Creating Accurate Plain Language Summaries With Artificial Intelligence

Plain language summaries (PLS) make research more understandable to non-experts [1], increase engagement in healthcare [1], and can be written at a reading level that accommodates a broader audience. It is important to considering the audience and use of content when determining the appropriate target reading level. Lower reading level broadens the audience reach but may reduce precision. An increased reading level allows for important nuances but may require too many words.

Recommended Reading Levels for Various Audiences

• 6th to 8th Grade

  This reading level is preferred for patients and caregivers. It is best for patient handouts and administrative guides.

• 9th to 12th Grade

  This is an appropriate reading level for scientifically engaged patients, caregivers and policymakers. It is best for clinical trial summaries and plain language summaries (PLS).

• 13th to 16th Grade

  Clinicians and other healthcare experts are a good audience for this reading level. It is best used for review articles and trade journals.

Published PLS are Not Making the Grade [2]

In a survey of 153 published scientific abstracts and associated plain language summaries, the average reading level of the scientific abstracts was fifteenth-grade while the average reading level of the plain language summaries was actually higher at sixteenth-grade.

ChatGPT Reduces the Reading Level While Maintaining Key Points [2]

However, expert editors were able to ehance the plain language summaries with artificial intelligence tools. Using appropriate prompts, ChatGPT 3.5 reduced the reading level while maintaining the key points [2]. The average reading level was reduced to below the tenth-grade after only two prompts and the average reading level of 90% of PLS-generated abstracts was reduced by two grade levels compared with the published PLS.

In a random selection of 25 ChatGPT-generated PLS:

• 88% included almost all of the key points from the abstract.

• 84% accurately described the key points.

• 84% were sufficient to understand the abstract.

Five Ways to Use ChatGPT to Improve PLS for Different Reading Levels

1. Test prompts that describe specific health literacy principles such as simple language, active voice, minimal jargon [3].

2. Include the reading level of the target audience in the prompt [3].

3. Submit multiple prompts; different prompts have an impact on readability and accuracy [4].

4. Open a separate “chat” for each prompt to minimize the risk of interference from previous instructions [3].

5. Consider the prompt, “rewrite the text for people with low literacy,” to further lower the reading level [4].

Study Summary Download

References

1. Edgell C, Rosenberg A, Putting plain language summaries into perspective. Curr Med Res Opin. 2022; 38(6):871-874.

2. Dang E, Watson J, Bublitz S, et al. Improved readability of plain language summaries using ChatGPT. Poster presented at the 20th Annual Meeting of ISMPP, April 29-May 1, 2024, Washington, DC.

3. Ayre J, Mac O, McCaffery K, et al. New frontiers in health literacy: using ChatGPT to simplify health information for people in the community. J General Intern Med. 2024;39:573-577.

4. Haver HL, Gupta AK, Ambinder EB, et al. Evaluating the use of ChatGPT to accurately simplify patient-centered information about breast cancer prevention and screening. Radiology: Imaging Cancer. 2024;6(2):e23008