AI-Powered Patient Screening and Twin Clinical Decision Support
The AI-Powered Twin Clinical Decision Support, a co-pilot tool for physicians, is developed using Large Language Models (LLMs) by enhanced by Preceptor AI. This innovative solution integrates various data sources including Electronic Health Records (EHR), patient data, historical treatment information, outcomes of medical tests, medical imaging, and results from pre-screening. Its primary aim is to aid in decision-making and streamline medical operations. By leveraging this comprehensive data, the tool bolsters the physician’s ability to make decisions with greater confidence and efficiency. Ultimately, this leads to a reduction in medical errors and an improvement in diagnostic precision.
Thailand's Health Care Challenges:
"Higher risk of patient safety and physician’s burn out"
Thailand has been facing a healthcare crisis due to a higher demand and a shrinking workforce. This has led to strain on the healthcare service quality, patient safety, and physician burnout.
The combination of a strained screening system and a reduced medical workforce jeopardizes patient safety, while time pressures and hurried services impair decision-making and thorough data collection, diminishing diagnostic and analytical precision due to the physical and mental strain on healthcare providers. Consequently, there is a heightened risk to patient safety and a higher incidence of diagnostic errors, particularly in hospitals serving large patient populations.
Impacts on Patients & National Budget
Those challenges significantly affect patient, physicians and public health budget.
The provided graphic underscores critical issues within the healthcare system, emphasizing an increased risk to patient safety and physician burnout. It points out that 1 in 10 patients experience harm due to unsafe care, and this substandard care contributes to 7% of all healthcare visits, totaling 400,000 incidents. The pie charts reinforce this by showing that 7% of care provided is unsafe, which not only jeopardizes patient health but also has substantial economic consequences, inflicting a loss of 9.6 billion THB per year—amounting to 6% of the public health budget. The graphic suggests that factors such as unnecessary visits, which make up 60% of the total, lead to rushed service and poor quality of care. Additionally, it illustrates the challenge of a diminishing workforce in healthcare, which contributes to an increased workload for current staff and exacerbates the risk of burnout among physicians. This data, drawn from a study by Luankongsomchit et al., (2023), highlights the pressing need for systemic improvements to address these pain points.
The figure above illustrates the key pain points and primary causes that contribute to patient safety issues and physical burnout.
Segmentation
-
Hospital Seekers: These are individuals who tend to visit medical school hospitals whenever they experience symptoms, choosing to see doctors promptly without delay.
-
Hospital-Averse: This group generally avoids going to public or community hospitals and tends to wait until their symptoms worsen significantly before considering a visit.
-
Visiting for Serious Concerns: Patients in this segment typically use private hospitals and are inclined to research their symptoms online, choosing to visit the hospital only for serious health concerns and often foregoing medical advice for less severe issues.
Patient Journey & Pain Points
Healthcare Worker Journey & Pain Points
Insights
Patients often do not receive timely and appropriate care, partly due to inefficient spending, which can be a consequence of the lack of an efficient screening process. This situation contributes to a heightened risk of patient safety. For physicians, the graphic suggests that overcrowding and rushed decision-making are significant issues, likely exacerbated by the screening inefficiencies and potentially leading to compromised patient safety. The overlapping area between patients and physicians indicates that both experience the consequences of these systemic problems.
Use-case components
Symptom & Patient screening (Triage)
This Screening process integrated with multiple sources, such as EHR, patient data, medical test results, medical images etc. not only assists in initial health assessments, but also helps guide patients to the appropriate level of care.
Co-Pilot to empower physicians
Co-pilot tool for physicians is developed by utilizing LLMs by Preceptor AI to assist the decision-making and better operations.
Design System & Prototyping
The Design System
The user interface (UI) of Twin CDS is crafted with a strong emphasis on user accessibility and patient safety. Interactive elements such as buttons and call-to-action prompts are designed to be prominent and unambiguous to healthcare personnel. This clarity in design is crucial, as it can bolster the confidence of the staff, leading to decisive actions and reducing the likelihood of hesitation when dispensing medications. Such a thoughtful UI layout can contribute significantly to streamlining the medication distribution process, ensuring that healthcare workers can focus on providing care without the added concern of navigating a complex system.
Symptom Checker &
Assessment
Chief Complaints
Patients report their main symptoms and specify how long they have been experiencing this symptom. This process is similar to the initial inquiry often conducted by a physician.
Patients respond to multiple set of questions to detail their symptoms including high yield questions and alarm symptoms.
Symptom Assessment
Triage Results
& Digital Patient Transfer
Triage - Screening Results
The Screening result indicates level of urgency and displays guideline for further appropriate care and actions.
The preliminary screening data will be seamlessly consolidated into the doctor’s twin CDS and HIS System.They can present QR at the hospital. This enables nurse and doctors to quickly access preliminary assessment result, color codes, differential diagnosis (DDx), enhancing the efficiency of the diagnostic process.
Digital Patient Transfer
Twin CDS
A Co-Pilot Tool for Physicians
The Twin CDS Homepage features a user-friendly interface that prominently displays the brand's logo and AI capabilities. A navigation panel offers easy access to various functionalities such as patient lists and schedules. Central to the homepage is the workspace with sections for ASR order management, lab result interpretation, and differential diagnosis assistance, all aimed at enhancing clinical decision-making. Interactive elements like search bars and buttons facilitate efficient workflow management, while the footer reiterates the tool's purpose as a supportive co-pilot for physicians. Notifications and status updates are conveniently positioned to alert users to important information.
Patient Info & HIS Linkage
Seamless linkage between Twin CDS and HIS
The feature depicted focuses on the seamless linkage between the Clinical Decision Support (CDS) system and the Hospital Information System (HIS), enabling efficient retrieval of patient data. This integration facilitates quick access to comprehensive patient information, which is crucial for informed decision-making. Through the interface, healthcare professionals can view and manage doctor orders, with options for setting one-time or continuous orders, ensuring timely and accurate treatment schedules. A QR code scanning functionality suggests a straightforward method for data entry or retrieval, likely enhancing the speed and accuracy of information exchange. The system's design, with clear categorization and search functionality, appears user-friendly, aiming to minimize the time spent navigating the system. This allows for more time to be dedicated to patient care, demonstrating the CDS's commitment to improving workflow efficiency and clinical outcomes. Overall, this feature represents a key component in streamlining healthcare operations by bridging the gap between patient records and active clinical use.
Order ASR
Order ASR (Audio Speech Recognition)
The feature shown demonstrates an innovative approach to inpatient care, where doctors can leverage AI and Automated Speech Recognition (ASR) technology to place orders for patients. After accessing a patient's data from the Hospital Information System (HIS), the physician can use a voice command interface to issue orders, which the system records and transcribes in real-time. This hands-free method allows for more efficient operations, as doctors can quickly articulate care instructions without manual typing, reducing time spent on administrative tasks. The interface includes functionalities for both one-time and continuous orders, ensuring flexibility in care delivery. Visual cues and timers within the ASR interface assist physicians in tracking the duration of order entries. This ASR Order feature, augmented by AI, not only streamlines the workflow but also minimizes the potential for human error, contributing to safer and more precise patient care. The integration of such a system is a testament to the evolving role of technology in enhancing healthcare delivery and operational efficiency.
Lab Interpretation
Comprehensive investigation through the analysis of patient's lab results.
A co-pilot for doctors to increase patient safety consists of 4 sections including Complete patient’s profile, Abnormal point identification, High-Risk Disease Assessment, and Future Examination Guidance. It is an advanced part of Twin CDS co-pilot system for physicians that aids in elevating patient safety by providing a detailed analysis of laboratory results. It compiles a full patient profile, enabling doctors to quickly identify any abnormal lab values that may indicate potential health risks. The system is designed to assess the risk of severe diseases and suggest additional tests or follow-up exams, supporting clinicians in making informed decisions about patient care. Notable is the clear presentation of data, with abnormal results highlighted, streamlining the identification process for high-risk conditions. Furthermore, this tool offers guidance on subsequent steps for examination, thereby ensuring a comprehensive approach to patient health management. Through its intuitive design, the system empowers healthcare providers with critical insights that are essential for proactive and preventive medical care.
A/B Testing & Usability Testing
A/B Testing
To know user feedback towards usability of symptom screening, "Which variant is more effective?" and "Which version aligns best with specific user persona?", this helped us understand which variant was more effective and determine the user profiles that best resonated with each version.
A Version : Hybrid Format
A format that combines two-distinctive characteristic of both questionnaire and chat interaction.It enables users a sense of two-way communication and advancement of AI analytics.
B Version : Questionnaire Format
A traditional interaction of a symptom checker questionnaire involves a structured set of questions designed to gather information about an individual's symptoms and health condition.
Based on the usability test, users responded more favorably to Version A, appreciating its intelligent AI-chatbot approach. The advantages highlighted for this version include its user-friendly and familiar chat-style interface, allowing users to easily review previous responses by scrolling, contributing to a modern and sleek chat experience. Nevertheless, the ability to edit answers requires further refinement.
In contrast, Version B was also considered user-friendly but felt more traditional, akin to a conventional questionnaire. Its linear progression necessitates frequent clicking and restricts users from revisiting past responses, which is seen as a drawback compared to the more dynamic and flexible Version A.
Update In progress
Stay tuned!