Introduction
The organization of supportive home care varies across Canada. In contrast to both the growing involvement of the for-profit sector in the delivery of home care and the transfer of home support services from government to informal networks (Daly, Reference Daly2007; Skinner & Rosenberg, Reference Skinner and Rosenberg2006), the Québec government has, since 1997, offered subsidized supportive care through social economy organizations (specifically, non-profit and cooperative agencies). The presence of cooperatives in the delivery of instrumental activities of daily living is unique in the Canadian home care landscape, and proponents of this organizational form have argued that they are well positioned to deliver high-quality services compared to other types of organizations (Restakis & Lindquist, Reference Restakis, Lindquist, Restakis and Lindquist2001). Compared to for-profit agencies, researchers have argued that cooperatives focus on a social mission (i.e., delivering services); compared to non-profit agencies, it is argued that cooperatives place a greater emphasis on the involvement of key stakeholders in decision making. The purpose of this article is to determine whether the cooperative model, compared to non-profits, is a predictor of quality. The specific effects of consumer and worker participation within home support agencies will also be tested.
Care and the Social Economy
Social economyis a term used to describe organizations that function to meet social, rather than profit-motivated, objectives. These organizations focus on social and economic development, addressing needs and issues not being taken up by the market or the state (Laville, Lévesque, & Mendell, Reference Laville, Lévesque and Mendell2007). In Québec, there are 102 home support agencies that offer services such as housekeeping, meal preparation, and help with errands.Footnote 1 Slightly over half of these social economy organizations are incorporated as non-profit, while the rest are cooperatives.
Individuals 65 years of age or over and those younger than 65 with a referral from their local health authority receive a standard, minimum financial contribution from the provincial government and are eligible for greater financial assistance depending on their level of income. The difference between the amount charged by the home support agency and the amount covered by government is paid for by the “user” or consumer. All provincial residents 18 years of age and older are in fact eligible to receive services through the program, but do not qualify for the additional, means-tested financial support unless they are at least 65 years of age or obtain a referral. Note that beyond providing home support services to residents, a second objective of these social economy organizations is to create employment opportunities and workplace integration for people without jobs or for those who face multiple barriers to employment (Chantier de l’économie sociale, 2003).
But do cooperative and non-profit home support agencies perform differently? These agencies, in fact, share a number of characteristics. Both have a non-distribution constraint, meaning they are restricted from pursuing profit for profit’s sake, and both feature a volunteer board of directors elected by stakeholders. However, the governance structure of the two organizational forms is not the same: home care cooperatives must be governed exclusively by consumers or workers, or governance may be shared among workers, consumers, and community members using a multistakeholder or “solidarity” model. Regardless of the specific cooperative form chosen by the community, provincial statutes dictate these specific options for board composition (Éditeur officiel du Québec, 2007). Non-profits have less-specific requirements for who must govern home support agencies, and they feature consumer and worker representation on the board of directors to varying degrees.
Other differences between the two forms include that cooperatives are guided by seven internationally followed principles including a commitment to education about the cooperative model and “concern for community” (International Co-operative Alliance, 2009). Finally, the two types of agencies self-identify differently and, for the most part, associate differently, with one umbrella organization serving each group.
Despite the uniqueness of these Québec cooperatives with respect to the rest of the Canadian home care landscape, they are found elsewhere, including the United States, Italy, Japan, and the United Kingdom. Further, cooperatives are present in other elder care sectors, such as nursing homes (Oka, Reference Oka2000). However, little is known about how these organizations perform, most notably in comparison to other types of home support agencies. Research on auspice and home care has compared for-profit and non-profit agencies (Clarke & Estes, Reference Clarke and Estes1992; Doran & Pickard, Reference Doran and Pickard2004). Literature on elder care cooperatives, in turn, has described either the potential for, or development of, this model in particular jurisdictions (e.g., Borzaga, Reference Borzaga, Spear, Defourny, Favreau and Laville2001; Kahn, Reference Kahn2000). One exception includes a summary of 13 case studies on home support agencies, 7 of which are cooperatives, in Québec. Although the goal of the study was not to ask how non-profits and cooperatives may predict quality, they examined this construct (Jetté & Lévesque, Reference Jetté, Lévesque, Vaillancourt, Aubry and Jetté2003). The authors reviewed the systems in place to monitor services, which reveals dimensions of care important to consumers, including instrumental aspects as well as the relationship between the care provider and recipient.
Participation in Care Delivery
Although there is little research on cooperatives as a potential determinant of quality, some literature has examined how participation may affect quality. This was reviewed in our study because participation in governance is an integral aspect of home support cooperatives, and its specific effects were tested and discussed. Testing such a specific dimension has also been encouraged by researchers who argue that organizational effects may in part be captured by characteristics beyond legal structure (Bozeman & Bretschneider, Reference Bozeman and Bretschneider1994). Examples include the percentage of total revenue that comes from government sources and the proportion of a firm’s output that is public versus private. Our research used stakeholder involvement as one such characteristic.
Turning to the literature, we find that one example of consumer participation is self-directed home care, where seniors are able to select and manage a home care worker with funds from the state. Self-direction has been argued to better meet the needs of seniors because it allows recipients of services to have control over the home care worker’s schedule and the specific kinds of tasks to be done (Benjamin, Reference Benjamin2001), although not all older people are interested in taking up this arrangement (Glickman, Stocker, & Caro, Reference Glickman, Stocker and Caro1997). Evidence suggests that individuals who are able to self-arrange supportive home care are more satisfied with the services they receive than those who receive agency-directed care on a number of dimensions, including (a) the technical aspects of the care delivered, (b) the impact of the services (that is, the ability of the individual to live independently), (c) the quality of the relationship with the provider, and (d) general satisfaction with the home care worker (Benjamin, Matthias, & Franke, Reference Benjamin, Matthias and Franke2000). Long-term care facilities are also experimenting with participation by the elderly; for example, the Eden Alternative seeks to improve the quality of life in nursing homes by, in part, challenging the hierarchical administrative structures present in institutions and encouraging residents’ involvement in decision making to the greatest extent possible. Quasi-experimental research that compared quality of life in a nursing home before and after the implementation of the Eden Alternative found that residents in the experimental group felt less helpless, depressed, and alone than those in the control group (Parsons, Reference Parsons2004).
Research on the connection between worker participation and quality outcomes is more tenuous. A case study comparing the opportunities available to staff belonging to a worker-owned home care agency and a for-profit agency demonstrated that members of the cooperative had more professional development activities available, greater opportunities to take on leadership positions, and enhanced self-esteem (Majee, Reference Majee2007). Although the specific impact on quality of care was not tested, other home support researchers have found links between measures taken to improve the working conditions of frontline staff and job retention (Feldman, Reference Feldman1993). Retention, in turn, has been argued to improve quality of care since it provides for a pool of workers who have a developed skill set, and in-depth knowledge of the needs and preferences of clients. Research on empowerment teams for nursing aides in nursing homes has also found that weekly meetings between workers and management allowed staff to discuss the needs of residents based on their experiences on the front lines, thereby improving quality by providing more tailored and suitable assistance to residents (Yeatts & Cready, Reference Yeatts and Cready2007). Other researchers have found, however, that such teams may serve to increase the strain experienced by personal care workers often facing heavy workloads (Allan & Lovell, Reference Allan and Lovell2003). Regardless of the impact of empowerment teams, it is worth noting that they are not designed for workers to influence or change human resource policies.
It is important to note that although home support agencies in our research feature stakeholder involvement in governance, the control of stakeholders is not absolute. Like some self-directed home support programs, the state dictates what kinds of services are eligible for government subsidies, and also the extent to which they will reimburse the home care agencies for each hour of service delivered. As a result of these constraints, it is not possible for agencies to meet all the potential service needs identified by their users, or to be fully autonomous in their human resource practices (e.g., hourly wages may be lower than desired because of the reimbursement rates provided by government). Nevertheless, consumers and workers, through their involvement on these agencies’ board of directors, are able to review the agency’s financial statements, make decisions about how funds are allocated, oversee the work of management, and set goals, priorities, and policies.
Methods
Data Collection and Generalizability
We used a cross-sectional design for our research. Data collection began in June 2006 and ended in March 2007. The population included home support agencies offering services through the government’s Financial Assistance for Domestic Help Services, as well as the agencies’ clients. The sampling frame we used for agency selection was a complete list of home care non-profits and cooperatives compiled by the Régie de l’assurance maladie Québec. For client selection, agencies used their lists of care recipients.
We drew a random sample of agencies and called selected home support agencies to see if they would agree to complete a questionnaire and to distribute surveys to their users. In Québec, the inclusion criterion for distributing client surveys was that cooperatives were required to have consumer representation on the board of directors, meaning they needed to be incorporated as either a consumer or multistakeholder cooperative. This meant that worker cooperatives were ineligible to participate. All cooperatives that participated in this research were in fact multistakeholder. Regarding clients—because all Québec residents age 18 and over are eligible to receive home support services (e.g., young, busy families may purchase these services from home support agencies as well)—our inclusion criteria targeted research participants that require home support services to help them live independently. These are also individuals who qualify for additional financial support from the government to help cover the cost of services. Specifically, inclusion criteria were that individuals had to be younger than age 65 with a referral from their local health authority, or seniors 65 years of age or older. The mean age of participants was approximately 74 years, with 83 per cent of the sample being 65 or older.
The objective was to collect agency and client data from 10 cooperative and 10 non-profit organizations. In total, 27 Québec agencies were approached to distribute surveys to their users; seven declined. While 10 of each agency type were recruited to collect client data, client surveys received from one cooperative and one non-profit were subsequently dropped from the data set. The cooperative was, in fact, a worker-owned home support agency that was not properly screened during recruitment. The client surveys associated with one non-profit were removed from the data set because it became clear that they were not distributed randomly to clients. Further, workers in this agency may have helped care recipients fill out the surveys.Footnote 2
Home support agencies were instructed to distribute surveys by mail to every nth person on their client list, depending on their number of clients. Participating home support agencies were sent a box of postage-paid mail surveys. These envelopes also included a cover letter, a postage-paid return envelope (addressed to the researchers, not the home support agency), and two copies of a consent form. Home support agencies received the instructions on how to distribute questionnaires both by phone and in the letter accompanying the box of questionnaires. Further, agencies were called after questionnaires were distributed to obtain aggregate information on their clients (such as their distribution by age and gender); at this point, agencies were also asked to confirm that clients were selected randomly and that questionnaires were sent by mail.
Questionnaires were first distributed to the clients of two agencies to test whether using mail surveys would result in an adequate response rate. When the response rate reached 40 per cent, we decided to proceed with this method of data collection.
The overall client response rate in Québec was approximately 40 per cent (831/2,082). Response rates were calculated by dividing the number of usable returned questionnaires by the total number of questionnaires sent to clients. Of the returned questionnaires, 16 (or 1.85% of returned surveys) were not usable. “Not usable” was defined as having responded to fewer than 10 out of the 13 questions pertaining to the summated satisfaction score. If a questionnaire was sent to someone who did not meet the study criteria (e.g., to an individual younger than 65 in good health), then both the numerator and the denominator were decreased by one. This occurred in 17 instances.
Several strategies were used to increase client response rates (MacDonald, Newburn-Cook, Schopflocher, & Richter, Reference MacDonald, NewburnCook, Schopflocher and Richter2009). Attention was paid to the layout of the document—questions were short and mostly multiple choice—and a postage-paid return envelope was provided with each questionnaire. Reminder postcards and new questionnaires were not sent to non-respondents. To maintain the anonymity of home support clients, client lists were not solicited from participating home support agencies. Moreover, home support agencies were not asked to track non-respondents because such a request was felt to involve too great a time commitment on the part of their staff and might impact the number of organizations amenable to participating in the research.
To explore and assess non-response bias, we calculated chi-square goodness-of-fit statistics for all participating organizations that had data available on the distribution of clients based on age and gender. Data on age were received from 13 out of the 18 agencies, while data on gender were available from 11. Of the 13 agencies, 9 were representative with respect to client age, and 8 out of 11 agencies were representative with respect to client gender.
Further, other sources of potential survey error were minimized (MacDonald et al., Reference MacDonald, NewburnCook, Schopflocher and Richter2009). Sampling error was minimized through a large sample size for clients (n = 831); we avoided non-coverage error by choosing a sampling frame that included clients needing services to live independently, and we minimized measurement error by using a survey with sound psychometric properties.
We explored the Québec agencies’ representativeness using 2002–2003 statistics on the home support sector published by the provincial department of economic development (Chagnon, Reference Chagnon2004). The variable we used to assess representation was the total hours of care delivered by home support agencies, but because of missing cell frequencies, it was not possible to calculate a goodness-of-fit statistic for the agencies that collected client data. However, for cooperatives, small and large organizations, as compared to mid-sized agencies, seem to have participated disproportionately in the research. For non-profits, descriptive statistics suggest that participating agencies match the population of non-profits with respect to hours delivered.
Measures
The survey completed by home support agencies was developed by the researcher using literature on home care workers, agencies, and quality of care. In-depth interviews were also held with stakeholders including a manager of a home support agency and representatives of cooperative and non-profit umbrella associations. The survey was pre-tested by a multistakeholder home care cooperative that did not participate in the research and by two health care researchers.
The consumer survey we used in this research was a reliable and valid client satisfaction instrument developed by researchers in collaboration with health authorities, home care recipients, and home care agencies in the province of Ontario (Smaller World Communication, 2000). Dimensions of service quality overlapped with qualitative interviews with recipients of home support services in Québec (Jetté and Lévesque, Reference Jetté, Lévesque, Vaillancourt, Aubry and Jetté2003). Further, to establish face validity, the survey was reviewed by the executive director of a participating home support agency. Finally, the survey’s reliability was also assessed, using responses obtained during this study; the Cronbach’s alpha was 0.88.
The client survey asked participants to rate 13 dimensions of the home support services they received, with choices on 10 dimensions being excellent, good, fair, and poor. For three dimensions, possible answers were yes/no/don’t know. Dimensions included whether the client was treated in a caring and friendly manner by the home support worker, whether their home and belongings were respected by the home support worker, and whether the same person or team of people was providing care. [For a complete listing of these dimensions, see Table 3 on page 115]. The survey also asked participants to rate the overall quality of care they received, with choices ranging from excellent to poor. Finally, the survey asked clients for socio-demographic information.
Variables and Coding
Two dependent variables were used in this research. Both were quality focused and consumer centered. The first was a satisfaction score, used in the first multivariate analysis. This score ranged from 0 to 39, and was a summated score comprising 13 dimensions of home support services. The Likert scale responses were assigned integers corresponding to their ranks: that is, excellent was assigned a value of 3, good was assigned a value of 2, fair was assigned a value of 1, and poor was assigned a value of 0. The two yes/no questions were assigned similar values so that answers would have the same weight as the Likert responses. The survey also included one “yes, always,” “yes, most of the time,” “no,” “don’t know question.” Here, “yes, always” was assigned a 3, “yes, most of the time” was assigned a 2, and “no” was assigned a 0. If the respondent answered “don’t know” to any of these three questions, the denominator of their satisfaction score changed, and the score was then standardized so that it equaled 39. This occurred in 25 instances.
The second dependent variable was the overall quality rating provided by clients, which ranged from 0 (poor) to 3 (excellent). This was used in the second multivariate analyses.
Key independent variables included organizational status, grouped as cooperatives (1) and non-profits (0), and consumer and worker involvement in the agency. These latter two variables were operationalized as the percentage of board members that were consumers, and the percentage of board members that were workers.
Prior research on consumer assessments of care services have found that socio-demographic characteristics are related to quality outcomes (Aharony & Strasser, Reference Aharony and Strasser1993; Geron, Smith, Tennstedt, Jette, Chassler, & Kasten, Reference Geron, Smith, Tennstedt, Jette, Chassler and Kasten2000; Hall, Milburn & Epstein, Reference Hall, Milburn and Epstein1993). As a result, this research controlled for three client-level variables. Gender was grouped as female (1) and male (0). Age was operationalized as the date the survey was completed, subtracted by the client’s year of birth and left as a continuous variable. The third client-related variable was the consumer’s desire to be in a hospital or nursing home versus his or her own home. This was categorized as yes, no, and don’t know, and coded as dummy variables. Two other control variables, namely the participant’s perceived health status and the amount of co-payment, were present in the original research models but dropped due to high item non-response.
Other control variables included the number of services clients received, ranging from one to six, as well as the type or types of services received. This included services received by at least five per cent of participants, namely meal preparation, personal support and help running errands. Help with housekeeping was not included as a control variable because it was received by 98.3 per cent of participants. Services received by less than five per cent of participants included outdoor work (including help clearing snow or piling wood) and basic household repairs (such as installing safety bars). In Model 2, which used overall quality as the dependent variable, control variables also included participant ratings of the 13 dimensions of quality.
Finally, three agency-related control variables were included in the analysis. Staff wages were operationalized as the average hourly wage of full-time home support workers during the agency’s last fiscal year. If the agency did not employ full-time workers, then the average hourly wage of part-time workers was used instead. Staff benefits were operationalized as the number of benefits offered to full-time home support workers during the agency’s last fiscal year. Again, if the agency did not employ full-time workers, then the average number of benefits offered to part-time workers was used instead. Staff training hours was measured as the number of continuing training hours provided by each home care agency.
Data Analysis
Data were entered and analyzed using Stata 9.0. All survey data were entered once and then cleaned manually by comparing each observation to the corresponding paper survey. Frequency distributions and cross-tabulations were examined for unusual values.
Data were then analyzed for missing values. To examine missing values, we used t-tests and chi-square tests to see if they were missing completely at random, missing at random, or were non-ignorable (Allison, Reference Allison2002). In terms of the 13 questionnaire items aggregated to create the summated satisfaction score, 89.35 per cent of respondents had no missing items, while 8.12 per cent of respondents had one, 1.69 per cent had two missing items, and 0.85 per cent had three or four. The statistical software package normally performs listwise deletion for the summated score of participants if any of their 13 responses are missing, which would have resulted in 10.65 per cent (101/948) of the scores being lost. As a result, “horizontal” or “person mean imputation” was used to estimate the values of the missing items “saving” the summated scores from deletion (Huisman, Reference Huisman2000; Shrive, Stuart, Quan, & Ghali, Reference Shrive, Stuart, Quan and Ghali2006).
Two variables related to client characteristics had high numbers of missing data. The survey instrument included a question regarding how much clients co-pay for an hour of home care. However, this question was not included in the surveys sent to the first two agencies that distributed questionnaires to their clients (n = 150).Footnote 3 Second, 10.6 per cent of participants who were asked this question did not answer, possibly because co-payments are calculated based on income. The data were not missing completely at random given that a higher number of cooperative clients did not reveal how much they paid for an hour of service (p ≤ 0.05). It was also not missing at random: within the cooperative category, those who did not respond were older than those who did (p ≤ 0.05). Since data were missing from two agencies and not missing at random from those clients who were asked the question, we decided not to impute the values or to include the “pay” variable in the final multivariate analyses.Footnote 4
The survey also included a question about clients’ perceived health status; however, like the question addressing the hourly fee, it was not included in the survey distributed to the first two surveys of clients. Note that there is an association between perceived health status and an individual’s desire to be in one’s own home versus an institution (V = 0.05). Like the “pay” variable, perceived health status was not included in the final analysis; this would also have resulted in losing not only 150 client questionnaires but also two home support agencies from the sample. The majority of research participants who were asked the health status question, however, provided an answer (that is, only 0.76% of answers were missing). As a result, this variable was tested in an exploratory data analysis to see how it affected results. It was found that including perceived health status did not affect the significance or size of any other coefficients in Models 1 and 2, nor was it a significant predictor of the summated satisfaction score or the overall quality score.
In terms of comparing cooperatives to non-profits, we generated descriptive statistics on home support clients to provide summary information on the kinds of people receiving services. We generated bivariate statistics such as t-tests and chi-square tests of independence to see if there were any differences in client composition by agency type. Bivariate statistics were also generated on home support agencies for the same reasons: to provide summary information and to make comparisons by agency type.
For the multivariate analyses, we used ordered logistic regression in Models 1 and 2. This approach compares the odds of being in a higher category or categories of the dependent variable. The outcome variable in Model 2 featured a natural categorical ordering; for Model 1, we chose ordered logistic regression because there was limited variation in the dependent variable. The mean total satisfaction score was 34.4 out of 39 (SD 0.18), and 30.93 per cent of respondents had an overall satisfaction score of 39/39. This distribution is negatively skewed (–1.32, p ≤ 0.01) and demonstrates kurtosis (4.70, p ≤ 0.01). As a result, the 39-point score was transformed into a categorical variable based on its distribution. Perfect scores were assigned a 3, while the remaining 69.07 per cent of scores were divided into approximately equal thirds: scores ranging between 36 and 38 (or 23.59% of scores) were assigned a 2, scores ranging between 31 and 35 (also 23.59% of scores) were assigned a 1, and scores below 31 (21.90% of scores) were assigned a 0. For both models, the following regression assumptions were tested and met before proceeding: (a) the proportionality of odds (this was verified using the Brant test), (b) nonlinearity, (c) the presence of outliers and influential data, and (d) collinearity (Menard, Reference Menard2002). Because client data were collected within clusters (i.e., home support agencies), we adjusted standard errors.
It was not possible to include both organizational form and the two variables representing board composition in the same equation because of the small number of agencies that participated in the study (n = 18). The same is true for worker-related variables: separate equations were estimated that included only one such variable at a time (that is, either the hourly wages of home support workers, the number of benefits received, or the number of training hours provided).
Results
Table 1 provides descriptive statistics on users of support services. The mean client age was approximately 74 years, and the majority of care recipients (79.49%) were women. While approximately 83 per cent of clients were 65 years of age or older and thus eligible to arrange home care services without the approval of the local health authority, 17 per cent were younger than this cut-off age and had a referral. Most clients (82.79%) also preferred being in their own home rather than being admitted to a nursing home or hospital, although almost 10 per cent said they were not sure if they wanted to be admitted to an institution. Regarding the kinds of services obtained, almost all clients received housekeeping. Meal preparation was the second most popular service, although it was received by many fewer individuals compared to housekeeping (16.18% versus 98.30%). A minority of clients reported being provided with personal supportFootnote 5 and having errands done at their request. The mean number of services received by clients was 1.3.
Table 1: Characteristics of individuals receiving services from non-profit and cooperative home support agencies
Bivariate statistics revealed no significant association between client-related variables and agency type. We anticipated no differences since individuals do not choose between home support agencies. Instead, there is one agency per geographic region, and no chance that individuals self-select into a specific agency type. Further, 2006 census data were obtained on the communities in which the home support agencies were located, and we found no associations between organizational type and variables including median household income, the education level of local residents, the percentage of people in private households living below the low-income cut-off, and the local unemployment rateFootnote 6 (Statistics Canada, 2006).
Table 2 provides descriptive statistics on the nine non-profit and nine cooperative agencies that distributed questionnaires to their clients. It shows that non-profits are older than their cooperative counterparts. No statistically significant differences were found between agency types in terms of the kinds of services offered. All provided housekeeping, and almost 90 per cent offered meal preparation. Outdoor work, such as lawn care and snow removal, was the third most common service, with approximately 60 per cent of home support agencies offering assistance in this area, despite the fact that only a small number of clients (2.92%) reported receiving this service. Personal support was offered by fewer than half of the home support agencies. Note that running errands for clients is not featured in Table 2 because it was mistakenly omitted from the checklist in the survey distributed to agencies.
In terms of agency size, we found no statistically significant differences between non-profit and cooperative home support agencies with respect to the number of clients receiving home care, the number of home care hours delivered, and the number of home support workers. With regard to human resource practices, there were no differences in the wages offered by non-profits and cooperatives, with the average hourly wage being CDN $9.81 (SD 0.15). According to the minimum wage law in Québec at the time of data collection, employers were required to pay at least $7.75 (Commission des normes du travail, 2007). There were also no statistically significant differences in terms of training hours, with the mean amount being 8.27 hours (SD 2.73). One difference between the two agency types, however, was the average number of benefits offered to home support workers. Non-profits commonly offered drug plans as well as disability and life insurance, while no cooperatives featured these benefits (p ≤ 0.01). Non-profits and cooperatives had profit sharing, retirement savings, and dental insurance plans in approximately equal proportions.
Finally, in terms of the composition of the boards of directors, no differences were found in terms of the percentage of board members who were users of home support services: consumer representation on the board of non-profits was 24.26 per cent (SD 5.23), while in cooperatives, it was 35.13 per cent (SD 2.60). The percentage of board members who were workers was significantly greater (p ≤ 0.05) in the cooperatives, however, with 34.51 per cent (SD 3.30) of board members being workers, versus 18.62 per cent (SD 6.65) for non-profit boards.
Table 2: Characteristics of non-profit and cooperative home support agencies
a p ≤ 0.01
b p ≤ 0.05
Table 3 shows that, overall, clients gave high quality ratings. The three dimensions with the highest ratings included (a) that individuals received the same quality of care at every visit, (b) that the home support worker was punctual, and (c) that individuals were treated in a caring and friendly manner.
Table 3: Quality ratings
Regression results are presented in Table 4. Model 1 answered whether cooperatives offer higher quality home support services using a summated satisfaction score as the outcome variable. Organizational form did not help predict satisfaction with services; however, testing the specific effects of participation produced different results. As the percentage of workers on the board increased, so did the likelihood of being in a higher category of satisfaction with services, although consumer participation was not significant. Receiving help with meals was also significant in Model 1 (p ≤ 0.05).
Because of the small number of home care clusters, each specification was calculated using only one agency-related variable (i.e., average wages, training hours, and benefits provided to home support workers) at a time. Table 4 reports on the estimation that included average wages only because neither benefits nor training were significant, nor did they change the direction or significance of any other variables in the model. Also, because the sample size changed in the second specification of Model 1, the sample size was also reduced to 712. The direction and significance of the coefficients did not change with the smaller sample size.
Table 4: Regression models
a p ≤ 0.01
b p ≤ 0.05
c Reference category is Not Sure if Wants to be in an Institution
d Reference category is No Numbers in parentheses are standard errors
In Model 2, we used consumers’ assessment of overall quality as the dependent variable to explore whether cooperative home support agencies offer higher quality care. This dependent variable ranged from 0 to 3. Table 4 shows that cooperatives did not increase the odds of clients indicating a higher overall quality rating. However, consumer involvement in governance emerged as significant, increasing the odds of clients reporting a higher category of quality (p ≤ 0.01).
Results also showed that there are certain dimensions of satisfaction with home support services that increase the likelihood of a higher quality rating. These include (a) the skills of the home support worker (p ≤ 0.01), (b) that the home support worker is responsive to the care needs of the client (p ≤ 0.01), (c) the home support worker works independently with minimal supervision (p ≤ 0.01), and (d) the home support worker does the work the care recipient expects (p ≤ 0.01).
In Model 2, client characteristics had no effect on overall quality. Training hours were also significant, although the odds of being in a higher category of overall quality due to an increase in training hours were negligible (odds ratio 1.03, ≤ 0.05).
In the second specification of Model 2, again the sample size is smaller (n = 680). This did not change the direction and significance of the results. Overall, Model 2 performed well, with the pseudo-R2 value equaling approximately 0.69.
Discussion
In Québec, state-subsidized home support services are delivered by non-profit organizations and cooperatives. The presence of cooperatives is distinctive, and contrasts the changing composition of the home support sector elsewhere in Canada. For example, although the province of Ontario challenged the monopoly of non-profit organizations in home care delivery by instituting managed competition and the involvement of private, for-profit businesses (Skinner & Rosenberg, Reference Skinner and Rosenberg2006), the Québec government encouraged the development of social economy organizations in the home support sector.
Non-profits and cooperatives share some similarities: they both have social objectives that supersede profit making, and they both feature voluntary boards of directors chosen by stakeholders. However, cooperatives and non-profits incorporate, self-identify, and organize differently. Moreover, they feature stakeholder representation in governance positions to varying degrees. Proponents of the model speak of a “cooperative difference” in care giving, although to date, empirical evidence has not been gathered with respect to outcomes.
In this study, regression results showed that despite organizational differences, the cooperative form was not a predictor of satisfaction or overall quality. One common explanation for the lack of difference between agencies of any type is that regulatory environments constrain their behaviour and blur the lines between different organizational forms (Bager, Reference Bager1994), but contextually, there is no evidence that this “coercive isomorphism” has taken place. At the time of data collection, there was no standardized monitoring of quality by government, nor were agencies using uniform methods of evaluating the services they offered. Another potential explanation is that non-profit and cooperative home support agencies share more similarities than differences, meaning that organizational form would not emerge as a predictor of care. Quarter, Sousa, Richmond, and Carmichael (Reference Quarter, Sousa, Richmond and Carmichael2001) found substantial overlap in the social objectives of non-profits and cooperatives providing social services as well as the extent to which they rely on government funding and volunteers, and argued that the differences between these two legal structures are more semantic than factual. Other non-profit theorists, however, have disputed the notion that all organizations with the non-distribution constraint are the same, and point to board composition as a major difference. For example, Ben-Ner and Hoomissen (Reference Ben-Ner and Hoomissen1993) differentiated between organizations with and without key stakeholders on the board, and wrote that user or worker involvement in governance results in better monitoring of resources and a greater focus on meeting organizational objectives.
Examining board composition does reveal more interesting results. The multivariate analyses showed that worker participation on the board of directors was a determinant of satisfaction with services, and that consumer participation on the board was a determinant of overall quality. So while we did not find the cooperative, as a legal form, to be associated with quality, stakeholder involvement in governance had an effect. This needs to be cautiously interpreted, since it may be argued that these two dependent variables overlap. Both are quality focused and consumer centered, and the variables are correlated (r = 0.74). Arguably, results would have been more definitive if both variables were significant in both models.
Still, this evidence suggests that consumer and worker involvement in the governance of home support organizations may be important. These findings fit with the growing emphasis placed by social care organizations on the involvement of seniors in care delivery, and with the empirical evidence that demonstrates a relationship between consumer involvement and quality outcomes. Findings also concur with scholars researching the connection between efforts to improve the conditions of frontline home care workers and quality of care (for example, Feldman, Reference Feldman1993). Finally, because research participants assigned high quality scores overall, this study provides evidence that social economy organizations are an effective means through which to deliver home support services.
An unexpected research finding and unrelated to organizational form was the four dimensions of satisfaction that emerged as strong predictors of consumers’ assessment of overall quality in Model 2 (p ≤ 0.01). These include that (a) the worker is responsive to the client’s care needs, (b) she/he works independently with minimal supervision, and (c) she/he does the work the client expects. The fourth variable is the skills of the home support worker. Qualitative research on home care suggests that the relationship between workers and their clients can be as important to those receiving services as more technical dimensions such as workers’ skills (Eustis and Fischer, Reference Eustis and Fischer1994; Eustis, Kane and Fischer, Reference Eustis, Kane and Fischer1993). This is one of the reasons why the high turnover endemic to frontline home care delivery is viewed as so problematic; clients are unable to establish lasting relationships with caregivers (Dawson and Surpin, Reference Dawson and Surpin2001). Our research, however, found that more-instrumental dimensions of service delivery were predictors of overall quality. Variables that could serve as possible proxies for the quality of the relationship between clients and workers, such as whether a consistent worker or team of workers provides services, whether the client is treated in a caring and friendly manner by the home support worker, and whether the worker takes time to answer questions, were not significant in Model 2.
Limitations
This research had two main limitations. The first concerns the measurement of quality. Although many researches use Likert responses to assess the quality of home support services and the survey instrument was carefully constructed, the consistently high ratings it generated suggest that our research might have been more conclusive had we used another measure of quality or another method of data collection (Williams, Coyle and Healy, Reference Williams, Coyle and Healy1998). For example, seniors may frame their experiences in terms of positive and negative aspects of the services they receive or in terms of how services could be improved. These tendencies have been supported through qualitative research, where interviewers have asked open-ended questions to care recipients about their experiences with a provider. Some researchers have proposed that it is through qualitative methods, such as interviews or diaries, that consumers’ perspectives on quality can most adequately be captured (Avis, Bond and Arthur, Reference Avis, Bond and Arthur1997). Others have proposed triangulation, where both qualitative and quantitative data are collected (Hyrkäs & Paunonen, Reference Hyrkäs and Paunonen2000).
The second limitation in terms of understanding organizational form is that private, for-profit organizations were not included as a comparison group. With the growing ubiquity of this type of organization in the Canadian home care landscape, comparing cooperatives to both non-profits and for-profits would have provided insight into whether quality outcomes differ among a wider range of organizational forms.
