Impact of delirium and dementia on 30-day readmission to hospital by place of residence in older adults: ORCHARD-EPR cohort
Abstract
Background: Cognitive frailty, defined as dementia, delirium or low cognitive test score, is prevalent in older in-patients. Current guidance (eg National Audit Dementia 2022) suggests hospital readmission as a performance metric in dementia care but reliable data are lacking. We therefore determined readmission risk by cognitive frailty status in care home residents vs those living at home.
Methods: ORCHARD-EPR (2017-2019) includes de-identified EPR data for patients ≥70 years with length of stay (LoS)≥1 day (2017–2019) admitted to four Oxfordshire, UK hospitals. Cognitive frailty was assessed using a standard cognitive screen mandated on admission comprising the 10-point Abbreviated Mental Test (AMT), dementia, and delirium diagnosis informed by the Confusion Assessment Method (CAM). Care home residence was determined by cross-referencing postal codes against Care Quality Commission repositories. Hazard ratios for 30-day readmission were adjusted for age, sex, comorbidity and illness severity.
Results: Among 28,590 patients (mean/SD age=81.8/7.4 years, 52% female), 3,569/26,596 (13%) were readmitted within 30 days of discharge: adjHR=1.18 (95% CI 1.07-1.30; 505/3,192 [16%], p<0.001) for delirium only, 1.17 (1.03-1.32; 301/1,993 [15%], p=0.013) for delirium on dementia, 1.10 (0.97-1.25; 279/1,930 [14%], p=0.13) for dementia only and 1.13 (0.93-1.38; 109/706 [15%], p=0.2) for low cognitive test score compared to no cognitive frailty (2,329/17,994 [13%]). After stratification by place of residence, in those living at home delirium but not dementia, predicted 30-day readmission: delirium alone adjHR=1.18 (95% CI 1.06-1.32; 411/2,1640 [16%], p=0.002), delirium on dementia 1.25 (1.08-1.45; 194/1,211 [15%], p=0.004), dementia alone (p>0.1). In contrast, cognitively frail vs robust patients previously resident in care homes had no increased risk of readmission (p>0.1).
Discussion: Only delirium in patients living at home prior to admission predicted 30-day readmission. Dementia alone was not associated with readmission irrespective of residence. Readmission may not therefore be a useful performance metric in assessing dementia care quality.
Comments
Interesting Poster
Really interesting poster, I like how you’ve separated outcomes by place of residence, that adds a lot of nuance to the findings.
I was wondering, do you think delirium is truly an independent predictor of readmission, or could it be acting more as a proxy for acute illness severity or underlying frailty?
And building on that, how confident are you that your adjustment, especially for illness severity and comorbidity, fully accounts for those factors?
So in that context, if delirium is partly a marker rather than a cause, how should we interpret the clinical implications? Should we be targeting delirium itself, or the underlying vulnerability, particularly in patients living at home?”
Good one. The findings…
Good one. The findings suggest that delirium rather than dementia alone is the key cognitive factor driving 30-day readmissions, particularly in patients living at home
Author's Response
Thank you for your comment. Although data are observational, we anticipate that delirium is likely both an independent predictor of readmission (for example, by leading to falls) and also a proxy or marker for underlying vulnerability (for example, secondary to acute illness or frailty). Associations between delirium and readmission have been shown to be robust to adjustment for confounders including ICU admission and comorbidity previously (LaHue et al., 2019), but associations between delirium and frailty are well established. Future studies could address this question more specifically.
The National Early Warning Score (NEWS) and Charlson Comorbidity Index (CCI) are useful constructs for capturing illness severity and comorbidity, respectively. Some residual confounding is expected given the study design. For example, only NEWS data on admission were used in the analysis so patients who deteriorated during their admission would not be fully accounted for. The magnitude of residual confounding is likely small as most patients have prevalent delirium, rather than incident delirium arising during admission (Gan et al., 2025).
We believe that the clinical implications of this work are first that delirium is important to identify as manage because it is predictive of readmissions (much more so than dementia) in addition to contributing to other adverse outcomes as described previously, and second that interventions aimed at addressing underlying vulnerabilities may be more effective in the group of patients living at home.