|Home | About | Journals | Submit | Contact Us | Français|
Psychiatric illness is associated with increased medical morbidity and mortality. Studies of primary care utilization by patients with psychiatric disorders have been limited by nonrepresentative samples and confounding by medical co-morbidity.
To determine whether patients with psychiatric disorders use primary care services differently than patients without these disorders, after controlling for medical co-morbidity.
Data from the 1999 Large Health Survey of Veterans (LHS) (=559,985) were linked to VA administrative data in order to identify veterans who received primary care. After adjusting for sociodemographic and clinical characteristics, medical co-morbidity, and facility characteristics, multivariate logistic regression was used to evaluate whether seven psychiatric diagnoses were associated with an increased or decreased likelihood of any primary care visit over 12 months.
Veterans with either schizophrenia, bipolar disorder or a drug use disorder were less likely to have had any primary care visit during the study period: [OR 0.61, 95% CI 0.59 to 0.63], [OR 0.63, 95% CI 0.60 to 0.67] and [OR 0.88, 95% CI 0.83 to 0.92], respectively, than veterans without these diagnoses, even after controlling for medical co-morbidity. Among patients with any primary care utilization, those with six of the seven psychiatric diagnoses had fewer visits in the study period.
Patients with schizophrenia, bipolar disorder or drug use disorders use less primary care than patients without these disorders. Interventions are needed to increase engagement in primary care by these vulnerable groups.
Serious mental illness (SMI), such as schizophrenia or bipolar disorder, has been associated with increased medical morbidity and mortality. Patients with schizophrenia have elevated rates of diabetes and emphysema1–4, and are twice as likely to die from cardiovascular disease as the general population5. Bipolar disorder is associated with elevated rates of hypertension and cardiovascular disease6,7. Several factors have been proposed to explain this increased morbidity and mortality, including high rates of obesity and tobacco dependence8,9, side effects of psychiatric medications, neglect of physical health because of psychosis, anxiety, depression, or social withdrawal10, and under-utilization of medical services11.
Several studies have suggested that persons with SMI are less likely to use medical care than individuals without SMI. Over 60% of patients with SMI report difficulty keeping medical appointments12. In a study of VA outpatients with chronic medical conditions, patients with co-morbid mental disorders were less likely to receive recommended preventive services than patients without such disorders13. In another study, only half of patients with SMI and co-morbid chronic hepatitis C infection had a primary care provider14. There may be disparities in the medical service use across psychiatric diagnoses: older persons with schizophrenia appear to receive fewer primary care and preventive health care services than patients with depression15. Patients with schizophrenia have very low rates of treatment of hypertension and hyperlipidemia16, and those with co-morbid diabetes receive less aggressive treatment of cardiovascular risk factors17.
But other studies suggest that patients with SMI may use as much or more medical services than the general population. Data from the 1994 National Health Interview Survey suggest that persons with a mental disorder are as likely to have a primary care provider as persons without a mental disorder18. Moreover, a study of 200 outpatients from mental health clinics in Baltimore found that the patients with schizophrenia and affective disorders were more likely to report receiving medical services in the past year than individuals in the general population19.
The inconsistencies in the existing literature reflect two important methodological limitations of previous studies. First, the samples in the previous studies may not be representative of the SMI population. Second, previous evaluations of disparities in outpatient medical service use have not controlled for potential confounding by medical co-morbidity, which may be a more important contributor to outpatient medical service utilization than any specific psychiatric diagnosis20.
Whether patients with schizophrenia and bipolar disorder under-utilize primary care in relation to their medical need is a critical question that needs to be answered to improve the quality of medical care for these vulnerable patients. The current study uses data from the 1999 Large Health Survey of Veteran Enrollees (LHS) to evaluate primary care service utilization among veterans with psychiatric illness while addressing the methodologic limitations of previous studies. The specific aims of this study were 1) to determine whether patients with psychiatric illness use primary care services differently than patients without psychiatric illness, after controlling for potential confounding by the presence of medical co-morbidity and facility characteristics; and 2) to quantify the magnitude of this effect, specifically the number of primary care visits once contact has been initiated. Seven psychiatric diagnoses were evaluated in order to determine whether any differential use of primary care services is generally true across a broad range of psychiatric illness.
Sample Analyses were conducted using data from the 1999 Large Health Survey of Veteran Enrollees (LHS), the largest and most detailed survey of users of VA health services ever conducted21. The LHS was a structured questionnaire on sociodemographic factors and health status, health behaviors and health care needs. It included a version of the Medical Outcomes Study Short Form 3622 specifically validated among veterans23, and a medical co-morbidity index which was based on patient self-report of diagnoses24. Patients were sampled from the March 1999–September 1999 enrollment files. A total of 1,406,049 enrollees were sent surveys and 887,775 returned the questionnaire, for a response rate of 63.14%21. A total of 559,985 respondents who used VA services in fiscal year 2000 responded to the survey between July 1999–December 1999 and had complete data on survey variables used in this study. The LHS dataset was linked to VA administrative data in order to identify whether the patient had any primary care visit in FY 2000 and the number of primary care visits patients received. The representativeness of the LHS sample was demonstrated by comparison of sociodemographic and clinical characteristics of the LHS sample to data on all patients who used VA health services nationally during FY2000 (=3,647,334 patients).
Psychiatric diagnoses Veterans with psychiatric illness were identified from administrative data by having at least one visit with an ICD-9-CM code25 for any of seven psychiatric diagnoses (as a primary or secondary diagnosis): schizophrenia or schizoaffective disorder (295.xx), bipolar disorder (296.0×, 296.1×, 296.40–296.89), major depressive disorder (296.2–296.39), other depression (300.4×, 296.9×, 311.xx, 301.10–301.19), posttraumatic stress disorder or PTSD (309.81), alcohol use disorder (303.xx or 305.00), or drug use disorder (292.01–292.99 or 304.xx or 305.20–305.99). The “other depression” category was comprised almost entirely of dysthymia (300.4×) and depressive disorder not otherwise specified (311.xx).
Medical co-morbidity Medical co-morbidity was evaluated based on patient self-report of medical diagnoses in the LHS survey. Veterans were asked about 12 common medical conditions in the survey, and the count of the diagnoses was used in the analyses.
Physical health status Physical health status was evaluated by the physical component scale (PCS) of the Short Form-36 for Veterans, an adapted form of the Medical Outcomes Study Short Form 3622 designed specifically for use with veterans23. It consists of the same 36 items and eight sections as the MOS SF-36: physical functioning, role limitations due to physical problems, bodily pain, general health perceptions, energy/vitality, social functioning, role limitations due to emotional problems, and mental health. Responses in the two role functioning scales are on a five-point ordinal scale, which differs from the dichotomous responses in the MOS SF-3626. The PCS is standardized to the national U.S. population with a mean of 50 and a standard deviation of 10, where higher scores denote better health status.
VA Site characteristics Characteristics of the VA facilities where subjects obtained their medical care were evaluated to identify differences in patterns of primary care utilization in terms of geographic location, size of the VA facility, and emphasis on mental health care. We used Rural-Urban Commuting Area (RUCA) codes developed in 1998 at the University of Washington (http://depts.washington.edu/uwruca/about.html) to identify veterans living in settings ranging from large urban locations to isolated rural ones, using the zip code of residence in the first outpatient encounter record. Facility characteristics included the size of the VA hospital (number of employees), and emphasis on mental health care (the percentage of the budget dedicated to mental health treatment, and the percentage of the mental health budget dedicated to research and education).
Primary care utilization The primary outcome of the study was receipt of any VA outpatient primary care services during FY2000. This was determined using clinic stop codes in the VA Outpatient Care File, a computerized database of all ambulatory care encounters at VA facilities represented by stop codes 301, 322, 323, 348, and 350. Sensitivity analyses expanded the definition of primary care to include all medical specialty clinics and mental health primary care clinics.
Statistical methods Four separate analyses were conducted. First, the representativeness of the LHS sample in terms of sociodemographic and diagnostic characteristics, and outpatient mental health and medical service utilization was evaluated by comparison to data on all VA health service users in FY 2000 (=3,647,334).
Second, multivariate logistic regression analyses were conducted to evaluate whether veterans with specific psychiatric diagnoses use primary care services differently than patients without psychiatric illness. Three regression models were evaluated, all with the outcome of any primary care visit in FY 2000. The first model evaluated the odds of any primary care visit across the seven psychiatric diagnoses controlling for sociodemographic characteristics (age, race, gender, education, and service connection). Service connection reflects compensation for a disability connected to military service, and may affect priority for services and motivation to seek services. In the second model, covariates for medical co-morbidity (the count of self-reported diagnoses in the LHS survey) and physical health status (SF-36V PCS) were added. In the third model, VA site characteristics were added: rurality of residence, size of VA facility, the facility’s emphasis on mental health care, and the percentage of the mental health budget spent on research and education.
Third, interaction terms between facility characteristics and mental health diagnoses were evaluated. These analyses required the use of general estimation equations (GEE) to adjust the standard errors of the coefficients because observations within facilities are not independent.
Fourth, the number of primary care visits was evaluated across the seven psychiatric diagnoses (among those veterans who had any primary care visit). Negative binomial regression technique was used because the count of the number of primary care encounters did not meet the distributional assumptions of ordinary least squares analysis. Analyses were conducted in Proc Genmod of SAS®.
The sample of 559,985 patients who had complete data on the LHS survey appear to be broadly representative of the 3,647,334 patients who used VA health services in FY 2000 (Table 1). They differed in that patients in the LHS subsample were, on average, older (mean age 64.1 compared to 58.9), used less outpatient mental health care (mean of 2.5 visits compared to 2.7), were more likely to have a primary care visit (85.1% compared to 72.1%), and had more outpatient medical visits (4.8 compared to 3.1 visits). All of these differences were statistically significant at <0.0001. The percentage of subjects with any primary care visit during the study period varied across the psychiatric diagnoses, from 72.8% of subjects with schizophrenia to 90.1% of subjects with other depression.
Multivariate logistic regression analyses were used to evaluate whether each of the seven psychiatric diagnoses was associated with an increased or decreased likelihood of any primary care visit in FY 2000. Three regression models were evaluated with a baseline model including only sociodemographic variables and psychiatric diagnoses and subsequent models controlling for potential confounding by medical co-morbidity and local VA facility characteristics (Table 2). Patients with either schizophrenia, drug use disorder, or bipolar disorder were less likely to have any primary care visit during the study period than those without a psychiatric diagnosis: [OR 0.61, 95% CI 0.59 to 0.63], [OR 0.63, 95% CI 0.60 to 0.67] and [OR 0.88, 95% CI 0.83 to 0.92], respectively. Patients with major depressive disorder, other depression, or PTSD were more likely than veterans without psychiatric illness to have a primary care visit (Table 2).
When logistic regression analyses were repeated with adjustment for medical co-morbidity, patients with schizophrenia or bipolar disorder were even less likely to have a primary care visit: [OR 0.55, 95% CI 0.53 to 0.58 for schizophrenia] and [OR 0.73, 95% CI 0.69 to 0.77 for bipolar disorder]. After adjustment for medical co-morbidity, patients with major depressive disorder, PTSD and alcohol use disorders were less likely to have had a primary care visit during the study period. The “other depression” category was the only diagnostic category that was associated with an increased odds of a primary care visit after adjustment for medical co-morbidity.
This under-utilization of primary care could be explained if veterans with psychiatric diagnoses were receiving primary care through either specialty medical clinics or mental health clinics. Repeat analyses with a broadened definition of primary care to include specialty medical clinics and mental health primary care clinics increased the utilization of services by a negligible amount: the percentage of subjects with schizophrenia who had any primary care visit increased by 0.7%.
The addition of VA facility characteristics did not substantially change the findings across the psychiatric diagnoses (Table 2). Patients in rural areas were more likely to have a primary care visit, as were patients treated at facilities with a greater emphasis on mental health treatment. Next, interaction terms were added to further explore utilization among patients with either schizophrenia, bipolar disorder or drug use disorders. Among patients with schizophrenia or bipolar disorder, those treated in large VA hospitals were less likely to receive primary care (<0.0001 for both interaction terms). These analyses required use of general estimation equations (GEE) to adjust the standard errors of the coefficients for the fact that observations within facilities are not independent.
Finally, the impact of each of the seven psychiatric diagnoses on the number of primary care visits during FY2000 was evaluated using a negative binomial regression analysis, controlling for medical co-morbidity, physical functioning and VA facility characteristics (Table 3). Among LHS subjects who had at least one primary care visit, those with any psychiatric diagnosis except other depression had fewer primary care visits when compared to patients with no psychiatric diagnosis. Subjects with other depression had slightly more visits, as did patients with increased medical co-morbidity (all <0.0001, Table 3). The addition of interaction terms to these analyses revealed that patients with schizophrenia or bipolar disorder receiving treatment at large VA facilities received fewer primary care visits (<0.0001).
This study utilized data from the 1999 Large Veterans Health Survey, the largest and most detailed survey of VA health services users ever conducted, in order to determine whether veterans with psychiatric illness utilize primary care services differently than veterans without psychiatric illness. Seven psychiatric diagnostic categories were examined, evaluating a broad spectrum of Axis I psychiatric disorders. Veterans with schizophrenia, bipolar disorder and drug use disorders were significantly less likely than other veterans to have a primary care visit during the one-year study period. And among those patients with any primary care visits, veterans with these diagnoses also had fewer primary care visits.
A major strength of the LHS dataset is that it provides the opportunity to control for potential confounding of the relationship between psychiatric illness and primary care utilization by medical co-morbidity, as the survey included self-report questions about medical diagnoses. After adjustment for medical co-morbidity, veterans with each of the psychiatric disorders except other depression had decreased odds of any primary care visit during the study period. An apparent increased odds of any primary care visit among patients with major depressive disorder, PTSD, or alcohol use disorders disappeared after adjustment for medical co-morbidity, suggesting either that increased utilization of services is explained by an increased burden of medical illness, or that these findings are less robust than the findings for patients with schizophrenia, bipolar disorder and drug use disorders. The other depression category is less precise than the other diagnostic categories, and may include patients with undiagnosed somatoform disorders.
This study provides clear evidence that veterans with schizophrenia and bipolar disorder have decreased utilization of needed primary care services. The national VA system is the largest integrated health care system in the U.S., and access to primary care could be expected to be easier for patients with serious mental illness (compared to community mental health settings). But in this study, patients with schizophrenia and bipolar disorder were less likely to have any primary care visit and had fewer visits, even after controlling for medical co-morbidity. Previous studies which suggested that patients with schizophrenia have as many (or more) primary care visits as patients without psychiatric illness19,20, did not adjust for medical co-morbidity and likely had less representative samples. Our study supports previous reports of under-utilization of primary care among patients with drug use disorders13,27.
Our study had several limitations. First, psychiatric diagnoses were based on administrative data, and previous literature has demonstrated significant under-diagnosis of psychiatric illness28,29. We attempted to diminish the impact of this under-diagnosis by identifying patients with psychiatric illness as having at least one encounter with either a primary or secondary diagnosis of the disorder. Moreover, any potential to bias our results would likely be to falsely attenuate the differences between the groups, as patients with disorders would have mistakenly been categorized as having no psychiatric illness.
Second, it is possible that some veterans received primary care services outside of the VA, and such visits would not have been captured by VA administrative data. But it is unlikely that veterans with schizophrenia, bipolar disorder or drug abuse would be especially likely to use non-VA services, since they have lower incomes and are less likely to have private insurance.
Third, in 2005 the VA implemented a Mental Health Strategic Plan30 to enhance the link between medical and mental health services which may be mitigating the disparities we identified. Moreover, veterans returning from the war in Afghanistan and the Iraq War may have different patterns of utilization of medical services. But as these groups comprise <10% of all users of VA mental health services, it is unlikely that their utilization patterns would significantly alter our findings at the current time.
Finally, our findings may have limited generalizability outside of the VA system. Our previous research suggests similar rates of medical service utilization among VA patients and non-VA patients31,32. But since the VA is an integrated system, if these groups do differ in their utilization patterns, our findings likely over-estimate the access of typical patients with schizophrenia, bipolar disorder and drug use disorders to primary care services.
In conclusion, this study suggests that patients diagnosed with schizophrenia, bipolar disorder or drug use disorders receive fewer primary care services than patients without these disorders. These findings are of particular concern because the VA is an integrated system which provides both mental health and medical services and thus may present a “best case” scenario for access to care among these vulnerable patient groups. Interventions are needed to facilitate the utilization of primary care by patients with serious mental illness, especially in light of the increased burden of medical illness in this population.
The research reported here was supported by a grant (K23 MH077824) from the National Institute of Mental Health.
Conflict of Interest None disclosed.
The SF-36® and SF-12® are registered trademarks of the Medical Outcomes Trust