Self-Report as an Indicator of Incident Disease

https://doi.org/10.1016/j.annepidem.2010.03.017Get rights and content

Purpose

Epidemiological studies use self-reports from repeated surveys to ascertain incident disease. However, the accuracy of such measurements remains unknown, as validity studies have typically relied on data from prevalent, rather than incident, disease. This study examined the validity of self-reports in the detection of new-onset disease with measurements at baseline and follow-up conditions.

Methods

We conducted a prospective cohort study of 34,616 Finnish public-sector employees. Data from self-reported, physician-diagnosed diseases from two surveys approximately 4 years apart were compared with corresponding records in comprehensive national health registers used as the validity criterion.

Results

There was a considerable degree of misclassification for self-reports as a measure of incident disease. The specificity of self-reports was equally high for the prevalent and incident diseases (range, 93%–99%), but the sensitivity of self-reports was considerably lower for incident than for prevalent diseases: hypertension (55% vs. 86%), diabetes (62% vs. 96%), asthma (63% vs. 91%), coronary heart disease (62% vs. 78%), and rheumatoid arthritis (63% vs. 83%).

Conclusions

This study suggests that the sensitivity of self-reports is substantially worse for incident than for prevalent diseases. Results from studies on self-reported incident chronic conditions should be interpreted with caution.

Introduction

In epidemiological studies, the prevalence of a disease (i.e., the proportion of people with a disease in a population) and its incidence (i.e., the proportion of new disease cases during a follow-up in an initially disease-free population) can be established in many ways, including questionnaires, interviews, clinical screening, and medical records. While prevalent diseases are often determined through a clinical examination in large cohort studies, this is not necessarily the case for incident events, which may be, in part, determined by self-reports. Furthermore, because of feasibility, economy, and convenience, many large-scale studies and health interview surveys have entirely relied on self-administered questionnaires 1, 2, 3, 4, 5.

Validation studies have supported the accuracy of self-reports as a measure of prevalent chronic diseases 1, 3, 6, 7. However, less is known about the accuracy of self-reports in ascertaining incident disease despite their frequent use in epidemiological studies 8, 9, 10, 11, 12. The measurement of prevalent disease needs to correctly identify the disease at one point in time only. By contrast, the accuracy of self-reported information on incident diseases is actually affected by the accuracy of self-report at two stages: baseline and follow-up. Thus the assessment of incident disease with self-reports may be more open to measurement error than the self-report assessment of prevalent disease, primarily because the measurement requires both an accurate determination of the disease-free population at baseline and an accurate detection of new-onset disease at follow-up.

The aim of our study was to examine the accuracy of self-reports as the sole source of information in detecting new cases of common chronic diseases of public health importance: hypertension, diabetes, asthma, coronary heart disease, and rheumatoid arthritis. Survey methods are typically used to study these chronic diseases. We compared self-reports from two repeated surveys to records from national health registers (considered as independent gold standard) in a large occupational cohort of Finnish public sector workers.

Section snippets

The Finnish Public Sector Study

Data were derived from the Finnish Public Sector Study 13, 14. The baseline survey in 2000–2002 was agreed to answer by 48,598 employees (response rate 68%). The follow-up survey targeted 46,414 identifiable employees who were still in the service of the target organizations and alive in 2004 through 2005 and 35,914 (77%) of them responded. We excluded those with missing data on the selected self-reported physician-diagnosed diseases in either of the surveys (n = 1,298). Thus, the final cohort

Results

As shown in Table 1, our cohort included 28,545 women and 6,071 men, of whom 488 (1.4%), 3,833 (11.1%), and 1403 (4.1%) fulfilled the set register-criteria for diabetes, hypertension, and asthma, respectively, at baseline. Coronary heart disease and rheumatoid arthritis were less frequent, 273 (0.8%) and 224 (0.7%) cases, respectively. Table 1 also indicates that 1% to 3% of the diabetes, hypertension, and asthma cases were documented in the Hospital Discharge registers. As expected, for the

Discussion

The accuracy of self-reported hypertension, diabetes, asthma, coronary heart disease, and rheumatoid arthritis was examined in a large occupational cohort by using records in comprehensive national health registers as an external reference. Data from repeated surveys showed equally good specificity but lower sensitivity for self-reported disease incidence at follow-up compared to self-reported prevalence at baseline. The sensitivity rates for incident diseases ranged from 55% to 63%, and these

Conclusions

Given the widespread use of self-administered questionnaires in epidemiology, it is important to understand the extent to which self-reports validly determine the prevalence and incidence of a given disease. Our prospective study indicates that the sensitivity of self-reports is substantially worse for incident than prevalent diseases. The low sensitivity (55% to 63%) of self-reports in determining incident disease is an important source of bias in epidemiological studies leading potentially

References (36)

  • A. Paganini-Hill et al.

    Accuracy of recall of hip fracture, heart attack and cancer: a comparison of postal survey data and medical records

    Am J Epidemiol

    (1993)
  • N. Haapanen et al.

    Agreement between questionnaire data and medical records of chronic diseases in middle-aged and elderly men and women

    Am J Epidemiol

    (1997)
  • B. Jacquemin et al.

    Home outdoor NO2 and new onset of self-reported asthma in adults

    Epidemiology

    (2009)
  • W.W. Eaton et al.

    Depression and risk for onset of type II diabetes. A prospective population-based study

    Diabetes Care

    (1996)
  • C.G. Solomon et al.

    A prospective study of pregravid determinants of gestational diabetes mellitus

    JAMA

    (1997)
  • C. Iribarren et al.

    Are patients with asthma at increased risk of coronary heart disease?

    Int J Epidemiol

    (2004)
  • T. Kurth et al.

    Smoking and the risk of hemorrhagic stroke in men

    Stroke

    (2003)
  • M. Kivimäki et al.

    Socioeconomic position, co-occurrence of behavior-related risk factors, and coronary heart disease: the Finnish Public Sector Study

    Am J Public Health

    (2007)
  • Cited by (125)

    • Asthma, chronic obstructive pulmonary disease, tooth loss, and edentulism among adults in the United States: 2016 Behavioral Risk Factor Surveillance System survey

      2020, Journal of the American Dental Association
      Citation Excerpt :

      In another study, researchers analyzed National Health and Nutrition Examination Survey I data (N = 23,808) to determine if there was an association between confirmed chronic respiratory disease (n = 41) and periodontal index, and they determined that there was no statistically significant association40; however, other researchers using National Health and Nutrition Examination Survey III data determined that there was an association between periodontal disease and COPD.2 Although our data are based on self-reported information, published studies support the use of self-reported data for COPD,41 asthma,42 and tooth loss.43 Although we observed that patients with asthma had lower unadjusted odds of missing 1 through 6 permanent teeth or all teeth, this association reversed once age, sex, and race were entered into the model.

    View all citing articles on Scopus
    View full text