.svg){kind=logo}
AI Solutions
Additional Services
Stranger Danger: How Well Do You Know Your AI?
Global street smarts through culturalization.
Case Study: Multilingual Retail Marketing
New AI Content Creation Solutions for a Sports and Apparel Giant
- RESOURCES
Reason 1: Readability and Health Literacy Defined
Reason 2: Readability Testing Methodologies and Their Limitations
Reason 3: AI-powered Readability Testing Superior to Readability Formulae
For years, the life sciences industry has applied formulae such as the Flesch-Kincaid Grade Level to assess the readability of layperson information. Readability testing with algorithmic scoring systems has limitations, even though tests are recommended in some regulatory guidance as an add-on or alternative to costly and time-consuming user consultations.
With the increased focus on patient engagement, transparency, and disclosure of research results, better technologies are needed to help improve content readability for people outside the scientific community. Content readability significantly influences an individual’s comprehension and isn’t frequently considered when developing health education, instructions, or information materials. Notably, effective communication between physicians and patients lowers patients’ anxiety, improves compliance, and enhances clinical outcomes.
This blog dives into readability and health literacy as key aspects of effective plain language communication and patient engagement. We present Lionbridge’s innovative AI-powered readability testing solution and explain how it drives outcomes superior to traditional readability testing formulas.
Readability and Health Literacy Defined
Readability refers to the ease with which a reader can understand written text, such as plain language writing. Aspects that impact the readability include length and complexity of words and sentences, as well as document structure. Readability tests are objective measures that typically return a score equivalent to a grade level of readability.
Health literacy is a broader notion beyond the written text and its understanding. It’s widely believed to strongly predict an individual’s health status and significantly influence patient outcomes. Health literacy focuses on more abilities, including reading, comprehension, analyzing information, decoding instructions, symbols, charts, and diagrams. As such, readability emphasizes the interaction between the reader and the text. It also captures how the reader can use the information to make informed decisions. The Good Lay Summary Practice guidance published on EudraLex, Volume 10, lists health literacy principles concerning language aspects. They include recommendations such as using:
- Simple, everyday language
- Short words and sentences
- Active voice (rather than passive)
- Nontechnical or nonscientific terms (or defining them in plain language)
- Neutral, non-promotional language
- Words and terms consistently
- Respectful language
Health literacy and readability are vital aspects of written content intended for patients and users of medicines and medical devices. When clinical trial sponsors obtain informed consent from clinical trial volunteers, volunteers must understand the risks and benefits of participating in a clinical investigation. Otherwise, they won’t be able to make an informed decision about attending. Likewise, medical devices’ instructions must enable users to understand and take appropriate action for safe and effective usage.
Readability Testing Methodologies and Their Limitations
Readability testing is recommended by the Clinical Trials Expert Group (CTEG) under the Clinical Trial Regulation for layperson results summaries. Where feasible, the CTEG also recommends testing with a small number of people who represent the intended readers. This type of readability testing, which is measured by quantitative scores correlating with grade-level equivalents, has limitations. A pure readability score lacks interaction between the text and the reader. Additionally, it doesn’t consider key stylistic aspects of language or specific terminologies for use cases and intended readers. This lack is a challenge for text translated from a source text written in medical, technical, or scientific language into plain language intended for a lay audience. See our white paper for more information on the stylistic differences between scientific and plain language.
At best, traditional readability formula can be a starting point for gauging health literacy. However, these formulas shouldn’t be applied as a standalone test. Some formulas assess only word and sentence length, which is no guarantee for a meaningful or readable text. Lionbridge recommends sponsors look for more advanced testing. Ideally, a test of a patient or lay users’ text accessibility should consider both readability scores and health literacy principles.
AI-powered Readability Testing Superior to Readability Formulae
Generative AI is known for its ability to generate and analyze content based on prompt engineering. Sophisticated, expert-led prompts and Large Language Models (LLMs) can replace readability formulae and offer more advanced and actionable solutions. Built-in principles of health literacy, language, style, and terminologies tailored for the use case and the reader will drive better readability outcomes and engagement with the intended users. Using LLMs can also cut time and burden from technical and medical writers during critical regulatory procedures. Notably, LLMs can become valuable co-authors for writers who find it unnatural or difficult to author content in conversational language (because of their training in scientific communication).
Lionbridge’s innovative readability solution offers advanced prompt templates that train our LLMs to segment, analyze, and improve written text. AI and life sciences translation services experts who understand the technology, clinical research methodology, and compliance aspects develop these templates. Within minutes, an instant automated readability report will be generated to provide a medical or technical writer with actionable recommendations during document development and review. Since the prompt contains specific language, style, and terminology instructions, it cuts hours (or days) from writing and improving plain language content. It can also be customized to cater to different therapeutic areas or glossaries.
Readability Validation
As with any other Machine Learning (ML) Model, the LLM must be governed to avoid hallucinations and ensure reliable plain language communication outcomes. Therefore, a medical or technical writer must review LLM output and determine how and when to implement the LLM-suggested text revisions. Such human-in-the-loop (HITL) test methodology anchors responsible use of AI in GxP and other regulatory-controlled settings:
- AI suggests improvements for process optimization and improved readability
- A human confirms and implements, ensuring contextual relevance and accurate outcomes.
- The LLM readability report and the final revised document will then create an auditable record of the human-machine interaction.
It’s critical to stress that technology-enabled readability testing cannot replace user consultation in all cases from a contextual, regulatory, and cultural perspective. However, they have many valuable applications for improving content readability and driving better patient engagement.
| Content Types | Use Cases |
|---|---|
| Plain Language Summaries (PLSs) of clinical research results | Provide clinical trial participants with aggregate results of the clinical study in plain language |
| Summaries of Safety and Clinical Performance (SSCPs) | Public access to information on safety and performance of medical devices in plain language |
| Summaries of Safety and Performance (SSPs) | Public access to information on safety and performance of In Vitro Diagnostic medical devices in plain language |
| Informed Consent Forms (ICFs) and Patient Information Sheets (PISs) | Help clinical trial volunteers ensure they understand the risks and benefits of trial participation and thus make an informed decision to participate in the trial |
| Instructions for Use (IFUs) intended for patients or product users | Instructions on product use to ensure safe and efficient use according to the intended purpose and the intended product users |
| Patient schedules or diaries | Help clinical trial participants understand study procedures or record information during trial participation |
| Patient newsletters | Inform trial participants during clinical trial execution for education, information, compliance, retention, etc. |
| Advertising content for patients or patient recruitment | Materials, brochures, or other written content for attracting potential trial participants in relevant studies |
| Information for patient associations, patient advocates, patient experts, etc. | Patient engagement initiatives to ensure patient input into R&D programs, disease insights, etc. |
| Website information to patients, trial participants, and the public | Web content for various purposes, e.g., information on disease, clinical trials, products, etc. |
| Information intended for nonscientific business professionals or suppliers | Inform investors, insurers, suppliers, and other stakeholders outside the scientific community |
Get in touch.
Need help with your readability assessments for plain language summaries or any other content intended for laypersons? Need assistance with your clinical trial translation? Have plain language communication or other life sciences translation needs? Let’s get in touch to discuss.
