Using a state authorized web-based platform to distribute survey invitations provides an opportunity to rapidly get access to great numbers of potential respondents while reducing research costs (6). The approach, however, can be challenged by issues relating to achieving a representative sample. In the invitation letter for our survey using the so-called digital mailbox, men were encouraged to participate to acquire knowledge about health care users’ preferences regarding participation in decisions about the medical care they receive. Unfortunately, however, only a minority chose to utilize this opportunity to have their voice heard, raising the question whether those who decided to respond are representative of our target population. Uneven representation of groups with different opinions regarding patient involvement could potentially introduce a significant non-responder bias requiring statistical adjustment be considered. We therefore previously reported on the representativeness of our sample in terms of socio-demographic characteristics and overall found our sample to fairly well represent the population of Danish adult men (7). However, RRs were statistically significantly higher among older men and among men living outside the capital region but lower in high-income areas (7). We finally concluded that our socio-demographic comparisons needed to be supplemented with studies of the representativeness regarding personality characteristics and respondents’ general view regarding patient involvement. In this paper, we compared the personality and control preference characteristics of our sample to previously collected international datasets. Despite minor variations, we found our sample to chiefly resemble international data. Below, we discuss findings in the context of the research literature.
Sample personality characteristics
On three scales (agreeableness, extraversion, and neuroticism scales) out of five, rating estimates were within the range of our comparison groups (a-c). Regarding item ratings, our estimates fell within the range of comparison group averages in five items (reserved, trusting, relaxed, outgoing, fault finding) out of ten. Another five item profiles of our sample, especially relating to ‘Conscientiousness’ and ‘Openness’ scales, fell outside previously published figures. At least in part, our data may reflect the fact that population personality figures may not be constant over time and may have changed since ISSP data collections from 2005 (30). It has been noted that the stability implied by the notion of ‘personality’ pertains to an individual life span and therefore does not preclude generational changes in personality trait distributions (30). Correspondingly, previous cohort studies agreed that, e.g., ‘Conscientiousness’ ratings seem to increase over time (30, 31). In spite of everything, the rating estimates in our sample mostly resembled those of the English-speaking countries rather than e.g. the Scandinavian countries group as a whole. The reason for this apparently greater similarity with English speaking countries remains unclear. It might reflect a greater similarity with English speaking countries but could also result from variation in the design of surveys and samples included in the Gesis data repository. In this regard, our comparisons in table 1 suggest that sample compositions in terms of socio-demographic characteristics differ amongst groups and that sample differences may be smallest between our sample and the group of English speaking countries. In other words, regarding the personality and decision preference measures under study, our sample seems more similar to samples from English speaking countries than to, e.g., previously reported Scandinavian samples. Among others, similarities were clear regarding the trust item (and associated ‘agreeableness’ factor) that may be a particularly important aspect of personality, when considering healthcare communication and decision-making (32). Physicians’ communication has an important impact on patients’ trust and trust generally is a crucial element of the healthcare provider-patient relationship. Correspondingly, trust may be particularly important when patients are in an exposed situation and, e.g., confront a potentially life-threatening illness such as cancer and therefore need to rely strongly on their care providers (33-35).
Regarding control preference profiles, our sample not only displays a distribution that pretty nicely reflects the (‘bell’) distribution that has been repeatedly reported in the literature (36-38), but also specifically resemble CPS profiles reported in studies on men's’ preferences for involvement in cancer care decisions. In this regard, it is remarkable that, if comparing to Degner and Sloan’s 27 year old data from the ‘pre-patient-involvement-era’, it appears that control preference figures have generally changed in favor of a more collaborative or active role (38, 39). Still, it must be remembered that most other studies have been conducted in patient samples. As such, Degner and Sloan’s original study is among the few studies including also non-patients (20). On the other hand, it probably could be claimed that similarity of our sample CPS figures with the control preferences of real life patients would be of great importance. Hence, it seems as if the preferences for involvement in decision-making of our sample is rather similar to patient preferences found in real life settings (27-29).
Consideration regarding limitations and strengths
This leads to considering limitations and strengths of our study in more detail. It would be undesirable to conclude from a survey with a thirty percent response rate that the population generally wants to participate in health care decision making if those seventy percent of the population not wanting to participate in the survey are people who would generally abstain from any participation in health care decision making (7, 40). Similarly, ‘norm data’ may not necessarily always be representative of the relevant population. For example, to the authors’ knowledge, it remains unclear to what extent, e.g., ISSP 2005 datasets were representative themselves of the countries studied. Findings from previous research suggested that study participation may itself depend on personality factors (41, 42). For example, students in a survey were found to be more likely to be socially engaged (‘investigative’) personality types while they were less likely to be ‘enterprising’ or ‘artistic’ types (43). In this regard, Holland personality typology was used with ‘enterprising’ or ‘artistic’ types partially correlating with Big Five’s extraversion and openness factors (9, 43). In other words, individuals who score high on extroversion and openness seem less willing to participate in research studies (9, 41). On the other hand, for example openness may be related to a lower probability of quitting a survey following recruitment and the relationship between personality and survey participation thus may not be that clear (42). Correspondingly, little is known about the association between decision-making preferences and survey participation, and it is still possible that those not participating in our survey may have dissimilar personalities and control preferences. This hardly can be ruled out without just comparing to responders in another survey. Hence, strictly speaking, we have demonstrated that to a considerable extent, our sample seems comparable to international survey samples and that it is therefore likely that our forthcoming survey findings regarding preferences for participation in decision-making can be replicated abroad. Specifically regarding our comparison groups, as we found no commonly accepted categories in which culturally the BFI scores are distinctly different, we chose to group countries on our own although with reference to some relatively well known ‘categories’ (Western world English speaking countries, Scandinavia, and European West Germanic area). The latter categorization also explains the specific selection of countries.