The Frederiksberg Heart Failure Study: Rationale, Design and Methodology, with Special Emphasis on the Sampling Procedure for the Study Population and Its Comparison to the Background Population

Background/Objective: The aim of the Frederiksberg Heart Failure Study was to describe the epidemiology of heart failure in the general population among persons older than 50 years. This paper describes some aspects of the methodology of this study, especially regarding population sampling. Methods: From 1997 to 2000, a random sample of 1,088 women and men aged 50–89 years from the municipality of Frederiksberg (city of Copenhagen, Denmark) were invited to participate in the study; 764 subjects (70.2%) were examined (57% female). To optimize scientific information across age groups, we used age-stratified sampling, attempting to have at least 150 subjects in each age decade, i.e. 50–59 years (n = 240), 60–69 years (n = 208), 70–79 years (n = 190) and 80–89 years (n = 126). Each participant filled in a questionnaire and was submitted to an echocardiographic examination, an ECG and collection of blood samples. The group of invited subjects (divided into participants and nonparticipants) was compared with the background population with regard to hospital admissions due to cardiovascular diseases and mortality 1 year before and 1 year after sampling. Results: As a result of the sampling strategy, the sampling proportion increased from 2.86% in the age class 50–59 years to 4.43% in the age class 80–89 years, with no evidence of a sex difference. The participation proportion systematically decreased with age, from 75.2% in the age class 50–59 years to 57.5% in the age class 80–89 years, with no sex difference. Both participants and nonparticipants had admission rates for cardiovascular diseases similar to that of the background population during the year before sampling. However, the rate of admission was significantly increased (p = 0.025) during the year after the time of sampling, and this applied equally to participants and nonparticipants. The 1-year mortality rate was increased in the sample after sampling (p = 0.014). However, participants had a 1-year mortality rate very close to that of the background population, whereas nonparticipants had a 1-year mortality rate more than three times higher than that of the participants (p < 0.001). In terms of methodology, the echocardiographic assessment was examined as follows: the interobserver variability between two independent observers in measuring the ejection fraction by means of echocardiography was calculated, and overall agreement was found to be 93.2% (kappa value = 0.786), corresponding to a very good level of agreement. The interobserver coefficient of variation was 4.9%. Conclusion: In this population-based study, the increased mortality and admission rates among nonparticipants suggest that subjects who agree to participate in a study on heart failure are a biased sample, in that they represent relatively healthy subjects. The fact that the participation rate declined with age indicates a systematic age effect in this bias. Standard methods to assess how representative of the corresponding background population a study population is should be agreed upon for future studies.