Decoding prognosis: Unraveling physicians' responses in electronic treatment plans.

Abstract

230 Background: Communicating prognosis to cancer patients is a critical aspect of shared decision making in healthcare. Practices in The US Oncology Network (The Network) use an electronic version of a treatment plan based on the Institute of Medicine (IOM) care management plan for patients receiving cancer treatment. Physicians document in their own words the prognosis section of the treatment plan. This study seeks to better understand the responses provided by physicians in the prognosis section. Methods: We evaluated word frequency analysis using FreqDist library, conducted sentiment analyses using the Natural Language Toolkit (NLTK) Sentiment Intensity Analyzer, and performed thematic analyses using Latent Dirichlet Allocation library on a random sample of 35,000 treatment plans developed in 2023 for practices in The Network. Results: 32% of the responses were 5 words or less, with 3.8% single word responses. Words like “excellent, good, fair, poor, guarded” or a combination thereof were commonly used to describe the prognosis. 26% of responses were ≥12 words in length, and generally were personalized to the patient’s disease, treatment, and/or contained situation specific narratives. Responses <12 words generally contained templated phrases, or minimally customized responses. 38% of the responses contained a numerical value, representing either a percentage probability of survival, chance of response to treatment, or a quantification of time (months, years). Thematic analyses conducted on responses ≥12 words revealed that the predominant theme focused on survival represented in different ways, followed by response to therapy, probability of recurrence or relapse, and disease control. Sentiment analyses revealed that 60.6% of responses denoted a negative sentiment. 71.7% of responses ≤5 words in length had a negative sentiment, compared to 58.3% of responses ≥12 words in length. Conclusions: This study reveals that prognosis responses from physicians used on treatment plans shared with patients are very diverse. The manner in which prognosis is presented to patients can greatly influence their understanding, engagement, and involvement in making treatment decisions. To optimize shared decision making in cancer care, healthcare providers must strive to effectively communicate prognosis by using clear and empathetic language, ensuring patient comprehension, and tailoring the information to individual patient needs and preferences. The data set has high potential for further study to better understand the role and impact of written treatment plans to help physicians refine this documentation for better patient understanding.