Original Article
Using the age at onset may increase the reliability of longitudinal asthma assessment

https://doi.org/10.1016/j.jclinepi.2006.10.010Get rights and content

Abstract

Objective

Recently, self-reported asthma was combined with reported age of onset to investigate the disease's natural history. To assess the validity of reported results, we investigated the reliability of the method.

Study Design and Settings

The European Community Respiratory Health Survey was a longitudinal study with interviews in 1991/93 and in 2000/02. Lifelong asthma and age of asthma onset were assessed through self-administered questionnaires. Responses of 10,933 participants in the follow-up were combined to separate true from false incident cases. The repeatability of questions was assessed and the bias in cumulative incidence (CI) estimation was quantified.

Results

Age at onset had excellent reliability (mean difference between the two interviews = −0.20, weighted κ = 0.88) allowing the differentiation of false and true incident cases. Given this information, lifelong asthma question's reliability was very high (agreement = 0.96, κ = 0.83). Misclassified subjects had respiratory conditions similar to the asthmatics. Baseline asthma was underreported and, if ignoring the onset age, the CI was severely overestimated (observed 5.82%, actual 3.02%).

Conclusion

Questionnaire-based longitudinal studies make more reliable estimates possible when all the retrospective information is used: the reported age of onset plays a key role and should be accounted for when investigating the natural history of diseases.

Introduction

In the absence of an agreed definition of asthma, epidemiological research mainly relies on questionnaires investigating the presence of asthma-like symptoms, asthma attacks, and/or a diagnosis of asthma. Accordingly, the natural history of asthma has recently been investigated by means of a filtering question on the presence/absence of lifelong asthma, followed by a nested question on the age of the first attack [1]. In cross-sectional studies, this second question was used both as a “control question,” to reinforce the first one, and to retrospectively estimate the incidence of asthma by considering the age of the first attack as a proxy of the age of asthma onset [2], [3]. In longitudinal studies, the question on the age at onset has been used to differentiate “true” incident cases that occurred during the follow-up, from “false incident cases” (i.e., subjects who only reported asthma during the follow-up, but who had had their first attack long before the start of the follow-up) [4]. Recently, the reliability of self-reported asthma and age of asthma onset were investigated: it was reported that the recall of asthma onset was accurate, whereas the reliability of the asthma question was dependent on the severity of the disease [5]. The consequence of an imperfect assessment of asthma on the incidence rate was also partially explored [6]. However, the joint reliability of the two questions has not been investigated so far, and the potential bias that could affect incidence estimates is still little known.

A low reliability of the ever asthma question would mean that some asthmatics do not report asthma and some healthy subjects report to have had asthma. However, the question on asthma was proved to be highly specific [7], [8], thus it is very unlikely that healthy subjects reply affirmatively to a question on the presence of asthma. Accordingly, misclassification should mainly concern asthmatics (who do not report their disease): the expected consequence should be an underestimation of prevalence in cross-sectional studies and an overestimation of incidence in longitudinal studies because asthmatic subjects who did not report asthma at the baseline are considered at risk during the follow-up.

The aim of our study was to estimate the long-term reliability of the questions on the age of asthma onset and on the presence of lifelong asthma and to evaluate the consequence of their joint use on the estimation of incidence. To address this goal, clinical and physiological characteristics of subjects reporting discordant answers to the same question on lifelong asthma at the two interviews were considered. For this purpose, we used the data from the European Community Respiratory Health Survey (ECRHS)-II, which was the 10-year follow-up of about 11,000 European subjects who participated in the clinical stage of the ECRHS in 1991/93.

Section snippets

Study design

The ECRHS was an international, multicenter survey on asthma and respiratory diseases, which was fully described elsewhere [9], [10]. In 1991/93 (ECRHS-I) a random sample of about 120,000 subjects from 29 centers, within 16 countries, aged 20–44 years (men/women = 1:1), replied to a mailed questionnaire on respiratory symptoms. Of the total sample, 18,356 subjects underwent a clinical investigation consisting in a standardized, structured interview and clinical tests (respiratory test,

The sample

The average time between the two interviews was 8.6 years (standard deviation, SD = 1.26). At the first interview, the mean age of the 9,459 eligible subjects was 34.3 years (SD = 7.13), 52.44% were women, 33.74% were smokers, 21.48% were ex-smokers, and 87.62% had a high level of education.

Reliability of the question on the age of asthma onset

Lifelong asthma was reported both in the 1991/93 and in the 2000/02 surveys by 1,174 subjects; of them, 1,154 provided the age at onset on both occasions. The reported mean age at the first attack was 15.9 (SD = 

Discussion

The ECRHS was a longitudinal study on asthma and respiratory diseases in adults; the information collected in its frame was, to a long extent, based on questionnaire. Like many other studies, ECRHS is open to the risk of bias due to the poor repeatability of questions over a period of time, which may have an impact on the measures of disease frequency.

The main findings of our analysis, aimed to investigate the reliability of the questions used to identify asthma cases and to estimate their

Acknowledgments

We owe a debt of gratitude to J. M. Antó (IMIM Universitat Pompeu Fabra—Barcelona, Spain) and S. Chinn (Department of Public Health Sciences—King's College, London, UK) for reading this manuscript and for their valuable comments.

References (25)

  • X. Basagana et al.

    Socioeconomic status and asthma prevalence in young adults: the European Community Respiratory Health Survey

    Am J Epidemiol

    (2004)
  • B. Järvholm et al.

    The association between epidemiological measures of the occurrence of asthma

    Int J Tuberc Lung Dis

    (1998)
  • Cited by (0)

    View full text