Abstract
Nordic countries have a long-standing tradition of using administrative healthcare data from specialist and hospital settings to study disease frequency and etiological associations. While similar data are increasingly available elsewhere, large-scale primary care data remain limited. Understanding how data source affects disease identification is therefore critical. We conducted a register-based study including all individuals aged ≥ 60 years residing in Region Stockholm, Sweden, from 2017 to 2022. ICD-10 codes from primary care, specialist outpatient, and hospital care were grouped into 60 disease categories. One-year cumulative incidence and prevalence were estimated and compared. Disease patterns were broadly similar across care settings, with hypertension being the most common diagnosis. However, underestimation was generally greater for incidence than for prevalence. Hospital and specialist care data primarily captured acute and severe conditions, e.g., stroke, ischemic heart disease, falls, and cancer, while primary care data more effectively identified chronic risk factors such as diabetes, hypertension, hyperlipidemia, and psychiatric disorders, particularly for incident cases. Age-related differences in underestimation varied by disease but showed no consistent pattern. Primary, specialist, and hospital care data each capture distinct aspects of the disease landscape in older adults. Excluding primary care data leads to systematic underestimation of many common and chronic conditions, especially for newly diagnosed cases. This study provides guidance for epidemiological research using administrative health registers and highlights the importance of integrating data from multiple levels of care to improve the accuracy of disease burden estimates in epidemiological research using administrative health registers.