Health Innovation East’s Post

Its International #clinicaltrialsday ! The BHF Data Science Centre have a number of projects aiming to support researchers to use electronic health records in clinical trials: https://lnkd.in/ep4tDheQ https://lnkd.in/e8rBSc4A Please get in touch if you would like to talk to us about: ✔ Accessing health systems data for cardiovascular or diabetes trials ✔ Developing a data utility comparison (DUCk) – a form of SWAT (‘Study within a trial’) that would allow assessment of healthcare systems data to traditional data collection ✔ Support materials

Dynamic risk prediction in LVAD patients holds immense promise—but are we truly advancing innovation, or simply refining the status quo? In my latest Letter to the Editor in JACC: Heart Failure, I build on the work by Shah et al. by addressing critical methodological refinements necessary for real-world impact. If these models are to guide clinical care, they must evolve beyond statistical benchmarks to deliver actionable insights. Key areas of focus: 1. Metrics Beyond AUC: Rare outcomes like stroke require precision-recall curves and clinically relevant thresholds to ensure models provide meaningful insights—not just statistical significance. 2. Time-Stratified Calibration: Event rates and risk factors change over time. Without robust calibration tools like time-stratified plots, we risk overestimating reliability in dynamic clinical environments. 3. Practical Decision Tools: Risk stratification must align with the dynamic nature of LVAD care. Decision curve analysis is critical for defining thresholds that maximize clinical utility. While the authors acknowledged the complexities of stroke prediction and perioperative clustering, this exchange underscores the need for continued refinement. Predictive models must not only identify risks but also meaningfully guide clinical decisions. Precision medicine demands no less. This conversation reflects the core of innovation in healthcare: rigorous questioning, collaborative improvement, and a commitment to patient-centered outcomes. Grateful to my co-author, Gavin Hickey, and the teams at UPMC, PITTSBURGH VA HEALTHCARE SYSTEMS, and the AI-HEART Lab for driving these discussions forward. Special thanks to Palak Shah, and their team for initiating this critical dialogue. Let’s ensure every step forward in predictive modeling meets the highest standards of rigor and relevance. Read our full exchange here https://lnkd.in/e2iYYV_A The future of LVAD care depends on bold thinking paired with real-world precision. #JACCHeartFailure #LVAD #PrecisionMedicine #RiskPrediction #PersonalizedMedicine

Thank you, Sir Muhammad Irfan Dr. Sheraz Naseer - (PhD Artificial Intelligence, Data Science) Muhammad Haris Tariq, for enabling me to learn about the importance of blood pressure variability and the role of the interquartile range (IQR) in healthcare research and practice. Your insights have provided valuable knowledge that will undoubtedly contribute to my understanding of cardiovascular health and patient care. Leveraging the Interquartile Range to Understand Blood Pressure Variability In clinical research and healthcare practice, understanding the variability of blood pressure measurements is crucial for assessing cardiovascular health and guiding treatment strategies. One statistical tool that aids in this understanding is the interquartile range (IQR). What is the Interquartile Range (IQR)? The IQR represents the spread of blood pressure measurements within a population or study sample. By calculating the IQR, healthcare professionals can gain insights into the range of blood pressure values and identify outliers that may require further investigation. Why Does Blood Pressure Variability Matter? Blood pressure variability is influenced by various factors such as age, gender, lifestyle, and underlying health conditions. Monitoring changes in blood pressure variability over time can provide valuable insights into cardiovascular health and help healthcare professionals tailor interventions to individual patient needs. How is the IQR Used in Blood Pressure Research? Describing Variability: The IQR describes the spread of blood pressure values within a sample. A larger IQR indicates greater variability, while a smaller IQR suggests more consistent readings. Identifying Outliers: Outliers in blood pressure measurements may indicate extreme values that could skew data analysis. Healthcare professionals use the IQR to identify and investigate outliers. Assessing Treatment Effects: Researchers use the IQR to assess the effectiveness of interventions aimed at managing blood pressure. Changes in the IQR before and after treatment can indicate treatment efficacy. Conclusion By leveraging the power of the interquartile range (IQR) to understand blood pressure variability, healthcare professionals can better assess cardiovascular health, identify areas for intervention, and improve patient outcomes. #xevensolutions #BloodPressure #CardiovascularHealth #HealthcareResearch #InterquartileRange #IQR #HealthcareProfessionals #LinkedInDiscussion

The Sequential Organ Failure Assessment (SOFA) Score is a staple of clinical research and machine learning as proxy of illness severity among those who are hospitalized. It is used to adjust for confounding by indication and to match patients according to propensity for treatment typically at the time of target emulation trial enrollment. In a previous study, we demonstrated that the SOFA score does not translate to the same risk of death across race-ethnicity. In this paper, we show that the contribution of the different SOFA components to the likelihood of death changes over time. The use of the SOFA score to (1) adjust for illness severity at the time of treatment to estimate its causal effect, or (2) capture illness severity at different patient states for machine learning, needs rethinking. https://lnkd.in/e2umHHD2

A Year of Journal Articles (Day 296/365) Kingsley, Viveck, Lisa Fox, David Simm, Glen P. Martin, Wendy Thompson, and Muhammad Faisal. "External validation of the computer aided risk scoring system in predicting in-hospital mortality following emergency medical admissions." International Journal of Medical Informatics (2024): 105497. Summary: Goal: Validate a previously developed CARSS model for predicting in-hospital mortality after emergency admissions. Methods: - Analyzed adult emergency admissions (2020-2022) from The Rotherham Foundation Trust (UK). - Evaluated model performance using c-statistic (discrimination) and calibration metrics. Results: - Included 20,422 admissions (out of 32,774). TRFT sample had similar demographics but higher mortality (6.1% vs 5.7%). CARSS model showed: - Good discrimination (c-statistic 0.87). - Good calibration after re-calibration for baseline mortality differences. Conclusion: - CARSS model is valid after adjusting for baseline risk. - Initially under-predicted mortality, but re-calibration improved performance.

📢 We’ve teamed up with NIHR (National Institute for Health and Care Research)’s OPTIMising Therapies programme to discover therapeutic targets and AI-assisted clinical management for patients living with complex multimorbidity.   🏥 As part of this project, we've curated a detailed dataset tracking the emergency healthcare needs of over 40,000 adult patients acutely admitted to hospital with an inpatient stay.   🔬 Are you a researcher doing important research into complex multimorbidities?   Our longitudinal data includes serial physiology readings, frailty scores, blood results, medications, comorbidities, and more.   💡 Discover more in our dataset on Health Data Research UK (HDR UK)’s Gateway: https://rb.gy/5pu37m   #HealthData #PIONEERdataset #ResearchCollaboration #OPTIMAL #NIHR #DataScience #Symptoms #Inpatient #Dataset #AI #Multimorbidity #Chronicconditions #LongtermHealthConditions #Polypharmacy #Comorbidity #Acuity #Disease 

𝗣𝗿𝗲𝘀𝗲𝗻𝘁𝗲𝗱 "𝗔 𝗛𝘆𝗯𝗿𝗶𝗱 𝗠𝗼𝗱𝗲𝗹 𝗳𝗼𝗿 𝗗𝗶𝗮𝗴𝗻𝗼𝘀𝗶𝘀 𝗼𝗳 𝗖𝗮𝗿𝗱𝗶𝗼𝘃𝗮𝘀𝗰𝘂𝗹𝗮𝗿 𝗗𝗶𝘀𝗲𝗮𝘀𝗲 𝘂𝘀𝗶𝗻𝗴 𝗖𝗹𝗶𝗻𝗶𝗰𝗮𝗹 𝗙𝗲𝗮𝘁𝘂𝗿𝗲𝘀, 𝗘𝗖𝗚, 𝗮𝗻𝗱 𝗠𝗥𝗜" at R𝗮𝗼 𝗕𝗮𝗵𝗮𝗱𝘂𝗿 𝗬. 𝗠𝗮𝗵𝗮𝗯𝗮𝗹𝗲𝘀𝘄𝗮𝗿𝗮𝗽𝗽𝗮 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗖𝗼𝗹𝗹𝗲𝗴𝗲 during the 𝗜𝗘𝗘𝗘 𝗙𝗶𝗿𝘀𝘁 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗖𝗼𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 on Ambient Intelligence, 𝗡𝗼𝘃𝗲𝗺𝗯𝗲𝗿 𝟮𝗻𝗱-𝟯𝗿𝗱, 𝟮𝟬𝟮𝟯. This innovative research integrates multiple data sources for accurate diagnosis, marking a significant advancement in cardiovascular healthcare. #IEEE #AmbientIntelligence #HealthcareInnovation #ResearchPresentation #post #paper #linkedin

Your weekly selection of #articles from the #ICMJournal: 🔹Systemic inflammation and delirium during critical illness 👉 https://rdcu.be/dFBvJ 🔹Noninvasive neuromonitoring in acute brain injured patients 👉 https://rdcu.be/dFBuR 🔹Serial lactate measurements to guide resuscitation: more evidence not to? 👉 https://rdcu.be/dFBvk 🔹OPEN ACCESS ~ Federated data access and federated learning: improved data sharing, AI model development, and learning in intensive care 👉 https://rdcu.be/dFBvt 🔹Less inappropriate medication: first steps in medication optimization to improve post-intensive care patient recovery 👉https://rdcu.be/dFBvy 🔹Visualizing ICP “Dose” of neurological critical care patients 👉 https://rdcu.be/dFBuU 🔹Mild hypocapnia and outcomes in mechanically ventilated acute brain-injured patients: another piece in the puzzle 👉 https://rdcu.be/dFBuV 🔹Relationship between the arterial partial pressure of carbon dioxide and outcomes in mechanically ventilated acute brain injured patients 👉 https://rdcu.be/dFBuW 🔹Concern for meta-analysis combining randomized parallel and cross-over trials 👉 https://rdcu.be/dFBuZ 🔹Positive or negative pressure: plus ça change, plus c'est la même chose 👉 https://rdcu.be/dFBvA 🔹OPEN ACCESS ~ Non-occlusive mesenteric ischemia: the wolf in sheep’s clothing 👉 https://rdcu.be/dFBuo Read more articles here! 👉 https://lnkd.in/eeBieiy

Exciting News! Our latest publication, “Interpretable data-driven approach based on feature selection methods and GAN-based models for cardiovascular risk prediction in diabetic patients,” presents an innovative machine learning (ML) model designed to predict 10-year CVD risk in older individuals with type 1 diabetes (T1D). Using data from the Steno Diabetes Center Copenhagen, we explored various ML models, including KNN, decision tree, random forest, and multilayer perceptron (MLP). The use of CTGAN-generated synthetic data significantly enhanced model performance, while feature selection techniques highlighted critical risk factors. The MLP model excelled with a mean absolute error of 0.0088. Key risk factors? Age, HbA1c, and albuminuria. This groundbreaking study paves the way for early intervention and targeted treatments to prevent CVDs in T1D patients. Discover more and download the full issue here. 🔗 https://lnkd.in/eiFdsduZ Norwegian Centre for E-health Research CiaoTech - Gruppo PNO INNOVATION PLACE Consiglio Nazionale delle Ricerche Universidad de Las Palmas de Gran Canaria UiT Norges arktiske universitet Fundación Canaria de Investigación y Salud y Servicio de Evaluación del Servicio Canario de la Salud Evaluation Unit of the Canary Islands Health Service Canary Islands Health Research Institute Foundation Universitetet i Oslo (UiO) Munster Technological University NetSun Software Universidad Rey Juan Carlos Sensotrend #health #ehealth #chronicconditions #artificialintelligence

The evidence underlying the use of advanced #DiagnosticImaging is based mainly on diagnostic accuracy studies and not on well-designed trials demonstrating improved patient outcomes. This has led to an expansion of low-value and potentially harmful patient care and raises ethical issues around the widespread implementation of tests with incompletely known benefits and harms. Randomized clinical trials are needed to support the safety and effectiveness of imaging tests and should be required for clearance of most new technologies. Large, diverse cohort studies are needed to quantify disease risk associated with many imaging findings, especially incidental findings, to enable evidence-based management. The responsibility to minimize the use of tests with unknown or low value requires engagement of clinicians, medical societies, and the public.    Read “Types of Evidence Needed to Assess the Clinical Value of Diagnostic Imaging,” the latest review in the Diagnostic Evidence Review series, by Carly Stewart, MHA, Matthew S. Davenport, MD, Diana L. Miglioretti, PhD, and Rebecca Smith-Bindman, MD: https://eviden.cc/3VtuZGk    #MedicalResearch