Quality of care for substance use disorders in patients with serious mental illness☆

1. Introduction 

Research suggests that nearly half of all individuals with serious mental illness develop a co-occurring substance use disorder (SUD) in their lifetime (Rach-Beisel, Scott, & Dixon, 1999). Co-occurring SUDs in patients with serious mental illness can disrupt psychiatric treatment and result in adverse health outcomes (Crome, 1999). These factors have lead to a growing national recognition of the importance of identifying and properly treating SUD in seriously mentally ill patients (Crome, 1999).

Recently, health-care policymakers and stakeholders such as the Washington Circle have developed and implemented indicators to monitor the quality of SUD care in the general population as a means to improve the quality of care and, ultimately, outcomes for individuals with SUD (Garnick et al., 2002). However, the extent to which seriously mentally ill patients are receiving adequate quality of SUD care has not been explored. Using the Washington Circle indicators, we assessed the quality of SUD care among patients with serious mental illness receiving mental health care in the VA.

2. Materials and methods 

We conducted a retrospective analysis of data from fiscal years (FY) 2001–2002 (October 1, 2000, to September 30, 2002) from the VA National Patient Care Database (NPCD). The NPCD includes information on all inpatient and outpatient visits and medical and psychiatric ICD-9 diagnoses. This study was reviewed and approved by local institutional review boards. Subjects were not required to give informed consent, as this was a secondary analysis of existing data.

3. Results 

A total of 8,083 patients in our sample received a diagnosis of serious mental illness in FY 2001. Of the 8,083, the mean age was 54 years (SD = 12, range = 19–100 years) and 6.9% (n = 554) were women (Table 1).

Table 1. Quality of SUD care for patients diagnosed with serious mental illness

		Overall sample (N = 8,083)		Identification of SUD (N = 8,083)a				Initiation of SUD treatment (n = 1,559)b				Engagement of SUD treatment (n = 1,559)c
		n	%	n	%	χ2	p	n	%	χ2	p	n	%	χ2	p
	Psychiatric diagnosis					26.02	<.001			59.29	<.001			72.25	<.001
	Schizophrenia	4,932	61.0	133	18.9			94	70.7	84	63.2
	Schizoaffective disorder	705	8.7	872	17.7			598	68.6	542	62.2
	Bipolar disorder	2,446	30.3	554	22.7			272	49.1	221	39.9
	Age					252.77	<.001			18.13	<.001			15.65	<.001
	<50 years	3,209	39.7	862	26.9			567	65.8	499	57.9
	50–64 years	3,161	39.1	551	17.4			326	59.2	288	52.3
	≥65 years	1,713	21.2	146	8.5			71	48.6	60	41.1
	Sex					7.08	.008			0.02	.88			0.06	.80
	Female	554	6.9	83	15.0			52	62.7	44	53.0
	Male	7,529	93.1	1,476	19.6			912	61.8	803	54.4
	Race					120.74	<.001			25.47	<.001			22.75	<.001
	Black	1,763	21.8	201	28.4			355	70.9	316	63.1
	Not Black	6,320	78.2	1,058	16.7			609	57.6	531	50.2
	Copayment waiverd					4.63	.03			7.63	.006			6.44	.01
	Yes	7,762	96.0	1,512	19.5			944	62.4	830	54.9
	No	321	4.0	47	14.6			20	42.6	17	36.2
	Marital status					80.01	<.001			12.25	<.001			7.99	.005
	Not married	6,035	74.7	1,302	21.6			830	63.8	728	55.9
	Married	2,048	25.3	257	12.6			134	52.1	119	46.3
	No. of general medical conditionse					395.36	<.001			10.13	.02			13.70	.003
	0	2,203	27.3	120	5.5			68	56.7		61	50.8
	1	2,149	26.6	470	21.9			269	57.2		228	48.5
	2	1,690		405	24.0			254	62.7		221	54.6
	≥3	2,041	29.4	564	27.6			373	66.1		337	59.8

Patients were considered to have a serious mental illness if they had at least one inpatient or two separate outpatient diagnosis of schizophrenia, schizoaffective disorder, or bipolar disorder in FY 2001. For patients with either one inpatient or two outpatient diagnoses for more than 1 condition, the following decision hierarchy was used: schizophrenia, schizoaffective disorder, bipolar disorder.

a Identification of alcohol or drug use disorder (SUD) is defined as an ICD-9 code for alcohol or drug psychosis, abuse, or dependence diagnosis. b Initiation of treatment for an SUD is defined as an inpatient or outpatient follow-up contact less than 14 days after the initial SUD diagnosis. c Engagement of SUD treatment is defined as 2 or more outpatient visits for substance use treatment less than 30 days after SUD diagnosis. d Patients required to pay a copayment are ineligible for low income-based copayment waiver. Patients not requiring a copayment are considered lower income. e Number of medical conditions diagnosed in FY 2001 is based on a count of medical conditions from the Agency for Healthcare Research and Quality Clinical Classifications Software and included hypertension, congestive heart failure, peripheral vascular disease, stroke, ischemic heart disease, diabetes, hyperlipidemia, pancreatitis, thyroid disorders, obesity, hepatitis, lower back pain, arthritis, chronic obstructive pulmonary disease, asthma, cancers, and other conditions.

Overall, 19.2% (n = 1,559) were identified as having a diagnosis of SUD. Patients diagnosed with bipolar disorder were more likely than those diagnosed with schizophrenia or schizoaffective disorder to be identified as having an SUD (22.7%, 18.9%, and 17.7%, respectively; χ2 = 26.02, df = 2, p < .001). Non-African American patients were less likely than African American patients to initiate treatment (Table 1).

Of the 1,559, 62.0% (n = 966) received treatment initiation. Patients diagnosed with bipolar disorder were less likely to initiate treatment than those diagnosed with schizophrenia or schizoaffective disorder (49.1%, 70.7%, and 68.6%, respectively; χ2 = 59.29, df = 2, p < .001). Non-African American patients were less likely than African American patients to initiate treatment (Table 1).

Of the 1,559 patients diagnosed with SUD, 54.3% (n = 847) continued treatment engagement. Patients diagnosed with bipolar disorder ere less likely to continue treatment engagement than those diagnosed with schizophrenia or schizoaffective disorder (39.9%, 63.2%, and 62.2%, respectively; χ2 = 72.25, df = 2, p < .001). Non-African American patients were less likely than their African American counterparts to engage in treatment (Table 1).

Confirmatory multivariable logistic regression analyses adjusting for patient and facility factors revealed that patients diagnosed with bipolar disorder were still more likely to be diagnosed with an SUD (adjusted OR = 1.90, p < .001), yet less likely to initiate SUD treatment (adjusted OR = .49, p < .001) and continue treatment engagement (adjusted OR = 0.44, p < .001). The association between race and identification, initiation, and treatment engagement was no longer significant (data not shown).

4. Discussion 

Fewer than 20% of patients with serious mental illness in our sample received a diagnosis of an SUD. The prevalence of SUD diagnosis in our sample was lower than what would be expected among patients with serious mental illness (Rach-Beisel et al., 1999). This finding may reflect potential underreporting of SUDs in patients with serious mental illness, thereby reflecting suboptimal quality of care.

Of those identified with an SUD, less than two thirds initiated treatment, and only about half continued SUD treatment engagement. We were surprised to find that patients diagnosed with bipolar disorder, although more likely to be identified as having an SUD, were less likely to initiate and engage in treatment. Previous research has suggested that patients with bipolar disorder are especially prone to substance abuse or dependence (Bauer, Unutzer, & Pincus, 2002). Perhaps, the cyclical nature of bipolar disorder episodes may have contributed to the lack of adherence to follow-up treatment.

However, treatment initiation and engagement estimates were higher than reported elsewhere, notably in studies that applied the Washington Circle measures to a general medical patient population outside the VA setting (e.g., 26% initiation and 14% engagement, respectively) (Garnick et al., 2002). Our results are likely attributed to the VA's ongoing efforts to organize mental health and SUD services under the same administrative entity. Specifically, mental health and SUD care are typically housed under the same administration across VA sites, which in turn operate under similar budget and treatment delivery protocols (e.g., inpatient and outpatient care). Still, in many cases, integration at the administrative level may not necessarily translate into “integrated” care at the provider level. For example, based on a survey of VA mental health facilities across the United States, only 20% provided both mental health and SUD care within the same treatment setting and less than half had on-site addiction specialists (Tracy, Trafton, & Humphreys, 2004). This lack of integration at the provider level may impede efforts to coordinate treatment initiation and engagement for individuals with serious mental illness.

Hence, the implementation of performance measures such as the Washington Circle criteria is an important first step in holding administrators and other stakeholders accountable for better integrating services and subsequently improving the quality of SUD care for patients with serious mental illness. Our results suggest potential gaps in quality of care for SUD among patients with serious mental illness, based on three established measures of processes of care for SUD. Quality indicators (i.e., measures of processes and outcomes of care) are important tools for improving access and quality of care for individuals with SUD, because they can be used benchmark good care and potentially hold service providers accountable for performance improvement (Garnick et al., 2002). Indicators can provide a summary or how care is being delivered and can be used to identify potential gaps in quality. Ultimately, these indicators can be applied to monitor and subsequently improve the quality of care for individuals with SUD.

The application of process measures that are easily derived from administrative data, such as the Washington Circle indicators used in this study, serve as potentially important markers for poor outcomes (Garnick et al., 2002, Moos et al., 2000). In measuring quality, two types of indicators are often employed: processes of care indicators, such as identification or engagement, and clinical outcomes (Donabedian, 1986, McGlynn, 1998). Outcome measures such as abstinence from substance abuse or decreased psychiatric symptoms proposed in the Substance Abuse and Mental Health Services Administration's (SAMHSA) National Outcome Measures (NOMs) Initiative (http://www.nationaloutcomemeasures.samhsa.gov/) are desirable because quality is ultimately demonstrated when a health-care system is achieving overall goals in improving outcomes. Nonetheless, processes of care measures such as the Washington Circle indicators are often preferred because they represent care for which providers are perceived to have the most control over, and because they can be derived from existing claims data rather than from primary data collection from patients, which can be costly (McGlynn, 1998). Moreover, observed differences in outcomes may be attributed to patient-level differences in clinical severity rather than differences in the care provided. The long-term use of outcome measures as quality indicators may also lead to potential “gaming” by providers (e.g., providers selecting healthier patients). Still, to identify patients who are vulnerable to poor outcomes, future efforts should focus on evaluating the validity of process measures such as the Washington Circle criteria to determine whether they are adequate indicators for “good” or “bad” care, when compared to clinical outcomes.

Key limitations to this study include the inability to determine potential reasons for gaps in quality (e.g., patient preferences) and the limited number of women. However, the VA patient population includes a disproportionate number of older, minority, and seriously mentally ill individuals, reflecting similar demographics seen in other publicly funded health-care providers (e.g., Medicaid). In addition, there is the possibility that patients received SUD treatment external to the VA and hence, was not observed in our quality data. Nonetheless, many veterans with serious mental illness are indigent and may rely on the VA for the majority of their mental health care, because of its generous outpatient and inpatient benefits.

Finally, this study was not designed to evaluate an equally important problem: gaps in quality of mental health care for patients receiving care for SUDs. Recent initiatives such as SAMHSA “No Wrong Door” policy (SAMHSA Co-Occurring Centers for Excellence, 2005) stress the need to better integrate care and reduce organizational barriers between traditional mental health and SUD providers (Donabedian, 1986). As a means to hold both mental health and SUD service providers accountable for improved quality of care, SAMHSA's NOMs stress the need to implement outcomes measures that are applicable to dually diagnosed individuals regardless of their point of entry.

Overall, SUD may be underreported in the VA among individuals with serious mental illness. Although the VA is a leader in integrating mental health and SUD treatment services, potential gaps in the initiation and engagement of SUD treatment for patients diagnosed with bipolar disorder still exist and should be addressed. Finally, as healthcare providers and policymakers call for efforts to improve SUD care, implementing performance measures based on available (administrative) data can help better identify patient subpopulations who are potentially vulnerable to suboptimal quality of care and can be used to benchmark and hold healthcare administrators accountable for improving quality.