Digital health and AI
WHO's Six Principles for Artificial Intelligence in Health
WHO's six AI ethics principles become useful when translated into concrete design, procurement, deployment, monitoring, and redress controls.
Health & Evidence Library
Digital health and AI
Clear explanations of clinical software, algorithms, data quality, validation, monitoring, privacy, and safe human oversight.
48 guides · Page 1 of 1
Digital health and AI
When More Health Data Is Not Better Evidence
Large health datasets shrink random error but can preserve bias, weak proxies, leakage, missingness, and drift with impressive precision.
Digital health and AI
When a Health Algorithm Should Say I Don't Know
Safe health AI needs a tested abstention pathway for unfamiliar, incomplete, conflicting, or uncertain cases, plus a reliable human fallback.
Digital health and AI
What WHO Guidance Says About Large Multimodal Models in Health
A practical reading of WHO's guidance on governing generative AI that can process text, images, audio, or video in health settings.
Digital health and AI
What Time in Range Means in Glucose Data
Time in range summarizes continuous glucose monitor readings within a defined interval. Learn the standard metrics, evidence, limits, and safe interpretation.
Digital health and AI
What Makes a Health AI Tool Explainable to a Patient
A patient-facing AI explanation should clarify purpose, inputs, basis, uncertainty, options, accountability, and recourse, not merely display model internals.
Digital health and AI
What Health Economics Asks
Health economics examines scarcity, opportunity cost, incentives, equity, and how costs and outcomes change when health systems choose among alternatives.
Digital health and AI
What Foundation Models Mean for Medicine
Foundation models can be adapted across medical tasks, but broad pretraining does not establish clinical validity, safety, regulation, or patient benefit.
Digital health and AI
What Decision-Curve Analysis Adds
Decision-curve analysis tests whether using a model could improve decisions across plausible thresholds while making assumptions and limits visible.
Digital health and AI
Understanding Overfitting in Clinical Models
Why clinical models overfit, how optimism appears, and which validation practices reveal whether performance will travel beyond development data.
Digital health and AI
Transportability of Clinical Prediction Models
Prediction models can fail elsewhere when populations, measurements, workflows, treatments, and outcome definitions change.
Digital health and AI
Prediction Versus Explanation in Clinical Models
Prediction estimates what is likely to happen; explanation asks why. Clinical models need a question, design, and validation matched to that purpose.
Digital health and AI
Why Digital Health Tools Survive Routine Care
Durable digital health depends on clinical need, evidence, workflow fit, operations, governance, adaptation, and continued measurement.
Digital health and AI
What an Explanation in Medical AI Must Prove
Medical AI explanations need a defined audience, tested fidelity, stability, limits, and evidence that they support appropriate human decisions.
Digital health and AI
What a Decision Threshold Does in a Clinical Model
A decision threshold turns a predicted risk into an action and encodes how a clinical setting weighs missed cases against unnecessary intervention.
Digital health and AI
The Difference Between Validation and Marketing
Validation tests a product claim under defined conditions. Marketing translates evidence into a message that can be accurate, selective, or misleading.
Digital health and AI
The Cost of a False Positive
False positives can trigger repeat tests, procedures, anxiety, labeling, alert burden, and delayed care. Their real cost depends on the full workflow.
Digital health and AI
Reporting Standards for Medical AI
How current reporting guidelines make medical AI claims checkable across prediction, diagnosis, trials, imaging, and early clinical evaluation.
Digital health and AI
Prospective vs Retrospective Validation in Clinical AI: What Each Design Proves
How retrospective, prospective silent-mode, and interventional studies answer different questions about clinical AI performance and value.
Digital health and AI
Predetermined Change Control Plans for AI Medical Devices: A Practical Guide
How a PCCP can authorize bounded future changes to an AI-enabled medical device, and what reviewers should expect from its methods, safeguards, and monitoring.
Digital health and AI
Monitoring Medical AI Performance After Deployment
A practical framework for monitoring medical AI after deployment, including drift, subgroup performance, workflow, updates, safety, and action thresholds.
Digital health and AI
Monitoring a Clinical Model After Deployment
Clinical model monitoring links data, performance, workflow, subgroup, and safety signals to predefined investigation and response paths.
Digital health and AI
How to Read an ROC Curve Without Being Misled by the AUC
How ROC thresholds trade sensitivity for specificity, what AUC measures, and why calibration, validation, prevalence, and utility still matter.
Digital health and AI
How to Read a Calibration Plot
Learn how calibration plots compare predicted risk with observed outcomes, what the diagonal means, and why discrimination alone is not enough.
Digital health and AI
How to Appraise a Clinical Prediction Model
Appraise a clinical prediction model by checking its purpose, data, calibration, discrimination, transportability, and decision value.
Digital health and AI
Classify EU Medical Software by Function and Consequence
Apply EU MDR Rule 11 only after qualification, then use intended purpose, clinical function, decision consequence, and the strictest applicable rule.
Digital health and AI
Why Interoperability Decides Whether Digital Health Works
What interoperability means in healthcare, why it is so hard, and why it often determines whether a digital health tool actually helps clinicians and patients.
Digital health and AI
When a Confident AI Makes a Good Clinician Worse
New experiments show a wrong but confident AI suggestion can drag a trained clinician's reasoning below no help at all. Here is what they measured.
Digital health and AI
What Continuous Glucose Monitoring Data Can and Cannot Tell You
CGM reliably improves glucose control for people on insulin. For most other claims the evidence is thinner. Here is how to read them.
Digital health and AI
What a Hospital's AI Governance Program Is Supposed to Do
What the Joint Commission and CHAI responsible-AI guidance asks hospitals to do, and the questions that reveal whether a program is real or on paper
Digital health and AI
Six Questions to Ask Before You Trust a Healthcare AI Tool
A six-question checklist for judging whether a clinical AI tool is safe to deploy: purpose, evidence, validation, transparency, monitoring, and accountability.
Digital health and AI
Model Cards and Nutrition Labels for Health AI, Explained
What a health AI model card is, and how a voluntary CHAI nutrition label differs from binding FDA device labeling.
Digital health and AI
Insulin Pumps and Closed-Loop Systems: How Automation Steadies Control
A plain-language guide to how insulin pumps and closed-loop artificial pancreas systems work, and why automating small dosing decisions holds glucose steadier.
Digital health and AI
Human in the Loop: Designing Clinical AI That Supports Judgment
Why clinical AI works best when a trained person keeps the authority to review, override, and own every decision before it reaches a patient.
Digital health and AI
How to Judge Whether an AI That Writes Clinical Notes Is Any Good
How to evaluate ambient AI scribe tools by omission, fabrication, certainty drift, and note bloat, using randomized and scoping evidence, not time saved.
Digital health and AI
How to Evaluate a Symptom Checker
How to judge an AI symptom checker: what it is for, how it treats uncertainty, and whether it points you toward care when it matters.
Digital health and AI
How Regulators Think About Generative-AI Mental-Health Chatbots
How FDA and its advisers weigh therapy-style AI chatbots, and why hallucination, sycophancy, and automation bias shape the oversight debate
Digital health and AI
How AI Is Changing Drug Discovery, and Where the Hype Outruns the Evidence
AI speeds up the front of the drug pipeline, targets, molecules, and property prediction. Here is where those gains stop and why human trials still decide.
Digital health and AI
Good Machine Learning Practice: 10 Principles for Trustworthy Medical AI
A plain explainer of the FDA, Health Canada, and MHRA joint Good Machine Learning Practice principles for medical AI.
Digital health and AI
External Validation: Why a Model Must Prove Itself Outside Its Training Data
Why clinical AI usually loses accuracy at new sites and in later years, and how to tell a real external test from a reassuring-sounding one.
Digital health and AI
Does a Clinical Algorithm Work for Every Patient Group? Three Ways to Check
How to check a clinical algorithm serves every patient group, using representative data, subgroup testing, and post-launch monitoring.
Digital health and AI
Decision support in the diabetes visit: what actually helps
How clinical decision support works inside a real diabetes visit, what the trial evidence shows, and how to tell a useful tool from a demo.
Digital health and AI
Data Quality for Clinical AI: The Model Is Only as Good as What It Learned From
Clinical AI rises or falls on its data. What trustworthy clinical data looks like, and why the flaws stay invisible until the tool reaches real patients.
Digital health and AI
Calibration vs Accuracy in Plain Terms, and Why Calibration Is the One That Keeps You Safe
Accuracy tells you how often a model is right. Calibration tells you whether its probabilities are true. Acting on a number needs calibration.
Digital health and AI
Your health records and privacy: a calm primer
What happens to your health data, the basic rights you have over it, and a sensible, non-anxious way to think about health privacy.
Digital health and AI
What good diabetes technology looks like
A plain, brand-neutral guide to the main diabetes technologies, what they do, and how to judge them by evidence and fit rather than marketing.
Digital health and AI
AI in medicine: separating progress from hype
Where the evidence for artificial intelligence in healthcare is genuinely strong, where it is still thin, and how to read claims about it honestly.
Digital health and AI
Clinical decision support: what it is and what works
What clinical decision support is, what the evidence shows, where alert fatigue bites, and why these tools support rather than replace a clinician's judgment.