Article Text

Download PDFPDF

100 Multimorbidity and disease-count as predictors of healthcare use and mortality in emergency department attenders: a cross-sectional secondary analysis of routinely-collected data
  1. Chris McParland,
  2. Mark Cooper,
  3. David Lowe,
  4. Bethany Stanley and
  5. Bridget Johnston
  1. University of Glasgow, School of Medicine, Dentistry and Nursing; NHS Greater Glasgow and Clyde; University of Glasgow, Institute of Health and Wellbeing


Background Having 2 or more chronic conditions (multimorbidity) is associated with increased mortality and healthcare use in community-dwelling populations. In order to develop a nurse-led intervention for people with multimorbidity and palliative conditions, we sought to explore whether multimorbidity and disease-count were significant predictors of mortality and healthcare use in emergency department (ED) attenders.

Methods We conducted secondary analyses of inpatient and ED records for Glasgow (Scotland) residents attending the ED between April 2019 and March 2020. We conducted binomial logistic regression and calculated adjusted/unadjusted odds ratios (ORs) with 95% confidence intervals (CIs). Age, sex, ethnicity and deprivation were included in adjusted models. To handle missing data, complete case analysis was conducted and compared with results from post-imputation analyses. Ethical approval obtained from Local Public Advisory Committee.

Results 126,158 attendances by 75,726 eligible persons occurred during the study period. Complete data was available for 63,331 persons. Multimorbidity and disease count were significant predictors of all outcomes in both adjusted and unadjusted models. Complete case and post-imputation analyses produced comparable results. Of particular relevance to palliative care, only a small number of individuals died during admission (n=1.031, 1.6%), but multimorbidity was a significant predictor of this in both crude (OR: 4.41, 95% CI: 3.90–5.00) and adjusted (adjusted OR: 1.80, 95% CI: 1.58–2.05) analyses.

Conclusions Significant associations were detected with access to only 2–3 years historical inpatient data, so further validation of these predictors with greater historic inpatient and primary care data is warranted. We have however shown that these predictors are significant and should be incorporated into models aimed at identifying people at risk of healthcare use and mortality. Improving end-of-life care for people with multimorbidity is an avenue for further research, and robust models which can handle major class imbalances (only 1.6% ED attenders died during admission) should be tested.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.