Mental Health Staffing at HRSA-Funded Health Centers May Improve Access to Care

Abstract

Objectives:

The study objective was to examine the association between mental health staffing at health centers funded by the Health Resources and Services Administration (HRSA) and patients receiving mental health treatment.

Methods:

Data were from the 2014 HRSA-funded Health Center Patient Survey and the 2013 Uniform Data System. Health Center staffing of mental health providers including psychiatrists, psychologists, and other licensed staff were examined. The outcomes of interest were if a patient (1) received any mental treatment, and (2) received any such treatment on-site (at their health center). Analyses were conducted using multilevel generalized structural equation logistic regression models for 4,575 patients ages 18–64.

Results:

Patients at health centers with at least one mental health full-time equivalent (FTE) per 2,000 patients had a higher predicted probability of receiving mental health treatment (32%) compared with those at health centers with fewer than one such FTE (24%) or no such staffing (22%). Among patients who received this treatment, those at health centers with no staffing had a significantly lower predicted probability of receiving such treatment on-site (28%) than patients of health centers with limited staffing (49%) and at least one mental health FTE (65%). The predicted probability of receiving such treatment on-site was significantly higher (58% vs. 40%) if there was a co-located psychiatrist.

Conclusion:

Co-locating mental health staff at health centers increases the probability of patients’ access to such treatment on-site and from providers elsewhere.

In 2018, 47.6 million adults in the United States had some form of mental illness, but only 43% received treatment for their condition.(1) Unmet mental health needs have been attributed to stigma avoidance, mistrust of the mental healthcare system, and lack of access to services due to cost of care and insufficient insurance coverage.(2, 3) This cost barrier is particularly concerning since uninsured and low-income groups are at a higher risk for mental illness.(4)

Research shows that most adults with mental health conditions end up seeking and receiving treatment within the primary care setting.(5) Integrating mental health providers into primary care settings can help primary care providers (PCPs) improve care coordination for these patients.(6) Other studies have shown that the availability of mental health services within primary care settings have enhanced mental health access and promoted treatment.(7–9) Yet, much of the existing literature has focused on specialized settings and partnerships that are not generalizable to community-based settings.(10, 11)

Health Resources and Services Administration (HRSA)-funded health centers are a primary source of providers of care for uninsured and low-income patients across the United States. Health centers provide comprehensive and affordable primary care to everyone, regardless of ability to pay, and in 2018, they delivered care to over 28 million Americans.(12, 13) Health centers also reported providing mental health services to over 2 million patients in the same year, including individual or group counseling/psychotherapy, 24-hour crisis services, and case management services. The strategic location of health centers in medically underserved areas and their experience providing culturally appropriate care highlight the potential of health centers in improving access to mental health services among low-income patients.(12) Although past evidence from unrepresentative samples indicated limited provision of mental health services, emerging evidence indicates an increase in service delivery.(14–17)

We aimed to address gaps in the literature to examine associations between co-location of mental health staff at health centers and (1) patients receiving such treatment anywhere, and (2) patients receiving mental health treatment on-site versus elsewhere. We examined both size as well as type of co-located mental health staff (i.e. psychiatrists, psychologists, other licensed staff).

We hypothesized that patients at health centers with co-located staff would have a higher predicted probability of receiving mental health treatment anywhere, because mental health staff would improve PCPs’ awareness of mental health conditions and provide guidance on screening and detection of these conditions. We also hypothesized that among patients who received such treatment, patients in co-located health centers would have a higher predicted probability of receiving mental health treatment on-site because of elimination of access barriers.

Methods

Data and sample.

We used the 2014 Health Center Patient Survey (HCPS), an in-person survey of health center patients conducted between October 2014 and April 2015. HCPS includes information on patient demographics, health care utilization, health conditions, and patient experiences at 169 health centers. The survey was designed using a three-stage sampling method to provide a nationally representative sample of parent organization, clinics within the parent organization, and patients at clinics. Patients were pre-screened to ensure they had at least one prior visit to their health center and were interviewed in the waiting room when they registered for an appointment. We merged the HCPS data with the 2013 Uniform Data System (UDS) report to include health center characteristics. UDS includes health center aggregated administrative data provided to HRSA for the past calendar year on staffing, revenues, patient demographics, and services delivered. Since UDS data reflects health center resources at the end of the calendar year, the 2013 UDS data was used to provide the best estimate of mental health staffing in place for the beginning of 2014 and the time period when patients reported receiving mental health treatment in HCPS. Our project was determined exempt by Our project was determined exempt by University of California, Los Angeles Institutional Review Board.

Due to differential service needs among children and the elderly, we restricted our sample to adult patients aged 18–64 years (n=5,040).(18) Because we were interested in the impact of co-located mental health staffing among health center patients specifically, we excluded 416 patients who did not identify the health center they were interviewed at as their usual source of care. We also excluded 49 patients with missing responses to questions about their mental health condition. Our final sample size was 4,575 patients.

Independent variables.

The primary predictor of interest for our analysis was mental health staffing at a patient’s health center, defined as the ratio of total full-time mental health employees (FTE) per 2,000 patients and categorized into more meaningful categories of none, less than one (part-time), and one or more (at least one FTE). Mental health staff included psychiatrists, licensed clinical psychologists, licensed clinical social workers, and other licensed mental health providers including psychiatric social workers, psychiatric nurse practitioners, family therapists, and other licensed master’s degree prepared clinicians.(19) In the absence of data on optimal or average panel sizes for mental health providers, we standardized this size of mental health FTE per 2,000 patients based on the calculation that in one year (50 weeks, excluding vacation) one FTE would be able to provide around 2,000 one-hour consultations. The second independent variable of interest was the type of co-located mental health providers. We created separate indicator variables for whether health centers had (1) any psychiatrist, (2) any psychologist, and (3) any other licensed mental health staff.

We controlled for the number of patients per PCP FTE to account for likely provision of mental health services by PCPs with smaller panel sizes. We created three categories indicating small (under 1,200, reference), medium (1,200–1,999), and large (2,000 or more) patients per PCP.

We also controlled for clinic support staff to PCP ratio as a measure reflecting the capacity of PCPs to collaborate with mental health staff in care of patients with mental health conditions. We categorized this ratio into two or less (reference), more than two and up to four, and greater than four clinical support staff to PCP. We also included an indicator for rural location of the health center since rural areas often have a shortage of mental health professionals. To reflect the overall capacity in size and service offerings of a health center, we used an indicator for the number of clinics within the health center organization as low (10 or fewer, reference), medium (11–19), and large (20 or more).

As a measure of patient demand for mental health services, we included the percent of patients diagnosed with depression (regardless of primary diagnosis) at the patient’s respective health center. We categorized this variable into low (5% or less, reference), medium (more than 5% to but less than 15%), and high (15% or more). We also included the percent of total revenue from Medicaid managed care contracts at the patient’s health center as a proxy for potential incentives to provide more integrated care and categorized this variable into no revenue vs. less than 25%, and 25% or more.

We used Andersen’s model of health care utilization and access to inform selection of relevant patient characteristics.(20) In line with pre-disposing factors, we controlled for socio-demographics such as patient’s race/ethnicity, gender, and age as proxies for perceived stigma.(21) We also included marital status, as a proxy for protective impact of patients’ social network from depression or anxiety. We included limited English language proficiency and education as proxies for patients’ familiarity with the healthcare system and their ability to communicate with health care professionals. We added a variable indicating whether a patient reported that they would definitely recommend the health center to their friends or family as a proxy for patients’ experience that may influence their decision to seek care in the future.

For enabling factors, we included a measure of poverty (at or below 100% FPL vs. higher) and insurance coverage (Medicaid or other vs. uninsured) as financial factors that enable patients to access mental health services. We used several indicators of need for services, including a variable to account for the patient’s level of psychological distress using the Kessler 6 (K6) diagnostic questions.(22) We identified patients with low (reference), moderate, and severe psychological distress. Since substance use disorders are highly correlated with mental health conditions, we also included an indicator variable for whether a patient reported wanting or needing substance use disorder treatment or counseling within the last year in the HCPS. We also used the patient’s self-reported health status, dichotomized as excellent or very good vs. good, fair or poor health (reference).

Dependent variables.

The first outcome of interest was whether a patient reported receiving any mental health treatment or counseling within the last year. The second outcome of interest was whether the patient received all, some, or no mental health services on-site (at their health center). We combined the first two categories (all and some) to indicate at least some mental health visits on-site due to small sample sizes. Further detail of the construction of dependent and independent variables is provided in the online supplement.

Statistical Analysis

We used STATA 15 for our statistical analyses. After running descriptive statistics and checking for collinearity problems, we developed multivariate logistic regression models to assess the independent association of co-location of any mental health staff and type of mental health staff with patients receiving any mental health treatment. We ran separate weighted multilevel generalized structural equation models to examine association of MH staffing with (1) patients receiving any mental health treatment and (2) patients receiving mental health services on-site at the health center, controlling for clustering of patients within health centers. For ease of interpreting results of these models, we used margins command to obtain absolute predicted probabilities. Both models included a single aggregated survey weight across health center organizations, sites, and patients within those sites to account for the complex survey design of HCPS. We conducted stratified analyses by the levels of K6 score and those with specific mental health conditions to assess if there were differences in results based on mental health status.

Results

Table 1 provides health center and patient characteristics of the study sample. The majority of patients (71%) went to health centers with limited mental health staff (less than one licensed mental health FTE per 2,000 patients) and 10% went to health centers with at least one licensed mental health FTE. In the study sample, most patients (73%) reported no mental health visits in the last 12 months, 16% had at least some mental health visits on-site at the health center, and 11% had all mental health visits off-site. Across all patients, 45% were scored as having mild or no psychological distress, 40% as moderate psychological distress, and 15% as severe psychological distress. There was a higher proportion of patients who reported moderate or severe psychological distress at health centers with at least one licensed mental health FTE (68%) than at health centers with fewer or no licensed mental health FTE (56% and 46% respectively).

TABLE 1.

Characteristics of patients and of the Health Resources and Services Administration–funded health centers that they attended, by the center’s level of licensed mental health FTE^a

				No mental health FTE			Licensed mental health FTE per 2,000 patientsb
	Total (N=4,575)			(N=547, 19%)			Fewer than one (N=3,543, 71%)			At least one (N=485, 10%)
Characteristics	N	%	SE	N	%	SE	N	%	SE	N	%	SE	pc
Patient mental health utilization (in past 12 months)													<.01
No mental health visits	3,419	73	2	459	81	3	2,695	74	2	291	50	11
All mental health visits off site	506	11	1	66	13	2	395	11	1	45	12	4
At least some mental health visits on site	650	16	2	22	6	2	453	15	2	149	37	8
Health Center
Organizational Capacity
PCP panel sized													.94
<1,200 patients (reference)	866	17	4	144	21	11	602	15	5	120	23	12
1,200–1,999 patients	2,565	58	6	221	60	15	1,990	58	7	354	50	19
≥2,000 patients	1,144	25	5	182	19	11	951	27	6	11	27	21
Ratio of clinic support staff to PCP FTE													.67
≤2 (reference)	1,306	34	6	170	43	15	971	33	7	165	26	13
>2 - ≤4	2,710	54	6	262	47	16	2,128	53	7	320	74	13
>4	559	12	4	115	10	8	444	14	5	0	—	—
Rural location	1,506	47	6	304	72	16	1,001	43	7	201	28	14	.13
Number of Health Center’s clinic sites													.15
≤10 (reference)	2,298	60	6	407	80	12	1,756	57	7	135	49	19
11–19	1,173	23	5	140	20	12	975	25	6	58	10	8
≥20	1,104	17	4	0	—	—	812	18	5	292	41	18
Demand for mental health services: % of patients at center with a diagnosis of depression													.05
≤5% (reference)	1,244	30	5	304	52	15	930	28	6	10	1	1
>5% - <15%	2,915	59	6	226	34	14	2,362	64	6	327	70	15
≥15%	416	11	4	17	14	13	251	8	3	148	30	15
Funding incentives for integrated care: % of total revenue from Medicaid managed care													.58
None (reference)	2,259	56	6	293	48	16	1,751	58	7	215	55	18
<25%	1,974	38	6	234	51	16	1,470	34	6	270	45	18
≥25%	342	6	2	20	2	2	322	8	3	0	—	—
Patient
Predisposing factors
Female	2,946	66	2	369	74	4	2,311	66	3	266	48	10	.04
Age													.06
18–25	427	16	2	265	62	4	1,691	50	2	247	51	4
26–49 (reference)	2,203	52	2	47	9	2	350	18	2	30	17	4
50–64	1,945	31	2	235	29	4	1,502	32	3	208	32	5
Race-ethnicity													.24
Non-Hispanic White (reference)	1,081	48	4	144	60	5	770	44	5	167	53	15
Hispanic, Latino	1,616	25	3	183	19	6	1,313	28	4	120	15	3
Non-Hispanic Black	1,074	20	2	159	18	6	813	21	3	102	18	9
Other	804	6	1	61	3	2	647	6	1	96	14	6
Not married and no domestic partner (reference: married or has a domestic partner)	2,717	59	2	308	49	5	2,075	61	2	334	61	3	.06
Education													.28
Less than high school (reference)	1,980	34	2	270	39	4	1,533	33	3	177	35	7
High school graduate	1,268	29	2	125	24	2	1,001	32	2	142	21	7
More than high school	1,327	37	2	152	37	4	1,009	35	2	166	44	6
Limited English proficiency	1,418	17	2	138	11	4	1,203	19	3	77	13	3	.13
Would not or would only somewhat recommend health center to family or friends (reference: would definitely recommend)	840	14	1	102	9	2	649	16	2	89	14	5	.19
Enabling factors
Income ≤100% of the federal poverty level (reference: >100%)	3,004	57	3	340	49	6	2,341	60	3	323	51	7	.13
Insurance coverage													.43
Uninsured (reference)	1,241	30	3	197	25	5	942	31	4	102	27	7
Medicaid	2,531	54	4	227	51	5	2,002	54	5	302	56	7
Other or missing response	803	16	2	123	24	6	599	15	2	81	17	4
Need factors
Level of psychological distress as measured by K6e													.08
Mild or none (reference)	2,029	45	3	270	55	7	1,585	44	3	174	32	7
Moderate	1,797	40	3	202	33	6	1,377	40	2	218	57	8
Severe	749	15	1	75	13	3	581	16	2	93	11	5
Reported past-year want of or need for counseling or treatment for substance use	356	5	1	40	4	2	263	6	1	53	5	3	.89
Excellent or very good self-reported health (reference: good, fair, or poor health)	775	20	2	107	16	3	580	19	2	88	28	4	.12

^aSource: Uniform Data System 2013 and Health Center Patient Survey 2014. Ns are unweighted, and percentages are weighted.

^bFTE, full time equivalent.

^cFrom chi-square test; reference group for comparison is indicated.

^dPCP, primary care provider.

^eK6 Score, Kessler K6 non-specific distress scale.

Logistic regression results indicated that the predicted probability of receiving any mental health treatment within the last year was 22% for patients at health centers with no licensed mental health staff compared to 32% for patients at health centers with at least one licensed mental health FTE (Table 2). This difference was significant at the p<.05 level. Our analysis showed that receiving care at a health center with limited mental health staff (less than one FTE vs. none) did not make a significant difference in predicting the probability of receiving mental health treatment. The type of mental health staff employed at the health center also did not make a significant difference in the probability of a patient receiving any mental health treatment.

TABLE 2.

Predicted probabilities (%) of patients’ receipt of mental health treatment on site at their health center or off site in the past 12 months (N=4,575)^a

Model and variableb	%	SEc	Marginal differenced	p>|z| for difference
Model 1: any licensed mental health staff (primary predictor)
No mental health staff (reference)	22	2
Fewer than one mental health FTE per 2,000 patients	24	1	2	.370
At least one mental health FTE per 2,000 patients	32	4	10	.049
Model 2: type of mental health staff (primary predictor)
No psychiatrist FTE on staff (reference)	24	2
Any psychiatrist FTE on staff	26	2	2	.507
No clinical psychologist FTE on staff (reference)	25	1
Any clinical psychologist FTE on staff	25	2	0	.881
No other licensed mental health provider FTE on staff (reference)	25	3
Any other licensed mental health provider FTE on staff	25	1	0	.909

^aSource: Uniform Data System 2013 and Health Center Patient Survey 2014

^bFTE, full-time equivalent.

^cDelta method SE.

^dDifferences are based on exact numbers, not rounded numbers displayed here

The predicted probabilities of on-site mental health treatment among patients who reported receiving any mental health treatment within the last year are displayed in Table 3. Patients at health centers with no licensed mental health FTE had a significantly lower predicted probability of receiving mental health treatment on-site (28%) than those at health centers with less than one licensed mental health FTE (49%) and patients at health centers with at least one licensed mental health FTE (65%).

TABLE 3.

Predicted probabilities (%) of patients’ receipt of mental health treatment on-site (vs. offsite) at their health center in the past 12 months (N=1,130)^a

Model and variableb	%	SEc	Difference from base leveld	p>|z| for difference
Model 1: any licensed mental health staff (primary predictor)
No mental health staff (reference)	28	6
Fewer than one mental health FTE per 2,000 patients	49	3	22	.001
At least one mental health FTE per 2,000 patients	65	7	38	<.001
Model 2: type of mental health staff (primary predictor)
No psychiatrist FTE on staff (reference)	40	4
Any psychiatrist FTE on staff	58	4	17	.002
No clinical psychologist FTE on staff (reference)	50	3
Any clinical psychologist FTE on staff	50	3	0	.935
No other licensed mental health provider FTE on staff (reference)	42	5
Any other licensed mental health provider FTE on staff	51	3	9	.133

^aSource: Uniform Data System 2013 and Health Center Patient Survey 2014. The sample only includes patients who reported they received mental health treatment within the past 12 months (on-site or offsite).

^bFTE, full-time equivalent.

^cDelta method SE.

^dDifferences are based on exact numbers, not rounded numbers displayed here

In the model assessing the relationship between type of mental health staff at a patient’s health center and where their mental health treatment took place, the predicted probability of receiving mental health treatment on-site was significantly higher (58% vs. 40%) if the health center had any psychiatrist on staff vs. no psychiatrist. However, no other mental health staff categories were significantly associated with this outcome. Full regression models are found in the online supplement. Sensitivity analyses showed that patients with similar results for patients with depression or generalized anxiety and those with moderate K6 scores but not for patients with panic disorder, schizophrenia, or bipolar disorder (n=1,013) or those with mild or severe K6 scores, (n=749, see online supplement).

Discussion

Our study found that although most health centers had some mental health staff in 2013, only a small proportion had at least one licensed mental health FTE per 2,000 patients in that year. We found the predicted probability of patients receiving MH treatment was higher only if the health center had at least one licensed mental health FTE, but the type of mental health FTE did not make a significant difference in this case. We also found that having a licensed mental health provider on staff increased the predicted probability of patients having at least some mental health treatment on-site. When different categories of mental health staff were considered separately, we found that having any co-located psychiatrist (part-time or full-time) increased this predicted probability, but the other types of staff did not.

Our results are consistent with previous research that suggests integrated care through co-location of mental health staff increases access to or utilization of mental health services.(23, 24) This increase is likely due to the greater capacity to diagnose and treat patients' mental health conditions, rather than having to refer them offsite. Our findings highlight the importance of having at least one mental health FTE to increase the predicted probability of patients receiving such treatment. This is likely because part-time capacity does not adequately support systematic screening of patients due to limited availability of mental health staff to engage in warm handoffs with primary care patients, and to diagnose and treat all patients that need such care.(25) The stratified analyses suggested that co-location of mental health staffing may have benefited patients with moderate psychological distress or less complex mental health conditions that could be treated in primary care settings. Those with mild psychological distress may have received care from PCPs.

We also found that co-location of psychiatrists significantly increased the predicted probability of receiving on-site treatment among patients who received any mental health treatment. This is likely because psychiatrists are able to prescribe medications for patients who would otherwise be referred for this care elsewhere and would have trouble finding providers willing to accept Medicaid coverage.(26)

The cross-sectional design of HCPS limits our ability to determine a causal relationship between mental health staffing at health centers and patients receiving treatment, even though data on such staffing were collected the year before patients reported receiving their mental health treatment. It is also possible that health centers’ staffing levels changed from 2013 to 2014 when patients were surveyed. Additionally, for health centers with multiple clinics, clinic-level co-location would have varied from that of the whole organization and patients may have received mental health services at a different location. Nevertheless, when health center providers can refer patients within the same organization, there are still fewer barriers to accessing services due to their mission to provide care to all patients regardless of the ability to pay. Consequently, our data highlight the value of having at least one co-located mental health FTE within the health center organization whether it is a single or multi-site health center. We also lacked data to assess continuity of care of patients with the same mental health provider, which may have played a role in their decision to seek mental health services at the health center.(27)

Finally, due to the sensitive nature of mental health issues, patients may have underreported such visits or felt uncomfortable responding to the K6 questions.(28) Our measure of psychological distress only captured problems within the last 30 days rather than the last year and may therefore under-represent level of mental health need in this population. Despite these limitations, our study is the first to use nationally representative data to examine the relationship between mental health staffing at health centers and low-income and uninsured patients receiving mental health treatment.

Conclusion

Our findings indicate that co-locating mental health staff at health centers can help low-income and uninsured patients receive mental health treatment. Yet, health centers face challenges in employing such staff in proportion to the needs of their patients. Health centers compete with other employers for mental health staff who are linguistically and culturally competent and are willing to work for less competitive salaries, particularly in the case of psychiatrists.(26, 29) As a potential model for improving co-location of these providers in community clinics, HRSA has implemented the Behavioral Health Workforce Education and Training Program to recruit mental health providers, as well as periodically disbursing mental health workforce grants to promote co-location of such staff.(30) In recent years, HRSA has further invested in increasing access to integrated mental health care by awarding $200 million through “Access Increases in Mental Health and Substance Abuse Services” in 2017; over $350 million through “Expanding Access to Quality Substance Use Disorder and Mental Health Services” in 2018; and nearly $200 million through “Integrated Behavioral Health Services” in 2019.(31, 32) However, co-location alone does not guarantee integrated care delivery at health centers, which is likely to improve overall outcomes of care and increase efficiencies. Full integration requires intensive efforts to promote close interaction between PCPs and mental health staff in patient care and requires commitment from the leadership, change in workflows, and provider buy-in.(33) Integration can be promoted by providing financial incentives and technical assistance to health centers.(34)

Our findings indicate further effort to develop mental health capacity at health centers is warranted, and that additional research is required to study use patterns of these patients to obtain insights into the role of co-located mental health staff in access to mental health services, quality of care provided, and improvements in mental and physical health.(29) Further studies could also quantitatively examine the application of tele-mental health, particularly in rural areas where there is a shortage of mental health providers.

Highlights.

• Hiring more licensed mental health staff at Health Resources and Services Administration-funded health centers may offer a promising way to increase patient access to mental health treatment.

• Attending a health center with at least one licensed mental health full-time equivalent (FTE) increased the probability of receiving mental health treatment compared to attending one with less than one licensed mental health FTE or no mental health staff.

• Attending a health center with licensed mental health staff on-site also increased the predicted probability of receiving mental health treatment, and this probability was significantly higher if the health center had a co-located psychiatrist.

• Co-located licensed mental health staff at health centers can help low-income and uninsured patients receive mental health treatment, and further efforts to develop co-located mental health capacity at health centers are warranted.