Diagnostic reasoning prompts reveal the potential for large language model interpretability in medicine.

One of the major barriers to using large language models (LLMs) in medicine is the perception they use uninterpretable methods to make clinical decisions that are inherently different from the cognitive processes of clinicians. In this manuscript we develop diagnostic reasoning prompts to study whether LLMs can imitate clinical reasoning while accurately forming a diagnosis. We find that GPT-4 can be prompted to mimic the common clinical reasoning processes of clinicians without sacrificing diagnostic accuracy. This is significant because an LLM that can imitate clinical reasoning to provide an interpretable rationale offers physicians a means to evaluate whether an LLMs response is likely correct and can be trusted for patient care. Prompting methods that use diagnostic reasoning have the potential to mitigate the "black box" limitations of LLMs, bringing them one step closer to safe and effective use in medicine.

Performance effectiveness of vital parameter combinations for early warning of sepsis-an exhaustive study using machine learning.

Objective: To carry out exhaustive data-driven computations for the performance of noninvasive vital signs heart rate (HR), respiratory rate (RR), peripheral oxygen saturation (SpO2), and temperature (Temp), considered both independently and in all possible combinations, for early detection of sepsis. Materials and methods: By extracting features interpretable by clinicians, we applied Gradient Boosted Decision Tree machine learning on a dataset of 2630 patients to build 240 models. Validation was performed on a geographically distinct dataset. Relative to onset, predictions were clocked as per 16 pairs of monitoring intervals and prediction times, and the outcomes were ranked. Results: The combination of HR and Temp was found to be a minimal feature set yielding maximal predictability with area under receiver operating curve 0.94, sensitivity of 0.85, and specificity of 0.90. Whereas HR and RR each directly enhance prediction, the effects of SpO2 and Temp are significant only when combined with HR or RR. In benchmarking relative to standard methods Systemic Inflammatory Response Syndrome (SIRS), National Early Warning Score (NEWS), and quick-Sequential Organ Failure Assessment (qSOFA), Vital-SEP outperformed all 3 of them. Conclusion: It can be concluded that using intensive care unit data even 2 vital signs are adequate to predict sepsis upto 6 h in advance with promising accuracy comparable to standard scoring methods and other sepsis predictive tools reported in literature. Vital-SEP can be used for fast-track prediction especially in limited resource hospital settings where laboratory based hematologic or biochemical assays may be unavailable, inaccurate, or entail clinically inordinate delays. A prospective study is essential to determine the clinical impact of the proposed sepsis prediction model and evaluate other outcomes such as mortality and duration of hospital stay.

Real-time alerting system for COVID-19 and other stress events using wearable data.

Early detection of infectious diseases is crucial for reducing transmission and facilitating early intervention. In this study, we built a real-time smartwatch-based alerting system that detects aberrant physiological and activity signals (heart rates and steps) associated with the onset of early infection and implemented this system in a prospective study. In a cohort of 3,318 participants, of whom 84 were infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), this system generated alerts for pre-symptomatic and asymptomatic SARS-CoV-2 infection in 67 (80%) of the infected individuals. Pre-symptomatic signals were observed at a median of 3 days before symptom onset. Examination of detailed survey responses provided by the participants revealed that other respiratory infections as well as events not associated with infection, such as stress, alcohol consumption and travel, could also trigger alerts, albeit at a much lower mean frequency (1.15 alert days per person compared to 3.42 alert days per person for coronavirus disease 2019 cases). Thus, analysis of smartwatch signals by an online detection algorithm provides advance warning of SARS-CoV-2 infection in a high percentage of cases. This study shows that a real-time alerting system can be used for early detection of infection and other stressors and employed on an open-source platform that is scalable to millions of users.

Real-time Alerting System for COVID-19 Using Wearable Data.

Early detection of infectious disease is crucial for reducing transmission and facilitating early intervention. We built a real-time smartwatch-based alerting system for the detection of aberrant physiological and activity signals (e.g. resting heart rate, steps) associated with early infection onset at the individual level. Upon applying this system to a cohort of 3,246 participants, we found that alerts were generated for pre-symptomatic and asymptomatic COVID-19 infections in 78% of cases, and pre-symptomatic signals were observed a median of three days prior to symptom onset. Furthermore, by examining over 100,000 survey annotations, we found that other respiratory infections as well as events not associated with COVID-19 (e.g. stress, alcohol consumption, travel) could trigger alerts, albeit at a lower mean period (1.9 days) than those observed in the COVID-19 cases (4.3 days). Thus this system has potential both for advanced warning of COVID-19 as well as a general system for measuring health via detection of physiological shifts from personal baselines. The system is open-source and scalable to millions of users, offering a personal health monitoring system that can operate in real time on a global scale.

Deriving High Performance Alerts from Reduced Sensor Data for Timely Intervention in Acute Hypotensive Episodes.

Alerting critical health conditions ahead of time leads to reduced mortality rates. Recently wirelessly enabled medical sensors have become pervasive in both hospital and ambulatory settings. These sensors pour out voluminous data that are generally not amenable to direct interpretation. For this data to be practically useful for patients, they must be translatable into alerts that enable doctors to intervene in a timely fashion. In this paper we present a novel three-step technique to derive high performance alerts from voluminous sensor data: A data reduction algorithm that takes into account the medical condition at personalized patient level and thereby converts raw multi-sensor data to patient and disease specific severity representation, which we call as the Personalized Health Motifs (PHM). The PHMs are then modulated by criticality factors derived from interventional time and severity frequency to yield a Criticality Measure Index (CMI). In the final step we generate alerts whenever the CMI crosses patientdisease-specific thresholds. We consider one medical condition called Acute Hypotensive Episode (AHE). We evaluate the performance of our CMI derived alerts using 7,200 minutes of data from the MIMIC II [7] database. We show that the CMI generates valid alerts up to 180 minutes prior to onset of AHE with accuracy, specificity, and sensitivity of 0.76, 1.0 and 0.67 respectively, outperforming alerts from raw data.

Data to diagnosis in global health: a 3P approach.

BACKGROUND: With connected medical devices fast becoming ubiquitous in healthcare monitoring there is a deluge of data coming from multiple body-attached sensors. Transforming this flood of data into effective and efficient diagnosis is a major challenge. METHODS: To address this challenge, we present a 3P approach: personalized patient monitoring, precision diagnostics, and preventive criticality alerts. In a collaborative work with doctors, we present the design, development, and testing of a healthcare data analytics and communication framework that we call RASPRO (Rapid Active Summarization for effective PROgnosis). The heart of RASPRO is Physician Assist Filters (PAF) that transform unwieldy multi-sensor time series data into summarized patient/disease specific trends in steps of progressive precision as demanded by the doctor for patient's personalized condition at hand and help in identifying and subsequently predictively alerting the onset of critical conditions. The output of PAFs is a clinically useful, yet extremely succinct summary of a patient's medical condition, represented as a motif, which could be sent to remote doctors even over SMS, reducing the need for data bandwidths. We evaluate the clinical validity of these techniques using SVM machine learning models measuring both the predictive power and its ability to classify disease condition. We used more than 16,000 min of patient data (N=70) from the openly available MIMIC II database for conducting these experiments. Furthermore, we also report the clinical utility of the system through doctor feedback from a large super-speciality hospital in India. RESULTS: The results show that the RASPRO motifs perform as well as (and in many cases better than) raw time series data. In addition, we also see improvement in diagnostic performance using optimized sensor severity threshold ranges set using the personalization PAF severity quantizer. CONCLUSION: The RASPRO-PAF system and the associated techniques are found to be useful in many healthcare applications, especially in remote patient monitoring. The personalization, precision, and prevention PAFs presented in the paper successfully shows remarkable performance in satisfying the goals of 3Ps, thereby providing the advantages of three A's: availability, affordability, and accessibility in the global health scenario.

Consensus motifs as adaptive and efficient predictors for acute hypotensive episodes.

Acute hypotensive episodes (AHE) are characterized by continuously low blood pressure for prolonged time, and could be potentially fatal. We present a novel AHE detection system, by first quantizing the blood pressure data into clinically accepted severity ranges and then identifying most frequently occurring blood pressure pattern among these which we call consensus motifs. We apply machine learning techniques (support vector machine) on these consensus motifs. The results show that the use of consensus motifs instead of raw time series data extends the predictability by 45 minutes beyond the 2 hours that is possible using only the raw data, yielding a significant improvement without compromising the clinical accuracy. The system has been implemented as part of a new framework called RASPRO (Rapid Summarization for Effective Prognosis) that we have developed for Wireless Remote Health Monitoring.

Severity Summarization and Just in Time Alert Computation in mHealth Monitoring.

Mobile health is fast evolving into a practical solution to remotely monitor high-risk patients and deliver timely intervention in case of emergencies. Building upon our previous work on a fast and power efficient summarization framework for remote health monitoring applications, called RASPRO (Rapid Alerts Summarization for Effective Prognosis), we have developed a real-time criticality detection technique, which ensures meeting physician defined interventional time. We also present the results from initial testing of this technique.