2025 Volume 48 Issue 5 Pages 641-649
Long coronavirus disease (COVID) is characterized by symptoms persisting or reappearing at least 2 months post-recovery from acute coronavirus disease 2019 (COVID-19). Although Long COVID symptoms have been widely studied, data on drug prescriptions for patients with Long COVID in Japan remain limited. Therefore, this study aimed to analyze drug utilization patterns for Long COVID treatment using a nationwide database in Japan, with the goal of providing basic data to support the establishment of standard treatments in the future. The Medical Data Vision COVID-19 dataset was used to identify patients diagnosed with Long COVID between January 15, 2020 and December 31, 2022. Symptoms and prescribed medications were extracted, and descriptive statistics were used to analyze the relationship between symptoms and drug prescriptions. Among 652016 patients with COVID-19, 3769 (0.6%) developed Long COVID. Common symptoms included fatigue, bronchial asthma-like symptoms, and insomnia. Acetaminophen was the most prescribed drug in the first month of diagnosis. Other frequently prescribed drugs included dextromethorphan, l-carbocisteine, and polaprezinc. From 3 months post-diagnosis, prescriptions for Hochu-ekki-to (a traditional Japanese herbal medicine; Kampo medicine) and polaprezinc increased, especially among patients aged 30–50 years. Long COVID in Japan is characterized by a wide range of symptoms, leading to symptom-based drug prescriptions, particularly fatigue, respiratory issues, and taste disturbances. These findings offer insights into the pharmacological management of Long COVID in Japan, highlighting the need for further research on optimal treatments in the future.
Symptoms of coronavirus disease 2019 (COVID-19) include not only flu-like symptoms, such as fever, cough, phlegm, and malaise, but also diverse symptoms, including olfactory disturbances, taste disorders, digestive issues, dizziness, and hair loss.1–3) Although most patients with COVID-19 recover, some continue to experience persistent or new symptoms beyond the acute phase, and this is why the disease is called Long COVID.4) The WHO defines Long COVID as follows. “Symptoms persisting for at least 2 months that cannot be explained by symptoms of other diseases. Symptoms may be new or persistent after recovery from the acute phase of COVID-19. Symptoms are variable in severity and may reappear after symptoms have resolved.”5)
Several hypotheses exist regarding the causes of COVID-19 sequelae. COVID-19 may directly enter and proliferate in cells by binding to the angiotensin-converting enzyme 2 (ACE2) receptor, a spike-like projection that leads to tissue injury. ACE2 is present in the lung, brain, nasal and oral mucosa, heart, vascular endothelium, and small intestine6); therefore, it has been suggested that the sequelae may cause various symptoms. Other hypotheses include the effects of cytokines, the active virus itself, and inadequate immune responses due to low antibody levels.7,8)
Common Long COVID symptoms include fatigue, shortness of breath, myalgia, arthralgia, headache, cough, chest pain, olfactory and taste disturbances, and diarrhea. Previous studies have suggested that treatment in the ICU may increase the rate of subsequent complaints.9,10)
Long COVID has become a worldwide issue; in Japan, reports show persistent symptoms such as dyspnea, fatigue, and cough in patients post-COVID-19 recovery.11–13) However, despite many reports on the symptoms of Long COVID, the status of prescribed treatments for these sequelae in Japan remains unclear. Furthermore, single-center or small multi-center studies make it challenging to capture a comprehensive picture of treatment trends in Japan. Therefore, in this study, we analyzed a nationwide database to provide basic data for further research on establishing a standard treatment for Long COVID in Japan.
This study utilized the COVID-19 dataset provided by Medical Data Vision Co., Ltd. (MDV), Tokyo, Japan. The MDV dataset is a comprehensive database that incorporates Diagnosis Procedure Combination (DPC) and health insurance claims data from approximately 40 million patients across 460 acute care hospitals in Japan. This dataset covers all the regions of Japan. All personal data in the dataset were anonymized. The dataset encompasses tables, such as PatientData (patient information), DiseaseData (disease information), ActData (medical practice information), FF1Data (discharge summary), and Drugs (prescription drugs). ATCCODE was used to extract patient medications and KUBUNCODE was used to extract medical procedures. Data were extracted using SQL via Snowflake (Snowflake Inc., San Mateo, CA, U.S.A.).
Ethics ApprovalThis study was approved by the Kitasato University Medical Center Ethics Committee (Approval No. 20222028). The ethics committee confirmed that all methods were performed in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan and the relevant regulations. The requirement for informed consent was waived due to the use of anonymized data.
Dataset OverviewThe dataset comprised 854114 patients diagnosed with or suspected of having COVID-19 in Japan between January 15, 2020 and December 31, 2022. Subsequently, 652016 patients with confirmed COVID-19 diagnosis (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision; ICD-10 CODE: U071) were included. Of the 652016 patients with a confirmed COVID-19 diagnosis (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision; ICD-10 CODE: U071), 3769 patients with Long COVID diagnosis or COVID-19 sequelae (ICD-10 CODE: U09) after the diagnosis date were included in the analysis (Fig. 1). Patients with missing age and sex data and those with suspected COVID-19 (ICD-10 CODE: U072) were excluded. To determine whether the patient composition of the dataset was comparable to the overall age distribution of Japanese patients, we compared the number of COVID-19 positive patients in Japan14) reported during the same time period as the present study.
Patients with U071 and U09 in the ICD-10 were defined as those diagnosed with Long COVID; those with U072 with suspected COVID-19 were excluded. The period covered was between January 15, 2020 and December 31, 2022, in Japan. COVID-19, coronavirus disease 2019.
To collect Long COVID symptoms and comorbidities, we extracted diagnostic names and symptoms using ICD-10 codes, following the methods used in the studies of Daugherty et al. and Kinugasa et al.15,16) Data on patient background and medical history include sex, age, hypertension, chronic obstructive pulmonary disease, chronic kidney disease, type 2 diabetes, cardiovascular disease history (angina pectoris, myocardial infarction, heart failure, and other ischemic heart diseases), cerebrovascular disease, and dyslipidemia as well as intensive care unit (ICU) or ventilator use for COVID-19. The data extraction methods were based on previous reports.17) Patient background information and medical history were collected based on the month in which the COVID-19 and positive diagnosis were newly recorded, and for the preceding 6 months.
Duration of Long COVID, Symptoms, and Extraction of MedicationsMedications prescribed post-Long COVID diagnosis were defined as Long COVID-related medications. Patients were categorized into four intervals from Long COVID diagnosis : within 28 d (first month), 29–56 d (second month), 57–84 d (third month), and 85–168 d (fourth to sixth months), and tabulated the number of patients per time period of illness and the number of patients, sequelae symptoms, and Long COVID-related drugs. The continuity of prescriptions for the same patient at each interval was not analyzed in this study. The top 10 sequelae symptoms with the highest number of affected patients were selected and classified based on illness duration.
Statistical AnalysisDescriptive statistics were used. Continuous variables were expressed as mean and standard deviation (S.D.). Long COVID symptoms and duration, age intervals, and drugs were used as categorical variables, and the number of matches for each variable was used as a continuous variable to create a heatmap. Heatmaps were generated using GraphPad Prism version 9.5.1, for Mac (GraphPad Software, Boston, MA, U.S.A.) with hierarchical clustering and a grayscale ranging from white (low) to black (high).
Table 1 presents patient information: among the 652016 patients diagnosed with COVID-19, 324953 (49.8%) were male, 327063 (50.2%) were female, and 196371 (30.1%) were aged ≥65 years, with a mean age of 52.1 ± 20.9 years. Of these, 3769 patients had Long COVID, representing 0.6% of the total; the proportion of males and females was equal among patients with Long COVID. The most common age group among patients with Long COVID was 40–49 years, with a high proportion of patients aged 20–60 years overall (Fig. 2). The number of COVID-19-positive patients in the database was approximately evenly distributed between age groups among patients aged 0–50 years (Supplementary Fig. 1). Compared with the age distribution of Japanese patients as a whole, there were slightly more elderly patients in the database; however, the trend was consistent with the general distribution of numbers (Supplementary Fig. 2).
| Overall number of patients |
Long COVID |
Not Long COVID |
|
|---|---|---|---|
| Number of patients, n (%) | 652016 | 3769 (0.6) | 648247 (99.4) |
| Mean age, years (S.D.) | 52.1 (20.9) | 52.1 (21.0) | 52.1 (20.9) |
| Aged ≥65 years, n (%) | 197396 | 1025 (27.2) | 196371 (30.3) |
| Males, n (%) | 324953 | 1871 (49.6) | 323082 (49.8) |
| Females, n (%) | 327063 | 1898 (50.4) | 325165 (50.2) |
| Hypertension, n (%) | 133173 | 863 (22.9) | 132310 (20.4) |
| COPD, n (%) | 11286 | 159 (4.2) | 11127 (1.7) |
| Chronic kidney disease, n (%) | 25042 | 169 (4.5) | 24873 (3.8) |
| Type 2 diabetes, n (%) | 117683 | 941 (25.0) | 116742 (18.0) |
| Cardiovascular disease, n (%) | 173829 | 1426 (37.8) | 172403 (26.6) |
| Cerebrovascular disease, n (%) | 84358 | 702 (18.6) | 83656 (12.9) |
| Lipid disorders, n (%) | 97743 | 710 (18.8) | 97033 (15) |
| None of the above diseases, n (%) | 384869 | 1504 (39.9) | 383365 (59.1) |
| History of ICU treatment, n (%) | 13220 | 87 (2.3) | 13133 (2.0) |
| Use of ventilator, n (%) | 11953 | 136 (3.6) | 11817 (1.8) |
COPD, Chronic obstructive pulmonary disease; ICU, intensive care unit. The data are shown as n (%) unless otherwise noted. The percentages are calculated based on the number of patients within the Long COVID and not Long COVID groups.
Age distribution of patients diagnosed with Long COVID is shown for both men and women.
Among the 3769 patients diagnosed with Long COVID, 3340 presented at least one documented symptom. Among these patients, the top 10 symptoms accounted for 2852 cases (85.4%), and the 11th and subsequent total was 488 (14.6%). The most common symptom was fatigue, followed by bronchial asthma-like symptoms, insomnia/sleep disturbance, cough, headache, neurosis, depression/depression, dizziness, and olfactory and/or taste disorders (Fig. 3a).
(a) The symptoms analyzed are restricted to the top 10 Long COVID symptoms with the highest number of patients shown. The first horizontal axis shows the number of patients with each symptom and the second axis shows the percentage of each symptom in the total Long COVID patients. (b) Number of patients per time period. Other Long COVID symptoms with smaller numbers are not included in this figure. The number of patients who experienced Long COVID symptoms for the first time is shown separately for each time period.
Similarly, many Long COVID-related symptoms, such as fatigue, bronchial asthma-like symptoms, insomnia, phlegm, and headache, appeared markedly within the first month following Long COVID diagnosis (Fig. 3b). By contrast, depressive symptoms, dizziness, and taste and smell disorders tended to appear regardless of time.
Long COVID Symptoms and Medications by Time PeriodOf the 2852 patients who experienced the top 10 symptoms, the number of those who first experienced symptoms during each interval was as follows: first month, 1344 patients; second month, 416 patients; third month, 260 patients; fourth to sixth months, 468 patients and 7 months or later, 364 patients. These numbers represent newly identified cases in each period, rather than cumulative counts. The cumulative number of patients was not calculated because there is no information on the continuation or remission of patients’ symptoms. In the first, second, third, and fourth to sixth months, the top 10 most prescribed drugs accounted for 29.0% (884/3050), 33.4% (427/1278), 33.7% (330/875), and 54.0% (605/1120) of the total prescriptions, respectively (Table 2). The denominator is the total number of drugs prescribed during each period for patients with experienced symptoms.
| Time period | Top 10 prescribed drugs, n (%) |
Total prescribed drugs, n |
|---|---|---|
| First month | 884 (29.0) | 3050 |
| Second month | 427 (33.4) | 1278 |
| Third month | 330 (33.7) | 875 |
| Fourth–sixth months | 605 (54.0%) | 1120 |
| Total (all periods) | 2246 (35.5%) | 6323 |
The proportion of the top 10 most prescribed drugs among the total prescribed drugs recorded for patients experiencing Long COVID symptoms in each period. The denominator represents the total number of prescribed drugs during each period for patients with experienced symptoms. The total row represents the cumulative sum across all periods.
Acetaminophen was the most commonly prescribed drug associated with Long COVID, within the first month, followed by dextromethorphan and l-carbocisteine as antitussive expectorants (Fig. 4a). In the second month, polaprezinc was prescribed more frequently (Fig. 4b), and in the third month onward, prescriptions for Hochu-ekki-to (a traditional Japanese herbal medicine; Kampo medicine) increased (Figs. 4c, 4d).
a First month, b Second month, c Third month, d Fourth–sixth months. Each drug is shown as the top 10 drugs prescribed in the highest number for each period. The horizontal axis shows the number of patients that received that specific prescription and the percentage of all drugs prescribed during the period.
Analysis of medications prescribed for each symptom over the entire Long COVID period showed that antipyretic analgesics, including acetaminophen, were prescribed most frequently for symptoms commonly associated with the initial symptoms of COVID-19 (Fig. 5). In addition, as a characteristic formulation, Kampo medicines were often prescribed in patients with malaise, with Hochu-ekki-to prescribed increasingly after the third month, particularly for patients aged 30–50 years (Fig. 6). Acetaminophen, which had the highest total number of prescriptions, was prescribed not only within 28 d of the long-term onset of Long COVID, but also over a longer period.
This diagram illustrates the relationship between various symptoms of Long COVID and the treatments prescribed to manage these symptoms. They are also used for medicinal purposes.
a First month, b Second month, c Third month, d Fourth–sixth months. The top 10 drugs with the highest number of prescriptions per period are shown for each drug.
This study aimed to provide foundational data for the establishment of standard treatment strategies for Long COVID in Japan in the future. By analyzing a nationwide database, we identified trends in prescription medications after COVID-19 infection across different time periods and age groups. The findings revealed that a wide variety of drugs were utilized to address diverse symptoms, with acetaminophen being the most frequently prescribed medication in the early phase. Additionally, treatments unique to Japan, such as Hochu-ekki-to, were prominently used, particularly for symptoms such as fatigue, highlighting their potential role in the management of Long COVID. Medications for respiratory symptoms, including gastric antisecretory agents and antitussive expectorants, are commonly prescribed. As these drugs target different symptoms, medication selection for Long COVID in clinical practice appeared to follow a symptomatic approach. However, as this study did not track non-covered drugs, a more detailed analysis of newer treatments is needed for a comprehensive understanding of current therapeutic strategies.
The number COVID-19-positive patients in the database was approximately evenly distributed among patients aged 0–50 years. This result is consistent with the general distribution of the incidence of COVID-19. However, there was a larger number of long COVID patients in the 40 to 50-year age group in comparison with the general population. Future studies should investigate whether age and sex are associated with specific long COVID symptoms or treatment choices.
A retrospective cohort study by Cohen et al. reported a 32% incidence of Long COVID among older adults aged ≥65 years diagnosed with COVID-19.18) In younger populations, Long COVID incidence is also increasing.19) In this study, a higher prevalence of Long COVID was observed in younger age groups compared with older adults. Consistent with other studies, malaise, respiratory symptoms, and psychiatric issues were more prevalent in younger patients than in older adults. This study did not focus on children; however, similar symptoms as those in adults are of concern in children <18 years, with potential impacts on learning.20,21) However, differences by country, region, environment, and race remain unclear and should be considered, factoring in virus virulence and epidemic context.
In this study, symptoms appearing early and persisting post-Long COVID diagnosis included fatigue, bronchial asthma-like symptoms, insomnia/sleep disturbances, cough, and headache. These symptoms affect a large number of people and develop relatively easily. By contrast, symptoms appearing after the acute phase include neurosis and depression. Symptoms such as headache, fatigue, and taste and smell disorders were present in the acute phase. While most olfactory and taste disorders improve early, a small percentage of patients experience persistent symptoms for months or even a year or more, with some suffering from dysosmia and dysgeusia.22,23)
Although most symptoms were present within the first month, the proportion of prescriptions attributed to Long COVID remained consistent throughout the study period. However, the increase in the proportion of the top 10 drugs prescribed in months 4–6 may indicate a shift toward addressing only persistent symptoms. Conversely, because symptoms change over time, the result of a constant prescription rate does not imply that a certain percentage of drugs were prescribed to patients who continued to experience symptoms. The persistence of specific symptoms and how the treatment approach was adapted over time need to be examined in the future.
Acetaminophen prescriptions were frequently observed in Long COVID, which may have been prescribed for a wide range of symptoms, mainly headache and prolonged low-grade fever.16) Acetaminophen was chosen because it is less likely to cause gastric mucosal damage than nonsteroidal anti-inflammatory drugs. The study was unable to identify and extract symptoms of low-grade fever from the dataset.
Gastric antisecretory agents have been prescribed to treat several symptoms. Drugs for respiratory symptoms such as antitussives and expectorants tend to be prescribed for relatively long periods. Polaprezinc was also included in this analysis. Polaprezine is commonly used to treat taste disorders caused by zinc deficiency.24) In this study, polaprezinc was mainly used for taste disorders in Long COVID; however, it is unclear whether it is as effective as the treatment of zinc deficiency, as taste disorders caused by severe acute respiratory syndrome coronavirus 2 infection are considered to be neurological complications.25) There has been a trend toward polaprezinc use in taste disorders due to Long COVID. Whether patients with underlying medical conditions are at increased risk of taste disorders remains to be investigated. It is also unclear whether patients with mild taste disorders are exacerbated by COVID-19 infection and whether prescribing polapresin is appropriate must be discussed. This is because the relationship needs to be clarified regarding the treatment of the underlying disease and the administration of drugs that may induce taste disorders.
We assume that the use of polaprezinc in this study was primarily related to taste disorders; however, owing to the limited data, we cannot rule out the possibility that it was intended to relieve gastrointestinal symptoms in addition to taste disorders. Future studies should obtain more detailed clinical information to identify the intended use of polaprezinc.
Regarding fatigue, which affected the largest number of patients, prescriptions for Kampo medicines were prominent. Tokumasu et al. reported that approximately 90% of Kampo medicines prescribed for fatigue in Long COVID were Hochu-ekki-to or similar.26)
Hochu-ekki-to has immunomodulatory properties that may be useful during the recovery phase of COVID-19. Regarding Kampo medicine use for Long COVID in Japan, a study of 102 patients with Long COVID reported that Japanese herbal preparations were used in 24% of prescriptions.27) Therefore, the use of Kampo medicines like Hochu-ekki-to for Long COVID in Japan may be distinctive. However, more detailed studies are needed as there are no clear reports of their use and effectiveness in other countries.
Patients treated for COVID-19 on ventilators continue to experience memory difficulties and daily activity limitations 1 year after leaving the ICU, attributed to Long COVID.28) While many Long COVID symptoms tend to improve over time, questions remain regarding the progression of residual symptoms over extended periods and the impact of different viral variants. Further studies are needed to address these uncertainties.
Multifaceted support is needed for patients with COVID-19, not only during the acute phase but also post-recovery, as residual symptoms can impair health-related QOL, increase anxiety, depression, and fear, and decrease sleep quality.
The study had several limitations: First, the use of ICD-10 code U09 enables the broad identification of post-COVID-19 sequelae but does not capture detailed symptom characteristics, such as the severity of specific symptoms (e.g., brain fog or the intensity of fatigue). This limitation stems from the absence of granular clinical descriptions in the dataset. Second, patient records often vary in the consistency and completeness of symptom documentation, which may lead to underreporting of less common symptoms. To address these gaps, future studies should integrate patient-reported outcomes, clinical evaluations, and diagnostic tests alongside ICD-10 codes to enhance the specificity and comprehensiveness of symptom identification. Third, while this study utilized prescription drug data to infer treatment intent, there was a lack of explicit linkage between symptoms and prescribed medications in the dataset. For instance, certain drugs may have been prescribed for comorbidities rather than for Long COVID symptoms, given the constraints of insurance coverage and off-label use practices. Future research should explore methodologies to refine the association between symptomatology and therapeutic interventions, such as incorporating longitudinal clinical data or conducting surveys with healthcare providers to clarify treatment intent. Fourth, this study did not examine the association between comorbidities and COVID-19 sequelae. Therefore, future studies examining the impact of comorbidities on the prevalence of COVID-19 sequelae and prescribing of medications need to clarify and analyze the definition of comorbidities. Fifth, the proportion of patients with Long COVID in this sample was only 0.6% of total patients with COVID-19, a low figure compared with known Long COVID incidence rates. Possible reasons for this discrepancy include the dataset's reliance on receipt data from DPC-eligible hospitals, excluding patients treated in non-covered hospitals or at home. Additionally, both patient and hospital data are anonymized, obscuring hospital size, regional factors, and hospital functions potentially introducing bias. Although the MDV database may contain hospital size information, these data were not included in our analysis. Future research should explore the impact of hospital size on treatment patterns. Furthermore, the improvement of Long COVID symptoms with the use of COVID-19 medication, differences by mutant strain, and even the influence of vaccination cannot be considered. Detailed information on the severity and time to improvement of each Long COVID symptom was not available.
Long COVID in Japan involves diverse symptoms, leading to symptom-based drug prescriptions, particularly for fatigue, respiratory issues, and taste disturbances. These findings offer insight into the pharmacological management of Long COVID in Japan. However, while the results of this study provide basic data on Long COVID treatment, further research is needed to establish a standard of care. In particular, it is important to evaluate the impact of comorbidities on COVID sequelae and treatment choices.
The authors declare no conflict of interest.
This article contains supplementary materials.
