[ad_1]
Discussion
The concept of mirror study has proven to be an effective method for validation of a novel technology to support data collection, in a relevant context of use: different EHRs, investigation sites, sponsors and studies.
The primary objective of the study was successfully met, with over 15% (16.9%) of the data points entered in the e-CRF correctly processed from EHR source records.
The four domains DM, VS, LB and CM selected by the consortium represent 46.4% of the data collected through the five trials in scope, this results validates the consortium choice.
A per study analysis demonstrates the major contribution of the local LB data followed to a lesser degree by the VS data to achieve an acceptable proportion of transferable data. This suggests that studies in oncology (ex: TED14856 and the AZ D19BC), with high volume of local LB data are best candidates for the early use of this digital data collection technology in the near future.24
The two domains LB and VS covers around 40% of the data in scope and represent more than 96% of accurately transferred data. This reflects the availability and good quality of these data at the hospitals EHRs.
The interoperability challenge has been successfully addressed through the implementation within the EHR2EDC module of a core list of data elements and its associated library of terminology mappings. The EHR2EDC module has been efficiently deployed in the four hospitals and the different users trained. The mapping and its implementation were designed to be reusable across studies, with limited (re)verification activities, to provide operational efficiencies, both for the sponsor and for site staff.
The limitations on the results for data in scope highlight a combination of factors affecting the ability to achieve higher performance. Among those factors, we have identified several root causes with possible remediations:
Regulations
For DM data (DM domain), legal limitations in collecting ethnicity in Europe produces an artefact as this information is collected during trials. When analysing only legally acceptable DM data, the result was 100%. This suggests that calculation methods and possible automatic quality controls must consider local regulations to be accurate.
Case report form design
The primary cause of missing data for the VS and LB domains arises for specific data points collected in study eCRFs to document the execution of the procedure. Most of the empty fields expect a ‘yes’ value for the question ‘Has the test been performed?’/‘Was the blood sample taken?’. This could be resolved by using auto populated fields (updated to a ‘yes’ value if results are present).
Local investigator’s team practices
Unlike IRST, other hospitals did not routinely train their staff to fill-in structured forms of the EHRs, and so the proportion of data accurately transferred was adversely affected by the proportion of data collected in EHR as free text or in paper source documents when running a clinical trial.
Special attention should be focused on staff using EHRs to collect patient data associated with a clinical study for preventing free text data entry or paper source. This includes training hospital staff in data quality standards, upgrading quality assurance measures and strengthening data governance activities, to enable EHR data to be trustworthy reused in research.
In the TransFAIR study, the low percentage of CM data correctly transferred reflects that they are more often recorded as free text, for example, in unstructured documents (eg, doctor’s letters) and a large part is prescribed outside of the investigational site and is consequently not captured in the EHR.
Clinical site maturity/readiness
Other factors influencing the level of performance include the site maturity in using their EHRs for clinical trials activities. Site organisational capabilities, best practices (EHR data quality assurance, use of EHRs as eSource in clinical trials, just-in-time data flow), skilled staff (data integration, data management) are essential to benefit from this new method of digital data collection.
Guided work effort is needed to augment the proportion of data recorded as eSource in EHRs to be collected using EHR2EDC solutions. Initial focus would expand transferability of structured data in EHRs, and work at rendering unstructured data to be collected. We envision this effort to be made possible through the development of consensus on ‘high-value data sets’, representing the data most commonly collected in clinical trials.
Nevertheless, not all data collected in clinical trials has its correspondence in patients’ EHRs sources. For example, specific forms in eCRFs collect data in relation with the management and evaluation of investigational medicinal products (tracking, patient’s compliance, pharmacokinetic data, etc).
[ad_2]
Source link
