American Journal of Community Psychology, VoL 8, No. 2, 1980

Evaluating the Impact of Community Mental Health Centers on Hospital Admissions: An Interrupted Time-Series Analysis I James L. Spearly 2 University of Texas at Austin

This article presents the results of a n evaluation o f the impact o f center services on first admissions and readmissions to state hospitals in a large Southwestern state using an interrupted time-series design with switching replications and sequential treatments. Of the five centers investigated, only one appeared to be effective in reducing both first admissions and readmissions from its catchment area," another center appeared to reduce first admissions from the area to which it expanded its services," and the initiation o f one center coincided with an increase in readmissions to state hospitals. As a culmination of the efforts and recommendations of the Joint Commission on Mental Health and Mental Illness (1961), the community mental health centers program was initiated to enable most mentally ill persons to be treated in their own communities. One of the new program's major objectives was to decrease the use of state mental hospitals. In over a decade of operation, community mental health centers have increased the availability of communitybased mental health services, but they have not been entirely successful in either reducing unnecessary admissions to state hospitals or in providing services to persons released from such hospitals. This is the verdict of the U.S. General Accounting Office (Note 1) in its most recent report to the Congress.

The author is indebted to Jack L. Franklin, PhD, Chief of Program Analysis and Statistical Research, Texas Department of Mental Health and Mental Retardation, who provided access to the data base for this study. The author also wishes to thank Mark Mason for his assistance with the computer analyses and Brian L. Wilcox for his helpful comments on earlier versions of this paper. 2All correspondence should be sent to James L. Spearly, Department of Psychology, University of Texas at Austin, Austin, Texas 78712. 229 0091-0562/80/0400-0243 $03.00/0 © 1980 Plenum Publishing Corporation

230

Speafly

State hospital populations have been reduced by some two-thirds since 1963, but there is little evidence that the centers program played a major role in the reduced utilization (Bassuk & Gerson, 1978; Scully & Windle, 1976). A substantial reduction in their populations has resulted from hospitals following a policy of early releases in response to economic pressures at the state and local levels, the tightening of their accreditation standards, and limitations on the use of involuntary commitment resulting from a series of recent federal court decisions. The GAO considers the development and use of psychiatric drugs and the expansion of federal assistance programs to have had a more dkect impact on state hospital utilization than the presence of community mental health centers. This early release policy combined with the failure to provide appropriate community-based alternatives to hospitalization has led to several unpleasant results. There has been criticism from several quarters concerning the large numbers of former patients found in the community without adequate services (Bassuk & Gerson, 1978; "The Night of the 'Deinstitutionalized'," 1978; Schumach, 1974). In some states a system of ungoverned shelter care has grown in response to the expanded federal assistance available; inadequate living and treatment conditions are common in many of these entrepreneurial efforts (Segal & Aviram, 1976). As a consequence of the lack of sufficient community supports, readmission rates for state hospitals have increased both in number and in proportion of the total admissions they represent (GAO, Note 1). The report of the recent President's Commission on Mental Health (1978) underscores these trends and relies upon the various investigations presented here in making its recommendations for future policy. The GAO draws support for its conclusion that the centers program has had little impact on the pattern of hospital use from large.scale studies conducted by the National Institute of Mental Health comparing hospital admissions and resident rates in areas with centers and areas without them (Scully & • Windle, 1976; Windle, Bass, & Taube, 1974). Data from these studies indicate that the community mental health centers program has helped to reduce admissions to state hospitals but has generally not had a significant impact on helping persons to return to the community from them. As reasons for this small impact on hospital rates, the GAO cites a lack of coordination between centers and hospitals and the use of federal resources to support former patients in inappropriate private shelter care and nursing homes. In addition, they note that center staff give low priority to the reduction of state hospital populations, while on the other hand, they have attracted a new type of patient who is not very ill and is not a candidate for hospitalization in a state institution. It has been demonstrated that former and potential state hospital patients can be maintained in the community for considerable periods of time, precluding the need for any or further hospitalization (Delaney, Seidman, & Willis, 1978; Fairweather, Sanders, Cressler, & Maynard, 1969), but it requires intensive and

Time-SeriesEvaluation of Community Mental Health Centers

231

aggressive community service efforts to do so (Davis, Dinitz, & Passamanick, 1972; 1974). The objective of the present study was to verify or contradict the implications of the GAO report, at least within the context of a large Southwestern state. In contrast to the studies relied upon in that report, more recent data were analyzed using the most powerful of the quasi-experimental designs for drawing valid inferences from retrospective archival data, an interrupted time-series design with sequential interventions and switching replications (Cook & Campbell, 1976; Glass, Wilson, & Gottman, 1975).

PREVIOUS INVESTIGATIONS An initial nationwide investigation revealed that the use of state mental hospitals by catchment area populations was lower than for the U.S. as a whole (Windle et al., 1974). But, as the authors point out, this cross-sectional comparison leaves open the question of whether these areas experienced different utilization rates prior to the initiation of the centers program. A longitudinal investigation using records from 16 states compared hospital admissions and resident rates of those counties wholly within to those counties wholly outside of catchment areas (Scully & Windle, 1976). No large consistent relationship was observed between the initiation of the centers and changes in the inpatient rate. There appeared to be a tendency, however, for counties with centers to decrease more (or increase less) in state hospital admissions rates. The authors concluded that centers have had some impact on the flow of persons from the community to the hospital but not on the flow of hospital patients to the community. It is possible, however, that events other than the initiation of center services affected the counties within catchment areas and not the counties outside catchment areas to produce the differential admissions patterns observed. For example, Windle, Goldsmith, Schambaugh, and Rosen (1975) present evidence of demographic differences between areas with and without federal community mental health center grants. Urban-rural differences alone could account for differential changes in services and policies across the different types of areas contributing to the study's findings. And, since relatively more areas that have centers also have higher percentages of families in poverty, it is highly plausible that the demands for mental health services have expanded at a different rate in center areas than they have in noncenter areas. Thus comparisons of the extent of change across two periods of time separated by a number of years leaves open to question whether a difference in the extent of change for center versus noncenter areas was due to an intervention in one type of area or to differential rates of expansion in demands for services.

232

Speatly

Doidge and Rogers (1976) capitalized on their state's mental health system being less complex in investigating the impact of federally funded centers on state hospital admissions in Wyoming. Counties which experienced the initiation of comprehensive service community mental health centers between 1969 and 1972 showed a decrease in total admissions to Wyoming's only state hospital across that span of time, while counties not experiencing a center start showed an increase in admissions across the same period. This study presents stronger findings of the impact of federally funded centers on state hospital admissions. But since counties were involved in selecting whether or not to have a mental health center, those factors that contribute to differential self-selection could also contribute to differential rates of growth in demand for mental health services. By observing only two points in time, 1969 and 1972, not enough evidence is available to determine if such differential demand maturation patterns existed prior to initiation of center services in certain areas. McNees, Hannah, Schnelle, and Bratton (1977) used single time-series designs for each of three counties in Tennessee which had initiated aftercare programs. Monthly readmission rates were observed for several years preceding the aftercare initiations and for several years following. No clear reductions in recidivism were noticed either as discontinuities coinciding with the program initiation date in the graphs of monthly data nor in a special time-series analysis. McNees et al. observed a similar series of data on first admissions, presumably not affected by the aftercare program, to see if any change in hospital admissions policies coincided with the initiation of their programs. No discontinuity in the first admissions series coincident with the interventions was observed. In addition, they examined a similar series of data on hospital discharges to see if a substantial increase in discharges just prior to the interventions could have masked any impact the programs may have had on readmissions, and none was observed. But as the authors point out, their failure to demonstrate an impact on readmissions could be due to the very small percentage of discharged patients actually served by their programs relative to the total discharge pool. Delaney et al. (1978)present an assessment ofa crisis interventionprogram using an interrupted time-series design with an untreated control group. Their program, aimed at avoiding state hospitalization by marshaling the resources and referrals needed by the individual and/or family, had as one of its primary functions a reduction in the number of state hospital admissions. Time-series were constructed for the number of quarterly admissions to the state hospital from the targeted and control zones for 2 years prior to and 2 years following the initiation of the crisis intervention program. Discontinuities in the graphed series coinciding with and continuing after the intervention were verified to be statistically significant. This study provides particularly strong evidence of the effectiveness of the crisis intervention program in reducing hospital admissions. Observation of the preintervention time trends controls for any differences in expansion of demand for services. And the use of quarterly data limits other events as plausible

Time-Series Evaluation of Community Mental Health Centers

233

explanations to the quarter in which the intervention takes place. The major limitation of this study involves the restriction of the intervention process to only one area thus limiting the generalizability of the results to other times and settings. Scully and Windle (1976) note that the reduction of state hospital populations is a gross program goal that requires increasing refinement. The first step would appear to be an investigation of centers' impact on first admissions and readmissions separately. A reduction in first admissions would indicate an improvement in eliminating the need for state hospitalization entirely for certain persons. Indeed, the fact that centers are serving a younger, less ill clientele does not preclude an effective early-intervention strategy wherein persons are being treated prior to any progressive deterioration to the point of requiring hospitalization. On the other hand, a reduction in readmissions would indicate improved community-based aftercare services for persons discharged from state hospitals permitting them to remain in their community, at least for longer periods of time. Previous studies have examined total admissions which confounds the two types of treatment processes. Inconsistent and inconclusive results could be the product of an unbalanced emphasis on a prevention versus aftercare strategy. In addition, the focus on resident rates to obtain information on the effectiveness of follow-up efforts is confounded with a variety of factors under the hospitals' control, e.g., the length of stay. METHOD

Design Cook and Campbell (1976) present the following definition, "Interrupted time-series designs involve repeated measurements of an effect both before and after a treatment is abruptly introduced and 'interrupts' previous data patterns" (p. 274). The causal inference in these designs is based on discovering an abrupt change in the values of the dependent variable coinciding with the introduction of the treatment. Interrupted time-series designs are particularly appropriate for unobtrusive measures such as institutional records where subjects are not responding to multiple testings. The major strengths of a time-series design include an assessment of a pretest time trend which permits a check on the plausibility of a deviantly high or low pretest score influencing the results of a simpler design, and an assessment of a posttest time trend which permits a determination of the persistence of an impact over time. The present analysis expands upon the untreated control group timeseries design used in Delaney et al. (1978) in order to add the potential for generalizability of an observed impact of centers operations on state hospital first admissions and readmissions. The design involves sequential interventions

234

Spearly

and switching replications rather than an untreated control group. Glass et al. (1975) label this a "Multiple Group Single Intervention Design" (p. 23) with "time-lagged control" (p. 25). In this design, there are several nonequivalent samples, each of which receives the intervention (a community mental health center start) at different times, such that when one area receives the treatment another serves as a control, and when that control area later receives the treatment, a previous treatment group serves as the control. The power of this design derives from its control of most of the threats to internal validity. It also reduces the plausibility of the areas having systematically different population characteristics since all areas involved are center areas. Cook and Campbell (1976) consider the present design to be the most powerful of the quasi-experimental designs for drawing valid implications. The potential for generalization derives from the effect being demonstrated with several different populations, settings, and moments in history and with different sets of irrelevancies associated with the application of each treatment. A bonus in this design for testing service innovations involves the need only to delay treatment for the controls rather than denying them the boon altogether. The investigation presented below involved various components of a large state mental health system located in the Southwest. Characteristics of the mental health service population in this state reflect national trends. The number of residents in state hospitals at any given time has sharply declined over the last decade. During this period, total admissions have remained fairly constant. However, the total admissions pool has been composed increasingly of persons multiply admitted. Indeed, readmissions have increased substantially in the last 6 years in both number and proportion of total admissions across all hospitals, geographic units, and specialty units. The intervention investigated in the present study involved federally funded community mental health centers providing the full complement of legally mandated services. Catchment areas were selected for investigation such that sufficient preintervention data points were available, no major reorganization of services occurred (with one exception to be discussed), and no hospital-sponsored outreach centers were serving the area (to eliminate the confounding effects of the impact of such programs on admissions). For those five catchment areas (denoted A, B, C, D, and E) which fit these conditions, first admissions and readmissions data were compiled over 9 years (September 1967 through September 1976) and in monthly intervals (representing all mental health first admissions and readmissions to state hospitals from these respective areas). Each center area was served by a different state hospital. The centers in these areas also differed as to their dates of initiation of services which were verified by both agency records and telephone consultations with the center directors. In addition to the admissions time-series, monthly data on hospital discharges to each area were compiled in order to observe any substantial increases (or decreases) just prior to the center intervention that may have influenced the observed impact on readmissions.

Time-Sedes Evaluation of Community Mental Health Centers

235

One of the five centers permitted a special contrast. When it first opened (June 1970), it served only two counties, and when it reorganized its service delivery 4 years later (September 1974), it began to serve three additional counties. For convenient notation, Area E-1 data refer to first admissions and readmissions to state hospitals of persons from the original two counties served, and Area E-2 data refer to admissions of persons from the other three counties. This provides a unique switching replication in that the same center was involved in both areas.

Analysis Cook and Campbell (1976) advocate the use of time-series designs even when no statistical test of the hypotheses can be carried out. They suggest that the data be plotted and "eyeballed" (p. 275) for a discontinuity in the time trend that cannot be explained readily in terms of the continuation of the trends that are observable in the preintervention time series. Statistical analysis of timeseries data involves fitting regression lines before and after the intervention date; an intervention effect would then appear as a change in the slope (direction) or intercept (level) of the series coincident with the initiation of the treatment or program. In contrast to an abrupt change in the level of admissions, a change in slope would indicate a change in the rate of increase or decrease in first admissions or readmissions. Since time-series data typically exhibit a serial dependence among the observations, ordinary least squares regression methods should not be employed. Box and Jenkins (1970)have developed an Auto-Regressive Moving Averages (ARIMA) modeling technique to describe and adjust for the serial dependence displayed in complex longitudinal processes (see Glass et al., 1975, for a detailed explanation of the use of this technique for analyzing time-series data). The TMS (Note 2) computer package, based on this ARIMA modeling technique, was used in the present study to provide an inferential statistical analysis of the intervention effects.

RESULTS

Only two of the six center areas (C and E-2) evidenced a significantly abrupt and constant change in the level of first admissions to state hospitals coinciding with the initiation of community mental health center services. The data for Center Area C produced a significant decrease in the level of first admissions, t = -2.77, df = 104, p < .01. Note the discontinuities in the timeseries graphs corresponding to the solid vertical lines in Figure 1. To establish the controls for those results using a switching replications design, the data for

236

Spearly

16~

l__initiation of "center services in area C

012

16r

"

t

I

centerinitiati°n of services in area E-2

+1°1 '7'12

I

t19681

il c. rOI

f19721

11974 1

t19761

Fig. 1. Monthly mental health first admissions to state hospitals by center catchment area. The dashed vertical lines indicate control analyses.

Center Area E-2 were reanalyzed, this time using the number of preinterven. tion points demarcating the opening of the center in Area C (well before the initiation of center services in the E-2 Area). As a control for the Center Area C results, the E-2 Center Area reanalysis produced no significant change in the level of first admissions, t = 0.00. Note in Figure 1 the absence of a discontinuity in the E-2 series coinciding with the opening of the center in Area C and vice versa. (Control analyses are indicated by dashed vertical lines in Figure 1 .) The data for the three counties comprising the E-2 Center Area also produced a significant decrease in the level of first admissions coinciding with the initiation of services to that area, t = -2.14, df = 104, p < .05. Note the discontinuity in the Area E-2 time-series graph corresponding to the solid vertical line in Figure 1. To establish a control for these results using a switching replications design, the data for Center Area C were reanalyzed using the number of preintervention and postintervention points coinciding with the extension of services to the additional three counties comprising the Center Area E-2. As a control for the E-2 Center Area results, the Center Area C reanalysis evidenced no significant level change, t = 0.64, df = 104, ns. Note in Figure 1 the absence of a discontinuity in the C series coinciding with the initiation of services to Area E-2.

Time-Series Evaluation of Community Mental Health Centers

237

An additional comparison was made using the data from the counties initially served by the Center Area E-1. An analysis was performed using the number of pre- and postintervention points coinciding with the extension of services to the additional three counties (rather than to the date of initiation of services in the original two counties, which had previously produced no significant change). As in the other control results, the E-1 Area reanalysis produced no significant change in the level of first admissions from its area coinciding with the extension of services to the other counties, t = -1.26, d f = 104, ns. Note in Figure 1 the absence of a discontinuity in the E-1 series coinciding with the initiation of services to Area E-2. Only two center areas evidenced a significantly abrupt and constant change in the level of readmissions to state hospitals coinciding with the initiation of community mental health center services. The data from Center Area C produced a significant decrease in the level of readmissions, t = -3.25, dr= 104, p < .01; while the data from Center Area A produced a significant increase in the level of readmissions, t = +2.05, d f = 104, p < .05 (see Figure 2). Delayed treatment controls were used instead of switching replications as the start dates for these two centers were too close to each other to permit a proper comparison. The readmissions data from the counties comprising the E-2 Center Area were

11Iin,ti.tiono center services in area C

12 ¢)

161[~

I/ initiation of P , center services

,(12 t

E

16

1196;

119701

11972 !

11974 I

11976,

Fig. 2. Monthly mental health readmissions to state hospitals by center catchment area. The dashed vertical lines indicate control analyses.

238

Spearly

reanalyzed as a control for each test area since those data represented the largest delay in initiation of service relative to Areas A and C (1969 to 1974). First, the number of pre- and postintervention data points coinciding with the start of services in Center Area C were used, and then those points coinciding with the opening of the center in Area A were used. In neither control run was a significant change in level observed, t = - . 3 2 , df = 104, ns.; and t = - . 6 9 , df = 104, ns. (see Figure 2). No significant changes in slope coinciding with the initiation of center services were observed for either first admissions or readmissions in any of the center areas. And a visual inspection of the plotted discharge data series for each of the center areas indicated no evidence of substantial increases or decreases in the discharge pool just prior to center start dates that might have influenced a center's efforts to reduce readmissions.

DISCUSSION The results of this investigation appear to be consistent with the conclu. sions of the GAO (Note 1), verified by a design that provides for increased internal validity. Overall, the centers investigated appear to have had a negligible impact on state hospital admissions. Only one center (C) appears to have been effective in reducing both the level of first admissions and readmissions from its catchment areas. Nevertheless, visual inspection of Area C's readmissions series indicates an increasing trend beginning about 5 years after the center opened. Center E, while seemingly unable to reduce admissions from the two-county area in which it first began delivering services appears to have been effective in reducing the level of first admissions from the three-county area to which it later extended its services. In addition to finding no reductions in the level of admissions in other center areas, no decreases in the rate of increase or decrease of first admissions or readmissions were observed. Although the initiation of Center A coincided with a significant increase in the level of readmissions from its catchment area, it seems less plausible to hypothesize that the center's activities caused the increase. Further evidence, for example, would have to be obtained that indicates that a selective case-find occurred wherein previously hospitalized persons were increasingly brought to their attention and referred to state hospitals, to a degree not occurring prior to the center start and with no similar increase in first admission referrals. The null results for the other centers are also difficult to interpret. Some areas have low total numbers of clients and extremely variable monthly admissions patterns which may hinder the production of statistically significant results. In any event, a visual inspection of the time series for the nonsignificant areas yields no observable changes coinciding with the intervention dates. Even if it were suggested that the centers' operations took considerable time to get underway before they

Time-Series Evaluation of Community Mental Health Centers

239

could demonstrate an effect, the long postintervention periods of data here available reveal no delayed or gradual reductions in admissions. One exception should be noted in that first admissions from Center Area D appear to show a decrease several years after the center start. It is open to speculation whether this is a delayed effect of the program or coincides with some other event. Investigation could include searching for relevant events at that point in time. Some unique pattern of impact other than a change in slope or an abrupt change in level of admissions might be hypothesized in a given case. For example, an initial increase in first admissions due to increased case-find could be followed by a decrease to below the preintervention level. Center Area A's first admissions series displays such a pattern upon visual inspection. TMS (Note 2) has an option for testing a user-specified linear function. A subsequent analysis could be run to test the existence of such a pattern in the data not attributable to the serial dependence. The potential for demonstrating the generalizabflity of the impact of the federally funded center model was not realized. Several explanations for the differential results may be considered. It may be that the intervention is effective with one type of population and not with another. But the comparisons here involve areas that are all receiving center services, so there is not as systematic a difference as between center and noncenter areas. Nevertheless, Center E only demonstrated a significant impact on first admissions from the three-county population to which it expanded its services and not from the original counties served. In order to test the plausibility of such an explanation for these results, social indicators related to mental health problems could be compiled for the various areas to see if they differ in any manner consistent with the differential impacts observed. While this study isolated treatment variables such that other mental health system components (outreach centers) would not confound the results, it could not guarantee that the same services would be delivered in the same manner across the various center areas. Future investigations could include a comparison of various service delivery characteristics to see if any plausibly related differences coincide with the differential impacts observed. Factors such as total personnel hours, extent of temporary alternative inpatient services available, and number of hours of crisis intervention, consultation and education, and/or aftercare services have been linked to the utilization of state hospitals (Anthony, Buell, Sharratt, & Althoff, 1972; Kirk, 1976; Siguel, 1974). Several additional characteristics of the mental health system were investigated in the present study, including: the proximity of the catchment area counties to a state hospital; the percentage of the mental health dischargees from state hospitals to the center areas who were contacted by the center in that area; and the total per capita mental health expenditures for each center. None of these variables was able to distinguish between centers exhibiting an impact and those failing to do so. While data on each center's operations back to the date of

240

Spearly

their initiation were not available for this study, future evaluations should examine these variables throughout the period being investigated. Comparisons between centers with varying impacts could give some indications as to which center services are effective and under what conditions. The findings of the present study yield strong case-by-case validity for the presence or absence of an impact on first admissions and/or readmissions for several community mental health centers located in a large Southwestern state. As such, it represents only one component of a comprehensive evaluation of center services. Centers have many other goals which patient-movement statistics cannot assess. Indeed, the validity of using readmissions as a criterion of successful community-based mental health services has been questioned (Rosenblatt & Mayer, 1974). As the average length of stay decreases, it may be desirable to provide an opportunity for periods of respite for the mentally ill as well as to compensate for the briefer exposure to treatment therapies (Rabiner & Lurie, 1974). But if the GAO is correct, considerable inapporpirate and unnecessary hospitalization is taking place which could be reduced by expanded screening, referral, and crisis intervention and more vigorous aftercare efforts. The report of the President's Commision on Mental Health (1978) calls for a comprehensive and integrated system of mental health services with community mental health centers playing a key role. Specifically, the Commission recommends the development of a national plan for "the continued phasing down and where appropriate closing of large state hospitals" (p. 22). Thus, the effectiveness of centers in reducing first admissions and readmissions to state hospitals will continue to be an important social policy issue. Future investigations should not only continue to assess the impact of centers on hospital populations but also attempt to determine what factors contribute to the reduction of state hospital admissions.

REFERENCE NOTES 1. U.S. General Accounting Office. Returning the mentally disabled to the community: Government needs to do more. Washington, D.C.: U.S. General Accounting Office (Distribution Section, Room 4522, 441 G Street NW), January 1977. 2. Bower, C. P., Padia, W. L., & Glass, G. V. TMS: Two fortran IV programs for analysis of time-series experiments. Boulder: University of Colorado, Laboratory of Educational Research, October 1974.

REFERENCES Anthony, W. A., Buell, G. J., Sharratt, S., & Althoff, M. E. Efficacy of psychiatric rehabilitation. Psychological Bulletin, 1972, 78, 447-456. Bassuk, E. L., & Gerson, S. Deinstitutionalization and mental health services. Scientific American, 1978, 238, 46-53. Box, G., & Jenkins, G. Time-seriesanalysis: Forecasting and control. San Francisco: Holden Day, 1970.

Time-Series Evaluation of Community Mental Health Centers

241

Cook, T. D., & Campbell, D. T. The design and conduct of quasi-experiments and true experiments in field settings. In M. D. Dunnette (Ed.), Handbook of industrial and organizational psychology. Chicago: Rand-MeNaUy, 1976. Davis, A., Dinitz, S., & Pasamanick, B. The prevention of hospitalization in schizophrenia: Five years after an experimental program. American Journal of Orthopsychiatry, 1972, 42, 375-388. Davis, A., Dinitz, S., & Pasamanick, B. Schizophrenics in the new custodial community. Columbus: Ohio State University Press, 1974. Delaney, J., Seidman, E., & Willis, G. Crisis intervention and the prevention of institutionalization: An interrupted time-series analysis. American Journal of Community Psychology, 1978, 6, 33-45. Doidge, J., & Rogers, C. Is NIMH's dream coming true? Wyoming centers reduce state hospital admissions. Community Mental Health Journal, 1976, 12, 399-404. Fairweather, G., Sanders, D., Cressler, D., & Maynard, H. Community life for the mentally ill: An alternative to institutional care. Chicago: Aldine, 1969. Glass, G., Willson, V., & Gottman, J. Design and analysis of time-series experiments. Boulder: Colorado Associated University Press, 1975. Joint Commission on Mental Health and Mental Illness. Action for mental health. New York: Wiley, 1961. Kirk, S. Effectiveness of community services for discharged mental hospital patients. American Journal of Orthopsychiatry, 1976, 46, 646-659. McNees, M., Hannah, J., Schnelle, J., & Bratton, K. The effects of aftercare programs on institutional recidivism. Journal of Community Psychology, 1977, 5, 128-133. The plight of the "deinstitutionalized" mental patient. Science, 1978, 200, 1366. President's Commission on Mental Health. Report to the President from the President's Commission on Mental Health. Washington, D.C.: U.S. Government Printing Office, 1978. Rabiner, C., & Lurie, A. The case for psychiatric hospitalization. American Journal of Psychiatry, 1974, 131, 761-764. Rosenblatt, A., & Mayer, J. The recidivism of mental patients: A review of past studies. American Journal of Orthopsychiatry, 1974, 44, 697-706. Schumach, M. State policy on mental patients scored for creating "revolving door" effect. New York Times, March 18, 1974, p. 1. Scully, D., & Windle, C. Community mental health centers and the decreasing use of state mental hospitals. Community Mental Health Journal, 1976, 12, 239-243. Segal, S., & Aviram, U. Community-based shelter care. In P. Ahmed & S. Plog (Eds.), State mental hospitals. New York: Plenum, 1976. Siguel, E. The impact of the development of community mental health centers on the utilization of state and county mental health hospitals (Doctoral dissertation, University of Michigan, 1974). Dissertation Abstracts International, 1974, 35, 2414B2415B. (University Microfilms No. 74-25,327) Windle, C., Bass, R., & Taube, C. PR aside: Initial results from NIMH's service program evaluation studies. American Journal of Community Psychology, 1974, 2, 311-327. Windle, C., Goldsmith, J., Schambaugh, J., & Rosen, B. Demographic differences between areas with and without federal community mental health center grants (Mental Health Demographic Profile System Working Paper No. 22). Washington, D.C.: U.S. Government Printing Office, 1975.