Team Performance and Risk-Adjusted Health Outcomes in the Program of All-Inclusive Care for the Elderly (PACE)

The Gerontologist Vol. 46, No. 2, 227 237 Copyright 2006 by The Gerontological Society of America Team Performance and Risk-Adjusted Health Outcomes in the Program of All-Inclusive Care for the Elderly (PACE) Dana B. Mukamel, PhD, 1 Helena Temkin-Greener, PhD, 2 Rachel Delavan, MS, 2 Derick R. Peterson, PhD, 2 Diane Gross, PhD, 2 Stephen Kunitz, MD, 2 and T. Franklin Williams, MD 2 Purpose: The Program of All-Inclusive Care for the Elderly (PACE) is a community-based program providing primary, acute, and long-term care to frail elderly individuals. A central component of the PACE model is the interdisciplinary care team, which includes both professionals and non-professionals. The purpose of this study was to examine the association between the team s overall performance and the risk-adjusted health outcomes of program enrollees. Design and Methods: The study included interdisciplinary teams in 26 PACE programs and 3,401 individuals enrolled in them. We combined information about individuals health, functional, and mental status from DataPACE with an overall measure of team performance. We used multivariate regression techniques to test the hypothesis that better team performance is associated with better risk-adjusted health outcomes: survival and short-term (within 3 months of enrollment) and long-term (within 12 months of enrollment) improvements in functional status and in urinary incontinence. Results: Team performance was significantly associated with better functional outcomes (both short and long term) and with better long-term urinary incontinence outcomes. There was no significant association with survival. Implications: This study provides empirical evidence for the relationship between team performance and patient outcomes in long-term care. It suggests that PACE programs can improve patient outcomes by improving the functioning of care teams. We gratefully acknowledge funding from the National Institute on Aging, Grant RO1AG17555. Address correspondence to Dana B. Mukamel, PhD, University of California, Irvine, 111 Academy Way, Suite 220, Irvine, CA 92697. E-mail: dmukamel@uci.edu 1 University of California, Irvine. 2 University of Rochester, NY. Key Words: Interdisciplinary teams, Risk-adjusted health outcomes, PACE, Long-term care, Managed care Long-term-care patients have multiple needs, requiring a complex set of services provided by many individuals with different training. There is a general perception among many health care providers and health policy makers that interdisciplinary teams are better able to coordinate and provide such services, resulting in better health care and outcomes (Heinemann & Zeiss, 2002; Wagner, 2004). Several studies provide empirical evidence to support this expectation (Weiland, Kramer, Waite, & Rubenstein, 1995). For example, Sommers, Marton, Barbaccia, and Randolph (2000) examined the effect of a physician, nurse, and social worker team for community-dwelling seniors with chronic diseases. They showed that, compared with a control group receiving regular care, individuals treated by the team had fewer hospitalizations and rehospitalizations, fewer physician visits, and increased social activity. Similarly, studies have found that teams are important in improving care in nursing homes. Rantz and colleagues (2004) examined 92 nursing homes in Missouri and attributed better patient outcomes in a subset of these facilities to, among other factors, the use of team and group processes. Yeatts, Cready, Ray, DeWitt, and Queen (2004) offered qualitative information from a pilot study of nurse aide teams in nursing homes, suggesting that teams facilitated improved interaction and communication. Two meta-analyses of studies of teams (Stuck, Siu, Wieland, Adams, & Rubenstein, 1993; Wieland, Stuck, Siu, Adams, & Rubenstein, 1995) found that, although individual studies may not always show improved outcomes for patients, when data are combined across studies, patients treated by Vol. 46, No. 2, 2006 227

teams have better survival, functional, and cognitive outcomes, and lower institutionalization rates. The expectation that teams provide better care is a cornerstone of the Program of All-Inclusive Care for the Elderly (PACE). This innovative managedcare program, which combines primary, acute, and long-term care in terms of both services and financing, is designed around the interdisciplinary team concept. In PACE, individual care plans for each patient are developed and implemented by a team that includes a primary care physician, nurses, social workers, therapists (physical, occupational), aides, nutritionists, and, in some instances, a pharmacist, a chaplain, and others who come in contact with the patient. As this article will discuss, the PACE team is more extensive and comprehensive than most teams found in medical care settings, and, because payments to PACE are capitated, the programs have more control and latitude in directing care and tailoring it to the needs of each individual patient. (In this article, when we refer to the PACE program, we refer to the PACE model of care. When we refer to the PACE programs, we refer to the individual PACE organizations that provide care based on the PACE model but are independent organizations. Each of these PACE programs may have several sites at which care is provided.) Most prior studies have examined the impact of teams by comparing outcomes for patients who are cared for by teams with patients who are not. They have not examined variations across teams. Teams may function differently, and, even when structured similarly, some may perform better than others. The latter issue is the focus of this study, which (a) examines the variation in the way PACE teams perform, and (b) investigates the association between an overall measure of team performance and several specific risk-adjusted health outcomes of the individuals who are cared for by those team members. The measure of overall team performance is based on an evaluation made by the team members of their team (see further discussion in the Data Sources section). This measure has been shown to be associated with better team processes including communication, leadership, and conflict resolution in prior studies conducted in both acute care and long-term care settings (Shortell, Rousseau, Gillies, Devers, & Simons, 1991; Shortell et al., 2004; Temkin-Greener, Gross, Kunitz, & Mukamel, 2004). In this article, we examine whether it is also associated with better individual patient outcomes. Description of the PACE Program PACE is a community-based managed care program for frail, chronically ill older individuals whose significant functional and cognitive impairments make them eligible for nursing home care. The objective of the program is to enable individuals to continue living independently in the community at the highest possible functional level and with the best quality of life. The program is responsible for all of its participants health care needs, ranging from primary to acute to long-term care. To this end, it receives both Medicare and Medicaid capitation funding, and it has the ability to use these resources creatively, tailoring services to the needs of individuals, often in ways not possible under usual Medicare and Medicaid rules. At the core of the program is the adult day care center, augmented by home care and meals at home. An interdisciplinary team develops, carries out, and monitors individual care plans. Today there are 40 PACE sites serving more than 12,000 individuals in 21 states across the country (for more details about the program, see Bodenhiemer, 1999; Eng, Pedulla, Eleazer, McCann, & Fox, 1997; Gross, Temkin-Greener, Kunitz, & Mukamel, 2004). Teams in PACE The PACE model of care centers on the team concept. Federal regulations mandate the existence of PACE teams and set forth the criteria for their composition and organization, as well as the roles and responsibilities of their members. At a minimum, PACE teams must include: primary care physicians, registered nurses, social workers, physical and occupational therapists, recreational therapists or activity coordinators, dieticians, PACE day center coordinators, home care coordinators, personal care attendants (or their representative), and drivers (or their representative). Many PACE sites choose to include in their teams others who come into direct contact with the patient (e.g., pharmacists or chaplains). As such, these teams are large and can include more than 50 staff (although teams of 10 interact in smaller groups at the weekly team meetings described in the next paragraph). This is in contrast to the typical teams one encounters in medical care settings, which usually have fewer than 10 members. Furthermore, most typical teams in medical care settings are composed of professionals (i.e., physicians, nurses, social workers, therapists), whereas the PACE teams include paraprofessionals as well (i.e., aides and drivers). The PACE philosophy recognizes that often it is these care providers who are the most intimately acquainted with patients and their families and are therefore most knowledgeable about their needs and the barriers and difficulties they face. Thus, their contribution to understanding the needs of the patient and to developing an optimal care plan is valued and is considered central to the treatment process. The interactions between team members in PACE are both formal and informal. The formal interdisciplinary team interactions occur at the weekly team meetings, during which team members discuss 228 The Gerontologist

individual participants care needs. The objective of these meetings is to review the status of individuals who may be experiencing a particular medical crisis (e.g., hospitalization, placement in a nursing home) or who are due for a health status reassessment, and then to adjust their care plans appropriately. Therefore, attendance at these meetings is determined by which participants are being discussed: Only those team members caring for the participants whose care is being discussed will attend a particular meeting. Team members who care for all participants (e.g., the physician, the director of nursing, the physical therapist if there is only one on staff) will, by necessity, attend all meetings. However, staff members who care for a small subset of individuals (e.g., aides, drivers, and therapists, if there are more than one) will attend only if their participant is on the agenda. This fluid meeting composition ensures that all staff who provide direct patient care participate in these meetings on a regular basis. In addition to formal interaction through weekly staff meetings, staff interact with each other informally on all issues related to patient care. The weekly team meeting is typically run by a member of the team who is considered to be a skillful facilitator and who may have received facilitator training. The identity of the facilitator varies across, as well as within, programs, when a program operates multiple locations. The facilitator makes sure that all represented disciplines provide input and that the information presented about each participant is comprehensive, thus ensuring that the care plan addresses all aspects related to services that the individual requires. The large size of the PACE team presents particular challenges and might make team operations more complex, relative to a smaller team. So although all PACE programs share the team concept and regard it as central and important to the care that they provide, the implementation of the concept and the integration of team processes into day-today operations can vary substantially. If the teams are indeed an important factor influencing the quality of the care that PACE participants experience, then we would expect that persons enrolled in programs with better functioning teams will experience better health outcomes. Quality of Care in PACE Quality of care in general is a complex, multidimensional concept (Mukamel & Spector, 2003). It is particularly complex in a program such as PACE, which provides a wide gamut of health and supportive services (from primary to acute to longterm care) and that influences all aspects of life for its participants. Furthermore, because different aspects of quality do not seem to be correlated (Mukamel et al., 2004), assessment of quality requires an examination of different dimensions of care. In this study, we focused on two health outcomes that are important determinants of an individual s ability to maintain independent living in the community a central goal of PACE. These were deterioration in functional status as measured by increasing limitations of activities of daily living (ADLs), and urinary incontinence (UI). ADLs (bathing, dressing, grooming, toileting, transferring, walking, and feeding) provide a comprehensive measure of the ability of an individual to function independently. We examined the impact of teams on these health outcomes both in the short term (at 3 months post enrollment in the program) and in the long term (at 12 months post enrollment). This multiple timeframe approach was motivated by the possibility that different processes of care are involved over the course of enrollment in PACE. Short-term outcomes reflect the team s efforts to right the wrongs of the regular care system (i.e., to address any health issues that may have gone unattended prior to the individual s enrollment in PACE). Once the team would have optimized the condition of the enrollee, further efforts would have focused on long-term maintenance. It is possible that team efforts may have a different impact in these different timeframes. We also examined the impact of teams on mortality. The goal of the PACE program is to improve participants quality of life, not to increase survival at the expense of quality of life (in compliance with the participant s wishes). Therefore, we do not expect to find better survival rates for participants who are enrolled in programs with better teams. Methods Sample The study included new enrollees in 26 PACE sites who joined the program during a 2-year period between July 1998 and June 2001 (the dates differed somewhat across programs). We limited the sample to new enrollees in order to obtain baseline scores for the risk factors at the time of admission to PACE so that we could evaluate changes in outcomes at 3 and 12 months post enrollment. The initial sample included 3,401 new enrollees. For the mortality analyses, we excluded enrollees who (a) did not have a health and functional assessment completed within 31 days of enrollment, (b) had chemotherapy treatment without a cancer diagnosis, or (c) had more than three missing risk factors. Analysis of the pattern of missing data suggested that data were missing randomly, and, therefore, exclusion of these records was not likely to introduce a bias. Self-assessed health variables were not missing randomly, and, therefore, we did not delete records with missing self-assessed health data, as will be described below. The remaining 2,747 new enrollees were retained for the mortality analysis. Vol. 46, No. 2, 2006 229

For the ADL limitations and incontinence analyses, which were based on outcomes at 3 and 12 months post enrollment, we further restricted the sample to include only individuals who had been observed for at least 12 months (i.e., those enrolled in the second year could not be included). We also excluded enrollees with catheter, renal failure, or dialysis at the time of enrollment from the analyses of incontinence outcomes. The sample sizes for the analyses were 1,856 individuals for the 3-month ADL analysis, 1,871 for the 12-month ADL analysis, 1,790 for the 3-month incontinence analysis, and 1,775 for the 12-month incontinence analysis. Data Sources We obtained data from several sources. We obtained individual risk factors at admission and hospitalization data from DataPACE. DataPACE is an individual-level, administrative database that contains information about each participant at the time of enrollment and at specified periods post enrollment. All sites use the same variables definitions, and all receive training in data collection (Mukamel, Temkin-Greener, & Clark, 1998). This data set includes information about enrollees demographics, socioeconomic status, health status and disability, medical history, utilization of health services, and date of death. We used these data to create variables measuring health outcomes and individual risk factors. We obtained information about staffing levels (full-time equivalent [FTEs] per 100 participants) from the National PACE Association (2002). In order to measure team performance, we surveyed all team members in 26 PACE sites. In line with the PACE philosophy and work organization, we considered all staff with direct care responsibilities to be members of the team and therefore included them in the survey. Team members were identified by each PACE site and included both professionals and non-professionals (e.g., physicians, nurses, therapists, nutritionists, aides, and others if the PACE site considered them to have direct care responsibilities). The mailed survey administered a validated team performance tool that was developed and tested in the PACE program (Temkin-Greener et al., 2004). Everyone identified as a team member was asked to indicate his or her degree of agreement with the following seven statements: (a) Our team does a good job in meeting family member needs; (b) All team members contribute based on their experience and expertise to produce a quality solution; (c) Our team does a good job in meeting patient care needs; (d) Our team responds well to emergencies; (e) Our team almost always meets its patients care needs; (f) Although there are a variety of patients, our team s outcomes are very good; and (g) Overall, our team functions very well together. We measured the responses on a 5-point Likert scale (1 = strongly disagree, to 5 = strongly agree). We constructed the measure as the average of the responses of all team members, with team performance improving as the average score increased. As reported by Temkin-Greener and colleagues (2004), this measure has good psychometric properties (with a standardized Cronbach s a =.89) and good discrimination (with significantly less variability [p,.0001] within teams than across teams). It also meets criterion-related validity because it is highly correlated with team processes (namely leadership, communication, conflict resolution, and coordination) that are likely to lead to better performance. In order to increase the response rate among aides, we offered a $10 remuneration and an 800 telephone number to answer questions. We conducted the surveys between June 2001 and September 2001. We administered surveys to all 1,860 direct care, full-, and part-time staff members, resulting in a 65% response rate (1,209 responses). Dependent Variables By using DataPACE, we constructed five health outcomes: time from enrollment to death; decline in functional status at 3 and 12 months post enrollment (defined as change in number of ADL limitations between enrollment and 3 months post enrollment, and between enrollment and 12 months post enrollment); and deterioration in UI at 3 and 12 months post enrollment (defined as UI levels at 3 and 12 months post enrollment, adjusted for baseline UI). Change in number of ADLs ranged from -7 to 7, with higher numbers indicating an adverse outcome (i.e., deterioration in functioning). We measured UI by one of three values (1 = none, 2 = some, and 3 = often). Thus, higher values for incontinence also indicated an adverse outcome (i.e., deterioration in UI status). Independent Variables Team Performance. We constructed a team performance score from the responses obtained in the survey of interdisciplinary team members as the average score for all team members within each PACE program. Individual Risk Factors. We obtained individual risk factors from DataPACE. We denoted the lack of independence in ADLs (bathing, dressing, grooming, toileting, transferring, walking, and feeding) by dichotomous variables. We summed instrumental ADLs (IADLs) which include limitations in preparing meals, shopping, doing housework, doing laundry, performing heavy chores, managing money, taking medications, and using transportation in order to indicate the number of IADLs present. We 230 The Gerontologist

measured cognitive status by the number of errors in responding to the Short Portable Mental Status Questionnaire (range 0 10; Pfeiffer, 1975). We defined self-assessed health by using five categorical variables, corresponding to excellent (reference category), good, fair, poor, and missing. We treated missing self-assessed health as another level of the self-assessed health variable, because discussions with experts at PACE sites suggested that missing values were correlated with poor health, an assumption that was borne out by their correlation with mortality (Mukamel et al., 2004). Additional Site-Level Variables In order to avoid potential bias due to omitted variables, we included in the models several site-level variables that were likely to be associated with individual health outcomes. Staffing levels measured the resources available in each PACE site. We computed the number of FTEs per 100 participants for nursing personnel and nonprofessionals (e.g., nurses aides and personal care aides) by using the National Pace Association report for each site (National PACE Association, 2002). Site propensity to hospitalize reflected an important practice-style variable with implications for the efficiency with which PACE sites utilized their resources. As PACE sites are capitated, their revenues are fixed. Hospitalizations are an expensive resource and high rates of hospitalization may affect the availability of funds to provide other services. In order to create a variable measuring the propensity of each site to hospitalize, we used information about hospitalizations, diagnoses, and other individual risk factors obtained from Data- PACE. We estimated a Poisson regression model predicting the number of times that an individual will be hospitalized within one year. This model included individual risk factors and indicator variables for each site (the model is shown in the Appendix). The coefficients for the indicator variables measured the propensity of each site to hospitalize, controlling for differences in case mix across sites. We used these variables in the main analysis as measures of the propensity to hospitalize. Burden of disability may influence site effectiveness in two ways. On one hand, sites with a large percentage of enrollees with similar disabilities may acquire expertise in treating that particular disability, and, therefore, individuals enrolled in such sites may experience better outcomes. On the other hand, a heavy disability burden in a fixed capitated revenues environment may limit the resources available per enrollee, thus potentially leading to worse outcomes. We created two variables, one measuring the ADL burden and one measuring the incontinence burden for each site. We defined the ADL burden variable as the average number of ADL limitations for all non-new enrollees in each site. We defined the UI burden as the percentage of all non-new enrollees with UI. Higher values for these variables indicated a higher burden of the specific disability at the site. Analyses We estimated five multivariate regression models corresponding to each of the five health outcomes defined in the Dependent Variables section. We estimated the models on the individual-level data, with the dependent variables being the health outcome and the independent variables being the individual risk factors and the site-level variables. We modeled changes in functional status at 3 and 12 months by using linear regressions, we modeled UI at 3 and 12 months as an ordered logit, and we estimated mortality by using a Cox proportional hazards model. In order to account for the potential correlation between observations for individuals enrolled in the same PACE site, we estimated all models as hierarchical models (i.e., models that recognize individuals are clustered within PACE programs). We estimated all models as random site intercepts (frailty for the Cox hazard model), thus allowing the intercept to be different for each program and the predicted outcome for each individual to depend on the PACE program in which he or she was enrolled. The initial models included all the individual risk factors listed in Table 1 in addition to site-level variables. We excluded risk factors that were not significant at the p =.1 level from the final models. An F test indicated that the excluded risk factors as a group were not significantly associated with the outcome and were thus appropriately excluded. Results Table 1 reports means and standard deviations for all variables included in the analyses. The sample is similar to the PACE population in general. The population is old (average age 77.6 years old) and mostly female (73.5%), with multiple ADL and IADL limitations. A large percentage (50.6%) has diagnosed dementia, and many are bladder and bowl incontinent (60.3% and 27.7%, respectively); 40.8% report their health as being fair or poor, and an additional 14.9% have no self-assessed health reported, which is typically indicative of poor health (Mukamel et al., 2004). The measure of overall team performance had an average value of 4.19. This suggests that most team members view their team as performing well (recall that the measure ranges from 1 5, with higher numbers indicating better performance). There was, however, substantial variation in this measure, with a standard deviation of 0.69 (i.e., 66% of teams rated themselves between 3.5 and 4.9). This suggests that Vol. 46, No. 2, 2006 231

Table 1. Sample Characteristics Table 1. (Continued) Variable M (SD) Variable M (SD) Program level Team effectiveness (1 ¼ least effective, 5 ¼ most effective) 4.19 (0.69) FTEs per 100 participants Nonprofessionals 22.08 (10.94) Professionals 4.14 (1.57) Propensity to hospitalize 0.226 (0.49) ADL burden (average no. of ADLs for all nonnew enrollees in program) 4.56 (0.76) UI burden (% of non-new enrollees with urinary incontinence) 63.31% (14.57) Individual level Changes in ADLs at 3 months (range from 7 to 7) 0.086 (1.47) Changes in ADLs at 12 months (range from 7 to 7) 0.198 (1.94) UI at 3 months, often 34.0% UI at 12 months, often 33.3% Age 77.6 (9.58) Female 73.5% Living arrangements Alone 29.7% With spouse 10.8% With other relatives 37.5% In nursing home 9.9% Other 12.1% Race or ethnicity White 45.5% Black 35.9% Hispanic 14.6% Asian 2.8% Other 1.2% Education, 8 years 44.2% ADLs Bathing 81.2% Dressing 68.5% Grooming 71.6% Toileting 47.9% Transferring 44.8% Feeding 30.1% Walking 49.8% Sum of IADLs 7.342 (1.35) Bowel incontinence 27.7% Errors in SPMSQ 4.15 (3.17) Diseases Hypertension 61.8% Cerebrovascular disease 28.3% Cancer 10.4% Dementia 50.6% Other cardiac diseases 46.8% Chronic obstructive pulmonary disease 13.6% Other respiratory diseases 12.4% Diabetes 33.3% Renal failure 8.2% Treatments Oxygen daily 4.7% Chemotherapy 0.4% Behavioral disorders Wandering 11.6% Regressive 7.6% Dependent on appliance 7.2% Communication difficulties Expressive 23.2% Receptive 23.0% Self-reported health Excellent 6.4% Good 37.9% Fair 30.0% Poor 10.8% Missing 14.9% Notes: FTE = full-time equivalent; ADL = activity of daily living; UI = urinary incontinence; IADL = instrumental ADL; SPMSQ = Short Portable Mental Status Questionnaire. PACE programs differ in the way their teams function and in the way their staff evaluate their own performance. Table 2 presents the five regression models predicting each of the health outcomes. For the functional status models, the table presents the increase in number of ADL limitations due to the specific risk factor. For example, individuals in a nursing home were likely to experience on average an increase of 0.36 ADL limitations (i.e., worsening functional status) during the first 3 months (second column), compared with individuals living alone. For the UI models, the numbers shown are the odds ratio for worsening UI. For individuals living in a nursing home, the odds ratio at 12 months was 1.75, indicating that the odds of UI deterioration within 12 months of enrollment were 75% greater compared with the reference category of persons living alone (odds ratio = 1). For the mortality model, the table reports the hazard ratio for mortality at any given time within one year post enrollment. Persons living in a nursing home had a lower mortality hazard, at 0.72, compared with those living alone. Different individual risk factors were associated with the outcomes, although, as expected, the models for the functional decline outcomes at 3 and 12 months were similar, and the models for incontinence at 3 and 12 months also were similar. We found different patterns for the associations between health outcomes and site-level variables. The measure of overall team performance was significantly associated with ADL outcomes at 3 months, and even more so at 12 months. It was also significantly associated with UI outcomes at 12 months. In all three instances, individuals who were enrolled in sites where team performance was ranked higher had better risk-adjusted outcomes. A similar relationship existed for 3-month UI outcomes, but the association 232 The Gerontologist

was not statistically significant. We also found no significant association with mortality outcomes. Sites with higher nursing FTE intensity appeared to have lower mortality (i.e., better outcome) but not better ADL or UI outcomes. Propensity to hospitalize was associated with worse ADL outcomes but not with mortality or UI outcomes. Discussion PACE often serves as a model for those who believe that interdisciplinary teams are essential for effective management of complex patients. Keough, Field, and Gurwitz (2002) described the use of a PACE site as training grounds for medical and family practice residents, because team definition, interdisciplinary role appreciation, and responsibility are central components of these training programs. PACE also offers an excellent opportunity to examine the hypothesis that team performance is associated with better patient outcomes. Teams have been recognized as particularly important in caring for frail elders because of the complexity of the medical and supportive services these individuals require, the diversity of the settings in which care is available, and the patients need to access different types of providers (Heinemann & Zeiss, 2002). This complexity demands coordination of services across disciplines and settings, as well as management of the different funding sources available through Medicare, Medicaid, and private insurance. Interdisciplinary teams, which are empowered to collaborate in making decisions that cross disciplinary lines (Garner, 1994), are particularly well suited to manage such complex patients. The PACE teams are interdisciplinary to a much larger degree than teams found in most other settings because they include both professionals and non-professionals and because they are comprehensive and incorporate all staff providing patient care. Furthermore, these teams are not only empowered to collaborate, but are required by their programs and by federal regulations to make decisions concerning their patients well-being. The existence of such comprehensive and empowered teams is an advantage, as it allows for a much more complete assessment of patient needs and encourages input in care planning from all those who implement the plan and are therefore in a position to impact its outcome. On the flip side, managing large and very diverse teams in particular teams that include both professionals and non-professionals may be more challenging. Effective team performance may be more difficult to achieve. Indeed, despite the many similarities between PACE programs and their common philosophy of care, we found that the performance of their teams varied significantly. The data suggest that individual patients receiving care from teams that rank their performance higher had better health outcomes in some important dimensions short- and long-term functional outcomes and long-term UI outcomes. The lack of association between mortality and team members perceptions of their performance may be explained by the fact that the primary objective of PACE is to improve quality of life by enhancing function and people s ability to maintain independence in the community for as long as possible. In a frail, elderly, and disabled population such as that found in PACE, prolonging survival is not always compatible with enhancing quality of life or with patients treatment preferences. Indeed, PACE programs actively engage their participants in discussions about end-of-life treatment and solicit advance directives. This has resulted in a much higher percentage of individuals with such directives compared with that from among the general elderly population, as well as a very high proportion of persons who prefer not to be kept alive by any form of life support (Temkin-Greener & Mukamel, 2002). It is, therefore, not surprising that team effectiveness was not a factor in mortality. It is interesting to note, however, that individuals within PACE programs with a higher number of professional FTEs per 100 participants did have lower mortality rates, presumably reflecting a more medicalized model of care. The finding that health outcomes were associated with overall team performance suggests that qualityimprovement efforts for PACE sites should focus on ways to improve the performance of their care teams. Several dimensions of team function have been identified as contributing to better operations (Fried, Topping, & Rundall, 2000; Heinemann & Zeiss, 2002; Nichols, DeFriese, & Malone, 2002; Shortell et al., 1991; Shortell et al., 1994; Shortell et al., 2004). Team leadership is important as an enabler of positive and effective team processes. The leadership not only sets the goals and standards for the team, but determines the agenda on which the team will focus. It can also facilitate or hinder the participation of team members in the team process. It is up to the leadership to create a culture that will encourage all team members across disciplines, professionals and non-professionals, to take an active role in team discussions and to feel comfortable contributing to them. Also important is developing a sense of ownership and belonging among team members, and it is up to the leadership to foster an environment that will encourage participation by all. Communication among team members is another important aspect of good team functioning, as studies of teams in acute-care settings have found (Baggs, Ryan, Phelps, Richeson, & Johnson, 1992; Gavett, Drucker, McCrum, & Dickinson, 1985; Knaus, Draper, Wagner, & Zimmerman, 1986). An accurate, open, and timely flow of information to all who require it to perform their tasks is crucial (Shortell et al., 1991). An atmosphere that encourages open and respectful communication will facilitate accurate and complete transfer of information, Vol. 46, No. 2, 2006 233

Table 2. Models Predicting Health Outcomes Variable Decline in Functional Status (Increase in No. of ADLs) Within 3 Within 12 Deterioration in Urinary Incontinence (Odds Ratio) Within 3 Within 12 Mortality Within 1 Year of (Hazard Ratio) Mortality Within 2 Years of (Hazard Ratio) Program effects Team effectiveness 0.71** 1.77*** 0.61 0.23** 1.77 FTEs per 100 participants Nonprofessionals 0.01 0.02* 0.99 0.98 1.01* Professionals 0.03 0.03 1.02 1.01 0.90*** Propensity to hospitalize 0.18* 0.41*** 0.67* 1.12 0.89 ADL burden 0.16*** 0.26*** NA NA NA UI burden NA NA 0.99 1.02*** NA Individual risk factors at enrollment Age 0.01** 0.01*** 1.01** 1.03*** Female 0.82** Living arrangement Alone (reference) With spouse 0.07 0.25* 1.10 With other relative 0.22** 0.28** 0.79** In a nursing home 0.36*** 1.44*** 1.75*** 0.72** Other 0.17 0.43*** 0.68** Race or Ethnicity White (reference) Black 0.09 0.10 0.93 0.76* 0.66** Hispanic 0.24* 0.33** 0.51** 0.52*** 0.66*** Asian 0.48** 0.50* 0.68 0.53* 0.61* Other 0.06 0.69* 0.57 0.92 0.56 Education, 8 years 0.15** 1.28* 0.73** ADLs Bathing 0.40*** 0.38*** Dressing 0.32*** 0.54*** 1.27* Grooming 0.60*** 1.99*** 1.46*** 1.27* Toileting 0.25** 0.29** Transferring 0.40*** 0.34*** 1.58*** Walking 0.55*** 1.28** Feeding 0.45*** 0.51*** 1.19* Sum of IADLs 0.10*** 0.13*** Incontinence Bowel 0.19** 0.45*** 1.39** 1.20* Bladder (yes or no) 0.17** 0.29*** Bladder seldom 14.01*** 4.18*** Bladder often 208.15*** 23.10*** SPMSQ 0.04*** 0.08*** 1.03 1.06*** 1.06*** Diagnosis Hypertension 0.16** Cerebrovascular disease 0.23*** 0.16* Cancer 0.21** 1.32** Dementia 1.30* 1.31** 0.83* Cardiac 1.19** COPD 1.32** Renal failure 2.29*** Other renal 1.28** Neurogenito, bladder 2.27** Decubitus 2.89** Treatments Daily oxygen 2.10*** Chemotherapy 8.99*** (Table continues on next page) 234 The Gerontologist

Table 2. (Continued) Variable Decline in Functional Status (Increase in No. of ADLs) Within 3 Within 12 Deterioration in Urinary Incontinence (Odds Ratio) Within 3 Within 12 Mortality Within 1 Year of (Hazard Ratio) Mortality Within 2 Years of (Hazard Ratio) Receiving 1 or more of parenteral nutrition, suctioning, oxygen, chemotherapy or dialysis 0.28** Behavioral disorders Wandering 0.66*** Regressive 2.53*** Communication difficulties Expressive 0.47*** Receptive 0.35* Sample size 1856 1871 1790 1775 2747 F test for H o : all excluded variables (as a group) can be excluded 0.42 0.92 0.85 0.23 0.77 Notes: ADLs = activities of daily living; FTE = full-time equivalent; UI = urinary incontinence; IADLs = instrumental ADLs; SPMSQ = Short Portable Mental Status Questionnaire; COPD = chronic obstructive pulmonary disease; NA = not applicable to this model. Only individual risk factors significant at the 0.1 level were retained in the final models presented here. Blank cells indicate that the variable was not significant in that model. For example, age was not significant in the model predicting urinary incontinence deterioration within 3 months. *.05, p.10; **.01, p.05; ***p.01. thus enabling all members of the team to make fully informed decisions. Team members often need to coordinate their activities. This can be achieved through shared planning and protocol development activities (Gavett et al.). Another important aspect of team functioning is conflict management. It is inevitable that groups of people who work together and who need to reach a consensus may have differences of opinion. In particular, in a diverse group made up of individuals from different disciplines (including both professionals and non-professionals), conflicts and disagreements due to different perspectives, different information about the patient, and possibly different goals will occur. Although some conflicts might be avoided by improving communication (such as ensuring that everyone operates with the same knowledge base about the patient), others may need to be managed. Well-functioning teams have identified processes such as collaborative problem solving that allow for both prevention of conflicts and amicable resolution of conflicts so that no team members feel alienated or slighted. Although teams should strive to improve in all of these dimensions, not all of them seem to be equally important for overall team performance. Temkin- Greener and colleagues (2004) reported that good communication among team members contributes the most to the effectiveness of the team. They found that on a 5-point Likert scale (with 5 being the most positive), an increase of 1 point in communication scores led to almost a 0.50-point increase in the overall performance score. Coordination and conflict resolution also were found to be important, although their impact on overall performance was half as large, leading to only about a 0.25-point increase in the overall performance score. Surprisingly, leadership was the least important dimension, with almost a negligible, albeit statistically significant, effect (of 0.07). It is possible that the small leadership effect is an artifact of the analysis, which included all four dimensions in the same multivariate analysis. To the degree that effective leadership enables communication, coordination, and conflict resolution among team members, the effect of leadership measured in this analysis was the residual effect, above and beyond its impact on communication, coordination, and conflict resolution. Thus, efforts at improving team performance cannot ignore the role of the team leaders. These findings suggest that if there is a need to prioritize team-building activities, communication should be the first area addressed. It is also important to note that teams are embedded in their organizations, and although common characteristics identify more effective teams across all programs, improving the functioning of a team has to be approached within the culture of the organization (Gibson & Barsade, 2004). Furthermore, it is important to be sensitive to the possibility that the organization is not homogenous and may in fact have several subcultures, requiring different approaches (O Reilly, 1989). Indeed, Temkin-Greener and associates (2004) reported significant differences in the perception of team functioning and effectiveness between the professional and non-professional members of the team, with non-professionals viewing the team as less effective than the professionals. This suggests that particular attention Vol. 46, No. 2, 2006 235

should be paid to including non-professionals in the team and ensuring their sense of participation and ownership. One caveat should be noted. PACE sites are unique in many respects, and even though the team concept is central to their functioning, it is possible that other aspects of the program that are correlated with team performance and omitted from our analysis contribute to participant health outcomes. For example, it might be the case that programs with professional staff (physicians and nurses) trained in geriatrics also have better performing teams, perhaps because professionals trained in geriatrics may be more cognizant of the importance of teams and more inclined to be good team members and leaders. If this is the case, the analyses presented here may be overestimating the magnitude of the direct team effects. Further research to examine in more detail the components of team performance (such as communication and coordination) is needed in order to obtain more specific information that can guide efforts to improve team performance. References Baggs, J. G., Ryan, S. A., Phelps, C. E., Richeson, J. F., & Johnson, J. E. (1992). The association between interdisciplinary collaboration and patient outcomes in a medical intensive care unit. Heart Lung, 21, 18 24. Bodenheimer, T. (1999). Long-term care for frail elderly people: The On Lok Model. New England Journal of Medicine, 341, 1324 1328. Eng, C., Pedulla, J., Eleazer, G. P., McCann, R., & Fox, N. (1997). Program of All-inclusive Care for the Elderly (PACE): An innovative model of integrated geriatric care and financing. Journal of the American Geriatrics Society, 45, 223 232. Fried, B., Topping, S., & Rundall, T. G. (2000). Groups and teams in health services organizations. In S. M. Shortell & A. D. Kaluzny (Eds.), Health care management: Organization, design and behavior (4th ed.). Albany, NY: Delmar. Garner, H. G. (1994). Multidisciplinary versus interdisciplinary teamwork. In H. G. Garner & F. P. Orelove (Eds.), Teamwork in human services: Models and applications across the life span (pp.19 36). Boston: Butterworth-Heinemann. Gavett, J. W., Drucker, W. R., McCrum, M. S., & Dickinson, J. C. (1985). A study of high cost inpatients in Strong Memorial Hospital. Rochester, NY: Rochester Area Hospital Corporation and the University of Rochester. Gibson, D. E., & Barsade, S. G. (2004). Managing organizational culture change: The case of long-term care. In A. S. Weiner & J. L. Ronch (Eds.), Cultural change in long-term care (pp. 11 34). New York: Haworth Social Work Practice Press. Gross, D., Temkin-Greener, H., Kunitz, S., & Mukamel, D. B. (2004). The growing pains of integrated health care for the elderly: Lessons from the expansion of PACE. Milbank Quarterly, 82, 257 282. Heinemann, G. D., & Zeiss, A. M. (2002). Team performance in health care: Assessment and development. New York: Kluwer Academic/Plenum. Keough, M. E., Field, T. S., & Gurwitz, J. H. (2002). A model of communitybased interdisciplinary team training in the care of the frail elderly. Academic Medicine, 77, 936. Knaus, W. A., Draper, E. A., Wagner, D. P., & Zimmerman, J. E. (1986). An evaluation of outcome from intensive care in major medical centers. Annals of Internal Medicine, 104, 410 418. Mukamel, D. B., Peterson, D. R., Bajorska, A., Temkin-Greener, H., Kunitz, S., Gross, D., et al. (2004). Variations in risk-adjusted outcomes in a managed acute/long-term care program for frail elderly individuals. International Journal of Quality Health Care, 16, 293 301. Mukamel, D. B., & Spector, W. D. (2003). Quality report cards and nursing home quality. The Gerontologist, 43(Special Issue II), 58 66. Mukamel, D. B., Temkin-Greener, H., & Clark, M. (1998). Stability of disability among PACE enrollees: Financial and programmatic implications. Health Care Financing Review, 19, 83 100. National PACE Association. (2002). 2001 PACE staff ratios report. Alexandria, VA: Author. Nichols, L., DeFriese, A. M., & Malone, C. C. (2002). Team process. In G. D. Heinemann & A. M. Zeiss (Eds.), Team performance in health care: Assessment and development (pp.71 88). New York: Kluwer Academic/ Plenum. O Reilly, C. (1989, Summer). Corporation, culture, and commitment: Motivation and social control in organizations. California Management Review, 31, 9 25. Pfeiffer, E. A. (1975). A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients. Journal of the American Geriatrics Society, 23, 433 441. Rantz, M. J., Hicks, L., Grando, V., Petroski, G. F., Madsen, R. W., Mehr, D. R., et al. (2004). Nursing home quality, cost, staffing, and staff mix. The Gerontologist, 44, 24 38. Shortell, S. M., Marsteller, J. A., Lin, M., Pearson, M. L., Wu, S. Y., Mendel, P., et al. (2004). The role of perceived team effectiveness in improving chronic illness care. Medical Care, 42, 1040 1048. Shortell, S. M., Rousseau, D. M., Gillies, R. R., Devers, K. J., & Simons, T. L. (1991). Organizational assessment in intensive care units (ICUs): Construct development, reliability, and validity of the ICU nurse physician questionnaire. Medical Care, 29, 709 726. Shortell, S. M., Zimmerman, J. E., Rousseau, D. M., Gillies, R. R., Wagner, D. P., Draper, E. A., et al. (1994). The performance of intensive care units: Does good management make a difference? Medical Care, 32, 508 525. Sommers, L. S., Marton, K. I., Barbaccia, J. C., & Randolph, J. (2000). Physician, nurse, and social worker collaboration in primary care for chronically ill seniors. Archives of Internal Medicine, 160, 1825 1833. Stuck, A. E., Siu, A. L., Wieland, G. D., Adams, J., & Rubenstein, L. Z. (1993). Comprehensive geriatric assessment: A meta-analysis of controlled trials. Lancet, 342, 1032 1036. Temkin-Greener, H., Gross, D., Kunitz, S., & Mukamel, D. B. (2004). Measuring interdisciplinary team performance in a long-term care setting. Medical Care, 42, 472 481. Temkin-Greener, H., & Mukamel, D. B. (2002). Predicting place of death in the Program of All-Inclusive Care for the Elderly (PACE): Participant vs. program characteristics. Journal of the American Geriatrics Society, 50, 125 135. Wagner, E. H. (2004). Effective teamwork and quality of care. Medical Care, 42, 1037 1039. Weiland, D., Kramer, B. J., Waite, M. S., & Rubenstein, L. Z. (1995). The interdisciplinary team in geriatric care. American Behavioral Scientist, 39, 655 664. Wieland, D., Stuck, A. E., Siu, A. L., Adams, J., & Rubenstein, L. Z. (1995). Meta-analytic methods for health services research: An example from geriatrics. Evaluation and the Health Professions, 18, 252 282. Yeatts, D. E., Cready, C., Ray, B., DeWitt, A., & Queen, C. (2004). Selfmanaged work teams in nursing homes: Implementing and empowering nurse aide teams. The Gerontologist, 44, 256 261. Received April 19, 2005 Accepted October 10, 2005 Decision Editor: Linda S. Noelker, PhD 236 The Gerontologist

Appendix Model Predicting Number of Hospitalization in 1 Year Age 0.00 Female 0.17*** Living arrangement Living alone Reference category Living with spouse 0.24* Living with other relative 0.18** Living in nursing home 0.29*** Other 0.20* Race or ethnicity White Reference category Black 0.14* Hispanic 0.17 Asian 0.22 Other 0.39* Education, 8 years 0.14** ADLs Bathing 0.12 Dressing 0.04 Grooming 0.05 Toileting 0.03 Transferring 0.15 Walking 0.07 Feeding 0.01 Sum of IADLs 0.03 Incontinence Bladder 0.07 Bowel 0.06 SPMSQ 0.01 Diagnosis CHF 0.40*** Hypertension 0.18*** Other cardiac 0.19*** COPD 0.09 Other respiratory 0.14** Cerebrovascular 0.03 Infectious diseases 0.39*** Dementia 0.04 Depression 0.09 Cancer 0.01 Diabetes 0.22*** Renal failure 0.31*** Other 0.37* Behavioral disorders Wandering 0.09 Verbal 0.02 Physical 0.33** Regressive 0.12 Dependent on appliance 0.61*** Self-reported health Excellent Reference category Good 0.21 Fair 0.43*** Poor 0.51*** Missing 0.33** Site A 0.47*** Site B 0.02 Site C 0.14 Site D 0.36** Appendix (Continued) Site E 0.07 Site F 0.53*** Site G 0.00 Site H 0.91*** Site I 0.48** Site J 0.14 Site K 1.02*** Site L 1.11*** Site M 0.60 Site N 0.88*** Site O 0.69*** Site P 0.01 Site Q 0.20 Site R 0.28 Site S 0.49** Site T 0.89*** Site U 0.29* Site V 0.14 Site W 0.40* Site X 0.18 Site Y 0.44* Notes: ADLs = activities of daily living; IADLs = instrumental ADLs; SPMSQ = Short Portable Mental Status Questionnaire; CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease. Vol. 46, No. 2, 2006 237