Neurological complications after first dose of COVID-19 vaccines and SARS-CoV-2 infection

Emerging reports of rare neurological complications associated with COVID-19 infection and vaccinations are leading to regulatory, clinical and public health concerns. We undertook a self-controlled case series study to investigate hospital admissions from neurological complications in the 28 days after a first dose of ChAdOx1nCoV-19 (n = 20,417,752) or BNT162b2 (n = 12,134,782), and after a SARS-CoV-2-positive test (n = 2,005,280). There was an increased risk of Guillain–Barré syndrome (incidence rate ratio (IRR), 2.90; 95% confidence interval (CI): 2.15–3.92 at 15–21 days after vaccination) and Bell’s palsy (IRR, 1.29; 95% CI: 1.08–1.56 at 15–21 days) with ChAdOx1nCoV-19. There was an increased risk of hemorrhagic stroke (IRR, 1.38; 95% CI: 1.12–1.71 at 15–21 days) with BNT162b2. An independent Scottish cohort provided further support for the association between ChAdOx1nCoV and Guillain–Barré syndrome (IRR, 2.32; 95% CI: 1.08–5.02 at 1–28 days). There was a substantially higher risk of all neurological outcomes in the 28 days after a positive SARS-CoV-2 test including Guillain–Barré syndrome (IRR, 5.25; 95% CI: 3.00–9.18). Overall, we estimated 38 excess cases of Guillain–Barré syndrome per 10 million people receiving ChAdOx1nCoV-19 and 145 excess cases per 10 million people after a positive SARS-CoV-2 test. In summary, although we find an increased risk of neurological complications in those who received COVID-19 vaccines, the risk of these complications is greater following a positive SARS-CoV-2 test.

system (CNS) demyelinating events, encephalitis meningitis and myelitis, Guillain-Barré syndrome, Bell's palsy, myasthenic disorders, hemorrhagic stroke and subarachnoid hemorrhage. We use the same population to investigate the associations between a positive SARS-CoV-2 test as a secondary exposure and the same neurological conditions. Incidence rate ratios (IRRs), that is, the rates of hospital admission or death from each neurological outcome in the risk period after vaccination or after a positive test relative to the baseline period, were estimated using the self-controlled case series (SCCS) methodology ( Fig. 2) 14,15 . A relative incidence greater than 1 indicates an increased risk after vaccination or a SARS-CoV-2-positive test. We also carried out an independent replication of the risk of neurological outcomes in a national cohort from Scotland using the SCCS design.

Results
Overall, 32,552,534 people received their first dose of COVID19 vaccine (ChAdOx1nCoV-19, n = 20,417,752; BNT162b2, n = 12,134,782) in England between 1 December 2020 and 31 May 2021. In the population of vaccinated people, 2,005,280 (~6%) had a SARS-CoV-2-positive test. Of those with a positive test, 1,838,628 (~91%) had their SARS-CoV-2 test prior to vaccination. The baseline characteristics of the study population are listed in Table 1.
The results for the investigated neurological outcomes following each exposure (ChAdOx1nCoV-19, BNT162b2 and positive SARS-CoV-2 test) are given for pre-specified time periods: 0, 1-7, 8-14, 15-21 and 22-28 days, as well as collectively for the 1-28 day period. The baseline characteristics of people who had these neurological outcomes are listed in Table 2. The IRRs for each outcome in the risk periods immediately before and after each exposure are given in Table 3 and Fig. 1.
Neurological outcomes. Acute CNS demyelinating events. A total of 1,105 people had a hospital admission with an acute CNS demyelinating event. Of these, 131 (11.9%) had a positive SARS-CoV-2 test including 121 (10.9%) who had a positive test prior to vaccination. There were fewer than five deaths overall. There was no association between the ChAdOx1nCoV-19 or BNT162b2 vaccination and admission with an acute CNS demyelinating event in any of the pre-defined 7 day risk periods.   (10) After vaccination 3.6 (11) 6.0 (9) *Cells with <5 are suppressed.   Table 3b). The findings for SARS-CoV-2 test positivity were similar to the English data, with an increased risk of neurological outcomes following a positive test, although the estimates were generally imprecise due to the smaller population and lower rates of infection.

Subgroup analyses by age group and sex.
We tested for interactions between vaccination and age group (>50 years and ≤50 years) and sex for the outcomes for which we observed an increased risk after vaccination. Supplementary Table 4a lists the results of the interaction tests. When interaction terms were statistically significant, the IRR for subgroup analyses by age group or sex have been reported in Supplementary Table 4b. There was a significant interaction (P = 0.016) between age group and risk of myasthenic disorder associated with the ChAdOx1nCoV-19 vaccine (IRR, 2.44; 95% CI: 1.07-5.55 at 15-21 days) in people aged ≤50 years. The increased risk of hemorrhagic stroke associated with the BNT162b2 vaccine was significantly higher in female participants (P = 0.007): the IRR for hemorrhagic stroke in female participants was 1.44 (95% CI: 1.05-1.96) at 1-7 days and 1.84 (95% CI: 1.40-2.42) at 15-21 days after BNT162b2 vaccination compared with 1.13 (95% CI: 0.82-1.57) at 1-7 days and 0.98 (95% CI: 0.70-1.38) at 15-21 days in male participants.
Excess events due to exposures. We estimated the number of exposures needed for one excess event and the excess number of events per 10 million exposed for each outcome (Supplementary Table 5 Associations with negative and positive control outcomes. We examined the associations of exposures with celiac disease as a negative control outcome and with anaphylaxis as a positive control outcome. We found no increased risk of celiac disease (negative control) across the pre-specified time periods for the vaccine exposures, but a decreased risk on the day of vaccination. Anaphylaxis (positive control) showed the expected increased risk in the  0-7 days after the first dose for both vaccines, but not in later periods (Supplementary Table 6).

Sensitivity analyses.
We conducted a number of sensitivity analyses to ascertain the robustness of our results. Overall, the main findings were not sensitive to censoring due to death (with the exception of hemorrhagic stroke), removal of patients who had a second dose of the vaccine or adjustment for potential delays in recording. Results for these sensitivity analyses are shown in Supplementary Table 7a and Extended Data Figs. 1 and 2. We also extended the exposure risk period to 5 weeks, to test the suitability of our 28 day risk period and to avoid missing events due to lag in diagnosis. Results were similar to the main analysis (Supplementary Table 7b). An increased risk of Guillain-Barré syndrome was observed in the 29-35 days after ChAdOx1nCoV-19 vaccination (IRR, 1.55; 95% CI: 1.03-2.34).
We have further restricted the analyses to patients who had a positive SARS-CoV-2 test before vaccination. The results did not change from the main analysis for the SARS-CoV-2 test exposure (Supplementary Table 7c), suggesting that there is a higher risk of neurological complications after a positive SARS-CoV-2 test for all outcomes. The number of people infected after vaccination was too small to carry out any additional meaningful analyses. We also restricted the analyses to those who did not have a positive SARS-CoV-2 test. The results did not change from the main analysis for both vaccine exposures (Supplementary Table 7d).

Discussion
This large population-based study of more than 32 million people investigated the neurological adverse events associated with the ChAdOx1nCoV-19 and BNT162b2 vaccines as well as SARS-CoV-2 infection. We identified several key findings that are timely and of high importance to the public, health policy makers and clinicians across the world. First, we found an increased risk of hospital admission for Guillain-Barré syndrome (15-21 days and 22-28 days), Bell's palsy (15-21 days) and myasthenic disorders (15-21 days) in those who received the ChAdOx1nCoV-19 vaccine. Second, an increased risk of hospital admission for hemorrhagic stroke (1-7 days and 15-21 days) was observed in those who received the BNT162b2 vaccine. Last, we identified a much greater increase in the risk of neurological outcomes following a positive SARS-CoV-2 test, such as acute CNS demyelinating events, encephalitis meningitis and myelitis, Guillain-Barré syndrome, Bell's palsy, myasthenic disorders, hemorrhagic stroke and subarachnoid hemorrhage. Apart from hemorrhagic stroke, the results were robust to sensitivity analyses that assessed the assumption that outcomes did not influence subsequent exposures, providing credence to our results.
In those who received the ChAdOx1nCoV-19 vaccine, we estimated 38 excess cases of Guillain-Barré syndrome per 10 million exposed in the 1-28 days risk period. However, we did not observe an increased risk of Guillain-Barré syndrome in those who received the BNT162b2 vaccine. We replicated these findings using an independent national cohort of more than 3 million people from Scotland, which provides strong support for the association between ChAdOx1nCoV-19 vaccine and Guillain-Barré syndrome. The study design and sensitivity analyses addressed some of the key issues related to this association, including confounding through fixed covariates, seasonality and variable exposures to predisposing infections during the pre-and post-vaccination periods. It remains unclear why the ChAdOx1nCoV-19 vaccine appears to contribute to the pathogenesis of Guillain-Barré syndrome while BNT162b2 vaccine does not. Further studies are needed to assess whether antibodies against adenovirus vector-based ChAdOx1nCoV-19 can cross-react with components of the peripheral nerves. Furthermore, we found that Guillain-Barré syndrome and Bell's palsy co-occur in those who received the ChAdOx1nCoV-19 vaccine. Clinically, this is likely to represent a variant of Guillain-Barré syndrome, in line with the emerging case reports of Guillain-Barré syndrome variants with facial weakness after the ChAdOx1nCoV-19 vaccine 16,17 .
In those who received the BNT162b2 vaccine, we estimated 60 excess cases of hemorrhagic strokes per 10 million exposed in the 28 days after vaccination. We did not observe an increased risk of hemorrhagic stroke in those who received the ChAdOx1nCoV-19 vaccine. However, the magnitude of this association between BNT162b2 and hemorrhagic stroke was reduced in sensitivity analyses that accounted for fatal events. Intriguingly, the increased risk of hemorrhagic stroke associated with BNT162b2 was substantially higher in female participants than in male participants, suggesting that sex may further modify this risk. The mechanisms underlying this disparity in risk between the two vaccines is not clear but several reports have suggested an increased risk of immune thrombocytopenic purpura in individuals receiving mRNA vaccines 18,19 , which in turn could contribute to major bleeding events. It is worth highlighting that the association between the BNT162b2 vaccine and hemorrhagic stroke was not replicated using Scottish data in the context of a low number of events, and thus further validation of this finding is warranted.
In line with previous surveillance studies 14,15 , we identified that all investigated neurological conditions were linked to SARS-CoV-2 infection itself. Specifically, we noted an excess of inflammatory disorders including encephalitis meningitis and myelitis (123 excess cases per 10 million exposed), myasthenic disorders (163 excess cases per 10 million exposed) and Guillain-Barré syndrome (145 excess cases per 10 million exposed) in the 1-28 days period after a positive test. Our post-hoc analysis restricting to those who had a positive test prior to vaccination did not change the results. Unfortunately, the number of people with a SARS-CoV-2-positive test after vaccination was too small to evaluate the risk of neurological complications in this group. The highest IRRs for these neurological outcomes were typically seen on day 0 (the day of the positive test). We included day 0 on its own as a risk period because hospital admission can trigger SARS-CoV-2 testing; however, this could overestimate or underestimate the overall risks associated with infection. The number of people infected in Scotland (3%) was much smaller than in England (~6%), reflecting differences in infection rates and containment measures in the two countries.
Findings from this study have clear clinical and public health implications. Crucially, we found that the risk of neurological complications from infection was substantially higher than the risk of adverse events from vaccinations in our population (for example, 145 excess cases versus 38 excess cases of Guillain-Barré syndrome per 10 million exposed in those who had a positive SARS-CoV-2 test and ChAdOx1nCoV-19 vaccine, respectively). This will need ongoing analysis and monitoring as younger people are vaccinated. Clearly, neurological complications from vaccination and infection reported in this study are rare. However, these rare complications can cause lifelong disability requiring long-term care. Collectively, these results provide timely and valuable information that can help to inform clinical decision making, including facilitating earlier diagnosis, as well as resource allocation for health-care provision. This may be particularly relevant for intensive care unit resource planning, given the potential need for prolonged admission to intensive care units for a proportion of patients with Guillain-Barré syndrome 20 .
This study had several strengths. First, the United Kingdom was an ideal place to carry out this study given that both the ChAdOx1nCoV-19 and BNT162b2 vaccines have been rolled out at speed and scale. Second, this was a population-based study of prospectively recorded medical data and it avoided the recall and selection biases linked to case reports. Third, the large sample size provided sufficient power to investigate these rare neurological outcomes that could not be assessed through clinical trials. Fourth, the SCCS study design (Fig. 2) removes potential confounding from fixed characteristics, and the breakdown of the study period into weekly blocks accounted for temporal confounding. Last, the key finding that the ChAdOx1nCoV-19 vaccine was associated with an increased risk of subsequent Guillain-Barré syndrome was independently replicated using Scottish national data.
There were several limitations to this study. We examined the risks associated only with the first vaccine dose because the data on outcomes following second doses were limited at the time of this study. Furthermore, it was not possible to distinguish between Guillain-Barré syndrome variants including Miller Fisher syndrome in view of the coding used in the Hospital Episode Statistics database. Also, given that we investigated only hospital admissions and mortality, a proportion of patients with milder neurological disease may not have been included and the overall burden of neurological adverse events from vaccination and infection could be underestimated. Some hospital admissions for myasthenic disorders could represent exacerbations of previously well-controlled myasthenic illness given that we excluded only those with a prior hospital contact for these neurological conditions in the preceding 2 years. The Scottish replication cohort was smaller than the study cohort, which meant that estimates were less precise and we did not have sufficient statistical power to replicate findings with smaller effect sizes.
In summary, this population-based study identifies and quantifies several rare neurological adverse events that are specific to the ChAdOx1nCoV-19 and BNT162b2 vaccines, as well as SARS-CoV-2 infection. We believe that these findings are likely to be of relevance to other countries using these vaccines and it would be useful to replicate these results in similarly large datasets internationally. Clinicians should be aware of these rare complications, and the findings from this study will be paramount to policy makers in risk-benefit evaluations and health-care resource allocation. Importantly, the risks of adverse neurological events following SARS-CoV-2 infection are much greater than those associated with vaccinations, highlighting the benefits of ongoing vaccination programs.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41591-021-01556-7.
Study design. The SCCS design was originally developed to examine vaccine safety 14,15 . The method includes only individuals who have had the outcome of interest (cases) and compares the incidence rate (in each case) of the outcome of interest during a time-limited exposed period (for example, immediately after vaccination) with rates during unexposed periods (that is, baseline). A relative incidence greater than 1 indicates an increased risk after vaccination or a positive SARS-CoV-2 test. Given that the analyses are conditional on each case, any fixed characteristics during the study period, such as sex, ethnicity or chronic conditions, are inherently controlled for. Any time-varying factors, such as seasonal variation, need to be adjusted for in the analyses.
A flowchart showing how each dataset used for the SCCS analysis of each outcome was created is presented in Extended Data Fig. 3.

Study period and population.
We examined the associations between ChAdOx1nCoV-19 or BNT162b2 vaccines and neurological complications during the ongoing COVID-19 vaccination program in England (commenced on 8 December 2020). We also investigated the association between a SARS-CoV-2-positive test and the neurological diseases of interest in these vaccinated individuals. Separate analyses were carried out for each neurological outcome of interest. People were considered eligible for inclusion in each study cohort if they had received at least one vaccine dose, were at least 16 years old, and were admitted to hospital or died from the outcome of interest between 1 December 2020 and 31 May 2021 (last data update). Patients were followed up from the start (1 December 2020) to the earliest of either the end of the study period (31 May 2021), the date of second dose or when they died. Only the first outcome event during the study period was included. Patients with a hospital admission for the same outcome in the 2 years prior to the start of the study period were excluded. We excluded patients who received the mRNA-1273 (Moderna) vaccine given that the number of these patients was too small for meaningful analysis.
Outcomes. The neurological outcomes in this study were selected a priori based on a literature search for any reports of emerging neurological complications from SARS-CoV-2 infection and vaccination. These included acute CNS demyelinating events, encephalitis meningitis and myelitis (excluding encephalopathy), Guillain-Barré syndrome (encapsulating all Guillain-Barré syndrome variants including Miller Fisher syndrome), Bell's palsy, myasthenic disorders, hemorrhagic stroke and subarachnoid hemorrhage. Other outcomes, such as peripheral neuropathies, were excluded because these conditions were unlikely to result in hospital admission (particularly within the 28 day risk period in this study). We used the ICD-10 codes to define each outcome, as listed in Supplementary Table 8. The outcomes were identified as the first hospital admission due to the event of interest or as a cause of death recorded on the death certificate with the event.
Exposures. The exposure variables were a first dose of the ChAdOx1nCoV-19 or BNT162b2 vaccines and first infection with SARS-CoV-2, defined as a positive COVID-19 polymerase chain reaction (PCR) test during the study period (either before or after vaccination). For all outcomes we defined the exposure risk intervals as the following pre-specified time periods: 0, 1-7, 8-14, 15-21 and 22-28 days after each exposure date, under the assumption that the adverse events under consideration are unlikely to be related to exposure beyond 28 days after the exposure. We assumed that patients who experienced an outcome before vaccination would probably delay vaccination until symptoms had improved. Therefore we included a pre-risk period in the analyses, defined as the 1-28 days immediately before the vaccination or SARS-CoV-2 test. The baseline period was then defined as extending from 1 December 2020 to 29 days before the exposure date, and from 29 days after the exposure date to 31 May 2021 or the censored date, if earlier. Hospital admissions for the events of interest can trigger COVID-19 testing. Such events may well be caused by SARS-CoV-2 infection, but the reverse causality involved in their detection induces bias. To reduce the bias, which could overestimate or underestimate the effect of infection, we allocated day 0 to a risk period on its own. Seasonality and COVID-19 pandemic period. Some outcomes of interest have seasonal variation: for example, Guillain-Barré syndrome is more common in the winter months 21 . Moreover, hospital admissions were probably influenced by the pressure on the health systems due to COVID-19, which was not uniform during the pandemic period. To allow for these underlying seasonal effects, we split the observational study period into weeks and adjusted for week as a factor variable in the statistical models. In doing so, we estimated the relative incidence in exposure periods up to 28 days after vaccination compared with the unexposed periods, adjusted for underlying seasonality, which accounted for changes in hospital admission rates caused by the pandemic. If an exposure period is long in relation to the observation period, it might be impossible to distinguish between seasonal variation and vaccine effects. However, given that the exposure period was relatively short (28 days), this was unlikely to happen given that after 28 days a vaccinated patient returns to the baseline period, ensuring that in each interval both exposed and unexposed cases appear (Extended Data Fig. 4). Statistical analysis. We described the characteristics of each cohort (vaccinated patients with the outcomes of interest) in terms of age and sex. The SCCS models were fitted using a conditional Poisson regression model with an offset for the length of the risk period. IRRs, the rates of hospital admission or death due to each outcome of interest in risk periods relative to baseline periods, and their 95% CI were estimated using the SCCS model adjusted for week number as time-varying covariates. Exposure terms for both vaccines and for infection with SARS-CoV-2 were included in the same model.
We conducted sensitivity analyses to assess the robustness of results to assumptions such as that the occurrence of an outcome event did not influence the probability of subsequent exposures, by (1) excluding those who died from the outcome and (2) restricting analysis to the post-vaccination period, without censoring at death due to the outcome; to assess potential reporting delays in the data by (3) restricting the study period until 1 May 2021; and to test the choice of the follow-up period by (4) restricting analysis to patients who received only one vaccine dose and (5) censoring at 12 weeks after vaccination or at the date of a second dose if earlier. To test our assumption that events after 28 days are not related to the vaccines, we did an additional analysis with a 35 day risk period after exposure.
To further validate the estimated IRR associated with a positive SARS-CoV-2 test, we also restricted the analyses to only patients who had a positive SARS-CoV-2 test before vaccination.
Stata version 17 was used for these analyses.
Absolute measures of effect. Absolute risk differences cannot be obtained using the SCCS methodology. We supplemented our estimates of IRRs with measures of the effect of each exposure in absolute terms, using a method 22 developed to estimate the number of exposures needed to produce one excess adverse outcome and the excess number of events per 10 million exposed for each outcome.
Interactions with age and sex. We investigated whether the association between vaccine exposures and outcomes is sex or age dependent by including, in the analysis, interaction terms between these covariates and the exposures. We used the likelihood ratio test to assess the interaction terms. We present results in separate subgroups when an interaction was significant.
Negative and positive control. We examined the association of exposures with celiac disease as a negative control outcome 23 , which is assumed not to be associated with exposure to vaccination or SARS-CoV-2 infection; and with anaphylaxis as a positive control outcome given that it could occur shortly after vaccination with either vaccine 24 .
Assessing the SCCS assumptions. To further assess the assumptions of the SCCS and our modeling choices, we visually examined the data. We plotted a histogram of the number of occurrences of an event by time prior to or since vaccination for each outcome to assess the possibility that a hospital admission for that event might affect subsequent vaccination (Extended Data Fig. 5). We plotted a histogram of the time from the event to the actual end of observation in censored and uncensored cases to assess whether event-dependent observation periods may be a problem for the analysis (Extended Data Fig. 6).
Event-dependent exposures. Extended Data Fig. 5 shows the number of occurrences of an event by time before or since vaccination. We notice that there is a decrease in the 28 days immediately before vaccination, indicating that occurrence of an event might have reduced the likelihood of vaccination. This pattern is similar for most of the outcomes and for both vaccines. Therefore, we have added the pre-risk period of 28 days. Fig. 6 shows the frequency of days from the event to the actual end of observation in censored and uncensored cases. A spike close to zero is apparent in the censored data histogram for hemorrhagic stroke and subarachnoid hemorrhage. This indicates the presence of event-dependent observation periods (censoring on death date due to outcome), algorithms (outside this work) into clinical computer systems. J.H.-C. is chair of the NERVTAG risk stratification subgroup and a member of SAGE COVID-19 groups and the NHS group advising on prioritization of use of monoclonal antibodies in COVID-19 infection. All other authors have no competing interests.
Correspondence and requests for materials should be addressed to Julia Hippisley-Cox.
Peer review information Nature Medicine thanks Avindra Nath and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Jerome Staal was the primary editor on this article and managed its editorial process and peer review in collaboration with the rest of the editorial team.
Reprints and permissions information is available at www.nature.com/reprints. which we tested further with sensitivity analyses 1 and 2. These additional analyses were in agreement with the main analysis, suggesting that there should be little concern for these outcomes.

Replication using national Scottish data.
To assess the robustness of the findings, we carried out an independent replication of the risks of neurological outcomes in vaccine recipients using a similar SCCS design. NHS Scotland provides comprehensive health care that is free at the point of care for all residents, with a base population that includes 5.4 million residents (~95% of the population) registered with a general medical practice in Scotland. We used deterministic linkage (based on the Community Health Index) to link vaccination information (extracted from general medical practice records and the Turas Vaccination Management Tool) and hospitalization records (from the Scottish Morbidity Records 01), mortality data (from National Records of Scotland) and SARS-Cov-2 infection data (from Electronic Communication of Surveillance in Scotland). We classified exposures and outcomes in a similar way to the main analysis. An individual was defined as exposed to either the ChAdOx1nCoV-19 or BNT162b2 vaccine on the date that they received their first dose between 8 December 2020 and 31 May 2021. We defined outcomes of interest using the same ICD-10 codes as for the main analysis. We excluded patients who experienced the outcome in the 2 years prior to the study start date and excluded patients under 16 years of age. We fitted conditional Poisson regression models with an offset for the length of the risk period and adjusted for the weekly time period as a time-varying covariate. The control (baseline) period was the period from 1 December 2020 to 29 days before vaccination, allowing for a 28 day clearance (that is, pre-risk) period, and from 29 days after the exposure date to 31 May 2021 or the censored date, if earlier. We also stratified the post-vaccination follow-up period using the same time periods as the analysis in England when the number of events allowed. We also repeated the analysis by replacing vaccination with SARS-CoV-2 infection, which was defined as a positive PCR test within the study period. Given the smaller population size of Scotland, we focused on assessing the consistency of the point estimates of associations observed in the main analyses.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support the findings of this study, that is, the National Immunisation (NIMS) Database of COVID-19, mortality (Office for National Statistics, ONS), hospital admissions (Hospital Episode Statistics, HES) and SARS-CoV-2 infection data (Public Health England, PHE) are not publicly available because they are based on de-identified national clinical records. Due to national and organizational data privacy regulations, individual-level data such as those used for this study cannot be shared openly. Approval to access PHE data needs to be provided by the PHE Office for Data Release. Approval to access NIMS, HES and ONS data requires permission from NHS Digital, who retain the copyright in that data.