Just out todayAugmented reality: The AR trainees were faster on the model they trained on, and no better anywhere elseVirtual reality: They matched the exercise dose exactly, and the VR advantage fell below the line that mattersVirtual reality: Both groups used the same headset. One watched a video first and picked up 10.8 more blocks

Cybersecurity/United States

316 patient harms from hospital cyberattacks, and nobody can tell you how often they happen

A systematic review of hospital cyberattacks found 316 patient harms, two distinct at-risk groups, and no way to say how often any of it happens.

What the paper found

A PRISMA systematic review of 59 studies published 2004 to 2026, combined with a Python pipeline mining 3,408 healthcare-related subreddits, identified 316 patient-level harms during hospital cyberattacks. Two at-risk categories emerged: patients harmed by time-criticality, and patients harmed by digital dependence with rapid decline. The resulting CIPHER dataset is open source.

Grade 13 reading level6 min read

What happened

Cyberattacks on healthcare organisations have changed character. They used to be about stealing records. Increasingly they are about stopping care, and the authors of this review describe the shift as moving from data theft to intentional denial of care.

What that does to patients has been reported case by case, hospital by hospital, with no systematic account of who gets hurt or how.

A team at the Department of Computer Science and Engineering at the University of California San Diego, the Department of Emergency Medicine at George Washington University, and the Department of Emergency Medicine at UC San Diego ran the review that was missing – and then went somewhere the literature does not reach.

The test

The systematic review followed PRISMA 2020, PRISMA-S and SWiM reporting guidelines, searching MEDLINE, Embase, PubMed, Scopus and Web of Science for peer-reviewed studies published between 1 January 2004 and 1 July 2026 describing cyberattacks affecting clinical care in healthcare delivery organisations.

Quantitative synthesis was not performed, because the included studies reported structurally incompatible outcome types for which no common effect measure existed. Synthesis without meta-analysis was conducted instead, following SWiM.

In parallel, a Python pipeline mined 3,408 healthcare-related subreddits for cyberattack terms extracted during the academic review. Posts were reviewed by clinical team members to ensure relevance and coded to clinical specialties and technical failure domains.

Using social media alongside the literature is a deliberate response to a known problem: hospitals have legal and commercial reasons not to publish what happened to their patients.

The numbers

Element Detail
Studies included 59
Search period 1 January 2004 to 1 July 2026
Databases searched MEDLINE, Embase, PubMed, Scopus, Web of Science
Subreddits mined 3,408
Patient-level harms identified 316
Quantitative synthesis not possible – incompatible outcomes
At-risk category 1 time-criticality: stroke, cardiac arrest, major trauma
At-risk category 2 digital dependence with rapid decline: cancer, diabetes, community prescribing
Output open-source CIPHER dataset

The result

Fifty-nine studies yielded 316 patient-level harms across the combined datasets.

Two distinct categories of at-risk patients emerged. The first is defined by time-criticality: patients with stroke, cardiac arrest and major trauma, for whom delays of minutes to hours are clinically decisive. The second is defined by the intersection of digital dependence and potential for rapid clinical decline: patients with cancer, patients with diabetes, and those dependent on community prescribing.

Specific technical failures were linked to distinct clinical consequences. Hardware failures produced oncological harms through compromised linear accelerators. Software failures created risks for patients with fractures through failures in digital templating software. Network and connectivity failures produced fetal deaths through telemetry failures.

Social media data exposed harms the literature did not capture: compromised door access systems preventing life-saving treatment, failures in ward-based alarms, closed-circuit television affecting staff and patient safety, and disrupted health-at-home services preventing community therapies.

The resulting open-source CIPHER dataset links technical failure domains to downstream clinical risks across care settings.

How this compares to what came before

Healthcare cybersecurity literature is substantial and mostly operational: how to prevent attacks, how to recover, what the regulatory requirements are. Patient harm appears as a justification rather than an object of study.

Individual incident reports exist and are valuable – including one in this same batch, a national audit showing hip fracture surgery within 48 hours falling from 75.3 to 69.5 percent after Ireland’s 2021 attack. Those give precision for one event and one pathway.

What this review adds is structure. Instead of a list of bad things that happened, it produces a mapping from technical failure domain to clinical consequence, and an identification of which patients are most exposed. That is the form in which the finding becomes usable for preparedness planning.

The social media component is methodologically unusual and defensible given the reporting problem. It is also where several entire categories of harm came from.

What it means

The two at-risk categories are directly actionable. A hospital planning for a cyber incident now has a defined set of patients to plan around: the time-critical ones who need minutes, and the digitally dependent ones who decline without their systems.

The failure-to-harm mapping is equally practical. Knowing that linear accelerator compromise produces oncological harm, or that telemetry failure has been associated with fetal deaths, tells a risk assessment where to concentrate.

The honest limitation runs through everything. 316 harms is a catalogue, not a rate. We still do not know how often any of this happens, because the denominator does not exist. Hospitals do not systematically report clinical harm from cyber incidents, and the legal and commercial incentives run the other way.

That is a policy problem. No amount of better research design fixes a reporting gap, and the authors are clear that the boundary between cybersecurity and patient safety has become blurred in a way that demands unified frameworks.

Business ideas from this paper

Cyber incident clinical preparedness planning. The two at-risk categories give a hospital a defined planning target. Translating that into downtime procedures for specific pathways is consulting work with a clear scope.

Failure-domain risk mapping. The CIPHER dataset links technical failures to clinical consequences. Applying that map to a specific hospital’s technology inventory produces a prioritised risk register.

Harm reporting infrastructure. The missing denominator is a reporting problem. Building the mechanism – anonymised, legally safe, clinically coded – would serve regulators and enable the research that cannot currently be done.

Resilience for non-clinical infrastructure. The social media data surfaced door access, alarms and CCTV as harm pathways. Those sit outside most healthcare IT risk assessments entirely.

What could be wrong with it

No quantitative synthesis was possible, so the 316 harms are a catalogue with no incidence attached. The review cannot tell you how likely any of this is.

Social media posts are unverified accounts by definition. The authors use them as a complement rather than evidence of incidence, which is the right treatment, and they remain self-reported narratives.

Publication of cyberattack harms is suppressed by legal and commercial incentives, which biases the base of 59 studies in ways that cannot be corrected for.

The review spans 2004 to 2026, across which healthcare technology has changed enormously. Harms from a 2008 attack and a 2025 attack are not comparable events.

Reddit skews towards English-speaking, North American and younger users, so the social media component has its own demographic bias.

How sure can you be?

We score this 10 out of 10 – the second article on Just Out Tech to reach the top of our scale.

It is a peer-reviewed systematic review in the Journal of Medical Internet Research, following PRISMA 2020, PRISMA-S and SWiM guidelines across five databases, synthesising 59 independent studies, with named institutions at UC San Diego and George Washington University, and an open-source dataset released as output.

A perfect score on our rubric does not mean the findings are certain. It means that on everything we can check – peer review, named institutions, evidence type, open materials and independent checking – nothing is missing. The review’s central limitation, that it cannot produce a rate, is a property of the underlying evidence base rather than a flaw in the work, and the authors state it clearly.

Do this today

If you run a hospital, identify your time-critical pathways and your digitally dependent patient groups. Those are the two populations this review says will be harmed first, and the list is shorter than a general risk assessment.

If you do cyber risk assessment in healthcare, look at the non-clinical systems. Door access, ward alarms and CCTV appeared as harm pathways here and are usually out of scope.

If you work in health policy, the missing denominator is the finding. Mandatory clinical-harm reporting after cyber incidents would make the next version of this review quantitative, and nothing else will.

Source: Unveiling Patient-Level Harms and At-Risk Clinical Groups During Hospital Cyberattacks: Systematic Review of Global Case Studies and Social Media Data, Journal of Medical Internet Research, September 2026. PMID 42809354 · doi.org ().

Just Out Tech explains new research in plain language. This article was drafted with AI assistance and checked by a human against the original source.

What to remember
  • No quantitative synthesis was possible - the 59 studies reported structurally incompatible outcomes.
  • Two at-risk groups: time-critical patients (stroke, cardiac arrest, major trauma) and digitally dependent patients with rapid decline potential (cancer, diabetes, community prescribing).
  • Specific failures mapped to specific harms: compromised linear accelerators to oncological harm, telemetry failures to fetal deaths.
  • Social media revealed harms absent from the literature: door access systems, ward alarms, home care disruption.
Where this really is

The review followed PRISMA 2020, PRISMA-S and SWiM guidelines across five databases, covering studies from 2004 to 2026 describing cyberattacks affecting clinical care.

Questions people ask

Why no meta-analysis?

The included studies reported structurally incompatible outcome types with no common effect measure, so the authors used synthesis without meta-analysis following SWiM guidelines rather than forcing a pooled estimate.

Why use social media data?

Because legal and commercial incentives suppress hospital reporting of cyberattack harms. A Python pipeline mined 3,408 healthcare-related subreddits, and clinical team members reviewed and coded the posts.

Which patients are most at risk?

Two groups. Those for whom delays of minutes to hours are clinically decisive - stroke, cardiac arrest, major trauma. And those whose digital dependence intersects with potential for rapid decline - cancer, diabetes, community prescribing.

What is CIPHER?

The Cyberattack Impacts, Patient Harms and Emergency Response dataset, produced by this review and released open source, linking technical failure domains to downstream clinical risks.

About the author

Mark Alex

Mark Alex is the founder and Managing Director of Real Biz Digital, a technology company operating out of Nairobi since 2018. He works in agentic AI and the Model Context Protocol, AI governance, enterprise software architecture and cybersecurity. He holds an MSc in Mechatronical Engineering from Obuda University in Budapest and a BSc in IT, Forensic Technology and Cybercrime, from USIU-Africa in Nairobi, and has published IEEE conference research on an AI-powered digital twin for greenhouse systems. He is the author of seven books. Between 2020 and 2024 he mentored more than 200 university students and interns in Nairobi. He writes every Just Out Tech article from the original research paper.