Calderdale and Huddersfield NHS Foundation Trust is using predictive analytics to accelerate understanding of mortality rates, which will give the trust the data it needs to determine which cases to review in light of death.

The work is being carried out by the Health Informatics Service (THIS), with the organization collecting data from the trust’s Cerner Millennium EPR and other wide-ranging clinical systems and then using the information to populate predictive R analysis software.

The result is that Caldersale and Huddersfield benefit from near-real-time death rates. Prior to this data, the trust would have relied on the national publication of Hospital Standardized Mortality Ratio (HSMR) data. Although the HSMR data is closely monitored by health professionals across the UK, it is delivered three months late.

Initial findings using predictive analytics yielded an 85% accuracy level compared to HSMR.

Oliver Hutchinson, TOVA’s head of performance information for primary care, who led the mortality rate analysis, said: “The benefits are twofold. The first is to pre-empt the national HSMR data by three months so that when the national publication arrives we have a forecast to help dispel any unnecessary concerns.

“The second is that it will help us identify potential clinical areas to improve.” For example, if we have a patient with a low probability of death and they die, we can study that case and do a clinical review to see if there is anything we can learn to prevent death under similar circumstances in the future.

Everyone who arrives at one of Calderdale and Huddersfield NHS Foundation Trust’s hospitals will be assessed and then the expected death rate (EDC) will be calculated. Criteria used for this assessment included the patient’s age, sex, exposure to deprivation, how they were admitted, co-morbidities, diagnosis and whether they had a palliative care code.

Predictive analysis yields a calculation between zero and one – the latter indicating an increased likelihood of death. Calculations can also be expressed as a percentage – for example, an EDC of 0.376 would mean that the patient has a 37.6% chance of dying.

The EDC will help the trust determine which cases may need a post-mortem review. It can also be used to potentially highlight errors in recording a patient’s medical history. If someone who had a low EDC has died, a clinical examination can investigate whether certain factors were missed from their medical records that could have caused a higher EDC if known.

Gawain Carter, from TOVA’s information structure team, said: “The trust has always looked at a good proportion of deaths, but there was no indication of which ones to really focus on to learn from. The difference now is that it helps focus a limited resource. Before that, we didn’t know where to focus until long after the event.”

Earlier in the year, Calderdale and Huddersfield NHS Foundation Trust implemented the Healthcare Gateway medical interoperability portal to give healthcare staff wider access to patient data.

Calderdale and Huddersfield uses predictive analytics for mortality data

Previous articleConsiderations for Digital Therapy Coverage – Health Economist
Next articleAllen Institute for AI Researchers offers PROCTHOR: A Machine Learning Framework for Procedural Generation of Embodied AI Environments