Skip to main content

Chris Peter Subbe

Betsi Cadwaladr University Health Board


Can I go home? Will I be ok?

We finally have an answer using routine NHS data that is available today at every bedside in Welsh hospitals.

Discharge decisions are complex but involve the diagnosis of ‘stability’ prior to transfer from an acute hospital to the community, be it the patients’ own home or a rehabilitation facility.

Defining in-stability:

The National Early Warning Score (NEWS) is generated from vital signs that are being collected by nurses in hospitalised patients. NEWS is being used throughout all hospitals in Wales as a tool to identify patients at risk of catastrophic deterioration on general wards. Patients with higher scores have a higher risk to die, to suffer cardiac arrests, to be admitted to Critical Care or to require a prolonged hospital stay

How about the reverse?

If high scores identify patients who require high levels of care in hospital, can low scores identify patients who are likely to be able to leave hospital safely? After what period of time with a low value of NEWS is the risk to suffer adverse events so low that allowing patients to return to their own home or in other community settings would represent a more patient centred and cost-effective way to care?

The National Early Warning Score (NEWS):

Planning and development:

Good NEWS 4 Home: Machine learning at bedside: As part of a collaboration with Philips Healthcare Betsi Cadwaladr University Health Board has got clinical areas which record electronically all NEWS scores and basic patient characteristics. These provide the setting for a high-impact low-cost intervention.

We used Random Forest Analysis, a form of machine learning, to review data from a sub-group of 1,451 patients to generate a stability indicator that takes values from -1 (very unstable) to +1 (very stable). Mean age of patients was 69 years (+/- 18). 866 patients achieved a stability indicator value of +1. Of these 473 patients had a stability indicator of +1 for more than 24 hours, 318 achieved the value for 24-48 hours, 273 to 72 hours and 257 for 92 hours or more. No patient had a period of deterioration with a NEWS score of 6 or more after a value of +1 of the indicator.

Together with a team from Philips Healthcare we have:

  • Analysed routine data from patients admittedto hospital to generate a novel ‘stability index’ that takes value from -1 (very unstable) to +1 (very stable);
  • Tested the algorithm in simulation in a sampleof patients admitted to the Ysbyty Gwynedd in Bangor;
  • Confirmed acceptability of the algorithm to nursing and medical team members in two focus groups; and,
  • Planned prospective implementation as part of a digital healthcare solution.

Fit with Prudent Healthcare:

Prudent Principle 1: Good NEWS 4 Home was co-designed with an award winning team of clinicians from secondary care;

Prudent Principle 2: Good NEWS 4 Home allows to make the most effective use of the limited bed-base in secondary care;

Prudent Principle 3: Good NEWS 4 Home permits the limiting of hospital stay to a time when it is needed for safety of patients; and,

Prudent Principle 4: Good NEWS 4 Home uses the universally available National Early Warning Score to introduce objective criteria for transfer of patients from acute hospitals.