Bourgeois et al.  explored a variety of techniques for improving the understanding of verbal interactions between NAs and residents with dementia. They found that after a single 10-minute conversation about quality of life topics supported by written cue cards (augmented verbal procedure), residents’ and nurse aides’ ratings demonstrated greater agreement on quality of life indicators as measured by the Dementia Quality of Life scale (DQoL)  compared to their pre-test ratings. In a follow-up study, Bourgeois, Camp, & Zeisel   tested two procedures delivered by nursing staff for eliciting opinions about quality of life indicators from long-term care residents with dementia. These involved: 1) a verbal enhanced condition consisting of giving printed questions to NAs to guide verbal discussion of QoL topics; and 2) a category sorting activity in which visual cues (labeled photos and a preference template indicating amount of preference for items/topics represented in the photos) were used to elicit discussion between NA-resident with dementia dyads.
During coding of the experimental sessions’ videotapes, it was observed that some residents did not appear to understand the instructions and questions posed to them by their NA during their interactions. Upon further review of the sessions it became apparent that some interactions reflected repetitions of instructions, repeated questioning by the resident, and very little actual information transfer. The NAs in these sessions were observed to speak in heavily accented dialects, mostly other languages (e.g., Swahili, Haitian Creole). A total of 19 of 60 NAs were determined to be non-native English speakers. Data then were analyzed using the additional factor of language matched (NA and resident were native English speakers) and language mismatched (NA was a non-native English speaker and resident was a native English speaker).
Participants were recruited from two long-term care facilities in the Greater Cleveland, Ohio area and one long-term care facility in Columbus, Ohio. The sample consisted of both assisted living and skilled nursing care residents. The research team met with the facility administration to outline the study inclusion criteria and then sent a targeted mailing to the families of residents that explained the study in detail. Resident enrollment criteria included: documented dementia diagnosis, between the ages of 65 - 85, and passing scores on functional vision, hearing, communication, and reading screenings. Exclusion criteria include: documented neurological or psychiatric condition that would impede communication (e.g., aphasia, psychosis, tardive dyskinesia), blind, deaf, does not speak English, or scores higher than 24 on the Mini-Mental Status Exam (MMSE; Folstein, Folstein, & McHugh, 1975)). When the research team received a signed consent form from the family or guardian, the resident was screened using the instrumentation described below. Residents who passed the screening were then randomly assigned to either the VMC or the Control condition. A nursing assistant regularly assigned to the resident was identified and research staff met with him/her to explain the study and obtain consent to participate. At all but one of the facilities the nurse assistants participated in the VMC activity during regular working hours. One facility insisted that the nurse assistants complete the activity on their own time. These nurse assistants worked with the residents either before or after their shift and were compensated for their time.
A total of 37 residents with dementia were recruited for the study and completed all four assessments, (along with their NA dyad members). Their mean MMSE score was 15 (range = 7 - 24); 78% were female, 97% were Caucasian, and 3% were African American; their mean age was 86 years (range = 67 - 96). All were native English speakers and passed the screening procedures.
2.3. Nursing Assistants (NAs)
A total of 33 nursing assistants took part in the study (a small number of NAs worked with two resident participants). Their mean age was 42 years (range 23 - 65); 97% were female and had worked as a nursing assistant for an average of 13 years (range 1 - 30) and at the facility for an average of 8 years (range 1 - 27). Native English speaking aides accounted for 54% (N = 20) of the sample and the non-native English speaking aides (N = 17) listed the following as their native languages: Jamaican, Ghanaian, Spanish, Philippine, Ukrainian, Tagalog, Visaya, Twi, Italian, French, Yoruba, and Haitian.
2.4. Screening Procedures
Resident screening measures consisted of a vision screen, hearing screen, communication screen  to determine functional abilities sufficient for participating in the subsequent interview procedures. This was followed by the MMSE  and the WRAT 4  to screen for reading ability at the word level.
There were no screening measures developed for the nurse assistants; a willingness to participate and a signed informed consent document were the only requirements, along with the ability to read and speak English.
The Card Sorting Activity for the Control condition consisted of two copies of each of the VMC™ cards; to begin, 12 cards (6 sets of matching cards) were placed face down on the table in front of the resident. NAs engaged the resident in a matching game, taking turns to uncover two cards at a time until all cards had been matched. At that point, 12 new cards were used from the VMC cards. This continued until 10 minutes had elapsed.
2.6. Dependent Measure
A Preference Assessment Questionnaire (PAQ) was created for use as the primary outcome measure. As shown in Appendix 1, this measure contained five domains (food, activities, daily living, socializing, and pain). Within each of the five domains there were five items listed, and each item was associated with three possible response choices, which were scored as: Always (2), Sometimes (1), and Never (0). This measure reflected the items and content of VMC.
The PAQ was administered separately by researchers, and concurrently, to each member of resident-NA dyads on four occasions. For the NAs on each of these occasions, they were handed a form with the 25 items printed in English along with the rating choices (Always; Sometimes; Never) for each item. For residents on each of these occasions, a printed version (42-point type size) of the rating choices was placed in front of the residents and each item was read verbally to the residents, who then indicated which rating reflected their opinion of the item.
For each occasion, a Convergence Score was created by taking the absolute value of the difference between the resident’s and NA’s responses on each item of the Preference Assessment Questionnaire, and then adding these values to create a Convergence Score. For example, if a resident responded that Bingo was never liked (a score of 0) and the resident’s NA thought that the resident always liked Bingo (a score of 2) then the Convergence Score for this NA-resident dyad would be 2 (absolute value of [0 - 2]) for this item. This total Convergence Score thus could range from 0 to 50. Lower total Convergence Scores reflect greater overall convergence between members of a dyad.
When residents passed the screening procedures and their NA consented to participate, research staff administered the Preference Assessment Questionnaire (Pretest 1) to both members of the dyad as described previously. Research staff asked residents to verbally indicate how often they enjoyed or experienced each of the 5 items in the 5 categories; NAs were asked to indicate how often they thought their resident enjoyed or experienced each of the same 25 items. Next, the experimental group NAs prompted their resident to sort labeled VMC™ picture cue cards into rating categories for up to 10 minutes; Control group NAs engaged their resident in the card matching activity for 10 minutes. The Preference Assessment Questionnaire was then re-administered to dyad members separately (Posttest 1). Thus, data were gathered for Pretest 1 and Posttest 1 immediately before and after either the NAs presented VMC or control procedures to their residents.
2.9. Treatment Fidelity Measures
Conversations between nurse assistants and residents were audio recorded during the application of VoiceMyChoice™ to ensure that all of the items of were provided appropriately during treatment sessions. Similarly, conversations were recorded between dyads during the control condition to ensure that the NAs were all following the protocol procedures. Both the nurse assistants and residents were given the opportunity to opt out and not have the session recorded. While only three residents declined to have their sessions recorded, 16 of the 33 nurse assistants would not allow us to record their sessions. To further insure that the protocol was being followed research staff observed from a distance all sessions and made corrective recommendations if the NA was not following the protocol.
3.1. Convergence between NA-Resident Dyads on the PAQ
Means and standard deviations of total convergence scores, as described earlier, are shown in Table 1. Results of the 2 (Treatment: VMC vs Control) × 2 (NA Native Language: English vs non-English) × 2 (Time: Pretest vs Posttest) × 2 (Week: Week 1 vs Week 2) Analysis of Variance (ANOVA) revealed significant effects for Treatment, [F(1,33) = 6.34, p < 0.02], as well as Treatment × Time [F(1,33) = 6.53, p < 0.02], and Treatment × Week [F(1,33) = 12.54, p < 0.001]. No significant effects were detected for NA Native Language. Significant within-subjects effects included those for Week [F(1,33) = 25.81, p < 0.001]; and Week × Time [F(1,33) = 4.86, p < 0.03].
These results, combined with an examination of the means associated with these outcomes, lead to the following explanations. Overall, NAs using VMC generated scores that were more convergent with those of their resident dyad members than NAs in the control condition. Convergence levels in the control condition remained stable over time. Convergence levels in the VMC condition improved from Pretest to Posttest in both Week 1 and in Week 2, though the largest improvement occurred after the first use of VMC. Effects of VMC were similar for both native English speaking NAs and non-native English speaking NAs.
3.2. Consistency in Residents’ PAQ Scores
For conclusions regarding convergence between resident-NA dyads to be valid regarding PAQ scores, it was necessary to demonstrate that residents’ PAQ scores were consistent and reliable. We approached this issue in a number of ways. First, we examined residents’ overall PAQ scores for all four occasions of testing, broken down by NAs’ native language and experimental condition. This resulted in an analysis using the 2-level between-groups factors of NAs’ Native Language (English vs non-English) and Experimental Condition (VMC vs Control), along with a 4-level within-subjects factor of Time of Testing (Pretest 1, Posttest 1, Prerest 2, and Posttest 2). A Multivariate Analysis of Variance approach to repeated measures was used to compare these four times of measurement based on linearly independent pairwise comparisons (using Greenhouse-Geisser adjustment to degrees of freedom) among the estimated marginal means along with a test of the main effects of Experimental Condition, of NA’s Native Language, and the interactions among these factors. None of these factors were significant. We then combined all residents into a single group, and the resulting analysis also resulted in a nonsignificant Time of Testing effect, F(3,34) = 1.79, p > 0.15, and no Bonferroni pairwise comparisons between means were significant. Thus, overall PAQ scores of residents remained stable over time, and were not significantly different in either NA language group nor in the experimental conditions. These means are shown in Table 2 (N = 37). In addition, correlations at Week 1 and Week 2 between pretest and posttest scores across all 37 residents were strong and significant, r = 0.8, p < 0.001, and r = 0.8, p < 0.001, respectively. Thus, residents’ overall PAQ scores were consistent and reliable.
Table 1. Mean (and S.D.) convergence scores by treatment, NA native language, and time of test on the preference assessment questionnaire*.
*Lower scores = better convergence between dyad members.
Table 2. Mean (and S.D.) Preference assessment questionnaire scores for residents.
We also examined whether, for residents, consistency in PAQ responses was stable across individual PAQ items and across time, and whether consistency in PAQ scores was related to mental status. To assess this, for each item of the PAQ, we took the difference between the response to the item at pretest and at posttest, then took the absolute value of that difference, and summed all of these absolute values across the 25 items of the PAQ. We did this separately for Week 1 and for Week 2 scores. These summed scores represented total overall consistency in responding by each resident for the PAQ. The means (and standard deviations) for these residents’ consistency scores were M = 8.5 (3.6) for Week 1 and M = 8.1 (3.1) for Week 2. A paired t-test comparison of these scores was not significant. These consistency scores then were correlated with MMSE scores for Week 1 and again for Week 2. No significant correlations were obtained, signifying that residents’ responses were consistent across items and across time, and that mental status was not related to residents’ consistency in responding to PAQ items.
These results are consistent with Bourgeois et al.  , who showed improved congruence between NAs and residents on the DQoL after an enhanced verbal condition, and with Bourgeois, Camp, and Zeisel  , who demonstrated improved conversational interactions when using visual stimuli, but not with enhanced verbal stimuli, for both native and nonnative English speaking NAs. However, results of the current study are different than those of Bourgeois, Camp, & Zeisel  , who did not show greater congruence on the DQoL after either verbal or visual treatment conditions; in this study the VMC treatment condition did lead to greater convergence on the PAQ than the control condition. This suggests that the choice of a QoL measure that is better aligned with the treatment condition has greater potential to measure the effects of treatment. This study also showed that improvements in convergence are possible with additional exposure to the VMC treatment condition, suggesting that facilities that want to address the PCC needs of their residents with dementia should plan time for NAs and residents to converse about QoL topics using visual supports.
Thanks to Veronica Rose, Emily Reifsteck, Caitlin Stewart, Diana Chang, Jeanette Benigas, and Evan Shelton for data collection and data management.
Conflict of Interest
VoiceMyChoice™ was developed through an SBIR grant to the Center for Applied Research in Dementia. The purpose of an SBIR grant is to create evidence-based interventions and products that will be taken to market.
This work was supported by the National Institute of Aging [R43AG043188-01 to C.C (PI).].
VoiceMyChoice™ Preference Assessment Form
Name: ________________ Date: ________________ PRE POST
Please Circle One
a. Roasted chicken ______________ always sometimes never
b. Pancakes ____________________ always sometimes never
c. Apple pie ____________________ always sometimes never
d. Soup and crackers_____________ always sometimes never
e. Bananas ____________________ always sometimes never
a. Bingo ______________________ always sometimes never
b. Reading ____________________ always sometimes never
c. Gardening __________________ always sometimes never
d. Keeping pets ________________ always sometimes never
e. Word games ________________ always sometimes never
3. Daily Living:
a. Taking a shower______________ always sometimes never
b. Exercising ___________________ always sometimes never
c. Napping _____________________ always sometimes never
d. Eating meals __________________ always sometimes never
e. Going for walks _______________ always sometimes never
a. Talking on the phone ___________ always sometimes never
b. Talking with residents __________ always sometimes never
c. Family visits __________________ always sometimes never
d. Holding hands ________________ always sometimes never
e. Going to church/religious _______ always sometimes never
a. Headache ____________________ always sometimes never
b. Arthritis/ joint pain ____________ always sometimes never
c. Stomach ache _________________ always sometimes never
d. Toothache ____________________ always sometimes never
e. Chest pain____________________ always sometimes never