Abstract
Objective:
To estimate the contribution of insurance on rates of inpatient admission for emergency department visits with depression diagnoses.
Methods:
We identified 3,681,581 visits for depression in the National Emergency Department Sample (2007–2018). We classified them by concurrent injury, suicidal ideation, or neither. Payer categories were defined, non-exclusively, as Medicare, Medicaid, private insurance, and no insurance. Logistic regression models, adjusted for age, year, and comorbidities, were used to describe differences in rates of inpatient admission by payer type, stratified by visit features.
Results:
Rates of inpatient admission for visits with neither injury nor suicidal ideation (31.9%; 95%CI, 30.8–33.0) were lower than for visits with injury (37.9%; 95%CI, 36.7–39.1) or with suicidal ideation (39.7%; 95%CI, 37.3–42.1). Rates of admission were significantly lower for those without insurance (26.6%; 95%CI, 25.5–27.8) than for those with insurance (37.1%; 95%CI, 36.1–38.1). In adjusted models, insurance was associated with increased likelihood (OR=1.81, 95%CI, 1.69–1.94) of admission. Insurance continued to be a significant predictor of admission among ED visits for depression with concurrent injury (OR=1.39; 95%CI, 1.29–1.51).
Conclusion:
After controlling for demographic characteristics and medical comorbidities, patients with depression who have insurance are significantly more likely to be admitted to the hospital compared to those without insurance.
Keywords: health policy, mental health services, suicide, psychiatric emergency services
1. Introduction
The decision to admit emergency department (ED) patients for inpatient care is a pivotal one in the care of individuals with depression [1]. Although most patients with depression can be managed effectively in the outpatient setting [2], worsening depression combined with multiple acute risk factors such as severe and persistent suicidal ideation with plan or intent, or even non-fatal suicide attempt, may prompt patients to present to the ED for immediate evaluation and possible psychiatric admission [3]. However, while EDs in the U.S. are required by law to stabilize patients without regard to the patient’s ability to pay [4], there is no such requirement for inpatient facilities to accept patients from the ED. As such, despite ED access, access to psychiatric inpatient services for depression may be unequal.
Some evidence suggests that payment rates may influence access to inpatient care. It is known that payment rates differ by payer even within diagnosis group and hospital readmissions are higher among patients with Medicaid [5,6], perhaps making this population less financially desirable for hospitals compared to privately-insured patients [7]. Among patients who will be admitted, ED lengths of stay may be one indicator of difficulties locating inpatient beds. Prior literature within mental health has reported long lengths of ED stay for patients presenting with psychiatric symptoms [8–12] and several studies have described that ED lengths of stay were longer still for patients who have Medicaid or are uninsured [10,11]. However, financial constraints may also affect clinical decisions: one study comparing Medicaid and private insurance patients presenting to the ED with deliberate self-harm found that Medicaid-insured patients were less likely to be admitted and, among those discharged, were less likely to have received a mental health assessment in the ED compared to privately-insured patients [13]. There is also some evidence that resource constraints can have aggregate-level effects. One study found that general inpatient hospitals experienced increased psychiatric burden after sustained, abrupt decreases in the level of local psychiatric bed supply [14].
The stakes for psychiatric admission decisions are high. Suicide is a leading cause of death in the United States [15]. For many patients with depression and suicidality, the ED remains an important point of contact [16–18]. Approximately 40% of patients who die of suicide have at least one visit to the ED over the last 12 months of their lives [19,20]. Thus, it is important to understand how ED disposition might be influenced by non-clinical factors such as payment.
To better understand whether and how financial and resource pressures may influence psychiatric admission decisions, we examined a nationally representative sample of ED visits from 2007–2018 presenting with depression. We stratified our sample by ED visit features that have been shown to be correlated with suicide risk: concurrent injury or suicidal ideation (SI) [21]. By describing rates of inpatient admission by features of clinical severity and by patient insurance status, we comment on the extent to which these non-clinical criteria may influence the clinical decision to admit. Prior to conducting these analyses, we hypothesized that payment status would be significantly correlated with inpatient admission with uninsured patients having the lowest likelihood of admission and that these correlations would be strongest for visits with lower risk, i.e., without injury or SI. Within the subset of insured patients, we hypothesized that admission rates would be positively correlated with average inpatient charges.
2. Methods
2.1. Data Sources
The 2007–2018 National Emergency Department Same (NEDS) is a stratified sample of discharges from U.S. community hospitals, describing approximately 30 million ED visits per year[22]. Strata and sample weights are provided by NEDS to produce nationally representative estimates.
2.2. Sample Assembly
Previous literature has identified a bimodal age distribution of ED visits for depression, peaking in the teenage years and again in middle age [23]. Because we wished to focus on inpatient admission for psychiatric reasons, we restricted the sample to ED visits for those aged 10–90 years from 2007–2018 to include both groups while excluding age groups for which psychiatric admission was less likely. We selected study visits by the presence of a depression diagnosis (ICD-9: 296.2x; 296.3x; 300.4; 301.12; 311) [23] in any of diagnosis codes associated with the visit. Because this sample still contained many visits with a primary diagnosis for an acute medical disorder, we further restricted our sample to only patients in this subset who presented with a primary diagnosis that was a mental health diagnosis [24] or an injury [25,26] (Table S1).
For ED visits 2015–2018 following ICD-10 transition, ICD-10-CM diagnoses were cross-walked to ICD-9-CM equivalents using the Centers for Medicare and Medicaid Services ICD-10-CM General Equivalence Mappings acquired through the National Bureau of Economic Research [27,28]. Due to the 2015 ICD-9 to ICD-10 transition, diagnosis data were found to contain inconsistencies with adjacent years and therefore 2015 data were excluded from the logistic regression models.
2.3. Dependent and Independent Variables
The primary outcome of interest was ED disposition coded as admission or discharge. Because not all hospitals have psychiatric units, psychiatric admission may require a hospital transfer [29]. This analysis considers transfers as admissions. Secondary outcomes among inpatients included inpatient length of stay and total inpatient charges. Payer variables were the primary independent variable of interest. The any Medicare indicator was defined by having either a primary or secondary payer of Medicare. Similar payer indicators were defined for Medicaid and private insurance. Any insurance was defined as including any of these payers or any other insurance, such as Worker’s Compensation, CHAMPUS, CHAMPVA, or Title V, as either a primary or secondary payer. No insurance was defined as the complement of any insurance. Other independent variables included patient sex, 10-year age group, visit year, and medical comorbidities as defined by the Elixhauser Comorbidity Index (ECI) [30–32].
2.4. Stratification
Because patients presenting with self-harm, suicide attempt, other injuries, or SI are at elevated suicide risk compared to patients presenting for mental health reasons without these features, analyses were stratified by these risk features [21]. Although deliberate self-harm is often assessed in EDs, there is evidence that under-ascertainment is prevalent [33–35]. Among our sample of visits with a discharge diagnosis of depression, we, therefore, consider three strata of risk: visits presenting with concurrent injury, visits with SI, and visits without either of these high-risk features.
2.5. Analyses
Analyses were performed in three stages. First, descriptive variables including rates of inpatient admission, length of inpatient stay conditional on admission, demographic, and comorbidity data were summarized using NEDS sample weights and hierarchically stratified by visit risk features: 1) with concurrent injury, 2) with SI, but without injury 3) without either injury or SI. Mean admission rates were summarized by these risk strata and by payer type.
Second, logistic regression was used to model the odds of admission as a function of having any insurance, insurance type, patient sex, age group, ECI comorbidities, and visit year. In these models, visits with no insurance were the reference group, captured by the any insurance variable. Payment type variables including any Medicaid, any Medicare, and any private insurance were nested within the any insurance category.
Third, we sought to better understand how the average per diem reimbursement by payer might correlate with rates of admission by analyzing per diem charges. Total charge amounts reported both for the inpatient stay and the emergency department were adjusted using annual Consumer Price Index data [36]. The total inpatient charge, excluding ED charges, was divided by inpatient length of stay to generate an average per diem charge by payer. Mean per diem charge by length of stay by payer was then computed. This allowed us to estimate the average per diem rates by payer.
All analyses were done in STATA 17.0. The New York State Psychiatric Institute Institutional Review Board deemed this study to be exempt from human subjects research review.
3. Results
3.1. Demographic characteristics
The study sample included 3,681,581 visits for depression of which 44.0% were visits with concurrent injury, 8.2% were visits with SI without injury, and 47.8% were visits with neither (Table 1). The total unweighted number of inpatient admissions across all groups was 1,159,777 (31.5%). Visits with concurrent injury were more likely to be by women and had a higher mean age compared to visits with SI or to those with neither risk feature. Diagnoses of alcohol use, drug use, or psychotic disorders were overrepresented in visits without concurrent injury or SI. Although payer status differed somewhat by risk feature, they did not vary appreciably: 82.2%, (95%CI 81.6–82.8) of visits with concurrent injury had any insurance, compared to 84.2% (95%CI, 82.7–85.6) in the group with SI, and 80.6% (95%CI, 79.8–81.4) for those without either risk feature.
Table 1.
Demographic and clinical characteristics for ED depression visits by concurrent injury, suicidal ideation, or neither, United States, 2007–2018
| Overall (N=3,681,581) | With Concurrent Injury (N=1,620,793) | No, Injury, With SI (N=300,128) | Without Injury or SI (N=1,760,660) | |||||
|---|---|---|---|---|---|---|---|---|
| % | 95% CI | % | 95% CI | % | 95% CI | % | 95% CI | |
| Admit | 35.2% | [34.2–36.1] | 37.9% | [36.7–39.1] | 39.7% | [37.3–42.1] | 31.9% | [30.8–33.0] |
| LOS, days (mean) | 5.33 | [5.24–5.43] | 4.79 | [4.71–4.87] | 5.81 | [5.60–6.02] | 5.82 | [5.68–5.96] |
| Female | 54.0% | [53.7–54.4] | 58.9% | [58.5–59.4] | 45.8% | [45.3–46.3] | 50.8% | [50.4–51.1] |
| Age | 43.90 | [43.74–44.07] | 45.91 | [45.69–46.13] | 39.30 | [39.11–39.49] | 42.79 | [42.65–42.94] |
| Any Insurance | 81.6% | [80.9–82.2] | 82.2% | [81.6–82.8] | 84.2% | [82.7–85.6] | 80.6% | [79.8–81.4] |
| Any Medicare | 24.7% | [24.3–25.0] | 28.2% | [27.8–28.6] | 18.6% | [18.1–19.0] | 22.4% | [22.0–22.7] |
| Any Private Insurance | 30.2% | [29.6–30.9] | 31.9% | [31.2–32.6] | 27.7% | [26.7–28.6] | 29.1% | [28.4–29.9] |
| Any Medicaid | 33.5% | [32.7–34.2] | 29.2% | [28.4–29.9] | 43.2% | [41.7–44.6] | 35.9% | [35.0–36.8] |
| Elixhauser Comorbidities | ||||||||
| Congestive Heart Failure | 1.9% | [1.9–2.0] | 2.6% | [2.5–2.7] | 1.3% | [1.2–1.3] | 1.4% | [1.3–1.4] |
| Cardiac Arrhythmia | 4.5% | [4.4–4.6] | 5.9% | [5.7–6.0] | 3.0% | [2.8–3.1] | 3.4% | [3.3–3.5] |
| Hypertension, uncomplicated | 24.1% | [23.8–24.5] | 27.6% | [27.2–28.0] | 19.6% | [19.1–20.1] | 21.6% | [21.2–22.0] |
| Neurological Disorder, not paralysis | 5.4% | [5.3–5.5] | 5.9% | [5.8–6.0] | 4.0% | [3.8–4.2] | 5.2% | [5.1–5.3] |
| Chronic Pulmonary Disease | 11.0% | [10.8–11.3] | 13.1% | [12.8–13.4] | 10.3% | [9.9–10.7] | 9.3% | [9.0–9.5] |
| Diabetes, uncomplicated | 8.5% | [8.3–8.6] | 10.1% | [9.9–10.3] | 6.1% | [6.0–6.3] | 7.3% | [7.2–7.4] |
| Diabetes, complicated | 1.2% | [1.2–1.3] | 1.2% | [1.1–1.3] | 2.3% | [2.2–2.4] | 1.0% | [1.0–1.1] |
| Hypothyroidism | 5.2% | [5.0–5.3] | 6.7% | [6.5–6.8] | 3.6% | [3.5–3.8] | 4.0% | [3.9–4.1] |
| Renal Failure | 1.6% | [1.6–1.7] | 2.2% | [2.1–2.3] | 1.2% | [1.1–1.3] | 1.2% | [1.1–1.2] |
| Liver Disease | 2.3% | [2.2–2.4] | 1.9% | [1.8–1.9] | 1.8% | [1.7–1.9] | 2.8% | [2.7–2.9] |
| Obesity | 3.5% | [3.4–3.6] | 4.1% | [3.9–4.3] | 4.0% | [3.7–4.3] | 2.8% | [2.7–3.0] |
| Fluid and Electrolyte Disorders | 6.5% | [6.4–6.7] | 7.3% | [7.1–7.6] | 4.4% | [4.1–4.7] | 6.1% | [5.9–6.3] |
| Alcohol Abuse | 22.1% | [21.7–22.5] | 16.9% | [16.4–17.3] | 21.1% | [20.4–21.7] | 27.2% | [26.7–27.7] |
| Drug Abuse | 19.7% | [19.2–20.1] | 15.8% | [15.3–16.3] | 27.3% | [26.5–28.2] | 21.9% | [21.5–22.4] |
| Psychoses | 7.5% | [7.3–7.7] | 4.9% | [4.7–5.1] | 8.9% | [8.4–9.3] | 9.6% | [9.3–9.9] |
| Hypertension, complicated | 1.5% | [1.5–1.6] | 1.9% | [1.8–2.0] | 1.5% | [1.5–1.6] | 1.2% | [1.1–1.2] |
| Elixhauser Weighted Sum | 2.34 | [2.32–2.35] | 2.37 | [2.35–2.39] | 2.25 | [2.22–2.28] | 2.32 | [2.30–2.34] |
National Emergency Department Sample, 2007–2018, analysis limited to those age 10–90 years. N=3,681,581. SI=suicidal ideation. LOS=Length of stay. LOS variable taken over the subset of n=1,159,777 admissions. Results based on weighted sampling.
By payer, visits with Medicare coverage had higher mean age compared to other payer groups (Table S2). Patient visits with Medicare had more ECI comorbid diagnoses and were more likely to have congestive heart failure (5.4%), hypertension (42.0%), chronic pulmonary disease (17.6%), diabetes, either complicated (17.6%) or uncomplicated (16.3%), hyperthyroidism (11.1%), and renal failure (4.9%) diagnoses compared to other groups (Figure S1). Alcohol (27.8%) or substance use comorbidities (24.6%) were overrepresented in visits with no insurance.
3.2. Admission Rates
Inpatient admission rates varied by payer group. Overall admission rates were highest for Medicare visits (43.2%) and lowest for visits without insurance (26.6%) (Table 2). Within payer category, admission rates were higher for visits with concurrent injury or with SI compared to visits without these features (Table 2). Within each risk group, Medicare was associated with the highest rates of admission and no insurance the lowest.
Table 2.
Admission rates among depression ED visits by payer, by risk features of concurrent Injury, suicidal ideation, or neither.
| Overall (N=3,681,581) |
With Concurrent Injury (N=1,620,793) |
No Injury, With SI (N=300,128) |
Without Injury or SI (N=1,760,660) |
|||||
|---|---|---|---|---|---|---|---|---|
| % | 95% CI | % | 95% CI | % | 95% CI | % | 95% CI | |
| Any Medicare | 43.2% | [42.2–44.2] | 45.5% | [44.3–46.7] | 44.1% | [41.5–46.6] | 40.4% | [39.2–41.6] |
| Any Private Insurance | 37.2% | [36.1–38.4] | 39.4% | [38.0–40.7] | 41.9% | [39.1–44.7] | 34.3% | [33.0–35.7] |
| Any Medicaid | 34.3% | [33.1–35.5] | 36.5% | [35.1–38.0] | 41.2% | [38.4–43.9] | 31.3% | [29.9–32.6] |
| No Insurance | 26.6% | [25.5–27.8] | 31.2% | [29.7–32.6] | 29.7% | [26.3–33.1] | 22.4% | [21.1–23.7] |
| Total | 35.2% | [34.2–36.1] | 37.9% | [36.7–39.1] | 39.7% | [37.3–42.1] | 31.9% | [30.8–33.0] |
National Emergency Department Sample, 2007–2018, analysis limited to those age 10–90 years. N=3,681,581. SI=suicidal ideation. Any Medicare, Any Private and Any Medicaid are not mutually exclusive categories reflective of both primary and secondary expected payers. Percentages based on weighted sampling.
Because demographic characteristics differed across payer types, we modelled the role of insurance using logistic regression (Table 3). The presence of any insurance was consistently associated with higher odds of admission, however, the magnitude of this effect varied by visit risk features. For visits with concurrent injury diagnosis, any insurance was associated with OR=1.39 (95%CI, 1.29–1.51) of admission. For visits with concurrent SI, any insurance was associated with OR=1.59 (95%CI, 1.31–1.92) of admission. For visits with neither concurrent injury nor SI diagnosis, any insurance was associated with OR=1.81 (95%CI, 1.65–1.99) of admission.
Table 3.
Associations of payer status and patient sex with disposition among depression emergency department visits.
| Subset Analyses | With Concurrent Injury | No Injury, With SI | Without Injury, without SI | |||
|---|---|---|---|---|---|---|
| OR | 95%CI | OR | 95% CI | OR | 95% CI | |
| Any Insurance | 1.39** | [1.29–1.51] | 1.59** | [1.31–1.92] | 1.81** | [1.65–1.99] |
| Any Medicare | 1.04 | [1.00–1.09] | 1 | [0.92–1.09] | 0.98 | [0.93–1.03] |
| Any Private Insurance | 1.20** | [1.13–1.27] | 1.23** | [1.08–1.41] | 1.05 | [0.97–1.14] |
| Any Medicaid | 0.95 | [0.89–1.01] | 0.97 | [0.85–1.10] | 0.87** | [0.80–0.94] |
| Female | 0.90** | [0.89–0.92] | 1.13** | [1.09–1.17] | 1.00 | [0.98–1.02] |
| N | 1,494,646 | 286,904 | 1,655,090 | |||
| Combined Cohort | ||||||
| OR | 95%CI | |||||
| Any Insurance | 1.81** | [1.69–1.94] | ||||
| Risk Category | ||||||
| With SI | 1.87** | [1.61–2.16] | ||||
| With Injury | 1.83** | [1.69–1.97] | ||||
| Interactions | ||||||
| Any Insurance*With SI | 0.96 | [0.83–1.10] | ||||
| Any Insurance*With Injury | 0.76** | [0.71–0.81] | ||||
| Female | 0.97** | [0.95–0.98] | ||||
| N | 3,436,640 | |||||
Four logistic regressions with outcome variable indicator of admission. The three models given by the first line are subset analyses stratified by risk. The bottom analysis reports interactions between level of risk and any insurance in the cohort that combines these subsets. Private=private insurance. Any Medicare, Any Private and Any Medicaid are not mutually exclusive categories reflective of both primary and secondary expected payers, nested within Any Insurance. The odds ratios for Any Insurance reflects a ratio of odds between those with insurance and those with no insurance. The odds ratios for the three nested insurance subgroups reflect a ratio of odds comparing payers among those with any insurance. All regressions include 10-year age group, indicator for visit year, and 31-category Elixhauser comorbidity. The year 2015 is excluded due to poor match due to ICD-9 to ICD-10 transition. Each regression contains an indicator for any insurance with nested indicators for insurance subtype.
indicates that the Wald statistic is significant at the 0.01 level.
In an aggregated model, injury (OR=1.83; 95%CI, 1.69–1.97) and SI (OR=1.87; 95%CI, 1.61–2.16) were each associated with increased odds of admission. Any insurance was associated with admission to a similar degree, OR=1.81 (95%CI, 1.69–1.94). Interacted with risk feature variables, insurance status was less predictive of admission for visits with concurrent injury.
Among visits with any insurance, private insurance was most strongly associated with admission, OR=1.20 (95%CI, 1.13–1.27) among visits with concurrent injury, and Medicaid was least strongly associated with admission, OR=0.95 (95%CI, 0.89–1.01) among visits with concurrent injury (Table 3). Regardless of risk features, the difference between private insurance and Medicaid was statistically significant. However, after controlling for age, year, and comorbidities, there was no longer a statistically significant difference between Medicare and Medicaid.
Admission rates by sex differed by risk group. For visits with concurrent injury, female sex was associated with decreased odds of admission (OR=0.90; 95%CI, 0.89–0.92), however, for visits either with concurrent SI or neither risk feature, female sex was correlated with increased odds of admission or no significant effect, respectively. The presence of ECI comorbidities was associated with significantly increased odds of admission across comorbidities.
3.3. Average Lengths of Stay and Charges
For patients who were admitted, the average inpatient lengths of stay were less than one week across insurance types and risk strata (Table 1 and Table S1), though they were on average longer for patients with Medicare (6.38 days) compared to patients with Medicaid (5.88 days), private insurance (4.90 days), or no insurance (4.12 days). Inpatient lengths of stay were on average shorter when patients were admitted after a presentation with concurrent injury (4.79 days), compared to admissions after presenting with SI (5.81 days) or with neither (5.82 days).
For all three payers, per diem charges were higher in the first week of admission compared to subsequent days (Figure S1). In the first week of admission, per diem rates were highest for patients with Medicare compared to other payer types, though Medicare payment rates converged with payment rates for private insurance for longer stays. Medicaid charges were lower compared to Medicare or private insurance. Accounting for differences in average lengths of stay, we estimated that Medicare was associated with an average per diem charge of $5,653 in 2021 dollars, excluding ED charges. The average per diem charge associated with private insurance was $4,186 (74% of the Medicare rate) and the average per diem charge was $3,453 for Medicaid (61% of the Medicare rate).
4. Discussion
We found that the likelihood of inpatient admission for patients presenting to the ED with depression varied depending on visit risk features and patient insurance. The relationships with insurance held after adjustments for patient comorbidities, age, and presenting features of risk, such as concurrent injury or SI. The most notable differences in admission were between those with and without any insurance and represented both statistically significant and clinically significant differences.
Other things being equal, the estimated odds ratios suggest that 21,600 fewer admissions occur annually for depression ED visits in the U.S. among patients who are uninsured compared to what we might predict based on comorbidities and demographic characteristics. Further, 35% of these missing admissions are accounted for by ED visits associated with injuries, which are the type of visits associated with the highest suicide risk [21]. Patients presenting without insurance were also disproportionately male and were more likely to have concurrent alcohol or substance use disorders, both risk factors for death by suicide [37,38]. Because uninsured patients compared to insured patients also experience greater financial barriers to accessing outpatient mental healthcare [39,40], they may have longer wait times for outpatient services after ED discharge compared, which might exacerbate existing risk. In sum, our analyses suggest that insurance status may pose a barrier to mental health service access even in the types of acute circumstances that might prompt patients to seek psychiatric care in the ED.
The strength of correlation between insurance status and admission was mitigated somewhat by features of risk such as concurrent injury or SI. Admission was least correlated with insurance status for ED visits for depression with concurrent injury and most correlated with insurance status for ED visits for depression with neither injury nor SI. These results suggest providers respond to level of risk when making disposition decisions in the ED, however, that the measured risk features did not fully mitigate insurance barriers.
Among patients with insurance, these findings suggest a differential between public and private insurance. Though Medicare was associated with the highest overall rates of inpatient admission, after adjusting for covariates, private insurance was more strongly associated with admission. It is possible that high charge amounts and high rates of admission for patients with Medicare are related to comorbidities or other risk factors such as age, which may also raise the cost of providing care. Despite differences in charge values, we did not find significant differences between Medicare and Medicaid in the likelihood of admission after adjusting for covariates.
Except in the case of visits with concurrent injury, male sex was not associated with admission. Although rates of suicide are higher among men than women [37], ED samples may not reflect rates of population-level risk. For example, women presenting to the ED for depression may have higher rates of other risk factors such as history of suicide attempt or self-injury higher than the population. Male sex could also be correlated with unobserved clinical features such as intoxication at presentation, symptom minimization, or stronger patient preference for discharge that might contribute to lower admission rates.
Although we could not explore disparities by race or ethnicity with our data, there are reasons to believe that persons of Black race or Hispanic ethnicity may be disproportionately affected by differential admission status by insurance. Rates of uninsurance are 11% and 20% among persons of Black race and Hispanic ethnicity, respectively compared to 8% among those of White race [41]. Rates of public insurance are 37% and 32% among persons of Black race and Hispanic ethnicity, respectively compared to 19% among those of White race [41]. Exploring these disparities may be an important topic for future study.
This study has several limitations. First, we lacked information on the history, symptoms, and other relevant clinical data available to clinical teams making ED disposition decisions. We use diagnostic codes associated with ED records to stratify patients by their characteristics and level of psychiatric risk and acknowledge heterogeneity in coding practices. Under-ascertainment of diagnoses, including suicidal ideation, may also contribute to imprecision of our estimates. Second, illness severity may be under-ascertained in our data. In particular, uninsured compared to privately-insured patients may be less likely to receive mental health screening or outpatient diagnoses [13]. If this is true, our findings would underestimate the association of insurance with the likelihood of admission controlling for severity. Third, we cannot distinguish medical admissions from psychiatric inpatient admissions, nor can we determine whether admissions were voluntary or involuntary. Consequently, we consider our findings to be broadly exploratory of differences between insurance groups. Third, our analyses were performed at the visit-level, which overweights higher acuity patient who may have presented repeatedly within a year relative to a person-level analysis.
5. Conclusions
In acute crises, some patients with depression, who are at high risk for suicide, seek care through the ED. For adult depression ED patients, we found that, controlling for demographic characteristics and medical comorbidities, those with insurance were significantly more likely to be admitted compared to those without insurance. Although differences in the likelihood of admission were lower for visits with features of high suicide risk, discrepancies in rates of admission remained. While risk assessment is complex and tailored to each patient’s individual history, presenting symptoms, and preferences, it is important to remember that it is also practiced within the context of real-world clinical resource constraints. Learning from evaluation of these practices may help us understand how to provide more equitable access to inpatient psychiatric services.
Supplementary Material
Figure 1.
Admission rates for depression by payer, by risk features of concurrent Injury, suicidal ideation or neither.
N=3,681,581. SI=suicidal ideation. Any Medicare, Any Private and Any Medicaid are not mutually exclusive categories reflective of both primary and secondary expected payers. 95% confidence interval given by error bars. Bar captions are shown only for Any Medicare and No insurance categories. Percentages based on weighted sampling.
Acknowledgements
YNG was supported in part by a Moynihan Clinical Research Fellowship from the Leon Levy Foundation and Award Number R25MH086466 from the National Institute of Mental Health.
Footnotes
COI
The authors have no conflicts of interest to disclose.
References
- [1].Mezzich JE, Evanczuk KJ, Mathias RJ, Coffman GA. Symptoms and hospitalization decisions. Am J Psychiatry 1984;141:764–9. 10.1176/ajp.141.6.764. [DOI] [PubMed] [Google Scholar]
- [2].Denchev P, Pearson J, Allen MH, Claassen C, Currier GW, Zatzick D, et al. Modeling the Cost-Effectiveness of Interventions for Suicide Risk in Hospital Emergency Department Patients. Psychiatr Serv 2018;69:23–31. 10.1176/appi.ps.201600351. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Weber AN, Michail M, Thompson A, Fiedorowicz JG. Psychiatric Emergencies: Assessing and Managing Suicidal Ideation. Med Clin North Am 2017;101:553–71. 10.1016/j.mcna.2016.12.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Emergency Medical Treatment & Labor Act (EMTALA) | CMS n.d https://www.cms.gov/Regulations-and-Guidance/Legislation/EMTALA (accessed April 20, 2022).
- [5].Maeda J, Nelson L. An Analysis of Private-Sector Prices for Hospital Admissions. Washington, D.C: Congressional Budget Office; 2017. [Google Scholar]
- [6].Fuller RL, Atkinson G, McCullough EC, Hughes JS. Hospital Readmission Rates: The Impacts of Age, Payer, and Mental Health Diagnoses. The Journal of Ambulatory Care Management 2013;36:147–55. 10.1097/JAC.0b013e3182866c1c. [DOI] [PubMed] [Google Scholar]
- [7].Venkatesh AK, Chou S-C, Li S-X, Choi J, Ross JS, D’Onofrio G, et al. Association Between Insurance Status and Access to Hospital Care in Emergency Department Disposition. JAMA Intern Med 2019;179:686–93. 10.1001/jamainternmed.2019.0037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Pearlmutter MD, Dwyer KH, Burke LG, Rathlev N, Maranda L, Volturo G. Analysis of Emergency Department Length of Stay for Mental Health Patients at Ten Massachusetts Emergency Departments. Annals of Emergency Medicine 2017;70:193–202.e16. 10.1016/j.annemergmed.2016.10.005. [DOI] [PubMed] [Google Scholar]
- [9].Zhu JM, Singhal A, Hsia RY. Emergency Department Length-Of-Stay For Psychiatric Visits Was Significantly Longer Than For Nonpsychiatric Visits, 2002–11. Health Affairs 2016;35:1698–706. 10.1377/hlthaff.2016.0344. [DOI] [PubMed] [Google Scholar]
- [10].Kraft CM, Morea P, Teresi B, Platts-Mills TF, Blazer NL, Brice JH, et al. Characteristics, clinical care, and disposition barriers for mental health patients boarding in the emergency department. The American Journal of Emergency Medicine 2021;46:550–5. 10.1016/j.ajem.2020.11.021. [DOI] [PubMed] [Google Scholar]
- [11].Misek RK, Magda AD, Margaritis S, Long R, Frost E. Psychiatric Patient Length of Stay in the Emergency Department Following Closure of a Public Psychiatric Hospital. J Emerg Med 2017;53:85–90. 10.1016/j.jemermed.2017.03.032. [DOI] [PubMed] [Google Scholar]
- [12].Bernstein SL, Aronsky D, Duseja R, Epstein S, Handel D, Hwang U, et al. The Effect of Emergency Department Crowding on Clinically Oriented Outcomes. Academic Emergency Medicine 2009;16:1–10. 10.1111/j.1553-2712.2008.00295.x. [DOI] [PubMed] [Google Scholar]
- [13].Marcus SC, Bridge JA, Olfson M. Payment source and emergency management of deliberate self-harm. Am J Public Health 2012;102:1145–53. 10.2105/AJPH.2011.300598. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Gao YN. Relationship Between Psychiatric Inpatient Beds and Jail Populations in the United States. J Psychiatr Pract 2021;27:33–42. 10.1097/PRA.0000000000000524. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Ehlman DC. Changes in Suicide Rates — United States, 2019 and 2020. MMWR Morb Mortal Wkly Rep 2022;71. 10.15585/mmwr.mm7108a5. [DOI] [PubMed] [Google Scholar]
- [16].Betz ME, Wintersteen M, Boudreaux ED, Brown G, Capoccia L, Currier G, et al. Reducing Suicide Risk: Challenges and Opportunities in the Emergency Department. Annals of Emergency Medicine 2016;68:758–65. 10.1016/j.annemergmed.2016.05.030. [DOI] [PubMed] [Google Scholar]
- [17].Olfson M, Wall M, Wang S, Crystal S, Gerhard T, Blanco C. Suicide Following Deliberate Self-Harm. Am J Psychiatry 2017;174:765–74. 10.1176/appi.ajp.2017.16111288. [DOI] [PubMed] [Google Scholar]
- [18].Doupnik SK, Rudd B, Schmutte T, Worsley D, Bowden CF, McCarthy E, et al. Association of Suicide Prevention Interventions With Subsequent Suicide Attempts, Linkage to Follow-up Care, and Depression Symptoms for Acute Care Settings: A Systematic Review and Meta-analysis. JAMA Psychiatry 2020;77:1021–30. 10.1001/jamapsychiatry.2020.1586. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Leavey G, Rosato M, Galway K, Hughes L, Mallon S, Rondon J. Patterns and predictors of help-seeking contacts with health services and general practitioner detection of suicidality prior to suicide: a cohort analysis of suicides occurring over a two-year period. BMC Psychiatry 2016;16:120. 10.1186/s12888-016-0824-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Cruz DD, Pearson A, Saini P, Miles C, While D, Swinson N, et al. Emergency department contact prior to suicide in mental health patients. Emergency Medicine Journal 2011;28:467–71. 10.1136/emj.2009.081869. [DOI] [PubMed] [Google Scholar]
- [21].Olfson M, Gao YN, Xie M, Wiesel Cullen S, Marcus SC. Suicide Risk Among Adults With Mental Health Emergency Department Visits With and Without Suicidal Symptoms. J Clin Psychiatry 2021;82:20m13833. 10.4088/JCP.20m13833. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].NEDS Overview n.d https://www.hcup-us.ahrq.gov/nedsoverview.jsp (accessed June 1, 2022).
- [23].Ballou S, Mitsuhashi S, Sankin LS, Petersen TS, Zubiago J, Lembo C, et al. Emergency department visits for depression in the United States from 2006 to 2014. Gen Hosp Psychiatry 2019;59:14–9. 10.1016/j.genhosppsych.2019.04.015. [DOI] [PubMed] [Google Scholar]
- [24].Pandya A, Larkin GL, Randles R, Beautrais AL, Smith RP. Epidemiological trends in psychosis-related Emergency Department visits in the United States, 1992–2001. Schizophr Res 2009;110:28–32. 10.1016/j.schres.2008.12.015. [DOI] [PubMed] [Google Scholar]
- [25].Healthcare Cost and Utilization Project (HCUP) NEDS Notes n.d https://www.hcup-us.ahrq.gov/db/vars/injury/nedsnote.jsp (accessed April 21, 2022).
- [26].Healthcare Cost and Utilization Project (HCUP) NEDS Notes n.d https://www.hcup-us.ahrq.gov/db/vars/i10_injury/nedsnote.jsp (accessed April 21, 2022).
- [27].ICD-9-CM to and from ICD-10-CM and ICD-10-PCS Crosswalk or General Equivalence Mappings. NBER n.d https://www.nber.org/research/data/icd-9-cm-and-icd-10-cm-and-icd-10-pcs-crosswalk-or-general-equivalence-mappings (accessed October 27, 2021).
- [28].ICD-10 | CMS n.d https://www.cms.gov/medicare/coding/icd10 (accessed October 27, 2021).
- [29].Zhu JM, Singhal A, Hsia RY. Emergency Department Length-Of-Stay For Psychiatric Visits Was Significantly Longer Than For Nonpsychiatric Visits, 2002–11. Health Affairs 2016;35:1698–706. 10.1377/hlthaff.2016.0344. [DOI] [PubMed] [Google Scholar]
- [30].Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care 1998;36:8–27. 10.1097/00005650-199801000-00004. [DOI] [PubMed] [Google Scholar]
- [31].Manitoba Centre for Health Policy, Garland A. The epidemiology and outcomes of critical illness in Manitoba. Winnipeg, Man.: University of Manitoba, Faculty of Medicine, Dept. of Community Health Sciences; 2012. [Google Scholar]
- [32].Stagg V ELIXHAUSER: Stata module to calculate Elixhauser index of comorbidity. Boston College Department of Economics; 2015. [Google Scholar]
- [33].Crandall C, Fullerton-Gleason L, Aguero R, LaValley J. Subsequent suicide mortality among emergency department patients seen for suicidal behavior. Acad Emerg Med 2006;13:435–42. 10.1197/j.aem.2005.11.072. [DOI] [PubMed] [Google Scholar]
- [34].Rockett IRH, Regier MD, Kapusta ND, Coben JH, Miller TR, Hanzlick RL, et al. Leading Causes of Unintentional and Intentional Injury Mortality: United States, 2000–2009. Am J Public Health 2012;102:e84–92. 10.2105/AJPH.2012.300960. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [35].Bridge JA, Olfson M, Caterino JM, Cullen SW, Diana A, Frankel M, et al. Emergency Department Management of Deliberate Self-harm: A National Survey. JAMA Psychiatry 2019;76:652–4. 10.1001/jamapsychiatry.2019.0063. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [36].Consumer Price Index, 1913- | Federal Reserve Bank of Minneapolis n.d https://www.minneapolisfed.org:443/about-us/monetary-policy/inflation-calculator/consumer-price-index-1913- (accessed May 6, 2022).
- [37].CDC. CDC WONDER: Underlying cause of death, 1999–2019 2020. https://wonder.cdc.gov/Deaths-by-Underlying-Cause.html (accessed April 20, 2022).
- [38].Chesney E, Goodwin GM, Fazel S. Risks of all-cause and suicide mortality in mental disorders: a meta-review. World Psychiatry 2014;13:153–60. 10.1002/wps.20128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [39].Cunningham P, McKenzie K, Taylor EF. The Struggle To Provide Community-Based Care To Low-Income People With Serious Mental Illnesses. Health Affairs 2006;25:694–705. 10.1377/hlthaff.25.3.694. [DOI] [PubMed] [Google Scholar]
- [40].Cunningham PJ. Beyond Parity: Primary Care Physicians’ Perspectives On Access To Mental Health Care. Health Affairs 2009;28:w490–501. 10.1377/hlthaff.28.3.w490. [DOI] [PubMed] [Google Scholar]
- [41].Jul 16 ADP, 2021. Health Coverage by Race and Ethnicity, 2010–2019. KFF 2021. https://www.kff.org/racial-equity-and-health-policy/issue-brief/health-coverage-by-race-and-ethnicity/ (accessed April 28, 2022).
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.