Eighty percent of participants supported a tax increase, under at least one of four scenarios, to fund unrestricted access to a hypothetical, new AD medication. Support was highest for the most favourable scenario, in which the medication was described as modifying the course of disease without entailing adverse effects.
A majority of participants would be willing to pay an additional $150 per annum in personal income taxes to fund unrestricted access to the medication. Just under half of the participants would also be willing to pay more than $225 per annum. For an average, single Canadian without children who earns $40,000 per annum and has an annual tax burden of 25% of income (i.e., $10,000 paid in taxes) , an extra $150 would translate into a 1.5% increase in tax burden.
This is the first study in Canada or elsewhere to assess whether the general public would pay more taxes to liberalize public reimbursement regimes for AD medications. In a recent systematic review , two of the authors (MO, JET) could only find one other study of taxation and publicly-subsidized programs in AD. This study, by Nocera et al. , involved 1,240 German-speaking persons in Switzerland who were aged 18 years or more and selected randomly from a telephone book. Three AD programs were assessed: two-day training for unpaid caregivers and free access to a nurse professional for four weeks annually; early screening; and intensified research in Swiss universities. Participants' support was assessed for various specific levels of taxation, e.g., would a participant support a tax increase of 'X' Swiss francs? Support was highest for the 'training-and-nurse' program (specific level of support not reported), followed by research (80% support) and screening (20% support). Nocera et al. did not report the percentage of participants who supported at least one program. The Swiss study, like the current study, shows that members of the general public do support increased taxation for AD programs, with the level of support varying depending on the perceived value of the program.
Subsequent to the review, a study by Negrín et al.  was published involving 598 participants selected randomly from the adult population of the Canary Islands. The participants were presented with a mix of three distinct programs (i.e., home care, access to day care centres, access to medium- and long-stay centres), combined with a mix of 'monthly contributions' (tax increases) to fund each program (i.e., €12, €24, €48, €72). The mix of programs and contributions was varied for each participant, who was asked to select the most preferred option from the mix. Home care was preferred by the majority of participants. Only 77 participants (13%) said they would not pay anything for these programs.
From a policy making perspective, the results suggest that Canadians would be likely to support (and pay taxes for) at least some relaxation of reimbursement restrictions on AD medications. The most support would flow when medications modify disease, especially in the absence of adverse effects. This finding is particularly relevant because the first disease modifying medications are due on the market within the next few years. Of course, if a disease modifying medication without adverse effects were ever to come on the market, then policy makers would likely approve it and may even offer unrestricted reimbursement, regardless of public opinion.
This study of the Canadian general public has numerous strengths that lend credence to its findings. First, the participants were recruited from a pan-Canadian sampling frame using a random sampling methodology, thus eliminating selection biases associated with region of residence or location of recruitment (e.g., recruiting visitors to doctors' offices instead of recruiting a broader, national sample). Second, data were collected with a standardized interview that was conducted by trained interviewers using CATI software to lessen potential information bias. Third, the order of scenarios, starting bids, and bid ranges were randomized to prevent ordering or starting point bias. Ordering bias can occur when participants' answers to a scenario are influenced by their answers to preceding scenarios, while starting point bias can occur when respondents express WTP values that are close to the initial bids. In an effort to reduce yeah saying and protest answers, respondents were first asked if they would support a tax increase. Fourth, the results for supporting a tax increase for specific scenarios made intuitive sense. Support was highest for the optimal drug scenario (disease modification, no adverse effects) and lowest for the least favourable scenario (symptomatic treatment, 30% chance of adverse effects). Support was also highest for the smallest tax increase ($75) and lower for larger tax increases (> $75) (Figure 2; p < 0.05 for all dollar-value comparisons versus $75).
In this study, the variable that was most consistently associated with support for a tax increase to fund unrestricted access to AD medications was participants' perceptions about whether family members or friends would somewhat or strongly approve of their decision to support a tax increase. Perhaps the favourable views of persons who are at the core of one's social network could exert a positive influence on support for a tax increase. Werner et al.  refer to this as 'subjective norm' and found similar results in a study of 220 caregivers of family members with AD. The caregivers in the Werner et al. study were found to be willing to pay more for an AD medication when they believed their close associates would approve of their paying for the medication.
Income was not associated with support for a tax increase. Three reasons could explain this finding. First, effects were undetectable because the sample was stratified into five equal-sized income categories to control for possible selection bias on income. Second, participants' said 'yes' to the tax support questions to 'do the right thing' or please the interviewer by giving a socially desirable response. Third, participants found the program affordable regardless of income.
Interestingly, 42% of the participants in the Canadian general public survey knew someone with AD. This may not be representative of the general Canadian population. Additionally, participants' median age was 51 years and 23% were 65 years or older, which indicates that the participants were an older subgroup of the Canadian population (2006 Census: median age = 39.5 years; percentage aged 65 years or more = 13.7% ). Participant characteristics on both variables could suggest that the study's findings overestimated the percentage of Canadians who would support a tax increase. For example, participants who know someone with AD will be aware of the devastation caused by the disease, thus enhancing their predisposition to support a tax-funded program of unrestricted access to AD medications. An older sample has an elevated risk for AD. These persons may recognize the challenges of being on a fixed income when their risk is highest, so they may be more likely to support a program of free access to medications.
The potential for overestimation of support for a tax increase is not actually born out by the data. In multivariable regression analyses, the crude association between knowing someone with AD and support for a tax increase was eliminated after adjustment for other possible explanatory variables. Age is inconsistently associated with support and it is significant at the 5% level in only three models that were built for the exploratory analyses. In these models, age was inversely associated with support. Perhaps people in upper middle age (who were saving for retirement) and seniors on fixed incomes were reticent to support a program that would reduce their personal financial resources. These persons might have made implicit risk-benefit calculations and concluded that the risk of actually getting AD was outweighed by the need to save money for present or future personal expenditures.
The data in this study are cross-sectional. Therefore, the direction of effect between some of the variables may be the reverse of what is hypothesized in the study (reverse causality). For example, it is possible that a participant's close family or friends might know that she or he supports a tax increase. Such knowledge could prompt the acquaintances to approve the participant's choice after it has been made. This is the opposite of the hypothesized direction of effect, which suggests the participant's perception of family or friends' approval influences the decision to support a tax increase.
The issue of reverse causality does not apply to all of the variables, simply due to their nature. For example, a participant's age and sex will not be affected by whether she or he supports a tax increase. Likewise, the number of relatives or friends with AD will probably be unaffected by one's support, or lack of support, for a tax increase. The same would be the case for health status on the EQ-5D, employment status, education, knowledge of AD, region of residence, or income.
The 4% overall interview completion rate (more than 13,000 telephone numbers dialled to generate 500 interviews) must be viewed in the context of the random digit dialling method. The ASDE Survey Sampler dialled random telephone numbers without guarantee that someone would answer the phone. Out of 13,195 dialled numbers, 66% were not answered, were busy, or were not in service and did not result in a contact. Since public opinion on tax support for AD medications is unlikely to be associated with answering a telephone, the potential for bias in this regard is likely to be minimal. For similar reasons, bias is also unlikely in the case of persons who were rejected because their income fell within a stratum for which the 100-participant quota was reached or because they were unneeded after the overall sample size requirement was met. Greater potential for bias might result from persons who refused participation or terminated an in-progress interview (n = 2,460), or who refused to provide their income (n = 92). Inclusion of these two groups in the denominator along with persons who agreed to participate (n = 540) creates a more realistic response proportion of 17%. Reasons for refusing to participate or for hanging-up are many and may well be unrelated to support for a tax increase (e.g., poor timing of the telephone call, person who answers the phone never as a rule participates in surveys, person perceives that the interview is taking too long and terminates the call). Although the participants were more likely to know someone with AD or be older than the average Canadian, these traits did not appear to bias the study results. Similarly, the decision to refuse to provide an income may not be associated with one's level of support for a tax increase. Indeed, the potential confounding effects of income were addressed by enrolling an equal number of participants in five income strata.
Certain segments of the population will be excluded from the sampling frame of a random digit dialling survey because they do not have a land line-based telephone. These persons include the homeless, people living in institutions, and people who use only cellular telephones. The possibility of bias in excluding these persons is likely to be minimal. The homeless or persons who are institutionalized due to mental or cognitive impairments would not be in a position to make decisions about supporting a tax increase, nor might they even pay taxes as a result of their circumstances. The exclusion of persons who use only cellular telephones would bias the results if they are more or less likely as a group, compared to people with land lines, to support a tax increase. Currently, there is no evidence to suggest whether the 'cellular only' group would be different from the 'land-line' group in this respect.
A potential issue in many surveys is social desirability bias, where participants provide what they believe are socially acceptable answers to questions or provide responses that they believe the interviewer wants to hear. While it is possible that this bias led to an overestimate of support for a tax increase, it should be noted that participants were identified through a telephone number, not by name, and they remained anonymous throughout the survey. Also, participants had no in-person contact with interviewers. The anonymous, impersonal nature of the interviews may have lessened participants' impetus to provide socially desirable responses because they would not have had to share unpopular views with peers. Additionally, non-verbal cues from the interviewer (e.g., nods of approval, frowns) would not be evident to participants over the telephone.
The models shown in Table 3 are dissimilar from one another. Some of the dissimilarity may be the result of the different sample sizes used to estimate the scenario-specific models. Sample sizes were based on the number of participants who said they would support a tax increase under each scenario. It is also possible that the determinants of support for a tax increase differ across scenarios. Further research is required to elucidate the nature of the differences across scenarios.
From a policy making perspective, the data in Figure 2 communicate two important points. First, support for a tax-funded program of unrestricted access to AD medications is likely to increase as the efficacy of AD medications improves. Second, supporters of the program would generally prefer to pay lower, rather than higher, taxes for the program, regardless of drug efficacy.
Interestingly, at higher taxation amounts (i.e., $225 or greater [Figure 2]), the data suggest a 'floor effect' in each scenario. A subgroup of the population is insensitive to changes in the range of taxation measured in the study. Also, another subgroup exists that would balk at paying ever higher amounts of tax for a program that they support in principle.
Future research would be advisable to set the study findings into a comparative policy context. For example, a group of people who are similar to the participants in this study could be asked whether they would support tax increases for unrestricted access to medications for AD, as well as to medications for other diseases, e.g., cardiovascular disease, diabetes. Policy makers require such comparative data to guide decision making. Another area of future research would be to investigate the determinants of support for specific dollar values of tax increases.