These analyses performed on a representative random sample of the population of the Paris metropolitan area confirm the main factors (age, level of income and education) discriminating Internet use in the general population on the one hand [2, 3, 30, 31] and Internet use for seeking answers to health questions in the Internet user population on the other [7, 32–34]. Our study shows simultaneously – in a single representative sample of the general population – an association with the three variables mentioned above for both outcomes. Researchers sometimes present ethnicity as a factor associated with health information seeking . In France, it is illegal to ask people for their ethnicity, but it is allowed to ask for their nationality. Even if both do not explore the same dimensions, obviously, our data show that nationality is associated with Internet access but no longer with health information seeking among Internet users. In addition, new discriminating factors were observed. Specifically, our analyses show that social isolation is associated with a lower probability of Internet use. In the subsample of Internet users, women, people concerned about a health problem (their own or that of a close relative), those worried about their health or those who had difficulty understanding advice from physicians were more likely to have previously sought health information online than the others. The same is true for the individuals who were well integrated socially and who had Internet experience.
Even if the random sample from the Paris metropolitan area exhibits good representativeness, these results cannot be extended to the entire population of France because the Paris metropolitan area has specific characteristics not shared by the rest of the country. On average, its inhabitants are younger and have a higher level of education and a higher socio-occupational status . Certain contextual factors themselves are different. Urban density, physician density  and the high-speed Internet connection rate  are notably higher there than in the other parts of France.
In addition, certain analyses performed on this sample of 3023 people may have suffered from a lack of statistical power because of the small sample size in some of the subgroups. However, the modelling performed on the 200 bootstrap samples generated from the initial sample shows that the results presented here are robust and satisfactorily adjusted to the data.
Our study confirms that the Internet penetration rate is higher in the Paris metropolitan area than it is nationally. Close to 60% of the respondents had a home connection, and 70% had previously personally gone online. In another national survey, the same rates were observed in Paris, while the estimates for France as a whole were respectively 43% and 54% during the same survey year . These estimates are of the same order of magnitude as in all the other countries with high Internet penetration rates (65 to 75% in the United States, Japan and Sweden, for example ).
Furthermore, this study shows the existence, in the Paris metropolitan area, of a digital divide previously identified in France [2, 30] and in other industrialised countries [4, 31]. The probability of having Internet access decreases with age but increases with the level of education and income. The gender effect sometimes reported in Europe or the United States [38, 39] was not observed in our study. These disparities are reinforced by the individuals' perception of their own socioeconomic situation. For a given level of education and income, the probability of having Internet access is lower in cases where there are perceived financial problems. Lastly, difficulty reading French also seems to be a barrier to Internet use. In general, people who do not have Internet access are those who have the least favourable social and economic characteristics: low income, no or few degrees, unemployed, foreign nationality and social isolation. Yet, these populations are also recognised as being more on the fringes of the health-care system and in less good health [40–42]. In our study, when adjusted for all of the socioeconomic characteristics, sick people also use the Internet less than others.
The analysis performed on the subsample of Internet users shows that nearly 49% of the individuals surveyed had previously searched online for information on a health-related topic during the previous three years. It is difficult to compare this estimate with those in other studies because the definition of the term "health" or the time period considered differs substantially from study to study. In late 2005, 28% of the Internet users in France had used the Internet to search "for information on health, a disease or diet and nutrition during the previous month" , while 58% of Norwegians  or 71% of European Internet users  had previously conducted online searches "for health purposes". In the United States, even though the total number of Internet users increased between 2004 and 2006, the proportion of "health seekers" remained stable (around 80% of American Internet users) .
With regard to socioeconomic status, certain determining factors of Internet access also discriminate health information seeking within the Internet user population itself. As has been shown in several international studies (often by descriptive or univariate analyses), the probability of having used the Internet to obtain health information decreases with age but seems to increase with the level of education and income . On the other hand, the effect of the level of income seemed to distinguish between households with the poorest quartile of income and the others. While both the reported level of income and the subjective perception of financial difficulties, as well as occupational status, nationality and origin were factors discriminating Internet access, the second analysis revealed a significant association only between health information seeking and the actual level of household income. The data did not show a significant association with the subjective perception of one's socioeconomic status, occupational status or nationality.
As in other studies [7, 30, 33, 34] (although some of them report discordant results ), our results show that women Internet users are the most active online health seekers. In general, women can be considered the ones who usually look after health matters in their families . Although our data do not show a significant association with the number of children, a positive relationship is observed with living in a couple relationship or there being a sick individual in one's close circle. The gender effect appears perhaps to manifest as a different level of interest in health.
Some authors suggest that, in fact, the health information available online particularly benefits the already privileged in terms of health and health-care utilisation and/or the well-educated [10–12]. Indeed, a European study found that the probability of health information seeking grows with the number of visits to a physician . In the French context of universal health insurance, our multivariate analysis did not show the four variables concerning health-care system utilisation or health insurance coverage to have a significant effect. On the other hand, our analyses did show strong associations between Internet use for seeking answers to health questions and health status, experiences and perceptions. As some studies tend to show with perceived health status [47, 48] or long-term illness [7, 8], the probability of previously having searched for health information online was greater in people who were sick and in those with a poor perception of their mental health status and/or who had a close relative with an illness. A positive association was found among individuals who feel that they worry more about their health than others, which confirms the notion in a previous American study which found that health seekers were "more likely to be health-oriented" . As for health status (reported), the opposite effect was observed for our first outcome. While the sick were less likely to have Internet access, it was they who, among the Internet users, used this tool more often to obtain health information. The Internet therefore seems to be an important information-seeking tool for people dealing with an illness, when they have an opportunity to go online.
To understand and interpret our results more globally, we can refer to the concept of eHealth literacy [18, 19], which combines the dimensions that underlie health literacy (functional, critical and interactive)  and online skills . Although having reading difficulties appears to be a barrier to Internet use, we did not observe any effect associated with health information seeking in the Internet user population. The analysis of each of the outcomes shows postsecondary education to have a positive effect. Several studies do, in fact, show that the contents of health-related websites are written in a language geared to a high level of education .
All factors otherwise being equal (and especially for a given level of education), the probability of having used the Internet to seek answers to health questions is greater in individuals who find information provided by physicians difficult to understand. On the other hand, having difficulty applying health advice in daily life (adjusted only for gender, age, level of education and level of income) no longer appears to be significantly associated with online health information seeking in the final model. It may thus be asked to what extent the Internet is useful to people with poor health literacy.
Lastly, a double socioeconomic divide was previously reported in another study concerning Internet access and general Internet use (but not specifically health information seeking) . Our data show that similar factors (age, level of education and level of income) are associated both with Internet access and the use of the Internet for health information searching. Furthermore, our study shows that Internet experience has a positive effect. Health information seeking appears to be more common when the frequency of use and the number of years of Internet use is higher. It seems that, with experience, the Internet assumes an increasingly important role in Internet users' lives in terms of how they obtain information and can even become an integral part of their daily lives. Searching on the Internet would thus become a habit or even an automatic reflex when searching for any information in general, just as when searching for health information in particular.