National Early Warning Score systems that alert to deteriorating adult patients in hospital

Study size, design and location

A before-and-after, prospective, observational study, in which 577 nurse events were observed across 3 wards in 2 university teaching hospitals in 1 NHS trust (Oxford, UK).

Intervention and comparator(s)

Intervention: SEND system (n=296). Comprising manual data entry at the patient's bedside, on a roll stand mounted tablet, alongside the vital signs monitor.

Comparator: paper-based observation chart (n=281).

Key outcomes

The observation recording process was divided into 'view chart' and 'take vital signs' tasks. The former was defined as the task of locating and opening the chart, the latter was defined as measuring and documenting vital signs. Any interruptions were timed, classified and excluded from the overall comparison.

Overall, the mean time to complete a set of physiological observations and output a NEWS2-based early warning score was lower using SEND (150 seconds; 95% confidence interval [CI]: 130 seconds to 172 seconds) than paper-based observation charts (215 seconds; 95% CI: 177 seconds to 262 seconds). The treatment effect ratio was 0.70 (95% CI: 0.57 to 0.85, p<0.001), equivalent to a 30% reduction in time.

Of the 2 tasks, 'view chart' and 'take vital signs', the greatest time savings were observed in the latter. The mean time to locate and open the chart was 13 seconds (95% CI: 10 seconds to 17 seconds) using SEND and 18 seconds (95% CI: 13 seconds to 27 seconds) using paper (treatment effect ratio 0.36, p=0.052). The mean time to measure and document a set of physiological observations was 140 seconds (95% CI: 120 seconds to 164 seconds) using SEND and 194 seconds (95% CI: 156 seconds to 241 seconds) using a paper chart (treatment effect ratio 0.72, p=0.005).

Strengths and limitations

There were a relatively large number of nurse events observed, but the study was only in 1 NHS trust, and may not be generalisable across the wider NHS.

Measurements were taken between 9am and 5pm on weekdays only, and may not reflect practice outside those times.

Nurses were aware that they were being measured, which may have altered their normal behaviour (the Hawthorne Effect).

Study size, design and location

18 wards were observed in a prospective study in 1 NHS trust (Croydon, UK).

Intervention and comparator(s)

Intervention: Vitalpac system.

Comparator: unclear.

Key outcomes

The mean number of observations per patient increased from less than 3 to more than 4 per day. The completeness of observations was sustained at more than 99%. Observations taken on time increased from 55% to 85%. Night-time observations increased from 40% of those expected to 100%. Unplanned admissions to the intensive care unit (ICU) reduced by 54%. Cardiac arrests reduced by 70%, equivalent to 120 fewer per year.

Strengths and limitations

A relatively large number of wards were observed, but the study was only in 1 NHS trust, and may not be generalisable across the wider NHS.

This study was conducted using the earlier version of NEWS scoring, but the alert system and care pathway would still be generalisable to NHS practice.

There is limited further information to make a judgement on the study's strengths and limitations. However, resource requirements such as increased availability of advanced nurse practitioners and medical response teams were not reported in this study.

Study size, design and location

A multicentre, retrospective observational study of 251,266 admissions at 5 acute hospitals in 2 NHS trusts (Oxford and Portsmouth, UK).

Intervention and comparator(s)

Intervention: NEWS2 scores using the SEND system in Oxford and Vitalpac system in Portsmouth.

Comparator: NEWS scores using the SEND system in Oxford and Vitalpac system in Portsmouth.

Key outcomes

Primary outcome: in-hospital death within 24 hours of an observation set.

Secondary outcomes: cardiac arrest, unanticipated ICU admission, and either cardiac arrest, unanticipated ICU admission or death within 24 hours of an observation set.

The NEWS2 adjustment for patients with or at risk of type 2 respiratory failure (T2RF) differs from NEWS in the assignment of weights to measured SpO₂ (NEWS weights SpO₂ values below 96%; NEWS2 below 88%). Additionally, for patients with or at risk of T2RF when receiving oxygen, NEWS2 assigns weights for SpO₂ values above 92%.

Among 251,266 adult admissions, 48,898 were identified to be at risk of T2RF by diagnostic coding. In this group, NEWS2 showed statistically significant lower discrimination (c-statistic; 95% CI) for identifying in-hospital mortality within 24 hours (0.860; 0.857 to 0.864) than NEWS (0.881; 0.878 to 0.884). For 1,394 admissions with documented T2RF, discrimination was similar for both systems: NEWS2 (0.841; 0.827 to 0.855), NEWS (0.862; 0.848 to 0.875). For all secondary endpoints, NEWS2 showed no improvements in discrimination.

Strengths and limitations

This study was clearly reported and focused on patient subgroups relevant to the new SpO₂ scoring method in NEWS2.

Large and clinically relevant cohorts of patients admitted with, or at risk of, T2RF were generated for analysis through the retrospective datasets. Vital signs measured throughout the patient's hospital journey were incorporated in complete observation sets. Multiple imputation methods were used to account for missing values.

Limitations identified by the authors include potential misclassification of the diagnostic codes and records of oxygen prescription that the study relied on to categorise patients with, or at risk of, T2RF. The local database did not include documentation of the 'new confusion' part of the assessment of consciousness in NEWS2, so this was omitted from the analyses. Evaluation of the cardiac arrest and unanticipated ICU admission secondary outcomes in the documented T2RF group should be interpreted with caution, given the small numbers of these events (fewer than 100).

In the documented T2RF group, NEWS and NEWS2 were comparable in terms of composite outcomes, but NEWS2 reduced the alerts by up to 39%, without increasing mortality. This may illustrate the impact of limitations of using diagnostic codes (including chronic obstructive pulmonary disease and obesity) to identify the 'at risk of T2RF' group.