Search the hub

This short film shows a fictional scenario of a handover between two healthcare workers. It has been created by Patient Safety Learning to help facilitate a group discussion around bias. Please read the guidance below (and attached) when using this within your teams. How to use this resource: exploring bias in handover This short video is designed to help you recognise how biases can influence clinical handovers and, ultimately, patient safety. It works best as a group learning activity. Step 1: Watch the video Watch the handover between Celia and Doreen all the way through once without interruption. As you watch, think about: What information is emphasised or dismissed. How decisions are explained. Whether anything feels “off” or incomplete. Or: Play “Bias Bingo”. Before watching again, either individually or in small groups, use a simple “bias bingo” card (you can create one using common biases such as confirmation bias, anchoring bias, etc.). Your task: Spot where different biases occur in the handover. Tick them off as you notice them. Note down the exact words or behaviours that suggest the bias. You may spot more than one bias in a single patient discussion. Step 2: Group discussion In small groups, discuss: Which biases did you identify? Did everyone spot the same ones? Where did opinions differ? How might these biases affect patient care or outcomes? Encourage open discussion—there are no “trick answers”. Step 3: Feed Back The group feeds back: One example of a bias they identified. Why they think it is that bias. What the potential risk to the patient could be. Have you seen similar situations in real handovers? What strategies could reduce bias? (e.g. structured handovers, questioning assumptions, using checklists). What would you do differently in Doreen’s position? Share your feedback If you use this resource, we'd love to hear from you. Was it useful? Did anything in the discussion surprise you or spark wider action? Please comment below or get in touch with us at [email protected].

Last month, Public Policy Projects hosted their annual Patient Safety Forum in partnership with Patient Safety Learning. Held at the Royal College of Surgeons of England in London, it was attended by senior healthcare leaders, patient safety experts, representatives from the HealthTech industry, frontline healthcare professionals and patients.  In this article, Patient Safety Learning reflects on one of the panel discussions—AI for patient safety: Innovation, assurance and strengthening communication. From AI-enabled ambient scribing tools that reduce the burden of administration, to predictive systems capable of detecting early warning signs before harm occurs, AI has significant potential to improve patient care and outcomes. Yet, alongside these benefits come risks—algorithmic errors, data bias, and challenges in maintaining trust, governance and oversight. At the Patient Safety Forum 2026 an expert panel was convened to discuss this topic, with the following members: Clive Flashman, Chief Digital Officer, Patient Safety Learning Dr Alison Cave, Chief Safety Officer, Medicines and Healthcare products Regulatory Agency (MHRA) Anil Mistry, AI Safety Lead, Guy’s and St Thomas’ NHS Foundation Trust Dr Basil Bekdash, Clinical Safety Officer, Sheffield Children’s NHS Foundation Trust Aleksander Alski, Head of Region – USA, Canada and UK, Vasco Electronics Panellists had a lively discussion with each other and the audience about how to balance innovation with assurance, to ensure that the use of AI in healthcare enhances safety rather than undermines it. They spoke about how AI should be understood as a support tool for healthcare professionals—it provides information and removes burden but, ultimately, staff treat patients. In this blog we highlight several key topics that emerged from this debate. Importance of patient safety A key theme running throughout the panel’s discussion was the importance of patient safety being built into AI development at the outset. Clive Flashman from Patient Safety Learning reflected on this point, suggesting that too often this is seen as a compliance ‘tick box’ or treated as an afterthought. Speaking to digital innovators, his message was that “you need to think about this from the very start when you are conceptualising the product”. Panellists also recognised that putting safety at the centre of discussions around AI and healthcare means involving all stakeholders, not just the healthcare professionals using these technologies but suppliers too. Alexander Alski from Vasco Electronics emphasised the importance of this being an area of shared responsibility between suppliers and healthcare providers. Getting regulation right Alison Cave from the MHRA spoke about the ongoing work of the National Commission into the Regulation of AI. This Commission was established by the MHRA to review current regulations and provide recommendations for a new regulatory framework for AI in healthcare. It held a public call for evidence which Patient Safety Learning responded to earlier this year. Discussing how to approach future regulation, she highlighted the importance of ensuring that “the risk is associated with the decision, not the technology itself”. It was noted that in some cases there may be very complex pieces of software in use, but these may be making very low-risk decisions. Panellists underlined the importance of having a risk-proportionate regulatory framework to support safe innovation. Predicting future harm The potential to use AI to identify patient safety issues is understandably an area of significant interest. Last year the Department of Health and Social Care announced that it planned to develop a world-first artificial intelligence (AI) early warning system to automatically identify safety concerns across the NHS. Panellists were asked to consider what examples they had seen of AI moving from reacting to incidents, to predicting and preventing future harm. They spoke about the value of AI as a support tool for clinicians and more broadly how it might be used to identify emerging patient safety issues. Basil Bekdash from Sheffield Children’s NHS Foundation Trust spoke about work that had been trialled in this area, but noted that currently there have not been many examples where these have been proven on a significant scale, stating: “None of them have really quite got to the point where they're proven in widespread deployment and so I'm not going to predict that's going to happen in the next five years.” Tackling bias While an AI tool may be safe when properly implemented and used by a well-trained healthcare professional, it could be potentially dangerous if such training and support is absent. Panellists concurred that having appropriate training and tackling bias were issues of critical importance in ensuring the safety of AI in healthcare. In particularly they discussed risks presented by: Confirmation bias—healthcare professionals favouring AI outputs that align with their pre-existing view and overlooking signals that may challenge this. Automation bias—over-reliance on AI systems and accepting their recommendations without sufficient critical evaluation. Alison Cave from the MHRA said that part of the training should be ensuring that healthcare professionals understand the devices they are using and where there are trade-offs between sensitivity and a specificity. Basil Bekdash from Sheffield Children’s NHS Foundation Trust noted the importance of having in mind the different levels of digital competence of staff, stating that when designing AI systems: “It is best to test by using your least capable people who are the least digitally enabled and that's not a criticism that's just the reality of the normal spread of what people do, and their primary function is to look after patients.” Transparency and patient communication As use of AI grows in healthcare, it is vital that patients understand how this is being applied if they are to have confidence in its safety. Panellists discussed issues around how to inform patients when AI influences their care, particularly when it affects clinical judgments. Anil Mistry from Guy’s and St Thomas’ NHS Foundation Trust suggested that: “If the AI result is going to affect their patient’s care, and it's going to limit their access to finite resources like a waiting list or appointments or ICU beds, then absolutely have that sort of communication.” However, he also spoke about some of the challenges this raises; for example, if a patient asked about whether AI has been used in their care. In practice this could cover a very broad range of areas, from the use of ambient scribes to take notes to tools that analyse images from scans. Panellists indicated that transparency needed to be balanced and proportionate to both the risk and impact on individual care. Governance requirements AI healthcare technologies have significant scope to evolve and change over time. When they iterate rapidly (with new versions being released at regular intervals) it can be difficult for existing governance frameworks, designed for other types of medical devices, to keep up. Panellists discussed the importance of having flexibility to governance arrangements. There was the suggestion that lower risks tools (such as those in Class 1 for Medical Devices under the MHRA framework in the UK) should have greater flexibility, with higher levels of scrutiny reserved for decision-influencing tools. It was also made clear that any new regulation will need to carefully consider the level of ongoing evaluation that will be required to account for these systems evolving and changing over time. This may be much longer than for other medical devices and change at significant pace. One audience member commented that with these tools becoming increasingly complex, in the future “realistically there is going to be a need for an AI tool that assesses AI tools”. Panellists also considered how procurement processes could act as potential leverage mechanisms for AI technologies in healthcare. It was noted they offer the potential opportunity to embed the open standards we want to see being used by AI technologies in the earliest stages of their design, putting safety concerns at the centre of the product before it ever reaches patients. Improving the quality of data Data accuracy, completeness and representativeness is key to ensuring AI technologies work safely in health and care environments. Panellists noted that poor foundational data standards undermine AI model training and lead to unreliable outputs. Their discussion reflected that a significant proportion of time is often spent on data cleaning before even applying AI. Improving this would have wider benefits for research, operational efficiency and public healthcare. As we increase the use of AI health technologies, it is vital that we do not embed existing health inequalities. Following on from comments in an earlier session from Professor Bola Owolabi from the Care Quality Commission, Alison Cave from the MHRA noted a “perennial challenge in all of our areas is to ensure that the training data is representative”. Training data for AI systems must be representative of diverse populations and care settings. Sharing insights from the frontline If healthcare organisations, professionals and suppliers are to share responsibility for the safe implementation of AI technologies in healthcare, this must go hand in hand with shared learning. Panellists discussed the need for sustained and transparent feedback loops between suppliers, regulators and healthcare organisations. On this point an audience member asked: “How do we ensure our learning keeps pace so that existing insight from frontline teams that really know the business can optimally inform the evolution of products, but without stifling the pace?” Panellists highlighted the absence of standardised mechanisms for frontline staff to provide real-time, structured feedback to AI suppliers on safety issues. One proposed suggestion to this was the potential to mandate native feedback functionality within AI health technologies. This would mean that feedback mechanisms are built directly into the AI tool’s user interface and workflow, allowing those using them to provide input about the AI’s output without leaving the system. Find out more about the Patient Safety Forum 2026 You can read more about different discussions and panel sessions at this year’s event in the below: Patient voice, safety and the NHS 10 Year Plan: Reflections from the Patient Safety Forum 2026 Safe systems, safe cultures: reflections from the Patient Safety Forum 2026

There are many different types of bias, some more commonly known than others. This resource has been created to help explain different types of bias and to provide some practical examples of how some of these can impact patient safety. The content has been developed following a Patient Safety Education Network session led by Samia Sakuma, lead Quality Governance Lead for Paediatrics at West Hertfordshire Teaching Hospitals NHS Trust. Types of bias and practical examples Anchoring bias – Sticking with your initial impression. Example: "I was right the last time". Aggregate bias –- Assuming evidence from population groups applies equally to an individual patient. Example one: A frailty pathway recommends conservative management for older adults with pneumonia. An individual patient who is usually very active and independent is not considered for escalation early, despite clinical deterioration. Example two: Pain assessment guidance based on average recovery patterns following surgery leads staff to underestimate significant postoperative pain experienced by one patient whose response differs from expected norms. Ascertainment bias – Judgements influenced by prior expectations or contextual information. Example one: A patient known to attend frequently with abdominal pain is initially assessed as having another functional episode, delaying recognition of acute appendicitis. Example two: Documentation describing a patient as “anxious” influences subsequent assessments, resulting in physical symptoms initially being attributed to anxiety rather than investigated further. Availability bias – Where people overestimate the importance or likelihood of events based on how easily examples come to mind. Example: A patient comes in with flu-like symptoms, it must be flu as its flu season. The patient had strep A infection that was unresolved but this was not treated as the flu diagnosis took precedence. Base rate neglect – Ignoring how common or uncommon conditions are when making decisions. Example one: A a rare neurological diagnosis is prioritised in a patient with headache, while more common causes such as medication side effects or dehydration are considered later. Example two: Chest pain in a young adult is assumed to be musculoskeletal without structured assessment, despite cardiac conditions still occurring at a measurable background rate. Commission bias – Preference for action rather than watchful waiting, even when intervention may not help. Example one: antibiotics are prescribed for likely viral infection because active treatment feels safer than observation, exposing the patient to avoidable side effects. Example two: Additional imaging is requested despite low clinical indication, contributing to unnecessary radiation exposure and incidental findings. Confirmation bias/belief bias – the tendency to search for, interpret, favour and recall information in a way that confirms or supports one's prior beliefs or values or decisions. Example: Labelling a child at handover as a ‘drama queen’, thus anything that child does is interpreted through this lens. The child’s abnormal saturations were felt due to her being anxious and hyperventilating, however there was a genuine medical nonanxiety related need for oxygen, the child then had a respiratory arrest. Diagnostic momentum – A diagnostic label becomes accepted and passed along without reassessment. Example one: A patient admitted with a presumed urinary tract infection continues to be treated for this diagnosis despite lack of supporting results, delaying identification of sepsis from another source. Example two: An ambulance handover describing “stroke” leads teams to continue that pathway even after features inconsistent with stroke emerge. Framing effect – Where people’s decisions are influenced more by how information is presented than by the information itself. Example: What order do you present things. The first things you discuss are what stick in peoples minds. The language you use also frames something in a particular way. Calling a follow up protocol “Active surveillance” as opposed to “watchful waiting” can really make a big difference in whether people agree to this or not. Gamblers fallacy – The mistaken belief that past random events can influence the probability of future independent events. Example: sepsis is relatively rare. If you have treated two patients in a row with sepsis, when you see a third patient you don’t believe the sequence can continue so you will go out of your way to find a diagnosis that isn’t sepsis, whereas each patient should be assessed afresh. Over valuing bias/endowment effect – Causes individuals to overvalue what they own, often irrationally. Example: Spending time reading in depth articles on a medical condition such as mesenteric adenitis and reviewing guidance on managing this. Therefore diagnosing patient as having mesenteric adenitis because of the time expended on gathering and reviewing information on this thereby potentially missing another diagnosis. Psych-out error - Physical illness incorrectly attributed to mental health or behavioural causes. Example one: Agitation in a patient with known mental health needs is attributed to psychiatric relapse before delirium secondary to infection is recognised. Example two: Shortness of breath in a patient with anxiety history is initially managed as panic symptoms, delaying diagnosis of pulmonary embolism. Sutton’s slip – Focusing on the most obvious or common explanation without adequate verification. Example one: a patient with recurrent falls is assumed to have mechanical instability, while medication-related hypotension is identified later. Example two: Hyperglycaemia in a person with diabetes is attributed to poor control, delaying recognition of steroid-induced glucose elevation. Visceral bias – Emotional reactions influencing clinical judgement. Example one: Challenging interactions during previous admissions unintentionally influence the urgency of reassessment when the patient re-attends unwell. Example two: A highly likeable patient’s reassurance that they feel “fine” reduces concern despite abnormal observations requiring escalation. Yin–yang out – Belief that a patient has already had extensive assessment, so further evaluation is unlikely to help. Example one: A patient with multiple previous admissions for chest pain receives limited reassessment because earlier investigations were normal, despite new symptoms. Example two: Repeated attendance with headaches leads to reduced diagnostic curiosity when new neurological signs develop. Zebra retreat – Avoiding consideration of rare diagnoses after being discouraged or corrected previously. Example one: After earlier feedback about over-investigating rare conditions, clinicians hesitate to pursue an uncommon metabolic disorder despite suggestive features. Example two: A rare drug reaction is not revisited because previous similar concerns were felt to be unlikely, delaying recognition when it genuinely occurs.

For healthcare to be safe it needs to be accessible. But what does this look like for people with ME (myalgic encephalomyelitis) and Long Covid? This blog from #ThereForME explores the barriers that impact access to NHS care for people with ME and Long Covid, and encourages the patient community to share their experiences. What is ME and why is accessing care difficult? ME (myalgic encephalomyelitis, sometimes referred to as ME/CFS) is a complex, chronic condition affecting multiple body systems.[1] Symptoms include debilitating cognitive dysfunction and post exertional malaise (PEM)—the exacerbation of symptoms following exertion, which can sometimes lead to a long-term deterioration—the cardinal symptom of ME. Patients with ME have one of the worst qualities of life of any disease: lower than various forms of cancer, multiple sclerosis or chronic renal failure.[2] The most severely affected patients are reliant on full-time care, sometimes becoming unable to speak or swallow, and may require hospital care to avoid dehydration and malnutrition. Since 2020 at least two million people in the UK have been affected by Long Covid. Approximately half of those affected meet the criteria for ME (though not all have been formally diagnosed), alongside those who have developed other long-term health issues following Covid infections.[3] For people with ME and Long Covid, accessing healthcare, whether for these or other conditions, can be challenging. PEM means that it can be difficult to receive care without risking a deterioration in symptoms, especially when reasonable adjustments are not made to minimise the exertion involved. A lack of knowledge, misunderstanding and stigma around the conditions exacerbate the issue, sometimes making patients reluctant to seek care and clinicians unlikely to understand the adjustments that are needed. Together, these and other barriers mean that people with ME and Long Covid may avoid, delay or be completely unable to seek the care they need, creating risks for patient safety. Difficulties accessing care at home A 2023 public consultation highlighted failures in the health service that included the accessibility of NHS care for people with ME—particularly for housebound or bedbound patients.[4] This was echoed by a 2024 #ThereForME survey of over 300 people with ME and Long Covid (and their carers).[5] Two-thirds of people responding to our survey said that the NHS had not been there for them when they needed it. The overall accessibility of care was highlighted as a core concern. Housebound patients answering our survey reported struggling to get access to home visits for monitoring and routine screenings or even remote/phone appointments. Patients reported delaying or avoiding seeking care as a result, or in some cases turning to private care as the only option to facilitate routine investigations. Learnings from care for other conditions can show how similar barriers have been addressed—for example, progress in care for people with learning disabilities.[6] Hospital systems and environments People with ME and Long Covid often experience difficulties navigating energy-intensive NHS systems and hospital environments. For many, the process of arranging and receiving medical care may go well beyond their limited energy envelope. This includes challenges like inflexible booking systems, appointments that are changed or cancelled at short notice, long journeys to medical appointments or needing to coordinate with multiple referrals and clinicians. Patients may delay seeking care, even in emergencies, due to the toll that a busy hospital environment is likely to take on their chronic symptoms. Particularly in A&E and inpatient care, busy waiting rooms and hospital wards may exacerbate sensitivity to noise, light and movement. Patients may be unable to sit upright in waiting rooms for long periods of time without their symptoms being exacerbated. While reasonable adjustments are key to accessibility,[7] and the 2021 NICE Guideline for ME/CFS outlines some adjustments that may be needed,[1] knowledge of the Guideline is limited in the NHS and the majority of NHS Trusts and Integrated Care Boards are not implementing it.[8] More widely, limited knowledge about ME, and similarly Long Covid,[9] means that patients don’t receive treatment that is sensitive to their symptoms—and, crucially, that avoids exacerbating them—because clinicians lack basic knowledge. People with ME and Long Covid, who are often particularly vulnerable to infections, may also avoid seeking healthcare due to concerns about acquiring infections. Many people with Long Covid report deterioration after Covid reinfections,[10] as the pandemic continues far from the headlines and with few measures in place to prevent airborne transmission. This may also impact the ability of family carers to access healthcare themselves, fearing acquiring an infection which could set back their loved one’s health. Trauma in healthcare Traumatic experiences in healthcare also play a role. Many patients with ME and Long Covid have experienced feeling dismissed or disbelieved, sometimes discouraging them from seeking care in future. The 2024 #ThereForME survey documented multiple cases of patients who said that, due to such experiences, they would be reluctant to seek NHS care even if experiencing life-threatening symptoms, expressing a sentiment that they would ‘rather die at home’ than seek healthcare in an emergency.[5] ME is significantly more common among women,[11] meaning that experiences of stigma linked to the condition overlap with gendered experiences of healthcare,[12] including how pain among women is routinely dismissed. Sharing your experiences We hope this blog has shone a spotlight on some of the challenges people with ME and Long Covid face when accessing care. If you have ME or Long Covid, or care for someone who does, we’re keen to hear about your experiences: Have there been times where you delayed or were unable to access the care you needed due to these or other challenges? Have you or the person you care for experienced an exacerbation of symptoms due to exertion involved in seeking healthcare? What would make the biggest difference to you to make care more accessible? Do you have any experiences to share where reasonable adjustments were made or a member of staff went out of their way to make it easier for you to access care? You can share your experience by posting in the Comments field below or join our conversation in the Community area of the hub. Related reading Improving healthcare services for people with ME and Long Covid: Patients share their challenges, and the actions needed References NICE. Myalgic encephalomyelitis (or encephalopathy)/chronic794457 fatigue syndrome: diagnosis and management. NICE guideline [NG206], 29 October 2021. Falk Hvidberg M, et al. The Health-Related Quality of Life for Patients with Myalgic Encephalomyelitis / Chronic Fatigue Syndrome (ME/CFS). PLOS One, 2015; https://doi.org/10.1371/journal.pone.0132421. Dehlia MA, Guthridge MA. The persistence of myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) after SARS-CoV-2 infection: A systematic review and meta-analysis. J Infection, 2024. Department of Health and Social Care, Department for Education and Department for Work and Pensions. Consultation outcome. Improving the experiences of people with ME/CFS: interim delivery plan, 9 August 2023. ThereForME. Building an NHS that’s there for Long Covid and ME, July 2024. Anderton M. Exploring deep sedation at home to support people with learning disabilities to access medical investigations with minimal distress. Patient Safety Learning, 17 July 2023. Brar P. Diagnostic safety: accessibility and adaptations–a (un)reasonable adjustment? Patient Safety Learning, 19 September 2024. Action for M.E. Patchy, Misunderstood and Overlooked Implementation of the NICE Guideline [NG206] on Myalgic Encephalomyelitis/ Chronic Fatigue Syndrome in England Freedom of Information Findings Report, May 2023. Patient Safety Learning. Long Covid: Information gaps and the safety implications. Patient Safety Learning, 7 June 2021. WHO. Knocked back by COVID-19 reinfection – the experience of Abbie, a British nurse living with long COVID. World Health Organization, 30 November 2023. DecodeME. Initial findings from the DecodeME questionnaire data published, 24 August 2023. Anonymous. One hour with a women's health expert and finally I felt seen. Patient Safety Learning, 7 November 2024.

Orthopaedic surgeon Sunny Deo has spent three decades diagnosing and treating knee joint issues. In this blog, Sunny argues that the healthcare community needs to take a more nuanced approach to diagnosis and decision making so that it can provide patients with safer, more appropriate treatment options. He reflects on why medicine prefers simple answers and looks at how this affects patient care. He goes on to explore how better data collection and the use of artificial intelligence (AI) could provide a more accurate picture of complexity and allow treatment options to be better tailored to individual patients’ needs. "To know the patient that has the disease is more important than to know the disease that the patient has." William Osler, father of modern medicine, 1849-1919. Diagnosis is the process of identifying the nature of an illness or other problem by examining the symptoms and objective findings from investigations. In modern medicine, it is a key focal point of the assessment and management of all patients. A huge amount of clinical medicine training is focused on the art and science of obtaining a diagnosis, and this focus continues into medical practice. The ease of getting to a diagnosis ranges from the glaringly obvious, the so-called ‘spot diagnosis’, through to cases that are very difficult to solve. In between these extremes there is a range from delayed to missed to incorrect diagnosis. The aim of doctors over the centuries has been to work out diagnoses from patients’ symptoms, presenting features (clinical signs) and, in the past century or so, from the evidence of clinical investigations. Quite often, symptoms, signs and investigations produce consistent patterns, and it is these patterns that are taught to medical and other healthcare professionals. This is how diagnoses and outcomes are portrayed in television series or films—just think back to the last episode of Casualty or Grey’s Anatomy you watched. It's also how things often appear in internet searches and on websites and social media. Seeking simple answers to complex questions However, the reality is different. When a patient is sitting in front of me, what I hear and observe may not exactly be what the textbooks, evidence or research tells me I should be seeing. But because we are wired and trained to recognise patterns, we tend to look for diagnoses and solutions that fit within the well-worn narrative. What if the pattern doesn’t fit the actual diagnosis? There are classic presentations for nearly every condition, and these are what you tend to find at the start of a Google search or when using NHS Choices. The expectation of typical symptoms sometimes means we ignore what we might see as annoying variance, superfluous detail or the patient embellishing the truth. This discordance then causes tension with a very basic trait of humans: when we’re faced with a difficult problem, we still seek the simplest solution. This is an evolutionary feature hardwired into us to optimise survival chances. It means we often believe there is a truth to be found that will provide us with a definite answer. From this answer we will come to the best, and ideally only, ‘correct’ solution. Patients who don’t fit the set patterns of diagnosis may then run into trouble when we offer them what is considered to be the ideal treatment. This is an important problem in clinical thinking, language and practice. As a medical community, we tend to create oversimplified approaches based on research that looks for binary answers to complex questions. This research evidence may be based on a small, highly selective ‘typical’ patient cohort, but its findings and conclusions are then translated on to the entire population. This approach results in poor patient outcomes and experience for a small but significant proportion of patients. Pathways designed for ideal diagnoses can cause harm to patients Over my 30 years as an orthopaedic surgeon, 15 as a knee specialist, I have seen that the assessment and treatment of any given condition isn't quite as predictable as we would like it to be. While many patients fit the pattern we are expecting, some do not. I would empirically put the proportion at 60:40, but some unpublished research we did a decade ago suggested the proportion of truly ‘typical’ case presentations for a common condition is much lower. For example, we found that in the case of suspected meniscal tear, this diagnosis actually applied to only 33% of patients with a variety of other diagnoses accounting for the rest. It gets worse when large organisations start to lump patients into a category by condition in a ‘one diagnosis fits all’ strategy. When this approach is taken, there are winners and losers. The winners are those patients whose condition very closely matches the classic presentation of a given condition in isolation. Let’s take the example of knee osteoarthritis—patients with the ‘right type’ of symptoms, physical signs and x-ray changes are generally more likely to do well. Their recovery is more likely to sit within the knowledge base of treating the condition that has evolved over the past half-century. In contrast, patients whose symptoms and test results fall outside of this category may be less likely to do well or recover in the predicted timeframe. This also applies to patients with additional diagnoses or conditions, often termed comorbidities, which may interact, usually in a bad way, with the condition at hand. Failure to consider other diagnoses, either by over-focus on one condition causing wilful ignorance, inattention or lack of attention, may lead to unexpected poor outcomes from a given treatment. It may also mean that the symptoms from the condition that the patient presents with are worse than expected. This doesn’t mean that they won't gain any benefit from a particular treatment, but the risks and potential outcomes may not be communicated adequately by the patient’s healthcare team, if at all. For example, for patients with painful knee osteoarthritis, the current diagnosis to treatment logic runs like this: Knee osteoarthritis is a painful condition. Total knee replacement surgery is a validated safe procedure with significant improvements in quality of life. Other treatment options do not produce as much positive therapeutic benefit compared to total knee replacement surgery. Therefore, total knee replacement surgery is the only treatment for painful knee osteoarthritis. However, there are patients for whom knee replacement surgery is not a safe or practical option, and these patients may benefit from alternative treatments that are not currently offered as they are seen as providing limited benefit. This may be because the participants in trials undertaken over the years had varying diagnoses, meaning that true comparisons of alternative options may have had additional interacting diagnoses or failed to account for differing severity. Understanding the spectrum of complexity As healthcare professionals, we have a duty to diagnose patients as accurately as possible. In orthopaedics, if treatments go wrong or are poorly undertaken, it may lead to prolonged or permanent pain or disability, and we obviously want to avoid this as much as possible. Incomplete identification and documentation of all relevant symptoms and health conditions can potentially lead to an increased risk of treatment failure and complications. Our priority should be to identify these diagnoses or diagnostic clusters as accurately as possible. I think these are basic principles we need to apply to create better systems and improved care for as many patients as possible. In my view, there are grades of ‘atypical patients’ and I have devoted the past decade to trying to demonstrate this, with surprisingly stiff resistance from peer-reviewed journals and funding organisations. I have tried to move away from lumping all patients into a single category. I have done some research on seemingly straightforward soft tissue problems and osteoarthritis in the knee. My initial analysis suggests that we need to collect more detailed and accurate data, rather than simplifying data into minimum datasets. This is where AI can really come into its own, not as a diagnostic tool initially, but as a powerful aid to unlocking and interpreting some of the diagnostic interactions that create problems for patients. However, the use of AI does need to be undertaken with extreme care and consideration, and this isn’t always happening currently. To offer healthcare that is truly person-centred, we need to look beyond our well-worn simple answers and solutions. By using better data and new machine learning tools to understand the nuances of each person’s condition and how it relates to their wider health, we can offer treatment options that are safer, kinder and more cost-effective. Share your views We would love to hear your views on the issues highlighted in Sunny’s blog Are you a clinician who would like to share your experiences? Do these challenges resonate with you? Or are you a patient who has experienced complications because of poor, missed or inadequate diagnosis? Add your comment below (you will need to be a hub member and signed in) or contact us at [email protected] and we can share your story anonymously. Related content on the hub: Using data to improve decision making and person-centred care in surgery: An interview with Sunny Deo and Matthew Bacon Diagnostic errors and delays: why quality investigations are key Patient safety and the role of AI in a cautiously optimistic future: A blog by Ian Fearnley