Focus on Care Delivery Problems to Improve Productivity with AI

Hospital + Healthcare Magazine · an hour ago · 11 views

Healthcare has spent years digitising clinical work without necessarily making it easier. New technologies have transformed what medicine can do, but the process of delivering care has often become more complex. Clinicians experience that complexity as more data entry, more screens, more clicks — and less attention for the person in front of them. Artificial intelligence (AI), however, bucks this trend. Applied well, AI can reduce administrative work, make relevant clinical information easier to find and release capacity within a constrained workforce. Those gains may improve throughput and lower the unit cost of care. But this is not inevitable. Making one task faster does not automatically make the whole system more productive. The workflow around it must change too. I’ve worked with a number of hospitals to implement AI-powered Electronic Health Record (EHR) systems that embed technologies such as ambient listening and AI search into every clinical workflow. We are already seeing streamlined data entry, faster access to clinical information, and less time wasted on navigating systems. This frees up clinicians to focus on what matters most, providing excellent patient care. I think it’s inevitable that every healthcare provider will go down a similar path. So, let me share a few observations to help you get there, and perhaps avoid a few pitfalls along the way. What care delivery problems are we trying to solve? I am enthusiastic about AI and its potential, but we also need to see it for what it is: another tool we can use to help clinicians. Too often, healthcare providers begin by asking, “What AI tool should we deploy?” rather than, “What care delivery problem are we trying to solve?” That order matters. The problem should determine the tool, not the other way around. In my experience, the choice is less about which AI model or algorithm to use or which tool performs best, and more about whether the health service can use it safely and whether it’s built on a trusted data foundation. We already have many technology solutions fragmented across our clinical workflows. Clinicians tell me we just want to get on with it. We don’t want something we’ll have to log into repeatedly. We want a tool that fits within our workflow, so we don’t have to look for it or use multiple applications. Streamlining clinical work by reducing data entry One problem at the top of every clinician’s and care provider’s list is reducing the burden of using clinical information systems. Increasingly, we expect clinical staff to enter data because the information captured is necessary or useful to the provider or the healthcare system. At first, we captured information to support clinical care. But increasingly, we have asked clinicians to enter data for clinical coding, reimbursement, statistics or statutory reporting. With AI, clinical work can become less about data entry and searching through fragmented information, and more about surfacing relevant context when clinicians need it. For example, AI can produce a draft clinical note summarising a patient encounter and help automate additional data capture, such as coding. The objective, however, is not simply to generate more documentation more quickly. The test is whether the technology removes work overall. If clinicians spend as much time checking and correcting AI-generated content as they previously spent creating it, the productivity gain disappears. Recovering the patient’s story from all the noise Another obvious care delivery problem is how to help clinicians recover the patient’s story from large volumes of health data — ideally inside their clinical workflow so that they can make better decisions with a smaller cognitive burden. When doctors first assess a patient, we develop a differential diagnosis: a list of possible explanations that we test and refine as more information becomes available. We learnt at medical school that if you take the patient history properly, you will probably arrive at the right diagnosis about eight times out of ten even before examining a patient or ordering further investigations. But in recent years the patient’s story has become much harder to see. It’s typically fragmented across overlapping sources, including referrals, clinical notes, imaging reports, medication lists, discharge summaries, and handover notes. Clinicians also commonly review and accept patient messages via a patient portal. AI is a powerful tool for unscrambling relevant information hidden among the noise and making the patient’s story accessible in an instant. Considering the human element of AI solutions I have seen many pilots involving exciting AI tools, but not all of them go on to solve care delivery problems at scale. The successful ones are often led by people who ask: Can we govern and scale this beyond the enthusiastic, technology-literate clinicians who volunteered for the pilot? Would it still work for a busy clinician who did not participate in the design process, has had limited training and is using it under real-world pressure? Again, we need to think less about the AI model or algorithm — in fact, less about the AI itself — and more about the clinical workflow, the conditions in which care is delivered and the people expected to use it. Technology suppliers must also consider the human element. For example, with InterSystems IntelliCare™, we created an EHR solution with built-in AI and interoperability capabilities, embedding intelligence directly within the electronic health record system and the clinical workflows it supports. A unified AI solution — as opposed to multiple bolt-on AI tools — is more consistent and easier to use, and eliminates the workflow friction that often drives clinician resistance. Unified governance also simplifies oversight and enables consistent application of AI policies. And AI algorithms have direct access to comprehensive patient data, enabling accurate insights and recommendations. Achieving significant productivity improvements We still have a long way to go with AI. Beyond these early use cases, however, there are many other opportunities to improve the delivery of care. For example, AI can help identify patients at risk of deterioration and support clinicians in deciding who needs attention first. I also expect AI to help health services understand which patients are receiving best-practice care and where people may have deviated from an agreed clinical pathway. Hospitals will increasingly use AI to optimise patient flow, manage capacity and improve resource allocation. Over time, agentic AI may also perform a growing number of bounded administrative tasks, with appropriate oversight and escalation when human judgement is required. But these capabilities do not automatically translate into productivity. A few minutes saved on an individual task will only create additional capacity if the surrounding workflow changes too. Health services need to decide what work they expect AI to remove, how they will use the released time, and how they will measure the result. That time might allow clinicians to see more patients, reduce waiting times, spend longer with complex patients or finish their work on time. These are all valuable outcomes, but they are not the same outcome. Right now, AI can enable meaningful productivity improvements if we focus on common care delivery problems, integrate it into clinical workflows and consider the human element from the outset. The measure of success should not be how much AI a health service deploys, but whether it removes unnecessary work, releases useful capacity and makes care easier to deliver. Disclaimer Any AI tool or AI functionality provided by InterSystems® is subject to regulatory and clinical safety requirements and is not made fully available to all global markets. Please consult the InterSystems AI Ethics webpage for more information on the company’s approach to Responsible AI and your InterSystems representative for any specific details on jurisdictional availability. About the Author Dr Emeline Ramos is a Physician Executive at InterSystems, a creative data technology provider managing over one billion healthcare records globally. With a medical career spanning over 20 years, Emeline has served as a Surgeon for the UK National Health Service, a Clinical Researcher for AstraZeneca EU, and an Information Management Clinical and Research Advisor for the Victorian Comprehensive Cancer Centre in Australia. Top image credit: iStock.com/zeljkosantrac