This study examined 124 private for-profit and not-for profit alcohol and drug treatment centers in California, Florida, Georgia, Minnesota, and New York. The sample was selected from the population of inpatient treatment centers that was identified in each state. The population was constructed from various state and national directories, and from telephone directories. During the 1988-1989 year, on site interviews with the chief administrator of the 124 centers were conducted. The data used in this study is based on codings of these interviews. This approach is similar to that of D’Aunno and Price (1985) and D’Aunno and Sutton (1992).

The response rate to the study was high. Ninety percent of the centers approached ultimately participated in the study. More than half of the nonparticipants failed to participate because of difficulties in coordinating interview schedules, with the remaining portion (less than five percent of the total) refusing to participate. Participation rates did not vary by state, size, age, or for-profit status.

The original study design did not include a longitudinal component and the control variables and the independent variables were all measured during the initial interview in 1988-89. The opportunity for analysis of organizational death and survival through the collection of additional data points about this outcome emerged as reports of treatment center demise were surfacing during the discussions surrounding health care reform. Indeed, fewer than 10% of the respondents indicated, during the initial on-site interviews, that their treatment centers were in a period of decline in profits or in number of clients.

The organizations in the sample were contacted by telephone to ascertain whether they were still in existence at four times subsequent to the initial on-site interviews. If a center could not be reached through information gathered through telephone directories, the center closure was verified by other treatment centers or by key informants located in the immediate geographic area. The survival or closure of the centers was ascertained in 1991, 1992, and 1994. Telephone calls to the surviving centers during the fall of 1995 indicated that there were no new closures. More refined information about the dates of closure was not ascertained. The time interval during which a center closed is known, but the timing within the intervals is unknown.


Because the independent and control variables were only measured during the initial interview, these data do not have time-varying covariates. Consequently, binary logistic regression analysis comparing those organizations that survived the entire time period to those that did not is more appropriate for this data set than event history analysis. In addition, a multinomial logistic regression with 4 ordered categories of the periods through which the organizations survived is estimated.


Dependent Variable

Survival. Survival was operationalized as the existence of the treatment center 6 years after the initial interview was conducted in 1988-1989. Sixty-six percent of the centers were operating in 1994, while thirty-four percent were closed. While the exact month and date of closure were not ascertained, 18 (14.5%) of the centers closed by 1991, another 17 (13.7% ) closed by 1992, and another 7 (5.6%) closed by 1994. The organizations that closed during the study period were coded 0 and those that survived were coded 1. In addition, the order of the time period of closure was also coded so that 1=center still open; 2= recent closure, center closed between 1992 and 1994; 3=middle closure, center closed between 1991 and 1992; 4=early closure, center closed between 1989 and 1991.


Independent Variables

Autonomy. Autonomy over two aspects of alcoholism treatment center management was measured: Decision making concerning the core technology of alcoholism treatment; and personnel management related to alcoholism treatment center staffing. These variables were ascertained during the initial on-site interview. Center administrators were asked during the interview to use a five point scale where a one represented no autonomy and a five represented total autonomy, to indicate how much autonomy the center has over decisions about its treatment practices and about its staffing.

Dispersed interorganizational linkages. Since multiple linkages can increase external control of organizations and their subsequent survival, the number of categories from which clients were referred was ascertained from the administrator during the interview. The respondent was asked to report the percentage of patients who were referred from the courts or criminal justice system, from community mental health centers or agencies, from private employee assistance programs associated with privately held workplaces, from private practicing physicians, from other treatment sources, from former patients, from recovering members of Alcoholics Anonymous, from family members, from self-referrals and other. If there were any patients from a source, it contributed a 1 toward the sum of the index of the number of referral sources. A lower number of referral sources is indicative of concentrated linkages, while a higher number is indicative of dispersed linkages.

Munificent interorganizational linkages. Two indicators of munificent referral linkages were measured: Referrals from public sector organizations are indicative of more munificent linkages and with those from private sector organizations indicative of less munificent linkages. The percent of patients who were referred from the courts or criminal justice system and the percent of patients who were referred from community mental health centers indicates the linkage with public sector organizations. The measure was transformed into its natural logarithm for the analyses.

The percentage of patients who were referred from employee assistance programs (EAPs) indicates the linkage with private sector organizations. EAPs are designed to motivate both supervisory and self-referrals of employees with alcohol and other behavioral health problems (Sonnenstuhl and Trice, 1986). They balance work performance and benefits management issues. The core technology of EAPs includes management of health care benefits by linking clients with appropriate treatment sources, on the micro level, as well as at the macro level where the aggregate benefit is managed (Blum and Roman, 1989). The macro function is analogous to what is now known as managed care. The measure was transformed into its natural logarithm for the analyses.


Control Variables

Size. Resource dependence theory holds that larger firms have a stronger need for interorganizational linkages because their size demands more financial resources and raises their visibility (Boyd, 1990; Pfeffer and Salancik, 1978). In addition, liabilities of size arguments suggest that larger organizations are less likely to die than smaller organizations (Hannan and Freeman, 1984). For these reasons, size was controlled in this study. Size was measured as the number of beds in the treatment facility based on data obtained from interviews with the center administrator. The use of this measure of size follows the model of Pfeffer and Salancik (1978), D’Aunno and Sutton (1992) and Gapenski, Vogue and Langland-Orban (1992).

Age. Liabilities of newness arguments suggest that new organizations are more likely to die than are older organizations (Hannan and Freeman, 1984). For this reason, age was included as a control variable. Age was measured as the length of time the organization had been in existence based on the year the center was opened and was obtained from interviews with the center administrator. This approach followed the model of D’Aunno et al (1991) and D’Aunno and Sutton (1992).

For-Profit Status. Healthcare organizations are often distinguished according to profit and non-profit status and may differ on a number of important dimensions that might influence survival, such as sources of funds, financial management, and customer characteristics (Cf. Kenkel, 1995). Further, the profit motive has been suggested as an incentive for greater efficiency in for-profit as compared to non-profit hospitals (Institute of Medicine, 1986). A dummy variable of 1 to indicate for-profit organizations and a 0 to indicate not-for-profit status was used. Information on the organization’s status as a for-profit or not-for-profit organization was obtained from interviews with the chief administrator.

Occupancy Rate. Much research has shown that firms that are performing well are less likely to die than firms performing poorly even though performance and survival may be related to different processes (Meyer and Zucker, 1989). For this reason, we controlled for occupancy rate. In the health care sector, occupancy of hospital beds is an important measure of performance. The occupancy rate of the treatment centers during the initial data collection was measured and used in these analyses (Bell, 1994). The occupancy rate during the initial data collection correlates 0.87 with the average of the occupancy rate measured at up to four other times between the initial interview in 1988 and 1994 if the center survived. The average census rate of these centers is comparable to the average census rates of hospitals generally in 1989 (National Center for Health Statistics, 1993).

Cost. Since the amount charged for goods and services can affect the survival of firms (Kenkel, 1995; Ehreth, 1993), the cost charged per day of treatment was measured during the interview with the administrator and is included in the analyses.



The results of the logistic regressions predicting the survival of the treatment centers are shown in Table 1. SAS software was used to estimate the models and the coefficients should be interpreted in the direction of survival. The binary logistic model, reported in Panel A, that includes the five control variables and the five independent variables has a chi-square of 38.2, with ten degrees of freedom and a significance of p=.0004, while the multinomial ordered logistic regression model, reported in Panel C, also has a significant fit with chi-square of 34.34, with ten degrees of freedom and a significance of p=.0001.

The results of the binary and multinomial ordered logistic regression analyses, shown in Panel A and Panel C, are consistent. Two of the control variables, year of founding and occupancy rate, are significant as expected. Older alcoholism treatment centers and treatment centers with higher census are more likely to survive. The control variables of size, for-profit status, and cost per day are not significantly associated with survival in these analyses. The coefficients for the independent variables are significant at p<.05 in both the binary and ordered logistic regression models, except for autonomy over staffing. These results indicate support for hypotheses 1a, 2, 3a and 3b. Linkages that support autonomy over treatment policies enhance the likelihood of survival, in support of hypothesis 1a. Contrary to hypothesis 1b, autonomy over staffing is not associated with survival. Concentrated interorganizational linkages enhance survival, while linkages with dispersed referral sources inhibit survival, consistent with hypothesis 2. The munificence of organized public sector linkages, such as those from the criminal justice system or courts and community mental health centers enhance survival, while linkages that increase external control, such as private sector organizational referrals through EAPs and worksites inhibit survival, consistent with hypotheses 3a and 3b.

The tests of the hypotheses yield consistent results for the binary and multinomial logistic regression analyses. However, the parallel lines or proportional odds assumption for the four category logistic regression is not met when all of the 124 cases are in the model. Alcoholism treatment centers that closed in the most recent period (e.g. survived the longest, but ultimately closed) appear to be more like the ones that survived than the ones that closed during earlier time periods for the variables included in the models. Binary logistic regression analysis, excluding the 7 cases of most recent organizational closures, reported in panel B, confirm the conclusions about the tests of the hypotheses, with the coefficients and the variance accounted for in the reduced case model being somewhat larger as expected. More importantly, the analogous multinomial regression, reported in Panel D, comparing survivals, middle closures (between 1991 and 1992) and early closures (prior to 1991) meets the parallel lines or proportional odds assumption of multinomial logistic regression, and also yields the same conclusions about the tests of the hypotheses as the full binary model, the reduced case binary model, and the full multinomial ordered regression model.



Overall, the findings confirm that organizational linkages that impose costly and conflicting demands may have a detrimental effect on organizational survival. Alcoholism treatment centers which are organized in ways that minimize interorganizational linkages which reduce the ability of organizations to adapt to change are more likely to survive than are treatment centers which are not structured this way.

This study shows that different types of autonomy have different effects on organizational survival. While autonomy over treatment policies had the predicted effect on survival, autonomy over human resource staffing practices did not. One reason that autonomy might not be advantageous with respect to human resource practices is the ambiguity surrounding what makes for effective human resource practices in this context. While alcoholism treatment centers tended to be isomorphic (D'Aunno, Sutton and Price, 1991) in their staffing philosophy during the period that these data were collected, they were forced to adapt to local environments. Since their pay scales for counselors and related staff did not permit flow within a national or regional labor market, many centers may have been in a relative vacuum in having direction in how to deal with different local environments. Given the emphasis on recovery experience and counseling skills, a center's hiring skills would be very "soft" and difficult to validate. It is also evident from the interviews that the center administrators did not bring substantial experience in either strategic management or human resource management to their positions. Their prior experience in alcoholism treatment was mainly as counselors or supervisors, and they may not have been well equipped to deal with the ambiguities of selecting center staff in a challenging environment. Under these circumstances, autonomy in dealing with personnel staffing decisions would appear to be a negative rather than a positive complement to organizational effectiveness. By contrast, access to direction in dealing with personnel issues from experienced managers in superordinate organizations in dealing with personnel issues would be associated with greater effectiveness. The implications of these findings are consistent with D'Aunno and Sutton's (1992) findings that treatment center management exhibits threat-rigidity responses to external threats, with such responses being more likely than alternative management responses to lead to organizational failure. Such rigidity can easily translate into inflexible personnel decisions. Therefore, autonomy over different activities may have different effects on organizational performance, and may even be moderated by different environmental conditions.

The present study shows that fewer interorganizational linkages is advantageous for survival presumably because it reduces the impacts of demands from a multitude of conflicting organizations. By contrast, D’Aunno and Sutton (1992) found that fewer interorganizational linkages reduces the probability of treatment center survival. D'Aunno and Sutton argued that this pattern of linkage adversely alters the power balance between external constituents and the organization. The differences in the two studies may be due to differences in research methodology, the period of time during which the research was conducted, or sample construction(Weisner and Morgan, 1992), since the D'Aunno and Sutton sample was primarily composed of publicly funded treatment centers, geared heavily toward treating problems associated with illegal drug use.

Alternatively, further post-hoc examination of our data may reconcile these disparate findings. While the dispersion of linkages does not have a significant quadratic association with survival, an examination of the number of linkages in surviving and dying organizations indicates that those centers that had fewer than 5 linkages were 1.7 times more likely to survive than to die, while those with 7 linkages were 3 times more likely to die than to survive. Thus, even though the statistical analysis did not ascertain an inverted U shape relationship between the number of linkages and treatment center survival, the number of interorganizational linkages appears to be important to survival to a point. Nevertheless, at some empirically determined point in the distribution, the negative effect of additional linkages overwhelmingly diminishes the probability of organizational survival.

The present data also indicate that the establishment of interorganizational linkages that are non-munificent may be detrimental to organizational survival. This finding is important because under conditions of uncertainty, which characterize many industries today, many organizations seek to "exert greater influence to reestablish the illusion or reality of control and stability over future organizational outcomes" (Oliver, 1991: 170). The present findings indicate that such a strategy may be detrimental to organizational survival in certain environments and has important implications for research on interorganizational linkages.



This study has shown that alcoholism treatment centers were more likely to survive a period of threats to their legitimacy and the turbulent reimbursement environmental if they were organized to minimize interorganizational linkages that reduced their ability to adapt to change. Since specialty treatment centers are an important type of organization in the health care industry, these findings provide additional understanding of an important dimension of today’s health care industry. The survival of health care organizations in general may depend on their ability to manage interorganizational linkages. Hopefully, this study will spur other researchers to examine how interorganizational linkages influence the performance of health care organizations. Such an understanding would be useful because health care is accounting for an ever increasing portion of our economy, and the cost of health care, to the public and private sectors, may affect our global competitiveness.


Previous Page | Main Menu | Next Page



1997 Babson College All Rights Reserved
Last Updated 1/15/97 by Geoff Goldman & Dennis Valencia

To sign-up for the Center for Entrepreneurial Studies' publication lists,
please register with the
Entrepreneurship WebTeam.