Jump to comment:

• Overinterpretation of the possible effects of health status at baseline on mortality rates
  Marja Y Veenstra
  Published on: 17 October 2005

• Published on: 17 October 2005

  Overinterpretation of the possible effects of health status at baseline on mortality rates

  • Marja Y Veenstra
  • Other Contributors:
    • Paul HHM Lemmens

  Dear Editor,

  In their recent article, Jousilahti et al.[1] compare mortality rates among participants and non-participants of population surveys on health in Finland, and report higher mortality rates among male and female non-participants. Since information regarding non-response is usually limited to a few background characteristics available from population registers,[2] the study by Jousilahti et al. is a valuab...

  Show More

  Dear Editor,

  In their recent article, Jousilahti et al.[1] compare mortality rates among participants and non-participants of population surveys on health in Finland, and report higher mortality rates among male and female non-participants. Since information regarding non-response is usually limited to a few background characteristics available from population registers,[2] the study by Jousilahti et al. is a valuable contribution to our knowledge on possible bias in population health studies due to non- participation. However relevant and interesting their effort, we object to some of the conclusions reached by the authors.

  In particular, we take issue with the authors’ fourth key point on page 313: “Differences in mortality show that health behaviour and health status substantially differ between people who participate and who do not participate in health surveys.” This point is drawn from page 314, second paragraph: “The observed difference in mortality … is caused by differences … in health status at baseline”, and a bit further “A difference in health status and existing disease at baseline also seems to be evident”. We think that this conclusion could not be substantiated by the results from their study simply because they do not have (collected) any information on health status at baseline of non-participants. ‘Evidence’ on health behaviour is also indirect, assessed by data on ‘mortality’ attributed by the authors to smoking and alcohol consumption. What is presented as evident facts and key outcomes of the study is a speculation on part of the authors, inferred as a possible explanation for the observed differences in mortality between participants and non- participants. That this inference may be flawed is illustrated in our Dutch prospective cohort study.[3] We found the paradoxical results that although mortality rates at follow up were lower among participants, as in the Finish study, coronary heart disease was found to be more prevalent at baseline for participants compared with non-participants.

  Another, yet minor point is their contradictory statements regarding the potential impact of participation rate, in the last two paragraphs of the article. They concluded that “… low participation rate may considerably bias the results of population based health surveys”. We agree that one way to reduce bias is to increase the response rate. But this is, obviously, not a general rule. Stang and Jockel[4] for example, showed that under certain conditions studies with low response rates may be less biased than studies with high response rates. Austin et al.[5] showed that when subjects’ participation was related only to either the exposure or to the outcome of the study the odds ratio was not biased. So, besides trying to increase the response rates it remains necessary to compare respondents with non-respondents on both exposure and outcome[6] to determine whether non-response is random and, hence, ignorable.

  MY Veenstra,
  Addiction Research Institute, Erasmus University Rotterdam, Rotterdam, the Netherlands

  PHHM Lemmens
  Department of Health Care Studies, Division Medical Sociology, Care and Public Health research Institute, University of Maastricht, Maastricht, the Netherlands

  Correspondence to:
  Marja Y. Veenstra
  Department of Health Care Studies
  Division Medical Sociology
  P.O. box 616
  6200 MD Maastricht
  Netherlands
  Telephone: +31-(0)43-388 22 23
  Fax: +31-(0)43-388 41 69
  email: m.veenstra@zw.unimaas.nl

  References

  [1] Jousilahti P, Salomaa V, Kuulasmaa K, Niemela M, Vartiainen E. Total and cause specific mortality among participants and non-participants of population based health surveys: a comprehensive follow up of 54 372 Finnish men and women. J Epidemiol Community Health 2005;59:310-5.

  [2] Stang A. Nonresponse research--an underdeveloped field in epidemiology. Eur J Epidemiol 2003;18:929-31.

  [3] Veenstra MY, Friesema IHM, Zwietering PJ, Garretsen HFL, Knottnerus JA, Lemmens PHHM. Lower registered prevalence of heart disease among survey nonrespondents, but higher mortality risk during follow-up for the middle aged general population. Accepted for publication in J Clin Epidemiol.

  [4] Stang A, Jockel KH. Studies with low response proportions may be less biased than studies with high response proportions. Am J Epidemiol 2004;159:204-10.

  [5] Austin MA, Criqui MH, Barrett Connor E, Holdbrook MJ. The effect of response bias on the odds ratio. Am J Epidemiol 1981;114:137-43.

  [6] Sackett DL. Bias in analytic research. J Chronic Dis 1979;32:51-63.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top