The European MSM Internet Survey

The European MSM Internet Survey (EMIS) is a multi-country, multi-language, anonymous online survey for gay, bisexual, and other men who have sex with men (MSM).

EMIS is a joint project of academic, community, and governmental partners across Europe, to inform sexual health interventions for MSM. It occured in 2010 and 2017 with core-funding from the European Union Health Programme. We are planning to do it again in 2023, but no core-funding is yet available.

EMIS aims to describe the differences in sexual health needs across different groups as men, as well as to estimate the relative levels of different risk and precautionary behaviours.

EMIS has resulted in a wide variety of technical and community reports, very many scientific publications, and 26 indicators for 60 countries across 4 continents in the UNAIDS Key Population Atlas.

All this has changed the landscape for gay health in many European countries.


Planning EMIS-2023

At present, we are fundraising to repeat EMIS in autum 2023. The project is led by Axel J. Schmidt @ German AIDS Federation (Deutsche Aidshilfe), Kai J. Jonas @ Maastricht University, and Ulrich Marcus @ Robert Koch Institute.

EMIS-2017

EMIS-2017 included all of the above and Canada, Israel, Lebanon, and the Philippines (50 countries, or 43 countries with viable samples). EMIS-2017 initially also included 18 Latin American countries that were subsequently surveyed separatedly as the Latin American Internet Survey (LAMIS).

EMIS-2010

EMIS-2010 included all European Economic Area countries and Switzerland, Turkey, Western Balkans, Belarus, Moldova, Russia, and Ukraine (38 countries with viable samples).


FAQ

As with all surveys, there are limitations to Internet surveys. We minimized those limitations during questionnaire design (aiming not only at self-identified gay men, but at all men who have sex with men or who feel attracted to men), during recruitment (using multiple online social networks and websites), and during the analyses (e.g. by adjusting for age, place of residence, source of recruitment, etc.).

Internet sample are not representative, because the sampling frame is unclear.
No, they are not representative. It is very difficult to recruit a representative sample of MSM because there is no sampling frame for this population. The absolute size of the MSM population is unknown and its size depends on how the construct ‘MSM’ is defined. Only very small representative samples of MSM (within the bounds of disclosure) have been recruited during the course of general population surveys, and these are much too small to use for programme planning purposes.
This is why in EMIS we are more concerned with differences across the sample (e.g. by age, area of residence, etc.) than the absolute levels of any variables. We can treat the EMIS samples as the population of MSM that public health and health promotion are able to reach and interact with. In this sense the samples ARE the population of concern. The HIV prevention needs assessment carried out with the MSM in EMIS describes the needs of the service user group of the EMIS partnership.
If results from convenience samples are not representative, why do we do such surveys anyway?
Absolute proportions from the EMIS study cannot be seen as being representative for all MSM in that country. However, they can serve as best guesses for a minimum or maximum proportion. E.g. if 30% in our samples have never been tested for HIV, the true value is likely to be higher, but not lower, because of oversampling men with an interest in HIV prevention, and men with higher education. Instead of basing conclusions on absolute proportions, convenience samples can best be used to explore associations between different factors. This can be done both on an individual and on a country level.
Why don’t you use random samples based on the general population to then filter out those who are not attracted to men or never had sex with men?
Using random samples drawn from the general population is not cost effective for sampling MSM, because extremely large samples would be needed. To recruit the number of European MSM who responded to EMIS-2010 using a general population approach, would require about 6 million men, or 12 million people to participate.
When recruiting through particular websites, a large bias is introduced.
Yes it is. However we used a lot of different websites to recruit respondents to our survey. Furthermore, we included all major European dating apps (websites) for gay men, using individual messages to ensure maximum participation. Bias is something you always consider when doing research, and what you take into consideration when performing the data analysis.
Not all gay men have access to the Internet. By using the Internet, you introduce selection bias towards the more privileged MSM.
Yes we do. In 2008, access to the Internet via privately owned computers still varied substantially across Europe. Nonetheless, the one thing all respondents had in common is that they had access to the Internet, whether it be at home or elsewhere. By 2017, the digital divide between eastern and western Europe which was still visible in 2008 has almost disappeared, and most respondents filled-in the survey on their smartphone.
Face‐to‐face interviews or print questionnaires are better instruments to recruit sexually active MSM.
By using the Internet, a broader range of the MSM population can be reached as compared to print questionnaires, venue based approaches, or even RDS. This is true for younger and older age, for less educated men, and for man living in small cities and in the countryside; or men less connected to established gay communities. The more important the Internet is for finding sexual partners, the more important it is to at least include Internet sampling.
How can you compare samples from different countries, if the relative response rates (i.e. the number of respondents per 1000 inhabitants) are so different?
If you look at the response rates of those who were personally invited to participate in the survey (e.g. by a message on their Grindr, GayRomeo, Hornet, Manhunt, or other gay social network Internet profile), the differences in response are much smaller.
If you don’t collect IP addresses, how can you be sure that people don’t fake questionnaires or enter their data multiple times in order to move the results into a particular direction?
Using such a large questionnaire makes it unlikely to fill it in repeatedly. Furthermore, we apply plausibility checks that exclude people who gave random answers to get through the questionnaire. Though we cannot exclude fake answers, particularly if they are authentically faked, we can minimize random answers. Given the large number of responses, a few double entries will have minimal effect on the main findings and associations.