Discussion

In this self-controlled case series study, which included more than 11 million individuals who received the COVID-19 vaccine in South Korea, we observed no association between COVID-19 vaccination and the risk of anxiety and nervousness, mood disorders, perceptual disturbances and psychoses, aggression and behavioural disturbances, cognitive impairments or sleep disorders. However, the risk of anxiety/nervousness or sleep disorders following COVID-19 vaccination increased modestly in individuals with no history of psychiatric disorders. According to stratified analyses, individuals vaccinated with the viral vector only were likely to experience anxiety/nervousness and sleep disorders after vaccination, and an increased risk of anxiety/nervousness, cognitive impairments and sleep disorders was observed in those aged 40–65 years old. Sensitivity analyses showed consistent results with our main findings.

To date, the profile and biological mechanisms underlying the psychiatric manifestations following COVID-19 vaccination remain unclear. Case reports of psychiatric symptoms after COVID-19 vaccination ^{Reference Reinfeld, Cáceda, Gil, Strom and Chacko20–Reference Uvais24} have suggested potential mechanisms involving immune activation through regulation of T-lymphocytes and proinflammatory cytokines. ^{Reference Tiyo, Schmitz, Ortega, da Silva, de Almeida and Oshiro25} Cytokines, which are mediators of central nervous system (CNS) activity caused by inflammation resulting from viral infections, ^{Reference Hosking and Lane26} are also involved in modulating neuronal activity in the amygdala, hippocampus, hypothalamus and cerebral cortex. ^{Reference Elenkov, Wilder, Chrousos and Vizi27} Proinflammatory cytokines such as interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF-α) have been mostly studied in relation to CNS and are known to be associated with cortical excitability. ^{Reference Galic, Riazi and Pittman28} IL-1β and TNF-α can be associated not only with seizures, but also with fear/panic attacks and sleep regulation. ^{Reference Irwin and Opp29,Reference Rosen and Schulkin30} SARS-CoV-2 infection also shares similar mechanism to psychiatric manifestations and is known to present with a wide range of neuropsychiatric symptoms. ^{Reference Wu, Xu, Chen, Duan, Hashimoto and Yang31–Reference De Berardis, Di Carlo, Di Giannantonio and Pettorruso33} A previous study that followed up on COVID-19 survivors for 1 month ^{Reference Mazza, De Lorenzo, Conte, Poletti, Vai and Bollettini34} showed that 56% of survivors exhibited at least one significant degree of psychopathology. Among these survivors, anxiety and insomnia were the most prevalent symptoms, with rates of 42 and 40%, respectively. However, it remains uncertain whether the risk of developing psychiatric symptoms in patients with COVID-19 is attributed to the viral infection itself, the host’s immune response or the psychological distress related to the diagnosis of COVID-19. Similarly, the association between COVID-19 vaccination and the risk of psychiatric adverse events may also be attributed to the stress caused by vaccination. ^{Reference Mopuru and Menon35} During the acute phase of the pandemic, COVID-19 vaccination in South Korea was mandated as a prerequisite for individuals to participate in community activities, suggesting that this situation might have induced psychiatric stress via high pressure on people who do not want to get vaccinated because of fear and uncertainty about the safety of COVID-19 vaccines.

Our study revealed significantly higher psychiatric adverse events in the middle-aged (40–65 years old) group than in the older group. Our findings are not consistent with the hypothesis that an immune response to COVID-19 vaccination can induce psychiatric symptoms through a mechanism similar to SARS-CoV-2 infection, considering that SARS-CoV-2 neurotransmission is known to pose higher risks with advancing age. ^{Reference Panariello, Cellini, Speciani, De Ronchi and Atti36} Regarding anxiety disorder in this study, the significant increase in the diagnoses following vaccination in the young and middle-aged adult group, in whom such diagnoses are most likely to occur, ^{Reference Angst, Paksarian, Cui, Merikangas, Hengartner and Ajdacic-Gross37–Reference Olaya, Moneta, Miret, Ayuso-Mateos and Haro40} suggests that vaccination may have triggered latent disorders. Regarding sleep disorders, younger adults are known to exhibit a heightened susceptibility to hypersomnolence, whereas the prevalence of insomnia tends to increase with advancing age. ^{Reference Ram, Seirawan, Kumar and Clark41,Reference Kerkhof42} However, the relevant evidence from prior literature reviews is limited; therefore, a more detailed interpretation of our results requires a thorough examination of each specific sleep disorder. In addition, these findings may reflect the unique vulnerabilities of this population during the COVID-19 pandemic, including increased psychosocial stress, occupational instability and caregiving responsibilities, which could have contributed to a higher risk of psychiatric symptoms. The differences from previous studies may be attributable to variations in study populations, settings or analytical methods. Further investigation is warranted to clarify these age-specific patterns and their underlying mechanisms.

It is known that adverse events may vary between different COVID-19 vaccines. For instance, BNT162b2 and mRNA-1273 vaccines have been associated with fewer neurological adverse events than the Ad26.COV2-S vaccine. ^{Reference Ahmad, Abbas, Ahmed, Mahmood, Kakamad and Baba43} In this study, we observed that viral vector vaccines were grossly associated with a higher risk (Supplementary Table 6); however, upon analysing individual vaccines, this trend was not consistent regarding psychiatric adverse events, as these manifestations seemed to differ across each individual vaccine (Supplementary Table 7).

In our study, we observed elevated incidence rates of perceptual disturbances and psychoses during the pre-vaccination period, particularly among individuals with a history of psychiatric disorders. These findings suggest the possibility that symptoms may have emerged before vaccination, or that increased healthcare interactions in anticipation of vaccination might have led to the identification of pre-existing conditions. Although the underlying biological mechanisms remain unclear, this highlights the importance of carefully defining pre-risk periods to detect and account for potential reverse causality in self-controlled case series analyses. Further research is warranted to better understand these patterns and their implications.

It seems that there is no mention of psychiatric-related symptoms and diagnoses related to suspected adverse events after COVID-19 vaccination in South Korea, within the main list of the Korea Centers for Disease Control and Prevention (KCDCP) reported by medical institutions. Additionally, the majority of adverse events reported to the KCDCP were observed within 1 day after vaccination. Anxiety disorder and sleep disorder, which are not common side-effects, might not have been well documented in the Vaccine Adverse Event Reporting Systems in the USA and Europe, unlike our study. ^{Reference Gallo, Goede, Mura, Abel, Moahamed and Preissner12} Previously, the Centers for Disease Control had announced that anxiety-related adverse events following immunisation may occur after COVID-19 vaccination. Associated symptoms include tachycardia, hyperventilation, dyspnoea, chest pain, paresthaesia, light-headedness, hypotension, headache, pallor and syncope. ^{Reference Loharikar, Suragh, MacDonald, Balakrishnan, Benes and Lamprianou44} However, it is crucial to note that anxiety-related adverse events following immunisation primarily occurs during the 15-min observation period after vaccination at the vaccination site, and it refers to a condition caused by observing and being influenced by the symptoms of others. This differs significantly from anxiety as an adverse event following vaccination, which is discussed in our study.

To the best of our knowledge, this is the first study to investigate whether COVID-19 vaccination is associated with an increased risk of developing a broad range of psychiatric adverse events. We implemented a self-controlled case series design to address the source of bias from time-invariant confounders by comparing vaccinated and unvaccinated periods within individuals. Our vaccine registry database contained precise information on the timing of vaccination, suggesting a minimal impact of misclassification of the unvaccinated window. In addition, claims data based on a single-payer healthcare system can reflect real-world clinical settings in South Korea. However, this study has several limitations. First, as our study outcomes were assessed based on routinely collected diagnostic codes without additional adjudication, the possibility of outcome misclassification cannot be ruled out, particularly as the data were collected for billing or claims operation, not clinical research. Moreover, the qualifications of assessors might not be available, which could influence the quality of information. However, the validity of diagnostic codes in the claims data has been proven with an 82% predictive positive value when compared with electronic medical records. ^{Reference Park, Jeon, Lee, Seo and Kim45} Second, a healthy vaccine effect may exist in our study population. However, our self-controlled case series design compared individuals with themselves; thus, we believe that the healthy vaccine was unlikely to affect our findings. Third, despite applying a self-controlled study design that can address time-invariant, unmeasured confounders or residual confounding by time-variant factors may have been present. Fourth, because of the short study period, our findings were largely attributed to the first and second doses, indicating that further studies should be conducted to evaluate the third and fourth doses. Fifth, as our study was based on a single ethnicity, it is unclear whether the findings may be extrapolated or replicated to other ethnic populations. Sixth, other side-effects following vaccination may cause patients to seek medical care in departments other than psychiatry, potentially reducing the likelihood of the visit of psychiatric clinics. However, we conducted a sensitivity analysis by dividing the 21-day risk window into 7-day intervals, and the results showed that the risk in each 1-week period was largely consistent (Supplementary Table 11). This finding suggests that the likelihood of patients visiting psychiatric clinics and having the outcome identified did not differ significantly within our defined risk window. Seventh, the diagnostic codes defined as study outcomes were broadly inclusive. However, certain conditions (such as ICD-10 codes F06, F19, F20 and F91) may raise concerns regarding the temporal link or the need for a longer duration of symptoms than the designated risk period. Therefore, we examined the number of patients diagnosed with ICD-10 codes F06, F19, F20, F91 and F93 from 2021 to 2022, using our database, which includes over 11 million individuals. The results are presented in Supplementary Tables 12 and 13. We found that the proportions of diagnoses of ICD-10 F19, F20, F91 and F93 were very low, ranging from 0.01 to 0.37%. Although the proportion of ICD-10 code F06 was also low at 1.44%, a breakdown of the four-digit codes revealed that most cases (92.1%) were classified as mild cognitive disorder. Additionally, our research team includes a psychiatrist (one of the corresponding authors) with substantial clinical experience treating patients during the COVID-19 pandemic, and their expertise contributed significantly to the definition of the study outcomes. Therefore, we believe that these diagnostic codes did not have a significant impact on our findings. Lastly, our study population included individuals with pre-existing psychiatric conditions, which introduces the potential for bias caused by concurrent use of psychotropic medications that was not adjusted for in the self-controlled case series model. Although this source of bias cannot be completely ruled out, psychotropic medications prescribed for psychiatric conditions are typically administered on a daily, long-term basis. Therefore, we believe that the self-controlled case series study design might minimise potential bias arising from this factor. However, since some psychotropic medications are intended for short-term use, cautious interpretation is needed. Future studies should consider these medications as potential confounding factors or clarify their potential impact on study outcomes.

In conclusion, we did not find an association between COVID-19 vaccines and the short-term risk of anxiety and nervousness, mood disorders, perceptual disturbances and psychoses, aggression and behavioural disturbances, cognitive impairments or sleep disorders. However, we found a modestly increased risk of anxiety/nervousness or sleep disorders following COVID-19 vaccination in individuals without any history of psychiatric disorders. As we wait for the results of confirmatory studies, our findings provide short-term safety profiles for COVID-19 vaccines regarding psychiatric adverse events, and support that individuals with psychiatric disorders should be included in the priority groups for vaccination. In the meantime, continuous monitoring of anxiety/nervousness or sleep disorders after COVID-19 vaccination is required, even when vaccinated individuals have no history of psychiatric comorbidities.