Study participants (Figure )
Intention to use. The phone interviewers screened 132 caregivers, and 63 of them (47.7%) qualified to participate. Out of the 63 qualified caregivers, 35 consented and 28 did not. When asked to explain why they were not interested, 19 reported the following reasons: they had no time, they were too stressed, they were too busy, or they knew enough about caregiving already. When the 28 uninterested caregivers were asked to complete the study questionnaire by phone, only 11 of them agreed to do so. The total sample in this stage was 46. Table presents the participant characteristics at baseline.
The characteristics of participants (N = 46) and uninterested caregivers who completed study measures (N = 21)
Initiation of use. Out of the 35 caregivers who gave consent, 31 initiated the services. The four caregivers did not activate the account and another four caregivers who took an unusually long time (i.e., more than 70 days) to activate the account were excluded from the sample. The sample size was 27 in this stage.
Service utilization (nonusage attrition and frequency of use). Among the 31 caregivers who initiated the service in Stage 2, 28 completed the postintervention questionnaire and 19 received the intervention. This represented a 3% dropout attrition rate and a 39% nonusage attrition rate.
Outcome of use. Twenty-eight participants completed the follow-up measures. Seven participants did not complete the follow-up measures (Figure ). One completer was excluded from the analysis because of an unusual increase in the BSFC score at the postintervention because of an extraordinary life stress. Therefore, at the postintervention, the sample size was 27.
Results from the Multivariate Analyses
Figure summarizes the factors tested and identified as predictors affecting usage behavior according to the four research objectives and stages.
Factors affecting usage behavior. 1 Significant in multivariate analysis 2 Significant in univariate analysis but excluded in multivariate analysis
Stage 1: Intention to use
In Stage 1, three factors affecting uptake were explored by comparing the consent and nonconsent groups (n = 46). The first factor was the perceived usefulness of the e-mail support service to the caregiver. The mean UTAUT performance expectancy scores of the consent and nonconsent groups were 19.54 and 19.63, respectively, and the difference was not statistically significant. The second factor was the caregivers' perception of the ease of use of the support service. The mean UTAUT effort expectancy score of the consent group was 19.03 and that of the nonconsent group was 21.27, and the difference was statistically significant (t-test, p = 0.04). The third factor was the caregivers' perceived burden of care. The mean BSFC score of the consent and nonconsent groups were 35.97 and 40.00, and the difference was not statistically significant. A multivariate logistic regression analysis (Table ) showed the UTAUT performance expectancy and BSFC were not significant (p = 0.72 and 0.35, respectively). The only significant factor was the UTAUT effort expectancy (p = 0.04), suggesting that caregivers who gave consent to participate gave higher ratings on ease of use.
A multivariate logistic analysis of the UTAUT subscales and the perceived burden on intention to use (n = 46)
Stage 2: Initiation of use
In this stage, another three factors were examined for their association with how quickly the caregivers began services. The outcome variable was the time taken to log onto the service the first time, and the average time was 12.26 days (SD = 8.38; range = 1 to 34; median = 11.00).
The factors were the caregivers' intention to use (intended duration), the caregivers' acceptance of technology, and the clinical needs of the caregivers. The multivariate linear regression analysis using a backward stepwise procedure showed that both intention to use and clinical needs were not significant (p = 0.89 and 0.85, respectively). The only significant variable was the UTAUT technology acceptance score. The results showed that caregivers who had a higher score on the measure started the service earlier than did those with lower scores (p = 0.04). Further analysis was conducted to assess whether the following variables might moderate the relationship between technology acceptance and initiation of use: gender, age group, employment, hours of care per week, years of care, age at immigration, language preference, living with care recipient, positive aspects of caregiving, social support, and technology aptitude. The results showed that none of these factors had an effect on the relationship.
Stage 3: Nonusage attrition and the frequency of use
Objective 3 explored factors associated with service utilization, which was divided into continuation and attrition. In this stage, the participants were divided into three groups based on the frequency of use: nonusers (no e-mail; n = 9), occasional users (1 to 2 e-mails; n = 8), and frequent users (3 or more e-mails; N = 10).
Univariate analysis showed that age group, CCM, and attitude toward technology (TPI) had a statistically significant group effect on usage. Older participants tended to be nonusers, whereas younger participants were more likely to be occasional or frequent users (p = 0.047). Nonusers mostly rated that they were competent to give good care, whereas occasional or frequent users felt they were not (p = 0.01). When usage frequency increased, the attitude toward technology (TPI) showed an increase in positive attitude (p = .012). The difference in TPI scores between nonusers and frequent users was 5.75 (t = 3.37, degree of freedom [df ] = 17, p = 0.01). However, no group effect was found for the following variables: gender, education, years of immigration, years of care, hours of care, relationship with care recipient, care-recipient functioning level (OARS), problem behavior frequency (RMBPC-Freq), caregiver reaction to problem behaviors (RMBPC-Reaction), caregiver perceived burden (BSFC), self-rated health (SRH), or UTAUT technology acceptance.
The correlations among age group, CCM, and TPI were explored. The results showed that age group and CCM were correlated (Spearman r = -4.03; p = 0.02), with an older age associated with greater perceived competence. Both older age and greater perceived competence were associated with less frequent use of e-mail. In addition, TPI and age group were found to be correlated (Spearman r = -0.40; p = 0.04), with an older age associated with a less positive attitude toward technology. Finally, TPI did not have a statistically significant correlation with CCM.
Logistic regression of usage frequency (frequent users vs. nonusers) showed that both CCM (p = 0.04) and TPI were associated with usage frequency (p = 0.04), but age group was not (p = 0.06) (Table ). Caregivers who were less competent were more likely to be frequent users. The TPI scores were higher among frequent users than among nonusers across both caregiver competence groups.
Logistic regression model for usage (frequent users vs. nonusers; n = 19)
Stage 4: Outcome of use
The final stage--the outcome stage--explored the factors associated with clinical outcomes. The mean pre- and postintervention BSFC change score was 1.07 (t = -0.63, df = 26, p = 0.54). Nonusers had an increase in perceived burden at postintervention, whereas occasional users had minimal changes in score and frequent users had a decrease of score. An analysis of variance showed that the differences among the three groups were not statistically significant (F = 1.78, p = 0.19). However, when the BSFC change scores were compared between nonusers and frequent users (Table ), the difference was 7.42 and was statistically significant (t = 2.50, df = 17, p = 0.02).
Besides frequency of use, the following variables were tested for their association with the BSFC score: age, gender, employment, education, year of immigration, English proficiency, years of care, hours of care, living together, OARS, SRH, TPI, UTAUT technology acceptance, RMBPC frequency and reaction subscales and CCM (binary). The results showed that none of these factors were statistically significantly associated with the BSFC score.
Intent-to-treat was analyzed by including caregivers who dropped out. Dropout caregivers were assumed to have a BSFC change score of 5.22 (the mean change score of nonusers). After including the seven dropouts, the mean BSFC change score was 2.76, which was not statistically significant (t = 1.75, df = 34, p = 0.09). However, a statistically significant difference between nonusers (N = 14) and frequent users (N = 10) was found. The difference in BSFC change scores was 7.42 (t = 3.15, df = 22, p = 0.005).