Psychiatric Emergency Service Use After Implementation of Managed ...

cla6.qxd

5/19/2005

9:20 AM

Page 691

Psychiatric Emergency Service Use After Implementation of Managed Care in a Public Mental Health System Cynthia A. Claassen, Ph.D. T. Michael Kashner, Ph.D., J.D. Saundra K. Gilfillan, M.D. Gregory L. Larkin, M.D., M.S. A. John Rush, M.D.

Objective: This study examined whether implementation of managed care in a public mental health system affected return visits to psychiatric emergency services within 180 days of an index visit. Methods: Data were taken from an administrative database of 75,815 patient visits made to a hospital-based psychiatric emergency service for mental health care between January 1, 1995, and December 31, 2002. Rates of return visits for patients whose index visit occurred at least 26 weeks before a system of managed care was implemented in 1999 were compared with rates for patients whose index visit occurred after the implementation but at least 26 weeks before the data collection period ended. Declining-effects modeling was used to adjust for patients’ gender, ethnicity, age, and admission status. Results: A total of 37,371 patients met study criteria for inclusion: 21,135 before managed care was implemented and 16,236 after managed care was implemented. In the pre–managed care group, 3,687 patients (17 percent) made a repeat visit within 26 weeks of their index visit; 2,369 patients (15 percent) in the post–managed care group made such a repeat visit. For any given index visit to the psychiatric emergency department, patients who presented for treatment after managed care were only 90 percent as likely as patients who presented before managed care to have a return visit within the first five weeks after the index visit. However, there was essentially no difference between groups in the likelihood of a return visit by week 26 after the index visit, suggesting that managed care delayed, but did not eliminate, return visits. In addition, the number of police-accompanied index visits continued to rise after managed care was implemented (from 32.0 to 52.6 percent of all index visits), suggesting that increasing numbers of patients with mental illness in need of treatment were coming to the attention of law enforcement officials after managed care was implemented. Conclusions: Managed care strategies are often used to reduce reliance on emergency services. In this study, managed care delayed, rather than prevented, return visits to the psychiatric emergency service. (Psychiatric Services 56:691–698, 2005)

Dr. Claassen, Dr. Kashner, Dr. Gilfillan, and Dr. Rush are affiliated with the department of psychiatry and Dr. Larkin is with the department of surgery at the University of Texas Southwestern Medical Center at Dallas. Dr. Gilfillan is also with psychiatric emergency services at Parkland Health and Hospital System in Dallas. Send correspondence to Dr. Claassen at 5323 Harry Hines Boulevard, Dallas, Texas 75390 (e-mail, cindy.claassen@ utsouthwestern.edu)

PSYCHIATRIC SERVICES

♦ http://ps.psychiatryonline.org ♦ June 2005 Vol. 56 No. 6

B

y 2003 public mental health services were being administered under some system of managed care in 27 states (1), and this structure was under consideration in another 21 states. Although the consumer’s experience in seeking and receiving treatment is assumed to have changed under these rapidly emerging new structures (2), assessing clinical outcomes continues to be a challenge (3–6). Since deinstitutionalization the use of psychiatric emergency services has been studied as one way of assessing outcome in communitybased systems of care (7–10). As in the past, examination of emergency service revisit patterns may provide information about the equity of, and access to, care provided under evolving managed behavioral health treatment systems. Even though psychiatric emergency services were originally conceptualized as an essential resource for persons with mental illness who reside in the community (11), frequent and repeated use of such services in lieu of alternative care has historically been viewed as undesirable (12–14). For example, a “bounceback” visit to the emergency department—that is, a visit occurring within days of inpatient discharge—is generally viewed as an indicator of inpatient treatment failure (15). If stringent limits placed on inpatient care result in the premature discharge of large numbers of marginally stabilized patients, an increase may be ob691

cla6.qxd

5/19/2005

9:20 AM

Page 692

served in the frequency of “bounceback” visits to the psychiatric emergency service (16). In addition, characterization of “revolving door” patients—that is, frequent users of psychiatric emergency services—has historically helped identify groups of consumers whose needs are not adequately met by outpatient systems of care (16–19). Some fear that diminished treatment dollars associated with managed care structures result in decreased availability of intensive, community-based services for persons who are the sickest and most vulnerable (20–22). If true, because the psychiatric emergency service is the most accessible point of care in the public mental health system, emergency service rosters should reflect increased numbers of repeat visits by patients with severe and persistent mental illness. Finally, reduction in resources after implementation of managed care systems is thought not only to leave vulnerable patients without adequate care but also to promote increasing “transinstitutionalization,” defined as the de facto management of mentally ill patients in the criminal justice system (23–25). If this concern is valid, an increase in both initial and repeat visits to the psychiatric emergency service during which the patient is accompanied by the police might be observable under managed care structures. The introduction of a regional system of managed mental health care in our public mental health catchment area in 1999 provided the opportunity for a natural experiment to test these assumptions. This study compared the rate of return visits to the psychiatric emergency service for patients presenting before and after adoption of the new system. In July 1999 the phase-in of a system of behavioral managed health care, called NorthSTAR, was begun in seven north Texas counties, including Dallas County where the study hospital (Parkland Hospital) is located. NorthSTAR is described by its architects as a blended-funding, integrated, behavioral health carve-out that eliminates the separation of treatment silos for substance abuse and mental health treatment. By De692

cember 1999 it functioned both as a full-risk, capitated, per-member, permonth 1915b Medicaid waiver program and a flat-fee reimbursement program for the medically indigent. These functions are rolled into a private sector–operated open system of care in which consumers are given a choice of providers (26). About 30 percent of the funding and patient base for the program came from Medicaid. At implementation, the NorthSTAR region ranked 35th out of 40 in Texas for mental health expenditures, and Texas ranked 42nd out of 50 states in per capita expenditures for mental health (27). Despite the underresourced system into which it was introduced, NorthSTAR had as a primary goal improved access for all covered patients—for example, it sought to eliminate long wait times before treatment enrollment. NorthSTAR divided local functions into organizations designated as treatment “providers” and one “authority” appointed to oversee how care was being provided. Existing care structures were absorbed into new provider networks, and a handful of new providers emerged. Pre– and post–managed care systems used essentially the same menu of services, including medication and case management; psychotherapy; assertive community treatment for individuals with severe illness, a history of multiple hospitalizations, or both; and supported housing and employment. However, the number of consumers who received assertive community treatment services increased from 79 to 450 in the first 18 months after NorthSTAR was implemented. According to NorthSTAR creators, between July 1999 and December 2002 the major change in provision of care associated with NorthSTAR implementation was more of a “culture shift”—a change in approach to care delivery—than a change in the quantity or type of care. The new system included prospective and concurrent clinical review and fixed rates of reimbursement, but eligible “priority populations” remained unchanged. These included patients in crisis with a Global Assessment of Functioning score of 50 or less; a primary diagnoPSYCHIATRIC SERVICES

sis of major depressive disorder, bipolar disorder, or schizophrenia was necessary to obtain ongoing services. The total population of consumers served increased from 6,000 to 8,000 per month in 1998 to more than 13,000 per month in 2001 (28), while the wait for outpatient service dropped from weeks to a mandated maximum of 96 hours after initial telephone contact. No change in policies regarding the use of mental health emergency services was apparent during the implementation of NorthSTAR, and examination of hospitalization and aftercare plans in the first 42 months under the new system revealed no major shifts in aftercare referral patterns for the psychiatric emergency service. In March 2001 (20 months after managed care implementation) NorthSTAR designated three local “front door” crisis sites that provided 23hour observation before hospital admission. After this change potential inpatients from the study hospital’s psychiatric emergency department were diverted to these sites before or instead of inpatient admission. For the nine months after this change, the study hospital’s psychiatric emergency census increased by an average of 33 patient visits per month. By using a hospital-based database of all patient visits made to Parkland Hospital’s psychiatric emergency service between January 1, 1995, and December 31, 2002, this descriptive study examined bounceback and revolving-door visits as well as visits initiated by the police. We sought to determine the frequency of each type of visit before and after NorthSTAR was implemented. In addition, we attempted to characterize subgroups of patients in the psychiatric emergency service who appeared to adapt less readily to the new system of care.

Methods Parkland Hospital’s psychiatric emergency service averaged 800 to 900 patient visits per month throughout the study period and served as one gateway into mental health services in the public sector. Other than changes in procedures related to the structure of public mental health care, there were no detectable major changes in men-

♦ http://ps.psychiatryonline.org ♦ June 2005 Vol. 56 No. 6

cla6.qxd

5/19/2005

9:20 AM

Page 693

tal health care policy, population distribution, substance abuse patterns, or legal policies for offenders with mental illness in Dallas County during the study period. Patterns of psychiatric emergency department visits were therefore assumed to be one reasonably sensitive indicator of the impact of NorthSTAR on help-seeking behavior among persons with mental illness receiving care in the public sector. This study was approved by the institutional review board of the University of Texas Southwestern Medical Center. Data source and major study variables The administrative database from which emergency service information was extracted for these analyses was created by medical record review and included both demographic characteristics (age, gender, and ethnicity) and information about the visit—for example, whether the visit was voluntary or whether the patient had ever been seen in the hospital system before the index visit. To study NorthSTAR’s impact on psychiatric emergency service revisit rates, two patient cohorts were established within the database. The pre–managed care cohort included patient visits between January 1, 1995, and June 30, 1999; the post–managed care cohort included patient visits between July 1, 1999, and December 31, 2002. For this study we assumed that a return visit within 26 weeks of the index visit constituted a repeat visit within the same illness episode and was thus related to the nature of the treatment received after the index emergency visit. In contrast, any visit by the same patient more than 26 weeks after an index visit was regarded as a second index visit—that is, we assumed that this second index visit was associated with a new episode of illness. Therefore, the same patient could have more than one index visit during each study period. The study sought to characterize patients’ help-seeking patterns during any given illness episode before and after managed care was implemented. We stopped identifying new index visits 26 weeks before the end PSYCHIATRIC SERVICES

of each data enrollment period. To avoid overpopulating the post–managed care group with patients who had a previous use history, the post–managed care working sample included patients who reentered the emergency department on or after July 1, 1999, with no psychiatric emergency service use for the previous 26 weeks. A declining-effects model (29–31, unpublished manuscript, Trivedi MH, Rush AJ, Crismon ML, 2004) was adapted to measure the impact of managed care on return visit rates. This strategy enabled us to test for the ability of the managed care system to reduce repeat psychiatric emergency visits within one week of the emergency index visit. This model also includes a measurement of growth that was used here to examine whether the initial effect of managed care on emergency revisit rates increased, remained constant, or decreased as time passed over the 26 weeks after the index visit. Unlike survival analyses (32,33), which would focus on the first return visit, declining-effects modeling takes into account all return visits and allows for patients who may have more than one index visit. In this within-patient nesting modeling, patients serve as their own controls. Therefore, the design has the ability to control for a wide range of information related to the demographic and visit characteristics that are potential covariates while analyzing the impact of managed care on return visits to the psychiatric emergency service. Mathematically, the Bernoulli hierarchical regression is 1n

() pt 1–pt

=β00+β01Iij+β10(t–1)+β11(t–1)Iij+

β20(t–1)2+β21(t–1)2Iij+β20(xij–xR)+uij

In the equation p is the probability that patient i will return on week t following the jth index visit. Iij=1 if the index visit occurred after managed care was implemented, and Iij=0 if the visit occurred before it was implemented. The symbol ui is a patientlevel variate. The vector xij represents patient characteristics, with mean values xR. When adjusted for mean characteristics (xR), the model allows investigators to estimate the expected proba-

♦ http://ps.psychiatryonline.org ♦ June 2005 Vol. 56 No. 6

bility that an average patient will have a return visit during the first week after the index visit either before (exp[β00]/[1+exp(β00)]) or after (exp [β00+ β01]/[1+exp (β00+β01)]) managed care was implemented. Expressed as an odds ratio, the difference in these adjusted rates measures the probability that managed care will affect rates of repeat visits during the first week after an index visit (exp[β01]). Rates of return visits are expected to decline (rate of exp[β10]) as time passes over the course of 26 weeks, both before and after managed care at a rate of (exp[β10+ β11]). The difference between the growth rates for pre– and post–managed care is the growth effect (exp[β11]), which accelerates or decelerates with time (rate of exp[β21]). The estimated parameters were used to compute an adjusted cumulative probability that the average patient would have had at least one revisit by week t: P(t)=1–[(1–p1)(1–p2)(1–p3) . . . (1–pt)]

With this model, multiple return visits for each patient over the 26 weeks after the index visit will lead to larger values of a cumulative probability.

Results Between January 1, 1995, and December 31, 2002, a total of 44,851 patients visited Parkland’s psychiatric emergency service, generating 75,815 separate visits. Among these patients, 11,669 (26.02 percent) made at least one return visit within 26 weeks of a previous visit, accounting for 30,964 (40.8 percent) of the 75,815 visits. The interval between the index and the first return visit ranged from one day to 2,882 days. Of all study-defined repeat visits, 13.7 percent, 20.8 percent, 30.5 percent, and 63.0 percent were made within 1, 4, 8, and 26 weeks, respectively. Using study-defined criteria for an index visit and requiring information on age to be present, we created a final analytic sample of 37,371 patients (83.3 percent of all patients): 21,135 in the pre–managed care group and 16,236 in the post–managed care group. Among the 37,371 patients, 693

cla6.qxd

5/19/2005

9:20 AM

Page 694

Table 1

Characteristics of persons who visited a psychiatric emergency service, by managed care status of the emergency department during the first index visita Total sample (N=37,371)

Ethnicity Caucasian African American Latino Other Male Age (years) Younger than 15 15 to 25 26 to 35 36 to 45 46 to 55 56 to 65 Older than 65 New hospital patient Escorted by police Presenting complaint Psychosis Suicidality or mood complaint Substance-related complaint Other a

Post–managed care (N=16,236)

Pre–managed care (N=21,135)

N

%

N

%

N

%

20,409 10,684 5,674 604 18,910

54.6 28.6 15.2 1.6 50.6

8,449 4,738 2,782 267 8,484

52.0 29.2 17.1 1.6 52.3

11,960 5,946 2,892 377 10,426

56.6 28.1 13.7 1.9 49.3

1,699 8,856 10,636 9,496 4,405 1,549 730 19,819 15,309

4.5 23.7 28.5 25.4 11.8 4.1 2.0 53.0 41.0

909 3,986 4,286 4,076 2,035 722 222 9,322 8,538

5.6 24.6 26.4 25.1 12.5 4.4 1.4 57.4 52.6

790 4,870 6,350 5,420 2,370 827 508 10,497 6,771

3.7 23.0 30.0 25.6 11.2 3.9 2.4 49.7 32.0

8,818

23.6

3,765

23.2

5,053

23.9

4,113

11.0

1,767

10.9

2,346

11.1

6,279 17,904

16.8 47.9

2,253 7,960

13.9 49.0

4,026 9,944

19.0 47.0

χ2†

df

p

110

3

.001

31.3

1