Role of AI in reducing medical error – Result of YouDiagnose Survey

Sep 24, 2020 | Clinical AI

Whether it is the case of a miscalculated dose of anaesthetics that killed a patient or an instance of wrong side surgery leading to limb loss on both sides, medical error is emerging as a serious public threat. We carried out a mixed input survey – comprising of a telephonic survey, face to face interview and web-based survey – to study people’s perception about safety in the healthcare industry and compared it with other industries. Out of 123 invitees, 93 participants took part in the survey in 2018.

Observation – 1 To a question, “Which industry is most hazardous?” most participants rated the nuclear industry as the most hazardous one. 69% of the participants labelled the nuclear industry as the most dangerous, followed by aviation (21%) and healthcare (10%).

Infographic showing survey results on the most hazardous industry: 69% chose Nuclear Industry, 21% chose Aviation, and 10% chose Healthcare, represented in three colorful donut charts.
  • facebook
  • reddit
  • LinkedIn
  • Pinterest

Observation – 2 When people were asked about the reason behind what they think, most people mentioned that nuclear explosion could wipe out the human race from the face of the Earth (~16%) or other such statements about the potential harm and damage from a possible nuclear explosion.

Observation – 3 Repetitive Work Driving Medical Error – One of the critical aspects which emerged as a startling observation was repetitive tasks, which increasing error proneness. Three participants mentioned that when someone is mindlessly repeating the tasks, the brain often switches off assuming full control over the situation. This may often lead to inaccurate decision-making. All participants mentioned that no-brainer work, when given in high volume, may lead to sluggish performance and vulnerability to more errors.

Observation – 4 How AI can help? – With regards to finding a solution to error-making, a large percentage of the people favoured the use of augmented intelligence or artificial intelligence in patient care to assist and support in the decision-making process. The joint model of working offers a human validation to a machine-led decision and machine validation to a human decision, thereby creating a man-machine synergy.

Observation – 5 Survey Results on AI’s potential role in medical error – 46% of the participants had a strong feeling that futuristic technologies will play a decisive role in solving the problems of medical error, 24% participants said they neither favour nor disfavour futuristic technologies and the remaining 30% of participants thought AI might play a decisive role.

Risk Perception Vs Ground Reality

However, ground realities are far away from people’s perception. The below-mentioned figures show that the most hazardous industry is the healthcare industry.

  • Aviation – 1 in a million chance of a person being harmed
  • Nuclear – 400 deaths per 20,000 years of reactor operation
  • Healthcare – 250,000 deaths annually in the USA
  • facebook
  • reddit
  • LinkedIn
  • Pinterest

Most of the respondents who had chosen differently were utterly surprised when they heard the real figures, which are as follows:

  • facebook
  • reddit
  • LinkedIn
  • Pinterest

Incidence of Medical Error

  • 750,000 harm-inflicting medical errors per year in the EU
  • 3.2 million days of hospitalisation due to medical error in EU
  • 260,000 fewer incidents of permanent disability in the EU
  • 95,000 fewer deaths per year in EU
  • 850,000 adverse events in a year in the UK
  • 10% of hospital admissions in the UK were medical errors and health-care related adverse events
  • WHO says that there is 1 in 300 chance of a patient being harmed when receiving healthcare.

Patient Survey on Medical Error Experience

  • 23% of European Union citizens claim to have been directly affected by a medical error
  • 18% claim to have experienced a serious medical error in a hospital and 11% claim to have been prescribed a wrong medication
  • 50% to 70.2% of such harm is preventable.

What are the ways to reduce medical errors?

Attention is the most valuable asset an individual has. In today’s complex and demanding healthcare environment, time is no longer an organisation’s or an individual’s most valuable asset; it’s attention.

In non-medical professions, attention is defined as what we each choose to focus on and what we decide to ignore. However, in a healthcare organisation, that is certainly not a huge possibility, given the sensitivity of the environment. Mental triaging of the circumstances and tasks is dependent on one’s experience and environmental factors. Thus, reducing medical error and maintaining patient safety standards needs a tremendous amount of attention which is always in limited supply.

Categories of Medical Errors

A circular infographic shows 8 categories of medical errors, including executive errors, knowledge gaps, lack of training, systemic issues, personal stress, equipment issues, work pressure, and intentional error, each described briefly.
  • facebook
  • reddit
  • LinkedIn
  • Pinterest

For simplification, we have categorised medical errors into the following groups:

  1. Intentional Error – A person wilfully commits a mistake due to exceptional circumstances.
  2. Human Errors
    • Executive Errors – Unintentional Errors, e.g. slips or lapses in memory or action
    • Mistakes
      • Gap in knowledge
      • Lack of Consistent Exposure or Training
      • Systemic Issues – a gap in rota, high patient waiting time, communication issues, lack of induction
      • Personal Stress – Illness or stress
      • Equipment Issues – Faulty instrument or inadequate instructions or training to operate
      • Heavy Work-Load

Root Cause Analysis of Medical Error

Every step involved in patient care requires active attention and decision-making, which reduces in a subtle way when the tasks are repeated, and the work is overwhelming. The medical errors are attributed to the following causes:

  1. Intensity of work (10%)
  2. Multiple parallel tasks (7%)
  3. Chronic staff deficit and a consequent need to cover the job of other people (14%).
  4. Gap in knowledge (8%)
  5. Repetitive work (15%)
  6. Chain reaction and cascade effect (7%)
  7. Career burn-out (12%)
  8. Lack of communication or coordination(5%)
  9. Technical failure of devices or systems(4%)
  10. Human error of judgement(18%)

When we analyse the responses, we find that the work-pressure to human ratio contributes about 31%, while the human error of judgement was the second most common cause (18%). Out of the rest, repetitive work and workplace frustration, for example, career burnout were the third and fourth most common cause behind medical errors.

How do we aim to reduce the incidence of medical errors?

The medication error was one of the critical issues affecting patient safety in nursing practices. The medication administration error was significantly reduced when the “double-checking of medication” practice was introduced. In this practice, two humans cross-check the medication content, expiry date and route of administration before administration. This is a routine practice in the National Health Services of the United Kingdom. If this practice can be adopted in patient care and decision making, it has a vast potential to reduce medication error.

Testimonials

Dr Arti Garg, Consultant Surgeon, Bart’s Health NHS Trust, says,

“Double-checking is a good practice to reduce the incidence of human errors in patient care. This is the main reason behind the cancer care MDT (multi-disciplinary team decision making) practice in cancer care in UK hospitals. But there is a paucity of doctors. NHS is already struggling with a staff shortage and therefore, getting additional human resources for double-checking is not possible.

The shortage of trained healthcare professionals is a global issue. The World Health Organisation predicts, “The global economy is projected to create around 40 million new health sector jobs by 2030, mostly in middle-income and high-income countries. Despite the anticipated growth, there will be a projected shortage of 18 million health workers needed to achieve the UN Sustainable Development Goals (SDGs) in low-income and lower middle-income countries, fuelled in part by labour mobility, both within and between nations.”

Prof Bijen Patel, Academic Surgeon from Queen Mary University and University College Hospital, London, says,

“It takes 15-20 years to produce a fully qualified and experienced consultant. We can’t resolve the huge staff shortage issue at this pace. The only way we can move forward is to make efficient use of technologies to develop a synergy between man and machine. This will not only address the issue of chronic shortage of healthcare staff but will also improve the healthcare practice standards by reducing medical errors.”

YouDiagnose and Medical Error

YouDiagnose is the world’s first artificial intelligent cancer-care solution which will provide a review of cancer diagnosis and treatment process, and thus, “double-check” the decision-making process. This will validate the decision taken, offer new options and alternatives, empower the patient, improve patient compliance and reduce the incidence of medical errors.

The Ada Paradox: Why a 5-Star Symptom Checker Can’t Fix the NHS

If you’ve ever searched online for health advice, you’ve likely encountered the dizzying, often terrifying, world of medical misinformation. It’s a landscape dominated by frantic Googling, questionable forum posts, and ads vying for your attention. Into this chaos steps Ada, a sophisticated AI-powered symptom checker app. On the surface, it’s a resounding success. With millions of downloads and a stellar 4.7–4.8★ rating across app stores, users consistently sing its praises. It represents a triumph of digital health design.

Yet, if we zoom out from the individual user experience to the system it claims to serve—the UK’s National Health Service—a puzzling paradox emerges. Despite its popularity, Ada’s presence has been virtually invisible in the metrics that matter most to the NHS. The Referral to Treatment (RTT) waiting list continues its relentless climb into the millions. A&E waiting times remain a political flashpoint. The chronic pressures of workforce shortages and underfunded services persist, seemingly untouched by this technological marvel.

This begs a central, critical question: If users love it so much, why hasn’t Ada moved the needle on the NHS’s most profound problems?

To solve this puzzle, we must first understand what Ada gets right, then explore the vast chasm between a beloved consumer app and a transformative healthcare system tool.

What Ada Gets Right for Patients

For the individual seeking answers, Ada is a revelation. Its strengths are specific, meaningful, and directly address the pain points of navigating modern healthcare.

Perceived Accuracy and Reassurance

Users frequently report that Ada’s top suggestion “matched my eventual diagnosis” or was “spot on.” There are powerful anecdotes of the AI picking up on rare conditions like herpangina or seizure risks that were initially missed by time-pressed clinicians. This perceived accuracy builds a crucial foundation of trust. Unlike the chaotic rabbit hole of a search engine, Ada provides a structured, calm line of questioning that mimics a clinical consultation, replacing anxiety with a sense of order and process.

Triage Guidance in an Overwhelmed System

Perhaps its most valued function is helping users decide if and where to seek care. For an anxious parent up at 3 a.m. with a sick child, or someone waiting weeks for a GP appointment, Ada acts as a digital arbiter. It can offer reassurance that symptoms may be self-limiting or provide the validation needed to seek urgent care. This “digital front door” function is a lifeline for those feeling lost in a complex system.

Superior User Experience

Ada is free, ad-free, and allows for multiple profiles, making it a true household health tool. Its assessments are fast, intuitive, and allow for follow-up questions. This seamless UX is a benchmark in an industry where clunky, outdated interfaces are often the norm.

What Users LikeRepresentative User Sentiment

What Users Like Representative User Sentiment
Accuracy “Ada’s top suggestion matched my eventual diagnosis.”
Triage Guidance “Helped me see when I should and shouldn’t go to the doctor.”
UX & Access “Lifesaver when I couldn’t get an appointment.”

 

Where Users Hit the Limits

However, even its biggest fans encounter the app’s seams. Ada’s model, built on a vast library of clinical knowledge, struggles with complexity. Users with multiple pre-existing conditions (multi-morbidities), those on complex medication regimens (polypharmacy), or individuals managing chronic mental health conditions like bipolar disorder often find the tool under-representative of their reality. Their health cannot be easily parsed into a linear Q&A format.

Furthermore, for unusual or nuanced presentations, Ada can hit a coverage gap, resulting in an inconclusive output. While its frequent disclaimer—“this is not a diagnosis”—is a necessary and responsible boundary for safety, it can leave users wanting more firm, actionable recommendations. Occasional friction points, like sign-up prompts or confusion over its free model, remind users they are still interacting with a product, not an integrated part of their care.

The NHS Problem Ada Says It Wants to Help With

This is the context that makes Ada’s potential so tantalizing. The NHS is in a state of perpetual crisis. The RTT backlog numbers in the millions. Four-hour A&E wait targets are routinely breached. A chronic shortage of GPs and specialists means the front door to care is often bolted shut. The promise of a “digital front door” that can triage patients efficiently, direct them to the right care faster, and alleviate pressure on clinicians sounds not just attractive, but essential.

On paper, a highly rated digital triage tool is exactly what the NHS needs. In practice, it hasn’t played out that way. The reason lies not in the quality of the app itself, but in a fundamental mismatch between consumer success and system-level adoption.

Why Ada Hasn’t “Excelled” in the UK

High app ratings do not equal system-level impact. Ada’s journey in the UK is a masterclass in this difficult truth, stemming from four core challenges.

1. Consumer App Success is Not Healthcare Integration

Ada’s primary strategy focused on winning the hearts and minds of patients directly. They chased downloads, ratings, and engagement—the metrics of a successful Silicon Valley-style app. But NHS problems don’t live on patients’ phones; they live in clinician workflows, referral systems (like the e-Referral Service), radiology networks (RIS/PACS), and capacity planning meetings. Without deep, seamless integration into these existing, often archaic, digital infrastructures, a symptom checker remains “advice on the side.” It’s a PDF printout a patient brings to a GP, not a structured data point that flows into a referral form. It is peripheral, not central.

2. Misreading NHS Procurement and Politics

The NHS is not a single entity but a complex, relationship-driven, and risk-averse ecosystem. Ada’s leadership has publicly expressed frustration with how NHS deals “come out of nowhere,” wishing for transparent “bake-offs.” While technically reasonable, this view is politically naive. Adoption in the NHS is rarely won on pure technical merit. It is won by aligning with massive, pre-existing programmes like Electronic Patient Record (EPR) rollouts or national platforms, and by navigating byzantine procurement processes dominated by large, entrenched incumbents. A standalone app, no matter how good, is an island in this ocean.

3. Regulatory and Data-Model Drag

Ada’s development path was EU-centric, aligned with the CE/MDR regulatory framework. Brexit created a dual regulatory nightmare, forcing compliance with both EU MDR and the UK’s new UKCA/MHRA rules. This complexity drained resources and made deep investment in the UK market harder and less appealing.

Simultaneously, sources suggest Ada’s underlying data model is large and monolithic, making it difficult to change quickly. This is a critical weakness in the NHS, where clinical pathways, guidelines, and local formularies can shift with a new health directive or hospital policy. An app that can’t adapt at the speed of the NHS is left behind.

4. No Closed Loop with Real-World Outcomes

From an NHS commissioner’s perspective, the question is not “Do users like it?” but “Does it reduce avoidable referrals, improve patient safety, or shorten pathway times?” To answer this, you need a closed loop: the app’s advice must be connected to the actual patient outcome. There is a stark absence of public, peer-reviewed evidence showing Ada has delivered at-scale impact on core NHS metrics like RTT lists or A&E wait times. Without this evidence, it remains an interesting pilot project, not a essential system tool.

LevelWhat Ada OffersWhat NHS NeedsGapPatientReassurance, self-triage adviceBetter access, understandingPartial matchClinicianPDF-style summariesStructured data in workflow, risk-sharingWeak integrationSystemNice app metricsRTT, capacity, equity improvementsNo evidence at scale

 

Level What Ada Offers What NHS Needs Gap
Patient Reassurance, self-triage advice Better access, understanding Partial match
Clinician PDF-style summaries Structured data in workflow, risk-sharing Weak integration
System Nice app metrics RTT, capacity, equity improvements No evidence at scale

Lessons from the Data: A Strong Patient Tool, A Weak System Lever

The conclusion from this analysis is clear: Ada is a powerful tool for individuals but a weak lever for system change.

For the anxious patient, the underserved individual, or the person waiting in limbo, Ada can be useful, reassuring, and sometimes genuinely life-saving. But this strength operates almost exclusively at the micro level—helping one person navigate their health on a given night.

The NHS’s most crushing problems—backlogs, pathway variation, mis-triage—are meso and system-level issues. They are not solved by millions of isolated individual actions, but by changing the structural flow of information, accountability, and patients through the system itself.

What Would a Tool That Could Move the Needle Look Like?

To genuinely impact NHS metrics, a decision-support tool must be fundamentally different. It must be a hospital-grade decision layer, not a consumer-facing app. This means:

  • Embedding Hospital Pathway Logic: It must be co-designed with specialist hospital teams, encoding their specific “hospital brain” and local pathways, not a one-size-fits-all global model.
  • Deep Integration: It must plug directly into GP clinical systems, populate referral forms, interface with booking systems and diagnostics, becoming an invisible part of the clinician’s workflow.
  • Closed-Loop Evaluation: It must be built from the ground up to answer the question: “Did this change the outcome?” It must prove it leads to fewer bounced referrals, a better mix of urgent cases, and shorter overall pathways.
  • Shared Accountability: It must move from being “an app that said so” to a shared, evidence-based, and auditable tool under clinical governance. It shares the risk and the responsibility.

Conclusion: Helpful for Individuals, Insufficient for Systems

The Ada paradox is ultimately a story about scale. We must celebrate digital health tools that provide genuine comfort and utility to individuals. Ada can, without a doubt, improve someone’s Tuesday night. But we must also be clear-eyed about their limitations. It cannot, on its own, fix a 7-million-patient waiting list.

The next wave of digital health innovation must learn this lesson. The goal cannot be to build better apps on the edge of the system. The imperative is to build smarter, integrated, and accountable brains that can be embedded directly into the heart of our clinical pathways. The future of healthcare tech isn’t a better front door; it’s a smarter, connected house.

Pin It on Pinterest

Share This

Share This

Share this post with your friends!

Share This

Share this post with your friends!