Non-interventional studies (NIS), i.e., studies without interventions, where a medication is prescribed by a physician according to the routine practice of the specialist, have been growing in popularity every year since recent times. One of the characteristic features of conducting this kind of research is the wide variety of data sources; because of the focus on studying routine clinical practice, it is necessary to constantly remember that it did not create the data sources for their subsequent analysis in a clinical trial. The goals and approaches to the maintenance of these medical records are beyond the objectives of the research protocol. This specificity requires separate attention when preparing the information collection system. In this note, we will review some methods for ensuring the integrity of data collection within the NIS.
It represents actual practice data by the broadest set of sources, interpreting which in either may be useful for studying the use of medications. In English-language practice, the term Real-World Data (RWD) is widely used to describe the variety of such data. Conventionally, “real-world” data can be divided into two groups: public data and closed data. Examples of public data could be social media data, discovery datasets on adverse events, and de-identified open data on clinical trials. Closed sources are represented, for example, by medical records, electronic databases of laboratory tests, data from continuous condition monitoring devices, etc. As you can see, the sources differ not only in the type of information collected but also in the means of their carriage. Sometimes, it is possible to carry out an automatic transfer of data from one database to another, and sometimes manual input is required.
However, more important is the fact that the goals set in the research protocol operate with integral parameters, sums, averages, changes in dynamics, and so on. The analysis process requires a certain data structure specially prepared for data fixation and interpretation. In contrast to classical clinical research, where the structure of records can be determined by the protocol, NIS cannot interfere with the course of routine clinical practice. That is where preparing the data collection system becomes critical, as it is only at this stage that there is an opportunity to group disparate sources into a coherent database.
Controlling Study Endpoints
To ensure data integrity, it is first necessary to focus on the data collection process by an endpoint. This is required foremost to ensure that the data collection system has at its core a highly understandable set of records for all study participants. Using this approach, it is possible, first, to provide simple control of the filling of the data, on the key parameters, and second, having such a rigid structure, to add new input fields, linking them to the main parameters. To do this, the following questions need to be answered:
- What are the characteristics of the endpoints?
- What statistical methods of analysis will apply to this data?
- How does the study design affect the collection of this data, is there retrospective data, is the data collection done vis-à-visor is the event a change in subject status?
Features of the eCRF design
Since a structured database needs to be organized to meet the objectives of the research protocol, the most convenient and logical way is to use Case Report Form (CRF), both electronic and paper, which would unify adapting real clinical practice records into a unified data collection system. Because of multiple sources, the local nature of medical record keeping, and the multicenter nature of studies, even the data entry process itself can gain a non-uniform nature, which can eventually lead to heterogeneity of the information got. This is especially true for paper-based CRFs since the absence of logical links between the forms, the absence of restrictions on entering data in the fields, and the longer and more complicated process of resolving requests for data clarification increase the risk of uncoordinated data collection. Thus, the most acceptable is the use of an electronic Data Acquisition System (DAS) and, electronic CRF.
The wide variety of tools for data verification at the stage of input in the CRF in its turn should not lead to the situation of the impossibility of data input. When constructing the CRF form in NIS, it is necessary to remember:
- The researcher still does the work of adapting your data.
- It is impossible to foresee all variants of the collected values in advance.
- There is a high probability of missing or partial data, which cannot be restored.
They must keep all these features in mind when designing the database structure, ensure the collection of data on the one hand, and not create a situation of complete inability to enter data.