Impact of dementia care resources on delirium screening in people with dementia in acute hospitals in Ireland

Background

Delirium is an acute confusion that commonly affects older people admitted to hospitals. Dementia is a chronic, progressive decline in cognitive function. Thus, dementia and delirium are distinct clinical syndromes though they frequently co-occur.¹ The reported point prevalence estimates of delirium range between 4-12% in community settings and about 6-74% in inpatient palliative care units.² Delirium is particularly common if the person has existing cognitive impairment, with a pooled prevalence of 48.9% for delirium superimposed on dementia (DSD).³ The presence of DSD is associated with significantly worse clinical outcomes, including accelerated functional decline, increased care dependency, prolonged hospitalisation, and higher mortality rates.³ Detecting delirium early can lessen mortality and morbidity.^4,5 Early recognition and management of delirium in individuals with dementia is therefore critical to improving prognosis. Thus, it is crucial that delirium screening is in place in acute hospitals, particularly for people at high risk, such as in-patients with dementia. However, this can be challenging in busy acute hospitals.

In this paper we aimed to explore whether delirium screening in in-patients with dementia, as recorded in INAD-2, related to specific dementia resources in an acute hospital, such as a dementia quality improvement team or staff training.

Methods of data collection

The data analysed in this study was collected during June-November 2019 as part of the INAD-2 audit, where 33 hospitals within the Republic of Ireland participated. For this study, data from the two orthopaedic hospitals and one elective hospital that had been included in INAD-2 were excluded, so that all included hospitals admitted people for unscheduled and scheduled care, giving a more homogenous sample. Data from two audit components was utilized- the hospital organizational audit tool and the healthcare records audit tool. The former focused on governance and training. The latter focussed on clinical practice, within 893 healthcare records (from 934 INAD-2 total records), where a patient had a length of stay of 3 or more days, and where dementia was recorded as a diagnosis in the routine hospital discharge dataset (ICD-10 coding, based on any documentation of dementia, of any type, in the medical chart), as per the INAD-2 inclusion criteria. More details about the method of data collection can be found in the INAD-2 report from 2020.⁵ INAD-2 was funded by the Health Service Executive; this secondary analysis is unfunded. Ethical approval for the secondary analysis was given by the Clinical Research Ethics Committee (CREC) of the Cork Teaching Hospitals, University College Cork.

Data analysis

Descriptive statistics were utilised, and as data was mostly categorical in nature, percentages and numbers were calculated. Univariate logistic regression analysis was conducted initially, to assess the impacts of six institution-level dementia resources on delirium screening performance. These resources were: a dementia quality improvement team within the hospital; one or more dementia specific roles (with protected time for dementia care); dementia champions (identified by senior hospital staff by their interest/ peer leadership; formal training not a criterion); receiving care on a geriatric ward; a hospital dementia care pathway or bundle in place; and dementia training provided to staff.

The significant variables in the univariate models were used in the multivariate model to calculate an adjusted odds ratio (aOR). We also included patient-level variables that might influence delirium screening such as the patient’s age, sex, length of stay in the hospital, and the main reason for admission, and the hospital’s complexity status (secondary, tertiary or quaternary care).

The OR and corresponding p-values were reported for both the univariate (OR) and the multiple logistic regression model (aOR). A p value less than 0.05 was considered significant. All data was analysed using IBM SPSS version 28.

Results

Of the 893 included healthcare records, 55% were females and the median age was 84 years (interquartile range [IQR] 79-88 years). The most common dementia diagnosis was Alzheimer-type (45%) and 93% were unscheduled admissions, most commonly for a respiratory tract infection (21%). Most of the patients (64%) were admitted from home and the median length of stay was 10 days (IQR=6-10days). Patients spent most of their in-patient stay on a general medical ward (69.8%), geriatric medicine ward (13.6%), surgical ward (7.8%), or an orthopaedic ward (2.8%), with very few patients on other types of ward.

Within this sample, 176 (19.7%) had evidence of delirium screening conducted at least once during admission and 715 (80%) did not. Within 24 hours of admission, 140 (15.7%) had screening performed, while 19 (2.1%) were screened between 25-48 hours after admission. Delirium screening occurring only after the first two days was rare, accounting for 17 (1.9%) of cases. The frequency or periodicity of screening was not recorded in the audit.

The most commonly used tool for delirium screening was the 4AT (76.4%) followed by the Single Question in Delirium (SQiD) (19%), months of the year backwards (4%) and the Confusion Assessment Method (CAM) (n=1). Of the 174 patients screened, the screen was positive in 103 patients (59.2%).

Although many factors influenced delirium screening on univariate analysis (Table 1), on multiple logistic regression modelling (Table 1), delirium screening was independently influenced by two service-level factors: staff dementia training in the last 12 months (aOR 3.5; 95% Confidence Intervals [CI] 2.2-5.6) and the presence of a dementia pathway (aOR 4.1; 95% CI 2.3-7.2). Among patient-level factors, although a longer length of stay gave more time for delirium screening, this didn’t impact delirium screening in either model, reflecting that almost all screening was done within 24 hours if done at all (aOR 1.01; 95% CI 1-1.02). Being admitted with dementia as the primary cause of admission increased the odds of delirium screening, (aOR=2.2; 95% CI 1.2-3.9). The hospital complexity status had no impact on delirium screening in the multivariate model. The variance inflation factor (VIF) values for all variables in the multivariable model were below the threshold of 5, indicating no significant multicollinearity.

The model was then repeated without the inclusion of ‘dementia specific roles’ as these had been non-significant in univariate analysis and indeed inversely associated with screening performance in univariate analysis (Table 1). Adjusted ORs changed slightly but independent influences remained.

Discussion

Delirium screening on all patients over 65 admitted to hospital is recommended. In the INAD-2 audit, more than 50% of included hospitals reported that they screen “all or some” patients with dementia for delirium, but the healthcare record audit found evidence of delirium screening at any time during admission (noting the inclusion criteria of at least 3 days stay) is only 19%.⁶ In fact, this represents a decline from the 2013 audit performance (29.7%; 196 out of 659), which may reflect little national focus or investment into acute hospital delirium screening/intervention in the intervening years. Of note, the positivity of screening was nearly 60%, higher than the reported occurrence of delirium in hospitalised patients with dementia of just under 50%.³ The accuracy of the screen result could not be explored due to the retrospective nature of the data, but the low frequency of screening performance and high screen positivity rate indicates that screening may be more likely in those who are displaying delirium features, thereby prompting staff to perform the ‘screen’.

It is valuable to understand the factors that may influence delirium screening, as these could potentially be targeted to support improved delirium screening. The INAD-2 audit proved an opportunity to explore this, notwithstanding the limitations of retrospective audit data. We were primarily interested in service-level factors, which we term “dementia resources”, as we hypothesised that these would facilitate delirium screening. One of these factors that was independently associated with delirium screening in patients with dementia was staff dementia training (aOR=2.8) which is intuitive as most care is done by non-dementia specific staff, and staff training would be expected to improve practice in this regard. It is important to highlight however that other confounding effects can’t be out ruled, such that hospitals that provide dementia training may also promote delirium screening in other ways, or may have other patient-level confounders that we could not measure in a retrospective audit. A previous study of implementing a delirium guideline found little change with a guideline in isolation, improved by staff training.⁷ Other studies have confirmed the value of staff training for delirium screening and assessment, across multiple settings.^8–12

Within service-level factors, the apparent negative impact of having dementia-specific staff on delirium screening is interesting. This may be due to confounding effects, but it is possible that having specialist dementia staff leads to other staff believing it is not their role to screen for delirium. Equally, the number of specialist staff was small, often just one dementia-specific nurse for a large hospital. There is contrary evidence from implementation studies that specialist (delirium) staff improves delirium screening.¹³ However, a recent study on delirium care practice in 15 hospitals in Ireland (any patient, any ward) also found an inverse relationship between delirium screening performance and having specialist delirium staff.¹⁴ More data is needed to explore the factors that may be influencing this association, noting that hospital complexity (size) did not play a role.

Staying within service-level factors, having a dementia pathway in place positively influenced delirium screening (aOR=2.8). Again, this is intuitive as most dementia pathways specify delirium screening, but there may be other factors relating to both pathway presence and delirium screening that wasn’t included in the model, as (delirium) guidelines alone are known to have little influence on care in the absence of staff engagement and training.¹⁵

Although our focus was on service-level predictors, our analysis found that being admitted directly due to dementia (e.g. wandering, carer stress, etc) was positively associated with delirium screening (aOR 2.4), whereas other patient-level factors such as age, gender and length of stay in the hospital had no significant association. It is important to note that the audit underlying this analysis was conducted in the charts of patients with well-documented dementia and may not reflect patient-level influences on delirium screening in general older or all-age adult cohorts. Equally, as 93% of admissions were for unscheduled care, it may not reflect practices in more elective services or hospitals.

In conclusion, this real-world data adds to the existing evidence from clinical trials/intervention implementation and short-term quality improvement projects. In practice, teasing out the effects of introducing a new pathway (guidance) versus the effects of staff delirium awareness raising and training is challenging, when these so often overlap, but our study shows that these two (dementia pathway and staff training) seem to be more independently associated with delirium screening than specialist dementia staff, dementia champions and dementia quality improvement teams.

Of note, the role of electronic healthcare records and audit/feedback cycles in influencing delirium screening was not explored in the study, as the former was not in place, and the latter was not part of the audit data. Next steps could explore how training and dementia pathways are used in practice, and identifying any barriers that may exist to routine screening. This could be achieved through frontline staff helping to identify how resources influence screening in practice, and where any gaps remain. To support sustained efforts in practice, the integration of screening across relevant care protocols, electronic records with “stop” functions when screening is not performed, and audit-feedback cycles could be explored.

Author Contributions

Shelly Chakraborty: Formal analysis, Writing – original draft

Suzanne Timmons: Conceptualization, Methodology, Funding acquisition, Investigation, Writing- original draft, Writing - review & editing.

Zahra Azizi Formal analysis, Validation, Visualization, Writing – review & editing

Lorna Kenny Visualization, Writing – review & editing

Emma O’Shea Conceptualization, Methodology, Writing – review & editing

Emer Begley Conceptualization, Supervision, Writing – review & editing

Mairead Bracken-Scally Methodology, Investigation, Data curation, Writing – review & editing