Original articles
Nonresponse in a community cohort study: Predictors and consequences for exposure–disease associations

https://doi.org/10.1016/S0895-4356(02)00431-6Get rights and content

Abstract

We have assessed predictors for response in a Norwegian community cohort study, with an 11-year follow-up. We also examined to what extent the association of gender, age, and smoking to the incidence of respiratory symptoms and asthma differed if the analyses were based on the 65% (n = 2,079) initial responders, or were based on the 89% (n = 2,819) who responded after three reminders. The associations between the six symptoms/asthma and the gender, age, and smoking groups amounted to 42 odds ratios. The adjusted odds ratio for responding at follow-up was 1.39 (95% CI: 1.01, 1.90) for those being middle aged at baseline compared to younger subjects. The adjusted odds ratios for responding at follow-up for those being students, unemployed, or retired at baseline were 0.50 (95% CI: 0.35, 0.73), 0.29 (95% CI: 0.16, 0.55), 0.21 (95% CI: 0.13, 0.36), respectively, compared to being employed. Of the 42 odds ratios mentioned above, 25 differed less than 10% when comparing the initial and all respondents. Twelve differed 10–20% and five differed 20–45%. The study indicates that to ensure a high participation rate in a follow-up study one should pay special attention to those being late responders, unemployed, retired, or students at baseline. No overt differences were observed in the gender, age, and, smoking associations to the respiratory disorders when the analyses were based on the initial compared to all respondents.

Introduction

Several cross-sectional surveys have assessed the possible influence of nonresponse on the study results, including studies on patient samples 1, 2, working groups [3], and general population samples 4, 5, 6, 7, 8. Limited knowledge is available on nonresponse to follow-up in community cohort studies. A 6-year cohort study on respiratory health of a rural community sample in Lebanon, CT, obtained a response rate at follow-up of 55% among those eligible [9]. The response rates at follow-up were significantly lower in smokers compared to nonsmokers, and in the age group 15–24 years compared to older age groups. In the follow-up survey of the first National Health Nutrition Examination Survey (NHANES I), data on responders and nonresponders by gender, age, and race were given, but no statistical analyses were performed [10].

It is not known whether employment status and readiness to respond at baseline might effect the response at follow-up in cohort studies of general population samples. To our knowledge, no study has examined how nonresponse at follow-up may effect the relationship between the exposure of interest and the incidence of the outcome variables.

We have performed an 11-year cohort study on respiratory health in a Norwegian general population. The objective of this report was to examine how response at follow-up varied with baseline data on gender, age, respiratory symptoms, and asthma, smoking habits, employment, and response. Furthermore, we wanted to assess how using two remind letters and one remind telephone call to the initial responders at follow-up affected the estimated relationships of gender, age, and smoking habits to the incidence of respiratory symptoms and asthma.

Section snippets

Materials and methods

The baseline survey of this follow-up study was conducted in 1985. A detailed description of the sampling procedures and population characteristics are given elsewhere [7]. Briefly, the survey population consisted of 204,952 inhabitants aged 15–70 years, living in the city of Bergen and 11 surrounding municipalities. A postal questionnaire was sent to a random sample of 3,786 subjects. After two remind letters a total of 89% had returned the questionnaire. The response rate was significantly

Results

Comparing the responders and nonresponders at follow-up by characteristics in 1985, there were no differences in gender and smoking habits (Table 1). There was a higher percentage of middle-aged subjects among the responders, and a lower percentage of the younger and older subjects, compared to the nonresponders. Among nonresponders in 1996/1997, 47.5% had been late responders in 1985 (i.e., responders to the first or second reminder), compared to 27.4% among the responders in 1996/1997 (Table

Discussion

The present study showed that subjects being employed and ready to respond early at baseline were the best responders at follow-up. However, increasing the response rate from 65 to 89% did not overtly alter the incidence rates of the five respiratory symptoms and asthma or their associations to gender, age, and smoking habits.

The high response rates, both at baseline and at follow-up, are in agreement with previous studies in Nordic general populations 14, 15, 16. Factors that might have

References (30)

  • C.A Macera et al.

    Patterns of non-response to a mail survey

    J Clin Epidemiol

    (1990)
  • M.H Criqui et al.

    The effect of non-response on risk ratios in a cardiovascular disease study

    J Chronic Dis

    (1979)
  • K Sheikh et al.

    Investigating non-response bias in mail surveys

    J Epidemiol Commun Health

    (1981)
  • J Barton et al.

    Characteristics of respondents and non-respondents to a mailed questionnarie

    Am J Public Health

    (1980)
  • M.H Criqui et al.

    Differences between respondents and non-respondents in a population-based cardiovascular disease study

    Am J Epidemiol

    (1978)
  • R.N Forthofer

    Investigation of nonresponse bias in NHANES II

    Am J Epidemiol

    (1983)
  • B.K Jacobsen et al.

    The Tromso Heart Studyresponders and non-responders to a health questionnarie, do they differ?

    Scand J Soc Med

    (1988)
  • P Bakke et al.

    Postal survey on airborne occupational exposure and respiratory disorders in Norwaycauses and consequences of non-response

    J Epidemiol Commun Health

    (1990)
  • A Hill et al.

    Non-response bias in a lifestyle survey

    J Public Health Med

    (1997)
  • G.J Beck et al.

    A longitudinal study of respiratory health in a rural community

    Am Rev Respir Dis

    (1982)
  • B.B Cohen et al.

    Plan and operation of the NHANES I Epidemiologic Followup Study1982–84

    Vital Health Stat

    (1987)
  • A Gulsvik et al.

    Measurements of respiratory symptoms and sample size to detect a given difference between treatment groups in obstructive lung disease

    Eur Respir J

    (1991)
  • K Bjartveit et al.

    The cardiovascular disease study in Norwegian counties. Background and organization

    Acta Med Scand Suppl

    (1979)
  • O.P Foss et al.

    Serum thiocyanate and smokinginterpretation of serum thiocyanate levels observed in a large health study

    Scand J Clin Lab Invest

    (1986)
  • A Gulsvik

    Prevalence of respiratory symptoms in the city of Oslo

    Scand J Respir Dis

    (1979)
  • Cited by (0)

    View full text