Participants were eligible for this study if they (1) were 55 years of age or older; (2) had 1 or more drinks containing alcohol in the past 3 months; (3) had an email account and were willing to share their email address so that staff could provide links to the educational program and study assessments; (4) were comfortable using the Internet; (5) had access to high-speed Internet; and (6) were willing to spend about 10 to 30 minutes on 2 separate occasions 4 weeks apart to complete an online alcohol education program and answer online questions in English about alcohol and health. Participants were recruited between September and October 2014 at a non-profit community-based social services organization located in Santa Monica, California and serving Los Angeles County. A trained study site coordinator used a combination of methods to recruit participants. These included in-person contacts with potential participants, mailings, and phone messages to the organization’s database of volunteers and users. The organization’s Chief Executive Officer also emailed qualified staff and colleagues with information about the study and requested their participation. Interested persons responded directly to the study’s site coordinator and the CEO was not told who responded.

The study site coordinator screened interested participants for their eligibility using a standardized script. Using a random number generator [53], we produced 200 numbers in random order. These numbers were allocated to each eligible study participant in the sequential order that they were generated. Participants with odd numbers were assigned to the intervention group, and those with even numbers were allocated to the control group. For example, the first 4 randomly generated numbers were 136, 172, 187, and 61. This meant that the first 2 eligible participants were assigned to the control group, whereas the next 2 participants were assigned to receive “A Toast to Health in Later Life!”

Participants in the intervention group received the URL for “A Toast to Health in Later Life!”, a study user name, and a password. Once in the site, intervention participants were given the choice of watching a video or reading a transcript that explained the study’s purposes and methods. If still interested, they clicked “submit” and were automatically directed to the informed consent form. After consenting, participants were automatically directed to the baseline survey, which collected information on drinking, health, medication use, functional status, lifestyle, and demographics. Participants were sent back to “A Toast to Health in Later Life!” automatically upon survey completion.

We programmed the website so that participants were required to review the entire site. Once they completed the review, however, participants were able to go back to any section that interested them. Intervention participants were given a brief post-survey focusing on the site’s usability.

Control group participants completed the same online informed consent and baseline survey as the intervention group. They were offered access to the site after the completion of data collection.

The study’s site coordinator emailed a link to the follow-up survey to all participants 4 weeks after each completed a baseline survey. Intervention and control participants completed identical follow-up surveys. Participants were reimbursed $35 after they completed all required data collection activities.

The conceptual frameworks for this study and for use in guiding website content, questionnaire development, data collection and data analysis were The Health Belief Model, models of adult learning, and the US Department of Health and Human Services (DHSS) guidelines for designing easy-to-use websites [54].

We selected The Health Belief Model because it provides a framework for specifying personal and situational factors likely to influence health behavior [55], and it is among the most rigorously studied and commonly used models to study how education changes behavior [55-59]. According to the model, changing behavior is contingent upon several key factors including knowledge of the problem and its consequences and self-efficacy to do something about the problem. Throughout the website, we included facts (“Did You Know?”) and practice exercises to reinforce learning. To strengthen self-efficacy, the site provides separate pages on how to cut down on drinking, how to speak to a doctor or other health professional about alcohol, and where to go for help. We used selected behavior change techniques, such as instruction on how to perform a behavior and demonstration of the behavior [60,61] to reinforce knowledge and self-efficacy.

To encourage learning, we relied on theories of adult learning which are often derived from or based on Knowles’ theory, which he called “andragogy” [62,63]. Knowles’ theory distinguishes adult from other learning in its emphasis on active learning, self-pacing, and problem-centered concepts [64,65]. The Internet is well suited to active learning and self-pacing. To facilitate active learning, “A Toast to Health in Later Life!” includes practice exercises and feedback.

We used DHSS guidelines [54] for easy-to-use websites (eg, keeping navigation simple and consistent; minimizing scrolling) to guide the website’s development and separated the graphics arts component of Web development from the technical one so as to ensure that the site’s graphics were colorful and age appropriate regardless of any technical challenge. The 2 study team geriatricians (JCB and DO), who have expertise in educating older learners, independently reviewed the website twice to make certain that it conformed to DHSS’ usability criteria.

Over the study’s 3-month recruitment period, we screened 137 interested participants (Figure 3), 20 of whom were not eligible for participation. There were no differences between control and intervention groups in enrollment or attrition. Of the 112 participants who were randomized, 91.0% (51/56) of the control and 89.0% (50/56) of the intervention group completed the baseline assessment. No differences were found in sources of recruitment for participants who completed the baseline assessment, with 72.5% (37/51) of the control and 72.0% (36/50) of the intervention having been recruited from the study site’s volunteers. Four weeks after enrollment, 92.0% (47/51) of the control and 98.0% (49/50) of the intervention group completed all study measures.

The intervention and control groups did not differ in gender; age; education; ethnicity; or alcohol use at baseline, and they were mostly female; college educated; and in good health.

On average, participants in each group were 70 years of age and White. In the intervention group 42.9% (21/49) participants reported that their doctor told them they had hypertension and 40.8% (20/49) were taking antihypertensives. Similarly, 44.7% (21/47) of the controls reported having been told they had hypertension, and 36.2% (12/47) were using antihypertensives. Of the 13 intervention participants 26.5% (13/49), and 12 controls 25.5% (12/47) used NSAID’s daily. Weekly drinking varied widely, with members of the intervention group drinking from 0 to 28 drinks, and members of the control group drinking from 0 to 21 drinks. Of the intervention group, 16.3%) (8/49) of the intervention and 23.4% (11/47) of the control group participants drank 4 or more drinks at one sitting in the past 12 months.

Drinking quantity did not change over time for the intervention (P=.31) or control group (P=.56) (Table 2).

At baseline, 89.8% (44/49) of the intervention and 97.8% (46/47) of the control group drank 2 drinks or less when they drank, and at follow-up, 96.0% (47/49) of the intervention and 93.6% (44/47) of the control group drank 2 drinks or less. Similarly, the frequency with which study participants drank did not change significantly from baseline to follow up for the intervention (P=.99) or the control group (P=.67). Paired data analyses yielded similar results for quantity and frequency of drinking (all P-values >.05).

The proportion of participants who exceeded NIAAA’s recommended weekly limits did not significantly change over time in the intervention (P=.99) or control (P=.53) group (Table 3). At follow-up, 30.6% (15/49) of the intervention and 48.9% (23/47) of the control group’s drinking exceeded the NIAAA’s limits. No differences were found in the paired data analyses.

At baseline, 44.9% (22/50) of intervention and 66.7% (31/51) of control group were harmful or hazardous drinkers (Table 4). No changes occurred from baseline to follow-up for either group in any category. There were also no differences in the risk categories between intervention and control at either baseline or follow-up (P=0.12 and P=0.11, respectively). This remained true in the sensitivity analysis where we excluded the participants that were nonhazardous drinkers at baseline (P=0.94 at baseline and P=0.58 at follow-up) (data not shown).

Participants were asked if they changed their drinking in the past 4 weeks, and if so, whether they drank more, the same, or less. Table 5 shows that although intervention group did not differ from control group at baseline (P=.74), the intervention group differed significantly from the control group at follow-up (P=.02), with intervention participants reporting that they drank less.

On average, the 49 intervention participants spent between 10 and 20 minutes on the website. The majority (93.9%, 46/49) reported having little or no difficulty using the site (Table 6). Most also reported (96.0%, 47/49) reported that they learned from the website, and 75.5% (37/49) stated being very or somewhat confident that they can answer questions about alcohol and health. A large proportion (77.6%, 38/49) indicated that the website helped them to become more confident in their ability to answer questions, and 93.9% (46/49) definitely or probably would recommend the site to others. Finally, 67.3% (33/49) said the use of the site will change the way they think about drinking.

We developed and tested the feasibility of a “Toast to Health in Later Life!” with older adults who were comfortable using the Internet and lived independently within the community. Most intervention participants had little or no difficulty navigating the website and recommended it to others. The participants reported that they learned from the site, could now confidently answer questions about alcohol and aging, and agreed that the information they gathered will change the way they think about drinking. Intervention participants reported drinking significantly less than controls 4 weeks after study enrollment.

Despite the intervention group’s favorable review of the website, and their perception that they drank less because of it, we found no difference between the intervention and control participants in their quantity and frequency of drinking, adherence to NIAAA weekly drinking guidelines, or drinking risks over the study’s 4-week data collection period.

There are several explanations for the discrepancy between participants’ favorable perceptions and the study’s statistical findings. First, notwithstanding the site’s appeal to participants, the study may not have provided the supportive resources necessary for them to change their drinking behavior. Other successful efforts to reduce hazardous and harmful drinking in older adults have provided access to screening, personalized feedback, health care provider advice and other services in addition to education. In Project GOAL [84], one of the first studies to target older adults, physicians used a work-book to give advice to patients, included a second reinforcement visit with the physician, and provided 2 follow-up calls from a clinic nurse 2 weeks after each physician visit. These features probably contributed to its success in reducing both 7-day alcohol use and binge drinking.

Using the ARPS screening and education system, a study of 665 older adults and 23 physicians in the same community as the present study provided screening and personalized feedback to patients and their physicians in addition to patient education. Patients significantly reduced harmful drinking at follow-up from a statistically expected 21% in usual care to 16% and increased nonhazardous drinking from 52% expected in usual care to 58% over a 12-month period [67].

Another study of 1186 community-dwelling older drinkers and 31 primary care providers [77] succeeded in reducing at-risk drinking as defined by a measure based on the ARPS. In addition to providing participants with educational materials, this study also provided participants with personalized reports, drinking diaries, physician advice during office visits, and telephone counseling delivered by a health educator. Similarly, a study [78] which included 631 older at-risk drinkers, also using measures derived from the ARPS algorithms, found that its multi-faceted intervention among older at-risk drinkers in primary care did not reduce the proportions of at-risk or heavy drinkers but the intervention was found to reduce the amount of drinking at 12 months [77]. This study’s participants were given personalized reports, a booklet on alcohol and aging, a drinking diary, and received advice from a primary care provider and telephone counseling from a health educator.

These studies took place in medical rather than community settings. The presence and support of health personnel is likely to have motivated participants, encouraging them to change their drinking behavior. The effective use of health personnel to motivate patients has been shown in clinical settings for problems such as preventive health care use, health behaviors, and control of diabetes and hyperlipidemia [85,86].

Another possible explanation of the present study’s inability to detect differences in quantity and frequency and in risks is that its brief duration may have been insufficient for change to occur. Many participants may not have seen a health care provider over the study’s 4-week data collection period, eliminating the potential for alcohol-related discussions on how to make changes in lifestyle or medication-use. Further, participants may not have encountered situations, such as a party or other festive occasion, in which to change their drinking pattern. Additionally, intervention participants may have reported drinking less because they perceived it to be the socially desirable or expected response.

This study’s findings must be interpreted cautiously. The study was an initial test of the website, and there were no comparable studies on which to base sample size calculations. Thus, we cannot be sure if the number of participants was sufficient to detect a true difference between groups. Further, we did not compare “A Toast to Health in Later Life’s feasibility with alternative approaches or programs designed to educate older people about the benefits and risks of alcohol drinking because, to our knowledge, none was available. In fact, it was the lack of existing age-appropriate education that was the impetus for developing and testing “A Toast to Health in Later Life.” We might have developed our own alternative version of the content (eg, audio or print), but that would have necessitated a systematic test of the feasibility and comparability of this new program (eg, for literacy level and age relevance of content), and we did not have the resources to conduct such a study. Instead we took the approach advocated by some researchers who have stated that even if evaluations find that one intervention is superior to an alternative intervention, they cannot claim that the superior treatment should be the new standard because the intervention has not been shown to be superior to the care commonly given by practitioners [87].

This study’s sample included nonhazardous drinkers who may not have had an incentive to change their drinking patterns, raising questions about the appropriateness of including them in the study. We included these participants because previous research has shown that nonhazardous drinking older adults may become risky drinkers over time. In a study of community-dwelling older drinkers, for example, 20 of 112 (18%) of nonhazardous drinkers who did not receive an intervention became hazardous or harmful consumers over the study's 12-month period [67]. Also, because this study aimed to test the website’s feasibility, we considered all current drinkers to be eligible, and the sensitivity analysis did not change the study’s findings. Given that there is evidence that without intervention, some older drinkers may increase their risks, it seems prudent to educate all older drinkers regardless of their current risk. Future research should consider evaluating the effectiveness of patient education in preventing alcohol-related risks in older people as well as reducing those that already exist. Enthusiasm for the applicability of the study’s findings must also be tempered by the fact that the participants were selected because they were comfortable using the Internet. In the United States, nearly 60% of persons 65 and older use the Internet, with persons 55 years and older using it even more frequently [47,88]. These statistics may not apply to other US samples or nations, thereby limiting the feasibility of “A Toast to Health in Later Life” among many older adults.

This study was designed to test “A Toast to Health in Later Life’s!” feasibility and determine if education has the potential to change older adults’ drinking behavior. We found that the site is usable and acceptable in a White, healthy, and well-educated sample of older people. We do not know if older adults with other demographic characteristics will react to the website in the same positive way. But in terms of its drinking, the study’s sample was comparable to other older adults [2,67] in its frequency of drinking risks and failure to adhere to NIAAA drinking limits. This finding supports the suggestion that the website’s objectives and content are appropriate for a more general US population.

Despite the study’s limitations in setting, duration, and sample characteristics, it is the first that we know of to provide evidence that older adults are willing and able to use online education to learn about alcohol use and aging. Its findings also suggest that for alcohol education to be most effective, it should be included as a component of a larger effort consisting of screening, education and counseling preferably in a health care setting. In the United States, Medicare reimburses health providers who screen older patients and counsel those who are at-risk but do not meet the criteria for alcohol use disorders. Given how many preventive activities are expected of primary care providers [89], online education programs, such as “A Toast to Health in Later Life!” can be used to supplement counseling, possibly ameliorating the health care provider’s burdens especially with respect to sensitive topics such as alcohol use. Important next steps consist of testing the usefulness of “A Toast to Health in Later Life!” or other Web-based alcohol education programs in larger studies on relatively diverse older populations and also evaluating their effectiveness and cost-effectiveness.