Is AI Patient Safety 4.0?

Jonathan B. Perlin

doi:10.1016/j.jcjq.2026.08.007

Is AI Patient Safety 4.0? | ResearchHub

Three decades after To Err Is Human catalyzed the modern patient safety movement, roughly one in four hospitalized patients still experiences an adverse event, and nearly one in four patients who die or deteriorate in U.S. hospitals does so in the setting of a missed or delayed diagnosis. These figures have prompted an appropriately skeptical question: has patient safety made any real progress? This special article argues that the skepticism mistakes an unfinished agenda for a failed one, and that the safety field has evolved through a recognizable sequence of framings. It also argues that artificial intelligence (AI) presents a "step-change" opportunity to accelerate progress in patient safety. Building on the foundational work of Jeffrey Braithwaite and colleagues, we are beginning to evolve from find-and-fix accounting for failure ("Safety 1.0"), to studying why care succeeds ("Safety 2.0"), to a combined, 360-degree view that examines failure and success together ("Safety 3.0"). None of these framings, however, were built to operate at the speed, scale, and complexity of contemporary care, where a single intensive care unit patient can generate more than 1,000 clinically relevant data points a day, which is several orders of magnitude beyond what unaided human cognition can reliably integrate. This article proposes "Safety 4.0": the disciplined use of AI, deployed across a defined autonomy spectrum and governed by explicit guardrails as continuous safety intelligence that senses risk earlier, synthesizes complexity, predicts deterioration, and reliably integrates care needs. Drawing on evidence organized around the National Academy of Medicine's six aims for healthcare (safety, timeliness, effectiveness, efficiency, equity, and patient-centeredness) together with case examples from sepsis surveillance, autonomous diagnostics and prescribing, among others, the article argues that the ultimate test of progress is not whether a system adopts AI, but whether care becomes reliably safe, effective, and compassionate without depending on a patient having an advocate.

External reference records

The Safety of Inpatient Health Care

10.1056/nejmsa2206117 · 2023

Diagnostic Errors in Hospitalized Adults Who Died or Were Transferred to Intensive Care

10.1001/jamainternmed.2023.7347 · 2024

Unresolved referenced work

2000

Changing how we think about healthcare improvement

10.1136/bmj.k2014 · 2018

False Dawns and New Horizons in Patient Safety Research and Practice

10.15171/ijhpm.2017.115 · 2017

Unresolved referenced work

2001

Coordinating Care — A Perilous Journey through the Health Care System

10.1056/nejmhpr0706165 · 2008

High-Reliability Healthcare: Getting There from Here

10.1111/1468-0009.12023 · 2013

When complexity science meets implementation science: a theoretical and empirical analysis of systems change

10.1186/s12916-018-1057-z · 2018

Unresolved referenced work

2014

Unresolved referenced work

Kept as external metadata until matched

The magical number seven, plus or minus two: some limits on our capacity for processing information

10.1037/h0043158 · 1956

The magical number 4 in short-term memory: a reconsideration of mental storage capacity

10.1017/s0140525x01003922 · 2001

Growth in Facts Affecting Provider Decisions Versus Human Cognitive Capacity [presentation]

2007

Three Epochs of Artificial Intelligence in Health Care

10.1001/jama.2023.25057 · 2024

The rise of agentic AI teammates in medicine

10.1016/s0140-6736(25)00202-8 · 2025

The role of agentic artificial intelligence in healthcare: a scoping review

10.1038/s41746-026-02517-5 · 2026

2019 John M. Eisenberg Patient Safety and Quality Awards: SPOTting Sepsis to Save Lives: A Nationwide Computer Algorithm for Early Detection of Sepsis — Innovation in Patient Safety and Quality at the National Level (Eisenberg Award)

2020

Unresolved referenced work

Kept as external metadata until matched

Unresolved referenced work

Kept as external metadata until matched

Unresolved referenced work

Kept as external metadata until matched

Unresolved referenced work

Kept as external metadata until matched

Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices

10.1038/s41746-018-0040-6 · 2018

Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis

10.1016/s1470-2045(23)00298-x · 2023

Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study

2025

Unresolved referenced work

Kept as external metadata until matched

10.1001/jamanetworkopen.2025.34976

10.1001/jamanetworkopen.2025.34976 · 2025

Stochastic programming for outpatient scheduling with flexible inpatient exam accommodation

10.1007/s10729-020-09527-z · 2021

Unresolved referenced work

Kept as external metadata until matched

Autonomous artificial intelligence for diabetic eye disease increases access and health equity in underserved populations

10.1038/s41746-024-01197-3 · 2024

Unresolved referenced work

Kept as external metadata until matched

Unresolved referenced work

2023

Rural-Urban Disparities in Cancer Outcomes: Opportunities for Future Research

10.1093/jnci/djac030 · 2022

Unresolved referenced work

Kept as external metadata until matched

Unresolved referenced work

2025

Evaluating a Nationally Localized AI Chatbot for Personalized Primary Care Guidance: Insights from the HomeDOCtor Deployment in Slovenia

10.3390/healthcare13151843 · 2025

Language proficiency and adverse events in US hospitals: a pilot study

10.1093/intqhc/mzl069 · 2007

Comparison of Waymo Rider-Only Crash Rates by Crash Type to Human Benchmarks at 56.7 Million Miles

10.1080/15389588.2025.2499887 · 2025

Dissecting racial bias in an algorithm used to manage the health of populations

10.1126/science.aax2342 · 2019

External Validation of a Widely Implemented Proprietary Sepsis Prediction Model in Hospitalized Patients

10.1001/jamainternmed.2021.2626 · 2021

Additional external references

No additional external references are available.

Is AI Patient Safety 4.0?

Referenced works

Cited by

External reference records

Additional external references