Using SDTM Data to Produce Patient Narratives

So, what is a patient narrative? It’s a summary describing serious adverse events, death, or other significant events experienced by a patient during a clinical trial. A patient narrative should provide a clinically relevant, chronological account of the events experienced during or immediately following a clinical study.

According to ICH guidelines, patient narratives must be included with the clinical study report (CSR). The primary consumers of these documents are Health Authorities, Safety Teams, and Data Safety Monitoring Boards.

Each narrative tells an individual patient story with data unique to that patient, and consistency is critical. However, authoring dozens – if not hundreds – of consistent patient narratives containing dozens of patient-specific data points takes a tremendous amount of time, which can be a challenge when teams are operating on tight timelines.

Picture this: A team of 4 medical writers needs to produce 2000 patient narratives across multiple trials in a year. At 4 hours per document, that’s 8000 hours, and that breaks down to all four writers working full-time year-round.

What’s the risk of producing a high volume of patient narratives when deadlines are tight? It may result in a reduction in quality.

We’re only human, and that means mistakes are inevitable, regardless of the root cause. And that means the solution to reducing instances of human error is to automate through technology.

Conventional medical writing typically starts with a tool like Microsoft Word by building a template that outlines the desired presentation of the patient narrative. But before a medical writing team can begin authoring a patient narrative, BioStats typically will need to create patient profiles first.

Traditionally, the data from patient profiles are then copied and pasted into a second document – the patient narrative – using the template created previously with Microsoft Word. This process entails a constant back and forth between BioStats and Medical Writing teams, generates additional work, creates opportunities to introduce errors, and adds to overall project timelines.

Another key element when producing hundreds of patient narratives across the study is consistency. Whether it’s the outline of the document, message development, presentation of information (i.e. dates), or formatting for all information throughout the patient narrative, consistency is essential. While human judgment is needed for key decision-making like determining medical relevance, consistency remains elusive in manual input.

A major component of creating a consistent, compliant patient narrative is the ability to apply standard formatting to text, tables, and data variables. This is critical when working with SDTM data as formatting can vary dramatically from study to domain to variable.

Prime examples of data variability include:

• fields in all caps
• fields that are abbreviated
• dates with various formats

The problem is that most medical writing teams don’t get to see any of the data until after database lock. That means it must be corrected manually, one instance at a time, in the Microsoft Word document.

Early access to draft datasets also aids medical writing teams in Quality Control measures.

When medical writers can work directly with the SDTM data, they have a better understanding of how the patient narrative should look, as well as an opportunity to “spot check” a live preview of the sample output. In the event a data domain or variable is missing, the medical writer can flag the data point, alter the presentation, or alert the BioStats team to the issue in question.

This live preview helps quickly identify if there are anomalies or abnormalities in the SDTM data. The best-case scenario is that a piece of data is simply poorly formatted, which can be addressed and corrected.

The worst case is missing or incomplete data. In this case, medical writers will have acted as an additional layer of data QC, which is especially helpful in the instance of working with pre-database lock SDTM datasets.

This live preview of patient-specific values in the data allows the team to pinpoint the issue so that the team can trace back to the SDTM and source datasets to help identify the root cause.

Organizations have begun to realize enormous gains due to the overall cost savings despite a dramatic uptick in productivity simply by adopting simple tech tools designed to assist in medical writing automation. These positive results can be attributed to five distinct reasons:

1. Automation allows medical writers to produce hundreds of patient narratives with consistent presentation and layout, regardless of the variability in the underlying SDTM data. This helps improve overall quality and helps to ensure they are submission ready.
2. Using SDTM data saves teams on average 63% more time versus traditional manual writing. Automating most steps for both writing and QC can dramatically reduce effort while improving turnaround times.
3. It allows medical writers to work directly with data. This enables full traceability of content and opens the door for medical writing teams to work ahead before the trial has concluded.
4. Using validated SDTM files from BioStats further ensures data accuracy and quality of narratives produced.
5. With the right automation tools and procedures, the need for patient profiles can be eliminated completely.

Let’s go back to our team of 4 medical writers producing 2000 patient narratives from earlier and apply the time savings reported above. 8000 hours is now down to 2960.

法规文件撰写团队面临压力，需要在资源短缺的情况下缩短递交时间线以满足这些需求。CoAuthor™ generative AI software provides transparency, consistency, and collaboration to allow regulatory writing teams to accelerate the drafting and submission of quality and compliant regulatory documents, reducing timelines. Learn more about CoAuthor™ in our recent webinar below.