Artificial Intelligence Spirometry

Lung disease is the third biggest cause of death in the UK, and a major reason for winter bed pressures in the NHS. Even before Covid-19, lung conditions cost the UK £9.9 billion each year. The NHS Long Term Plan (LTP), which outlines the nation’s planned health investments, identified lung disease as a national clinical priority. This is because millions have undiagnosed and untreated lung diseases, and others have been misdiagnosed and treated with unnecessary medications, leading to important health inequality.

Spirometry Accuracy

The LTP aims for earlier and more accurate diagnosis of lung conditions. An important diagnostic tool for lung conditions is spirometry (a forceful blowing test). Previous research has shown that spirometry performed in general practice is of poor quality, in particular the interpretation of tests. We have undertaken some work on this in previous years and the findings from this can be found at the foot of this linked page in the ‘Spirometry’ drop down.

The LTP plans to invest money into primary care networks (groups of GP practices working in partnership) to train staff to perform and interpret spirometry at a higher standard. Many thousands of patients are currently on the waiting list for spirometry in primary care.

Artificial Intelligence Spirometry Software

The proposed AI innovation is decision support software (ArtiQ.Spiro) which combines two sub-components – one focusing on quality assessment, and one on spirometry interpretation. The first element will support staff in evaluating immediately if the spirometry is of good enough quality for clinical interpretation. This immediate support allows to immediately add one or multiple trials to the spirometry session when the patient is still present instead of sending of bad quality data onto the GP. Immediately after the spirometry session is performed the interpretation element will provide a description of the results according to the latest international guidelines (which could save staff time and improve consistency). In addition, the interpretation support also uses artificial intelligence to calculate disease probabilities and options for next steps to support the diagnostic process.

Real World Evaluation on AI Spirometry

We worked on a real world evaluation case study to evaluate the experience and efficiencies that this solution may bring to the NHS. Three Primary care GP practices in Sunderland took part in the study:

  • The aim was to include a minimum of 25 patients (>18yrs old) per site, without a current diagnosis of a lung condition.
  • A Nurse with ARTP qualification performed and interpreted along with a GP the Spirometry without AI. (report proforma provided)
  • The AI Spirometry report was then applied to the same patients, a Nurse and GP compared and documented any differences and their confidence in report on the data sheet

We anticipate that supporting primary care spirometry pathways with AI software (ArtiQ.Spiro) will improve the quality of spirometry and accuracy of diagnostic reports, improve patient well-being and experience, reduce dependency on secondary/tertiary care support, improve operational efficiency, address workforce issues and bring cost-savings to the NHS.