Staff ratios and resident outcomes in special care units: Do activity aids make a difference?

by R. Colin Reid and Neena L. Chappell

Abstract

This study examines the effect of activity aide-to-resident ratios on special care unit resident outcomes over a one year period following admission. Existing studies focusing on the effects of staff ratios on resident outcomes tend not to separate activity aides from other types of staff, if they study them at all. However, research on best care practices in special care units for persons with dementia has established the importance of activities for optimal resident outcomes. Employing multiple regression analyses, this study finds evidence that activity aide-to-resident ratios have positive effects on three resident outcomes - expressive language skill, social skills, and cognitive function - when controlling for resident characteristics, facility characteristics, and environmental design. It is concluded that because activity aide-to-resident ratios tend to be very low, but nonetheless appear to have a significant positive effect on resident outcomes, that a modest increase in activity aide-to-resident ratios would be beneficial for residents with dementia in special care units.

Introduction

In the context of an aging population and the consequent increase in numbers of persons with advanced dementia, the search for effective caregiving strategies for these persons continues to intensify. The expectation that optimum quality of care will be provided in a special care unit (SCU) for persons with advanced dementia hinges on the ability of an SCU to implement a set of philosophical principles of dementia care. The lynchpin between these principles and their successful implementation is the facility staff. Appropriately trained staff in sufficient numbers are a necessary condition for best care. While this has been recognized in recent research (e.g. Davis, 1991; Castle and Fogel, 1998), attention has been focused primarily on registered nurses (RNs), licenced practical nurses (LPNs) and care aide staff-toresident ratios. Effects of other types of staff ratios on quality of care, if addressed at all, are typically included in an overall nursing staff-to-resident ratio (e.g. Johnson-Pawlson and Infeld, 1996). Because activities programming is considered a central principle of special dementia care (Holmes et al., 1994), and because almost all SCUs report having some type of specialized activity programming (Leon, Cheng, & Alvarez, 1997), it is of more than passing importance that the effect of those who are responsible for implementing such programming - activity staff - on outcomes be subjected to evaluative scientific enquiry. This study examines the effect of staff ratios, including and with particular interest in activity staff ratios, on resident outcomes over a one year period for residents of SCUs.

Review Of The Literature

Much research has focussed on the relationship between quality of care in nursing homes and nurse staffing levels, particularly professional staffing levels. Results are generally not in agreement. In a review of studies on the correlates of nursing home quality, Davis (1991) concluded that RNs or professional staff-to-patient ratios contribute more to quality of care in comparison to nonprofessional staff-to-patient ratios. In a longitudinal study of the effects of nursing home attributes including licensed (RN/LPN) and unlicensed (nursing aide) hours on functional ability, discharge home, and death, results showed that licensed nursing hours were associated with improved functional ability, increased probability of being discharged home, and decreased probability of death (Bliesmer, Smayling, Kane, & Shannon, 1998). However, among residents that did not die and were not discharged, no relationship was found between licensed nursing hours and functional dependency. Unlicensed nursing hours were unrelated to outcomes.

Johnson-Pawlson and Infeld (1996) examined whether nursing home facilities with higher staffing levels provided higher quality nursing care as measured by a multi-dimensional deficiency index. The authors found a relationship between total nursing staff ratio and quality of nursing care, however, no relationship was found between the ratio of RNs to residents and quality of nursing care. The authors suggested that no effects were found because of the limited range of RN staffing levels. Castle and Fogel (1998) studied the relationship between nursing home structural measures and outcomes for persons with mental disorders (including dementia). They found that out of six resident outcomes, a higher ratio of nurse aides to residents was associated with improvement in mood, social engagement, cognitive performance scale scores, and ADL scores. However, a lower ratio of RNs to residents was associated with improvement in mood. In addition, a lower ratio of licensed practical nurses (LPNs) to residents was associated with improvements in mood, social engagement and behaviour problem scores. The authors concluded that the beneficial effects of high staff-to-patient ratios may be dependent on the type of staff. Likewise, Spector and Takada (1991) suggest that staff levels related to needs are important for quality outcomes. They studied nursing home characteristics and resident outcomes, finding that residents in facilities with moderate or low nursing staff levels and high mean resident ADL levels were less likely to show functional improvement. These studies show how complex an accurate specification of the effect of staff-to-resident ratios on quality of care can be.

A further complication is added when it is recognized that activity programs can differ tremendously from one unit to another (Grant and Potthoff, 1997). Furthermore, while research supports the idea that specialized programs are essential for the optimal care of residents with dementia (Coons, 1991; Grower et al., 1994), the type of activity program is dependent upon both unit philosophy and availability of resources. According to Maas, Swanson, Specht and Buckwalter (1994: 178) “Activities may become the mainstay of SCU programming depending on the philosophy of the unit”. Activity programs guided by a rehabilitative or psychosocial model, they argue, should result in better outcomes. These models emphasize the provision of activities which are appropriate to the cognitive and functional abilities of residents, are relevant to the resident in terms of “interests, strengths, values and heritage” of the resident, and which promote “socialization, exercise, reminiscing and sensory enjoyment” (Maas et al., 1994: 178). These activities can take the place of wandering, pacing, and a variety of disruptive behaviours. Beuttner et al. (1996) found that, compared to those in traditional activity programs, dementia residents exposed to an enriched activity program showed improvement in strength and flexibility, and a decline in agitated behaviours. Rovner et al. (1996) report similar reductions in behaviour disorders following the implementation of activities programming under psychiatric supervision (see also Aronstein, Olsen, & Schulman, 1996; Cohen-Mansfield & Werner, 1995; Cohen-Mansfield, Werner, & Marx, 1992; Martichuski, Bell, & Bradshaw, 1996).

Whether activities are beneficial to residents with dementia is not a particularly contentious issue in the literature: theoretical and empirical evidence point to the value of activities. A key issue in the successful implementation of activity programs, however, is the availability of adequately trained staff. Insofar as both appropriate philosophy and standards of training can be assumed, the issue becomes one of activity staff-to-resident ratio. The purpose of this study is to compare the effect of activity staff, RN, and care aide-to-resident ratios on six quality of care outcomes within the SCU setting. Specifically, it is expected that the greater the activity staff-to-resident ratio, the less the decline in agitation, cognitive function, physical dependency, expressive language skills, social skills, and affect. The expected effects of RN, and care aide-to-resident ratios are unspecified, given the inconsistency of findings in past research.

Methods and Measures

Data were collected for 257 newly admitted SCU residents with dementia in long term care institutions throughout the province of British Columbia, Canada, as part of the Intermediate Care Facility (ICF) Project. Data collection took place from 1996 to 1998 at baseline and one year post-recruitment. Data were collected for newly admitted residents from the residents themselves, medical charts, family and staff by specially trained research personnel (RAs). Data on staff ratios were collected between 1997 and 1998 in a survey of facility administrators. RAs monitored facilities regularly for fundamental changes in care provided, such as physical environmental alterations, care philosophy improvements, and staff ratio changes. Two facilities were dropped from the ICF Project due to fundamental changes in care provision. Fifty of the 67 SCUs (74.6%) in the province participated in the study and are representative on measures of staff education, provision of flexible care routines, restraint use, and assessment procedures. The mean number of beds in the SCUs is 30, and the mean percentage of residents with dementia is 96 per cent. All newly admitted residents to the participating facilities who the contact person at the facility (typically the director of nursing) indicated suffered from moderate or severe dementia were screened for eligibility. To be eligible, residents were required to have a primary or secondary diagnosis of either Alzheimer’s or vascular dementia, have confirmatory evidence that they in fact had dementia, be unlikely to die or move from the unit in the 12 months following their admission, have the ability to communicate in English, and be at least 65 years of age. Forty eight per cent of all newly admitted residents who the contact person identified as having moderate or severe dementia were eligible. The most common reason for non-eligibility was a lack of specific diagnosis of dementia in the medical chart. Once eligibility was confirmed, signed consent was obtained from either a responsible family member or a guardian. A 91 per cent consent rate was obtained. Six outcomes were measured for residents. To assess change over time outcomes were measured at admission (t1) and at 12 months (t2). The Cohen Mansfield Agitation Inventory (CMAI - short form - Cohen-Mansfield & Marx, 1989) is a 14 item questionnaire that measures the occurrence of three types of agitation during the previous two weeks including:

  1. aggressive behaviours (hitting, kicking, pushing, scratching, tearing things, cursing or verbal aggression, grabbing);
  2. physically nonaggressive behaviours (pacing, inappropriate robing or disrobing, repetitious sentences or questions, trying to get to a different place, handling things inappropriately, general restlessness and repetitious mannerisms); and
  3. verbally agitated behaviours (complaining, constant requests for attention, negativism, repetitious sentences or questions and screaming).

Using a five point scale (1 = never, 5 = a few times a day or continuous for half an hour or more) nursing staff were asked to indicate the number of occurrences of each specified behaviour during the preceding two weeks. These items were summed. Scale range is 14-70 (alpha=.77 for t1; .87 for t2).

A measure of physical dependency, activities of daily living (ADL), was obtained using the Minimum Data Set (MDS), Item E. Item E monitors the level of independence of a resident on a five point scale (0 = independent, 4 = total dependence) for nine ADLs, including: bed mobility; transfer between surfaces; movement between locations; dressing; eating; toilet use; personal hygiene; walking; and bathing. Scale values can range from zero (independent) to 36 (completely dependent). Alpha for t1 is .84 and for t2 is .87.

Mood was measured using the Affect Scale of the Feeling Tone Questionnaire (FTQ). This instrument measures mood in persons with communication difficulties. It is composed of 16 questions and has two components. The first is the actual answer given by the respondent to mood-life satisfaction questions (e.g. Do you have any pain? Do you sleep well?); the second is the affect expressed by the respondent (rated by the RA) for each response on a five point scale from very positive to very negative). These are summed; scale totals range from 16 to 80, with 80 representing the most extremely negative affect possible (alpha=. 87 for t1 and .95 for t2).

The Multi-focus Assessment Scale - Revised (MAS-R) (Crockett et al., 1991) was administered by RAs to assess cognitive (discussed below) and behavioural functioning. Behavioural functioning was measured through social behaviour skills (0-11 point scale; alpha=.77 for t1 and .93 for t2) and expressive language skills (i.e. ability to effectively communicate based on speech, language production and comprehension assessment; 3- 15 point scale; alpha=.84 at t1 and .92 at t2). These are considered outcomes in that they can potentially be affected by the quality of care implemented. For all MAS-R scales, a higher score means higher competence, except for the expressive language skills scale, where a higher score means poorer language skills. Crockett et al. (1991: 15) found interrater reliability to be high (< .95).

Cognitive function is measured by the MAS-R using two scales: early memory and present orientation. Each consists of 11 items. Early memory refers to memories acquired at a relatively young age (e.g. When was WWII? How many years of schooling did you have?). Present orientation measures the ability to acquire and retain current information (e.g. Who is the Prime Minister of Canada? How many people share your room?). The two scores are summed to create an overall cognitive function score. Scale range is 0-22 (alpha=.84 at t1 and .99 at t2). Change scores (t1 minus t2) were calculated for each outcome. For the purpose of having all change scores coded in the same direction (i.e. higher=better), cognitive function and social skills scores are reverse coded.

In the first of the three sets of independent variables, resident characteristics (control) variables are age at admission to the SCU and gender (male = 1, female = 2). Facility characteristics include RN, care aide, activity aide-to-resident ratios, number of beds and profit/not-for-profit status (for profit = 0, not for profit = 1). The latter two variables are controls. Ratios were calculated by dividing the FTE for each type of staff over all shifts by the number of residents. The number of independent variables used in the model are limited by the sample size.

Because environmental features can have an effect on resident outcomes (Cohen-Mansfield & Werner, 1998), an environmental factor is included as a control. RAs completed a modified version of the Therapeutic Environment Screening Scale 2+ - Instrument C (TESS) in each of the participating facilities. The TESS is designed to “evaluate the appropriateness of a nursing home unit for residents with dementing disorders” (Sloane & Mathew, 1990: 22). Instrument C was the most current version of the TESS available at the time of study commencement, consisting of scales that assess general design, maintenance, lighting, noise, residents’ rooms, and programming orientation. Scales were summed in order to calculate a total environmental score for each facility. Possible total scores ranged from 1 to 154 (alpha = .82).

To analyse the data, a separate multiple linear regression equation is estimated for each outcome. Because the t1 score for each outcome variable should be highly predictive of change scores from t1 to t2, it is included as a control variable and forced in first for that outcome’s regression equation. To meet the assumption of normality, three positively skewed variables required log transformations (activity aide-to-resident ratio, number of beds, and t1 expressive language skills), and two negatively skewed variables required a square transformation (t1 affect and t1 social skills). Due to the possibility that the three types of staff ratios may interact, interaction terms were included in the initial equations. None was significant and were therefore not included in the final equations.

Missing cases were non-problematic (less than 6%) for all outcome variables except affect, for which 12.3 per cent (N=20) were missing. Analyses were re-run with only those cases for which there were complete data. Results were the same as when substitution of the mean was used. Results from the latter procedure were therefore used.

Results

Of the 257 residents participating in the study at t1, 162 (63%) remained in the same unit at t2. The majority (65%) of dropouts were due to death. Other notable reasons for dropping out included transfers to other facilities, discharges, hospitalizations and facility dropout (one facility dropped out of the study). At admission, the mean age of the 162 residents was 80 years and 36 per cent of the sample was male. Eighty-six per cent were born in Canada. Forty-five per cent were widowed and 42 per cent were married. Fifty-three per cent lived in their own or someone else’s home before being admitted to the SCU, 11 per cent came from a hospital, 32 per cent from another facility and four per cent from another unit in the same facility. Sixty-five per cent of participants had completed highschool or higher, and the first language of 85 per cent was English.

A comparison of residents remaining in the study and those dropping out before 12 months showed that staff-to-resident ra tios (RNs, care aides, and activity aides) are unrelated to whether or not a resident remains in the study or drops out before 12 months. Dropouts and survivors did not differ at t1 by age, gender, cognitive function, or agitation. Dropouts had statistically significantly poorer social skills, expressive language skills, and greater physical dependency scores at t1 than did survivors, but the differences were substantively small (for example, less than 1.0 difference on a 12 point scale for expressive language skills). Table 1 provides descriptive statistics for all variables in this analysis.

Overall, residents show decline in five of the six outcome variables during the year following admission (Table 2). Residents show decline in cognitive function, physical function, expressive language skills, social skills and affect, but not agitation.

Table 3 shows the results of the six multiple regressions. For physical function, only the time 1 physical function score is predictive of the outcome. In other words, age and gender of resident, RN, care aide and activity staff to resident ratios, number of facility beds, profit/non-profit status and environment have no effect on changes in physical functioning of residents over a one year period beginning at admission to the SCU. The same general results are shown for agitation and affect, with t1 scores showing the only significant effects among all of the independent variables for each respective equation.

Change in expressive language skills is affected by three independent variables: care aide and activity staff-to-resident ratios, and number of beds. The higher the care aide-to-resident ratio, the greater the decline in expressive language skills. Likewise, the more beds an SCU has, the greater the decline. Activity aide ratios are positively associated, indicating less decline in expressive language skills as the ratio of activity staff-to-residents increases. The t1 expressive language skills scores are unrelated to the change score.

Change in social skills is related to activity staff ratio only. The greater the activity aide ratio, the less the decline in social skills. The t1 score is unrelated. Cognitive function outcome scores are also related to activity aide ratio, with higher ratios being associated with better outcomes. Cognitive function at t1 shows a significant effect: the higher a resident’s cognitive function at admission, the greater the cognitive decline on average.

Among the independent variables, t1 scores are predictive of the rate of decline for most of their respective outcomes. Among the four instances where they emerge as the primary predictors of their respective outcomes - physical function, agitation, affect and cognitive function - three are the only significant predictors, the exception being cognitive function. Five independent variables - age, gender, RN-to-resident ratio, ownership status and environment - are not significantly related to any of the six outcomes. Other than the t1 scores, only three of the independent variables are related to any of the dependent variables. Increased care-aide-to-resident ratios and higher number of beds appear to have a deleterious effect on expressive language skills. Of the factors studied here, only activity staff-to-resident ratios appear to mitigate declines in social and expressive language skills, and cognitive function in the 12 months following admission.

Discussion And Conclusion

This study separates the effects of activity aide-to-resident ratios from the effects of other staff types, and from the effects of other structural, individual and environmental variables, on six SCU resident outcomes. Due to the theoretical importance of activity programs in SCUs, it was expected that higher activity staff ratios would result overall in less decline among residents as measured by the six outcomes. The findings reported in this study constitute partial support for this hypothesis. Residents in SCUs with higher activity aide-to-resident FTE ratios tend to show less decline in resident expressive language skills, social skills and cognitive function. RN and care aide-to-resident ratios have no effect on outcomes and the one statistically significant relationship (between care aide-to-resident ratio and expressive language) was inversely associated, that is, the higher the care aide-to-resident ratio, the greater the decline in expressive language skills. This finding contributes to the inconsistent findings reported in the research literature; whether more care aides contribute to increased dependency is unknown.

The causal mechanisms through which increased activity aideto- resident ratios work to slow the decline in social and expressive language skills, and cognitive function cannot be specified in this study. Such an assessment would require much more comprehensive information on program types, degree of implementation of the program, the propensity of individual staff to engage in “caring” rather than only in task performance, unit philosophy, and a host of other factors not available here. In addition, an understanding of the course of the biophysical progression of the disease and its behavioural manifestations in each individual resident (which should in theory facilitate appropriate care responses and therefore the best outcomes), is necessary. Leon (1996) found evidence of this approach in SCUs, reporting that of 214 care facilities studied (both SCU and non-SCU), it was only in the SCUs that staff were under instructions to conduct direct resident observations as a means of facilitating appropriate care. In other words, because residents of SCUs tend to have relatively limited means at their disposal to express their needs, it is up to the staff to monitor and compensate for such communication gaps as they ebb and flow in response to the underlying disease process. The data presented in this study suggest that activity aides may be essential to maintenance of the ability of moderate-to-late stage dementia residents in SCUs to resist the individual communicative decline inherent in the progression of the disease. Residents exposed to higher activity aide-to-resident ratios are thus able to maintain a relatively better quality of life.

The present study is subject to several important limitations. Although the SCUs included in this study are probably representative of all SCUs in the province, there was no opportunity to randomly assign residents to experimental (e.g. activity aides present in facility) and control groups (e.g. activity aides not present). Representativeness of this sample can therefore not be determined. Furthermore, information on program type, in association with unit philosophy, are essential components of any care program which should be included in future research.

Activities are considered central to the well-being of SCU residents with dementia (Holmes, 1996). In apparent recognition of this, SCUs tend to allocate more activity aide hours per resident than do non-SCUs (Grant, Potthoff, Ryden, & Kane, 1998), although wide variation in provision of activity programs (Morris & Emerson-Lombardo, 1994) and program types (Grant & Potthoff, 1997) exists. It is not the purpose of the present study to account for all the possible variations in activity programs. However, even though activity programs of facilities included in this study probably vary widely, it is not unreasonable to expect that the activity aide-to-resident effect on the outcomes examined here would be even stronger if one were able to account for these variations. Chappell and Reid (2000), for example, found that there were almost no relationships between five critical dimensions of care and six key resident outcomes among residents with dementia in long-term care facilities. They argue that none of the facilities implemented all of the dimensions of care that in theory will lead to optimal outcomes. It is therefore improbable that any facilities implement activity programs at the theoretical level required, both in terms of types of programs and staff ratios, to produce optimal outcomes. The associations shown in this study would likely be stronger if the full range of both activity aide-to resident ratios and program types (from none/nonexistent to theoretically ideal) existed and were accounted for in the present study.

The current study found virtually no relationship between care aide-to-resident ratios and outcomes. This finding highlights the theoretical and methodological importance of separating the effects of care aide- and activity aide-to-resident ratios on resident outcomes. The absence of a statistically significant relationship between RN-to-resident ratios and the six outcomes is not particularly surprising. RN staffing is maintained at a low level relative to some other staff types and, consistent with Johnson-Pawlson and Infeld’s (1996) argument, this may prevent the detection of a relationship between staff ratios and outcomes. However, this same argument reinforces the potential importance of activity aide-to-resident ratios for the improvement of resident outcomes since activity aide-to-resident ratios are at lower levels and exhibit less variation than RN-to-resident ratios. That is, despite the restricted levels and variation in activity aide ratios - a condition that tends to reduce the likelihood of detecting an effect - their presence appears to make a difference to residents.

Activity staff to resident ratios are not routinely considered by researchers to warrant independent consideration in the search for determinants of best quality care. This study suggests that a closer look at the benefits of such ratios, at least within the SCU setting, is needed. Compared to RN and care aide-to-resident ratios, activity staff-to-resident ratios are low, suggesting that a significant improvement in outcomes could be realized with minimum increases in staff costs.

An increase in activity aide ratios may also have indirect positive effects. For example, residents participating in organized activities are less likely to engage in agitated behaviours (Buettner et al., 1996; Cohen-Mansfield & Werner, 1995), thus reducing the need for supervision and redirection by other staff types. A reduction in agitation is an indicator of improved quality of life for the resident, and results in a better work place for all staff types. An improved work place should have implications for staff turnover, which in itself has an effect on resident outcomes. Additionally, the findings in the present study may have implications for optimal staff mix, suggesting that a proportional increase in activity aides might be warranted.