BRENNEN

Industries / Health / Healthcare / Telehealth

Specialist help at a distance

A clinician or care team near the patient gets help from specialists or nurses far away: live by video in emergencies such as stroke, by camera in intensive care and on wards, or later from images and written questions sent ahead.

7

Milestones achieved, from 1950 to 2011.

3

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1950

    X-rays sent by telephone

    J. Gershon-Cohen and A. G. Cooley described "telognosis", diagnosis from X-ray images sent over telephone lines, in the journal Radiology in 1950. A 1952 report by Gershon-Cohen and colleagues in the Journal of the American Medical Association covered "three years of experience" with it.

    europepmc.orgeuropepmc.org

    Credit

    J. Gershon-Cohen, A. G. Cooley and colleagues

  2. 2005

    Stroke experts on call

    Telestroke means a stroke specialist examines a patient by video and reviews the brain scan from far away. REACH, a web-based telestroke system built at the Medical College of Georgia, gave 8 rural community hospitals in Georgia stroke consults 24 hours a day, 7 days a week. In its first report, 194 consults led to 30 patients receiving tPA, a clot-dissolving drug, with no symptomatic brain bleeds, and the time from stroke onset to treatment fell from 143 minutes for the first 10 patients to 111 minutes for the last 20.

    europepmc.org

    Credit

    D. C. Hess, R. J. Adams and colleagues, Medical College of Georgia

  3. 2006

    Stroke network in Bavaria

    The TEMPiS project linked five community hospitals without specialized stroke care to two academic stroke centers by telemedicine. In a non-randomized comparison with five matched hospitals, covering 3,122 patients admitted from July 2003 to March 2005, 44% of patients treated in the network had a poor outcome at three months (death, institutional care or disability), against 54% in the comparison hospitals.

    europepmc.org

    Credit

    H. J. Audebert, R. L. Haberl and the TEMPiS group

  4. 2008

    Video beats the phone

    In the STRokE DOC trial, 222 patients with stroke symptoms at four remote hospitals were randomly assigned to a stroke specialist consult by two-way video with scan review, or by telephone. The decision on whether to give clot-dissolving drugs was correct in 98% of video consults and 82% of phone consults, by central review; outcomes at 90 days did not differ.

    europepmc.org

    Credit

    B. C. Meyer, P. D. Lyden and colleagues

  5. 2008

    Eye photos read by graders

    From 2003, and across all of England by 2008, the NHS has offered everyone with diabetes aged 12 and over a yearly photograph of the back of each eye; screeners take the photos and trained graders, about 1,500 of them, read them on screen. This is store-and-forward care, in which images are captured in one place and read later by someone else. In 2015–16 the program offered screening to 2.59 million people and screened 2.14 million. A 2014 analysis found that diabetic eye disease was no longer the leading cause of certified blindness in working-age adults in England and Wales, for the first time in at least five decades (14.4% of certifications in 2009–10, against 17.7% in 1999–2000); its authors name the screening programs as one possible reason.

    pmc.ncbi.nlm.nih.goveuropepmc.org

    Credit

    the English NHS Diabetic Eye Screening Programme, described by P. H. Scanlon; G. Liew, M. Michaelides and C. Bunce (blindness data)

  6. 2009

    Tele-ICU without overall gains

    In a tele-ICU, intensivists, doctors who specialize in intensive care, watch patients in several intensive care units from a central office over cameras and data links. A 2004 before-and-after study, written by staff of VISICU, the company that sold the eICU system, reported hospital mortality of 9.4% with remote coverage against 12.9% before. An outside study of 6 ICUs in 5 hospitals, published in JAMA in 2009, found no overall improvement in mortality or length of stay once patients' severity of illness was taken into account; sicker patients did better and less sick patients did not. A 2011 study at UMass Memorial, where the tele-ICU came with redesigned care routines, reported lower adjusted mortality (odds ratio 0.40) and shorter hospital stays.

    Lessonthe same remote setup gave different results in different hospitals, and none of these three studies was randomized.

    europepmc.orgeuropepmc.orgeuropepmc.org

    Credit

    M. J. Breslow and colleagues (VISICU); E. J. Thomas and colleagues (University of Texas Health Science Center at Houston); C. M. Lilly and colleagues (University of Massachusetts)

  7. 2011

    Rural doctors treat hepatitis C

    Project ECHO (Extension for Community Healthcare Outcomes) used video conferences with specialists to train primary care clinicians at 21 rural and prison sites in New Mexico to treat hepatitis C. In a study of 407 patients, 58.2% of those treated at ECHO sites had a sustained viral response, the study's measure of successful treatment, against 57.5% at the University of New Mexico's specialist clinic.

    europepmc.org

    Credit

    S. Arora and colleagues, University of New Mexico

Watch for

  1. Tele-ICU in safety-net hospitals

    Chidinma Chima-Melton, founder and chief executive of Tele-ICU Inc., and colleagues reported that after a 24-hour, intensivist-led tele-ICU program started at four Los Angeles safety-net hospitals with no intensivists on site, hospital stays fell by 1.7 days (geometric mean 8.8 to 7.1 days) across 6,386 ICU admissions, while ICU stays and ventilator use did not change and the drop in mortality across the network was not conclusive.

    Achieved whena peer-reviewed paper, or an evaluation by a group with no stake in the program, confirms the shorter stays against comparison hospitals.

    healio.com

    Claimed by

    Chima-Melton and colleagues, at the CHEST Annual Meeting, October 2026

  2. Virtual nurses for rural hospitals

    Guthrie Clinic, a six-hospital system, says its Pulse Center, where remote nurses watch and support patients by camera alongside bedside staff, cut pressure injuries by 20.3%, and that remote "tele-sitting", in which one person watches several patients at risk of falling, cut major-injury falls by 87.5%.

    Achieved whenan outside evaluator publishes before-and-after or controlled comparisons with methods and patient counts.

    hcinnovationgroup.com

    Claimed by

    Guthrie Clinic's chief nurse executive Debra Raupers, in Healthcare Innovation, May 2026

  3. AI answers to specialist questions

    NOHARM is a public benchmark of 1,100 tasks built from consultation cases sent by primary care clinicians to specialists, in 10 specialties, with 12,747 expert annotations of how harmful each of 4,249 management options would be. Its authors report that following the advice of 20 language models and 4 retrieval-based clinical AI tools directly carried potential for severe harm in up to 24.6% of cases, more than 80% of it from omissions, and that in a randomized study of 101 US generalist physicians, AI help improved their answers, though they often left out good AI suggestions.

    Achieved whena study inside a working consult service, run or audited by people outside the team, shows AI-assisted answers to real consult questions are at least as safe as specialists' own.

    arxiv.org

    Claimed by

    David Wu and 56 coauthors, on arXiv, December 2025, revised July 2026

Open

  1. 01

    Stroke exams over weak links

    A test that finds how poor a video link can get before a remote neurologist's stroke scoring stops matching the bedside. Actors trained to show stroke signs are examined with the NIH Stroke Scale, the standard bedside score of stroke severity, by a rater in the room and by certified remote raters over a video call whose bandwidth, delay and packet loss are set by a network emulator.

    The brief
    Credit · unclaimed
  2. 02

    Test virtual sitters with actors

    A public test of electronic sitters, the camera systems that let one remote observer, sometimes helped by software that spots movement, watch several hospital patients at risk of falling. In a furnished mock hospital room, actors follow scripts: sitting on the bed's edge, getting out of bed, sliding to a crash mat, pulling at a tube, and harmless movements such as a visitor leaning over the bed. Record whether each system or its remote observer flags each risky event, how many seconds before the actor stands, and how many false alerts come per hour.

    The brief
    Credit · unclaimed

Open 01 · the brief

Stroke exams over weak links

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A test that finds how poor a video link can get before a remote neurologist's stroke scoring stops matching the bedside. Actors trained to show stroke signs are examined with the NIH Stroke Scale, the standard bedside score of stroke severity, by a rater in the room and by certified remote raters over a video call whose bandwidth, delay and packet loss are set by a network emulator.

Why it is still open

Studies since 1999 show that remote and bedside stroke-scale scores agree over good links, and a 2016 study of video from moving ambulances (iTREAT) rated audio and video quality on a 6-point scale; its authors wrote that work on jitter and packet loss "may determine minimum performance standards". Searches found no published bandwidth or packet-loss limit tied to scoring agreement.

pmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.gov

Built with what exists

Linux netem to set bandwidth, delay and packet loss; an open-source WebRTC video app such as Jitsi Meet; actors trained from written deficit scripts; raters holding current NIH Stroke Scale certification; a speech-to-text model to log each exam (see the Note "Fewer word errors").

man7.org

  1. First step

    Write 10 deficit scripts, train two actors, and run them with two raters over a clean link and one degraded link, such as 500 kilobits per second with 2% packet loss.

  2. Done looks like

    20 scripted cases, each scored at 6 network settings by 4 remote raters; weighted kappa (a measure of agreement beyond chance) between remote and bedside scores for each item and the total; the lowest bandwidth and highest packet loss at which the remote total stays within 2 points of the bedside total in 95% of exams; scripts, settings and scores published.

Open 02 · the brief

Test virtual sitters with actors

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public test of electronic sitters, the camera systems that let one remote observer, sometimes helped by software that spots movement, watch several hospital patients at risk of falling. In a furnished mock hospital room, actors follow scripts: sitting on the bed's edge, getting out of bed, sliding to a crash mat, pulling at a tube, and harmless movements such as a visitor leaning over the bed. Record whether each system or its remote observer flags each risky event, how many seconds before the actor stands, and how many false alerts come per hour.

Why it is still open

A scoping review of 25 studies, published on 29 October 2026, found that most used uncontrolled before-and-after designs, only two had a comparison group, at least five had a financial tie between authors or sponsors and the vendor, and that the detection figures came from "few evaluations using heterogeneous metrics". A search found no independent test with staged events in a mock room.

researchonline.jcu.edu.au

Built with what exists

the systems the review names, such as Ocuvera, AvaSure, Oxehealth and Artisight, under trial or with the maker's permission; a hospital bed and crash mats in a nursing school's simulation lab; volunteer actors; two reference cameras to record what happened; a vision-language model such as Gemini or Claude as an open baseline that marks the same clips.

  1. First step

    Write 20 scripts with a nurse educator and run them in a nursing school's simulation lab against one system on trial.

  2. Done looks like

    400 scripted events, half risky and half harmless, across 3 room layouts in day and night lighting; for each system, the share of risky events flagged within 30 seconds with 95% confidence intervals, the median warning time before the actor stands, and false alerts per hour over 24 hours of harmless activity; scripts and labeled reference video published.