Points of interest


ITS’ comments to the report SOU 2015:6 on proposals for legislation to implement the EU Directive on Tobacco Products.


Mortality attributable to tobacco among men in Sweden and other European countries: an analysis of data in a WHO report.


Death by regulation: the EU ban on low-risk oral tobacco

Clive Bates makes strong arguments for lifting the EU-ban on smokeless tobacco.

Read more >


Swedish tobacco experts in double seats

Doctors help to lobby against snus – and for Pfizer’s drugs

Read more (eng) >

Read more (Swe) >


Cut out the smoke, reduce the risks


VAKT-YCYL (Tobacco data collected through ”FSI Inc” Your Country and Your Life)

Sample and response rate
The sample for each yearly study is retrieved from the Swedish population registry and is a national representative probability sample of people aged 18-79 years. Between 1971 and 1993 surveys were conducted once a year. Since the fall of 1993 a questionnaire is sent on a weekly basis to a sub-sample extracted from the main sample. The weekly sample size is 300-400 but can in certain periods be higher. Since the late 1990s, 7000-9000 responses are collected during each year. The “raw” response rate is estimated to approximately 60%. Non-response follow-up does, however, that the response rate is about 70%

Special non-response analysis
Each answered questionnaire has special code that tells the date the questionnaire were answered, the date when the returned questionnaire arrived in the mail and a code for how many and which measures was put in place (reminders, etc.) before the response is obtained. These codes provide a measure of the tendency to respond. The relationship between this measure and the answers to each question in the material are calculated. These calculations not only provide hints about possible non-response errors, but also on direction.

Quality Controls
It is well documented that the most serious sources of error in questionnaire surveys comes from properties in the questions wording, the questions focus and width, etc. These problems are handled by the presence of several indicators of the same or a similar focus and a large number of indicators and background variables that relate to the issue, offering a chance to read the entire response pattern rather than individual responses to individual questions. The comparative background that may be necessary to get an idea of what qualifies as ”high”, ”good” or ”low”, ”bad” values of the received responses is often available through comparisons several years back in time. The possibility to make comparisons with responses from the exact same query formulations are in many cases available for several decades back in time. Another source of error that has attracted attention is the so-called context-sensitivity. Query answers can be strongly influenced by the environment, which exists when the answers given. This controlled by measurements made of the response variability between days, weeks, months, etc. Measurements with this methodology show a long list of examples of query response, which would not have been representative of the year or even six months, if they have resulted from a single study limited in time. Another contextual effect is that the answers to a question may be affected (sometimes substantially) of the surrounding issues. This effect is minimized by having a very similar composition in the different all questionnaires.