Association Between Race/Ethnicity and Disparities in Health Care Use Before First-Episode Psychosis Among Privately Insured Young Patients

Key Points

Question  To what extent are racial and ethnic disparities in health care use present in the year before diagnosis of the first episode of psychosis?

Findings  In this cohort study of 3017 privately insured patients aged 10 to 21 years, Black and Hispanic patients were less likely than White patients to receive behavioral health diagnoses and treatment before the diagnosis of first-episode psychosis. Racial/ethnic disparities persisted after controlling for socioeconomic variables.

Meaning  Racial/ethnic disparities in diagnoses and treatment appeared to be visible before the first diagnosis of psychosis, even in this relatively homogeneous sample of individuals with continuous private health insurance; lower rates of outpatient behavioral health care use in the year before the first episode of psychosis may result in increased duration of untreated psychosis and worse long-term outcomes.

Importance  Racial/ethnic disparities in health care use and clinical outcomes for behavioral health disorders, including psychosis, are well documented, but less is known about these disparities during the period leading up to first-episode psychosis (FEP).

Objective  To describe the racial/ethnic disparities in behavioral health care use and prescription drug use of children and young adults before the diagnosis of FEP.

Design, Setting, and Participants  An observational cohort study was conducted using medical and prescription drug claims from January 1, 2007, to September 30, 2015, obtained from Optum’s deidentified Clinformatics Data Mart Database, a commercial claims database augmented with race/ethnicity and socioeconomic variables. Data analysis was performed from February 6, 2018, to October 10, 2020. First-episode psychosis was determined by the presence of psychosis diagnoses on claims for at least 1 hospitalization or 2 outpatient events, with a continuous enrollment requirement of at least 2 years before the first diagnosis. Participants included 3017 Black, Hispanic, or White patients who were continually enrolled in commercial insurance plans and received an FEP diagnosis between the ages of 10 and 21 years.

Main Outcomes and Measures  Race/ethnicity was determined from a commercial claims database. Rates of inpatient admission, emergency department presentation, and outpatient visits (including psychotherapy), behavioral health disorder diagnoses, and antipsychotic/antidepressant prescription fills were determined for the year before FEP. Race/ethnicity was also obtained from Optum’s claims database. With use of multivariable logistic regression, results were adjusted for covariates including estimated household income, age, sex, and geographic division in the US.

Results  Of the 3017 patients with FEP, 643 Black or Hispanic patients (343 [53.3%] Black, 300 [46.7%] Hispanic, 324 [50.4%] male, mean [SD] age, 17.2 [2.76] years) were less likely than 2374 White patients (1210 [51.0%] male, mean age, 17.0 [2.72] years) to receive comorbid behavioral health disorder diagnoses in the year before the diagnosis of FEP (410 [63.8%] vs 1806 [76.1%], χ² = 39.3; P < .001). Except for emergency care, behavioral health care use rates were lower in Black and Hispanic patients vs White patients (424 [65.9%] vs 1868 [78.7%]; χ² = 45.0; P < .001), particularly for outpatient visits with behavioral health care professionals (232 [36.1%] vs 1236 [52.1%]; χ² = 51.7; P < .001). After adjustment for socioeconomic covariates, behavioral health care use rates (68.9% vs 79.2%; P < .001), outpatient visits with behavioral health professionals (37.7% vs 51.2%; P < .001), and other outcomes remained significantly lower for Black and Hispanic patients vs White patients.

Conclusions and Relevance  The results of this study extend existing research findings of well-known racial/ethnic disparities in the population of patients who are diagnosed with FEP. These differences were apparent in young patients with continuous commercial health insurance and after controlling for household income. Providing equal access to preventive outpatient behavioral health care may increase opportunities for timely detection of psychotic symptoms and early intervention and improve differential outcomes after FEP.

It is well documented that there are racial, ethnic, and socioeconomic disparities in behavioral health care use in the US. White patients with psychosis have higher behavioral health care use than other racial/ethnic populations across a variety of settings,^1,2 including specialist care.³ There are also disparities in prescription drug use, including antipsychotics; generally, prescription drug use is lower among Black or Hispanic patients with psychosis,⁴ especially use of second-generation antipsychotics.⁵ These disparities reflect long-standing societal biases and institutional and interpersonal racism that create barriers for patients of color, including unequal access to care,⁶ language barriers,⁷ clinician biases in interpretation of symptoms,^8-10 and health care professional shortages in underserved neighborhoods.¹¹

Less is known about disparities during the period before the development of first-episode psychosis (FEP). There is consensus that FEP events are generally preceded by more subtle symptoms that result in high and increasing levels of health care use,^12-15 and there have been reports of racial/ethnic disparities in initiation and type of care during this time.^15-19 This evidence, however, is limited and based on retrospective recall in selected samples or is not representative of health systems and disparities unique to the US. Behavioral health–related visits during the time before FEP represent opportunities to decrease the duration of untreated psychosis after the onset of symptoms and thereby increase the probability of improved long-term outcomes,^20,21 even if those improvements may not fully persist after discontinuing specialized, early intervention services.^22,23 Thus, disparities in access to care before FEP could exacerbate disparities in long-term outcomes for patients after the diagnosis of FEP.

The goal of this study was to describe differences by race and ethnicity in health care use for young patients during the year before a diagnosis of FEP. Specifically, we compared emergency, outpatient, and inpatient care; psychotropic medication use; and diagnoses of other behavioral health disorders before diagnosis of FEP by patient race/ethnicity. We made these comparisons while controlling for health insurance status and other measures of socioeconomic status, including estimated household income. Although appropriateness of psychotropic medication use in youth for psychosis and other disorders is an ongoing topic of discussion,^24,25 understanding disparities in behavioral health care use in the US in the period before FEP is needed to mitigate the long-term impact of psychosis in vulnerable populations.

Medical and prescription claims data from January 1, 2007, to September 30, 2015, were obtained from Optum’s deidentified Clinformatics Data Mart Database, a private claims database with 60 million covered lives throughout the US. The data set contains all health care and prescription drug claims for enrollees, race/ethnicity, and socioeconomic variables, and a unique identification number to allow for longitudinal identification. Data analysis was performed from February 6, 2018, to October 10, 2020. The eMethods in the Supplement provides more details on the data set and our sample. Household income (based on census block median income) and geographic location were provided at the policyholders’ level. The individuals’ year of birth, sex, and race or ethnicity (Asian, Black, Hispanic, White, or unknown) were also provided in the claims database. Race/ethnicity was based on Optum’s proprietary algorithms relying on the policyholder’s zip code in combination with the individual’s first, middle, and last names (E-Tech 7.3, Ethnic Technologies), public records, and self-reported surveys. Thus, the race/ethnicity variable should not be construed as a biological indicator. Our study only included comparisons between patients with FEP identified as White vs Black or Hispanic, because the number of patients identified as Asian was too small to analyze separately. Patients with missing or unknown race/ethnicity data were also excluded.

The University of Southern California institutional review board approved the study and determined that it was not necessary to obtain informed consent from individuals or pursue a waiver of consent, because the key to identify individuals was not available to the researchers. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Patients experiencing FEP were identified if they had at least 1 inpatient or 2 outpatient claims with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) all-listed diagnosis codes 295, 297, 298, 293.81, 293.82, 296.04, 296.14, 296.24, 296.34, 296.44, 296.54, and 296.64. These codes include schizophrenia, delusional, and other nonorganic disorders and bipolar or major depressive disorder with psychotic behavior. The service date of the first psychosis encounter was designated as the index event. We excluded patients younger than 10 or older than 21 years during the index event. In addition, because of the discontinuation of ICD-9-CM diagnosis codes on September 30, 2015, patients with index events after this date were excluded. The findings of this study may vary when International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes are used, because these codes result in different psychosis prevalence estimates compared with ICD-9-CM codes,²⁶ although the prevalence estimates are presumed to vary equally within race/ethnicity groups. To reduce the possibility of selecting patients with preexisting psychosis, we included only those who were continuously enrolled for at least 24 months before the index date and excluded patients whose index diagnosis was schizophrenia in remission (295.x5).

Hospitalizations were identified from inpatient claims and further categorized as direct admissions or admissions through the emergency department (ED). Noninpatient claims were categorized as ED discharges based on relevant Current Procedural Terminology codes, place of service, provider category, type of service, or revenue codes. In addition, outpatient claims were characterized as psychotherapy or as a visit with a psychiatrist or other behavioral health professional (eg, psychologist, social worker, or addiction medicine specialist, collectively called behavioral health care professionals in this study) (eMethods in the Supplement). For each place of service, we determined the proportion of patients who had at least 1 event in 30-day increments before the start of the index event (not including the index event), for a total of 24 months.

Behavioral health comorbidities were identified by medical claims with ICD-9-CM diagnosis codes 291 to 319. Additional events, specifically for suicidal/homicidal ideation and alcohol- or drug-related disorders or misuse, were identified using E-codes and V-codes (eTable 1 in the Supplement). For ease of exposition, we divided all diagnoses into 14 categories. Patients in the episodic mood disorder with psychotic behavior category at the index event were, by definition, also included in the depression and/or bipolar disorder categories. In addition, because patients can receive multiple diagnoses during 1 encounter or during the year before the index event, patients were frequently being diagnosed with multiple comorbid behavioral health conditions (54.9% of patients in the year before and 69.8% at the time of FEP diagnosis).

We identified prescription fills for antipsychotics and antidepressants, commonly prescribed to patients with psychosis,²⁷ using generic drug names and additional codes provided in Optum’s data dictionary (eTable 2 in the Supplement provides further detail). The proportion of patients with medication use on any of the 360 days before and on the index date was based on estimated adherence to the medication (eMethods in the Supplement). Because typical antipsychotic prescriptions represented only 3.5% of all antipsychotic prescriptions filled, data on atypical and typical antipsychotics were combined.

Differences in outcomes between racial/ethnic groups were evaluated using χ² tests and Wilcoxon rank sum tests, as appropriate. We estimated adjusted differences in outcomes by race/ethnicity using multivariable logistic regression controlling for sex, age, year, type of psychosis diagnosis at the index event, geographic division, and household income. We report regression-adjusted estimated proportions, holding all control variables at their sample means. With 2-tailed testing, results were considered significant at P < .05. Sensitivity analyses were performed by comparing White with Black and White with Hispanic patients separately. Most outcomes for Black and Hispanic patients are not presented separately because most results were similar (eTable 3 and eTable 4 in the Supplement). All analyses were also repeated for patients with missing and unknown race/ethnicity; characteristics and outcomes of these patients were most similar to those of White patients. This similarity may reflect an approximately random sample of the overall, mostly White, population (eTables 5, 6, and 7, and eFigure 1 in the Supplement). Additional sensitivity analyses included removing patients who were diagnosed with FEP in the ED only (eTable 8 in the Supplement). Data sets were prepared using SAS, version 9.4 (SAS Institute Inc) and analyzed using Stata, version 16.0 (StataCorp LLC).

Of the 3017 patients included in the study, 643 were Black or Hispanic (21.3%; Black, 343 [53.3%]; Hispanic, 300 [46.7%]), and 2374 were White (78.7%). The Black and Hispanic cohort comprised 324 males (50.4%) and 319 females (49.6%); mean (SD) age at index diagnosis was 17.2 (2.76) years. The White cohort included 1210 males (51.0%) and 1164 females (49.0%); mean age at index diagnosis was 17.0 (2.72) years. Attrition sample sizes and race/ethnicity distributions are presented in eTable 9 and eFigure 2 in the Supplement. Black enrollees were more likely (343 of 27 336 patients [1.3%], χ² = 13.2; P < .001) and Hispanic enrollees less likely (300 of 42 467 patients [0.7%], χ² = 36.4; P < .001) to receive a probable FEP diagnosis than White enrollees (2374 of 233 077 patients [1.0%]) among all enrollees aged 10 to 21 years in the Optum database (n = 360 702). Most patients (2130 [70.6%]) were diagnosed in an inpatient setting, either direct admission or through the ED. No significant difference was observed between the proportion of White patients (1688 [71.1%]) and Black and Hispanic patients (442 [68.7%]) (P = .24) diagnosed in the hospital. When the qualifying index events consisted of 2 outpatient visits (including ED discharges), the median duration between these visits was 13 days (interquartile range, 4-35 days); the difference by race/ethnicity was not significant (Black or Hispanic: 10 days [interquartile range, 2-39 days], White: 14 days [interquartile range, 5-34 days]; P = .28). There were also no significant differences in age, sex, or enrollment duration between Black or Hispanic and White patients with FEP (Table 1). Black or Hispanic patients with FEP were more likely than White patients with FEP to be in the lowest household income category (129 [20.1%] vs 194 [8.2%]; P < .001) and less likely to be in the highest household income category (129 [20.1%] vs 1169 [49.2%]; P < .001) compared with White patients (Table 1). Separate analyses of family enrollment data from the claims database showed that 97% of patients with FEP who conformed to our inclusion criteria were likely dependents, based on linkage to an enrollee who was at least 15 years older than the patient.

White patients were more likely than Black or Hispanic patients to receive a diagnosis of another behavioral health disorder in the year before the FEP diagnosis (1806 [76.1%] vs 410 [63.8%], χ² = 39.3; P < .001). Regardless of race/ethnicity, patients most frequently received a diagnosis of depression; other common behavioral health disorders included bipolar disorders (without psychotic features) and anxiety disorders (Table 2).

Differences by race/ethnicity in the prevalence of comorbid behavioral health diagnoses were less pronounced at the time of the FEP diagnosis. This was especially true for adjustment reaction and transient disorders that, together with conduct and impulse control disorders and attention-deficit/hyperactivity disorder, were less likely to be diagnosed at the index event than in the year before FEP. Conversely, suicidal ideation or attempt increased substantially for both Black or Hispanic and White patients at the time of the FEP diagnosis, suggesting that the onset of suicidal symptoms may be a major reason for help-seeking among patients diagnosed with an FEP. Schizophrenia was diagnosed in a larger proportion of Black or Hispanic patients (76 [11.8%]) than White patients (201 [8.5%]) (P < .01). This difference was mostly attributable to a disparity between Black (45 [13.1%]) and White (201 [8.5%]) (P = .005) patients; the rate of schizophrenia diagnoses among Hispanic patients (31 [10.3%]) was not significantly different from either Black or White patients. Schizophrenia diagnoses were less common than other psychosis diagnoses, because many patients in general are diagnosed initially with unspecified psychosis.²⁸

In the year before the FEP event, Black or Hispanic patients were less likely than White patients to visit a behavioral health professional in an outpatient setting (232 [36.1%] vs 1236 [52.1%]; P < .001) and receive psychotherapy from a behavioral health professional (198 [30.8%] vs 1136 [47.9%]; P < .001). Both the gap and absolute proportions became larger as the index diagnosis approached (Figure, A; eFigure 3 in the Supplement). Black or Hispanic patients were also less likely to receive a behavioral health diagnosis from other outpatient professionals (eg, general practitioners) than White patients in the year before the diagnosis of FEP (199 [30.9%] vs 1003 [42.2%]; P < .001). However, this gap narrowed leading up to the index diagnosis and was not significantly different in the month before the index diagnosis (13.1% vs 11%; P = .11) (Figure, B). Black or Hispanic patients were not significantly less likely than White patients to have an ED visit in the year before FEP (337 [52.4%] vs 1340 [56.4%]; P = .07) (Figure, C). Patients who did not have a behavioral health professional visit in the year before the diagnosis of FEP had a higher probability of receiving an FEP diagnosis in the ED (823 of 1549 [53.1%]) than those who were seen by a behavioral health professional (685 of 1468 [46.7%]) (P < .001), regardless of race/ethnicity.

Despite having no diagnosis of psychosis before the diagnosis of FEP, 691 of all patients (22.9%) filled a prescription for antipsychotic drugs, 496 (71.8%) of whom also filled prescriptions for antidepressants. The percentage of Black and Hispanic patients who filled 1 or both prescriptions was consistently lower than the percentage of White patients (Table 3). Unadjusted overall behavioral health care use in the year before the FEP event was 12.8 percentage points lower in Black or Hispanic patients than in White patients (424 [65.9%] vs 1868 [78.7%]; χ² = 45.0; P < .001) (Table 4). After adjusting for covariates, this difference was slightly reduced to 10.3 percentage points (68.9% vs 79.2%; P < .001). Other observed differences in medication use and health care use rates also persisted between White and Black and Hispanic patients (Table 4).

A diagnosis of FEP was associated with a small reduction in disparities in behavioral health professional visits. In a random sample of individuals aged 10 to 21 years conforming to similar inclusion criteria but without FEP, White patients were 1.56 times more likely to have seen a behavioral health professional than Black or Hispanic patients (7.2% vs 4.6%; P < .001), compared with a ratio of 1.36 times in patients with FEP (51.2% vs 37.7%) (Table 4). Receipt of outpatient psychotherapy from a behavioral health professional exhibited a similar association, suggesting that, to a certain extent, FEP reduced disparities in this population.

Using a large national database of young, privately insured patients, we compared behavioral health care use by race/ethnicity for patients in the year before FEP. We found that Black and Hispanic patients appeared to be significantly less likely than White patients to receive care from a behavioral health professional in either an inpatient or outpatient setting. Moreover, the racial/ethnic disparities in access to behavioral health care became more pronounced closer to the FEP diagnosis. Black and Hispanic patients also had lower rates of comorbid behavioral health diagnoses and fewer patients had fills for psychotropic prescription drugs, which may be due, in part, to the reduction in health care use and/or interpretation of clinical symptoms. It is not clear whether current prescription drug rates for White patients with FEP are optimal considering increasing off-label use of antipsychotics in youth,²⁵ and thus it is harder to advocate for eliminating this specific disparity as opposed to, for example, the disparity observed in outpatient behavioral health services. Reduced outpatient health care use was not compensated for by more Black and Hispanic patients receiving emergency care, and differences in use could not be explained by the type of FEP diagnosis or patient characteristics, including estimated household income. In addition, disparities in behavioral health care use for patients with FEP were slightly reduced compared with enrollees without FEP.

Our results are generally consistent with previous findings of lower rates of behavioral health care,^17,18,29-31 differential diagnoses,^8,10,32 and psychotropic medication use^16,29,31,33 for Black and Hispanic patients. However, to our knowledge, no other study has looked at these racial/ethnic differences before the diagnosis of FEP in US-based health systems using large administrative databases, nor have there been any studies on rates of comorbid behavioral health diagnoses by race/ethnicity. The current findings merit consideration based on our analysis of a relatively homogeneous sample (young, dependent patients with private health insurance for at least 2 years) and the persistence of disparities even after controlling for differences in socioeconomic variables and type of FEP diagnosis. Thus, our estimates are likely conservative in light of well-known racial/ethnic disparities in health insurance coverage and access to care in the US.

From our data, it appears unlikely that the observed lower rates of outpatient care and antidepressant and antipsychotic prescription drug fills reflect a reduced need for behavioral health treatment in our study. We found that Black enrollees showed a higher overall rate of FEP diagnoses and Black patients with FEP were more likely to receive a diagnosis of schizophrenia than White patients with FEP. Our findings are consistent with previous reports^8,34,35 and supported by studies showing that Black patients face reduced access to high-quality care and differential treatment by clinicians,⁸ while simultaneously displaying a higher unmet need for behavioral health care.^36,37 For Hispanic patients in our sample, the contrast between severity and use of health care was not as evident, because schizophrenia diagnosis rates were similar to those of White patients and overall FEP diagnosis rates were lower. However, the correlation between mental health status, help-seeking behaviors, stigma, and clinical outcomes among Hispanic individuals is complex and mediated by other factors, including country of origin, English-language proficiency, and migrant status,³⁸ that we were not able to control for in our data set.

Reducing the duration of untreated psychosis has been reported to result in improved long-term behavioral health outcomes.^20,21 Our findings highlight that Black youth are being diagnosed with psychosis more often but are less likely to receive care before their psychosis manifests, suggesting that these patients have fewer opportunities to receive an earlier diagnosis. Interventions should add emphasis on improving access and increasing referral rates for Black and Hispanic patients with new mental health disturbances so psychotic symptoms can be assessed earlier. Health care professionals should be aware that there are relatively few opportunities to identify patients with incipient psychosis and adjust screening and follow-up procedures accordingly. Behavioral health professionals may be able to further increase appropriate preventive behavioral health care use before the occurrence of FEP by accepting more clients with a diverse background, practicing in underserved areas, being aware of differences in the attribution of psychotic symptoms, receiving antiracism training, and providing foreign language services.

This study has limitations. The results were adjusted for household income and US geographic division, but our data source does not provide detailed geographic information that would allow us to control for factors such as local availability of behavioral health care professionals. Another limitation of the data source is the lack of access to clinical records to confirm the FEP diagnosis, meaning that we likely included a broad range of psychiatric patients with diverging symptoms. We also lacked data to assess how the duration of untreated psychosis varied by race/ethnicity, which should be addressed in future research, because the association between the duration of untreated psychosis and race/ethnicity has not been the main focus of many US-based studies³⁹ and large subgroups are needed to account for variability in the duration of untreated psychosis.⁴⁰ Furthermore, our findings are susceptible to the possibility that we observed preexisting rather than new-onset psychosis, although we attempted to minimize this possibility by adopting a 24-month run-in period without psychosis diagnoses. Our data source also does not include school-based mental health visits if schools do not bill private insurance plans. Black and Hispanic adolescents receive mental health care in educational settings at relatively higher rates than their White counterparts, although this difference in rates has been observed to a greater extent for those with public insurance coverage.⁴¹ In addition, although we likely missed some care not paid for through private insurance, we were unlikely to miss acute services, such as ED visits or hospitalizations. We also did not observe patients who did not seek care, did not receive a diagnosis, or did not fill prescriptions. These concerns were minimized by selecting young patients who have a high likelihood of being dependents of primary enrollees. Being a dependent of a primary enrollee provides a presumed care network of family members who may encourage patients to seek necessary care and fill prescriptions on their behalf. The chance that selective attrition occurred based on FEP severity as a result of patient unemployment and subsequent loss of insurance coverage is also reduced. Thus, despite our database’s limitations, we believe our study reliably documents the existence of significant racial/ethnic disparities in the receipt of behavioral health care before FEP.

The results of this study extend previous findings of racial/ethnic disparities in behavioral health care use in a US-based population of patients who later are diagnosed with FEP. These differences are apparent even in young patients with commercial health insurance and after controlling for estimated household income and type of FEP diagnosis. Because Black and Hispanic patients have fewer behavioral health–related clinical contacts before the occurrence of FEP, these patients may have reduced opportunities for timely detection of psychotic symptoms and early interventions. Additional efforts to encourage behavioral health care use in Black and Hispanic populations and combat racism in health care are required to help break down real and perceived barriers, allow for equal access to high-quality behavioral health care, and eventually help reduce racial/ethnic differences in clinical outcomes before and after a diagnosis of FEP.

Accepted for Publication: October 20, 2020.

Published Online: December 23, 2020. doi:10.1001/jamapsychiatry.2020.3995

Corresponding Author: Hanke Heun-Johnson, PhD, Leonard D. Schaeffer Center for Health Policy and Economics, University of Southern California, USC University Park Campus, 635 Downey Way, VPD 312, Los Angeles, CA 90089 (heunjohn@usc.edu).

Author Contributions: Dr Heun-Johnson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Heun-Johnson, Menchine, Axeen, Claudius, Seabury.

Acquisition, analysis, or interpretation of data: Heun-Johnson, Menchine, Axeen, Lung, Wright, Seabury.

Drafting of the manuscript: Heun-Johnson, Menchine, Axeen, Lung, Seabury.

Critical revision of the manuscript for important intellectual content: Heun-Johnson, Menchine, Axeen, Claudius, Wright, Seabury.

Statistical analysis: Heun-Johnson, Menchine, Axeen, Lung, Seabury.

Administrative, technical, or material support: Menchine, Seabury.

Supervision: Menchine.

Conflict of Interest Disclosures: Dr Heun-Johnson reported receiving research support from Alkermes during the conduct of the study. Dr Seabury reported receiving personal fees from Precision Health Economics outside the submitted work. No other disclosures were reported.

Funding/Support: Our study was partially supported by an unrestricted gift to the Leonard D. Schaeffer Center for Health Policy and Economics from Alkermes (Drs Seabury and Heun-Johnson).

Role of the Funder/Sponsor: The funding organization had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Meeting Presentation: A portion of this paper was presented at the 2019 AcademyHealth Annual Research Meeting; June 3, 2019; Washington, DC.