How to get patients flowing: 12 uses of AI to unblock hospitals 

There is a great deal of focus given to the use of explainable AI in diagnostic applications, supporting clinicians in improving both the accuracy of diagnosis and in the speed to arrive at a diagnosis. A recent article titled ‘The AI will see you now’ published in the Wall Street Journal highlights the power of these developments. 

Whilst these incredible transformations continue to revolutionise the delivery of care, there is an often overlooked – but critical – capability in healthcare which focuses on ensuring that capacity remains available for patients as they arrive at the hospital.  

Maintaining available capacity in our healthcare systems is a constant battle. Every day, healthcare systems work to manage fluctuating demand for their services, and support complex care journeys that are by their nature, difficult to predict. Every day, there is a potential risk that tasks related to a patient’s care and subsequent discharge may be delayed due to urgent priority cases or activities. This can result in a patient taking longer to get home than they should, and a bed being occupied for longer than it needs to be. 

Studies have shown that hospitals with occupancy rates over 85% run the risk of access block, with a ‘delay in supplying inpatient beds inevitable’ at this level of occupancy. (Bagust el al, 1999) 

Getting this balance of demand and supply right for healthcare operations can be challenging to achieve on a case-by-case basis, however, this problem is perfectly suited to AI, where big data can be used to effectively predict demand for services by analysing large amounts of historical data. 

The team at Alcidion continue to work closely with our customers to identify patient flow challenges where AI can be applied to support clinicians and operational staff. Below, we have outlined a selection of examples that we are working on today: 

  • Emergency Department presentations – we have predictive models to support managers of Emergency Departments in understanding expected demand and therefore, what resources may be needed to meet this demand and manage performance. 
  • Unplanned admissions – we have developed a predictive model for unplanned admissions including admission type 
  • Hospital occupancy and access block – We have live capability to monitor and predict access block, alerting management and operational teams ahead of time so that action can be taken to avoid or minimise peak capacity levels. 
  • Readmission risk – a score given to each patient on their likelihood to readmit, based on behaviours such as patient history, adherence and comorbidities. 
  • Avoidable admissions – analysis and predictions on avoidable demand 
  • Hospital Acquired Complications likelihood – a score given to each patient on HAC likelihood in order to support care teams as they produce care plans. 
  • Patient deterioration – alerting for clinicians on patient deterioration risk 
  • Personalised care plan recommendations – recommended care plan options for clinicians to use for patients  
  • Elective waitlist prioritisation – identification of recommended actions that can be taken to reduce waitlist duration and financial impact 
  • Remote monitoring patient selection – identification of patients likely to respond well to virtual care and remote patient monitoring programs 
  • Bed management and outlier reduction – support for proactive bed management based on incoming demand and optimisation of outliers to minimise length of stay (LOS) impact 
  • Workforce and rostering requirements – predictive views of forward demand for workforce requirements 

In each of these examples, AI identifies patterns in patient data that may be too subtle for humans to recognise, allowing care teams to make more effective decisions that benefit patient care.  

However, while AI has the potential to transform healthcare, it cannot do so alone. AI algorithms are only as effective as the data they are trained on, and in healthcare, that data is generated by other systems and care teams. Typically, either a care team member or a monitoring device will collect patient data, then a clinician will make the diagnoses, and develop treatment plans, and their expertise is essential for training AI algorithms to support them effectively. 

Partnering with care teams helps to ensure that AI is used in a way that aligns with clinical best practice and does not compromise patient safety. Clinicians can help identify which tasks are best suited for AI application and ensure that the algorithms used are accurate and reliable. 

With the help of AI, healthcare organisations can release capacity and resources for care teams to focus on patient care. 

Nick White is Director of Analytics & Insights at Alcidion.