Logo image
Risk stratification for death in acutely unwell medical patients: A mixed methods study
Dissertation   Open access

Risk stratification for death in acutely unwell medical patients: A mixed methods study

Mark Holland
Professional Doctorate, University of Greater Manchester
04/2026

Abstract

Introduction Identifying acutely unwell patients at risk of deterioration and adverse outcomes is a clinical priority. The National Early Warning Score (NEWS) is an aggregate weighted track and trigger tool to support healthcare professionals in this endeavour. High-risk patients are readily identified by NEWS, but are in the minority, while patients with a low NEWS, who account for the majority of acute hospital admissions may still be at risk. Aim The aim of this thesis is to see if acutely unwell adult medical patients with either a low-risk (<5) or medium-risk NEWS (5-6) can be identified and stratified for their risk of in-hospital mortality. The thesis also aims to understand how consultant acute physicians use NEWS when making decisions. Methods The study employed a segregated mixed methods design. A retrospective observational study of routinely collected electronic patient data from a tertiary teaching hospital sought to identify presenting complaints and primary discharge diagnoses with the greatest risk of in-hospital mortality in acutely unwell medical patients, and to then identify which conditions have the greatest mortality risk when the admission NEWS is low or medium risk. All acute medicine admissions from 01/04/2014 to 30/06/2022 with complete data were included. An interview based phenomenological study drew on the experiences of consultant acute physicians to consider how they make clinical decisions and how they use NEWS. Results There were 104,301 acute medicine admissions with a primary discharge diagnosis and complete NEWS data, and 62,676 acute admissions with a documented presenting complaint and complete NEWS data. Of the 104,301 total admissions, within 24 hours of admission to hospital, 357 (0.3%) died. Of the patients who died, 102 (0.1%) had an admission NEWS <5, 53 (0.8%) a NEWS of 5-6 and 202 (4.1%) a NEWS of ≥7. While the relative mortality rate was lower in admissions with a low NEWS <5, 1 in 908, compared to those with a high NEWS ≥7, 1 in 24, in absolute terms patients with an admission NEWS <5 accounted for 28.6% of all deaths within 24 hours of hospital admission. Using the International Classification of Diseases 10th revision (ICD-10), in the 104,301 admissions, four ICD-10 diagnostic chapters were found to have an increased risk of 24-hour mortality when compared to all other chapters combined: infections (odds ratio (OR) 3.1, 95% confidence interval (95% CI) 2.4 4.0), respiratory (OR 2.2, 95% CI 1.8-2.8), neoplasms (OR 2.1, 95% CI 1.3-3.2) and cardiovascular (OR 1.4, 95% CI 1.1-1.8). Within diagnostic chapters, 24-hour mortality was lower when the admission NEWS was lower, although in relative terms there was variation in mortality between chapters, for example infections were more strongly associated with mortality when NEWS is elevated (OR 32.2, 95% CI 19.8 52.5, for NEWS ≥5) compared to respiratory conditions, where mortality is relatively more likely with a lower NEWS (OR 9.6, 95% CI 6.2-14.9, for NEWS ≥5). Using Summary Hospital-level Mortality (SHMI) diagnosis groups, 19 groups had an increased risk of 24 hour mortality when compared to all other groups combined. Of these groups, nine included ≥400 admissions and represented important acute medicine conditions: pneumonia, septicaemia, congestive heart failure, acute myocardial infarction, gastrointestinal haemorrhage, alcohol-related liver disease, stroke, bronchial cancer and respiratory failure. While a higher admission NEWS conferred an absolute increased mortality risk, in relative terms, respiratory failure (OR 1.3, 95% CI 0.3-6.4, for NEWS ≥5), congestive heart failure (OR 7.4, 95% CI 2.9-18.8, for NEWS ≥5) and pneumonia (OR 9.7, 95% CI 6.0-15.9, for NEWS ≥5) were more likely to die with a lower admission NEWS compared to all other SHMI groups combined (OR 20.4, 95% CI 16.2-25.7, for NEWS ≥5). There were 48 different presenting complaints. Shortness of breath (OR 3.8, 95% CI 2.9-5.1) and collapse (OR 2.4, 95% CI 1.5-3.9) were associated with a statistically significant increase in 24-hour mortality compared to all other complaints combined, while chest pain (OR 0.4, 95% CI 0.2-0.7) and falls (OR 0.3, 95% CI 0.1-0.8) were associated with a statistically significant decrease in 24-hour mortality. While a higher admission NEWS conferred an absolute increased mortality risk, there was variation in the risk of 24-hour mortality according to presenting complaint and admission NEWS. Although not statistically significant, those presenting with shortness of breath were relatively more likely to die with a lower admission NEWS compared to all other presenting complaints (OR 15.6, 95% CI 8.6-28.5, versus OR 20.0, 95% 14.5-27.6, for NEWS ≥5). Eight semi-structured interviews were completed with eight consultant acute physicians. Three themes were identified to understand how they make clinical decisions: the rituals of patient assessment, decision-making support and making the decision. Information gathering was vital, while reaching a decision was influenced by individual patient, environmental and personal factors. Cognitively, the consultants adopted a qualitative Bayesian approach, where each piece of new information gathered helped refine their prior clinical hypothesis. The consultants readily accepted an elevated NEWS as identifying patients at high risk of an adverse outcome; however, when the NEWS was low they sought further information to categorise each patient’s risk. Conclusion This mixed methods study has contributed to the understanding of the assessment and clinical decision making processes undertaken by consultant acute physicians. The quantitative study shows that a patient’s risk of in-hospital mortality is associated with their presenting complaint and final discharge diagnosis. Furthermore, seemingly physiologically stable patients still die following admission to hospital, and the mortality risk in low-risk patients varies according to their presenting complaint and primary discharge diagnosis. Therefore, acute medicine clinicians need to be aware of the risk associated with the conditions they see and the different levels of risk associated with these conditions in seemingly stable patients. However, the quantitative data shows there are no absolute cut-offs to provide information on which to make definitive decisions about a patient’s mortality risk. The qualitative study was therefore complementary, showing that consultants use NEWS alongside other sources of information when making decisions, either as qualitative Bayesian cognitive thinking, or as a binary judgement that a high NEWS always equates to increased risk, while a seemingly low-risk NEWS requires further information and evaluation.
pdf
Risk stratification for death in acutely unwell medical patients: A mixed methods study 12.59 MBDownloadView
PDF Version for Repository with Redactions Open Access In Copyright All Rights Reserved  — This item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation (such as any Fair Dealing allowances allowed by copyright law). For other uses you need to obtain permission from the rights-holder(s).

Metrics

1 File views/ downloads
2 Record Views

Details

Logo image

Usage Policy