National Institute of Public Health, University of Southern Denmark, Copenhagen, Denmark
Center for Clinical Research and Prevention, Bispebjerg and Frederiksberg Hospital, Frederiksberg, The Capital Region, Denmark.
J Epidemiol Community Health. 2023 Feb;77(2):116-122. doi: 10.1136/jech-2022-219944. Epub 2022 Nov 29.
Agreement may be low when comparing self-reported diseases in health surveys with registry data. The aim of the present study was to examine the agreement between seven self-reported diseases among a representative sample of Danish adults aged ≥16 years and data from medical records. Moreover, possible associations with sociodemographic variables were examined.
Nationally representative data on self-reported current or previous diabetes, asthma, rheumatoid arthritis, osteoporosis, myocardial infarction, stroke and cancer, respectively, were derived from the Danish National Health Survey in 2017 (N=183 372). Individual-level data were linked to data on the same diseases from medical records in registries. Logistic regression models were used to explore potential associations between sociodemographic variables and total agreement.
For all included diseases, specificity was >92% and sensitivity varied between 66% (cancer) and 95% (diabetes). Negative predictive value (NPV) was >96% for all diseases and positive predictive value (PPV) varied between 13% (rheumatoid arthritis) and 90% (cancer). Total agreement varied between 91% (asthma) and 99% (diabetes), whereas the kappa value was lowest for rheumatoid arthritis (0.21) and highest for diabetes (0.88). Sociodemographic variables were demonstrated to be significantly associated with total agreement for all diseases, with sex, age and educational level exhibiting the strongest associations. However, the directions of the associations were inconsistent across diseases.
Overall, self-reported data were accurate in identifying individuals without the specific disease (ie, specificity and NPV). However, sensitivity, PPV and kappa varied greatly between diseases. These findings should be considered when interpreting similar results from surveys.
在将健康调查中的自我报告疾病与登记数据进行比较时，可能会存在一致性低的情况。本研究旨在检验丹麦≥16 岁成年人代表性样本中七种自我报告疾病与病历数据之间的一致性，并探讨其与社会人口统计学变量之间的关联。
本研究使用丹麦国家健康调查 2017 年（N=183372）的数据，分别对当前或既往患有糖尿病、哮喘、类风湿关节炎、骨质疏松症、心肌梗死、卒中和癌症的人群进行代表性调查。个体层面的数据与登记处的病历中相同疾病的数据进行了关联。使用逻辑回归模型来探索社会人口统计学变量与总一致性之间的潜在关联。
对于所有纳入的疾病，特异性均>92%，而敏感性则在 66%（癌症）至 95%（糖尿病）之间波动。所有疾病的阴性预测值（NPV）均>96%，阳性预测值（PPV）则在 13%（类风湿关节炎）至 90%（癌症）之间波动。总一致性在 91%（哮喘）至 99%（糖尿病）之间波动，而类风湿关节炎的 Kappa 值最低（0.21），糖尿病的 Kappa 值最高（0.88）。社会人口统计学变量与所有疾病的总一致性均显著相关，性别、年龄和教育程度的相关性最强。然而，疾病之间的关联方向并不一致。
总体而言，自我报告的数据在识别未患特定疾病的个体时较为准确（即特异性和 NPV）。然而，敏感性、PPV 和 Kappa 在疾病之间存在较大差异。在解释类似调查结果时，应考虑这些发现。
