthcare navigation (e.g., language acquisition/deprivation and lack of exposure to incidental learning). ^6–8 These sociomedical differences in the DHH experience likely contribute to the differences in health outcomes among DHH ASL-users and English-speakers when compared to non-DHH English-speakers.

ED Utilization Among DHH Patients

Descriptive, self-report studies indicate that 56% of DHH ASL-users in Florida used the emergency department (ED) during the past year (in 2018),⁹ and 16% of DHH ASL-users in Rochester, NY used the ED two or more times in the past year (in 2013).¹⁰ In 2015, McKee and colleagues conducted the most comprehensive study of ED utilization among DHH ASL-users, to date.¹¹ In their chart review study, when adjusting for socio-demographic characteristics, DHH ASL-users in Rochester, NY had 1.6 times higher odds of using the ED in the past 36 months than their non-DHH peers. The data from Rochester, NY, however, are contextually unique when compared to other locations in the United States: DHH people in Rochester may have higher socioeconomic opportunity (due to the presence of linguistically accessible primary, secondary, and post-secondary education opportunities) and access to DHH and non-DHH healthcare providers who are fluent in ASL. Concordant communication among DHH patients and their providers increases preventive health behavior and improves patient health outcomes.¹² When considering DHH English-speakers, however, little information is available; existing research indicates that adults with hearing loss, particularly untreated hearing loss, have increased risk for using the ED.^13–15 However, to our knowledge, there are currently no other epidemiologic studies assessing ED utilization outcomes among this population. Such epidemiological studies would further detect hypothesized health disparities experienced by DHH people and inform the development of health promotion and quality improvement opportunities. Further, understanding the conditions for which DHH patients access ED care can help define health promotion program needs, particularly related to injury prevention.

Goals of This Investigation

To address the limitations in prior work in this area, the goal of the present study was to examine inequities with ED utilization outcomes among DHH patients compared to non-DHH English-speaking patients. First, we assessed ED utilization among DHH patients by segmenting across patient language categories (i.e., ASL-users and English-speakers) in comparison to their non-DHH English-speaking counterparts. We hypothesized that DHH ASL-users (H1a) and DHH English-speakers (H1b) would have higher odds of using the ED than non-DHH English-speakers. Then, we advanced previous approaches by using a modeling technique to better understand ED utilization rates, modeling the frequency of ED encounters. We hypothesized that DHH ASL-users (H2a) and DHH English-speakers (H2b) would have higher rates of ED utilization than non-DHH English-speakers. Finally, we conducted an exploratory aim to describe the conditions for which DHH patients seek ED care – an aim not yet assessed in the peer-reviewed literature.

Methods

Study Design and Setting

This study employed retrospective chart review methodology to gather patient ED utilization data from a large academic medical center in the southeastern United States. The medical center operates primary care, specialty clinics, and hospitals in the region, including an adult ED designated as a Level 1 Trauma Center and two free-standing full-service EDs. The setting is different than the Rochester, NY context with respect to characteristics associated with better socio-economic position and health outcomes among DHH ASL-users: the present setting had a smaller DHH ASL-using community, fewer ASL/English interpreters (and no on-staff ASL interpreters in the staff), and no healthcare providers who are ASL-fluent.^12,16,17 Upon approval of study methods from the University of Florida’s Institutional Review Board, all patient data were extracted from the electronic health record (EHR) by an honest broker affiliated with the medical center.

Selection of Participants

This study was an aim of a larger investigation of ED utilization among DHH patients; given the aim of the parent study, the sampling frame were patients within the academic medical center’s EHR system, between June 1, 2011, and April 3, 2020. Our analytic sample was extracted using the following strategy: 100% of DHH ASL-using patients (n = 277), 1,000 randomly sampled DHH English-speaking patients, and 1,000 randomly sampled non-DHH English-speaking patients. DHH ASL-users were identified based on their EHR language status (i.e., sign language) and ICD-9-CM/ICD-10-CM hearing loss diagnostic codes. DHH English-speakers were identified based on English language status and ICD-9-CM/ICD-10-CM diagnostic codes indicating hearing loss. A sample size calculation/power analysis was not conducted prior to sampling. Consistent with the sample size justifications described by Lakens,¹⁸ our sample size was informed by previous research on this topic conducted by McKee and colleagues, who included a total of 400 patients (200 DHH ASL-users and 200 non-DHH English-speakers).¹¹

As previously mentioned, age-related hearing loss is a cause of becoming DHH, with the prevalence of DHH people increasing with each age decade.⁵ Therefore, it is necessary to consider age as a confounder between the patient segment variable and health outcomes. To account for the confounding effect of age, the honest broker implemented an age-matching procedure: matching the proportion of patients within age strata to the DHH ASL-using patient group.

Measurements

As data were extracted by an honest broker, charts did not need to be abstracted and coded by the research team. Therefore, no data collection forms were used. (A list of the measures used in this study and their coding is available in the Web Appendices.)

Outcome and predictor measures.

The primary outcome variable was ED utilization in the past 12- and 36-months within the sampled health system. Two separate variables were defined for these outcomes: (1) a binary variable indicating ED utilization during the specified timeframe (consistent with McKee and colleagues’ study¹¹), and (2) a frequency variable with the number of ED encounters for each patient during the timeframe (an outcome not assessed in the present literature). The predictor measure was patient DHH status (i.e., DHH ASL-user, DHH English-speaker, or non-DHH English-speaker); DHH patient groups were each compared to non-DHH English-speakers.

Other variables.

Covariate selection was informed by previous ED utilization research, McKee and colleagues’ study, and a conceptual model describing ED utilization among DHH patients.^8,11,19,20 These variables were: gender, race/ethnicity, insurance payer, smoking status, age, and health status (as measured using the 1997 version of the Charlson Comorbidity Index [CCI]).²¹ A description of variable coding is available in Table S1.

Analysis

Primary analysis.

Descriptive statistics (e.g., frequency, means, etc.) were estimated using SAS version 9.4. The first aim sought to address the question of if DHH ASL-users and DHH English-speakers have higher odds of using the ED in the past 12- and 36-months than non-DHH patients. Logistic regression assumptions were tested, and multicollinearity was deemed non-problematic (all variance inflation factors < 1.250). Logistic regression models were estimated in Mplus version 8 using full information maximum likelihood (FIML) with robust standard errors and Monte Carlo integration to account for missing data.²² Previous studies have not accounted for health status as a covariate when modeling ED utilization among DHH patients.¹¹ We added health status as a covariate and assessed model fit between a null model (without health status) and expanded model. Models were compared using Satorra-Bentler chi-square testing procedures.²³ We report the adjusted odds ratios of patient segment and covariates on ED utilization in the past 12- and 36-months. We then extended the logistic regression model to use two-part modeling to model the frequency outcome. We used a negative binomial hurdle model to account for zero-inflation and overdispersion of the outcome, and ambiguity of the zero-generation process.²⁴ The 12-month ED count model did not converge.

Pooled and sensitivity analyses.

We used the R statistical package meta to conduct pooled fixed and random effects analysis using logistic regression results from McKee and colleagues’ study¹¹ and the current study. These two studies represent the only research available, to date, using a non-self-reported measure of ED utilization among DHH ASL-users. Therefore, a pooled analysis will provide a stronger estimate of the association between DHH ASL-user status (compared to non-DHH English-speakers) and ED utilization.

Prior to pooled analysis we assessed clinical and methodological heterogeneity. Clinical heterogeneity, focused on variation of the populations studied (i.e., DHH patients) and outcomes measured (ED utilization), was deemed acceptable given the intentional alignment between the present study’s methods and that of McKee and colleagues’ study,¹¹ and operational definition of DHH ASL-user. Methodological heterogeneity, focused on risk of bias and study design, was also acceptable; the studies were designed using the same methods, just at a different site.^25,26 Although clinical and methodological heterogeneity cannot be empirically tested, inacceptable heterogeneity is expected to influence statistical heterogeneity which is empirically tested.

In addition, we re-ran models to perform a sensitivity analysis, reducing the strict inclusion criteria for DHH ASL-users; specifically, removing the requirement of ICD-9-CM/ICD-10-CM hearing loss diagnosis as these individuals may not have a hearing loss diagnosis in their medical record. This led to the inclusion of 97 additional individuals.

ED encounter diagnoses.

We classified the first-listed diagnosis codes of all ED encounters using the Agency for Healthcare Research and Quality’s (AHRQ) Clinical Classification System (CCS); specifically, we used the 2015 ICD-9-CM CCS and the beta ICD-10-CM CCS.^27,28 CCS collapses ICD-9-CM and ICD-10-CM diagnostic codes into relevant categories that can be rank ordered and analyzed descriptively. In our case, we analyzed these data by patient segment (i.e., DHH ASL-users, DHH English-speakers, or non-DHH English-speakers) and patient age at the time of encounter. For this analysis, we used age stratification consistent with federal reports.^29,30

Results

Sample characteristics are described in Table 1. Demographic characteristics were similar across patient segments, except for insured status; a greater proportion of DHH ASL-users were Medicare insured. A greater proportion of DHH English-speaking and ASL-using patients used the ED than non-DHH English-speakers in the past 12- and 36-months; the effect sizes of these relations were small (Cramer’s Vs ≤ 0.077).

Table 1.

Descriptive characteristics.

Characteristic	DHH ASL-users (n = 277)	DHH English-speakers (n = 1,000)	non-DHH English-speakers (n = 1,000)
Mean age (SD), range	48.332 (17.055) Range: 18 to 94	48.421 (17.786) Range: 18 to 106	48.230 (17.456) Range: 18 to 101
Gender
Woman	58.484% (162)	54.300% (543)	56.000% (560)
Man	41.516% (115)	45.700% (457)	44.000% (440)
Race*
White	63.158% (168)	77.372% (742)	74.038% (693)
African American	24.060% (64)	12.304% (118)	16.026% (150)
Asian	1.504% (4)	1.773% (17)	2.244% (21)
Pacific Islander	None	None	0.214% (2)
Indigenous American	0.376% (1)	0.209% (2)	0.427% (4)
Hispanic	0.376% (1)	0.209% (2)	0.321% (3)
Multiracial or other	10.526% (28)	8.133% (78)	6.731% (63)
Hispanic ethnicity*
Unknown	4.332% (12)	3.800% (38)	6.800% (68)
Hispanic	8.664% (24)	5.000% (50)	6.000% (60)
Insurance payer*
Uninsured	3.971% (11)	8.408% (84)	26.497% (261)
Private	16.606% (46)	47.648% (476)	40.508% (399)
Medicaid	26.354% (73)	17.317% (173)	11.980% (118)
Medicare	49.458% (137)	21.922% (219)	17.766% (175)
Other	3.610% (10)	4.705% (47)	3.249% (32)
Smoking status
Unknown	5.054% (14)	3.600% (36)	19.700% (197)
Current	11.552% (32)	13.400% (134)	13.800% (138)
Prior	18.773% (52)	24.900% (249)	17.000% (170)
Never	64.621% (179)	58.100% (581)	49.500% (495)
Emergency department utilization
Past 12 months	10.830%	11.200%	7.000%
Past 36 months	24.910%	22.900%	16.300%

^*Some cases missing.

ED Utilization Models

Model comparisons – using Satorra-Bentler chi-square testing – indicated that the expanded logistic regression model, including CCI (i.e., health status) as a covariate, fit the binary utilization outcome data better than the model excluding it for ED utilization in the past 12-months. Model fit was not significantly better for the binary ED utilization outcome in the past 36-months. Table 2 reports the adjusted odds ratios of ED utilization for the expanded model. DHH English-speaking patients, but not ASL-users, had higher odds than non-DHH English-speaking patients of using the ED in the past 12-months (aOR = 1.790; 95% CI: 1.287 to 2.489). In the past 36-months, both DHH ASL-users (aOR = 1.644; 95% CI: 1.176 to 2.298) and DHH English-speakers (aOR = 1.608; 95% CI: 1.274 to 2.030) had higher adjusted odds of using the ED compared to non-DHH English-speakers.

Table 2.

Crude and adjusted odds of ED utilization among DHH ASL-users and DHH English-speakers, past 12- and 36-months.

Characteristic	Past 12 months		Past 36 months
	OR	aOR	OR	aOR
Patient segment (ref. non-DHH English-speakers)
DHH ASL-users	1.614 (1.029 to 2.531)	1.576 (0.988 to 2.514)	1.703 (1.237 to 2.345)	1.644 (1.176 to 2.298)
DHH English-speakers	1.676 (1.226 to 2.290)	1.790 (1.287 to 2.489)	1.525 (1.220 to 1.907)	1.608 (1.274 to 2.030)
Gender (ref. Men)
Women	1.166 (0.875 to 1.553)	1.132 (0.845 to 1.516)	1.010 (0.822 to 1.241)	0.985 (0.798 to 1.218)
Race (ref. White)
Black	2.315 (1.663 to 3.223)	2.232 (1.586 to 3.139)	2.460 (1.903 to 3.182)	2.396 (1.842 to 3.114)
Other	0.506 (0.263 to 0.976)	0.455 (0.234 to 0.884)	0.897 (0.617 to 1.303)	0.849 (0.579 to 1.246)
Payer (ref. insured, non-Medicaid)
Uninsured or Medicaid	1.609 (1.206 to 2.148)	1.490 (1.085 to 2.048)	1.313 (1.060 to 1.627)	1.149 (0.911 to 1.449)
Smoking status (ref. Never smokers)
Current	1.397 (0.940 to 2.075)	1.338 (0.883 to 2.026)	1.570 (1.182 to 2.087)	1.624 (1.212 to 2.179)
Prior	1.246 (0.881 to 1.763)	1.486 (1.007 to 2.192)	1.046 (0.806 to 1.358)	1.219 (0.919 to 1.619)
Age	0.988 (0.980 to 0.995)	0.990 (0.981 to 0.999)	0.989 (0.983 to 0.994)	0.990 (0.983 to 0.996)
Health status (Charlson Comorbidity Index 1987)	0.776 (0.578 to 1.042)	0.838 (0.704 to 0.997)	0.934 (0.840 to 1.039)	0.962 (0.868 to 1.065)

Note. Adjusted models estimated in Mplus using full information maximum likelihood with Monte Carlo integration to account for missing data. Analytic n = 2,277. DHH = Deaf and Hard-of-Hearing; ASL = American Sign Language.

Next, we modeled the frequency of ED encounters using zero-inflated modeling. The pattern of results of the zero versus non-zero model were the same as the previously interpreted logistic regression model (see Table 3). The count model, however, provides information on the rate of ED utilization encounters for patients. Results indicate that, among users, the number of ED encounters in the past 36 months increased for DHH English-speaking and DHH ASL-using patients when compared to non-DHH English-speakers (by 61.0% and 70.1%, respectively).

Table 3.

Negative binomial hurdle model results of count ED utilization frequency among DHH ASL-users, DHH English-speakers, and non-DHH English-speakers.

Characteristic	ED utilization – past 36-months
	Count model incident rate ratio	Zero model odds ratio
Patient segment (ref. non-DHH English-speakers)
DHH ASL-users	1.701 (1.037 to 2.790)	1.644 (1.176 to 2.298)
DHH English-speakers	1.610 (1.129 to 2.293)	1.608 (1.274 to 2.030)
Gender (ref. Men)
Women	1.174 (0.85 to 1.621)	0.985 (0.798 to 1.218)
Race (ref. White)
Black/African American	1.839 (1.307 to 2.588)	2.396 (1.842 to 3.114)
Other	0.710 (0.381 to 1.323)	0.849 (0.579 to 1.246)
Payer (ref. insured, non-Medicaid)
Uninsured or Medicaid	1.702 (1.231 to 2.351)	1.149 (0.911 to 1.449)
Smoking status (ref. Never smokers)
Current	1.358 (0.931 to 1.982)	1.624 (1.212 to 2.179)
Prior	1.380 (0.89 to 2.14)	1.219 (0.919 to 1.619)
Age	0.993 (0.984 to 1.003)	0.990 (0.983 to 0.996)
Health status (Charlson Comorbidity Index 1987)	1.053 (0.977 to 1.137)	0.962 (0.868 to 1.065)

Note. Models implemented with full information maximum likelihood (FIML) to account for missing data. DHH = Deaf and hard-of-hearing. ASL = American Sign Language.

Pooled and sensitivity analyses.

Clinical and methodological heterogeneity were deemed acceptable for pooled effects analysis, and statistical heterogeneity (χ² = 0.28, p = 0.59; I² = 0%) indicated the data were acceptable for pooled analysis. Pooled analyses indicated that DHH ASL-users had higher aOR of ED utilization than non-DHH English-speakers in the past 36 months (random effects aOR = 1.72, 95% CI: 1.28 to 2.32; see Figure S1).

A sensitivity analysis assessed changes in effects based on changes to the sample definition. Specifically, we assessed if lessening the strict inclusion criteria for DHH ASL-users influenced findings. We removed the requirement to have a hearing loss diagnosis code in the medical record, to assess ED utilization among ASL-users and non-DHH English-speakers. Findings indicated that the results were sensitive to this sample definition: ASL-users did not have higher adjusted odds than non-DHH English-speakers of using the ED in past 36-months (aOR = 1.338; 95% CI: 0.981 to 1.826). Pooled effect analysis with the sensitivity analysis effect provided attenuated, but significant effects similar to our central pooled findings (i.e., pooled random effects aOR with sensitivity analysis = 1.52, 95% CI: 1.04 to 2.21). That is, even with the change in the sample definition, ASL-using patients still have higher aORs of using the ED in the past 36 months.

Encounter Conditions

Clinical classifications of primary diagnosis codes from encounters between June 1, 2011 and April 3, 2020 are described in Table 4, with age and patient segment stratified tables in Table S2 to S4. The most common principal diagnosis code was abdominal pain. Abdominal pain was the number one charted condition for 18–44-year-old patients, and was a top five condition for 45–64-year-olds patients and patients aged 65 years and older. Among all patient ages, DHH English-speakers made up over half (~52%) of all abdominal pain encounters. Other top medical conditions included nonspecific chest pain; other connective tissue disease; and, spondylosis, intervertebral disc disorders, or other back problems.

Table 4.

Most frequent principal diagnoses of ED encounters, by patient segment.

Rank	Condition	Total	DHH ASL-users	DHH English-speakers	non-DHH English-speakers
Injury
1	Superficial injury; contusion [239]	72 (2.685%)	7 (1.651%)	41 (2.806%)	24 (3.011%)
2	Sprains and strains [232]	69 (2.573%)	7 (1.651%)	36 (2.464%)	26 (3.262%)
3	Open wounds of extremities [236]	62 (2.312%)	9 (2.123%)	32 (2.190%)	21 (2.635%)
4	Other injuries and conditions due to external causes [244]	56 (2.088%)	12 (2.830%)	20 (1.369%)	25 (3.137%)
5	Open wounds of the head, neck, and trunk [235]	24 (0.895%)	2 (0.472%)	13 (0.890%)	9 (1.129%)
Medical
1	Abdominal pain [251]	205 (7.644%)	46 (10.849%)	109 (7.461%)	50 (6.274%)
2	Nonspecific chest pain [102]	167 (6.227%)	27 (6.368%)	96 (6.571%)	44 (5.521%)
3	Spondylosis; intervertebral disc disorders; other back problems [205]	122 (4.549%)	19 (4.481%)	56 (3.833%)	47 (5.897%)
4	Other connective tissue disease [211]	120 (4.474%)	24 (5.660%)	70 (4.791%)	26 (3.262%)
5	Other lower respiratory disease [133]	102 (3.803%)	11 (2.594%)	62 (4.244%)	29 (3.639%)
Mental health/substance use
1	Alcohol-related disorders [660]	21 (0.783%)	1 (0.236%)	10 (0.684%)	10 (1.255%)
2	Anxiety disorders [651]	18 (0.671%)	0 (0.000%)	15 (1.027%)	3 (0.376%)
3	Substance-related disorders [661]	6 (0.224%)	1 (0.236%)	3 (0.205%)	2 (0.251%)
4	Mood disorders [657]	5 (0.186%)	0 (0.000%)	4 (0.274%)	1 (0.125%)
5	Suicide ideation and intentional self-inflicted injury [662]	5 (0.186%)	0 (0.000%)	2 (0.137%)	3 (0.376%)
Maternal/neonatal
1	Other complications of pregnancy [181]	32 (1.193%)	3 (0.708%)	25 (1.711%)	4 (0.502%)
2	Hemorrhage during pregnancy; abruptio placenta; placenta previa [182]	7 (0.261%)	0 (0.000%)	2 (0.137%)	5 (0.627%)
3	Other complications of birth; puerperium affecting management of mother [195]	3 (0.112%)	0 (0.000%)	3 (0.205%)	0 (0.000%)
4	Spontaneous abortion [177]	3 (0.112%)	0 (0.000%)	1 (0.068%)	2 (0.251%)

Note. Principal diagnosis codes are from June 1, 2011 to April 3, 2020. Clinical classifications were derived using AHRQ’s CCS for ICD-9-CM and the beta version for ICD-10-CM. The beta version was used for ICD-10-CM to allow ranking across time with the same single-level categories used in ICD-9-CM CCS. There was no condition category for the 5th rank of maternal/neonatal conditions. DHH = Deaf and hard-of-hearing. ASL = American Sign Language. Encounter totals: Total = 2,682; DHH ASL-users = 424; DHH English-speakers = 1,461; and non-DHH English-speakers = 797.

Superficial injuries and contusions, followed closely by sprains and strains, were the top conditions for injury-related ED visits. DHH ASL-users represented fewer ED encounters for these reasons (~9–10%) than DHH English-speakers and non-DHH English-speakers (~52–59% and ~33–38%, respectively). Alcohol-related disorders and anxiety disorders were the most frequent diagnoses for mental health/substance use-related visits. DHH and non-DHH English-speakers each had 10-of-the-21 encounters for alcohol-related disorders, while 15-of-the-18 encounters for anxiety disorders were from DHH English-speaking patients. Of the 55 encounters for the top five mental health/substance use-related visits, only two encounters were from DHH ASL-users (one for alcohol-related disorders, and one for substance use-related disorders). Maternal/neonatal-related encounters were primarily for other complications of pregnancy. Of the 32 encounters in this category, 25 encounters were among DHH English-speakers aged 18–44 years old.

Discussion

Results from the logistic regression analyses supported the hypothesis that DHH ASL-users have higher adjusted odds than non-DHH English-speakers to use the ED in the past 36-months but not in the past 12 months (H1a partially supported; see Table S5). Not only did DHH ASL-users have higher odds to use the ED in the past 36-months, but, among ED users, they also had 70.1% more ED encounters than non-DHH English-speakers (H2a supported).

In extending previous research, we included DHH English-speakers in the sampling frame. Previous research indicates that DHH English-speakers have more ED visits.^13–15 Our findings supported the hypothesis that DHH English-speakers had higher odds of using the ED than non-DHH English-speakers, in the past 12- and 36-months (H2a supported); in addition, among ED users, DHH English-speakers had 62% more encounters than their non-DHH counterparts (H2b supported).

The inclusion of a measure of health status (in this case, the CCI) significantly improved model fit for the 12-month ED utilization outcome model, but not the 36-month model. This finding suggests the relative utility of including health status as a covariate in ED utilization research. We observe that patients with higher CCI scores (i.e., more and/or more serious health comorbidities) had lower odds of ED utilization; however, this finding is undoubtedly moderated by the management of the condition(s). Presence of comorbidities increases the likelihood of primary and specialist care engagement, and, in the presence of socio-behavioral antecedents (e.g., those discussed in the conceptual model⁸), better management of the health condition and lower odds of using the ED. When compared to DHH patients, non-DHH English-speakers likely have better social and healthcare navigation resources, and better patient-provider communication, leading to fewer ED visits.^8,12 Future research should use expanded measures of health status incorporating health behavior, lifestyle, and genetics to better understand differences between DHH and non-DHH people.

Interestingly, only two of the 54 encounters for mental health and substance use were among DHH ASL-users: one for alcohol-related disorders and one for substance-use disorders. This finding is surprising for several reasons. First, although any child may experience communication neglect and language deprivation, it appears more common in DHH children.⁷ In adolescence and adulthood these early childhood experiences may present as mental health symptoms associated with Language Deprivation Syndrome.³¹ Secondly, mental health is a priority for the Florida DHH ASL-using community. A community-engaged survey in 2018 found that mental health was a significant area of need for DHH ASL-users: mental health was the most reported greatest health concern, 15.5% scored at-risk for depression on the PHQ-2, and DHH ASL-users had higher likelihood of engaging in binge drinking than non-DHH English-speakers.³² Further, it is well-established that mental health facilities are inaccessible to DHH patients. The Substance Abuse and Mental Health Services Administration reports that, in 2018, 44.5% of mental health facilities in the U.S. did not provide services to DHH ASL-users through an interpreter.³³ This leads to a fundamental question of if and where DHH ASL-users are receiving treatment for emergent mental health concerns.

The separate, statistically significant, findings among DHH ASL-users and DHH English-speakers highlights the importance of considering heterogeneity among the DHH population. It is crucial to recognize the structural and institutionalized oppression against DHH individuals as a fundamental cause of deleterious social determinants of health and increase of social needs – which are likely associated with the increase in ED utilization.⁸ The experience of oppression based on DHH status and other intersectional oppression (e.g., racism) and in/accessibility to mitigating resources is related to the DHH patients’ experiences in healthcare, making it imperative to consider DHH heterogeneity when discussing this group. In addition, the social position of DHH people differs geographically. McKee and colleagues observed that a smaller proportion of DHH ASL-users were Medicare and Medicaid insured (12% and 26%, respectively), with the majority having private insurance (61%).¹¹ In this study, a higher proportion of DHH ASL-users, than DHH and non-DHH English-speakers, were Medicare and Medicaid insured (25% and 47%, respectively). Therefore, it is crucial that the expanding literature on DHH health behavior and clinical outcomes also considers the unique role of geographic, DHH-specific contexts, in addition to the diversity in health outcomes based on language modality and other DHH-specific variables such as age of onset of hearing loss.

Limitations

There is potential model misspecification due to the lack of antecedents to ED utilization available within the EHR. As indicated by the Conceptual Model of Emergency Department Utilization Among DHH Patients,⁸ there are several constructs that likely have explanatory power when examining ED utilization among the DHH population. In addition, some variables that were available in the EHR were excluded due to overwhelming missingness; for example, educational attainment (a predisposing factor) was missing for over 95% of patients. Therefore, expanding covariates to include relevant patient and non-patient variables is crucial to better understanding DHH patient ED outcomes.

In addition, we are limited by using a single healthcare system for this study. By sampling at a single site, we introduce the presence of sampling zeros (i.e., patients considered as non-ED users but who may have sought ED care elsewhere). However, many health systems do not accurately classify ASL-users.³⁴ For example, the PCORnet Common Data Model classifies sign language as ‘Other.’³⁵ This is a barrier to conducting EHR-based studies to identify health inequities among this population. The sampled academic medical center explicitly identifying patients who use ASL is a strength of this study, despite the limitation on a lack of complete healthcare record for the sampled patients. Furthermore, it is still possible that there was misclassification of the exposure: DHH ASL-users may have self-classified as DHH English-speakers, been incorrectly classified by healthcare staff as DHH English-speakers, and/or DHH English-speakers may be incorrectly classified as non-DHH if their hearing loss diagnosis is not in their medical record. Nevertheless, the process for extracting DHH codes from chart- or claims-based records has been validated and widely used.^36,37 Still, we strongly recommended larger, multi-site studies that assess ED utilization among DHH patients to account for the bias of sampling a single site (and not having a patient’s full medical record). As PCORnet partners and other EHR developers begin to promote ASL-use as a language within the medical record, these data sources will provide an opportunity to capture equity-relevant variables with a larger sample of connected medical records, providing a more complete profile of DHH patient ED use.

Conclusions

In summary, results from the present study indicate that DHH ASL-users and DHH English-speakers were more likely to use the ED and, among ED users, more likely to have multiple encounters than non-DHH English-speakers in the past 36-months. These findings align with previous research focused on DHH ASL-users and patients with hearing loss. Although these findings indicate disparities in ED utilization among DHH and non-DHH patients, more research is needed to identify the reasons why DHH patients may be at higher risk of ED use. Additional patient-centered qualitative research is currently underway to better understand reasons DHH patients seek ED care, as opposed to other sources of care.