Good TREs work

Evidera Ltd projects

2 data files in total were disseminated unsafely (information about files used safely is missing for TRE/"system access" projects).


Health Burden of COVID-19 and Healthcare Resource Utilisation in England — DARS-NIC-561357-X0F3N

Opt outs honoured: Anonymised - ICO Code Compliant (Does not include the flow of confidential data)

Legal basis: Health and Social Care Act 2012 - s261(5)(d)

Purposes: Yes (Research)

Sensitive: Sensitive, and Non-Sensitive

When:DSA runs 2022-11-18 — 2023-11-17

Access method: System Access, One-Off
(System access exclusively means data was not disseminated, but was accessed under supervision on NHS Digital's systems)

Data-controller type: ASTRAZENECA UK LIMITED, HEALTH & SOCIAL CARE INFORMATION CENTRE

Sublicensing allowed: No

Datasets:

  1. Civil Registration - Deaths
  2. COVID-19 Second Generation Surveillance System
  3. COVID-19 Vaccination Status
  4. GPES Data for Pandemic Planning and Research (COVID-19)
  5. Hospital Episode Statistics Accident and Emergency
  6. Hospital Episode Statistics Admitted Patient Care
  7. Hospital Episode Statistics Critical Care
  8. Hospital Episode Statistics Outpatients
  9. Medicines dispensed in Primary Care (NHSBSA data)
  10. Uncurated Low Latency Hospital Data Sets - Admitted Patient Care
  11. Uncurated Low Latency Hospital Data Sets - Critical Care
  12. Uncurated Low Latency Hospital Data Sets - Emergency Care
  13. Uncurated Low Latency Hospital Data Sets - Outpatient
  14. Civil Registrations of Death
  15. COVID-19 General Practice Extraction Service (GPES) Data for Pandemic Planning and Research (GDPPR)
  16. COVID-19 Second Generation Surveillance System (SGSS)
  17. Hospital Episode Statistics Accident and Emergency (HES A and E)
  18. Hospital Episode Statistics Admitted Patient Care (HES APC)
  19. Hospital Episode Statistics Critical Care (HES Critical Care)
  20. Hospital Episode Statistics Outpatients (HES OP)

Expected Benefits:

Benefits from this study are expected to include the following
1. Benefits for regulators (e.g., MHRA): the findings from this study are hoped will support MHRA’s review of EVUSHELD for the use among patients who are immunocompromised and other vulnerable populations to supplement the trial evidence, based on which the Conditional Marketing Authorisation in PrEP (pre-exposure prophylaxis) was granted. Furthermore, the results of this study will provide a baseline against which to benchmark the overall impact of the use of EVUSHELD, until data accrual and maturity allow for a contemporaneous comparative effectiveness and safety assessment, following the administration of sufficient doses. Additionally, should the respective clinical trials confirm EVUSHELD’s safety and efficacy in outpatient and inpatient treatment indications, the results of this study will serve as foundation for the development of the submission dossier to ascertain the most accurate characterisation of the source population in England.

2. Its is anticipated that there will be benefits for Health Technology Assessment and endorsement bodies such as NICE and the Scottish Medicines Consortium (SMC): During the early scientific advice procedure in which AstraZeneca engaged, NICE requested more accurate information on the expected number of patients that would be eligible for EVUSHELD use. NICE also recommended that AstraZeneca conduct an observational study to continue identifying which populations do not respond to vaccinations, beyond patients who are immunocompromised, and expressed concern for the dynamic landscape due to the emergence of new variants, which should also be monitored. Lastly, NICE requested that the health-economic model be populated with efficacy data from the phase III trial PROVENT (NCT04625725, please see - https://clinicaltrials.gov/ct2/show/NCT04625725) but the baseline characteristics be adjusted to more accurately reflect the population in England as well as account for the substantial heterogeneity likely to exist in the target population (e.g., in terms of comorbidities, resource use, risk of severe COVID-19). These are the exact research questions that guided the design of the current study. The results from objectives 2 and 3 will be used to adjust the population characteristics for the cost-effectiveness model. Additionally, the patterns of HCRU and costs associated with an episode of COVID-19 will be an important input for the cost-effectiveness model. The accurate count of patients at risk from objective 1 will also be used in the budget impact model. The exploratory objectives to identify and quantify risk profiles with high unmet clinical need will be the basis for sensitivity analyses in both health-economic models. Furthermore, as the pandemic becomes endemic, regulators and policy makers will also benefit from country-specific estimates on the burden of long-COVID-19 to patients and to the healthcare system to be factored in policy decisions.

3. It is anticipated that there will be benefits for the UK Department of Health and Social Care, clinicians, and healthcare providers: The wide-spread vaccination of the British population has significantly improved the epidemiological situation, thus reducing its pressure on the limited healthcare resources and allowing to plan and act proactively (as opposed to reactively, like during the worst moments of the pandemic). To appropriately plan for the management and administration of resources all agents in the continuum of the healthcare provision will benefit from understanding the nature and magnitude of the outstanding unmet needs in the prevention and treatment of COVID-19. This study will provide these insights. Furthermore, the supply of EVUSHELD to address some of these needs is limited and it is expected that doses will become available in consecutive batches, which means that priorities will need to be established based on formal criteria. This study will also provide the evidence to inform those decisions.

4. It is anticipated that there will be benefits for payers (i.e., NHS and taxpayers in the UK): A thorough and evidence-based understanding of the health and economic burden of COVID-19 across different vulnerable populations enables effective planning and efficient resource allocation. Thus, the benefits described for NICE, healthcare providers, and patients may ultimately translate in cost savings or even costs being averted.

5. It is anticipated that there will be benefits for patients and the general public: Patients and the general public benefit from improved healthcare provision. By making the study results available in the public domain (see Section 5c), the general public can benefit from more accurate assessment of their risk of contracting COVID-19 and developing long-COVID-19 and make informed health decisions (e.g., taking up vaccine boosters or seek advice on eligibility for prophylaxis or treatment).

Outputs:

The initial results for this study are expected within a year following the access to the NHS Digital-linked datasets.

Evidera will be conducting all of the data processing and the analysis. They will send aggregated results that will be in the format of excel tables to AZ for review. The tables that they will send will include patient attrition cells (number of patients excluded during the patient selection process), baseline descriptive results of the study populations (i.e., number and percentage of patient demographics and clinical characteristics identified at baseline), the number and percentage of patients identified as ineligible for COVID-19 vaccine (different row will be provided for each ineligibility criteria), the number and percentage of patients at risk of COVID-19 infection (different row will be provided for each risk factor), the number of new COVID-19 infections and incidence of COVID-19 overall and in each time period of interest, the number of new long COVID-19 infections and incidence of long COVID-19, the rate of resource utilisation per patient and per-patient per COVID-19 episode. Only aggregated data with secondary suppression of cells will be send to AZ. At no point will the patient level data be transferred from Evidera to AZ. The results will also be presented in a study report and sent to AZ for review.

The planned study outputs include a study report, manuscripts in submission to peer-reviewed journals and presentations at scientific conferences. Only aggregated data with secondary suppression of cells will be presented in the planned study outputs. It is anticipated that high impact respiratory/infectious disease conferences/journals will be targeted. These include BMJ, New England Journal of Medicine, Lancet Infectious Disease and BMC Infectious Disease. Where possible, results will be published via the open access route to ensure that all clinicians, policy makers and members of the public can access the results freely. It is also anticipated that the results will be disseminated via presentations at key conferences (e.g., European Congress of Clinical Microbiology and Infectious Disease (ECCMID), International Society for Pharmacoeconomics and Outcomes Research (ISPOR)), webinars to Physicians using key AZ Medical Science staff to communicate results.

In addition, active engagement with charity organisations including the research communities (King’s College London, COVID Symptom Study Team) for topics like Long COVID is planned. AZ intend to work with Long-COVID clinics in the country to identify suitable interested patient groups to disseminate results in the form of presentation, newsletters or sharing of publication summaries. AZ will also be engaging immunocompromised patients, who are considered to not been adequately protected by COVID-19 vaccines due to their body unable to generate an optimal level of antibodies, to provide feedback on the current and future studies. Patients will provide input during study analysis planning, first results readout and publications.

PPIE
AZ will also be engaging immunocompromised patients, who are considered to not been adequately protected by COVID-19 vaccines due to their body unable to generate an optimal level of antibodies, to provide feedback on the current and future studies. Patients will provide input during study analysis planning, first results readout and publications.

Based on an assumed data delivery date in Nov 2022, Evidera estimates to complete data analysis in Jan 2023 and study report in August/September 2023. Once study report has been drafted, manuscripts will be prepared for submission to peer-reviewed journals and abstracts for conference presentation in Q3/Q4 2023.

These planned outputs will be designed with the intention of including sufficient information on methods to ensure research transparency and reproducibility. All types of results, including those sometimes seemed as unfavourable, will be published along with key code list used for case definition. The code lists refer to the SNOMED (https://termbrowser.nhs.uk/) and ICD-10 (https://icd.who.int/browse10/2019/en#/ ) diagnosis codes for identifying patients with COVID or long-COVID. The case definition refers to how cases (i.e., COVID and long COVID-19) are defined and identified from the requested datasets. The proposed study is descriptive in nature, e.g., estimating the size of populations that are not protected by COVID-19 vaccines. The incidence of COVID-19 and long COVID-19, and COVID-19-related healthcare service use, as opposed to estimating the effectiveness of certain treatments/vaccines. The aim is to understand the current health and economic burden of COVID-19. The plan is to publish results regardless of whether the estimates are high or low. With that said, given the data published on COVID dashboard (https://coronavirus.data.gov.uk/), it is unlikely to be low/unfavourable. COVID-19 research results need to and will be interpreted in the wider context, e.g., social restrictions and movement, infection rate, vaccination/booster roll-out and availability of an-viral treatments.

This study is also exploring whether machine learning based methods can help with identifying risk profiles of vaccinated patients who experienced a composite outcome of COVID-19 hospitalisation or COVID-19-related death. As previously mentioned, machine learning methods will first identify all patients that have a COVID-19 hospitalisation or COVID-19 related death after 14 days of a COVID-19 vaccination (i.e., break through infections). Then clustering methods and supervised learning with nested cross-validation will be used to classify these patients into k clusters with similar characteristics. These results, details on the development of algorithms and lessons learned are also planned to be disseminated.