The present study revealed an estimated 34% excess relative risk and an attributable risk of 155 per 1000 person-years for children and adolescents with documented SARS-CoV-2 infection in 2020 to suffer from unspecific PASC within the first three months after the index quarter. Considering a mean time-at-risk of 0.365 years per person during this period, this means 5–6% of the children and adolescents with a documented SARS-CoV-2 infection showed onset symptoms that are statistically attributable to COVID-19. The relative risk in the COVID-19 group to suffer from specific PASC such as smell/taste disturbance, and fatigue was almost 330% corresponding to an excess incidence of 21 per 1000 person-years (Supplementary Table S3). Put differently, almost 1% of the children with a documented SARS-CoV-2 infection had an onset diagnosis of smell/taste disturbance (0.3%) or fatigue (0.5%) attributable to COVID-19.

At 9–12 months of follow-up, IRRs for many common unspecific symptoms remained close to 1.5, which hints at changes in the immune response toward other infectious diseases following COVID-19. In line with other studies, we found a considerable drop in the relative risk of COVID-19-specific symptoms after one year. We noted a general pattern of increasing incidence over time for almost all health conditions in the control group as visualized in Supplementary Fig. 1. This could be attributed to the (accumulating) baseline risk of symptom onset in non-prevalent cases over time. In the COVID-19 group, this pattern only applied to unspecific conditions.

Age-stratified results suggested that adolescents were more affected by PCC-specific symptoms. Most likely, they can describe their symptoms much more elaborately than younger children. Adolescents were also more affected by respiratory problems following COVID-19 compared to children. Furthermore, we noticed a shift in the excess incidence of diagnoses after 0–3 months versus after 9–12 months pointing towards an aggravation of respiratory symptoms in many adolescent patients, e.g. ΔIR of dyspnea decreased by almost 50% (16.9 versus 8.7 per 1000 person-years) whereas ΔIR of respiratory insufficiency doubled (1.4 versus 2.8 per 1000 person-years).

We found a particular increase in the risk of inflammatory disorders in children. We suppose there are different aspects of inflammatory processes represented in our estimates reflecting immediate immune response (e.g. PIMS/MIS-C) as well as delayed immune response (e.g. juvenile arthritis). COVID-19 as a potential driver of inflammatory (auto-) immune response is broadly discussed. Previous electronic health records-based studies reported an excess risk of 20–40% for onset autoimmune disease after COVID-19 [28, 29].

ME/CFS is a serious disabling chronic disease. Patients suffer from overwhelming fatigue, which is not improved by rest and worsens after any kind of physical or mental activity. Current hypotheses on the interference of ME/CFS and COVID-19 include the possible exacerbation of latent pathogenic factors following acute COVID-19, or particular manifestations of PCC over the course of 6 months meeting the criteria of ME/CFS [30]. Both hypotheses are supported by our analyses. We found very high effects for ME/CFS at 0–3 months after the index quarter, albeit with broad confidence intervals especially in younger children (IRR: 7.50, 95% CI 0.59, 96.58). While we found significant excess relative risk in adolescents after 6–9 months (IRR: 4.80, 95% CI 1.23, 18.74), estimating the risk beyond 9 months proved unrealistic due to the limited number of observable cases at this point.

Following individuals with onset PASC from quarter 1 to quarter 5 after the index, we found most conditions showed similar persistence rates in the COVID-19 and control cohorts. For the majority of incident health outcomes our findings suggest a similar prognosis regarding symptom resolution in both study groups, although elevated persistence rates among COVID-19 patients were visible for some diagnoses (discussed below). From the patients with incident health outcomes, only a minority showed symptoms persisting for more than one quarter. Following patients for a further two quarters (6–9 months), and three quarters (9–12 months) after the first quarter with an onset diagnosis reduced the persistence rates of most outcomes to under 20%, and under 10%, respectively. This is considerably less compared to the persistence rates found by some studies relying on longitudinal survey data [14, 23, 31]. A Danish study including 6630 adolescents with a history of SARS-CoV-2 showed that for many of the symptoms commonly associated with long/post-COVID about half of the patients with PASC reported ongoing symptoms after 6 months and later [14]. In a comparable cohort from the UK, persistence rates after 6 months were in a similar range for typical symptoms such as tiredness or shortness of breath and at about 20% for several other incident symptoms. However, for many of the considered outcomes there were only small changes in symptom persistence after 6–12 months [23]. Another questionnaire-based study from Denmark including 15,041 children and adolescents reported ongoing symptoms in about 40% of the children and about 50% of the adolescents during at least four months after acute COVID-19 [31].

The finding of similar recovery rates in both cohorts in our study does not negate the general burden of disease among the pediatric population caused by post-acute effects of COVID-19. Similar symptom persistence in both groups also means a lasting gap of excess incident health problems in children and adolescents attributed to COVID-19, which results in a long-term impact on pediatric healthcare. The increased incidence of symptoms at more than 6 months post-infection, as was also noted elsewhere [23], further adds to a persisting gap of morbidity among children and adolescents.

Some outcomes, especially cardiac symptoms, gastrointestinal symptoms, fatigue, and inflammatory disorders showed both initial excess risk of onset and prolonged persistence, meaning children and adolescents with a history of COVID-19 are affected more often by these symptoms and take longer to recover from them. In comparison, most neuropsychiatric disorders showed smaller excess risk and almost equal symptom persistence, yet increasing incidence rates in both the COVID-19 and control cohort (Supplementary Fig. 1). This could point to mental health outcomes being raised by social restrictions and closures of childcare or educational infrastructure rather than PASC, which is in line with previous evidence from electronic health records [32].

Many conditions mentioned on the WHO “broad” list of PCC-related symptoms showed IRRs of at least 1.3 in the first quarter of follow-up in the present analysis (e.g. stomachache, headache, fever, cough, dyspnea, diarrhea, nausea, throat/chest pain, palpitations, joint pain, myalgia, mood swings, and dizziness/vertigo) [3]. Of the three most specific PCC symptoms named by the WHO, we can confirm smell/taste disturbance, and fatigue (in terms of ME/CFS and malaise/exhaustion) as relevant health outcomes showing very high IRRs in our data. In contrast, anxiety was not considerably more frequent in the COVID-19 cohort compared to controls (Supplementary Table S3). We therefore suggest reevaluating the PCC-specific symptom list based on accumulating evidence, potentially by including other relevant conditions such as respiratory insufficiency, and inflammatory disorders. Our findings confirm the relevance of cardiovascular symptoms, gastrointestinal symptoms, and fatigue as listed by the Delphi-based PC-COS. By contrast, we found respiratory problems and smell/taste disturbances, which are not included in the PC-COS, as equally important symptoms in children and adolescents based on physician-made diagnoses.

Major strengths of the present study include the large database of routinely documented healthcare data reflecting the evaluation made by physicians, and the matched control design. Therefore, potential bias due to recall bias or selective self-reporting of symptoms among patients is minimized. Nevertheless, there are several sources of potential bias in secondary analyses of routine healthcare data. First, physicians can only evaluate conditions in children and adolescents seeking medical care. In the COVID-19 cohort, this could have been more often the case, and diagnostic bias due to greater awareness of PASC cannot be excluded. We therefore included two negative outcome controls (prescription glasses, acne) which were consistently found unrelated to COVID-19 in the present analysis. There is however, an unquantifiable risk of underestimating the incidence and persistence of health outcomes, which is likely to affect both study groups equally. One potential reason are barriers in the access to healthcare imposed by the social distancing measures in Germany during the studied period. In addition, the individual burden of the disease, hesitation, or frustration could present other potential barriers for patients to (re-)visit a doctor. In the persistence analysis, we aimed to minimize this bias by allowing for one intermediate “missing” quarter between two recurrent diagnoses to fulfill the persistence criterion. Second, routine healthcare data are not primarily collected for scientific use and we cannot exclude misclassification bias due to uncertainty in the accuracy of diagnoses or diagnosis due to other reasons (e.g. access to further medical care due to the diagnosis of ME/CFS). Third, follow-up in the present study was limited until September 30, 2021. The results of the present study are thus confined to data documented during the earlier phases of the COVID-19 pandemic dominated by the wildtype and alpha variant of SARS-CoV-2.