zy zyxwvu Horizontal and Vertical Integration-Diversification in Rural Hospitals: A National Study of Strategic Activity, 1983-1988 Stephen S . Mick, PhD, Laura L. Morlock, PhD, David Sulkever, PhD, Gregory de Lissovoy, PhD, Faye E . Malitz, MAS, Christopher G. Wise, M H S A , and Alison S . Jones, PhD ABSTRACT This study examines both the magnitude of and factors influencing the adoption of 13 horizontal and vertical integration and diversification strategies in a national sample of 797 U S . rural hospitals during the period 1983-1988. Using organization theory, hypotheses were posed relating environmental and market factors, geographic location, and hospital characteristics to the adoption of horizontal and vertical integration and diversification. Results indicate that only one of 13 strategies was adopted by more than 50 percent of all rural hospitals during the study period, and that most of the directional hypotheses were not confirmed using Cox’s proportional hazards models. In particular, environmental and market factors were unrelated to the strategies studied. Issues of methodology and theory are discussed; however, during an historically turbulent period, both relatively low levels of rural hospital strategic activities and lack of predictive power of the theory suggest caution in relying heavily on a policy for rural hospital survival that is dependent on individual market-oriented strategic behavior. zyxwvu zyxwvutsr gies to reverse declining performance, and, in this vein, the professional and trade literature contains literally hundreds of articles promoting one kind of strategic remedy or another (Mick & Morlock, 1990). Yet, the extent to which rural hospitals have heeded this advice and what factors are associated with rural n view of the well-publicized difficulties of US. rural hospitals (Congressional Budget Office [CBOI, 1991; Ermann, 1990; Hart, Amundson, & Rosenblatt, 1990; Merlis, 1989; Moscovice, 1989; Office of Technology Assessment [OTA], 1990), widespread opinion exists that strategic activity, particularly horizontal and vertical integration, on the part of rural hospitals may improve both their financial performance and their chances of survival (Buada, Pomeranz, & Rosenberg, 1985; James Bell Associates, 1991; Weiss, Phillips, & Schuman, 1986).Indeed, great hope has been placed in individual rural hospital adoption of these strate- I h,lick, Morlock, Sulkever, de Lissouoy, Malitz, Wise, and /ones ~____ ~~~~~ The research for this paper was funded by grant no. ROI-HS 05998-02, Agency for Health Care Policy and Research, Department of Health and Human Services. An earlier version of this paper was presented at the annual meeting of the National Rural Health Association, May 7,1992, in Washington, DC. For further information, contact: Stephen S . Mick, PhD, Department of Health Services Management and Policy, School of Public Health, University of Michigan, 1420 Washington Heights, Ann Arbor, M I 48109. 99 Spring 1993 zy zyxwvu zyxwvutsrq zy hospital strategic activity have been two largely unanswered questions. Answers to these questions may help evaluate the wisdom of a “strategic management policy” for rural hospitals. This paper has two purposes: first, to quantify the extent of rural hospital involvement in various horizontal and vertical strategies during the period 1983-1988.The second purpose is to examine the effects of a set of factors thought to influence these strategic management activities. Strategic Management Activities Strategic management activity was defined according to contemporary views, both in health care (Kimberly & Zajac, 1985; Shortell, Morrison, & Friedman, 1990; Topping & Hernandez, 1991) and in the general management literature (Ansoff, 1979; Chakravarthy, 1986; Herbert & Deresky, 1987)as any managerial effort designed to influence rural hospital adaptation of its structure and behavior to changing environmental circumstances. This study focuses on one aspect of hospital strategic managementstrategy content. Smith & Piland (1990)differentiate this dimension from ”strategy planning processes.” Strategy content may be thought of as the actual substance of programs and activities to which the strategy process leads. The study covers a range of 13 horizontal and vertical strategies from group purchasing arrangements to outpatient services within and outside the service area, from multihospital system affiliation to conversion of acute care beds to swing beds. Horizontal and vertical integration are two generic forms of strategic activity that are widely viewed as relevant to rural hospitals (Buada, Pomeranz, & Rosenberg, 1986). Horizontal Integration. Horizontal integration involves any new interrelationship between or among two or more hospitals, a generalized strategy that has long been recommended for hospitals (Starkweather, 1971,1972).Among autonomous, independent hospitals, these horizontal arrangements cover a spectrum of activities from complete integration of services and facilities resulting from merger and consolidation to a variety of consortia, such as voluntary health planning councils (DeVries, 1978),to group purchasing arrangements. The most extreme form of horizontal integration is the merger or consolidation of two or more hospi- tals (Fottler, Schermerhorn, Wong, & Money, 1982). Technically, mergers differ from consolidations:in the former, one of the organizations retains its name am3 legal status whereas in the latter a completely new legal entity derives from the melding together of separate organizations (Finkler & Horowitz, 198s).For simplicity, this study groups both forms of horizontal integration into a single category. Economies of scale, concentration of management talent, increased access to capital markets, reduction of redundant services, and personnel are reasons motivating this form of horizontal integration (Rowland & Rowland, 1984a). Another horizontal arrangement is the multihospital system, defined as two or more acute care hospitals that are owned, leased, sponsored, or managed by a single corporate entity (Zuckerman, 1979).Rural hospitals have joined multihospital systems in increasing numbers (Lewis & Parent, 1986): recent studies have shown that about 40 percent of all rural hospitals are affiliated with a multihospital system (Rosenberg, 1989). Consortium affiliations are aggregations of hospitals that have agreed to engage in some joint cooperative venture. They qualify as instances of horizontal integration because they require linkage across hospital facilities. Increasingly popular in thr 1980s, consortia provided avenues for hospitals to keep their own autonomous corporate identity (Provan, 1984),while benefiting in joint activities such as staff education programs, staff recruitment, shared services, and legislative liaisons (Moscovice, Johnson, Finch, Grogan, & Kralewski, 1991). Group purchasing is an arrangement that seeks to reduce costs through bulk purchasing, competitive bidding, and contract negotiations for supplies and services (Rowland & Rowland, 1984b).Among cooperating hospitals it requires some minimal standardization, guarantee of inventory space, and information sharing among designated administrators. Group purchasing can exist among very loosely connected hospitals as well as in a tightly integrated multihospital system. Thus, even though there is undoubtedly overlap with some of the other categories of horizontal integration, group purchasing defines the lower boundary of horizontal integration because only a minimal amount of horizontal interchange is required by a given hospital. Vertical Integration. Vertical integration is the production by a single organizational entity of successive stages in the processing and distributing of goods and services (Mick, 1990a).For an acute care hospital, zyxwvutsrqponm zyxwvutsrqponm zyxwvutsrqpo zy The IOtfrl7d of Rural Health 100 Val. 9, h.0. 2 explicitly in this category because “unrelated“ vertical integration may refer to involvement with one suggests that these activities neither feed directly into or more ”input sources,” such as a primary care clinic nor flow out of the hospital, but relate in the way from which referrals to the inpatient facility might be argued by Clement (1988),noted earlier. made, or an ambulance service that brings patients to the facility. The organization can also extend itself into ”output” services or products such as a nursing home facility to which patients are discharged. Conceptual Approach and Hypotheses A recent controversy in the literature concerns whether instances of vertical integration are actually This study hypothesizes that there was an forms of diversification. Diversification refers to the increase in strategic management activities during the production of outputs that are neither substitutes for six-year period 1983-1988.This period encompasses each other nor complements of other services (Clemthe implementation of the Prospective Payment ent, 1988).A review of studies on diversification and System (PPS), an important, but not the only, instance vertical integration in health care demonstrates no of 1980s experimentation with incentives to effect cost crisp empirical distinction between the two phenomcontainment and organizational efficiency in the ena. Starkweather and Carman (1987),for example, health sector (Mick, 1990b).Although this study does use the terms vertical and horizontal diversification, not posit a singular causal influence of PPS as a soconcentration, and integration. This study amends called ”environmental jolt” on the strategic manageAlexander’s (1990) terminology of “closely related ment activities of rural hospitals during the study diversification,” “partially related diversification,” and period (Meyer, 19811, it does argue that it helped “unrelated diversification.” Additionally, a geographic define 1983 as a good beginning point for study (Note dimension concerning whether the strategic activity 1). The follow-up year is set at 1988 because that was was inside or outside the administrator-defined the latest date for which comprehensive financial hospital service area is added. Thus, categories for this data were available at the time this study began. study consist of (1) closely related, (2) partially related, Formally, the first hypothesis is: and ( 3 ) unrelated vertical integrated-diversified HI: The proportion of rural hospitals engaged in services. Vertical integration is considered to be the strategic management activities was higher in predominant characteristic of the first two categories; 1988 than in 1983. diversification predominates in the third. The hypothesis is based on the numerous exliorClosely related vertical integration-diversification tations in both the trade and scholarly literature to try includes services such as strategic activity (Mick & Morlock, 1590).There is the affiliated long-term care facilities (usually a nursing impression based on the trade literature that rural home), ambulance or emergency medical transport hospitals have everywhere engaged in strategic vehicles, and swing beds. activities that often required fundamental restructurPartially d a t e d vertical integrated-diversified ing of the organization. There are also compelling services include outpatient services or clinics in the theoretical reasons for the hypothesis that derive hospital’s service area, as well as affiliation with a from a resource dependence perspective (Pfeffer & health maintenance organization (HMO) or with a Salancik, 1978; Thompson, 1967) that argues organipreferred provider organization (PPO).HMOs have zations will seek to control or rationalize their shown promise for rural hospitals in some areas relationships with the environment through a variety (Christianson, 1989),and their ability to secure a stable of strategic mechanisms. The conceptualization of the patient base may be an important attraction. Unlike environment is a topic that has occupied organizaHMOs, which have existed for decades, preferred tional analysts since the early 1960s, and although provider organizations (PPOs) were part of the 1980s many formulations exist for characterizing environphenomenon of selective contracting (Rice, de ments (Aldrich, 1979; Staw & Szwajkowski, 19751, we Lissovoy, Gabel, & Ermann, 1585). rely on Dess and Beards (1584) summary of these Unrelated vertical integration-diversification refers concepts in terms of three fundamental dimensions: to other health care services, e.g., durable medical environmental munificence, complexity, and dynaequipment manufacturing, as well as non-health mism. services, e.g, parking lots, building maintenance, and Munificence refers to the resource level or capacoutpatient services or clinics outside the hospital’s ity of the environment to sustzin the needs of orgaxiservice area. Diversification is emphasized inore zations. Examples include the number of potential zyxwvutsrqp zyxwvuts zyxwvutsr Alick. Morlock, Snikeuer, de Lissovoy, Mnlitz, Wise, n l ~ dIories 101 zyx Spring 1993 patients, average income level, strength of the economy, and so forth. Complexity refers to the level of heterogeneity or the concentration of organizations, or both, in the environment. Examples include number and types of other health care organizations, especially competitors. Dynamism refers to the level of instability or turbulence in the environment. Indicators include changes in any of the relevant measures of munificence and complexity. These three environmental features all bear on the rural hospital, and at the most general level, as the environment becomes-and is perceived as becoming-characterized by scarcity, complexity, and change, one expects strategic activity as an organizational response. But because there are quite different strategic activities potentially available to rural hospitals, a more refined view than this is necessary to explain why these organizations engage in one or another strategy. It is posited that different effects of the three environmental dimensions on the likelihood of adopting horizontal and vertical integration-diversification strategies exist. Following Shortell and Zajac’s (1990) statement that strategies are functions of eclectic forces including environments, markets, organizations, and other factors, this study focuses on two other groups of pertinent variables: geographic location and hospital characteristics. zyxwvu zy rangements. Hoped for consequences of horizontal integration include economies of scale, concentratiim of managerial talent, reduction of redundant services, and access to capital markets, among others-all things that help a hospital react more favorably to resource constraints. Taken together, the following three hypotheses can be posed: H2: The munificence of rural hospital markets is negatively related to horizontal integration strategies, other things being equal. H3: The complexity of rural hospital markets is unrelated to horizontal integration strategies, other things being equal. H4: The dynamism of rural hospital markets is positively related to horizontal integration strategies, other things being equal. Vertical integration. Conrad and Dowling (19901; Conrad, Mick, Madden, and Hoare (1988);Mick (1990a);and Mick and Conrad (1988)have suggested that a transaction cost model explains why health care organizations integrate vertically. Simply stated, transaction costs are those that are encountered in the organization and management of health services production. If the costs of market exchanges increase, then the organization will attempt to integrate vertically to obtain these services internally to capture greater efficiencies made possible by the variety of internal managerial control mechanisms available to it (Williamson, 1975). A key reason why market transaction costs rise is increasing uncertainty. Two sources of uncertainty are environmental complexity and dynamism, both of which increase the transaction costs of conducting business in the environment. The organization can stabilize its relation to these markets, and lower its transaction costs, by incorporating the relevant input suppliers (e.g., primary care clinics, HMOs, PPOs, radiology and imaging services) or output destinations (e.g., nursing homes and other long-term care facilities, rehabilitation services).Note that the same general prediction is supported by resource dependence theory. For example, Pfeffer and Salancik (1’978) state that vertical integration is an important way that organizations reduce uncertainty and decrease the need to make risky adaptations that could threaten the organization’s survival. In sum, vertical integration is the rural hospital’s attempt to gain direct control over its local market when that market displays high complexity and high dynamism. On the other hand, munificence is not a force predicted to influence vertical integration. If the munificence of a market or environment is low, there zyxwvuts zyxwvuts Hypotheses Related to Environment and Market. Horizontal integration. Environmental and market pressures that favor horizontal arrangements include those that constrain the availability of funds and payments, constrict the demand for services, and otherwise limit the ability of the hospital to continue business as usual (Starkweather & Carman, 1987).In short, environmental munificence will be negatively related to horizontal integration. Second, the complexity of the environment alone creates neither an incentive nor a disincentive to integrate horizontally. Other things being equal, organizations like hospitals prefer autonomy and independence (Cook, Shortell, Conrad, & Morrisey, 19831, and a complex environmental mix of health care organizations does not in itself produce a rationale for surrendering these traits. Finally, as a rural hospital environment or market increases in dynamism, i.e., moves toward greater instability or turbulence, especially in regard to payment and demand for services, a rural hospital will be more likely to engage in horizontal integration ar- The Journal of Rural Health 102 zy VOl. 9, KO.2 is little reason for a rural hospital to extend itself any further because little is to be gained by spending scarce resources to tap scarce input or output sources. If munificence is high, there is even less reason to extend vertically because resources will be readily available to the hospital regardless of vertical arrangements. A key exception to this argument arises in the case of unrelated vertical integration-diversification in which diversification is probably the more appropriate way to describe activities like nonhealthrelated strategies and ambulatory clinics outside the service area. In this instance, it can be predicted that when munificence is low, the rural hospital-searching for new revenue opportunities-might attempt to enter alternative markets. Thus, one would expect a negative association between environmental munifi-. cence and unrelated vertical integration-diversification. On the other hand, positive associations between this form of vertical integration-diversification and complexity, as well as dynamism, can be posited. In the case of complexity, the options for diversification are much greater than in noncomplex environments. In the case of dynamism, diversification may be attractive as a defense against the vagaries of change and loss of current revenue sources. Given these expectations, six additional hypotheses are proposed: H5: The munificence of rural hospital markets is unrelated to closely and partially related vertical integration-diversificationstrategies, other things being equal. Hsa: The munificence of rural hospital markets is positively related to unrelated vertical integration-diversificationstrategies, other things being equal. H6: The complexity of rural hospital markets is positively related to closely and partially related vertical integration-diversification strategies, other things being equal. The complexity of rural hospital markets is positively related to unrelated vertical integration-diversificationstrategies, other things being equal. H7: The dynamism of rural hospital markets is positively related to closely and partially related vertical integration-diversification strategies, other things being equal. H7a: The dynamism of rural hospital markets is positively related to unrelated vertical integration-diversificationstrategies, other things being equal. Hypotheses Related to Geographic Location. Location is often cited as an important defining characteristic of rural hospitals (DeVries, 1988),with some stating that very remote rural hospitals (e.g., "frontier" hospitals) are dissimilar to those in more compactly settled regions. Implied in these arguments is that a remotely located hospital, such as one in Montana, will have little ability to engage in strategic activities whereas those in more densely settled areas (e.g., Vermont) will have more ability to do so. The problem is that location may combine a number of factors, any of which can be operating alone or in concert to influence the strategic management activities of interest. The study includes location because there is intense interest among the rural health policy and analysis communities in possible regional or situational differences, quite apart from the impact of more direct measures of each dimension of the environment or market. This interest is generated by a belief that regions have certain distinguishing traditions, norms, and customs, and may have different regulatory climates. For example, the New England and Mid-Atlantic census divisions are known for having deliberative and regulation-minded state governments, especially in health matters. This situation is in contrast to the East South Central or West South Central census divisions that consist of states not known for their regulatory oversight in health and social welfare issues. One might posit that rural hospital activities would be more constrained in the former divisions and less so in the latter. On the other hand, one could just as forcibly argue that the latter census divisions would not supply an encouraging state environment for experimenting with rural hospital innovation in the face of cost containment pressures felt nationally during the 1980s, whereas more support might be expected in the traditionally more liberal divisions of the U.S. Northeast. In sum, it is difficult to predict what impact, if any, location might have; therefore, the null hypothesis is posited for both horizontal and vertical integration: Hg: Regional location of rural hospitals is unrelated to horizontal integration strategies, other things being equal. H9: Regional location of rural hospitals is unrelated to vertical integration strategies-diversification other things being equal. zy zyxwvuts zyx zyxwvutsrqp Mick, Morlock, Salkever, de Lissovoy, Mulitz, Wise, and Jones Hypotheses Related to Hospital Characteristics. Size. The larger the facility, the more likely it is to have engaged in both horizontal and vertical strategic 103 Spring 1993 zyx zyxwvu zyxwvutsrqpon zyxwvutsrq management activities due to its greater resource base, organizational "slack," and economies of scale (Christianson, n.d.), with respect to managerial expertise, financing ability, likelihood of higher occupancy rates, and lower per unit fixed costs. These features may permit a larger hospital to adapt to its environment, either through horizontal or vertical arrangements, more easily than a small hospital. 'Thus, it is hypothesized that: Hlo: Hospital size is positively related to horizontal integration strategies, other things being equal. H11: Hospital size is positively related to vertical integration-diversificationstrategies, other things being equal. Case Mix. The case-mix index is a measure of the severity of illness and trauma among patients in an institution. A high case mix may indicate that a rural hospital has high resource consumption in dealing with such patients. This situation may erode a hospital's resource base, leading to financial strains that provide motivation to engage in horizontal integration strategies to effect cost efficiencies. Thus it is expected that: H12: Case mix is positively related to horizontal integration strategies, other things being equal. Vertical integration presents a somewhat different problem: a high case mix can mean high costs for a rural hospital, and one way to lower them is to have the hospital vertically linked with supplier and distributor entities. In this way, the focal unit, i.e., the acute care facility, has greater control over the entire treatment path of the patient and can actually reduce per unit costs through careful coordination of care. It is predicted that: H13: Case mix is positively related to vertical integration-diversificationstrategies, other things being equal. This study also argues that an increasingly complex case mix over time would lead to the same horizontal and vertical outcomes: H14: Change in case mix is positively related to horizontal integration strategies, other things being equal. H15: Change in case mix is positively related to vertical integration-diversificationstrategies, other things being equal. Ownership To predict the effect of ownership, this study uses the not-for-profit community hospital as the reference group. It posits that for-profit ownership will be positively correlated with strategic activity because most for-profit hospitals, especiallly those in large national chains, were the first to champion corporate-like organizational strategies (e.g., horizontal and vertical integration-diversification) in health care (Herzlinger & Krasker, 1987). Conversely, government-owned hospitals (e.g., municipal, county, and district hospitals), which usually have elected boards and statutory limitations, as well as reliance on tax revenues, require their administrators to be less discretionary in their decision making than their counterparts in the community not-for-profit sector. Therefore, little horizontal or vertical strategy is expected among this group. The not-for-profit religious or church hospitals are expected to engage in both horizontal and vertical strategies more often than communitynot-for-profit hospitals because many of the former have a long tradition of participating in loosely coupled multihospital arrangements run by religious orders. Building on this tradition would be relatively easy compared to rural hospitals with no such history. This study also argues that the churchaffiliated institutions, in keeping with their service missions, are more likely to engage in enterprises that involve vertical extensions into their communities, with less regard for financial consequences than other community not-for-profit rural hospitals. T h s notion is consistent with work by Ludke, Westhofl, and Flood (1992) in which they demonstrated that rural Catholic hospitals were more likely than other rural hospitals to be in an activist "prospector" category (Miles & Snow, 1978).The following hypotheses summarize these relationships: H16: For-profit ownership is positively related to horizontal integration strategies, other things being equal. H17: For-profit ownership is positively related to vertical integration-diversificationstrategies, other things being equal. His: Government ownership is negatively related to horizontal integration strategies,, other things being equal. H19: Government ownership is negatively related to vertical integration-diversification strategies, other things being equal. H20: Religious not-for-profit ownership is positively related to horizontal integration strategies, other things being equal. H21: Religious not-for-profit ownership is positively related to vertical integrationdiversification strategies, other things being equal. zyxwvut Tlw Journal of Rural Health 104 VOl. 9,[v-o. 2 zyxwvuts Table 1. Summary of Hypothesized Relationships Among Horizontal and Vertical Integration/ Diversification Strategies and Environmental, Geographic, and Hospital Characteristics, U.S. Rural Hospitals, 1983-1988. Characteristic Horizontal Integration Hypothesis Prediction Munificence H2 Negative Complexity H3 Dynamism Geographic location Environment Vertical Integration zyxwvut Supported’ Hypothesis Prediction Supported’ No H5 H5a 0 Negative Yes No 0 Yes H6 H6a Positive Positive No No H4 Positive No H7 H7a Positive Positive No No H8 0 Yes H9 0 Yes Log operating beds H10 Positive No H11 Positive Yes? Case-mix index 1983 H12 Positive No H13 Positive No Change case-mix 1983-1988 H14 Positive Yes? H15 Positive No For-profit H16 Positive No H17 Positive No Governmental H18 Negative Yes? H19 Negative No Not-for-profit, church H20 Positive Yes? H2 1 Positive No Total margin 1983 H22 0 Yes H23 0 Yes Change total margin 1983-1987 H24 0 Yes H25 0 Yes Number head administrators 1983-1988 H26 Negative No H27 Negative No Hospital 1. zyxwvuts “Yes” indicates that more than one half of the relevant coefficients are statistically significant in the predicted direction. “Yes?” indicates that one half or close to one half of the relevant coefficients are statistically siginificant in the predicted direction. ”No” indicates that less than one half of the relevant coefficients are statistically significant in the predicted direction. For all null hypotheses, the predictions are supported and are indicated by a ”yes.” Financial Performance. Attempting to predict the likely impact of financial performance presents a quandary: one would expect that as a rural hospital’s financial position declined, it would feel increasing pressure,to attempt different strategic management activities to improve its financial position. On the other hand, it seems just as likely that a rural hospital h l i c k , Movlock, Salkeuer, de Lissouoy, Malitz, Wise, arid Jones in deteriorating financial condition would not have the resources-operating funds, capital funds, managerial personnel-needed to make the investment required to undertake strategic management activities. In this study both horizontal and vertical integration strategies are grouped together and null hypotheses are expected to hold. Furthermore, it can zyxwvutsrq 105 Spring 1993 Sampling Procedure. The study identified mal hospitals (defined here as those not located in a Consolidated Metropolitan Statistical Area [CMSA] or a Metropolitan Statistical Area [MSAI) on both the 1986 AHA tape and the 1986 Medicare PPS data bases, the latest year available at that point in the study. All federal hospitals were excluded. Of the remaining hospitals, only those that had short-term lengths of stay were included in the sampling frame. These procedures resulted in records for 2,044 hospitals; from these a 50 percent proportionately stratified random sample of 1,021 hospitals was drawn. The strata were the nine U.S. Bureau of the Census divisions, and their inclusion in the sampling scheme ensured that the regional distribution of the universe was represented by the sampled hospitals. This overall design ensured a sufficient number of hospitals for the detailed statistical analysis planned for the study. be argued that change in financial performance will also be unrelated to either horizontal or vertical strategies: Hz: Financial performance is unrelated to horizontal integration strategies, other things being equal. H z : Financial performance is unrelated to vertical integration-diversificationstrategies, other things being equal. H24: Change in financial performance is unrelated to horizontal integration strategies, other things being equal. H25: Change in financial performance is unrelated to vertical integration-diversification strategies, other things being equal. Administrator Turnover. Finally, this study hypothesizes that the lower the amount of head administrator turnover experienced by the rural hospital, the more likely it would be that strategic management activities will be undertaken. This is based on the argument that leadership stability and followthrough are important characteristics for a rural hospital undertaking strategic organizational change (Hart, Robertson, Lishner, & Rosenblatt, 1992). Once again, this study considers that this logic holds for both horizontal integration and vertical integrationdiversification strategies: H26: Turnover of head administrators is negatively related to horizontal integration strategies, other things being equal. H27: Turnover of head administrators is negatively related to vertical integration-diversification strategies, other things being equal. All these hypotheses are summarized in Table 1. zyxwvut zyxwvutsrqpo zyxwvuts Methods Telephone Survey. The telephone survey instrument was developed through the advice of three rural hospital administrators, the assistance of a focus group organized by the National Rural Health Association, a review of questionnaires from other hospital-oriented studies, and the results of a pretest on 10 rural hospitals (Note 2). Key to the study was the comparison of service mix and strategic activity in FY 1982-83and FY 1987-88. Respondents were asked to remember events such as the addition or deletion of a service that had happened as much as (sometimes more than) five years earlier. In some cases, the chief administrator had not been at the hospital at the time a given service or strategy was adopted. To address the possibility of faulty recall, a procedure of advance telephoning and mailing of the survey instrument to the potential respondent was instituted, noting that if there was a question of not knowing some of the information, the respondent might want to consult with someone else in the hospital, e.g., a nurse or physician, to verify the information. The original sample of 1,021 was subsequently cut to 988 because of a loss of 33 hospitals. Twentytwo of these hospitals had closed between 1986 and the time the interviews were conducted, although several were trying to reopen. The remainder had converted to nursing homes, psychiatric facilities, or clinics. This 3.2 percent loss of the original sample was small enough that any resulting selection effert was considered insufficient to bias the results of the study. zyxwvutsr Overview. Data are derived from a telephone survey of 797 chief administrators of rural hospitals combined with secondary data from the American Hospital Association (AHA), the national Area Resource Files (ARF), and the Health Care Financing Administration (HCFA)Medicare cost reports. Survey questions were asked of each administrator to determine what the strategic and operational status of the hospital was during fiscal year 1987-88 (telephone interviews were conducted from June to December 1988), as well as what the hospital's status was five years earlier in fiscal year 1982-83. This approach allowed the creation of a retrospective panel study. zyxwvutsrqponm zyxwvutsrq zy zy The Journal of Rural Health 106 vor. 9,N ~2 . zyxwvut zy zyxwvutsr Based on the reduced sample, the completion of 797 interviews produced a response rate of 80.4 percent. Chi square tests performed on the reduced sample comparing nonresponding hospitals to responding hospitals revealed that for-profit hospitals were less likely to respond to the survey (67.0%, P<0.05), hospitals in the 100 to 199 bed size group were less likely to respond (75.0%,P<0.05), and hospitals with Joint Commission on Accreditation of Healthcare Organizations (JCAHO)accreditation were less likely to respond (77.3%,P<0.05). Examples of variables with no nonresponse/response differences included average daily census and occupancy rate. Because about one in 10 rural hospitals are forprofit and two in 10 are in the 100-199bed category, these sources of bias are probably not a major concern. The lower response rate of JCAHO-accredited hospitals is perhaps surprising, and caution should be taken when generalizing from study results to all rural hospitals. Measurement of Strategic Management Variables. Dependent variables in the study include the 13 horizontal and vertical strategies described earlier. Variables are defined in terms of the year of adoption of each of the strategies. In the statistical models to examine factors associated with adoption of the strategies, the number of years’ difference between the year of adoption and 1983was calculated, thus excluding hospitals that had adopted the strategy before 1983 (Note 3). This approach was required for the use of the chosen analytic procedure (Cox’s proportionate hazards model) explained below. Measurement of Independent Variables. Descriptive statistics for all independent variables are shown in Table 2. The characteristics of the rural hospital market or environment are measured at the county level (Note 4). Two approaches to their measurement are used. The concepts of munificence and complexity are measured by variables that describe the environment or market in or close to the baseline year 1983. The concept of dynamism is one of change, and hence a group of first difference variables, i.e., the value of 1983 variables subtracted from those in 1987 or 1988 are employed. Skewed variables are transformed using natural logarithms. The specific variables and their measurement are discussed in the appendix. describe the probability of strategic management activities during 1983-1988. A t test for paired comparisons was used to test hypothesis 1. Cox’s proportional hazards models are used to test Hypotheses 2 through 27 because the dependent variables of interest have the property of being “right censored.” That is, a certain number of hospitals will not have engaged in a strategy at the time their head administrators were interviewed, but adoption of the strategy might have occurred the following year or later. Not taking this possibility into account could produce biased estimates of the variables predictive of the particular dependent variable, and because hazards techniques account for this possibility, they are thus the appropriate analytic technique (Note 5). Findings Description of Sample. For the entire sample, the average number of licensed beds was 79.6 and the average number of operational beds was 6.5.4. Thirteen percent of the hospitals had 2.5 or fewer beds; 77.8 percent had 100 or fewer beds. The average occupancy rate of the sample in 1988 was 44.2 percent, and more than one fourth of the hospitals had a rate 30 percent or lower. About one fourth (24.7%)of the hospitals were located in the West North Central census division (Table 2). Strategies in 1983 and 1988. The average number of horizontal and vertical strategies in rural hospitals in 1983was 1.3 out of a possible total of 13, with 30.7 percent having no strategic activity. By 1988, the average number was 3.7, with only 0.6 percent (five hospitals) having no strategic activity. The 1983 to 1988 means were statistically different (t=44.870, P<O.OOl), and the findings support hypothesis 1. zyxw zyxwvutsrq Figure 1. Number of U.S. Rural Hospitals by Number of Strategic Activities, 4983-1988. zyxwvu 250. 5 P T f z zyxwvutsrqpo zyxwvutsrq 200 - ’50 t so; 0 Mick, Morlock, Salkever, de Lissovoy, Malitz, Wise,and Jones 1 2 3 4 5 6 7 E 9 10 zyxwvutsrq Statistical Approach. Descriptive statistics and cumulative frequency distributions are used to Number of Stnteper 107 Spring 1993 zyxwv zyxwvu The proportion of hospitals adopting one or another of the 13 strategies by 1988 varied considerably. In descending order, the proportion of strategies present was: group purchasing arrangements 80.4% swing-bed program 46.4% multihospital system membership 34.1% consortium affiliation 31.0% PPO affiliation 27.1% nursing home affiliation 25.6% HMO affiliation 22.0% ambulance services 21.2% outpatient services/clinics inside the service area 17.2% other health-related services 11.7% mergers/consolidations 5.8% nonhealth-related services 5.2% outpatient services/clinics outside the service area 1.8% Thus, except for group purchasing arrangements, less than one half of rural hospitals had engaged in any of the remaining 13 strategies. Figures 2 through 5 show the year of adoption of each strategy grouped into its generic category. For horizontal integration strategies, Figure 2 indicates that before 1983 group purchasing had been adopted by slightly more than 30 percent of rural hospitals; by 1983, it had nearly reached the 50 percent level and steadily grew to about 80 percent by 1988. Multihospital system membership showed a nearly constant linear increase from less than 20 percent to about 33 percent, a trend that was paralleled by consortium arrangements at only a slightly lower level. By contrast, mergers and consolidations were almost nonexistent before PPS as well as during and after the implementation of PPS. A slight upward movement approaching 5 percent was registered between 1987 and 1988. The adoption of closely related vertical integration strategies (Figure 3) was a phenomenon that started and ended at low levels except for one instance. Affiliation with nursing homes and ambulance services grew only slightly during the period studied, from about 10 percent to 15 percent of all rural hospitals to about 20 percent to 25 percent. By contrast, swing beds were virtually nonexistent at the beginning but had been adopted by about 40 percent of rural hospitals by 1988. As shown in Figure 4, rural hospital adoption of partially related vertically integrated strategies was less marked (Figure 4). Starting from a near zero point before PI’S, the adoption of outpatient services The ]ourrial of Rural Health Table 2. Descriptive Statistics for Independent Variables Included in the Proportional Hazard Models. Variable Mean Natural log of population per square mile, 1983 3.506 Natural log change in population per square mile, 1983-1987 -0.001 Per capita income, 1983($) 9279.509 Change in per capita income, 1983-1987 ($)I 1552.581 Percent Medicare eligibles, 1983 15.684 Change in percent Medicare eligibles, 1983-1987 0.777 Percent unemployment, 1981 8.600 Change in percent unemployment, 1981-1987 -0.700 Physicians per 1,000 population, 1983 0.865 Change in physisicians per 1,000 population, 1983-1987 0.059 Natural log miles to next nearest hospital 2.944 Natural log operating beds 3.886 Case-mix index, 1983 1.039 Change in case-mix index, 1983-1988 0.058 Total margin, 1983 0.025 Change in total margin, 1983.1987 -0.031 Number of head administrtors. 1983-1988 2.211 zyx zyx zyx zy Standard Deviation 1.133 0.096 1614.838 1220.730 3.901 0.993 0.12n 0.118 0.837 0.219 0.939 0.Z4 0.8347 0.9T3 0.m 0.215 1.568 Percent Voluntary not-for-profit church Voluntary not-for-profit community For profit Government (municipal, county, district) New England Middle Atlantic South Atlantic East North Central West North Central East South Central West South Central Mountain Pacific 1. 0.112 0.402 0.114 0.373 0.028 0.039 0.115 0.151 0.247 0.100 0.172 0.092 0.056 11.2 40.2 11.4 37.3 2.8 3.9 11.5 15.1 24.7 10.0 17.2 9.2 5.6 Real 1983dollars. in the service area and affiliation with either HMOs or PPOs was never high. By 1988, the prevalence hovered between 15 percent and 25 percent, with the strongest growth in PPOs. In regard to adoption of outpatient services in the service area, the most zyxwvutsrqponm zy 108 Vol. 9, ,vo. 2 zyxwvut __ Figure 4. U.S. Rural Hospitals, Distribution of Cumulative Proportions, Adoption of Partially Related Vertical Integration/ Diversification Strategies, 4983-1988. Figure 2. U.S. Rural Hospitals, Distribution of Cumulative Proportions, Horizontal Integration Strategies, <1983-1988. zyxwvutsrqpon - < 1983 1983 1984 1985 1986 1987 1988 Year ot Strategic Activity < 1983 1983 1984 Gp Purchasing - 1983 1984 -m Merged -p#) 1885 - 1987 1988 zyxw Consortium Figure 3. U.S. Rural Hospitals, Distribution of Cumulative Proportions, Adoption of Closely Related Vertical Integration/ Diversification Strategies, 4983-1988. < 1983 1986 Outpt in Sew Multi + - 1985 Year of Strategic Activity 1986 1987 Year of Strategic Activity 1988 common service offered was some form of outpatient/primary clinic care (78.1%).In all, the chief administrators reported offering 29 different outpatient services, with considerable diversity in responses among hospitals. By examining Figure 5, it can be seen that unrelated services, also beginning at nearly zero prevalence, never reached more than 10 percent (other health-related services), and in the case of ambulatory services offered outside the service area, remained close to zero by 1988. For other health-related services, the most common activity was laboratory services (28.2%),followed by home health agencies (11.7%),and durable medical equipment (9.7%).In all, 28 services were mentioned by the chief administrators. Nonhealth-related services that were tried included owning or operating, or both, apartment buildings (16.3%),office buildings (14.3%), and child day care centers (12.2%). Nursing Home - Swing Beds k f i c k , Morlock, Salkever, de Lissovoy, Malitz, Wise, nrzd \ones zy zyx Correlates of Strategies. The results of the Cox's proportional hazard models for all strategies are shown in Table 3. For ease of reading, only the sign (a) and statistical significance of coefficients at or beyond the 0.05 alpha level are shown (Note 6). Ambulance 109 Spring 1993 zyxwvutsrq zyxwvu Tests for Hypotheses 16,18, and 20 compare categories of ownership to not-for-profit community ownership with respect to horizontal integration. Hypothesis 16 is supported one time out of four; hypothesis 18, two times; and hypothesis 20, two times. Thus, these hypotheses receive weak support at best. Hypotheses 22 and 24, predicting no relationship between total margin and change in total margin and horizontal integration, are supported. Finallv, hypothesis 26, predicting a negative link between head administrator turnover and horizontal integration, is contradicted by two statistically significant positive coefficients; therefore, the hypothesis is rejected. Figure 5. U.S. Rural Hospitals, Distribution of Cumulative Proportions, Adoption of Unrelated Vertical Integration/ Diversification Strategies, 4983-1988. zyxwvutsrqpon - zyxwvutsrqp Vertical Integration-Diversification Strategies. < 1083 1083 1084 1085 1086 1087 Examination of relationships of closely and partially related vertical integration-diversificationand environmental munificence (hypothesis 5) reveals that out of a total of 18 coefficients, only three are statistically significant, supporting the hypothesis of no effect. Subhypothesis 5a-predicting a negative relationship between munificence and unrelated vertical integration-diversification-is not supported because only one of 12 coefficients is statistically significant. Hypotheses 6 and 7, environmental complexity and dynamism, respectively, predict positive associations with closely and partially related vertical integration-diversification.However, in the first case, only one of 12 coefficients is significant with the wrong sign, and in the second case, only four of 30 coefficients are significant, two in the predicted direction, two in the opposite direction. These two hypotheses are rejected. Subhypotheses 6a and 7a, which predict positive relationships between unrelated vertical integration-diversification,are supported. Hypothesis 9 predicts no link between census division and vertical integration: of 72 possible coefficients, 14 are statistically significant. Seven of these are for the specific strategy of HMO affiliation, but this is a function of the choice of the reference variable. Had a census division at or near the affiliition average been selected, fewer significant coefficients would have resulted. Hence, these study findings appear to support the hypothesis of no geographic effect. Hypothesis 11 links vertical integration positively to hospital size, and for four of nine strategies, the relationship holds. No relationship exists for an additional four strategies. Swing beds and size are negatively related, which is understandable because the program was originally established for rural 1988 Year of Strategic Activity Other Health Non-Health Outpt Outside Horizontal Integration. Only one of 16 coefficients linking horizontal integration strategies to environmental munificence is statistically significant, which leads to the rejection of hypothesis 2. None of the eight coefficients linking horizontal integration strategies to environmental complexity is significant. This finding supports hypothesis 3. Four of 20 coefficients relating horizontal integration to environmental dynamism are statistically significant, but one has the wrong sign. Thus, hypothesis 4 must be rejected. Of 24 coefficients relating geographic region (census divisions) to horizontal integration, only three are statistically significant, and hypothesis 8, predicting no relationship, is supported. With respect to hospital characteristics, only the horizontal strategy of merger is positively and significantly associated with size; therefore, hypothesis 10 is not supported. Hypothesis 12, predicting a positive association between case mix and horizontal strategies, also is not supported. Hypothesis 14, predicting a positive association between change in case mix and horizontal integration, is supported for multihospital system membership and for mergers. This constitutes tentative support for this linkage. The Journal of Rural Health zy zyxwvutsrqpo 110 VOl. 9, K O . 2 zyxwvuts zyxw zyxw Table 3. Summary of Cox’s Proportional Hazard Models, Strategic Management Activities, 1983-1988, Rural U.S.Hospitals, Signs and Probabilities of Coefficients at Alpha I 0.05. Close Vertical Integration/Diversification Horizontal Integration Group Purchasing MultiHospital Consortium Merger Nursing Home Swing Beds Ambulance ~~ sign prob Market-Environment Munificence Log population per square mile, 1983 Per capita income, 1983 % Medicare, 1983 % unemployment, 1983 Complexity Log miles nearest hospital MDs per 1,000 population, 1983 Dynamism Log change population per square mile, 1983-1987 Change per capita income, + 1983-1987 % change Medicare, 1983-1987 % change unemployment, 1981-1987 Change MDs per 1,000 population, 1983-1987 Geographic location East South Central West North Central Mountain West South Central South Atlantic New England Pacific Mid-Atlantic sign prob Hosp ita1 Log operating beds Case-mix index, 1983 Change case-mix index, 1983-1988 For profit Governmental + Not for profit, church + Total margin, 1983 Change total margin, 1983-1987 Number head administrators, 1983-1988 + sign prob sign prob sign prob sign prob + + 0.006 0.000 0.023 + 0.028 - 0.050 + 0.002 - 0.017 - 0.004 - 0.000 - - 0.034 + 0.032 0.049 0.031 + 0.039 0.035 + + + 0.043 + 0.006 0.008 + - 0.032 0.000 0.000 0.041 0.031 + 0.000 - 0.022 0.036 + + - sign prob 0.001 + + 0.047 0.033 0.005 - 0.021 0.001 - + 0.001 zyxwvutsrqpo zyxwvutsrqponm zyxwvutsrqp Probability Wald model chi square 0.029 0.000 0.001 0.003 0.168 0.305 0.000 Reference category for census divisions=East North Central. Reference category for hospital ownership variables=not for profit, non-church. Table 3 continued on next page. Mick, Morlock, Sulkever, de Lissovoy,Malitz, Wise,and Jones 111 Spring 1993 Table 3-continued from previous page. zyxwv Partial Vertical Integration/Diversification Unrelated Vertical Integration/Diversifica tion Outpatient Services Inside HMO Pro sign prob sign prob sign prob Other Health Services Non-health Services Outpahent Services Outside sign prob sign prob sign prob zyxwvutsrqp Market-Environment Munificence Log population per square mile, 1983 Per capita income, 1983 % Medicare, 1983 % unemployment, 1981 Complexity Log miles nearest hospital MDs/1,000 population, 1983 Dynamism Log change population per square mile, + 1983-1987 Change per capita income, 1983-1987 Change % Medicare, 1983-1987 Change % unemployment, 1981-1987 Change MDs per 1,000 population, 1983-1987 - - Probability Wald model chi square - + + + 0.024 0.041 Geographic location East South Central West North Central Mountain West South Central South Atlantic New England Pacific Mid-Atlantic Hospital Log operating beds Case mix index, 1983 Change case mix index, 1983-1988 For profit Governmental Not for profit, church Total margin, 1983 Change total margin, 1983-1987 Number head administrators, 1983-1988 0.010 0.012 0.004 0.000 0.000 0.019 0.016 0.005 0.010 0.002 + 0.012 0.045 0.000 Reference category for census divisions=East North Central. Reference category for hospital ownership variables=not for profit, non-church. hospitals with a maximum of 50 beds (Richardson & Kovner, 1986). Thus, there is some tentative support for the hypothesis. Hypotheses 13 and 15, case mix The Journal of Rural Health 0.021 - 0.017 - 0.000 - 0.004 + - 0.032 0.020 + 0.001 + 0.048 0.000 + 0.032 - 0.025 0.069 0.0642 0.0975 zy zyx and change in case mix, are not supported; together only three of 18 coefficients are statistically significant, although they are in the predicted direction. 112 Vol. 9, No. 2 Hypotheses 17,19, and 21, concerning for-profit, governmental, and not-for-profit religious ownership, respectively, are not supported. Predicted relationships with for-profit ownership hold in none of nine strategies; with respect to governmental ownership, four of nine associations are significant (twice in the predicted direction, twice in the opposite direction); and for not-for-profit religious ownership, two of nine are significant. Hypotheses 23 and 25, predicting no relationship between vertical integration-diversification and either total margin or change in total margin, are supported; no significant coefficients are registered. Finally, hypothesis 27, predicting a negative link between vertical integration-diversificationand head administrator turnover, is not supported, once again because of the absence of statistically significant coefficients. In sum, on the basis of 13 separate proportional hazards models, there is at best only tentative support for some of the hypotheses relating horizontal and vertical integration positively or negatively to environmental market and hospital variables. Where hypotheses of no relationship were specified, these were more often supported (e.g., geographic location and financial performance). A summary of whether empirical evidence supported the theoretical expectations is given in the third and sixth columns of Table 1. Specific Strategies. On the other hand, certain strategies are more strongly related to the specified model than are others. The horizontal strategies of group purchasing and multihospital system affiliation have at least five significant predictive variables. Among vertical strategies, swing-bed programs as well as HMO and PPO affiliations have more than five statistically significant predictors. Group purchasing is more likely for hospitals that are church-owned, have a high total margin at the baseline year, and have experienced head administrator turnover and are in environments with increasing per capita income. Location in the West North Central division has a depressing effect on group purchasing. Multihospital affiliation is positively associated with location in the East South Central division, an increasing case-mix index, status as either a church-owned or for-profit facility, and a higher level of head administrator turnover. It is negatively affected by an increase in population per square mile. Swing-bed programs are likely when the hospital is in a market with a low per capita income, a high percentage of Medicare enrollees, a high level of unemployment, or an increasing amount of unemployment and a decline in physicians per 1,000 popu- lation. Swing beds are more often found in the West North Central division, and less often in the West South Central and Mid-Atlantic divisions. Finally, as would be anticipated, smaller hospitals engage in this strategy. HMO affiliation is positively influenced by church ownership status. In seven of the eight census divisions, affiliation is depressed, but this is a function of the choice of the reference variable. HMO affiliation is also correlated with a smaller distance to the next nearest rural hospital. Finally, PPO affiliation occurs in markets where the proportion of Medicare eligibles is declining and in the Pacific division. Hospitals in the West North Central, West South Central, and Mid-Atlantic divisions are less likely to have PPO affiliations. Greater bed size and status as a church-owned facility are associated with PPOs. zy zyxwvutsr zyxwv Discussion General Growth of Strategies. Study findings indicate significant growth in the number of rural hospitals engaging in strategic activities during the 1983-88period. With the exception of group purchasing, at the beginning of the study period-the eve of PPS-none of these strategies was present in more than about 15 percent of all rural hospitals. The growth that did take place, however, was not substantial. Aside from group purchasing, all horizontal and vertical strategies were adopted by fewer than one half of all rural hospitals. Hence, it can be concluded that, despite the widespread urging of policymakers, professional associations, academics, and others, rural hospitals had not moved strongly into any but a few strategic activities. Weak Support for Directional Hypotheses. The evidence supporting Hypotheses 2 through 27 was mixed, with directional hypotheses supported weakly or not at all, and null hypotheses always supported (Table 1).Interestingly, the hypotheses concerning environmental and market effects were only slightly supported. In fact, a more straightforward conclusion is that environmental and market as well as geographic variables had little impact on horizontal and vertical integration strategies. Only in comparison to the environmental and geographic variables were hospital-related variables more frequently associated with strategies, but often in directions not predicted. Among specific strategies, some patterns zyxwvutsrqp Mick, Morlock, Salkever, de Lissovoy, Malitz, Wise, and Jones 113 Spring 1993 zy zyxwvut emerged: a nonmunificent environment (e.g., high proportion of Medicare beneficiaries) was associated with swing-bed adoption, a minimally disruptive activity geared to smaller facilities. But, in most other respects, munificence appeared to have no effect; the proportion of Medicare beneficiaries and changes in this proportion had little impact on most other rural hospital strategies. This pattern does not support the argument that PPS was harming rural hospitals, although this would be a very attenuated argument to make based on these findings. A corollary interpretation is that high potential Medicare loads did not, in these data, compel rural hospitals to adopt horizontal strategies, which is contrary to our theoretical predictions. . Geographic effects were rare. Considering the fact that we controlled for variables often implied in the use of geographic location, it can be concluded that a rural hospital’s location may be an overstated factor in understanding its behavior. Hospital characteristics yielded more significant effects, with operating bed size showing a tendency to increase the probability of several strategies. The same was partially the case for church-owned rural hospitals. Government run hospitals exhibited somewhat the opposite behavior. Higher and increasing case-mix indexes also influenced the adoption of some strategies: hospitals with patients who had more severe illness and injuries, often in larger facilities, were more likely to move in strategic directions. Profitability as measured by level of, and increases in, total margin had little impact in regard to the adoption of these strategies. weaknesses. However, significant associations with Area Resource File (ARF) data appear in social science and economic studies of all kinds, and these variables derive from a wide variety of data collection sources such as the U.S. Bureau of Labor Statistics and the American Medical Association. ]It IS improbable that measurement error is at fault. The use of the county as a proxy for the envircmment and market, although practical, is arbitrary. Morrisey, Sloan, and Valvona (1988), in a study of Nebraska, showed that hospital markets comprised an average of six counties and concluded that rural markets were not concentrated. But, McLaughlin, Normolle, Wolfe, MacMahon, & Griffith (1989) showed that small area variation use rates and their correlates differed only slightly whether counties versus more carefully defined market areas were used as units of analysis. This study concludes that counties as environmental and market boundaries are not unreasonable geographic units, although a finer-grained delineation of market areas might be desirable. Third, it is possible that examination of specific, albeit well understood, horizontal and vertical strategies misses important strategic clusters or associations among strategies. There may be underlying dimensions of rural hospital strategic behavior that are more sensitive to environmental market, geographic, and hospital variables. From preliminary work, it is believed there may be three such groups of strategies: a mix of horizontal and vertical integration strategies, managed care vertical integration, and multihospital affiliation combined with downsizing of selected services. Factor and cluster analyses are necessary to confirm these possibilities, and this work is in progress. A fourth possibility is that this study’s measwes of strategy are too global and that more specific service mix variables would be better indexes of rural hospital strategic activity and thus be more sensitive to the analysis performed in this paper. Examples include rehabilitation services like radiation, physical, and occupational therapy, either inpatient or home-based; high-technology imaging services; or a host of other specific activities. Analyses at this level of measurement are underway. Substantive Issues. The first of these concerns may stem from our assumption that the causal direction is one in which environmental-market, geographic, and hospital factors influence strategic: behavior. This may not always be the case. For example, because rural hospitals sometimes domi- zyxwvutsrq zyxwvut Lack of Environment-Market Linkage to Strategy Content. Not only does the theory we present argue in favor of environment-organization linkage, but also empirical studies such as that of Veney and Kahn (1973) strongly support the connection. Why have this study’s findings not been congruent with theory and previous research? Methodological Issues. This study may have omitted the variables that tap the most appropriate environmental and market characteristics. Yet, the variables used are standard and often powerful predictors of organizational change, and it seems unlikely that these variables are irrelevant as critical measures of the environment. Second, the variables used may be the correct ones, but are poorly measured either through faulty data collection processes or through poor delineation of relevant environmental and market boundaries. Both are plausible zyxwvutsrqponml zyxwvutsrqp zy The Journal of Rural Health 114 Vol. 9, No. 2 nate their markets as the single largest employer, they may themselves influence the environment. Thus, a strategically adapted rural hospital may improve environmental and market indicators rather, or in addition to, the converse. Indeed, part of the logic of the Hill-Burton Act of 1946 to build rural hospitals was to attract physicians to these areas. This study does not attempt to untangle this causal riddle primarily because proportional hazards models cannot accommodate a so-called ”endogeneity problem,” which is regarded as less important than the right censoring issue. It can also be argued that the problem of reciprocal causation is better confronted when factors related to rural hospital financial performance are assessed. Second, rural hospital horizontal and vertical strategic behavior may follow a logic different from that predicted by this study‘s mix of resource dependency and transaction cost theory. It is possible that rural hospitals engage in a kind of ”follow-theleader” behavior, adopting strategies for the most part because other rural and urban hospitals do. Or, rural hospitals may have been responding more to changing systemwide ”institutional” norms articulated by public and private health policy sources, and less to any particular market signals that might prompt horizontal and vertical integration. Such a perspective stems from institutional theory (Alexander & DAunno, 1990; Meyer & Rowan, 1977; Scott, 1987).Arndt and Bigelow have recently applied this line of reasoning to hospital vertical integration (1992). sive managerialism is quite beyond the means of most rural hospitals. Starkweather and Cook (1988) may be more accurate in depicting rural hospitals as organizations that can exist in deprived environments as much as anything else through various legitimizing activities-garnering community support and cohesion and engaging in other noneconomic activities like legislative lobbying. Because rural hospital size in terms of operating beds is associated positively with four of the strategic activities, it is reasonable to assume that the organizational resources available with increased size allow adoption of strategies independently of environmentmarket factors and geography, a pattern congruent with this argument. The reported research examined strategic management activity as a viable health policy avenue for rural hospitals. The perspective is borrowed almost wholesale from the business literature, and rural hospitals are somehow analogously compared to manufacturing and commercial firms. Much can be gained from such comparisons, as Carter (1990) has shown in her study of small physician PPOs in which proactive strategic moves were more effective than reactive strategies. Yet, damage can also result, especially if state and national policy operates on the premise that rural hospitals, without subsidization and by virtue of strategic activity, can engage in effective market-oriented behavior. Market forces have helped close 330 rural community hospitals between 1980 and 1989 (American Hospital Association, 1991),and notwithstanding arguments that access remains unaffected (Bronstein & Morrisey, 1991; Codman Research Group, Inc., 1990),there must be limits beyond which access to care becomes a genuine problem. Concern resulting from the findings of this study is that rural hospitals may not be effective market-oriented organizations with the necessary resources to engage in strategic behavior. The majority of the sample of rural hospitals did not engage in many strategies, although a minority of facilities did appear to move aggressively into such activities. This raises the question of whether the minority of strategically active hospitals improved their financial performance, a question addressed in a subsequent analysis. Thus, a cautionary note is necessary: more research is required before an unequivocal conclusion can be drawn. However, based on this study, it can be argued that health policy relying on individual rural hospital strategic adoption may not ensure a future network of viable rural hospitals (Note 7). zyxwvutsrq zyxwvutsrqp zyxwvu zy zyxwvutsrq Conclusion Despite these concerns, because it is believed that the strategies studied have face validity for the majority of rural administrators, that the cumulative proportions of rural hospitals engaging in them are relatively small, and that the analyses performed are reasonable within the context of the research problem, this study’s findings cannot easily be dismissed. As a class of organizations, it may be that rural hospitals define an outer boundary where strategic behavior is relatively rare and nearly random. This is the overriding conclusion from the findings displayed in Table 3. It may be that many rural environments are so nonsupportive compared to those of metropolitan areas, and the conditions of a majority of rural hospitals are so fragile, that strategic behavior practiced in the manner of Porter’s (1980) aggres- Mick, Morlock, Salkever, de Lissovoy, Malifz, Wise, and Jones 115 Spring 1993 Appendix zyxwvutsrq of operating acute care beds in the responding facility, the Health Care Financing Administration (HCFA) case-mix index measuring the severity of the patients treated, and the ownership status (not-forprofit community, not-for-profit church, for-profit or investor owned, and governmental (principally county, district, or municipal). This study used the not-for-profit community hospital as the reference category. For financial performance, it used total margin, a measure of profitability. Consideration was given to operating margin, but total margin was used instead because many of the strategic activities examined would not necessarily provide patient or other operating revenue. Finally, the study used the total number of head administrators that the rural hospital had since 1983 as its measure of hospital administrator turnover. Other variables such as distance of the responding hospital from the nearest Metropolitan Statistical Area, county population less than 25,000, status as a rural referral center, or hospital beds per 1,000 population were considered, but because of their high correlation with other variables such as population density or physicians per 1,000 population, they were dropped from the analysis to avoid collinearity problems. Munificence variables include population density, per capita income, percent Medicare enrollees, and percent unemployment. High population density (natural log of population per square mile) indicates greater potential demand for hospital services. The measure also taps a dimension of rurality, adjusting for rural areas in which the population is extremely dispersed versus more densely populated rural areas. The second variable is per capita income, a measure of potential demand for hospital services, for which there is wide variation across rural areas. The third variable is percent Medicare enrollees, an indicator that that shows wide variation across rural counties; a high value suggests lower munificence because of the restrictive payment scheme of PPS. The fourth variable is percent unemployment, an economic factor that varies widely and taps a number of dimensions such as strength of the local economy and burden of uncompensated care. This study used 1981 data as an approximation of the 1983 level because the ARF file did not contain the latter value. Complexity variables include physicians per 1,000 population and distance to the next nearest hospital (natural log miles). The number of competing hospital beds might seem a more logical choice than physicians per 1,000 population, but because this measure suffers from reporting error (e.g., subtracting the number of respondent hospital’s beds from the number of beds in a given county reported in the ARF data often results in negative beds) and correlates strongly with physicians per 1,000 population (r=0.434, P<O.OOl), this study used the physician measure. The distance measure derives from the telephone questionnaire and is the head administrator’s estimate of the miles to the next nearest hospital. Change variables selected to indicate market or environmental dynamism include changes in population density, per capita income, percent Medicare beneficiaries, percent unemployment, and physicians per 1,000 population. The second group of geographic variables are categories of the nine U.S. Bureau of the Census divisions. The East North Central census division was used as the reference category because its proportion of nonmetropolitan statistical area population most closely reflected the 1988 U.S. national average of 23.1 percent (Bureau of the Census, 1989). The rural hospital variables include the number zyxwvuts The Journal of Rural Health Notes Controversy exists regarding the impact PPS has had on rural hospitals. Some, like Cromwell(1989) argue that, except f i x certain intensive procedures, rural hospitals have not been generally underpaid. The General Accounting Office (GAD) (1991, p. 3) concluded that “losses on Medicare patients \$ere not a major factor underlying most closures.” On the other hand, studies of New York (Altman & Garfink, 1988), Texas (Bailey, Boff, & Rampmeier, 1988), Oklahoma (Oklahoma Medical Research Foundation, 1989), Kansas (David, Zeddies, Zimmerman, & McLean, 1990), and national data (Cleverley, 1988) show consistently declining operation and total margins for rural hospitals during the 1980s, always worse than urban hospitals. In another report, the GAO (1990) showed that financially distressed rural hospitals lost money on Medicare patients, but average losses on non-Medicare patients were even larger. This study concludes that, although other factors in addition to PPS have contributed to rural hospital problems, PPS was a major force defining the rural hospital environment. A copy of the telephone survey may be obtained by writing the primary author. The procedure omits hospital adoption of horizontal and vertical strategies before the study period, and hence the $data become ”left censored.” But, because there were only a few strategies (e.g., group purchasing, multihospital system membership, nursing home affiliation) that had been adopted by 15 percent or more of rural hospitals before PPS, left zyxw zyxwvut zyxwvutsrq zyx 116 Vol. 9,N o 2 censoring is not considered a major issue. In a separate analysis, characteristics of hospitals having multihospital affiliations before and after PPS were compared, and it was found that only two of the 26 comparisons were statistically signficant. 4. The county is the unit of observation in the national Area Resource File (ARF) database. Given the widespread use of the ARF in other studies of rural hospitals and given the enormous difficulties in determining actual hospital markets for nearly 800 rural hospitals, this study relied on the county as the most feasible approximation of a market. 5. Readers may consult Allison (1984)for technical details. Briefly, the technique relies on a baseline risk or hazard of adopting a given strategy, and a hospital’s risk of adoption of the strategy is in proportion to this baseline hazard. Each hospital’s risk is a function of the independent variables entered into the model. In reality, two variables are calculated when Cox’s procedure is used: the first is based on whether the hospital adopted a given strategy, and the second derives from the right censoring. Combining these produces a rate of adoption based on the number of years one observes a given hospital compared to the number of hospitals that adopt the strategy during the time period under study. As with other multivariate techniques, each coefficient calculated is the net of all others in the model, and the usual interpretation is applied (e.g., a positive coefficient means that the variable has a positive effect on the probability of adopting the strategy under consideration). 6 . The actual models for all 13 strategies may be obtained by writing the primary author. 7. The authors wish to thank John W. Seavey and six anonymous reviewers for helpful comments on an earlier version of this paper. The responsibility for the paper’s content, Bronstein, J.M., & Morrisey, M.A. (1991).Bypassing rural hospitals for obstetrics care. Journal ofHealth Politics, Policy and Law, 16, zyxwvutsrqponmlk 87-118. Buada, L., Pomeranz, W., & Rosenberg, S. (1985).Laying the groundwork Planning for change. Rural hospitals: Strategies for survival (Monograph No. 1).Kansas City, M O National Rural Health Association. Buada, L., Pomeranz. W., & Rosenberg, S. (1986).Horizontal and vertical integration management issues for rural hospitals. Rural hospitals: Strategies for sumival monograph series (Monograph No. 4). Kansas City, MO. National Rural Health Association. Bureau of the Census. (1989). Sfatistical abstract of the United States: 1989. Washington, DC: US.Government Printing Office. Carter, N.M. (1990). Small firm adaptation: Responses of physicians’ organizations to regulatory and competitive uncertainty. Academy ofManagement Journal, 33,307-333. Chakravarthy, B.S. (1986).Measuring strategic performance. Strategic Management Journal, 7,437-458. Christianson, J.B. (1989).Alternative delivery systems in rural areas. Health Services Research, 23,849-889. Christianson, J.B. (n.d.). Economic issues in the reduction of rural hospital capacity: A research summary (DHEW Publication No. lR03-H5-033745-01).Washington, DC: Department of Health, Education and Welfare. Clement, J.P. (1988).Vertical integration and diversification of acute care hospitals: Conceptual definitions. Hospital and Health Services Administration, 33,99-110. Cleverley, W.O. (1988).Hospital industry financial report, 1984-1988. Westchester, IL: Healthcare Financial Management Association. Codman Research Group, Inc. (1990). The relationship between declining use of rural hospitals and access to inpatient services for Medicare beneficiaries in rural areas. Washington, DC: Prospective Payment Assessment Commission. Congressional Budget Office. (1991). Rural hospitals and Medicare’s prospective payment system. Washington, DC. Author. Conrad, D.A., & Dowling W.L. (1990). Vertical integration in health services: Theory and managerial implications. Health Care Management Reviau, 15,9-22. Conrad, D.A., Mick, S.S., Madden, C.W., & Hoare, G. (1988) Vertical structures and control in health care markets: A conceptual framework and empirical review. Medical Care Review, 45,49-100. Cook, K., Shortell, S.M., Conrad, D.A., & Morrisey, M.A. (1983).A theory of organizational response to regulation: The case of hospitals. Academy of Management Review, 8,193-205. Cromwell, J . (1989). An analysis of the prospective payment system’s labor-nonlabor share by diagnosis-related group. Health Services Research, 24,213-236. Davis, R.G., Zeddies, T.C., Zimmerman, M.K., & McLean, R.A. (1990). Rural hospitals under PI’S A five-year study. Journal of Rural Health, 6,286-301. Dess, G.G., & Beard, D.W. (1984).Dimensions of organizational task environments. Administrative Science Quarterly, 29,52-73. DeVries, R.A. (1978, March). Strength in small numbers. Hospitals, pp. 81-84. DeVries, R.A. (1988, September/October ). When you’ve seen one, that’s all you’ve seen. Healthcare Forum Journal, pp. 38-39. Ermann, D.A. (1990).Rural health care: The future of the hospital. Medical Care Review, 47,33-73. Finkler, S.A., & Horowitz, S.L. (1985). Merger and consolidation: An overview of activity in healthcare organizations. Healthcare Financial Management, 39,19-28. zyxwvutsrqponml zyxwvutsr however, is the authors’ alone. References zyxwvutsrqpon Aldrich, H.E. (1979).Organizations and environments. Englewood Cliffs, NJ: Prentice-Hall. Alexander, J.A. (1990).Diversification behavior of multihospital systems: Patterns of change, 1983-1985.Hospital and Health Services Administration, 35,83-102. Alexander, J.A., & DAunno, T.A. (1990).Transformation of institutional environments: Perspectives on the corporatization of US. health care. In S.S. Mick & Associates (Eds.), Innovations in health care delivery: Insights for organization theory (pp. 53-85). San Francisco, C A Jossey-Bass. Allison, P.D. (1984). Event history analysis: Regression for longitudinal event data. Beverly Hills, CA: Sage Publications. Altman, S.H., & Garfink, C.M. (1990).How have PPS changes affected allocation of Medicare spending for hospital care? A case study of New York State. Inquiry, 27,319-331. American Hospital Association. (1991). Hospital closures 1980-1 990: Statistical profile. Chicago, IL: Author. Ansoff, H.I. (1979).Strategic Management. New York, NY: John Wiley & Sons. Arndt, M., & Bigelow, B. (1992). Vertical integration in hospitals: A framework for analysis. Medical Care Review, 49,93-115. Bailey, C.W., Boff, J.F., & Rampmeier, W.A. (1988).Report on Medicare inequities. Houston, TX: Texas Hospital Association. Mick, Morlock, Sulkever, de Lissovoy, Malitz, Wise, and Jones zyxw zyxwvuts zyxwvutsrq 117 Spring 1993 zyxwvutsrqpo zyxwvutsrqp zyxwvutsr zyxwvu zyxwvuts zy Fottler, M.D., Schermerhorn, J.R., Wong, J., & Money, W.H. (1982). Multi-institutional arrangements in health care: Review, analysis, and a proposal for future research. Academy of Management Rmiew, 7,67-79. General Accounting Office. (1991). Medicare: Further changes needed to reduce program and beneficiary costs. (GAO Publication No. GAO/HRD-91-67). Washington, DC:Author. General Accounting Office. (1990). Rural hospitals: Federal leadership and targeted programs needed. (GAO Publication No. GAO/ HRD-90-67). Washington, DC: Author. Hart, L.G., Amundson, B.A., & Rosenblatt, R.A. (1990). Is there a role for the small rural hospital? Journal of Rural Health, 6,101118. Hart, L.G., Robertson, D.G., Lishner, D.M., & Rosenblatt, R.A. (1992). CEO turnover in rural WAMI hospitals. Seattle, W A University of Washington, Department of Family Medicine, WAMI Rural Health Research Center. Herbert, T.T., & Deresky, H. (1987). Generic strategies: An empirical investigation of typology validation and strategy content. Strategic Management Journal, 8,135-147. Herzlinger, R., & Krasker, W.S. (1987). Who profits from nonprofits? Harvard Business Review, 65,93-106. James Bell Associates, Inc. (1991).On the move: Innovations in rural health care delivery. Rockville, M D Office of Rural Health Policy. Kimberly, J.R., & Zajac, E.J. (1985). Strategic adaptation in health care organizations: Implications for theory and research. Medical Care Review, 42,267-302. Lewis, B.L., & Parent, F.D. (1986).Acquisition of small rural hospitals by multihospital systems. Journal of Rural Health, 2, 55-65. Ludke, R.L., Westhoff, L.J., & Flood, B.M. (1992).Rural hospitals’ viability and strategic responses. Health Progress, 73,49-53,75. McLaughlin, C.G., Normolle, D.P., Wolfe, R.A., McMahon, L.F., & Griffith, J.R. (1989). Small-area variation in hospital discharge rates: Do socioeconomicvariables matter? Medical Care, 27, 507-521. Merlis, M. (1989).Rural hospitals. CRS report for Congress (Publication No. 89-296 EPW). Washington, DC: Library of Congress. Meyer, A. (1981).Reacting to surprises: Hospital strategy, structure, and ideology. Health Care Management Review, 6,2532. Meyer, J.W., & Rowan, B. (1977). Institutionalized organizations: Formal structure as myth and ceremony. American Iournal of Sociology, 83,340-363. Mick, S.S. (1990a).Explaining vertical integration in health care: An analysis and synthesis of transaction-cost economics and strategic-management theory. In S.S. Mick & Associates (Ed.), Innovations in health care delivery: Insights for organization theory (pp. 207-240). San Francisco, C A Jossey-Bass. Mick, S.S. (1990b).Themes, issues, and research avenues. In S.S. Mick & Associates (Ed.), Innovations in health care delivery: Insights for organization theory (pp. 1-19).San Francisco, C A Jossey-Bass. Mick, S.S., & Conrad, D.A. (1988). The decision to integrate vertically in health care organizations. Hospital and Health Services Administration, 33,345-360. Mick, S.S., & Morlock, L.L. (1990).America’s rural hospitals: A selective review of 1980s research. Journal of Rural Health, 6, 437-466. Miles, R.W., & Snow, C.C. (1978).Organization strategy, structure, and process. New York, N Y McGraw-Hill. Morrisey, M.A., Sloan, F.A., & Valvona, J. (1988). Defining geographic markets for hospital care. Law and Contemporary Problems, 51,165-194. The Journal of Rural Health Moscovice, I. (1989).A literature synthesis and health services agenda. Health Services Research, 23,892-930. Moscovice, I., Johnson, J., Finch, M., Grogan, C. & Kralewski, J. (1991).The structure and characteristics of rural hospital consortia. Journal of Rural Health, 7,575-588. Office of Technology Assessment. (1990). Health care in rural America, (OTA Publicaton No. OTA-H-434). Washington, Dc: US.Government Printing Office. Oklahoma Medical Research Foundation. (1989).Fiscal health qf Oklahoma hospitals in 1987. Tulsa, OK: Center for Health Policy Studies, Oklahoma Medical Research Foundation. Pfeffer, J., & Salancik, G.R. (1978).The external control of organiurtions: A resource dependence perspective. New York, NY: Harper & Row, Publishers. Porter, M.E. (1980).Competitive strategy: Techniques for analyzing industries and competitors. New York, Ny: The Free Press. Provan, K.G. (1984).Interorganizational cooperation and decision making autonomy in a consortium multihospital system. Academy of Management Review, 9,494-504. Rice, T., de Lissovoy, G., Gabel, J., & Ermann, D. (1985). The state of PPOs: Results from a national survey. Health Affairs, 4, LS40. Richardson, H., & Kovner, A.R. (1986).Implementing swing-lbed services in small rural hospitals. Journal of Rural Health, 2,4660. Rosenberg, S. (1989). The changing rural health care delivery system. Development of alternative models. Kansas City, M O National Rural Health Association. Rowland, H.S., & Rowland, B.L. (1984a). Hospital administration handbook. Rockville, MD: Aspen Systems Corporation. Rowland, H.S., & Rowland, B.L. (1984b). Hospital managemenf:A guide to departments. Rockville, M D Aspen Systems Corparation. Scott, W.R. (1987).The adolescence of institutional theory: Problems and potential for organizational analysis. Adrnin istrative Science Quarterly, 32,493-512. Shortell, S.M., Morrison, E.M., & Friedman, B. (1990). Strategic choices for America‘s hospitals: Managing change in turbulent times. San Francisco, CA: Jossey-Bass. Shortell, S.M., & Zajac, E.J. (1990).Health care organizations and the development of the strategic-management perspective. In S.S. Mick & Associates (Ed.), Innovations in health care d e l i z q : Insights for organization theory (pp. 144-180),San Francisco,,C A Jossey-Bass. Smith, H.L., & Piland, N.F. (1990). Strategic adaptations to PPSby rural hospitals: Implications for theory and research. jourrsnl of Rural Health, 6, 140-160. Starkweather, D.B. (1971). Health facility mergers: Some conceptualizations. Medical Care, 9,468-478. Starkweather, D.B. (1972).Beyond semantics of multihospital aggregations. Health Services Research, 20,58-61. Starkweather, D.B., & Carman, J.M. (1987). Horizontal and vertical concentrations in the evolution of hospital competition. In R. Scheffler & L. Rossiter (Eds.), Advances in health economics ,and health services research: Vol. 7. Mergers in health care: The performance of multi-institutional systems (pp. 179-194). Greenwich, CT.: JAI Press. Starkweather, D.B., & Cook, K.S. (1988). Organization-environment relations. In S.M. Shortell & A.D. Kaluzny (Eds.), Health care management: A text in organization theoy and behavior, (2nd ed.) (pp. 344-378). New York, N Y John Wiley & Sons. Staw, B.M., & Szwajkowski, E. (1975). The scarcity-munificence component of organizational environments and the commission of illegal acts. Administrative Science Quarterly, 12, 345354. zyx zyx zyxwvuts zyxwvutsrqpo 118 vor. 9, N ~2 . zyxwvutsrqpo zyxwvutsrqpo zyxwvutsrqp Thompson, J.D. (1967).Organizations in action: Social science bases of administrative theory. New York, NY McGraw-Hill. 'Topping, S., & Hernandez, S.R. (1991).Health care strategy research, 1985-1990:A critical review. Medical Cave Review, 48, 47-89. Veney, J.E., & Kahn, J. (1973).Causal paths in the elaboration of organizational structure: A case of hospital services. Health Mick, Morlock, Salkever, de Lissovoy, Malitz, Wise, and Jones Services Research, 8, 139-150. Weiss, S.L., Phillips, D.F., & Schuman, J.G. (1986). Management issues for rural hospitals. Chicago:, IL American Hospital Publishing, Inc. Williamson, O.E. (1975). Markets and hierarchies: Analysis and antitrust implications. New York, Ny: Free Press. Zuckerman, H.S. (1979).Multi-institutional systems: Promise and performance. Inquiry, 16,291-314. zyxwvutsrq 119 Spring 1993