Skip to content
English
Flinn.ai
  • There are no suggestions because the search field is empty.

📃 AI Clinical Report Writing

Find here the resources needed to master the AI report feature from Flinn and automate parts of your clinical writing activities.

💡 The AI Report feature helps you draft SOTA, CER, PMCF, and PSUR sections more efficiently, directly within your Flinn environment. It leverages the data extracted from your included papers to accelerate writing draft sections within seconds. To fully benefit and save hours of work, you’ll need to spend a short time learning how to make the most of this AI writing assistant. Read more to discover how.

 

The most important part of all AI feature is the input: Your extraction table.

1st step: Optimizing table configuration

  1. Be highly selective with columns: Only include columns directly relevant to your current section. Our testing shows that even comprehensive sections typically only require 3-4 columns. This significantly improves AI performance.
  2. Table size considerations: While we're increasing your publication limit to 150, please note that very large tables with numerous text-heavy columns may reduce effectiveness. Consider focusing on specific subsets of data for each section.

With an optimized table, you can now start using the AI to draft your paragraphs.

2nd step: Efficient AI workflow through iteration

  1. Be specific in your prompts: Clearly state exactly what you need. For example, instead of asking for "information about elderly patients" specify "summarize benefits for patients 65 or older"
  2. Targeted refinements: For small changes, copy-paste just the specific sentence or paragraph you want modified, rather than regenerating entire sections.
  3. Reference verification: All AI-generated content now includes reference numbers, making it easier to verify sources and accuracy.
  4. Providing an old report paragraph as an example in your prompt will help give context and writing guideline to the AI. You can then ask the AI to rewrite it with the latest data in the table.

  5. Multiple iterations: Expect to refine prompts and outputs several times to achieve desired results, especially during initial use. These interactions are now saved for future reference.

3rd: Know the strength and limitations

  1. The writing agent can identify ambiguities in your request and ask follow-up questions to maximize output quality.
  2. It can either draft multi-paragraph sections or rewrite specific sentences. You can tell it which you need.
  3. It remembers your previous interactions and uses them to inform future responses.
  4. It doesn’t currently have access to the full text of your publications, though this could be enabled in the future if needed.
  5. The writing agent is not optimized to build tables or edit existing tables. If you feel that this is needed throughout your testing, share this feedback with us.

Now you are ready to automate your next CER writing! Let's look at some prompts examples to inspire you.

Prompts examples

Example of a prompt to draft the SOTA Benchmark of technology section:

Using the data provided, write one cohesive paragraph for the Technology Benchmark section. Group the studies by device type or technology and summarize the main clinical or technical benefits each provides. Compare these technologies to highlight their relative advantages and limitations, and explain how could endoscopic treatments be superior or more advanced approach in terms of safety, effectiveness, or usability. Write in a clear, formal, regulatory tone without bullet points.

Example of a prompt to draft the Risk & Benefit analysis CER section:

Summarize the key benefits identified for adults patients in this dataset. Define ‘adult’ as patients aged 25 and older. Use information from the 'Benefits' column, as well as relevant findings from 'Outcomes' that indicate benefits if this column exists. If patient age is missing or unclear, exclude those rows from the summary. Present the results with clear paragraphs 1 as an introduction and 1 to list key findings.
Summarize the main complications identified by those studies grouped by the different device types. Specify for all complications found the main metrics such as bleeding rate or operation time mentioned. If complication is missing or unclear, exclude those rows from the summary. Present the results with clear paragraphs 1 as an introduction and 1 to list key findings.

Example of a prompt to update your previous SOTA Vulnerable population section:

Update the following Vulnerable Population paragraph using the new data provided. Keep the same structure and tone, but revise the content to reflect the most recent data found.
The reviewed literature did not identify any specific safety or performance concerns regarding the use of the subject device in vulnerable populations, including pediatric, geriatric, or pregnant patients. Most studies involved adult populations within the intended user group, and no differential outcomes were reported based on age, sex, or comorbidity status. Although limited data exist for pediatric and elderly subgroups, the available evidence indicates that the device functions as intended across patient types when used in accordance with the instructions for use. No contraindications or additional risks were identified for these populations beyond the general risk profile described for all users.