Change and Predictors of Change in Social Skills of Nursing Home Residents With Dementia

Abstract

Background

Social skills are of primary importance for those with dementia and their care providers, yet we know little about the extent to which basic social skills can be maintained over time and the predictors of change.

Methods

A total of 18 nursing homes with 149 newly admitted residents with moderate to severe dementia, 195 direct care staff, and 135 family members, in British Columbia, Canada, contributed data on change in social skills from admission to 6 months and 1 year later.

Results

Three-quarters of residents maintained or improved their basic social skills during both the time periods. Decline was explained primarily by cognitive status at the time of admission, notably present orientation. However, staff-to-resident communication becomes more important over time.

Conclusions

Social skills appear to present an opportunity to maintain interaction with these residents. The findings also suggest that a focus on the present orientation before and following admission and on staff-to-resident communication may be beneficial.

The fact that dementia is a progressive syndrome, which can be due to multiple pathologies with no known cure and is devastating for the afflicted individual needs no repetition here. One area affected by cognitive impairment is a person’s social skills and subsequently their ability to interact socially. A minimal level of social skill is necessary for meaningful interaction. Yet, as Lawton ^1,2 noted, interaction is a basic need of those with dementia, even though their context becomes more restricted as dementia progresses. ^3,4 A lack of interaction can also lead to loneliness, ^{5 –7} and we know that those with fewer social skills in nursing homes are less likely engaged with nurses, more likely socially isolated, ⁸ and more likely prescribed antipsychotic drugs. ⁹

Although we know that a decrease in the ability to communicate accompanies disease progression, ^10,11 we do not know the extent to which social skills decline or whether these skills can be maintained or even improved over time. And, as O’Connor and colleagues ¹² noted, there is growing evidence that some of the undesirable impacts of dementia might be slowed through the promotion and delivery of key dimensions of care (such as staff-to-resident communication, resident autonomy, involvement of family, etc). This article reports the findings on change in basic social skills at 6 and 12 months after admission and the predictors of change including measures of the dimensions of care among a sample of nursing home residents with moderate to severe dementia.

Review of the Literature

Social interaction is increasingly recognized as an important component of quality indicators within nursing homes. ¹³ McLean ¹⁴ and Robinson and colleagues ¹⁵ argued that such interaction must be prized over instrumental tasks and that without opportunities for meaningful social interaction, nursing homes are not person-supporting communities. Especially among residents who are cognitively impaired, social skills are important in terms of how others react to them, a consequence complicated by physical limitations, including those conditions that affect the ability to communicate. ⁸ Other well-known problems in the nursing home context include understaffing, which can leave little time for staff–resident social interaction, unsatisfactory resident–resident interaction due to dementia-related communication difficulties, and isolation from the outside world.

The literature ^{6 –8} is replete with terms that fall under the umbrella of social interaction including but not limited to social support, expressive language skills, social skills, relations, and relational interaction. Interaction can be measured in terms of face-to-face contact, all contact (including letter writing, telephoning, Skyping, and texting), quality of the interaction, and relationship-specific interaction (with spouses, daughters, etc). Most of this focuses on the extent and/or quality of interaction. Concern in this article is directed toward the individual’s basic social skills or, said another way, the person’s competence to engage in social interaction. For example, does she or he respond appropriately when spoken to, acknowledge another’s comment, look at the person speaking to him or her?

The extent to which care can improve or slow the deterioration in social skills among those with dementia is unknown. Much of the research focuses on other aspects of interaction. One study in the United Kingdom ¹⁶ found that a diagnosis of dementia is unrelated to social withdrawal while impairment in activities of daily living (ADL) is related. Similarly, the degree to which individual characteristics are related to change or lack of deterioration in social skills is also unknown. Chappell and Reid ¹⁷ reported that female residents with dementia are less likely than male residents to exhibit decline in expressive language skills over time. Although older age and lower education have been identified as potential risk factors for dementia, ^18,19 their relevance for the social skills of those with dementia has not been investigated. It is intuitively plausible that social support, such as number of friends/family visits per week, and their relationship to the resident (spouse, daughter, etc), is important for maintaining basic social skills.

Determining which aspects of the nursing home environment might also affect social skills is challenging. Basic social skills are seldom studied as a care outcome despite their obvious importance for person-centered care. ^2,20 In the United States ^{21 –23} and Canada, ¹⁷ all studies with rigorous scientific designs report little if any empirical relationship between the dimensions of care and a variety of other (ie, not social skills) outcomes (such as expressive language skills, functioning, agitation, etc). ²⁴

Most research linking dimensions of care with resident outcomes is restricted to the study of a single dimension of care with an emphasis on staffing issues being critical to the provision of optimal care for those with dementia. ²⁵ Studies ^26,27 have linked staff education to a reduction in physical aggression toward nursing staff and quality of care that is measured, for example, by completing the basic physical tasks of caring. Early research ²⁸ also linked professional staff-to-resident ratios to the quality of resident life. Castle and Fogel ²⁹ reported that a higher ratio of nurse aides but a lower ratio of licensed practical nurses (LPNs) to residents was associated with improvement in social engagement. Reid and Chappell ³⁰ however did study social skills and found that the ratio of nursing staff to residents is unrelated to change in social skills. They also reported a significant overall decline in social skills during the first year of residency.

Other aspects of the nursing home environment (such as for-profit status and percentage of residents with dementia) have not been studied in relation to social skills. Size of facility (number of beds) has been shown to affect resident outcomes in areas other than social skills. Day and Calkins, ³¹ for example, found that smaller units (6-12 vs 20-60 residents) led to decreased overstimulation for persons with dementia. Other research ^32,33 found that smaller units are associated with benefits such as reduced anxiety, depression, and psychotropic drug use as well as increased mobility.

In summary, basic social skills are important for residents with dementia in nursing homes if they are to optimally engage/interact with staff and others, yet surprisingly little research attention has been devoted to this area. Indeed, basic information concerning whether and the degree to which change in basic social skills takes place over time remains undocumented. Despite the lack of empirical evidence, there are many factors that could potentially be important for the maintenance or enhancement of residents’ social skills including individual factors such as gender, age, and social support and institutional factors such as dimensions of care, staffing, size of facility, and profit/not-for-profit status. The cognitive impairment itself may also be relevant for both level of and change in level of basic social skills.

Purpose of the Present Study

This study examines change, and predictors of change, in social skills among residents with moderate to severe dementia in nursing homes in British Columbia, Canada. Change is measured from admission to 6 and 12 months later.

Methods

Study Design

A 2-level (facility and individual) prospective design was implemented in an observational cohort study. ³⁴ Power analyses indicated that 14 facilities with 8 residents at each would be sufficient to detect a moderate relationship between variables, with statistical power of .80 at the .05 level of significance and taking into account a 10% attrition rate. Such a multimethod approach was necessary given the complexity of the questions being asked. ³⁵ Ethics approval was received from the university and regional health authorities.

The Institutional Level

In total, 18 nursing homes were recruited, chosen to reflect diversity in terms of size, type of ownership, and percentage of residents with dementia. Facility-level data were collected using interviews with facility staff and family members and averaging scores for each facility. As shown in Table 1 panel A, just over a fifth (22.2%) of the facilities were public, 44.4% private, nonprofit, and 33.3% private, for profit. Over three-quarter (77.8%) had specialized dementia care units; 61.1% permanently assigned their staff to residents, and most (77.8%) were unionized. Although few (11.1%) required staff to have dementia care training prior to employment, as is evident subsequently, most nevertheless had received prior dementia care training. Facilities ranged in size from 49 beds to 300, and ranged from 37% of their residents with dementia to 100%.

Table 1.

Selected Sample Characteristics.

	N	%
(A) Facilities (N = 18)
Ownership
Public	4	22.2
Private, nonprofit	8	44.4
Private, for profit	6	33.3
Have specialized dementia  unit	14	77.8
Staff permanently assigned  to units	18	100.0
Staff permanently assigned  to residents	11	61.1
Staff required to have  dementia training prior to  employment	2	11.1
Staff unionized	14	77.8
# of beds	mean = 127; range 49-300
# of units	mean = 5.2; range 1-16
% residents with dementia	mean = 75.1%; range 37-100
(B) Family members (N = 135)
Female	86	67.0
Relationship
Daughters	62	45.9
Sons	34	25.2
Wives	13	9.6
Husbands	10	7.4
Other (eg, nieces)	16	11.9
(C) Direct care staff (N = 195)a
Type of staff
Resident care aide	124	65.6
Health care workers/aides	28	14.8
Personal care attendants	18	9.5
Nurse’s aides/RAs/LPNs	19	10.0
Female	170	89.0
English first language	183	96.3
Years at facility
>7	87	45.5
1-7	84	44.0
<1	20	10.5
Type of shift
Day shift	78	40.8
Evening shift	46	24.1
Day and evening	54	28.3
Night shift	13	6.8
Full-time status
Full-time regular	114	47.0
Part time	47	24.6
Casual	30	15.7
Dementia training	178	93.2
Type of unit
Special care unit	67	35.3
Regular care unit	51	26.8
Both	72	37.9
(D) Residents (N = 149)
Female	111	74.5
Ever married	148	99.3
Born in Canada	108	72.5
Completed high school	101	67.8
Care subsidized	120	80.5
Moderate dementia	95	63.8
Moved from
Another facility, hospital,    or congregate living	99	66.4
Own home	41	27.4
Someone else’s home	7	4.7
Mean age at admission	85.1 years (SD = 6.7)

Abbreviations: LPN, licensed practical nurse; RAs, research aides; SD, standard deviation.

^a This is the total N, but some respondents did not respond to individual items; therefore, the percentages are based on the number of respondents to the particular item. Missing values range from 0% to 6%.

Upon admission, the resident’s family member (1 per resident) was given a letter describing the study, then approached to ask their interest in participating. These data were collected at 6 months postadmission, once the resident was settled in and the family member had become familiar with the facility. Of the 149 residents in the study, 135 (90.6%) family interviews were completed. Instances of noncompletion (9.4%) were due to refusal, inability to contact, or prior death of the resident. The mean age of family respondents was 60.7 years (standard deviation [SD] = 11.9 years), and most were female (63.7%). Almost half (45.9%) were daughters, approximately a quarter (25.2%) were sons, 9.6% were wives, and 7.4% were husbands (See Table 1 panel B).

Staff were recruited by research assistants who spent time inside facilities, approaching those who entered the lunchroom. The effect of this approach on representativeness of staff samples is unknown, but adequate numbers of staff participants were recruited. Information on dimensions of care was collected from a total of 195 direct care staff provided, including resident care aides (65.6%), health care workers/aides (14.8%), personal care attendants (9.5%), and nurse’s aides/registered nurses (RNs) and LPNs (10.0%). All participating staff provided personal care such as bathing, toileting, dressing, and feeding to the residents. The LPNs performed more medically oriented tasks such as dressing changes or injections. All staff except LPNs and RNs (who undergo a different training regimen) are typically required to complete a 6-month course offered through a public or private institution. The vast majority (89.0%) of these staff were women, English was their first language (96.3%). Almost half (45.5%) had worked at their current facility for more than 7 years. The largest category (40.7%) of respondents was comprised of regular day shift workers. Over half (59.7%) were full-time regular employees. Almost all (93.2%) had some dementia training. A third (35.3%) worked in a special care unit; 26.8% in a regular unit; and the remaining 37.9% in both (See Table 1 panel C).

The Individual Level

Residents were recruited in the order in which they were admitted to each facility, until the minimum required sample size was reached; the number of residents per facility ranged from 4 to 17 (mean [M] = 8.28, standard deviation [SD] = 3.68); in total, 149 newly admitted residents participated. Inclusion criteria for the residents were new admissions to nursing homes who had a primary or secondary diagnosis of moderate to severe dementia, including but not exclusive to Alzheimer’s disease, identified initially by directors of nursing and confirmed with the Mini-Mental Status Examination (MMSE; standard cutoff of 24), not having a comorbid condition likely to lead to rapid deterioration or death (such as advanced cancer), and not having select primary diagnoses such as schizophrenia or bipolar affective disorders. Data for individual-level variables were collected using direct observations of residents, review of medical charts, and standardized neuropsychological and functional examinations of the residents by trained research staff.

Table 1 panel D shows selected characteristics of the 149 residents. Most (74.5%) were female. Their average age at the time of admission was 85.1 years (SD = 6.7, range = 66-101). All but 1 (99.3%) were ever married. The majority (63.8%) scored in the moderate severity range on the MMSE (scores 10-20) and 35.6% of the residents scored less than 10. Most (72.5%) were Canadian born (18.8% in British Columbia). A majority (67.8%) completed at least high school education. For a large majority (80.5%), their stay in the facility was partially or fully (4.7%) subsidized, and for 12.8%, it was not subsidized. Most (66.4%) lived in a congregate living setting prior to admission; 27.4% lived in their own homes; and 4.7% lived in someone else’s home.

The Dependent Variable

The dependent variable is the individual resident’s basic social skills. It was measured with the social behavioral skills subscale of the Multi-Focus Assessment Scale Revised, ³⁶ a “user-friendly” instrument developed for evaluating older adult residents of nursing homes. It is designed to assess primary abilities rather than complex cognitive processes. The subscale consists of 11 items that assess the resident’s basic social skills and is scored based on the resident’s interaction with the interviewer in the resident’s own environment. The first 6 items (any discernible response to greeting; if response is verbal and appropriate; offers hand to interviewer; assents or acknowledges comment; any discernible response to any question; and if response is verbal and appropriate during assessment) can often be scored in the initial rapport-building period. The next 4 items (the absence of inappropriate grimaces and mannerisms; if the resident looks the interviewer in the eye at any time; the resident does not obviously avoid eye contact or stares fixedly at the interviewer; and absence of resident sitting in a bizarre position or in constant motion or is nearly motionless) are scored on the basis of the resident’s behavior throughout the interview. Another item (resident acknowledges interviewer’s departure either verbally or with a gesture) is scored after the interview has been terminated. All items are scored dichotomously, with a “1” assigned for appropriate social behavior and a “0” otherwise, for a possible score range of 0 to 11. Interrater stability ranged from .81 to 1.0, and interrater reliability ranged from .78 to .96. In the sample studied here, Cronbach’s α equaled .81 at admission, .76 at 6 months, and .90 at 12 months.

Institutional-Level Predictors

At the facility level, the independent variables included measures of 5 domains of care, using the individualized care (IC) instruments developed by Chappell and colleagues. ^37,38 One domain is staff-to-staff communication, consisting of 10 statements with responses on a 4-point Likert-type scale (never to always; eg, to what extent do you share personal information about residents that may help other staff members make sense of resident behavior; to what extent do you share care approaches that can help residents to do things for themselves, etc.; α = .80.); a second domain is staff-to-resident communication (α =.71), consisting of 3 items with a 4-point Likert-type response scale (never to always) asking staff whether they talk to residents about social events that are going on within the facility; talk to residents about what is happening outside of the facility; and talk to residents about their personal lives. The third domain is knowing the person, measured by 4-point Likert-type scale ratings on 11 statements such as I read the social histories of the resident care plans and I do not know the behavior patterns of individual residents (α = .68). The fourth domain is choice and autonomy, consisting of 11 items with responses forming a 5-point Likert-type scale (never to very frequently; eg, Do you feel that the facility you work in supports the independence of residents, do you feel that the facility you work in offers choice in activity programming, etc; α = .75.). All responses refer to the previous 7 days and are asked of care attendants.

The fifth domain, family involvement, ³⁹ is measured with 2 scales consisting of 20 items related to care provision and importance, rated on a 4-point Likert-type scale by family members. For each item (eg, I’ve been asked about my family member’s personal history), the individual rates the extent to which it is provided by the facility (the F-INVOLVE scale; α = .84), and then how important it is to them personally in the care of their family member regardless of whether the facility provides it or not (the F-IMPORT scale; α = .86) for 17 of the 20 items (3 items were deleted to increase internal consistency).

Other facility variables recorded were number and percentage of persons with dementia; for-profit/not-for-profit status; special care unit/nonspecial care unit status; number of beds; whether staff are permanently assigned to residents (yes/no); whether staff are required to have dementia training prior to employment (yes/no); number of hours of dementia training required by staff; whether staff are unionized (yes/no); hours of direct care per resident day; and staff turnover during a 12-month period (RNs, LPNs, and resident care attendants).

Individual-Level Predictors

In addition to basic demographic variables (age, gender, and number of years of formal schooling), a number of resident-level variables were measured. Two memory subscales within the Multi-focus Assessment Scale for the Frail Elderly Revised (MAS-R) were an 11-item subscale on early memory (eg, When was World War II?, How many years of schooling did you have? α = .78 at admission, .81 at 6 months, and .85 at 12 months) and an 11-item subscale on present orientation (eg, Who is the Prime Minister of Canada? How many people share your room? α = .61 at admission, .74 at 6 months; and .75 at 12 months). Functioning was measured in terms of ADL using the Minimum Data Set, version 2, item E, developed for nursing home populations. ⁴⁰ Residents are scored on a 5-point scale (ranging from independent to totally dependent) for each of the 9 ADLs—bed mobility, transfer between services, movement between locations, dressing, eating, toilet use, personal hygiene, walking, and bathing—and a total score is computed for each time period. Variables related to the resident’s social interactions were collected from family members and included size of the resident’s social network (number of different people visiting) and number of visits per year from family and friends from outside the facility. Within the facility, information was gathered on whether the resident participated in pet therapy (yes, no), exercise (times per week), music (times per week), and mental stimulation (times per week), all from chart audits.

Data Analyses

To analyze the relationships between changes in residents’ social skills and resident-level and facility-level predictors, we fitted hierarchical linear models (HLMs) using HLM 7.0. Although the intercept-only (baseline) model indicated that overall the residents’ maintenance of social skills did not differ across the facilities and were primarily due to within resident variation (the intraclass correlations were .00012 for the 6-month time interval and .00024 for the 12-month time interval), we nonetheless used HLMs for our main analyses, because we were interested in examining the relationships among variables at the 2 levels simultaneously. ⁴¹

To select relevant predictor/covariate variables at the resident level, bivariate correlations were estimated between changes in the social skills scores (from admission to 6 months and from admission to 12 months) and all the resident-level variables. To select potential predictor variables at the facility level, the residents’ changes in social skills scores were summarized for each facility (in 2 ways described later) and correlated with the facility-level predictor variables. The exploratory analysis feature in HLM 7.0 was then used to finalize the best-fitting model. All HLM model estimates used restricted maximum likelihood estimation and robust variance estimates. In general, all statistical tests were carried out at the .05 level of significance, without adjusting for type I error for multiple comparisons. As these were exploratory analyses, we were willing to risk a type I error in order not to miss a potentially real relationship between variables. Three nondiscriminating variables were not used in the analyses, marital status of the resident, first language of staff, and whether staff had dementia training.

Results

Residents’ average social skills scores from the MAS-R were 10.53 (SD = 1.29) at admission, 10.29 (SD = 1.43) at 6 months, and 10.16 (SD = 2.03) at 12 months. As the maximum possible score on the MAS-R is 11, the distributions were highly skewed. We therefore assessed changes over time in 2 ways. First, difference scores were computed for each resident for the time intervals from admission to 6 months and from admission to 12 months. Second, we computed a dichotomous score by assigning a 1 to a resident if his or her social skills scores remained the same or improved in a given time interval or zero if the resident deteriorated (note 1). Given that all residents had moderate to severe dementia at admission, maintaining social skills level was considered to be progress. The proportions in Table 2 show that in the first 6 months, the majority (63.8%) of residents improved, and at 12-months, the majority (59.8%) had maintained their social skills. Overall, most residents maintained or improved between admission and 6 months (75.4%) and between admission and 1 year (73.0%), but considerably fewer improved over 12 months compared to 6 months (13.2% vs 63.8%).

Table 2.

Changes in Social Skills Over Time: Number (Percentage) of Residents.

	Admission to 6 Months	Admission to 12 Months
% residents who improved	88 (63.8%)	16 (13.2%)
% residents who maintained	16 (11.6%)	73 (59.8%)
% residents who deteriorated	34 (24.6%)	33 (27.0%)
N	138	122

Correlations With Resident-Level Variables

For both time intervals, the difference scores and the dichotomized scores were uncorrelated with the demographic (age, gender, and education level) and the activities variables (pet therapy, exercise, music, and mental stimulation). The difference score was also uncorrelated with resident’s MMSE score, family involvement measures, and length of residence at a previous facility. The dichotomized change score was uncorrelated with size of the resident’s social network and total number of visits (note 2).

There were significant correlations between social skills difference scores and several resident-level variables. Between admission and 6 months, larger decreases in social skills were related to smaller resident social networks (r = .27, P < .01) and fewer visits per year (r = .22, P < .05). Also, residents with larger decreases in social skills scores at 12 months had fewer total visits (r = .28, P <.01). It appears that regardless of the length of the time period, a higher level of interaction was correlated with a smaller decline in social skills (see Table 3 panel A).

Table 3.

Bivariate Correlations of Changes in Residents’ Social Skills.

Correlation With	Admission to 6 Months	Admission to 12 Months
(A) Differences scores
Resident-level characteristics
Smaller social networks	.27 (P = .010)	ns
Fewer visits/year	.22 (P = .039)	.28 (P = .014)
(B) Dichotomized scores
Resident-level characteristics
MMSE	.33 (P = .001)	.16 (P = .080)
MAS-R early memory	.26 (P = .003)	ns
MAS-R present orientation	.29 (P = .001)	.27 (P = .003)
MAS-R total orientation	.30 (P = .001)	.17 (P = .070)
Facility-level characteristics
IC SR comm	ns	.51 (P = .031)

Abbreviations: IC, individualized care; MAS-R, Multi-focus Assessment Scale for the Frail Elderly-Revised Form; MMSE, Mini-Mental Status Examination; ns, not significant; SR comm, staff-to-resident communication.

Table 3 panel B shows the significant correlations between the dichotomized deterioration scores and the resident-level variables. For the 6-month time interval, significant correlations were found with the MMSE (r = .33, P < .001), early memory (r = .26, P < .01), present orientation (r = .29, P < .001), and total orientation (r = .30, P < .001). The worse residents scored the more likely their social skills deteriorated. From admission to 12 months, the MMSE approached significance (r = .16, P = .08), as did total orientation (r = .17, P = .07). Early memory was not significant but present orientation was (r = .27, P < .01). All these scales measure cognitive function, but the different scales in the MAS-R unpack the aspects of cognitive function, telling us what type of cognitive function is related to change in social skills. In the longer term (ie, 12 months), it is present orientation.

Correlations With Facility-Level Variables

To examine the bivariate relationships between social skills changes and the facility-level variables, changes in socials skills were summarized in 2 ways for each facility (1) the social skills change scores were averaged across residents at each facility and (2) the proportion of residents who did not decline was computed for each time interval. None of the mean difference scores was correlated with any of the facility-level variables assessed. The proportion of residents whose social skills had not declined, however, was related to staff-to-resident communication (Table 3 panel B; note 3). Facilities where fewer residents exhibited decline in their social skills tended to have higher IC staff-to-resident communication domain scores (r = .52, P < .05, from admission to 12 months). Significant correlations with the proportions, and not the mean differences, are not surprising. Averaging the residents’ difference scores within a facility likely cancels any observed changes, as some residents improve while others worsen resulting in an average close to zero, reducing any systematic variances across facilities.

Hierarchical Linear Model Fitting

The resident- and facility-level variables correlated with changes in social skills scores (with P < .10, see Table 4) were entered as predictors in a series of HLMs. The (binary, uncentered) outcome variable was the dichotomized social skills scores. To aid comparisons in a systematic manner, if a predictor was significantly related to social skills in either time interval, a model was fit with that predictor for both time intervals.

Table 4.

Results of Best-Fitting Hierarchical Linear Models for Predicting Changes in Social Skills Across Time.

	Admission to 6 Months			Admission to 12 Months
	Odds Ratio	95% CI	P Value	Odds Ratio	95% CI	P Value
IC staff-to-resident communication	1.551	0.930-2.588	.087	1.748	1.054-2.897	.032
MMSE	1.156	1.077-1.241	<.001	1.076	0.972-1.191	.157

Abbreviations: CI, confidence interval; IC, individualized care; MMSE, Mini-Mental Status Examination.

Including the intercept-only (baseline) models, a total of 10 HLM models were tested for each time interval. Because both the MMSE and the MAS-R are measures of cognitive functioning, separate models were tested using these as covariates. In addition, because we have a particular interest in the dimensions of care, we also examined models with the 4 IC domain facility-level variables. No interactions between variables across levels were examined (note 4). In all cases, the models with the MMSE as the resident-level covariate produced better fitting models. Table 4 shows the results of the best-fitting HLM.

For the 6-month time interval, the resident-level MMSE was a stronger predictor of maintenance of social skills (t ₁₃₀ = 4.056, P <.001). The odds ratio of 1.156 indicates that for each increase in the resident’s MMSE score upon admission, the resident is about 16% more likely to maintain his or her social skill for 6 months following admission to the facility. The facility’s staff-to-resident communication score was marginally significant (t ₁₆ = 1.820, P = .087). An increase in the facility’s communication score makes it about 55% more likely that the residents will maintain their social skills.

The picture changes slightly at 1 year. Twelve months following admission, residents’ MMSE scores no longer predict whether the resident will maintain his or her social skills (t ₁₃₀ = 1.424, P = .157). However, the facility’s staff-to-resident communication does (t ₁₆ = 2.341, P = .032). For each increase in the communication score, the residents are 75% more likely to maintain their social skills.

Although we expected that family involvement would be related to changes in social skills over time, it was not a significant predictor of change in social skills in this sample of residents.

Discussion

For the sample of residents studied here, approximately three-fourth either maintained or improved their basic social skills from admission to 6 months and from admission to 12 months. Only one-quarter deteriorated in their social skills. However, the proportion exhibiting improvement slowed by 12 months, and the proportion displaying decline increased. This suggests that, among those with moderate to severe dementia, basic social skills do not deteriorate especially quickly. This presents an opportunity to maintain interactions with residents, suggesting the feasibility of a person-centered approach that incorporates relations with others. This finding constitutes support for those who argue that meaningful interaction with these residents is required for the provision of adequate care (2, 11, 20), recognizing that the social skills measured here are very basic.

The decline in social skills which did occur, especially between admission and 6 months, was primarily explained by the residents’ cognitive status at the time of admission. Some of the decline may be slowed by staff interacting with residents, indicating potential lines of inquiry for the future. Once cognitive status was taken into account, maintenance of social skills was uninfluenced by the following individual-level variables: social network size; number of visits; and family involvement. This may be related to the fact that these are measures of social interactions between residents and persons from outside the facility and therefore are less salient for the maintenance of basic resident social skills. Communication with staff represents ongoing daily life, so it makes sense that it would be more important for basic social skills. It also appears that staff-to-resident communication becomes more important over time. These results, however, do not tell us the directionality of this relationship, whether staff communicate more with residents who have maintained their social skills, or residents maintain their social skills when staff interact with them more? However, the stronger relationship at 12 months postadmission suggests it is important that staff continue to interact with residents whose social skills may be at risk of deteriorating.

Given the importance of the resident’s present orientation for change in social skills, these data suggest the need to focus in this area. Given that transfers into facilities are notoriously difficult for many with dementia (and others), resulting in negative outcomes including increased risk of mortality particularly during the first 3 to 6 months following admission, ⁴² attention targeted toward resident orientation before and following admission may have the added benefit of maintaining resident social skills over the longer term.

Given the high-social skill scores obtained on the MAS-R in this population of residents with moderate to severe dementia, more research is required on the measurement of basic social skills. On the one hand, it may be that very basic social skills such as those measured by the MAS-R are relatively intact even among those with more advanced dementia. If this is the case, it is good news for those advocating more person-centered care that requires considerable interaction with the resident. On the other hand, the instrument itself may not be sufficiently sensitive to pick up important declines in basic social skills. It is also possible, especially given the greater declines at 12 months, that a longer follow-up period would have seen much greater maintenance (rather than improvement) and even greater declines. This area is in need of more research.

Some might object to the use of the MMSE, given concerns expressed over its influence by certain sociodemographic and health variables such as lower socioeconomic levels, lower education or literacy, depression, and memory complaints. ⁴³ It is, however, widely used as a reasonable measure of cognitive impairment, especially at the moderate and more severe levels. ⁴⁴ In our study, similar results were obtained with the memory subscales of the MAS-R, strengthening confidence in the results obtained with the MMSE. That is, our results appear not to be dependent on the particular measure.

Research with persons with dementia is known to be difficult; this study is no exception. The sample is small and from 1 part of the world. A further limitation is the unknown degree of selection bias in each of the samples—family, resident, and staff. Although the response rate of family members was excellent (90.6%) and should serve to mitigate selection bias, nonparticipants may nevertheless represent a qualitatively distinct group. The staff sample was one of convenience. Future research should investigate the means to obtain a representative sample in a cost-effective manner among a workforce that is routinely understaffed and highly stressed.

Future research should also focus attention on inclusion of the direct voice of the resident. The present study relies on indirect measures, but recent progress has been made on methods that allow residents a voice. ⁴⁵ The finding that most residents maintain at least basic social skills suggests this is feasible for many.