We used three nationwide population-based cohorts for this study: a Korean claim-based cohort (K-COV-N cohort; main cohort; n=10,027,506), a Japanese claim-based cohort (JMDC cohort; replication cohort A; n=12,218,680), and a prospective cohort from the UK Biobank (UKB cohort; replication cohort B; n=468,617). The study was approved by Kyung Hee University (KHSIRB-23-241), the Korea Disease Control and Prevention Agency (KDCA), the National Health Insurance Ser vice (NHIS; KDCANHIS-2022-1-632), JMDC (PHP-00002201-04), and UKB (94075). The requirement for written informed consent was waived because all personal identifiers in the administrative data were anonymized.

The K-COV-N cohort is a population-based cohort in South Korea conducted by the NHIS of South Korea, KDCA, and Korea National Statistical Office (Supplementary Table 1) [9]. The K-COV-N contains a comprehensive set of information regarding COVID-19 (results of diagnostic tests, vaccination status, and clinical outcomes), national health examinations, healthcare access (outpatient and inpatient), and death. The International Classification of Diseases, Tenth Revision (ICD-10) was used for disease categorization, and the overall positive predictive value of the diagnosis code was 82% in the South Korean health insurance claims database [10]. The registry was described in detail previously. Several distinctive features were observed in this cohort. First, it was conducted within the framework of the comprehensive healthcare system established by the Korean government, which extended its coverage to include complementary healthcare services for individuals diagnosed with COVID-19. Second, the study relied on data derived from the national COVID-19-related registry, which was meticulously built by the KDCA. Finally, to uphold the principles of confidentiality and protect the privacy of individuals involved, the Korean government has adopted measures to anonymize patient-related information [9].

The study sample comprised participants aged ≥20 years registered between January 1, 2018, and December 31, 2021. To focus on patients with long COVID, we structured the data starting from January 1, 2018, to exclude individuals with a history of gastrointestinal disease, hepatobiliary disease, or other digestive abnormalities in the lookback period (2018–individual index data) or within 30 days of SARS-CoV-2 infection; those with insufficient information on socioeconomic status; and those with co-infection or reinfection with COVID-19 (excluded n=3,004,933).

Primary exposure, ICD-10 codes for defining outcomes, observational periods, and statistical methods used in the discovery cohort were consistently applied to the validation cohorts, except for the vaccination records which were not available in validation cohorts A and B. In addition, detailed descriptions of the JMDC cohort and UKB cohort are provided in the Supplementary Methods. Finally, a 1:4 propensity score matching was conducted for validation cohorts A and B, resulting in their alignment.

The study populations comprised patients (≥20 years of age) with SARS-CoV-2 infection and non-infected individuals, which were randomly selected as controls. To reduce immortal time bias, we assigned an individual index date as the date of the first diagnosis of COVID-19 in patients with COVID-19, or as the matched date of the corresponding infected case in non-infected individuals.

SARS-CoV-2 infections were defined as exposure and confirmed through real-time reverse transcriptase polymerase chain reaction assays or antigen testing of nasal and pharyngeal swabs, following the guidelines of the World Health Organization [11]. The primary outcome of the study was the development of 17 gastrointestinal diseases, 8 hepatobiliary diseases, and 9 other digestive abnormalities in adults (≥20 years of age), defined as at least one claim after 30 days of the index date (Supplementary Table 1). Outcomes were also defined using ICD-10 codes (Supplementary Table 2) [6]. In addition, a gastroenterologist with more than 5 years of clinical experience (KL) and a hepatologist with more than 10 years of clinical experience (YH) reviewed and selected the ICD codes for this study.

The baseline characteristics of the individuals before propensity score matching show that the discovery cohort comprised 10,027,506 individuals (mean age 48.4 years; 49.9% female), as presented in Supplementary Tables 4 to 6. After applying the exclusion criteria and propensity score matching (Fig. 1), we identified 90,399 patients with COVID-19 (mean age, 46 years; 57.7% male) and 386,787 noninfected individuals (mean age, 47 years; 58.3% male) in the discovery cohort (Table 1). The final sample sizes after matching were 2,654,945, with a mean age of 43.26 years and 64.16% male, comprising 602,058 patients with COVID-19 and 2,052,887 non-infected individuals for validation cohort A (Supplementary Table 7) and 322,920 (75,814 patients with COVID-19 versus 247,106 non-infected individuals) for validation cohort B (Supplementary Table 8). The density and box plots of the three nationwide cohorts before and after matching are presented in Supplementary Figures 1 to 3. After matching, no significant differences were identified in the controlled covariates in the discovery cohort (all SMDs <0.1).

HRs were calculated using two methods: adjusted for age and sex (Model 1) and maximally adjusted for all variables (Model 2). Model 2 was used to present the results. In all three cohorts, we identified a higher risk of gastrointestinal disease (HR 1.15; 95% CI 1.08–1.22; discovery cohort) in individuals who had acute SARS-CoV-2 infection than non-infected individuals (Table 2). Specifically, the incidence of gastroesophageal reflux disease (GERD; HR 1.36; 95% CI 1.18–1.56), and functional intestinal disorders (HR 1.21; 95% CI 1.00–1.47) was significantly increased during the post-acute phase of SARS-CoV-2 infection (Supplementary Table 9). We found that SARS-CoV-2 infection was consistently associated with a higher risk of GERD, and functional intestinal disorders in validation cohort A (Supplementary Table 10). Additionally, in an alternative cohort where a more stringent definition was applied, we observed similar patterns (Supplementary Tables 11 and 12).

SARS-CoV-2 infection was associated with a higher risk of developing hepatobiliary disease after 30 days of initial diagnosis in the discovery cohort (HR 1.30; 95% CI 1.09– 1.55), as well as in validation cohorts A and B (Table 2). In particular, the HR for the inflammatory liver disease was 1.73 (95% CI 1.31–2.26) during the post-acute phase of COVID-19 (Supplementary Table 9). A significant correlation between COVID-19 and incidence of inflammatory liver disease was also identified in the validation cohort A (HR 1.62; 95% CI 1.56–1.69; Supplementary Table 10). Additionally, in an alternative cohort where a more stringent definition was applied, we observed similar patterns (Supplementary Tables 11 and 12).

The risk of other digestive abnormalities was higher in those with SARS-CoV-2 infection than in the non-infected individuals in the discovery cohort (HR 1.05; 95% CI 1.01– 1.10), validation cohort A (HR 2.08; 95% CI 2.05–2.12), and validation cohort B (HR 1.06; 95% CI 1.00–1.12; Table 2). The risk of developing symptoms of ‘nausea, vomiting, and dysphagia’ following SARS-CoV-2 infection was significantly higher in the discovery cohort than in the other cohorts (HR 1.28; 95% CI 1.17–1.40; Supplementary Table 9). Analysis of the validation cohort A revealed consistent results (HR for ‘nausea, vomiting, and dysphagia’ 2.58; 95% CI 2.51–2.64; Supplementary Table 10). Additionally, in an alternative cohort where a more stringent definition was applied, we observed similar patterns (Supplementary Tables 11–18).

In this multinational population-based cohort study, we investigated the incidence of gastrointestinal and hepatobiliary disorders and other digestive abnormalities >30 days after SARS-CoV-2 infection. Digestive disorders encompass a comprehensive spectrum of relevant diseases and symptoms and are categorized into gastrointestinal diseases, hepatobiliary diseases, and other digestive abnormalities. This study had several major findings. We identified an increased risk for incident gastrointestinal diseases (such as gastroesophageal reflux disease, functional intestinal disorders, and gastritis and duodenitis), hepatobiliary diseases (such as inflammatory liver diseases), and other digestive abnormalities (such as nausea, vomiting, and dysphagia) in patients with long COVID compared with uninfected control individuals. The risk was pronounced according to the COVID-19 severity. SARS-CoV-2 vaccination reduced the risk of gastrointestinal diseases but not hepatobiliary diseases and other digestive abnormalities. Over time, the risk profile was most pronounced during the initial 3 months; however, it persisted for >6 months in validation cohorts A and B, but not in the discovery cohort.

Evidence indicates that SARS-CoV-2 can directly and indirectly affect the digestive system, including the gastrointestinal and hepatobiliary tracts [17-20]. The potential mechanisms of damage to the digestive system during and after SARS-CoV-2 infection might include increase in intestinal permeability and altered gut microbiota, thus disrupting immune homeostasis [21-23]. Accordingly, the clinical impact of acute SARS-CoV-2 infection on digestive disorders has been demonstrated in previous studies during the COVID-19 pandemic [17,19,24]; however, the evidence was insufficient in patients who recovered from acute infection, despite persistent or new-onset digestive disorders that many of these patients experience. We hypothesized that the risk of digestive disorders after the acute phase (>30 days) of SARS-CoV-2 infection differs between infected and non-infected individuals. Some studies have previously assessed the incidence of digestive outcomes in patients who recovered from acute COVID-19 [6]. A meta-analysis of ten studies (until 2022) identified an increased risk of irritable bowel syndrome during the post-acute phase of COVID-19 [25]; however, the results cannot be generalizable considering the limited number of patient population and primary outcomes that were evaluated, relatively short duration of follow-up, as well as the lack of external validation of these studies.

Our analysis included a large number of individuals from multinational population-based cohorts, observed the incidence of outcomes for longer periods of time (>6 months), and used a comprehensive list of ICD-10 code-based digestive diseases and abnormalities. We set the observation period to assess the time-attenuation effect at 3 and 6 months, in accordance with a previous study on long COVID and functional gastrointestinal disorders [25]. In the discovery cohort from South Korea, the risk persisted for three months. However, in the cohorts from Japan and the UK, the increased risk persisted for ≥6 months. These differences could be attributed to variations in culture and ethnicity among the countries [26-28]. We used robust and rigorous statistical methodologies to determine the validity and reliability of our findings. However, further analysis will be necessary to explore these differences in more detail. We used a robust and rigorous statistical methodology to determine the validity and reliability of our findings. Our findings were generally in accordance with those of previously published literature and therefore confirm that our study provides clinically valid estimates of the post-acute impact of COVID-19 on adverse digestive outcomes [6].

Our study had some limitations. First, our data did not include the epidemic period of the Omicron variant. Second, outcome variables were defined from ICD-10 codes derived from nationwide registries. Although most of the codes used have been previously validated with high reliability, the possibility of misclassification exists, considering the differences in the primary outcomes among the three cohorts. Moreover, although the positive predictive values have been validated for most of the codes used, the negative predictive value has been studied less extensively [29]. The nationwide nature of the study population increases the generalizability of our findings; however, if systematic misclassification exists in patients not receiving the code yet having the disease, the validity of the study would decrease. Third, the observational nature of the study necessitates that the conclusions relate only to associations and not causal relations, although we analyzed the effect across a time span to reduce the risk of reverse causation. Fourth, although the study sample was assessed using hospital coding and pharmacy prescriptions, the proportion of comorbidities treated with lifestyle interventions might not have been identified. Fifth, exclusion of reinfected individuals hinders the detailed assessment of dose-response relationship. However, the risk of reinfection with SARSCoV-2 is very low, estimated to be around 0.3% [30]; therefore, such analysis might not be suitable for this dataset. Sixth, it is not completely determined whether the digestive outcomes were caused by COVID-19 per se or by medications, such as corticosteroids. Seventh, information on SARS-CoV-2 vaccination status was not available for validation cohorts A and B; therefore, analysis in this regard could not be performed. Eighth, claims data from Japan and South Korea were constructed without participant consent owing to single-payer healthcare systems; by contrast, the UK Biobank dataset was created through voluntary participation. Despite the diversity of these datasets, demonstrating similar outcomes enhances the reproducibility of our findings. Nineth, the prevalence of each disease varies among countries. However, when considering the specific characteristics of each country, similar outcomes related to long COVID were evident. This indicates the generalizability of our findings across multiple countries. Tenth, although all potential covariates were adjusted for in the analyses, the possibility of residual potential confounders such as dietary patterns cannot be excluded [31]. Furthermore, original viral strain and delta were inferred by the timing of confirmed infection, without biological evidence confirming the strain. Lastly, monitoring patients with COVID-19 could result in higher rate of referral and detection of digestive disorder outcomes compared to the general population. To address potential detection biases related to differential healthcare access, baseline risk differences, and outcome ascertainment between patients with COIVD-19 and general population, we used several strategies. These included external validation of our findings in other national cohorts to enhance reliability and offset potential issues with digestive disorder diagnosis codes; confirmation of consistent trends even with diverse definition conditions by altering the conditions of the ICD codes; ensuring consistent association by altering the conditions of ICD codes; and considering and adjusting for hospital admissions and outpatient contacts before the index date to balance healthcare-seeking behaviors between groups. Despite these efforts, detection biases cannot be entirely ruled out. However, due to our robust methodology and extensive supplementary analyses, we expect the impact of these biases has been mitigated. Further research, including prospective cohort studies, is necessary to corroborate our findings.

Despite these limitations, to the best of our knowledge, this is the first study to provide comprehensive evidence of the differences in digestive disorders, including gastrointestinal and hepatobiliary diseases as well as other digestive abnormalities, during the post-acute phase of SARSCoV-2 infection. Stratified analyses based on vaccination status further intensified the impact of our study. Based on our study, the burden of digestive disorders may continue to increase even after the end of the COVID-19 pandemic. Vaccination against SARS-CoV-2 should be advised to the general population not only to prevent COVID-19 per se but also to reduce the risk of developing subsequent digestive disorders during the post-acute phase. Our study could help clinicians provide patient guidance regarding the impact of SARS-CoV-2 vaccination on the long-term digestive outcomes and could facilitate clinical decision on performing further blood and/or imaging tests. For greater generalizability, our results should be reproduced across other racial groups. To strengthen the findings of our study, additional studies using other infections as comparators are required. In addition, further research on COVID-19-reinfected patients is necessary to elucidate the risk of reinfection.

In conclusion, in this multinational population-based cohort study, we identified an increased risk of gastrointestinal and hepatobiliary diseases as well as other digestive abnormalities in patients with long COVID compared with uninfected control individuals. The risk was pronounced according to the COVID-19 severity. The SARS-CoV-2 vaccination was shown to reduce the risk of gastrointestinal diseases but had no impact on hepatobiliary diseases or other digestive disorders. Over time, the risk profile was most pronounced during the initial 3 months; however, it persisted for >6 months in validation cohorts A and B, but not in the discovery cohort. Our findings underscore the importance of recognizing digestive disorders following recovery from COVID-19.