“The mission of the Mental Health Division is to promote recovery and safety.”
The Prevalence of Serious Mental Illness in Washington State
Copy of the final report to the Legislature required by Chapter 7, Laws of 2001, E2, Section 204(5) (c) and Chapter 25, Laws of 2002, E1, Section 204(5)(b), December 1, 2003 is now available. Questions should be addressed to Judy Hall, Ph.D.(halljj@dshs.wa.gov) or call her at (360) 902-0874.
[5,146KB]
[819KB]This report is the Mental Health Division (MHD) response to Chapter 7, Laws of 2001, E2, Section 204(5)(c) and Chapter 25, Laws of 2003, E1 Section 204(5)(b). That legislation mandated that a study "shall examine how reasonable estimates of the prevalence of mental illness relate to the incidence of persons enrolled in medical assistance programs in each regional support network area."
To meet this charge, the Department of Social and Health Services (DSHS) MHD convened a Prevalence Advisory Committee (PAC), consisting of Regional Support Network (RSN), provider, consumer, research, and Joint Legislative Audit Review Committee (JLARC) representatives. This group met monthly with project staff for two years to design the study, guide implementation, and review results. In addition, MHD convened an Expert Panel, consisting of leading mental health epidemiology researchers that reviewed study issues and assisted in design, implementation, and interpretation. Working with project staff, PAC and the Expert Panel first identified project goals and decided upon groups for further study to develop reasonable estimates of the prevalence of serious mental illness (SMI) in adults and serious emotional disturbance (SED) in children.
The development of the study plan and the results of a series of ten separate sub-studies are detailed in Chapters 1 through 9 of the full report. The last chapter of the full report integrates the results of all studies, compares results to other prevalence studies, examines how these estimates relate to Medicaid eligibility, and makes recommendations for future studies.
This study revises a prior study looking at the prevalence of mental illness in Washington State. The 1998 study, Prevalence Estimation of Mental Illness and Need for Services (PEMINS) study, used a telephone survey of approximately 7,000 Washington residents to calculate prevalence estimates statewide, by county, and by region. The current study (hereafter referred to as PEMINS 2000) differs from the 1998 PEMINS study in four major respects:
- It is based on the most current full U.S. Census (2000).
- Separate smaller studies were conducted to produce estimates for several non-household target groups overlooked or underestimated in the original PEMINS study. These results are added to the household prevalence estimates to estimate the prevalence of SMI in all adults.
- Estimates for children with SED are included.
- Race and ethnicity were used as predictors in the earlier study to generate estimates by region and county for 1998. Race and ethnicity were removed as predictors from estimation equations in the current study. This decision by PAC results from an extensive literature review that did not reveal any significant differences in the prevalence of SMI in minority populations or ethnic groups.
Methodology
Preliminary discussions of the first PEMINS study led PAC to focus on three areas of concern:
- Groups requiring more specific estimates, such as the homeless and those living in institutional settings.
- Groups not specifically addressed in PEMINS 1998, including children, minorities, immigrants, and refugees who may not have been adequately counted in the household survey.
- Drift. The possibility that persons with mental illness may move for reasons related to the location of social and mental health services, or access to those services, resulting in disproportionate numbers in some regions.
The following target groups were identified for further study: children, the homeless, jail and prison populations, children in juvenile facilities, hospital populations, and residents of rehabilitation and group homes in communities. Each target group study is presented as a chapter in this report.
PAC developed a plan with the following steps:
- Recalculate PEMINS 1998 household estimates using current 2000 U.S. Census data.
- Use the research literature and conduct additional studies where needed to develop specific prevalence rate estimates for each target group.
- Add the target group calculations for each RSN to the revised household estimates (PEMINS 2000) to generate prevalence estimates for each RSN.
- Address additional groups of interest in the report.
Serious Mental Illness Operationally Defined
PEMINS 1998 had provided regional estimates according to 13 different models of mental health need that varied according to diagnosis and functional impairment. PAC decided that the populations served by the RSNs, and the mandate of the enabling legislation, were most closely matched by the following medium-band definition of mental health need used in the original PEMINS study:
Respondent has a major disorder (such as depression, psychosis, or manic episodes) and meets at least one of these additional criteria:
- Functional limitation that limits major life activities, ability to work, or taking care of personal needs such as bathing;
- Mental health (MH) services use or desire for MH services;
- Danger to self or others;
- Dependence, i.e., inability to support one's self or provide for one's own medical care.
More detail on definitions of mental illness is offered in Chapter 2 of the report.
Other Groups Considered
A few groups were the focus of extensive PAC discussions, but no effective methodology was adopted to estimate their influence on prevalence rates. These groups were not studied further because of a lack of published SMI estimates or population estimates or both. The resources that would have been required to study these groups would have far exceeded the resources of the current study. These included migration and drift of mentally ill persons, recent immigrants, and race and ethnic issues in prevalence studies. These are addressed in more detail in Chapter 9 of the full report and are summarized briefly below.
Results
PEMINS Recalculations
The current study applied the methodology of PEMINS 1998 to U.S. Census 2000 data. In addition, PAC requested that MHD staff work with Dr. Charles Holzer of the Psychiatry and Behavioral Sciences Department at the University of Texas Medical Branch in Galveston, Texas to produce alternative household prevalence estimates:
• PAC requested separate estimates for all households and for
households with incomes at or below 200% of the Federal poverty level
(FPL) as a proxy measure of those in need of public mental health services.
•
PAC requested results from estimation models that excluded race and ethnicity
as predictors in estimation equations.
Household and Target Group Estimates
Results of this analysis indicate that rates of SMI in households increased slightly between 1998 and 2000. Using race-neutral methods led to higher estimates of household rates. The following conclusions are drawn from these analyses:
- Choice of estimation method made little difference to each Regional Support Network’s proportion of the total number of persons with serious mental illness (hereafter called shares), except for King County.
- RSN shares remained relatively stable between the 1998 and 2000 estimates.
- RSN shares of persons with mental illness in households closely tracked each RSN’s share of the total state population.
- Using proportions of the state population as the standard, shares of persons with mental illness in households at <200% of FPL were disproportionately high for Greater Columbia and North Central RSNs. King County and North Sound RSNs showed disproportionately low shares, with the effect being marked in King County.
Following the study plan developed by PAC, the race-neutral household prevalence estimates from PEMINS 2000 were combined with target group estimates to yield prevalence counts, by RSN, displayed in Table ES.1. The total number of persons with SMI/SED (all income levels) was estimated at 295,884, compared to 157,070 estimated for adults in households and institutions by the original (PEMINS 1998) study. Accounting for most of this gain was the addition of 105,969 children with SED, and some increase in the institutional and homeless estimates of persons with SMI. Combining the estimated number of adults with SMI in households with incomes at or below 200% FPL with estimates of the number of children with SED living in households at or below 250% FPL, and the relevant target group estimates, yields a total of 148,732 persons likely to be dependent upon publicly supported mental health services. The estimates for each RSN are shown in Table ES.1.
Comparisons with Other Estimates
Table ES.2 compares results of this study with those from other recent studies of the prevalence of SMI in Washington State and with the number of Medicaid Eligibles:
- Blueprints for an Effective Mental Health System in Washington State (Blueprints), produced by the Washington Community Health Council worked in conjunction with NAMI Washington (2000).
- The number of Medicaid Eligibles as calculated for MHD financial services purposes.
The data presented in Table ES.2 are percentages of all known cases, to provide a common metric for comparing the studies. The bottom row of the table contains the statewide population counts for each method. The first two data columns present a comparison of the general population prevalence results of this study and the Blueprints study. This study identified 11% fewer cases, likely due to more restrictive criteria for identifying adults with serious mental illness. The relative shares for the RSNs are very similar between the two studies.
Table ES.1
Integration of Estimates from All Studies
RSNs |
Household Estimate 1) |
Target Group Studies |
Total Estimated Number of SMI |
|||||
Community Residential 2) |
Jails and Prisons 3) |
Homeless 4) |
Incarcerated Children 5) |
State Hospitals 6) |
Children 7) |
|||
| Chelan-Douglas | 2,588 |
194 |
73 |
98 |
22 |
26 |
1,977 |
4,978 |
| Clark County | 9,487 |
363 |
218 |
375 |
39 |
78 |
6,929 |
17,489 |
| Grays Harbor | 1,924 |
112 |
53 |
66 |
15 |
33 |
1,208 |
3,411 |
| Greater Columbia | 15,348 |
837 |
447 |
599 |
83 |
146 |
12,084 |
29,544 |
| King County | 52,941 |
3,254 |
1,025 |
2,793 |
144 |
642 |
27,345 |
88,144 |
| North Central | 3,357 |
251 |
94 |
129 |
25 |
40 |
2,835 |
6,731 |
| North Sound | 25,730 |
1,425 |
469 |
949 |
119 |
259 |
17,808 |
46,759 |
| Northeast | 1,872 |
97 |
34 |
68 |
7 |
21 |
1,337 |
3,436 |
| Peninsula | 8,870 |
382 |
171 |
350 |
34 |
113 |
5,696 |
15,616 |
| Pierce County | 19,442 |
1,537 |
548 |
944 |
109 |
335 |
13,340 |
36,255 |
| Southwest | 2,598 |
104 |
114 |
92 |
20 |
44 |
1,743 |
4,715 |
| Spokane County | 11,936 |
1,047 |
220 |
1,295 |
26 |
239 |
7,525 |
22,288 |
| Thurston-Mason | 7,180 |
253 |
211 |
253 |
62 |
69 |
4,490 |
12,518 |
| Timberlands | 2,420 |
170 |
107 |
92 |
25 |
27 |
1,652 |
4,493 |
| Other/Unknown | 0 |
0 |
43 |
0 |
0 |
4 |
0 |
47 |
| Total | 165,154 |
10,025 |
3,826 |
8,104 |
730 |
2,076 |
105,969 |
295,884 |
| 1) PEMINS 2000 estimate of the number of household members who meet criteria for SMI (Medium Need- Race Neutral Method). With the indirect estimation method employed in the PEMINS studies, the model is applied to each RSN and to the state totals separately. This results in small differences between the statewide PEMINS totals and the sum of the values for each of the 14 RSNs. See Chapter 2 for description of how estimates were derived. | ||||||||
| 2) See Chapter 8 for study details. | ||||||||
| 3) Based on Jail Average Daily Population data provided by the Washington Association of Sheriffs and Police Chiefs for calendar year 2001and prison data provided by the State of Washington Department of Corrections Planning and Research Section for June 30, 2002; applies rate of 12% to jail population and 15% to prison population (see Chapter 5). | ||||||||
| 4) Uses estimate of 35% applied to estimated number of homeless based on one-night-counts and a Key Informant Survey (see Chapter 4). | ||||||||
| 5) Uses estimate of 60% applied to data provided by the State of Washington Juvenile Rehabilitation Administration for calendar year 2001. Does not include youth in community facilities or tribally adjudicated youth (see Chapter 6). | ||||||||
| 6) Applies estimate of 100% prevalence for all persons in beds on May 29, 2002. See Chapter 7 for description of how estimates were derived. | ||||||||
| 7) Source: Census 2000, SF-1data file, 100% data, applying a rate of 7%. See Chapter 3 for description of how estimates were derived. | ||||||||
Table ES.2
Comparison of Estimates
RSNs |
Estimated SMI (Households-Race Neutral) + MiniStudies1 |
SMI Estimates from Blueprints |
PEMINS SMI <200/250% FPL + MiniStudies1 |
# Needing Public MH Services (Blueprints) |
Number of Medicaid Eligibles |
WA State Population |
| Chelan-Douglas | 1.70% |
1.70% |
2.00% |
1.60% |
2.10% | 1.70% |
| Clark County | 5.90% |
5.70% |
5.80% |
5.70% |
6.10% | 5.90% |
| Grays Harbor | 1.20% |
1.30% |
1.30% |
1.30% |
1.70% | 1.10% |
| Greater Columbia | 10.00% |
11.00% |
11.40% |
11.00% |
14.90% | 10.20% |
| King County | 29.70% |
28.20% |
26.60% |
28.20% |
21.20% | 29.50% |
| North Central | 2.30% |
2.40% |
2.80% |
2.60% |
3.90% | 2.20% |
| North Sound | 15.80% |
15.70% |
15.10% |
15.60% |
13.90% | 16.30% |
| Northeast | 1.20% |
1.30% |
1.30% |
1.30% |
1.80% | 1.20% |
| Peninsula | 5.30% |
5.40% |
5.30% |
5.40% |
4.70% | 5.50% |
| Pierce County | 12.20% |
12.50% |
12.50% |
12.40% |
12.20% | 11.90% |
| Southwest | 1.60% |
1.70% |
1.70% |
1.70% |
2.10% | 1.60% |
| Spokane County | 7.50% |
7.40% |
8.40% |
7.30% |
8.90% | 7.10% |
| Thurston-Mason | 4.20% |
4.20% |
4.20% |
4.20% |
4.00% | 4.40% |
| Timberlands | 1.50% |
1.70% |
1.70% |
1.70% |
2.20% | 1.60% |
| Other/Unknown | 0.00% |
0.00% |
0.00% |
0.00% |
0.40% | 0.00% |
| Total | 295,884 |
331,617 |
148,732 |
133,406 |
829,508 | 5,894,121 |
The next two columns of Table ES.2 compare these two studies by the percent of persons who meet FPL criteria for Medicaid (or very similar criteria) and who are SMI/SED. These percentages and the actual counts on which they are based serve as a proxy estimate of persons needing public mental health services. This study estimates 148,732 persons in this category, compared to 133,406 in Blueprints. The next column indicates the distribution of the 829,508 Medicaid Eligibles across the RSNs. The last column indicates each RSN’s percentage of the state population (2000 Census data).
Most RSN shares of total SMI populations closely track their shares of the state population. It was noted above that among households at or below 200% FPL, King County and North Sound had disproportionately low shares, while North Central and Greater Columbia had disproportionately high shares. These disproportions were attributed to regional differences in employment and income levels. The addition of estimates from the target groups dampened but did not eliminate these disproportionate shares. By disproportionate, we mean that they deviate from population proportions, not that they are suspect.
SMI Estimates and Medicaid Eligibility
In looking at the relationship between the number of Medicaid Eligibles and various prevalence estimates, the following conclusions can be drawn:
- The addition of the target groups, while increasing the overall number of individuals estimated to have SMI or SED, results in very little change to the relationship between the proportion of individuals with SMI/SED and the proportion of Medicaid Eligibles.
- For most RSNs, the proportion of individuals with SMI/SED and the proportion of Medicaid Eligibles show a very close association. The exceptions are King RSN, Greater Columbia RSN, and to a lesser extent North Central RSN and North Sound RSN.
The difference in the shape of the distributions can also be represented in terms of the ratio of Medicaid Eligibles to SMI in each region. These ratios are presented in Table ES.3.
Table ES.3
Ratios of Medicaid-Eligible Persons to Estimates of Persons with SMI, by RSN
| RSN | All SMI |
SMI <200/250% FPL |
Medicaid Eligibles |
Elig:SMI |
Elig:SMI <200/250% FPL |
| Chelan-Douglas | 4,978 |
2,902 |
17,282 |
3.5 |
6 |
| Clark County | 17,489 |
8,613 |
50,556 |
2.9 |
5.9 |
| Grays Harbor | 3,411 |
1,906 |
13,885 |
4.1 |
7.3 |
| Greater Columbia | 29,544 |
16,945 |
123,341 |
4.2 |
7.3 |
| King County | 88,144 |
39,477 |
176,077 |
2 |
4.5 |
| North Central | 6,731 |
4,158 |
32,372 |
4.8 |
7.8 |
| North Sound | 46,759 |
22,376 |
115,091 |
2.5 |
5.1 |
| Northeast | 3,436 |
1,959 |
14,867 |
4.3 |
7.6 |
| Peninsula | 15,616 |
7,935 |
38,741 |
2.5 |
4.9 |
| Pierce County | 36,255 |
18,628 |
101,139 |
2.8 |
5.4 |
| Southwest | 4,715 |
2,519 |
17,599 |
3.7 |
7 |
| Spokane County | 22,288 |
12,425 |
73,500 |
3.3 |
5.9 |
| Thurston-Mason | 12,518 |
6,262 |
33,396 |
2.7 |
5.3 |
| Timberlands | 4,493 |
2,485 |
18,132 |
4 |
7.3 |
| Total | 295,884 |
148,732 |
829,508 |
2.8 |
5.6 |
Table ES.3 shows the considerable variation in ratios of Medicaid Eligibles to SMI, ranging from a low of 2.0 in King County to a high of 4.8 in North Central RSN. For example, in King County RSN there are two Medicaid Eligibles for every SMI/SED person in the general population. There are 4.5 Medicaid Eligibles in King County RSN for every person with SMI or SED that met the FPL criteria for the study (200% for adults, 250% for children). These variations help describe the shifts in percentages between King County and Greater Columbia that were demonstrated in Table ES.2.
The percentages or shares of the total in Table ES.2 provide information about the relative proportion of the population estimated to be in each RSN. The ratios in Table ES.3 provide additional information about the relationship between Medicaid eligibility and prevalence of SMI/SED. Closer examination of ratios sheds additional light by showing that the relationship between Medicaid eligibility and serious mental illness is more complex than just the share-of-total issue.
Some RSNs, King County primarily and to a lesser extent North Sound, have lower ratios of Eligibles to SMI than do other RSNs. This could be interpreted to mean that Medicaid eligibility is not a good proxy for these regions as it may underestimate the prevalence of SMI/SED in these regions. Visual inspection of these data suggests a linear relationship between size (in terms of population) and these Medicaid-to-SMI ratios.
Similarly, the rural RSNs, consisting primarily of counties known to have lower median incomes and more poverty, tend to have higher ratios. This does not necessarily mean they have fewer persons with SMI/SED, but due to economic issues in the region, they may have more persons who are eligible for Medicaid. It may be that they simply have higher proportions that are eligible for Medicaid for economic reasons rather than being eligible due to disability. It might prove fruitful to look closer at the subtypes of Medicaid eligibility to determine whether threshold criteria reflecting disability as well as economic status might more closely reflect the rates of SMI/SED found in this and other prevalence studies.
Conclusions and Recommendations
The results of this study provide reasonable estimates of SMI and SED in Washington State and address the relationship of these estimates to the number of Medicaid Eligibles in the state. These were the primary purposes of the study.
In a recent article by David Mechanic (2003) on the use of prevalence estimates as a measure of need for services, parity, and the expert management of mental health benefits, he states, “it is an illusion to believe that we can avoid muddling through to some extent. The hope is that we can do so thoughtfully.” This serves as a good summary of the efforts of the current study over the last two years. We did muddle through—there was little to guide us. However, we did so thoughtfully. PAC, the Expert Panel, and project staff grappled with the issues, debated perspectives at every step, and sought solutions within our budget. The estimates generated, while not perfect, represent significant progress. All participants learned much from participation in this study. The following recommendations are offered to guide future efforts.
1. Conservative, transparent and defensible prevalence estimates are critical for studies that use complex estimation methodologies and when the results may be used in policy, planning, and funding decisions. This yardstick is recommended for future efforts to estimate prevalence in Washington State.
2. Studies in which results might be contentious or challenged should engage a stakeholder group and provide real opportunity for input. The active participation of PAC in this study was invaluable in guiding the process. Much was learned and a common conceptualization of the issues emerged, which informed the resulting product. We would urge participation by stakeholders at all levels in future studies.
3. When key data are going to be used in policy and resource allocation decisions, regenerating estimates every two or three years is advisable, especially when methods depend upon shifting demographic data, such as economic indicators. New methods, federally funded studies, and routine data collection activities are evolving rapidly and are quickly disseminated. Revisiting studies periodically can capitalize on these enhancements. This can be done cost-effectively if the focus is maintained on easily accessible aggregate data from unbiased sources such as the Office of Financial Management, U.S. Census Bureau, and a variety of other Federal, state, and local data repositories.
Revisiting the topic regularly will continue to contribute to the sophistication and understanding of all stakeholders. The use of consistent methods over time can provide comparison data and opportunities to continually refine estimates. Because capitation is a critical component of virtually all managed care, understanding precisely how we define and count people needing services will remain vitally important.
4. The results of this study suggest that Medicaid eligibility in and of itself is an adequate proxy estimator of the number of persons with SMI/SED for most RSNs, but not all. For this reason it is not an ideal proxy, and in some regions the use of Medicaid eligibility may underestimate the number in need of services. Medicaid eligibility does have a strong relationship with the prevalence of SMI/SED but should not be used exclusively to estimate prevalence or to guide decisions about the funding and administration of mental health programs. It might be that some subtypes of Medicaid eligibility, such as those that reflect disability criteria as well as economic criteria, may prove a better proxy measure of SMI than does the broader category of Medicaid eligibility.
Data that are going to be used to guide public mental heath administration, policy, and funding should be thoroughly understood. Examination of the Medicaid Eligibles numbers should be subjected to similar scrutiny if they are to be used in this context.
Composite indicators are often preferable to indicators taken singly, when the issues are complex and there are competing interests and interpretations. The Consumer Price Index, the Dow Jones Industrial Average, and the system for rating the efficiency of National Football League quarterbacks are examples. Emphasis on a single count or statistic can be misleading and may not take all relevant factors into account.
5. The current estimation models are based on the original Washington State Needs Assessment Household Survey (WANAHS), conducted in 1993-1994 on approximately 7000 households. The empirical relationships found in that survey may still hold, but that is an empirical question. The Substance Abuse and Mental Health Services Administration (SAMHSA) has funded the more current National Co-morbidity Survey II (NCS-II), and the Western Interstate Commission on Higher Education has developed similar prevalence estimates for a number of states. Because surveys are very expensive, “piggy-backing” on existing or new efforts can lead to improvements in estimation models without bearing the cost of re-surveying. Another option is to combine and coordinate surveys being conducted by state agencies for various purposes.
6. The race-neutral approach used in this study satisfied some of the concerns about epidemiological research methods and cultural bias, but not all. The methods used here are consistent with current literature and as such are defensible. However, to assume that the race-neutral methods employed satisfy all concerns or answer all questions about this very important aspect of epidemiological research would be a mistake. Further studies are needed to address the unique needs and issues in estimating prevalence for racial and ethnic groups.
7. With regard to the study of the prevalence of SED in children, the new federally funded NCS-II study is near release. A hybrid approach, taking the best of newly-released efforts and combining these with the best attributes of local studies, like this one and Blueprints, could lead to significant improvements in estimating prevalence of SED in children.
Equally important, more careful consideration and more clarity are needed in discerning the subset of children who are dependent upon publicly funded systems for mental health care.
8. Confidence intervals need to be calculated for the estimates derived in this study. Although methodologically challenging and costly, these parameters would permit assessment of the statistical significance of the differences observed between RSNs and the precision of these estimates. Confidence intervals have been provided in the large, well-funded national prevalence surveys as well as in the previous PEMINS study. The current study has been criticized for not including confidence intervals to date.
Adobe Acrobat Reader is required to view PDF files and you can download it for free. If you do not have an application that can display Microsoft Word documents (.doc), you can download a free Word Viewer.
For more ways to get in touch with the Department of Mental Health Services, go to the DSHS Contact Information Web page. Mental Health Related Questions Contact:
