A hospital-based special care unit for dementia decreased hospital readmission rates for behaviour while reducing rates of falls and occupational violence across medical wards

Abstract

Background

Hospital-based Special Care Units (SCU) for dementia show promise as effective models of care. However, limited research describes hospital-wide benefits.

Objective

To describe patient-level outcomes and hospital service-level outcomes of a SCU.

Design

Pre-post analyses of SCU-patient data and hospital service-unit incident report data.

Setting, participants

2-years of SCU-patient admissions and 4-years of hospital incident-reports from four medical wards (2-years pre-post SCU).

Methods

Admission and discharge severity of SCU-patients’ behaviour was prospectively measured by specialist SCU nurses. One-year hospitalisation rates, length-of-stay, diagnosis and patient demographics were retrospectively obtained from medical records. Hospital service-unit data included 4 years of monthly incident report rates for falls, pressure injury and occupational violence (OV) across four medical wards. Analysis of count data used Poisson and Negative Binomial Regression models.

Results

121 SCU admissions involved 107 unique patients. Median SCU LoS was 23 days (interquartile range [IQR], 13–50), and stabilisation of behaviour severity took 11 days (IQR 6–12). Barriers to discharge related to substitute decision-making and care facility availability. After SCU discharge, yearly hospitalisation rates for ‘all-reasons’ decreased by 68% (Incident Rate Ratio [IRR], 0.32, 95% CI, 0.23–0.43), and 83% for behaviour-related admissions (IRR 0.17, 95% CI, 0.11–0.28). For hospital service-unit outcomes, falls-per-month decreased by 21% (IRR 0.79, 95% CI 0.64–0.99) after SCU implementation and OV by 26% (IRR, 0.74, 95% CI 0.59–0.94).

Conclusions

A hospital-based SCU reduced hospital health service demand through decreased SCU-patient readmissions and was associated with decreased falls and OV rates across hospital medical wards.

Key Points

• Targeted evidence-based models of care for inpatients with behaviours and psychological symptoms of dementia are lacking.

• Special Care Units (SCU) hold promise but a barrier to adoption is a lack of research limiting the evidence-base.

• An SCU reduced readmissions for behaviour symptoms by 83% and decreased falls and occupational violence across general wards.

• SCUs not only reduce behavioural and psychological symptoms of dementia (BPSD) severity, they may decrease the impact of BPSD on the hospital service at large.

Introduction

In hospital, one-in-four medical inpatients are likely to have dementia [1]. Their average length-of-stay (LoS) is 14.8 days, five times higher than the 2.7-day average for all hospitalisations [2]. Associated admission costs are ~50% higher [3, 4]. The presence of complex changed behaviours and neuropsychiatric symptoms is a significant contributing factor [5–7]. Clinically referred to as behavioural and psychological symptoms of dementia (BPSD), ~75%–94% of this vulnerable hospitalised cohort will experience symptoms including agitation, aggression, resistance to care, delusions and hallucinations, sleep disturbance, anxiety, wandering and exit seeking [6, 8].

BPSD challenge care-delivery, disrupt planned medical treatments and are associated with adverse outcomes [6, 9], physical restraint and psychotropic sedation [10–12] and patient-to-nurse violence [13–15]. The direct impact on LoS from BPSD has been estimated to be double that of patients with dementia who do not have BPSD [8]. The presence of BPSD also increases mortality [8], 30-day readmissions and decreases discharges home [8, 16].

BPSD often manifest from disease-induced vulnerabilities to every-day environmental, social and physiological stressors [8, 10, 16]. Busy hospital wards, staffed by clinicians unskilled in dementia care, working to system priorities incongruent with the needs of patients with dementia, perpetuate and exacerbate BPSD [6, 17–20]. This suggests BPSD severity in hospital inpatients could be modifiable through better care practices and environmental redesign [20, 21]. Architecturally designed as dementia-enabling, Special Care Units (SCUs) are serviced by staff specialised in geriatric medicine and/or psychogeriatrics and are situated in hospitals internationally [20–24]. Germany has 44 SCUs and the United Kingdom (UK) include SCUs in their dementia-friendly hospital initiative [20]. Australia has ~25 hospital ‘units’ caring for patients with BPSD, but it is unclear how many have the built-design of an SCU [25]. Most Australian hospitals rely on standard ward-based care, staff training in dementia care and designated quiet areas within wards [1, 26].

A barrier to hospital uptake of SCUs is an absence of research on their design, operation and impact on hospital services [21, 27]. Following on from our previously published work comparing SCU behaviour outcomes with standard ward outcomes in an Australian context [28], this paper adds to the research-base.

It aims to report on:

1. The hospitalisation outcomes for SCU patients (e.g. LoS, admission/discharge, BPSD) and

2. The impact of implementing an SCU on other hospital service-unit outcomes (e.g. hospital acquired complications and occupational violence).

Methods

We undertook retrospective pre-post analyses on data related to the implementation of a SCU in January 2020 at a large metropolitan hospital in Brisbane, Australia. Data included a two-year longitudinal dataset of all SCU patient admissions and a 4-year dataset of hospital incident reports for all patients from four medical wards (including the SCU ward). At the time of this study, no other hospital-based SCUs existed in the Hospital Health Service.

SCU model of care

The design of the SCU model of care in this study has been described in detail in a previous paper [28]. In brief, it is an eight-bed secure unit with a dementia-enabling architectural design for independently mobile older patients (≥ 65 years) with clinically significant BPSD. The SCU is located within a subacute geriatric and rehabilitation ward and receives referrals hospital-wide, although most are from the three general medical wards. It is staffed by a geriatrician-led interprofessional team expert in managing changed behaviours and formulating effective Behaviour Support Plans. The SCU provides discharge support and community follow-up with either a Nurse Navigator (NN), Hospital Geriatrics Outreach Service, or referral to Dementia Services Australia.

Hospitalisation outcomes for SCU-patients

A longitudinal dataset of all SCU admissions between January 2020 and January 2022, was prospectively collected through normal SCU clinical operations. It included repeated measures of BPSD severity, treatments and outcomes. Our analysis of hospitalisation outcomes for SCU-admissions retrospectively examined this dataset.

SCU patient characteristics

Patient’s medical records established patient demographics for age, sex, Charlson Comorbidity Index, dementia subtype, presence/absence of dementia diagnosis on hospitalisation, delirium on SCU admission, hospital presentation reason, place of residence.

Length-of-stay

At weekly case-conferences BPSD acuity was recorded i.e. whether acute or stable enough for discharge, enabling calculation of an ‘acute BPSD LoS (i.e. an active treatment phase) and a ‘stable BPSD LoS’. Delays and barriers to discharge were also recorded, noting whether these were related to involvement of the Queensland Civil and Administrative Tribunal (QCAT), Public Guardian, Public Trustee, and National Disability Insurance Services (NDIS), or were related to availability of a suitable aged-care facility, family inaction or the logistics of organising patient transfers back to the community.

Discharge destination

The discharge destination of each admission episode (including patients’ readmissions) was recorded including whether it was a change from the place of residence prior to that hospital admission.

Admission rates

The number of hospital presentations/admissions to any public hospital across the hospital health service one-year prior to SCU admission and within one-year of SCU discharge were counted retrospectively using the patients’ electronic medical records. For those patients with multiple SCU readmissions, the first admission was used as the point of measure. Presentations were classified as follows: (i) medical presentation with no admission, (ii) BPSD presentation with no admission, (iii) medical admission and (iv) BPSD admission. Medical reasons encompassed any illness, injury, trauma and surgical or medical procedures. Patients who died in hospital during the first admission or within a year from their first discharge were excluded from the admission-rates analysis.

BPSD severity

BPSD severity scores taken on hospital admission, SCU admission and SCU discharge were extracted from the dataset for analysis. BPSD severity was rated by specialist dementia nurses every 4 (±1) day on three validated instruments: (i) The rating Scale for Aggressive Behaviour in the Elderly (RAGE) which rates aggression severity on 21-items (scores 0–61) [29], (ii) The Neuropsychiatric Inventory (NPI) which rates 12 neuropsychiatric symptoms in severity (scores 0–144) [30], and (iii) a modified version of Cohen Mansfield Inventory (CMAI-9) for use in acute-care which rates non-aggressive agitation on 5-items (scores 5–35) and aggressive agitation on 4-items (scores 4–28) [31]. NPI subcategory scores were calculated for Frontal, Hyperactivity, Mood, Psychosis domains [16, 30, 32]. All scales rated BPSD over the previous 4 (±1) days and higher scores equate to higher symptom severity.

Hospital service-unit outcomes

Monthly rates of hospital-acquired complications and occupational violence incidents were extracted from a four-year incident-report dataset pre-SCU implementation (2018–2020) and post-SCU implementation (2020–2022). Ninety-two percent of patients admitted to the SCU were from the four medical wards included in this dataset.

Hospital-acquired complications

Hospital-acquired complications (HACs) for falls and pressure injury were examined in the dataset. Staff report incidents when they observe a patient pressure injury or fall. These reports include patient identifiers and characterises related to the incident. For falls, injury-level was categorised as follows: (i) death, (ii) sprain/fracture/laceration/contusion/internal-bleeding, (iii) abrasion/bruising and (iv) nil. For pressure injury, injury-level was categorised as per staff’s staging of the injury: (i) open-wound extension to muscle/tendon/bone or closed-wound involving deep-tissue injury or unstageable, (ii) open-wound extension through skin and fat but no further, (iii) open-wound extension through skin but no further, and (iv) nil.

Occupational violence

Occupational violence (OV) reports are made when staff witness or perceive verbal threat, abuse and aggression and/or physical threat and assault. Both verbal and physical OV was included. We categorised level of injury as follows: (i) sprain/fracture, (ii) bruise/abrasion, (iii) psychological, and (iv) nil.

Two researchers independently reviewed the medical records of all patients linked to incidents reported in the dataset and categorised them with regards to cognitive impairment and mobility: (i) ‘mobile with dementia and/or delirium’; (ii) ‘poor mobility with dementia and/or delirium and/or substance/intoxication/withdrawal’; (iii) ‘other neurological syndromes and cognitive disability’; (iv) ‘mental health’; and (v) ‘cognitively intact’. A third researcher arbitrated differences in classification.

Analysis

Data analyses were conducted using Stata 18 (Statacorp Texas). Descriptive statistics were presented as frequency counts, percentages, means with standard deviations (SD) and medians with interquartile ranges (IQR). Fisher’s exact test was used for correlations between LoS and barriers to discharge. Paired-samples t-tests compared changes in behaviour scores between admission and discharge for individual SCU patients. Cohen’s d was used determine if effect-size was small (d = 0.2), medium (d = 0.5) and large (d ≥ 0.8). Wilcoxon signed-rank tests were used for non-normal data. Statistical significance was estimated at the 0.05 level. Missing incident data was due to reporting staff omissions for some reporting-fields and was treated as Missing Completely at Random (MCAR) and removed by listwise deletion when under 10%. For multivariate analysis of count data and rates, Poisson regression was undertaken. Where overdispersion of data was detected, negative binomial regressions were used. For count data related to falls, pressure injury and occupational violence, Poisson regression was undertaken using the exposure variable of fractional accrued patient bed-days to adjust for variances in inpatient-bed-occupancy. Results were reported with 95% confidence intervals (CI) for associations between pre-post SCU counts and for risk-adjusted variables.

Ethical approval with a waiver of consent was granted from a Metro South Health Human Research Ethics Committee HREC/2022/QMS/81684). This research as funded by a Metro South Health Research Support Scheme Innovation (seed) grant.

Results

There were 121 SCU-patient discharges between January 2020 and January 2022. These discharges involved 107 unique patients, with fourteen same-patient readmissions. Three patients were admitted three times and eight patients, two times. There were nine SCU inpatient deaths.

Patient characteristics are presented in Table 1 alongside their corresponding NPI scores. The average patient age was 78.6 years (SD 7.8; range 60–97) and the majority (60%) were male. Almost half were admitted from home. While 60% presented to hospital because of BPSD, another 20% presented with an initially unclear behavioural aetiology as 30% of presentations had no prior dementia diagnosis. The most common types of dementia were Alzheimer’s disease, mixed dementia (Alzheimer’s/Vascular) and Vascular dementia.

Hospitalisation outcomes for SCU patients

Table 2 presents SCU outcomes for the 121 discharges. Modes of separation describe 60% of patient discharges to new or modified community-care situations, with 21% requiring a different aged-care facility and 32% transitioning to their first aged-care facility. The median LoS in the SCU was 23 days (IQR 13–50) of which 11 days (IQR 6–21) involved active treatment for BPSD (i.e. the acute-BPSD phase). Overall median hospital LoS was 34 days (IQR 17–65). Analysis of LoS for patients with stable BPSD (see supplemental table) revealed that 44% of stable LoS > 4 weeks occurred when discharges depended on issues involving QCAT and 63% related to RACF capacity for BPSD. Conversely, 76% of stable LoS of one week or less related to logistics of organising patient transfers i.e. RACF preparations, transportation and equipment set-up. These associations were statistically significant (P < .05).

SCU patients experienced statistically significant reductions in symptom severity across all behaviour scales with large effect sizes suggested by Cohen’s d (Table 3). The difference between NPI-score on SCU admission (M 28.9, SD 25.6) and discharge (M 15.7, SD 17) was 13-points (95% CI 10.26–16.13), a 45% decrease in severity from admission to discharge. For aggression, the difference was 7.3-points (95% CI, 5.99–8.67), a 60% decrease in severity. Similarly, CMAI-9 scores for non-aggressive agitation decreased by 40%, or 7.1-points (95% CI, 5.96–8.17) points and for aggressive agitation by 37%, or 3.5-points (95% CI, 2.67–4.27). Decreases in severity of NPI clusters for Psychosis, Hyperactivity and Mood were statistically significant while changes in the Frontal cluster, were not.

Eighty-five patients were included in the admission-rates analysis as nine died in hospital and thirteen died in the first year of discharge. Comparison between one-year hospital admission rates for SCU-patients before and after first SCU-discharge revealed a decrease from an average of two hospital admissions-per-year (2.0, SD 0.18) to less than one admission-per-year (0.65 SD 0.10). Table 4 reveals statistically significant decreases in admissions for ‘any reason’ by 68% (IRR 0.32, 95% CI 0.23–0.43), while admission rates for ‘medical reasons’ decreased by 45% (IRR 0.55, 95% CI 0.37–0.83) and admission rates for ‘BPSD reasons’ decreased by 83% (IRR 0.17, 95% CI 0.11–0.28) after adjusting for age, sex, comorbidity, type-of-dementia, NPI-subtype-severity, residence, discharge destination and NN discharge support. Although Emergency Department presentation rates increased slightly by 4% (IRR 1.04 95% CI 0.75–1.56), this was not statistically significant.

Hospital service-unit outcomes

Monthly reported rates of falls, pressure injury and OV, across four medical wards for incidents two-years prior the SCU and two-years post were compared (Table 5). As missing cases were 0.6% for falls, 7.2% for OV and nil for pressure injury, they were removed by listwise deletion. Of the total 1273 falls reported, 668 falls were pre-SCU. After adjusting for age (≥65 yrs), cognitive diagnosis, ward, level-of-injury and the exposure variable of fractionally-accrued-bed-days per month, negative-binomial regression estimated a statistically significant decrease in number of falls-per-month by 21% (IRR 0.79, 95% CI 0.64–0.99) with the model also estimating a significant increase in SCU-falls and minor injury; and a significant decrease in falls involving patients with mental health diagnoses. Of a total of 1778 incidents across 4-wards, 930 were post-SCU. When adjusted for age (≥65 yrs), cognitive diagnosis, ward, sex, level-of-injury and the exposure variable of fractionally-accrued-bed-days per month, negative-binomial regression estimated a statistically significant decrease in the monthly rate of OV incidents of 26% (0.76-per-month, 95% CI 0.61–0.96), with lower rates in patients with dementia and/or delirium with poor mobility, cognitively intact patients and for all injury classification levels, while increased rates for SCU patients, those aged >65-years and males, for the period following the introduction of the SCU. No pre-post changes in monthly pressure injury rates were identified.

Discussion

The introduction of a hospital based SCU reduced impact on hospital health services. Mean yearly-hospitalisation-rates for ‘any reason’ decreased by 68% for each SCU-patient after their first discharge. Further, yearly-hospitalisation-rates for BPSD reduced by 83%, providing confidence that the behaviour support plans formulated by the interprofessional team in the SCU were sustainable and effective in community care settings. The lack of difference between the adjusted and unadjusted statistical models in our analysis gives us further confidence that BPSD-management outcomes were not dependent upon diagnosis-type, presentation and discharge destination but instead, related to SCU-care and formulation of an effective behaviour support plan.

The SCU improved hospitalisation outcomes for patients with complex BPSD, reducing overall severity of neuropsychiatric symptoms by 45%, aggression severity by 60% and non-aggressive agitation severity by 40%. These changes were statistically and clinically significant with large effect sizes. Changes over 4-points on the NPI are considered clinically significant [33]. We found a 13-point decrease between SCU-admission and discharge, and a 19-point decrease when comparing SCU-discharge to hospital admission, demonstrating therapeutic BPSD management. Our previously published paper comparing SCU care to standard-ward care found this degree of therapeutic management of BPSD was achieved with reduced psychotropics, physical restraints and security-callout rates [28]. Our results are similar to a recent controlled trial in a Singapore 10-bed SCU, which reported a 50% decrease in agitation [34], providing increasing evidence that a SCU model-of-care provides therapeutic management of BPSD that is more effective than standard-ward care.

Historically, the implementation of hospital based SCUs has been driven by the goal of reducing harm for hospitalised patients with dementia. However, we speculate that SCUs provide broader benefits to the hospital health service at large. A recent Aged Care Royal Commission in Australia [35] and Governmental Report in the United Kingdom [36] highlight an underfunded community-care system struggling with basic dementia care supports and resources. Consequentially, when BPSD threatens safety in community settings, hospital admissions are often the default response. Approximately 77% of patients in our study presented to hospital because of safety issues related to BPSD. While 50% lived at home, another 50% were from Residential Aged Care Facilities (RACFs) and 30% had no previous dementia diagnosis. Yet, general hospital wards are not well suited to managing BPSD [17–19, 21]. Clinical practice guidelines assert that all patients have a right to appropriate assessment and management of their BPSD [37]. However, providing such specialisation requires significant investment in time and resources, the economics of which have largely been overlooked [38]. We argue that hospital based SCUs are uniquely positioned to provide the type of safe, controlled and specialised care setting that is required for assessing, developing and evaluating individualised management of complex BPSD. Hospital-based SCUs can take advantage of existing clinical resources in geriatric medicine and psychiatry and existing safety-response structures within hospitals. Our results suggest the SCU successfully developed BPSD management plans that were individualised, effective, transferable and sustainable in community settings.

Our unit benchmarks well with operation of a well-established SCU in a large Sydney hospital which published almost identical demographics and unit-activity, suggesting results are generalizable [39]. Average patient age for each unit was within 4 years, 62% were male, 52% were from independent living, 32% had Alzheimer’s disease and 16% had Vascular dementia. The average number of NPI per patient was a little over six for both units indicating complex behavioural presentations. Median LoS (23 days versus 21 days) and mortality (nine patients versus seven) were similar. To our knowledge, we are the first publication to differentiate acute from stable BPSD phases, revealing a median acute treatment phase of 11 days (IQR 6–21) and stable phase of seven days (IQR 1–26). Stable-BPSD LoS greater than 4 weeks was mostly due to a lack of RACFs willing or able to accommodate the severity of BPSD. A shortage in RACFs adept at managing complex BPSD may also explain why 20% of SCU-discharges did not return to their pre-admission RACF, with 9% requiring a different Memory Support Unit (i.e. dementia-care unit).

Our results suggest an added benefit of improved outcomes for other inpatients and staff in hospital service-units outside of the SCU. Alongside OV, reducing HACs like falls and pressure injuries is a major priority for hospitals. Our 21% decrease in the rate of falls across general medicine wards was unexpected and suggests that mobile patients with BPSD may negatively impact the ability of general-ward nurses to deliver quality care for all inpatients. By transferring those patients to the SCU, rates of OV and falls decreased even though the SCU-ward was included in the dataset. While a decrease in OV rates was anticipated, the magnitude-of-change (26%) was limited due to other inpatients not targeted by the intervention (e.g. mental health and cognitive impairment with poor mobility). Nevertheless, our results suggest positive financial consequences, as non-injurious falls alone cost $6669 per patient [40] and OV is associated with substantive costs due to workplace injury, absenteeism and nursing turnover [41]. Future studies should explore these variables in an economic analysis. This is the first study to analyse the impact of a hospital-based SCU on general-ward outcomes.

Limitations

Our study has several limitations. First, it is an observational study and cannot assume causality. Second, the falls, pressure injury and occupational violence data was derived from staff incident reporting alone and subject to bias. Third, results are limited to independent mobile patients with dementia. Fourth, being a single site study limits generalizability. To our knowledge, there were no other local hospital or health system initiatives that may have influenced results, and the comparability of our results with a similar unit in a separate city and state, increases confidence that our results are generalisable in the Australian context. Finally, it is not known whether the impacts reported in this study persist beyond the initial 2-year period of SCU operation. To mitigate the risk of dwindling effectiveness, we have developed a staffing structure of rotating clinicians through the SCU (such as geriatricians undertaking 6-month SCU rotations). In this way, a hospital-based gerontological workforce can develop while the model’s success is decoupled from dependencies on any individual clinician. Future evaluations are necessary to establish sustainability of this model in improving outcomes for patients with BPSD and the hospital system.

Conclusion

A hospital based SCU demonstrated positive outcomes in managing independent patients with complex BPSD, a vulnerable patient cohort usually hospitalised for safety concerns related to behaviour. Developing effective BPSD management is often challenging in community and hospital settings. Hospital-based SCUs may provide the requisite specialisation and controlled environment for successful development of individualised plans appropriate to community settings, reducing future hospital health service demand and admissions. Hospital operation of an SCU may also have the added benefit of decreasing patient falls and OV across general medical units.

Declaration of Conflicts of Interest:

None.

Declaration of Sources of Funding:

This Research was funded by a Metro South Health Research Support Scheme Innovation (seed) grant: Metro South Study, Education and Research Trust Account.

The funders played no role in the design and execution of the research.

References