Who is Looking After Mom and Dad? Unregulated Workers in Canadian Long-Term Care Homes*

Older adults living in residential long-term care or nursing homes have increasingly complex needs, including more dementia than in the past, yet we know little about the unregulated workforce providing care. We surveyed 1,381 care aides in a representative sample of 30 urban nursing homes in the three Canadian Prairie provinces and report demographic, health and well-being, and work-related characteristics. Over 50 per cent of respondents were not born in Canada and did not speak English as their first language. They reported moderately high levels of burnout and a strong sense of their work’s worth. Few respondents reported attending educational sessions. This direct caregiver workforce is poorly understood, has limited training or standards for minimum education, and training varies widely across provinces. Workplace characteristics affecting care aides reflect factors that precipitate burnout in allied health professions, with implications for quality of care, staff health, and staff retention.

This article describes the workforce providing the majority of direct care to the frail elderly in nursing homes. We report on health, well-being, job and continuing education characteristics of this workforce and discuss policy and management implications that arise. The current lack of information about the numbers and characteristics of this occupational group reduces our ability to undertake effective workforce planning and to monitor progress toward achieving workforce improvements that result in acceptable quality of care. Our research question on the current characteristics of the care aide workforce in long term care (LTC) homes illuminates effective routes to modify those characteristics. This analysis highlights factors that, if modifi ed, could signifi cantly improve quality of care and quality of life for LTC residents, and quality of work life for care aides.
By 2036, the number of Canadians 65 and older will more than double to 10.4 million, with growth most rapid in those over 75 years of age (Statistics Canada, 2010 ). After age 65, age-related dementias (ARDs) and other neurodegenerative diseases begin to take their greatest toll on quality of life and produce the greatest proportional costs to the health system. Presently, one in 40 Canadians aged 65-74 and one in three Canadians over age 85 have ARD (Canadian Institutes of Health Research, 2010 ). By 2038, 1.125 million Canadians, or 2.8 per cent of the population, are projected to have an ARD (Alzheimer Society of Canada, 2010 ). Recent prevalence fi gures in the United States are higher (Alzheimer's Association, 2013 ). In the United States, 70 per cent of people with dementia die in a nursing home (Mitchell, Teno, Miller, & Mor, 2005 ); in Europe, this fi gure ranges from 50 per cent (Wales) to 92 per cent (Netherlands) (Houttekier et al., 2010 ). As the number of older adults with dementia increases, so will the need for supportive living and residential long-term care, without dramatic breakthroughs in either prevention or treatment of ARD. Recent Canadian projections estimate that the need for LTC beds will rise 10-fold by 2038 (Alzheimer Society of Canada).
In addition to being in greater demand, residential LTC facilities (nursing homes) in Canada have become more complex care environments (Hirdes, Mitchell, Maxwell, & White, 2011 ). Increasingly, older adults are admitted later in the trajectories of their dementia and other chronic diseases and are thus more dependent, with more-complex needs and in a greater state of vulnerability. Few new residents to long-term care have low care needs (Ikegami, Morris, & Fries, 1997 ;Mor et al., 2007 ) while 60 per cent have signifi cant and often co-morbid care challenges (Doupe et al., 2012 ); more than 70 per cent have an ARD (Doupe et al., 2011 ;Gruber-Baldini et al., 2009 ). These residents with their advancing age, loss of family and other supports, and their severe communication diffi culties are among our most vulnerable and at-risk citizens.
In the United States, the majority of care aides are female (86%), over 40 years of age, born in the United States (77%), earn less than half the US national median annual earnings, and have a high school education or less (55%) (Fredriksen-Goldsen & Bonifas, 2013 ;PHI National, 2011 ;Potter, Churilla, & Smith, 2006 ). Less is known about the Canadian care aide workforce. We were unable to identify a single comprehensive study or report that focused on this workforce and associated demographic characteristics. Individual Canadian studies report that these workers are predominantly women (> 90%), and on average are just over 46 years old (Morgan, Stewart, D'Arcy, Forbes, & Lawson, 2005 ). About half are Canadian-born (Chappell & Novak, 1992 ), two-thirds speak English as their native language (McGilton et al., 2007 ), and in 2003, 63 per cent held a post-secondary certifi cate or diploma (Statistics Canada, 2004 ).
In Canada, there is no national education standard for care aides to enter practice (Berta, Laporte, Deber, Table 1 for more detailed information). With a potentially vulnerable and variably trained workforce providing the majority of direct care for a group of frail, older Canadians with highly complex care and dependency needs, quality of care is a signifi cant concern. For 40 years, reports at the international (OECD, 2005 ;Tolson et al., 2011 ), national (Baum, 1977  In this study, we answered the research question: What are the demographics and select health and work-related outcomes of the unregulated workforce in western Canadian nursing homes? We were also interested in whether these demographic and other characteristics differed by province or owner-operator model in our sample.

Methods
All data were collected within the Translating Research in Elder Care (TREC) study, a fi ve-year research program seeking to identify modifi able features of organizational context that are associated with better resident and staff outcomes in LTC home settings (Estabrooks, Hutchinson, et al., 2009 ;Estabrooks, Squires, Cummings, Teare, & Norton, 2009 ). The breadth and depth of the TREC data collection makes possible secondary studies such as ours that ask a defi ned research question about a select subset of the TREC data.

Sampling and Measures (Nursing Homes)
TREC is situated in 36 nursing homes (30 urban, six rural) in Alberta, Saskatchewan, and Manitoba. All nursing homes that met the TREC inclusion criteria were eligible to participate  ). The TREC team selected nursing homes using stratifi ed (health region, owner-operator model, size) random sampling in order to ensure representation of these dimensions in our sample. For this study, we used data from the 30 urban LTC homes. LTC homes fell into three categories of owner-operator models: (1) public -a facility supported primarily through public funds, owned and operated by the local government; (2) voluntary -a facility run by a voluntary, cultural, or religious organization; and (3) private (for profi t) -a facility in which the individuals or agency in control receive compensation (other than wages, rent, or other expenses) for the services they provide. Our sample of urban LTC homes included 22 small homes (35 to149 beds) and 8 large homes ( ≥ 150 beds). We report here on data collected in 2009-2010.

Sampling and Data Collection (Care Aides)
Care aides completed the TREC survey, a suite of instruments designed to measure (a) demographic characteristics, (b) organizational context, (c) use of best practices, (d) staff outcomes, and (e) factors believed to infl uence the use of best practices. To sample, we used a census of eligible care aides. All care aides who met the inclusion criteria (see Table 2 ) were invited to participate in the survey, and we collected data from all who accepted the invitation . Trained data collectors administered the survey to care aides using computerassisted, structured personal interviews . Care managers and facility administrators completed unit and facility surveys respectively.

Care Aide Measures
In this article, we report on the following variables collected in our study from care aides: demographics, work-related variables, health status and burnout, and continuing education.

Work-Related Variables
We measured (a) job and vocational satisfaction, and (b) whether care aides had enough orientation and job knowledge. We did so by using single items scored on a fi ve-point Likert agreement scale ranging from strongly disagree (1) to strongly agree (5). These single items have produced consistent fi ndings in our previous pilot work with care aides (Boström, Squires, Mitchell, Sales, & Estabrooks, 2012 ) and past studies with nurses (Estabrooks, Squires, Adachi, Kong, & Norton, 2008 ;Squires et al., 2013 ) indicating reliability.

Dementia-Related Responsive Behaviours
We measured these behaviours, exhibited towards staff by residents, by using six items (threat of assault, emotional abuse, physical abuse, verbal sexual harassments, sexual assault, and forced sexual intercourse). Each item was scored as yes or no.

Health Status
The SF-8™ health survey (Ware, Kosinski, Dewey, & Gandek, 2001 ) assesses mental and physical health status using eight items. The eight items have been selected from pools of empirically tested items, and b No = Alberta has a "directory" of care aides. It is the responsibility of the employer to submit information to the directory with regards to those they employ as a care aide, but it is not legislated or included in contracts with service providers. c Home Care is the fi rst setting Ontario is pursuing with its newly developed ( scored on the same norm-based metric as the original larger SF-36 ® scale (Carr, 2003 ). Responses were on a fi ve-or six-point scale. Scoring was done using a proprietary algorithm obtained when permission to use the scale was granted us.

Burnout
The Maslach Burnout Inventory General Survey (MBI-GS) (Maslach & Jackson, 1981 ;Maslach, Jackson, & Leiter, 1996 ) has three subscales (emotional exhaustion, cynicism, job effi cacy). The original MBI-GS contained 16 items. In this study, we used the MBI-GS (short form), which consists of nine items (three items for each of the three subscales, each scored on a seven-point Likert scale). The mean was taken for each subscale. A low risk for burnout is refl ected by one or more of the following: emotional exhaustion score of < 1.67, cynicism < 1.00, and effi cacy > 4.00. A high risk for burnout is refl ected by one or more of the following: emotional exhaustion > 3.00, cynicism > 2.33, and effi cacy < 3.30.

Continuing Education
Care aides were asked how often they attended in-services, workshops, or courses in the past year, and managers were asked if there were one or more clinical educators in the facility.
Additional details on each of these variables are presented in Additional File 1.

Statistical Analyses
We calculated means and standard deviations for interval data and frequency counts and proportions for categorical data. To assess differences between provinces, owner-operator models (private, public, voluntary), and facility size (small, medium, large) on categorical and interval-level variables, we used contingency chisquare tests and one-way ANOVA respectively with a post-hoc test (Bonferroni correction for interval-level variables; e.g., years worked as a care aide, and binary or multinomial logistic regression for categorical variables; e.g., age). To assess for a possible province effect, we calculated statistical signifi cance (using ANOVA p value and effect size) of all work-related and staff outcome variables before and after adjusting for three factors: sex, education (care aide certifi cate), and whether the care aide was born in Canada.
We were also interested to observe if any one of these three sampling dimensions (province, owner-operator model, facility size) were more strongly associated with the three educational and/or professional development opportunities for care aides that were available to us.
To determine this possible association, we conducted two-level individual logistic regressions (one model for having a care aide certifi cate and one for in-services attended) using each of our three educational variables as dependent variables) that controlled for the three sampling dimensions. We compared these same sampling dimensions using the presence of a clinical educator (facility level) with Fisher's exact test.

Ethics
Ethics approvals were obtained from the research ethics boards of all investigator-affi liated universities. Operational approvals were obtained from participating organizations.

Sample Characteristics
A total of 1,381 health care aides (representing approximately 70% of those eligible to participate during the period in which we collected data in each facility) completed the TREC survey in year two (July 2009-June 2010) (see Table 3 ). The majority ( n = 837, 60.6%) were from Alberta, followed by Manitoba ( n = 336, 24.3%) and Saskatchewan ( n = 208, 15.1%). Many worked primarily day shifts (665, 48.2%), with 548 (39.7%) working evening shifts and 168 (12.2%) working night shifts. Just over half of the sample reported speaking English as their fi rst language. We found statistically signifi cant differences between provinces on all demographic characteristics with the exception of age and whether the care aide had completed a high school diploma. Select (but fewer) demographic variables also differed by owner-operator model (e.g., being born in Canada, hours worked in two weeks, and years on unit) and facility size (sex, fi rst language spoken, and being born in Canada) (See Additional File 2).

Work-Related Health and Burnout Characteristics
Mean scores by province for the work-related variables and health and well-being outcomes are presented in Table 4 ; mean scores by owner-operator model and facility size can be found in Additional File 2. The care aides overall reported being satisfi ed with their job and with being a care aide, and perceived that they had adequate knowledge and orientation to carry out their job. They reported experiencing, on average, three types of dementia-related responsive behaviours during the last fi ve shifts worked. Mental and physical health subscale scores on the SF-8 health survey were 51.0 ( SD = 8.7) and 49.4 ( SD = 8.0) respectively. Care aides had moderate risk levels for burnout on two of the three burnout subscales: exhaustion and cynicism. On the third subscale, job effi cacy, aides reported unusually high levels (higher is better).
After adjusting for sex, education (care aide certifi cate), and whether or not the care aide was born in Canada, all variables except physical health status were significantly different ( p < .05) between provinces (see Table 4 ). However, all effect sizes (less sensitive to large sample sizes) were small. To further assess whether a "true" province effect existed, we selected a fi ve per cent random sample. Our sampling strategy ensured suffi cient variation to permit inclusion of province, facility size, and owner-operator model as substantive predictor variables (see Table 4 and Table 5 ). Analysis on this sample resulted in only two variables (adequate knowledge and adequate orientation) displaying statistically signifi cant differences between provinces. Only one variable (MBI Cynicism-Involvement) was statistically different between owner-operator model and facility size after adjusting for sex, education (care aide certificate), and whether or not the care aide was born in Canada (see Additional File 2). Cynicism was higher in private facilities than in voluntary or public facilities; cynicism was also higher in medium-sized facilities than in small or large facilities.

Continuing Education
In Table 6 , we report three variables related to educational opportunities for care aides. The majority (83.6%) of care aides reported having a care aide certifi cate while fewer than 50 per cent reported attending in-services, workshops, or courses regularly in the past year. Signifi cant differences between provinces (see Table 6 ), owner-operator models (see Additional File 2), and facility size (see Additional fi le 2) were noted with respect to both variables: having a care aide certifi cate and attending in-service sessions regularly. In addition, most (73.3%) of the TREC nursing homes had a clinical educator; however, signifi cant differences were present only between provinces (28.6% Saskatchewan, 86.7% Alberta, 87.5 % Manitoba).

SD = standard deviation a Chi-square test for categorical variables and one-way ANOVA for quantitative variables. b The post-hoc test was examined using Bonferroni correction for continuous outcomes and (binary or multinomial) logistic regression for categorical outcomes. Letters a , b , and c denote the post-hoc test (multiple comparison) result for AB-SK, AB-MB, and SK-MB respectively (e.g., a implies that a difference exists between AB and SK).
Table 5 presents fi ndings from the logistic regression models that we ran to determine which sampling dimensions of province, owner operator model, and facility size were more strongly associated with having a care aide certifi cate and attending in-services. Findings show that all three dimensions (province, owner operator model, facility size) were signifi cant predictors of having a care aide certifi cate as well as attending in-services. Owner-operator model (public vs. voluntary) displayed the strongest beta coeffi cient in both models. Only province was a signifi cant predictor of a facility's having a clinical educator in place (see Table 7 ).

Discussion
We report the fi rst data, to our knowledge, for Canadian nursing home care aides on demographic, health status, burnout, job and vocational satisfaction, and continuing educational opportunities. Care aides in Canadian prairie provinces tend to be middle-aged and older women, with a high school diploma; some have care aide certifi cate-level education. In urban centers, half of these workers were born outside of Canada and do not speak English as a fi rst language. Care aides in this study report high levels of job and vocational satisfaction. They have mean scores for physical and mental health from the SF-8 survey consistent with U.S. general population norms (Ware, Kosinski, & Keller, 1996 ). Consistent with other Canadian reports (Boström et al., 2012 ;Morgan et al., 2005 ;Morgan et al., 2012 ), they report regularly experiencing dementia-related responsive behaviours from residents. They are at moderate risk of burnout, consistent with other reports of nurses' burnout in the literature  ). However, their reported job effi cacy -the sense that their work is meaningful and has purpose -is unusually high. This fi nding is consistent with discussions of

Each variable was asked with a single item, scored on a fi ve-point Likert scale (1 = strongly disagree to 5 = strongly agree). d Physical and mental health status were measured using the Health Status Short Form (SF-8) which contains eight items. Responses are on a fi ve-or six-point scale, and scoring is done using a proprietary algorithm obtained when permission to use the scale is granted.
Higher scores indicate better perceived health status. These fi gures are an average across age groupings. e Burnout was measured using the Maslach Burnout Inventory General Survey (MBI-GS), which consists of three subscales (emotional exhaustion, cynicism, job effi cacy), each containing three items. All items are scored on a seven-point frequency Likert scale (0 = never to 6 = daily). A mean is calculated for each subscale. High scores on exhaustion and cynicism with low scores on effi cacy indicate high risk for burnout. f Dementia-related responsive behaviour towards staff is measured by asking care aides to report whether or not they have experienced six kinds of responsive behaviours by a resident in their last fi ve shifts. A count of the kinds of responsive behaviours they indicated experiencing is taken for a total score between 0 and 6.
intrinsic rewards in the direct care workforce by Morgan, Dill, and Kalleberg ( 2013 ) and Rose ( 2003 ).
These fi ndings are similar to those reported in the U.S. studies and in several small Canadian studies. However, our sample reported higher educational levels than U.S. samples -93 per cent of our sample had high school education versus less than 55 per cent in the United States (PHI National, 2011 ). We believe this is the fi rst study to report the health status of the care aide workforce. Although it is positive that these care aides report high levels of job effi cacy in the face of relatively high levels of burnout, their burnout levels are worrisome and are higher than reported levels for regulated workers (e.g., registered nurses) in our larger study .
Burnout presents a threat to staff and has been the focus of intense research for over 35 years (Schaufeli, Leiter, & Maslach, 2009 ). Workplace characteristics affecting care aides -such as frequent exposure to dementiarelated responsive behaviours, high workload, high acuity of residents, and little time to perform tasks for residents -mirror the environmental factors that are reported to precipitate burnout in allied health professions (Josefsson, Sonde, Winblad, & Robins Wahlin, 2007;Stevens, 2008 ). While no review has yet been published examining burnout in nursing home care aides specifi cally, burnout in the nursing and allied health professions presents a threat to quality of care, staff health (Kerr, Laschinger, Severin, Almost, & Shamian, 2005 ), and staff retention . Resource constraints, reductions in proportions of regulated nursing care staff, and a resident population with increasingly complex and high medical and social needs are conditions that will continue to be exacerbated as the longer-living baby boomers experience increasing rates of dementia and move through the health care, social, and residential care systems.
Although efforts are underway in most provinces to establish registries and educational strategies for the nursing home care aide workforce and to stabilize the work environment, an acute need exists to accelerate this work. Effective health human resource planning for the sector cannot proceed in Canada until we have the capacity to count these workers and are aware of their qualifi cations in each provincial jurisdiction.
Our lack of information about the numbers and characteristics of this occupational group reduces our ability to monitor change or progress in achieving workforce improvements .
From our data, we identifi ed potential areas for improvement. We found signifi cant provincial differences in workforce educational levels, continuing education opportunity in the workplace, and the presence of clinical education support. Each of these is modifi able and has the potential to improve quality of care for residents and quality of work life for staff in residential nursing homes. Indeed, the recent report concerning retention, recruitment, and expansion of the capacity of this workforce in the OECD countries (Fujisawa & Colombo, 2009 ) highlighted increasing vocational training and continuing education for non-registered health professionals working in elder care as a promising strategy for retention, recruitment, and enhancement of this workforce.
The following are the strategies, based on our data and others, that will likely have positive results for both residents of nursing homes and their caregivers if implemented: • Mandatory registries of care aides working in residential long-term care in all provinces, and assurance that registration and training/educational requirements (both initial and ongoing) are suffi ciently compatible across provinces to facilitate workforce migration across provincial boundaries. • Processes whereby a percentage of public funding going to nursing homes is earmarked for training and ongoing education of care aides. • Inter-provincial cooperation in developing and implementing health human resource plans -designed to ensure that we have adequate numbers of suffi ciently trained care aides working with appropriate professional skill mixes in nursing homes to provide care that is not  only safe, but at least meets minimum quality standards and, at best, is person-centred and an international exemplar of how care ought to be provided. • A national discussion that addresses the complex and challenging issue of regulation of this workforce and whether it would contribute to safer and higher quality of care for nursing home residents. • Special attention to issues of gender -this workforce is overwhelmingly female (e.g., 93% in our sample), as is the informal caregiver workforce (e.g., family members, partners) and the disproportionate and deleterious effects of long-term care on female caregivers are well-known globally (Fredriksen-Goldsen & Bonifas, 2013 ). However, little evidence exists in Canada or elsewhere that policy planning efforts take gender into account for workforce, service delivery, or other areas. • Special attention to ethnicity and culture as factors in planning for workforce and service delivery needs. The majority of this workforce is non-Caucasian and multi-lingual, with English not the dominant native language. Further, the care aide workforce in most urban nursing homes is a blend of various cultural, ethnic, and linguistic groups, which is in stark contrast to the majority of baby boomers who were born between 1946 and 1964 and are of western European (predominately United Kingdom) extraction (Statistics Canada, 2013a, 2013b, 2013c. Such an ambitious program of reform will require political, administrative, and public will, as well as cooperation across multiple sectors, actors, and groups. However, we are facing rapid increases in the aging population (with accompanying dementia) requiring residential care, with increasing complexity and dependency levels on admission. Admission occurs later in the trajectory of decline, creating much shorter lengths of stay and, consequently, higher levels of resident turnover in the system. All of these factors combine to create a pressing need for consistent and high-quality of life and quality of end-of-life care for these highly vulnerable residents. We must act urgently and with determination.

Strengths and Limitations
While our sample is representative of urban nursing homes in the prairie provinces, we have no assurance that our fi ndings are representative of urban nursing homes in Canada in general, and our care aide sample was not drawn randomly from a known population. We saw provincial differences suggesting that important provincial variations may exist across the country with respect to a number of characteristics. Our fi ndings do not refl ect the situation in rural Canada. Although the surveys were completed using in-person structured interviews and a rigorous quality assurance process , the data are self-reported and therefore subject to the challenges inherent in selfreported data.

Conclusion
At present in Canada, we can offer only a partial and unsatisfying response to the question, "Who is looking after Mom and Dad?" We have an even sparser picture of care aide working conditions, health indicators, and work-life quality indicators -all areas that infl uence the quality of care. Further research will provide partial information; full information and the necessary initiatives to optimize this workforce will require action on the part of policy makers. We entrust the care of the most complex, frail, vulnerable, and challenging charges in our health and social system to this occupational group of care aides 24 hours a day -inaction is not a rational response.