J Med Syst (2007) 31:365–374 DOI 10.1007/s10916-007-9078-8

ORIGINAL PAPER

Factors Influencing Community Health Centers’ Efficiency: A Latent Growth Curve Modeling Approach Shriram Marathe & Thomas T. H. Wan & Jackie Zhang & Kevin Sherin

Received: 27 April 2007 / Accepted: 28 June 2007 / Published online: 25 July 2007 # Springer Science + Business Media, LLC 2007

Abstract The objective of study is to examine factors affecting the variation in technical and cost efficiency of community health centers (CHCs). A panel study design was formulated to examine the relationships among the contextual, organizational structural, and performance variables. Data Envelopment Analysis (DEA) of technical efficiency and latent growth curve modeling of multi-wave technical and cost efficiency were performed. Regardless of the efficiency measures, CHC efficiency was influenced more by contextual factors than organizational structural factors. The study confirms the independent and additive influences of contextual and organizational predictors on efficiency. The change in CHC technical efficiency positively affects the change in CHC cost efficiency. The practical implication of this finding is that healthcare managers can simultaneously optimize both technical and cost efficiency through appropriate use of inputs to generate optimal outputs. An innovative solution is to employ decision support software to prepare an expert system to assist poorly performing CHCs to achieve better cost efficiency through optimizing technical efficiency. Keywords Technical efficiency . Cost efficiency . Community health centers . Growth curve modeling Introduction The Community Health Center (CHC) program began in the mid-1960s as part of President Lyndon Johnson’s War S. Marathe : T. T. H. Wan (*) : J. Zhang : K. Sherin Public Affairs Doctoral Program, College of Health and Public Affairs, University of Central Florida, 3280 Progress Drive, Orlando, FL 32826, USA e-mail: [email protected]

on Poverty and evolved from a simple, two-neighborhood health center demonstration project into a complex system with over 3,000 clinic sites comprising of community and migrant health centers, primary care programs for public housing residents, and programs of health care for the homeless [1, 2]. The CHC program provides a unique model of health care that couples traditional primary care services with preventive and enabling (support) services. CHCs, primarily serve the areas where economic, geographic, or cultural barriers limit access to primary health care for a substantial portion of the population. The uninsured patients served by health centers have been shown to be less likely to postpone seeking care and more likely to receive counseling on health issues than patients who seek care in other health care settings [3].Throughout its 40-year history, the health center program has focused on certain key goals and features: to reduce disparities in health care, to involve the community in providing services and management through community participation on health center boards, and to provide universal access to high-quality health care. As of 2001, CHCs had provided preventive and primary care to one-fifth of the nation’s underserved [4]. Organizational analysis of CHC efficiency is needed to identify the effects of organizational structure (design) characteristics and contextual (environmental) characteristics on CHC efficiency. Scientific knowledge gained from analysis of organizational efficiency can guide performance improvement. The major research questions for the study are: Q1 Are there any patterns and trends of technical and cost efficiency observed among CHCs over a period of 5 years? Q2 What are the predictors of technical and cost efficiency observed among CHCs over a period of 5 years?

366

J Med Syst (2007) 31:365–374

Q3 Does technical efficiency positively affect cost efficiency?

Related studies Despite success in providing primary care, many CHCs are of insolvency [5]. During years 1998–2000, the Health Resources and Services Administration (HRSA), a division of Health and Human Services (HHS), reported that about 10% of all health centers were in major financial difficulties [6]. The increase in the need to care for the uninsured, homeless and immigrants have raised the cost of providing services [7]. More than half of CHCs reported deficits every year between 1997 and 1999 [5]. The factors contributing to poor financial performance included inadequate management, the burden of the uninsured, increasingly competitive health care markets, and insufficient funding. Examples of poor management practices included the inability to control expenditures, unfavorable contracts with other providers and managed care organizations, inappropriate or inadequate responses to market changes, and ineffective business operations [6]. Some studies have concluded that CHCs are less costly as compared to other providers [8, 9]. The GAO (1976) study of CHCs found inefficiencies including overstaffing given the number of patients [10]. Brecher and Forman [11] compared costs of nine CHCs to those of private, for-profit group practices and found that CHCs had higher expenditures for their non-medical staff, which contributed to overall cost increase. In contrast, some studies found the cost of care per patient provided by CHCs to be less than that of other providers [12–15]. Since 1994, integration among CHCs and between CHCs and other safety-net providers has been facilitated through the Federal Integrated Services Development Initiative (ISDI). Expected outcomes for the resulting networks included, cost efficiencies, economies of scale, sharing of expertise and staff among collaborators, and a “value-added aspect of higher performance” in areas such as revenue, staff utilization, and data capture [16]. Ortiz, Fottler, and Hofler, (2005) found that as of 2000, approximately 36% of all health centers were in ISDIs and that there was no evidence of cost efficiencies or higher performance in staff utilization in ISDIs network member CHCs [17]. Forty percent of all CHC patients are “self pay” and are likely to be uninsured [18]. The increasing numbers of uninsured are likely to increase the demand for CHC services and compel CHCs to provide more charity care. To maintain financial viability, CHCs might provide fewer enabling services. However, McAlearney (2002), in his

study of CHCs from 1996–1999, found the opposite: more centers had increased the number of enabling services they offered. A large portion of the total CHC revenue comes from Medicaid [5]. It represented 34.6% of the total revenue in 1997 [18]. Between September, 1998 and January, 2000 the GAO conducted an exploratory study of CHCs, using health center data, interviews and case studies. The GAO study found that successful centers adapt to changes in the health environment and that the contributing factors to success were: integration into networks, participating in managed care, acquiring JCAHO accreditation, having patients with diverse payment sources, private donations, and strong billing and collections systems [6]. A number of earlier case studies found CHCs to have weathered the changes in the health care market [19]. However, conclusions from these case studies may not be generalizable to other CHCs. The weaknesses of previous research include absence of theoretical specifications in identifying characteristics of high-performance CHCs, lack of methodological rigor and longitudinal analysis of CHC performance, and lack of guidance to improve performance.

The context-design-performance framework The study uses the context-design-performance (CDP) framework that allows observation and measurement of the interrelationships among a health center’s environment (context), organizational structure (design), and performance. CDP is derived as a strategic adaptation of the contingency theory (Fig. 1). A longitudinal, panel design can ascertain the causal influences of contextual and organizational factors on CHC performance. This study fills the gap of absence of empirical longitudinal studies examining CHC efficiency patterns and trends with their predictors that assess CHC efficiency.

Context % Medicare % Medicaid Percentage of % Hispanics Region Of Rurality CHC Crude Mortality Rate Population/MD ratio County

Design/ Organizational Structure Staffing Mix Payer Mix Integration/network

Performance

Q1

Technical Efficiency (InputTotal Cost, OutputsProvider Encounters) Q3

Q2 Cost Efficiency (an inverse ratio of total cost/encounters)

Fig. 1 The context-design-performance framework for assessing CHC performance

J Med Syst (2007) 31:365–374

367

Contextual factors contributing to efficiency Variables representing environmental characteristics of CHCs are treated as contextual variables constituting multiple contingencies within organizations that can either facilitate or impede their performance. The environmental characteristics should include poverty, physician–population ratio, birth rate, uninsurance, crude mortality rate, minority population, region, and rurality. These variables are important indicators of the demand for care. For instance, minority populations are more likely to be poor and to endure poverty-related conditions such as chronic illnesses, poor health behavior, and stress. Organizational structural factors contributing to efficiency Organizational variables are often considered as design factors. They include: (1) size, i.e. the number of medical, dental, mental health, and substance abuse, and other professional FTEs; (2) staffing, i.e. administrative, facility, and patient services support non-professional staff FTEs expressed as a percentage of total FTEs; (3) payer mix, i.e. financial resources expressed as a percentage obtained by dividing grant revenue by total revenue; and (4) integration, i.e. participation in a network funded by the Integrated Services Development Initiative. Relationships of CHCs’ contextual and organizational factors to efficiency In this study, the structural relationships of contextual and organizational (design) factors to CHCs’ efficiency are

Table 1 Variables and operational definitions

The variables are measured for each of the 5 years (2000–2004)

Variable Context Medicare Poverty Physicians Minority Region Rurality–Location Design Size (of staff) Staff Mix Integration Financial resources Federal grants Total revenue Performance Cost efficiency Technical efficiency

examined by both cross-sectional and longitudinal analyses. Design factors such as management style, leadership, strategic activities, care coordination, and other centerbased patient care activities were not available for inclusion in the study.

Materials and methods The study is a non-experimental panel study of 493 CHCs, with repeated measures of efficiency indicators for 5 years (N=2,465 observations). Data on all variables are compiled from the administrative data system for the period between 2000 and 2004. The CHC data file was merged with the Area Resource File to constitute a comprehensive research data set for exploring the relationships among the contextual, organizational structural (design), and performance variables. Analytical techniques included data envelopment analysis (DEA) of technical efficiency, and latent growth curve modeling of multi-wave technical efficiency and cost efficiency. The classification of study variables and their measurements Table 1 shows the measurement variables classified into the contextual/ environmental factors, design/organizational structural factors, and CHC efficiency performance. In this study, CHC efficiency is measured by two indicators of efficiency: technical efficiency and cost efficiency. This study assumes that CHCs attempt to maximize the number of patient encounters. Technical efficiency is calculated as a

Operational definition

Percentage of the population that is Medicare eligible Percentage of the population that is at or below 200% poverty level Number of physicians per thousand population Percentage of population that is African American; percentage that is Hispanic US Census Region (1. Northeast, 2. Midwest, 3. South and 4. West) Urban/rural location Number of physicians + NPs + PAs Size/total staff Member of an ISDI network (1=member; 0=non-member) Dollars revenue expressed as a percentage of total revenue Total revenue in dollars Cost per encounter Number of encounters (physician visits and PA/NP visits) per total FTEs for three medical staff groups (physicians, PAs, and NPs)

368

ratio of the number of encounters relative to three categories of clinicians/care providers: physicians, physician assistants (PA), and nurse practitioners (NP). Technical efficiency is the extent to which a CHC is able to achieve the maximum possible number of encounters. Cost efficiency is computed by an inverse ratio of the total cost of CHC operations divided by the total number of encounters. Analytical methods Analytical techniques include data envelopment analysis (DEA) of technical efficiency, and latent growth curve modeling of multi-wave performance indicators of technical efficiency and cost efficiency. Data envelopment analysis (DEA) The complex concept of Efficiency consists of cost, process and technical aspects of production [20]. Data envelopment analysis (DEA) is a nonparametric technique requiring no presupposition regarding the form of production [21, 22]. In order to measure technical efficiency for each CHC, DEA uses a linear programming (LP) algorithm to calculate the radial distance from the actual production position of the CHC to the (fully efficient) position of that same CHC on the efficient production function curve. The availability of multiple waves of CHC performance data enables specification of both input- and outputoriented DEA scores and measurement of relative efficiency by using the ratio of weighted sum of outputs to weighted sum of inputs [23, 24]. This study used an input oriented model with the assumption that CHCs have more control over inputs (resources) than outputs (provider encounters). DEA uses the frontier approach to estimate technical efficiency of decision making units abbreviated as DMUs. DEA calculates the maximum relative efficiency score of each decision-making unit (DMU) [22]. DMUs assigned an efficiency score of unity are deemed technically efficient in comparison to their peers [25]. Inefficient DMUs score between zero and one. Theoretically, the technically inefficient DMUs need more inputs to produce the same output in comparison to their more efficient counterparts [25]. Efficiency scores, whose values depend on the choice of peers, are relative and not absolute [25]. DEA can be used not only to ascertain the relative efficiency of scores but also to judge which inputs are used or outputs produced technically inefficiently; which in turn can guide performance improvement of inefficient units [25]. DEA incorporates multiple outputs and inputs and can account for the multidimensional character of production by healthcare entities such as hospitals [20]. CHCs, like hospitals, are also multidimensional production facilities. DEA examines

J Med Syst (2007) 31:365–374

how resources (supplies, labor, and capital) produce a variety of outputs (research, teaching, patient care in a hospital setting) [20]. In this study, the potential for biased efficiency scores due to nonlinear relationships between service production (provider encounters), and the relative nature of efficiency scores obtained based on only the CHCs in the sample, led to the choice of the more conservative variable return to scale (VRS) model. The limitations of DEA include the influence of measurement error on the shape and position of the frontier and the potential for bias due to exclusion of an important input or output. Latent growth curve modeling The dynamic relationships among multiple causes and effects of CHC efficiency cannot be adequately explained by conventional regression methods and are best assessed by growth curve modeling [22]. Growth curve modeling is used for the following reasons: (1) the means, variances, and covariances of repeated measures of a continuous variable can be investigated by latent growth curve modeling; (2) random coefficients are used to capture individual CHC differences in the initial observation period and the growth trend; (3) both time-constant and timevarying covariates can be included as predictors or control variables for an endogenous variable; and (4) the change patterns of CHCs’ efficiency over the time span of 5 years can be delineated so that we can test the contribution of multiple concomitant factors to efficiency. The latent growth curve model can be extended to include parallel processing factors when investigating change trajectories (patterns and trends) of performance. In this analysis, only statistically significant results (p