Background
Delirium is an acute neuropsychiatric disorder characterised by an acute onset, fluctuation, deficits in attention, and other cognitive impairments.1 It is defined by reduced ability to focus, sustain, or shift attention and altered (increased or decreased) alertness.2 Causes of delirium vary considerably and may include substance withdrawal, infection, metabolic derangements, and surgery.2 The estimated pooled occurrence is 23% amongst adults admitted onto acute medical or geriatric wards,3 25% after stroke, 6-74% in palliative care wards,4 up to 50% in postoperative patients,5,6 and up to 89% in Intensive Care Units (ICU).7 Older patients are most frequently impacted by delirium, with delirium present in 11-25% of older general medical patients on admission to hospitals,8–10 and up to 42% during hospital admission.11–13
In Ireland, a national delirium algorithm (last updated in 2021) states that patients older than 65 presenting to emergency departments (ED) or acute medical assessment units (AMAU) should be screened for delirium,14 using a formal assessment tool, the 4AT. This result should be checked on inpatient transfer, and performed if either missed or performed more than 48 hours previously (ward delirium algorithm; last updated in 2021). In Ireland’s second round of the National Audit of Dementia Care (INAD-2), conducted in 2019, only 19% of patients with known dementia (who are at particularly high risk of delirium) received any delirium screening during hospital admission. This was worse than in 2013,15 indicating a need to improve the national implementation effort.
World Delirium Awareness Day takes place each year in March. In 2023, a global survey was conducted across multiple countries to examine delirium prevalence, outline ward practices for delirium prevention, assessment and management, and describe the barriers to improvement (Start | wdad-study.center). We report prevalence data from Irish hospitals, representing more sites than participated in the national dementia audits.
Methods
Participants
We included patients in clinical sites in the Republic of Ireland, including acute hospitals, residential care facilities, and inpatient rehabilitation units. No direct patient data, only pooled ward data, were collected. The survey respondents (i.e. participants) were clinical or research staff who acted as data collectors and completed the online survey (one per ward).
Data Collection and Sharing
On a standardised paper-based collection sheet, hospital-level data included the a) number of beds and b) affiliation (e.g. university-linked, private, etc). Unit/ward-based data included a) patient age category (paediatric, adult, over-75, mixed), b) main discipline of the ward (e.g. medical/surgical, etc), c) ward type (e.g. ICU, general ward etc), and d) number of beds. Delirium-specific data included verbal reports of delirium assessments performed on the ward/unit (multiple formal tools listed, with an option for none, other tool, and personal judgement).
On World Delirium Awareness Day (WDAD; 15th March 2023), the participants visited each participating ward twice, at 8AM +/- 4 hours and 8PM +/- 4 hours and interviewed the ward managers who reported the delirium point-prevalence on the ward (number with and without delirium, and number with uncertain delirium status, mostly due to not having been tested, though some were known to be untestable - e.g. due to aphasia), based on the ward’s usual clinical practice for delirium detection. We excluded data from wards where the total number of reported patients (delirious and non-delirious) exceeded the total number of ward beds (two wards for 8PM data and one for 8AM data). The paper-based data was entered into the global online survey within 3 days of WDAD, i.e., by midnight on 18th March 2023. The Irish data were retrieved separately using a pre-determined code.
Ethical approval
Kiel University in Germany granted ethics permission for the worldwide study (AZ_D 519/22_Aug 22). The Clinical Research Ethics Committee of the Cork Teaching Hospitals gave ethics permission for secondary data analysis of Irish data (ECM 4 (n) 13/12/2022).
Data analysis
We used R, version 4.2.016 for all analyses. Nominal data were reported as percentages; the numerator/denominator value was also reported whenever this was less than 90% of the total data set. To compare categorical data between distinct groups, chi-square tests were performed. In addition, we examined the association between delirium prevalence and ward or hospital characteristics using generalised logistic mixed-effect models by glmer in the lme4 package (family = binomial),16 with delirium prevalence as the dependent variable, and ward speciality, time points, and hospital/ward factors as predictors. Statistical significance level was set at 95% (p < 0.05).
Results
In total, 132 wards from 15 hospitals across the Republic of Ireland participated; the South-Southwest hospital group was predominant (Supplementary Figure 1). The size of the participating hospitals varied, from <250 beds (20.5%) to >750 beds (17.4%). Most were university-linked (92%), while four were rehabilitation or long-term care hospitals. Wards were diverse, most commonly either medical or mixed medical-surgical (including one oncology ward), geriatric wards, and intensive care units. A smaller number were EDs or AMAUs (combined as ED/AMAU), and surgical wards. There were few long care wards (n=7), rehabilitation wards (n=5) or transitional care (step-down) wards (n=1); combined as non-acute wards.
In total, 59.8% of wards used personal judgment for delirium assessment, while 26.5% used 4AT, and a few used CAM-ICU (7.6%), NU-DESC (0.8%) or any other formal tools (5.3%). Delirium data included the number of assessed patients, those delirious/not, and those who were untestable (e.g., aphasia) or had an unclear result (dementia affecting performance), at 8AM and 8PM. The data did not vary between these two-time points, so we combined them (Supplementary Figure 2). Delirium status unavailability was different when using personal judgment (8.9%, 304/3497) versus formal tools (26.7%, 600/2249) (p<0.001) (Figure 1), showing that formal tools were not always used where this was reportedly normal practice. Excluding those not assessed at all, a formal tool determined that 15.9% (222/1395) had delirium, compared with 11.5% prevalence (348/3018) when only personal judgment was used (p<0.001). While personal judgment can be easily applied, it appears to miss cases (Figure 1).
The prevalence of delirium also varied based on ward type. While more than half of patients assessed with a formal tool (54.2%; 26/48) were diagnosed as having delirium in non-acute wards (noting the small sample size), this was only 9.6% (24/249) in intensive care wards, 14.2% (35/246) in general surgical wards, and 20.4% (22/108) in the ED/AMAU. Using a formal assessment tool, 13.4% (70/524) and 20.5% (45/220) of patients in general medical/mixed medical-surgical wards and in geriatrics wards, respectively, were identified as delirious (Supplementary Figure 3).
The association between delirium prevalence and: using a formal assessment tool; ward specialty; time points (8AM/8PM); and number of beds in the ward/hospital is reported in Figure 2 and Supplementary Table S1. The prevalence of delirium was higher when using a formal tool for assessment, while non-acute wards reported fewer delirium cases when general medical wards were set as the reference in the model. Other predictors did not show any association.
Discussion
This is the first Irish study to assess the prevalence of delirium in multiple hospital wards in multiple hospitals over a single day. Delirium prevalence was 12.9% in the entire sample and 15.9% in the sub-sample where a formal assessment tool was used.
While early and precise detection of delirium enables early intervention and treatment of the underlying cause,17 it is reported that up to 80% of delirium is undetected by healthcare professionals.18 This finding was replicated in more recent studies and a wide range of facilities, including an acute hospital (72%),19 emergency departments (84.6)18 and palliative care centres (61%).20 Under-detection of delirium may occur for a number of reasons; symptoms are heterogeneous and transient,13,21 and diagnosis is subjective, relying on clinical skills in the absence of a objective ‘test’.22,23 Our findings mirror these results as formal assessment was associated with higher detection, although this could be confounded by wards where delirium is more common being more likely to implement formal assessment. Of concern, 26.7% of patients did not receive any delirium assessment on wards reportedly using a formal assessment tool. The issues that lead to the non-performance of delirium screening need to be explored further.
In high-income countries, the prevalence of delirium in general medical wards is between 18-35%, and 17% in surgical wards.17 The prevalence of delirium in our findings was lower than in a prior single site point prevalence investigation in Ireland (15.9% [in our formally assessed patients only] vs. 19.6%).12 However, that latter study used a highly systematic multi-step screening process of all adult in-patients, with initial high-sensitivity and low-specificity screening by junior doctors, followed by more in-depth assessment by more senior staff, and then formal diagnosis by highly trained experts. A prior multihospital study in the UK had reported delirium prevalence to be 14.7%24 among in-patients aged 65 years and older screened using the 4AT by the study team, a finding very similar to ours.
The rate of delirium across ward types was different. Using formal assessment, non-acute wards had the highest prevalence, compared with much lower reports in rehabilitation hospitals (14.0%).11 However, these differences may be confounded by study assessment methods not being directly comparable. We note that previous reports suggest that delirium in non-acute wards is common, but with estimated prevalence ranging between 1.4% and 70%, depending on diagnostic criteria and the prevalence of dementia25–27; equally, care workers often fail to recognise it.26 Our findings showed that when delirium was formally assessed, prevalence in geriatrics wards was lower than the previous single site point prevalence study (53%),12 noting a very small sample size (n=15 patients in the geriatric ward) but more rigorous detection methods in that previous study. Other studies reported the prevalence of delirium in geriatric wards ranging from 20% to 30%,11,24,28 similar to ours.
The strengths of the present study include the large sample and the inclusion of more than 130 wards from different settings across Ireland. This study has certain limitations. In this study, the South-Southwest hospital group in Ireland was predominantly represented, possibly due to the existence of a dementia quality improvement steering committee in this group, or the senior author’s clinical post (acting as a study champion). Additionally, the study did not collect patient data, limiting reporting of the prevalence of delirium based on patient characteristics. Finally, the study relied on reports from busy clinical staff, which were not checked for accuracy, and many wards did not have formal delirium screening in situ, limiting the sample size for tool-detected delirium.
The reported prevalence of delirium in Irish hospitals on wards that used a formal assessment tool was 15.9%, with the highest prevalence in geriatric wards (20.5%). Where formal assessment tools were used on a ward, the rate of known delirium was higher, but more than a quarter of patients on these wards were not assessed and a further 11% were indeterminate or untestable, also indicating missed cases. Thus, we must conclude that delirium remains significantly under-detected as yet in Irish hospitals.
Statements and Declarations
The authors have no competing interests to declare.
Author contributions
ST and NO’R contributed to the study conception and design. Data collection was performed by ST, NO’R, TD, TB, EH, CD, IC, MC, SO’R, CN and EW and analysis by ZA. The first draft of the manuscript was written by ZA and ST and all authors commented on this version of the manuscript. All authors read and approved the final manuscript.
