Authoring and generation of individualized patient education materials.

Chrysanne DiMarco, Peter Bray, H. Dominic Covvey, D D Cowan, V Diciccio, Eduard Hovy, Joan E. Lipa, Carl YangPubMed 2006

TL;DR

This paper tackles the challenge of creating personalized patient education materials, which are often generic and ineffective due to the sheer complexity of medical procedures and individual patient needs. The authors propose a novel "generation-by-selection-and-repair" approach using Natural Language Generation (NLG) to automatically tailor educational content, moving beyond static brochures to dynamic, patient-specific documents.

What Problem Does This Paper Address?

Providing effective patient education is crucial, especially before surgery. However, current methods, primarily static brochures, fall short. The problem is twofold: the vast number of possible surgical variations and patient-specific concerns (like literacy, medical history, and personality) make it impossible to create a comprehensive set of pre-written brochures. This leads to either overwhelming patients with irrelevant information or failing to address their specific needs, ultimately impacting compliance and outcomes. The paper aims to bridge this gap by enabling the automated generation of truly individualized educational materials.

Methods

The core of the proposed solution is a Natural Language Generation (NLG) system based on the "generation-by-selection-and-repair" paradigm. This approach utilizes a "master document" containing all potential text segments for a given topic. The system then selects relevant pieces based on patient-specific criteria and automatically "repairs" them for coherence, style, and form.

Key methodological components include:

  • Corpus of Textual Variants: A structured collection of text segments organized by surgical procedure components (e.g., preoperative workup, postoperative course) and patient modifiers (e.g., smoking status, diabetes).
  • Authoring Tool: A new approach to guide healthcare providers (specifically surgeons) in creating these textual variants. Unlike previous tools requiring deep linguistic expertise, this tool aims to allow surgeons to input text in ordinary English, guided by a cognitive model that considers various stakeholder concerns. It emphasizes a Constructivist approach to learning, helping authors organize knowledge and build the underlying discourse structure of the master document.
  • NLG Tailoring Engine: The software kernel that implements the generation-by-selection-and-repair process. It uses XML for master documents and XSL transformations to produce customized output, supporting both web and paper formats.

Datasets

Not explicitly discussed in the paper. The authors mention creating a "corpus of textual variants" and a "master document" as input for their NLG system, but do not detail specific datasets used for training or evaluation.

Key Insights and Ideas

  • The "Combinatorial Explosion" of Patient Needs: The paper vividly illustrates why static brochures are insufficient. The sheer number of combinations of surgical procedures, patient characteristics, and concerns creates an unmanageable number of potential documents.
  • Beyond Simple Customization: True tailoring goes beyond just creating a few versions. It requires a sophisticated system that can dynamically assemble and refine content based on a multitude of factors.
  • Empowering Domain Experts: The development of an authoring tool that allows surgeons to contribute in plain English is a significant step towards making such systems practical. This shifts the burden from computational linguists to the actual subject matter experts.
  • Constructivism in Authoring: Applying a Constructivist learning model to the authoring process helps structure the knowledge acquisition and ensures that the generated content addresses patient concerns effectively.
  • "Generation-by-Selection-and-Repair" as a Viable Paradigm: The paper champions this approach as a robust method for generating complex, polished texts from pre-existing components, with a focus on automated text repair for coherence and style.

Key Results

The paper focuses on the methodology and system design rather than presenting quantitative results from a deployed system. However, it highlights the potential benefits of their approach:

  • Enhanced Patient Education: Tailored materials are expected to be significantly more effective than generic ones, leading to better patient understanding and compliance.
  • Improved Patient Outcomes: This enhanced education is predicted to result in decreased perioperative anxiety, fewer and less serious complications, faster recovery, and better recognition of postoperative issues.
  • Increased Efficiency: Potentially less time required in perioperative discussions due to more effective communication.
  • Demonstrated Feasibility: The earlier HealthDoc Project demonstrated that complex, stylistically polished texts could be crafted from pre-existing texts using their paradigm.

Limitations and Future Work

  • Authoring Tool Complexity: While the new authoring tool aims to simplify input for surgeons, the paper acknowledges that the underlying linguistic representation and knowledge modeling are still complex.
  • Automated Repair Challenges: The paper mentions the ongoing development of repair algorithms to recognize and revise various infelicities in ill-formed texts, implying that this is an area requiring continued research.
  • Future Work: The authors envision extending the content to other procedures and surgical subspecialties. They also suggest the potential for application beyond surgical interventions to any medical treatment requiring cohesive educational content, such as in medical and radiation oncology. The long-term goal is to further develop and test the theory of automated text repair.

Summary updated

Abstract

Although the pre-surgical patient-surgeon encounter is the opportunity to educate the patient, it is essential that the patient be given educational materials to complement the face-to-face exchange. This is virtu-ally impossible to do well with brochures, because many combinations of procedures are possible, dif-ferent patients have different concerns, and patients have varying levels of literacy and knowledge. In the extreme, a patient would either be given a set of bro-chures selected from 100s of variants, or all patients would be given the same set of brochures without regard for differing needs. We have been developing an information brochure generator that customizes material for every individual patient regardless of the complexity of the surgical intervention.
BibTeX
@inproceedings{dimarco2006authoring,
  author = {Chrysanne DiMarco and Peter Bray and H. Dominic Covvey and D D Cowan and V Diciccio and Eduard Hovy and Joan E. Lipa and Carl Yang},
  title = {{Authoring and generation of individualized patient education materials.}},
  booktitle = {PubMed},
  year = {2006},
}

Loading related content...

Figure