What to Do When a Global Pandemic Resurrects Paper-Based PROs Creating Data Collection Headaches for Data Managers
By Hayley Jeffery, Terek Peterson, MBA, and Justin Jaeschke
Electronic clinical outcomes assessments (eCOA) are the fastest growing technology in the eClinical space. It allows subject and site data to be captured in real time, accurately, and with languages included in the protocol. In the rare event of device failure, a workaround is necessary to ensure data is still collected. These backup solutions may require capturing data on paper and manually entering the data into the system. This process requires more effort, is time consuming, and can also increase the risk to data integrity.
With the unexpected arrival of COVID-19, the requirement for paper assessments suddenly increased. Global lockdowns resulted in sites closing and subject visits being cancelled. To ensure continuity of data collection, subjects either completed paper assessments at home or provided responses to sites via telephone interviews, requiring sites to transcribe those responses onto paper.
Paper assessments were no longer the backup solution – they were the primary solution—so using the existing, inefficient process of entering paper data into the eCOA database was not a viable option. An alternative approach was needed, one that was scalable, could support several languages, while maintaining high data integrity. The process of submitting the assessments also needed to be simple and easy, ensuring any additional burden on the site staff was kept to a minimum.
This case study explores the overall process and technical solutions implemented to allow a large volume of critical paper assessments to be entered.
KEY REQUIREMENTS
When devising a solution, several key requirements needed to be considered from both a site/sponsor perspective, as well as from a data entry viewpoint.
Site/Sponsor requirements
Minimize burden to site staff
One of the many benefits of utilizing an eCOA platform in clinical trials is that the devices do all the hard work. The site staff review the questionnaires with the subjects and provide guidance on how to use the tools, but the data is captured and synced automatically to the system with minimal effort from the site staff. When capturing the data on paper, it was important that the data submission process was kept simple and easy, avoiding too much additional work for the site staff.
Maintain high data integrity
When capturing data electronically, it is captured in real time, stored securely on the device, and uploaded to the central server during a manual or automatic sync process. When captured on paper, the data is not automatically uploaded, but manually entered. This manual interaction can increase the risk of error, therefore, the solution needed to use automation, where possible, and implement a verification step to ensure accurate data entry.
No change in how data is stored or transferred
Although the process for capturing and entering the data is different, how the data is stored once in the system and then transferred to external systems should remain the same. The sponsor should not ‘see’ any difference in the overall process.
Data available quickly
As data cannot be entered in real time, it is important to ensure that, once submitted, the data is entered quickly to avoid any issues with subject eligibility, randomization, or delays to subject visits. The solution needed to consider how paper assessments can be entered quickly to avoid downstream issues.
Data Entry Requirements
Scalable across multiple studies and PROs
With a high volume of paper assessments expected across several studies, all utilizing different patient-reported outcomes (PROs), the solution needed to be scalable to ensure efficient data entry. It should also be logical and simple to use.
Eliminate language barriers
With COVID-19 being a global pandemic and countries worldwide being impacted, it was essential that subjects and sites could still complete the paper assessments in the subject’s native language. The solution needed to consider how English-speaking data managers would be able to read and transcribe those answers accurately.
Cost effective
With all new solutions, cost is a factor that needed to be considered. The solution needed to leverage existing resources and tools, where possible, to keep cost to a minimum.
THE SOLUTION
To minimize the burden on the site, paper eCOA screen reports, which mimicked the electronic devices, were provided. These were available in the subject’s native language so questions could be understood, and answers could be documented quickly and accurately.
The existing data correction process was used for the submission of the assessments, again keeping the approach consistent and minimizing any inconvenience to the site.
The solution used native connectors between SQL and dynamic worksheets, allowing live data to be pulled down and used. Like other clinical data systems, eCOA systems often utilize a relationship database model to drive data relationships. This parent/child relationship is driven by Protocol > Questionnaire > Question > Choices and is maintained in a RTSM SQL database. To maintain data integrity, and to provide a logical means for the data managers to process the paper assessments, the solution utilizes the data relationships to ask questions to our data management team, in logical order.
Figure 1 Data entry process flow
This approach, while very simple, allows the data manager to have clear and distinct choices without viewing or seeing data that is unrelated to the paper questionnaire.
Answer data is not usually stored in free text, meaning choice responses are saved in the database as an ID value. When processing paper questionnaires, it is critical that data can be entered using human readable prompts. Manually mapping IDs to choices would be highly inefficient and vastly heighten the risk of error to the process. To mitigate this risk, the solution provides an automatic mapping of human readable choices to ID, so that as data is captured, the relationship between the human readable text and system IDs is maintained.
Another key requirement was allowing subjects to complete native language assessments; however, this posed a unique challenge to the English-speaking data managers. To reduce the risks, the solution allows for the question-and-answer text to appear in both English and the patient-specific language. This allows the data manager to quickly identify the desired response even for symbol-based languages.
The industry’s expectation is that any manual data entry be double-entered or verified, prior to being committed to the dataset. It is never appropriate to write data directly to the database without this verification being completed. As a response, a custom macro was created which loops through the selections and pushes the data back into SQL. This is stored in a staging table.
Linking this to a verification worksheet allows error checks and manual verification to be performed on the data, prior to it being committed to the database. Verification can also happen in both English and the native language, allowing the verifier to see the data side by side with the paper questionnaire with native language text.
CONCLUSION
Thinking back to the key requirements, the solution met the needs of the sponsor, site, and data managers. Due to the simple collection and submission process, the burden on the sites and subjects was kept to a minimum. The validation process ensured that the data integrity was maintained and the process for storing and transferring the data was unaltered. As the solution was logical and used worksheets, data could be entered quickly, resulting in a time saving improvement of 120% compared to the previous paper backup solution. It could also be used by multiple data managers, allowing data entry 24/7.
The implementation of dynamic mapping from study specific SQL databases ensured the solution was scalable across multiple studies and PROs. The option to change the question-and-answer text to the required language, for both data entry and verification, eliminated any translation concerns and improved overall data integrity. Building the tool in house, using existing tools and resources, kept the solution effective.
