Publication Information
ISSN 2691-8803
Frequency: Continuous
Format: PDF and HTML
Versions: Online (Open Access)
Year first Published: 2019
Language: English
Journal Menu |
Editorial Board |
Reviewer Board |
Articles |
Open Access |
Special Issue Proposals |
Guidelines for Authors |
Guidelines for Editors |
Guidelines for Reviewers |
Membership |
Fee and Guidelines |
A Patient’s Visit to the Clinic: Using Structured AI to Create a Socratic Tutor to Promote Critical Thinking about COVID-19 and Congestive Heart Failure
Taylor Mulvey1, Stephen Rappaport2, Yehoshua Deitel3, Tim Morford4, Angelica DiLorenzo5, Sunaina Saharan5, Muhammad Yaseen5, Howard Moskowitz4*
1The Thomas More School, Oakdale, CT;
2SDR Consulting Inc., Norwalk, CT;
3Sifra Digital Inc., Jerusalem, Israel;
4Cognitive Behavioral Insights LLC., Albany, NY;
5Global Population Healthcare Management Forum, Brooklyn, NY.
Received Date: January 17, 2024; Accepted Date: February 19, 2024; Published Date: March 06, 2024;
*Corresponding author: Howard Moskowitz, Cognitive Behavioral Insights LLC., Albany, NY. Mailing Address: 11 Sherman Ave., White Plains, NY, USA, 10605; Email: mjihrm@gmail.com.
Citation: Mulvey T, Rappaport S, Deitel Y, Morford T, DiLorenzo A, Saharan S, Yaseen M, Moskowitz H (2024) A Patient’s Visit to the Clinic: Using Structured AI to Create a Socratic Tutor to Promote Critical Thinking about COVID-19 and Congestive Heart Failure. Adv Pub Health Com Trop Med: APCTM-195.
DOI: 10.37722/APHCTM.2024102
Abstract
Through the use of AI embedded in a research platform, the paper shows how a novice can use AI to 'learn' the questions and answers that might be relevant to ask a patient who has either COVID-19 or CHF (Congestive Heart Failure). The strategy to learn uses AI as a Socratic Tutor, with the user identifying the diseases, the age, and gender of the presumed patient. The user types in a request to identify relevant questions to ask the patient, the reason why the question is being asked of the patient, and then likely thoughts that the patient might be thinking before answering the question. The approach enables the user to quickly explore many types of diseases, types of patients, all with the goal of learning about what might be encountered in the patient-doctor interaction. The approach is presented as a work-in-progress, to demonstrate the contribution of AI to medical education.
Introduction
As part of their education and training, medical students are taught appropriate questions to ask when talking to a patient. Knowing how to talk to a patient, how to listen, how to recommend, constitute three different aspects of the softer side of medicine, in previous years called the doctor’s bedside manner (Kanofsky, 2020). Much of the knowledge has been codified (Silverman, 2012), but it is still a truism to say that the best teacher is experience. It is with experience that the novice doctor becomes a professional, 'hearing more' by understanding what patients are trying to say.
The recent advance in AI, artificial intelligence, opens up the possibility that there may be a faster way to understand the patient. To some, this faster way might be an esoteric system of measuring physiological responses and interpreting them (Wu et al., 2023), but that esoteric system of giving meaning to the cacophony of physiological responses is far in the future. Rather, it might well be that AI can 'teach' a student how to listen, or perhaps simulate what a patient might be thinking. It is this latter way of understanding the patient which constitutes the topic of this paper (Afzal et al., 2020; Baidoo-Anu & Ansah, 2023; Chan & Zary, 2019).
The objective of this paper is to introduce and demonstrate a recently developed approach, using a combination of design thinking from technology called Mind Genomics (Moskowitz, 2012; Zemel et al., 2019), along with AI to provide some of the output to queries called for by this Mind Genomics technology (Le, 2019; Liao et al., 2020), and finally with sub-text, providing what might happen in terms of doctor-patient interaction, and what the patient might be thinking (Narayanan et al., 2023).
The importance of knowing the patient cannot be overestimated. The question is not the importance, but rather whether the 'new' world of AI can be adapted to a rapid, scalable, cost-effective system to teach and to train. Or are we for the next while to remain with training and experience.
The approach shown in this paper originated decades ago, when users began to realize that the subjective world cannot be well understood by isolating a topic and studying it, such as one does with questionnaires (Rowley, 2014). An example of this one-at-a-time topic comes from questionnaires about the medical experience, such as the Press Ganey score (Wilson et al., 2016). The method is deceptively easy but fails to take into account the complexity of the situation.
During the early 1990s, author HRM and colleagues began to experiment with more 'ecologically relevant' test stimuli in the effort to understand how people make decisions. The work involved creating small descriptive phrases, mixing them together to create vignettes, presenting these vignettes to people, obtaining ratings, and then later using statistical methods to deconstruct the ratings into the contribution of the elements. The process seems clunky and overly complicated compared to the simple method of presenting single general ideas, instructing the respondent to rate the idea, and then averaging the ratings. The reality captured by the vignettes turned out to be more instructive, however (Moskowitz & Silcher, 2006). The user could modify these ideas, also known as messages or elements. The ideas could be polished, shaped, expanded or contracted, changed in meaning, changed in their very nature and tonality. As a consequence, it soon became obvious that users could understand and study more complex situations, and better understand the minds of people.
Methods
The simulation approach used a Mind Genomics platform, Socrates as a Service (https://socratesasaservice.com/u/dashboard). The idea was to create a scenario instructing AI (Idea Coach) to simulate a visit of a patient to a doctor. The AI was told that the person to be simulated was a medical professional. The objective was to deal with asking questions to the patient. The AI was to suggest a question that the medical professional might ask, provide a reason for asking the question, and then come up with four thoughts that the patient might have upon being asked the question. The AI was given the background as shown below:
“I am interested to learn how to ask questions in a clinic. My patients often come in with complaints that are not clearly expressed. I want to learn what questions to ask my patients about their condition, without alarming them. For each question that I ask, I want to understand WHY I am asking it. I also want to get four thoughts that are going through the mind of my patient, when my patient hears me ask the question. This will give me the question, the reason why I ask the question, and four things that my patient might be thinking before answering the question.”
In addition to the standard or ‘boilerplate’ background, the AI was also provided with a short description of the patient.
It should be noted that the information that will be provided in this paper comes from only two of 20 different iterations, where the nature of the patient varies (e.g., age, gender, illness). The actual material which emerges from the effort is embedded in the so-called ‘Idea Book’ for the project. Each iteration with a new definition of the patient requires about 30 seconds, allowing the user to create a reference book for any indication. The Idea Book is emailed to the user about 20-30 minutes after the first part of the Mind Genomics study has been completed, where questions and answers are selected.
Table 1 shows the results for two individuals, all other prompts to the AI being the same. The first individual is a 50-year-old male suffering from COVID-19. The individual is specified as being American. The second individual is the same 50-year-old male, this time suffering from CHF, congestive heart failure.
The important thing to learn from the top Table 1 is the introduction to critical thinking offered by creating an ordinary dialectic in Idea Coach. The effort does not require anything in the way of deep understanding. Rather, the AI prompt is set up in the form of a simple request that one might ask of a personalized tutor. The Idea Coach AI is able to transform the ‘almost chatty request’ into the proper format.
Four things are relevant here:
- As just noted, the request is easy to create. One need not become an expert in so-called prompt-engineering. Natural language suffices, as long as the description of what is desired is crisp, concise, and the language used is meaningful.
- The user is able to configure the language to achieve the correct tonality. One example is the request that ‘I want to learn what questions to ask my patients about their condition, without alarming them.’ This language sets the tone of what emerges.
- The user can ask the Idea Coach AI to provide what ideas may be coursing through the mind of the patient.
- The user can specify the amount of information provided, such as the request to provide ‘four things that the patient might be thinking before answering the question.’ The results in Table 1 will show that occasionally fewer than four things are provided.
- Once the request is made, the user simply presses the Idea Coach button, to generate the necessary information. The answers are returned in approximately 10-30 seconds, sufficient to allow time for the user to review, and either do the task again, edit the squib and then do the task, or move on and end the process.
What could be causing these symptoms? Will it get worse? How long will this last? Can I manage it at home? Should I be concerned about a high temperature? How can I bring the fever down? Will it come back eventually? Can I still enjoy food? Is it just part of having COVID-19? Will it go away on its own? How long will it last? Can I treat it at home? How long until I start feeling better? Is my recovery progressing as expected? How can I protect my loved ones? Will I be isolated? Should I have avoided certain places? Will I need to quarantine? Will they interfere with my COVID-19 treatment? Should I stop any medications? Am I at higher risk of severe illness? How can I protect myself? How can I ensure their safety? Will I be responsible for their illness? How can I cope with the stress? Will it impact my recovery? What other precautions should I take? How can I avoid contracting it again? How can I better understand COVID-19? Will I ever fully recover? Will the doctor understand what I mean? How serious is it? How will it affect my daily life? Is it related to something specific? Should I have seen a doctor earlier? Will it be a long-term issue? Should I avoid certain things? Will the doctor recommend any lifestyle changes? Is it a serious problem? Is my current treatment working? Will the doctor need to adjust my medication? Is my condition worsening? Will the doctor recommend alternative treatments? Will my medications be changed? Should I mention minor illnesses? Is it important to disclose past surgeries? Will it affect my treatment options? Can it be related to my symptoms? Should I inform the doctor about past reactions? Will it affect the prescribed medications? Will it affect my diagnosis? Should I inform about relatives' chronic conditions? Is it just a coincidence? Should I mention work-related stress? Will the doctor recommend stress management techniques? Is my condition purely stress-related? Is it a serious symptom? Will I need an inhaler? Is it related to my heart condition? Will it go away on its own? Will the doctor prescribe diuretics? Can it lead to other complications? Will it improve with treatment? Will I need to modify my lifestyle? Can it indicate something more severe? Will it pass on its own? Should I mention it if it's only occasional? Will I need a cardiac catheterization? Can I still be physically active? Will it worsen my condition? Will the doctor provide exercise guidelines? Is it just a normal part of aging? Will the doctor prescribe sleep aids? Can it lead to other complications?
I am interested in learning how to ask questions in a clinic. My patients often come in with complaints that are not clearly expressed. I want to learn what questions to ask my patients about their condition, without alarming them. For each question that I ask, I want to understand WHY I am asking it. I also want to get four thoughts that are going through the mind of my patient when my patient hears me ask the question. This will give me the question, the reason why I ask the question, and four things that my patient might be thinking before answering the question.
My Patient is a 50-Year-Old Male Suffering from COVID-19
My Patient is American Born
Questions
Reason for the question
Patient's thoughts (often 3 of 4)
“How would you describe your current symptoms?” (To assess the severity and nature of their complaints)
Understanding the symptoms presents an overall picture of the patient's condition.
Am I experiencing symptoms that others with COVID-19 have reported?
“Have you experienced any difficulty breathing?” (To assess respiratory distress)
Shortness of breath is a common symptom of COVID-19 and can indicate the severity of the infection.
Could this difficulty breathing be a sign of something serious?
“Have you had a fever? If so, what was the highest temperature you reached?” (To assess if fever is present and its severity)
Fever is commonly associated with COVID-19 and can indicate the body's response to the infection.
Is my fever within normal limits?
“Have you experienced any loss of taste or smell?” (To assess possible COVID-19 symptoms)
Loss of taste or smell is a distinct symptom of COVID-19 and aids in diagnosing the condition.
Is it common for people with COVID-19 to lose taste or smell?
“Are you experiencing any chest pain or pressure?” (To assess potential heart or lung involvement)
Chest pain can indicate complications from COVID-19, such as pneumonia or heart problems.
Could this chest pain be life-threatening?
“Have you noticed any rash or skin changes?” (To assess if COVID-19 is causing dermatological symptoms)
Skin manifestations have been observed in some COVID-19 cases and may provide valuable diagnostic information.
Is it common for COVID-19 patients to have skin issues?
“How long have you had these symptoms?” (To understand the duration of illness)
Duration of symptoms helps determine the stage of the infection and possible complications.
Is the duration of my symptoms normal?
“Have you had any known exposure to individuals diagnosed with COVID-19?” (To assess potential transmission sources)
Contact tracing is crucial in preventing further spread of the virus and identifying potential risk factors.
Did I contract the virus from someone I know?
“Any recent international travel or exposure to high-risk areas?” (To assess travel-related exposure)
Travel history may help identify potential sources of infection and determine the need for further testing or monitoring.
Could I have picked up the virus during my recent travels?
“Are you currently taking any medications?” (To evaluate possible drug interactions)
Determining the patient's current medication regimen is essential to prevent any adverse effects or interactions with COVID-19 treatments.
Can I continue taking my regular medications?
“Do you have any pre-existing medical conditions?” (To assess potential risks or complications)
Some pre-existing conditions can increase the patient's vulnerability to severe outcomes or complications from COVID-19.
Will my existing health condition worsen due to COVID-19?
“Have you been in contact with any immunocompromised individuals?” (To assess potential risk to vulnerable populations)
Identifying contact with immunocompromised individuals is crucial to prevent transmission to those at higher risk.
Could I have unintentionally spread the virus to someone vulnerable?
“Have you experienced any mental health changes or increased stress during this time?” (To address mental well-being)
COVID-19 can have psychological impacts, and assessing mental health helps provide comprehensive care
Is it common to experience worsened mental health during COVID-19?
“Are you practicing recommended hygiene measures and following public health guidelines?” (To promote safe behavior)
Encouraging adherence to preventive measures is crucial in preventing further transmission and protecting the patient and others
Am I doing enough to protect myself and others?
“Do you have any questions or concerns about your current condition or COVID-19 in general?” (To address patient education)
Providing an opportunity for the patient to voice their concerns or seek clarification helps with informed decision-making
Are there any important aspects of my condition that I'm missing?
My Patient is a 50-Year-Old Male Suffering from Congestive Heart Failure
My Patient is American Born
Question
Reason for the question
Patient’s thoughts (often 4 of 4)
“How would you describe the symptoms you're experiencing?"
To understand the nature and severity of the patient's complaints.
Am I describing it accurately?
“When did you first notice these symptoms?"
To determine the onset and duration of the condition.
Do I remember the exact date?
"Do your symptoms have any triggers or worsen in certain situations?"
To identify any environmental factors that contribute to the symptoms
What could be causing this?
"Have you noticed any improvements or changes in your symptoms recently?"
To assess the progress of the condition and the effectiveness of any ongoing treatments.
Is it getting better on its own?
"Are you taking any medications, supplements, or herbal remedies?"
To determine if any medications might be interacting with their condition or worsening the symptoms.
Should I mention even over-the-counter medications?
"Have you had any recent illnesses, surgeries, or hospitalizations?"
To assess the patient's medical history and potential factors that may have contributed to the present condition
Is my recent illness related?
“Do you have any known allergies?"
To ensure that potential allergens are considered during the diagnosis and treatment process.
Should I mention all types of allergies?
"Has anyone in your family experienced similar health conditions?"
To assess the familial risk factors and potential hereditary component contributing to the condition
Should I mention extended family?
"How would you rate your overall stress levels?"
To understand if stress plays a role in exacerbating the symptoms or overall health.
Does stress affect my condition?
"Do you experience any difficulty with breathing or shortness of breath?"
To assess the impact of the condition on respiratory function and potential complications.
Will I require additional tests?
"Have you noticed any swelling in your ankles, legs, or abdomen?"
To evaluate the presence of fluid retention, a common symptom of congestive heart failure.
Is it normal to have swelling?
"How would you describe your energy levels and fatigue?"
To understand the impact of the condition on daily activities and overall quality of life.
Is fatigue a normal symptom?
"Have you experienced any chest pain, discomfort, or tightness?"
To assess the presence of angina, a common symptom in congestive heart failure.
Is it a heart attack?
"Are you currently engaging in any physical exercise or activities?"
To determine the patient's activity level and discuss any necessary modifications to their routine
Is exercise recommended?
"Are you experiencing any difficulties sleeping, such as waking up frequently or shortness of breath while lying down?"
To evaluate the presence of sleep disturbances related to congestive heart failure.
Should I mention it if it's occasional?
Table 1: Results from the study.
AI to Broaden Horizons and to Teach Rapidly and with Targeted Instruction
Beyond the creation of questions lies the possibility of uncovering patterns in the material offered. After the user has completed the acquisition of questions, viz., iterations with or without changing the squib provided to the AI, the AI program moves to the second phase. As noted above, the user can request iteration after iteration, changing the squib at will. The AI program treats each iteration as a totally separate set of information to be analyzed. The AI program ‘summarizes’ each iteration with the following AI-based summaries:
- Themes – which identify the recurrent issues across all of the material generated in the iteration.
- Perspectives – which identify what is provided, and a critique of both the good and the bad aspects of the provided material.
- What is missing from the information provided by that iteration?
- Alternative viewpoints – different interpretations emerging from the same question or answer.
- Interested audience – who would be interested in the material.
- Opposing audience – reasons to doubt the validity of the material or its applicability to the situation.
- Innovation – ‘new’ ideas, or at least hints about new ideas. The innovations are based only on the AI and are simply points for discussion.
Symptoms assessment and severity evaluation: Exposure assessment: Medical and health assessment: Education and prevention: Themes - 50-Year-Old Male Suffering from Congestive Heart Failure Understanding symptom characteristics: Assessing medical history and current condition: Evaluating overall health and well-being: Assessing lifestyle factors: Perspectives - 50-Year-Old Male Suffering from Congestive Heart Failure 1. Understanding symptom characteristics: a. Plus factors: i. Helps in identifying the nature and severity of symptoms, leading to appropriate treatment. ii. Allows for tracking the progression and changes in symptoms over time. b. Minus factors: i. Relies on accurate and detailed reporting from patients, which can be subjective and prone to errors. ii. Doesn't provide a definitive diagnosis on its own, further medical evaluations may be required. c. Interesting factors: i. Triggers or situations that worsen symptoms can reveal specific environmental or lifestyle factors to address. 2. Assessing medical history and current condition: a. Plus factors: i. Provides important information on previous treatments and outcomes for better decision-making. ii. Enables identification of potential risks and vulnerabilities based on familial or hereditary factors. b. Minus factors: i. Relies on accurate recall and reporting of medical history, which can be challenging for some patients. ii. Limited by the availability and accessibility of medical records from different healthcare providers or institutions. c. Interesting factors: i. Allergies can reveal possible interactions or contraindications for certain medications or treatments. 3. Evaluating overall health and well-being: a. Plus factors: i. Considers the broader impact of lifestyle and emotional well-being on overall health. ii. Helps in assessing the effectiveness of stress management techniques on symptom management. b. Minus factors: i. Difficulties in quantifying or measuring subjective experiences like energy levels or stress impact. ii. Doesn't provide a complete picture of physical health and may require other assessments or tests. c. Interesting factors: i. Swelling in different body parts could indicate underlying issues related to circulation or organ function. 4. Assessing lifestyle factors: a. Plus factors: i. Allows for identification of potential lifestyle changes that could positively impact symptoms. ii. Provides insights into patients' physical activity levels and exercise habits. b. Minus factors: i. Relies on accurate self-reporting, which may be influenced by social desirability bias. ii. Doesn't account for other potential lifestyle factors (e.g., nutrition, smoking) that could impact symptoms. c. Interesting factors: i. Difficulties sleeping while lying down could suggest potential sleep disorders or respiratory issues. What is Missing - 50-Year-Old Male Suffering from Congestive Heart Failure Alternative Viewpoints - 50-Year-Old Male Suffering from Congestive Heart Failure There may be several audiences that could potentially oppose the topic of discussing COVID-19 symptoms and related questions. These could include: It is important to address these opposing viewpoints with accurate information, empathy, and open dialogue to promote understanding and encourage adherence to public health guidelines. It is important for healthcare providers to create a safe and non-judgmental environment where patients feel comfortable and encouraged to share their symptoms openly, ensuring accurate diagnosis and appropriate treatment. Discussion and Conclusions This paper demonstrates that the introduction of AI into the world of attitudinal research may accelerate the search for facts by prompting the user with questions. Rather than allowing the researcher to learn to think in a trial-and-error fashion, the approach presented here forces the user, especially the novice, to formulate the situation, such as the problem, as well as a variety of relevant side issues, such as how the problem is expressed, what is the underlying thought, how will people of a certain type respond to the problem. The approach guides the researcher in a different direction, thinking about the meaning of different outcomes to the question before the research effort begins. That outcome itself is interesting. Even more interesting, perhaps, is the summarization by AI of the questions presented by AI. When all is said and done, the outcome of the effort is a rapid, Socratic tutor, AI-based, with this tutor ready to work tirelessly, accepting query after query, doing its job, and then providing a summary. What this paper presents is one effort, an effort lasting approximately 1-2 minutes of the user’s time. For that investment, the payback is enormous. What is even more promising is the easily demonstrated extension of the effort so that the enterprising user can parametrically explore the topic, varying perhaps the WHO (who is the person) and the WHAT (the reason driving the individual to visit the doctor or the clinic). In a matter of one simple hour, the enterprising, and even without more programming, the enterprising user can explore age, gender, and disease state, with perhaps 30-40 options. It is this type of generative power that makes the AI effort an easy jump into the future. References
Themes - 50-Year-Old Male with COVID-19
Perspectives - 50-Year-Old Male with COVID-19
What is Missing - 50-Year-Old Male with COVID-19
Alternative Viewpoints - 50-Year-Old Male with COVID-19
Interested Audiences - 50-Year-Old Male with COVID-19
Interested Audiences - 50-Year-Old Male Suffering from Congestive Heart Failure
Opposing Audiences - 50-Year-Old Male with COVID-19
Opposing Audiences - 50-Year-Old Male Suffering from Congestive Heart Failure
Audiences that might oppose the topic would include:
Innovations - 50-Year-Old Male with COVID-19
Innovations - 50-Year-Old Male Suffering from Congestive Heart Failure
Researchers and consumers of research have accustomed themselves to a world of facts and, perhaps, hypotheses about how the world works. Advances in science come from new facts, new pieces of information about how the world works. The Popperian view of science is that the hypotheses emerging from these facts can be subject to falsification (Nola, 1987). From this hypothetico-deductive system, we begin to think that the job of the researcher is to use science to establish 'what’s really happening.' And, in the words of S.S. Stevens of Harvard University, 'theories abound in the absence of facts' (Stevens, 1939). In other words, the focus is on fact, individual facts bound together by the sinews of hypotheses.