Falls in focus: an analysis of the rate of falls in 25 Australian residential aged care facilities from 2019 to 2021, pre- and during COVID-19 lockdowns

Abstract

Introduction

During 2020–2021 Australia maintained comparatively low rates of COVID-19 in the community and residential aged care facilities (RAC) due to stringent public health measures such as lockdowns. However, the public health measures implemented may have had unintended impacts on critical RAC resident health outcomes, such as falls, due to routine care disruptions and aged care resident isolation. We utilised a longitudinal data to assess the association between COVID-19 lockdowns and the rate of falls in RAC settings.

Methods

A longitudinal cohort study was conduct using routinely collected data from 25 RAC facilities from one non-profit aged care provider in Sydney, Australia. The study included 2,996 long term residents between March 2019 and March 2021. The outcome measures were all falls, injurious falls, and falls assessed as requiring hospitalisation. Generalised estimating equations (GEE) were applied to determine the association between COVID-19 lockdown periods and fall-related outcomes while adjusting for confounders and seasonality.

Results

During the study period 11,658 falls were recorded. Residents frequently experienced at least one fall during the study period (median: 1, interquartile range: 0–4). During Lockdown 1 (March-June 2020) the rate of all falls increased 32% (IRR 1.32, 95% CI 1.19–1.46, p < 0.01) and the rate of injurious falls increased by 28% (IRR 1.28, 95% CI 1.12–1.46, p < 0.01) compared to pre-pandemic rates. The rate of falls assessed as requiring hospitalisation remained unchanged during Lockdown 1 (IRR 1.07, 95% CI 0.86–1.32, p = 0.519). During Lockdown 2 (Dec 2020-Jan 2021) the rate of all falls, injurious falls, and falls assessed as requiring hospitalisation did not change significantly compared to pre-pandemic rates.

Conclusion

These findings suggest that the consequences of stringent COVID-19 restrictions, as seen in Lockdown 1, produced changes in residents’ care which contributed to more falls and associated harm. The subsequent lockdown, which were less restrictive and occurred after staff had gained experience, was associated with no significant increase in falls rate. The nature and extent of lockdowns implemented for infection control in RAC need to balance multiple potential adverse effects. Factors which facilitated resilience during this period require exploration in future research.

Introduction

The COVID-19 pandemic triggered a crisis in residential aged care facilities (RAC) (i.e., nursing homes and long-term facilities) internationally as RAC residents were more susceptible to the negative outcomes associated with COVID-19 and aged care services were often poorly positioned to cope with an increase in care needs [1,2,3]. Globally, aged care services have long faced systemic issues such as chronic under resourcing, limited governance, poor design, and infection control standards [4, 5]. The scale of the disaster in RAC is demonstrated in global COVID-19 mortality statistics. In 2021, it was estimated that COVID-19 related deaths in RAC accounted for 30% of the COVID-19 death toll in 38 Organisation for Economic Co-operation and Development nations even though RAC populations frequently make up less than 1% of the overall population in these countries [3].

The Australian experience of COVID-19 in RAC between 2020 and 2021 differed from the international experience due to comparatively low rates of COVID-19. Between 2020 and late 2021 Australia, a country with a population of over 25 million people including 180,000 living in RAC [6], reported fewer than 400,000 COVID-19 cases approximately 3,700 of these occurred in RAC facilities [7]. By comparison, in December 2021 regions of the United Kingdom (UK) estimated that 5% of the population had an active COVID-19 infection in a single week [8]. Australia was able to maintain low rates of COVID-19 in 2020–2021 largely due to strict public health measures which included: heavily restricted international and inter-state travel; work from home orders; mask mandates; high levels of testing and isolation requirements for COVID-19 close contracts and COVID-19 cases [9,10,11]. In Australian RAC facilities, key public health restrictive measures included facility lockdowns (i.e., limiting community access and visitors), routine COVID-19 testing of staff and residents, isolation requirement for COVID-19 cases and close contracts, and the use of personal protective equipment (PPE) for residents, staff, and visitors [9, 10].

While public health measures implemented in Australian RAC from 2020 to 2021 protected residents and the health system from COVID-19, they may have inadvertently harmed RAC residents by disrupting care routines and isolating residents from their social networks. During lockdowns, Australian RAC facilities paused allied health and lifestyle and leisure services, such as group exercise, music, and mental stimulation activities and community outings, sometimes for months at a time [12, 13]. Services which continued during Australian RAC lockdowns, such as support for activities of daily living (i.e., feeding and showering), were reportedly carried out with poorer quality (e.g., delayed care and limited flexibility, and choice for residents) due to staffing limitations [14, 15]. Residents also spent more time alone as family and friends were not permitted to enter facilities, communal activities were restricted, and other forms of social connection were limited [12, 16, 17]. While these measures were effective in reducing COVID-19 outbreaks, these public health measures are also believed to have negatively impacted the quality and quantity of care, and physical, and psychological health of residents in RAC facilities [16].

The rate of falls is a critical measure of care quality and resident harm in RAC settings that may have been impacted by the pandemic. Research conducted in community settings indicates that falls increased in older adults during waves of COVID-19 cases due to an increase in fall risk factors including loneliness and physical inactivity [18]. However, limited research has been conducted on falls in RAC settings. The association between public health restrictions and falls in RAC settings is critical to explore to understand the potential risks and benefits of implementing facility lockdowns in future infectious disease outbreaks. To fill this knowledge gap, we aimed to investigate rates of fall and their outcomes during COVID-19 lockdowns in Australian RAC facilities from 2020 to 2021 compared to pre-pandemic using routinely collected data.

Methods

Setting and design

We conducted this study using a longitudinal retrospective cohort analysis before and after COVID-19 lockdown periods in Sydney, Australia (March 2019 to March 2021) using electronic health record data of 25 RAC facilities. All facilities belonged to the same not-for-profit aged care provider and were in Sydney.

This study is reported according to the REporting of studies Conducting using Observational Routinely-collected health Data (RECORD) guidelines [19]. The study forms part of a broader research project which investigates the application of predictive analytics and decision support to prevent falls in aged care [20,21,22]. Ethics approval, including a waiver of participant consent for the use of routinely collected data, was received from Macquarie University Human Research Ethics Committee (Project ID: 6144).

Participants

Study inclusion criteria were limited to aged care residents who resided in the participating RAC facilities for > 30 days between March 2019 and March 2021. Respite residents were excluded from the analysis.

Data source

De-identified resident electronic health records were obtained from the aged care provider. Data sets included in this study were resident demographic data (e.g., gender, health conditions at admission, age, date of entry and exist into aged care) and incident report datasets containing information related to falls (e.g., time and date of fall, injurious/non-injurious fall, and need to attend hospital following the fall). Health conditions in this data set were recorded as free text, and we applied a health macro, developed by our research team, to categorise the health conditions [23]. Information on the dates of COVID-19 restrictions was sourced from national and state government websites [9, 10].

Outcome measures

The outcomes were all falls, falls resulting in injury, and falls requiring hospitalisation. All falls was defined as the occurrence of any fall, as entered in the resident’s incident report, regardless of whether they resulted in injury. Injurious falls were falls resulting in any injury as assessed and entered in the incident report by the responding RAC staff member. Injurious falls include minor injury (e.g., skin tears), and major injury, (e.g., hip fracture). Falls assessed as requiring hospitalisation were falls assessed as requiring hospitalisation by RAC staff. Falls requiring hospitalisation is a check box in the fall’s incident form at the RAC provider and not the actual number of people who attend hospital. All outcomes were expressed as counts in the dataset.

COVID-19 lockdown periods

A monthly categorical variable was generated to describe pre-COVID-19, COVID-19 lockdown periods, and periods of eased restriction during COVID-19 in Sydney, Australia (Fig. 1). The dates of COVID-19 lockdowns and periods of eased restrictions were derived from New South Wales (NSW) Health public health orders [9, 10].

Fig. 1

COVID-19 lockdowns in Sydney Australia

In the community in Sydney, Australia, during Lockdown 1 businesses such as gyms and restaurants closed, public events and gatherings were banned, businesses transitioned to work from home, and social distancing was implemented in all settings. During this time in RAC, staff wore PPE, group activities (e.g., group meals and social and exercise groups) were limited, residents spent more time in their rooms, and residents had restricted or no access to external visitors [12, 16]. By the end of Lockdown 1 in NSW Australia, there had been 61 COVID-19 cases in RAC settings [24].

Lockdown 2 was triggered by an outbreak of COVID-19 in the community in the Northern Beaches Local Government Area of greater Sydney. While none of the facilities included in this study were in the Northern Beaches, COVID-19 restrictions were tightened in all health facilities, including RAC facilities, across NSW. NSW public health orders during this time stated that RAC facilities must exclude all visits from family and friends, conduct routine COVID-19 testing on staff and residents, and restrict staff to work in only one health facility. RAC facilities however were encouraged to increase allied health staff ratios to facilitate social, mental, and physical stimulation [25]. No COVID-19 cases were recorded in NSW RAC facilities during Lockdown 2 [9, 10].

In between COVID-19 lockdown periods in 2020 and 2021 COVID-19 restrictions eased in the community and RAC facilities. However, some restrictions remained in place. For example, RAC facilities often continued to limit the number of external visitors, and while group activities resumed, they often were modified to enforce infection control policies such as social distancing and mask wearing [26]. Additionally, staff needed to comply with increased use of PPE, compared to pre-pandemic, and all staff, visitors, and residents needed to isolate for two weeks if they tested positive to COVID-19 or were in close contact with someone who had tested positive.

Statistical analysis

We used descriptive statistics such as medians, interquartile ranges (IQR), and counts to report cohort characteristics. We assessed the distribution of time-invariant covariates during COVID-19 lockdowns and pre-COVID-19 using chi-squared analysis for categorical variables and Kruskal-Wallis tests for continuous variables as continuous variables, such as age, are not normally distributed in this cohort. Multicollinearity between variables was assessed with a 0.80 correlation coefficient and at a 95% confidence level (P < 0.05). Unadjusted fall outcome descriptive statistics are presented either over the total study period or monthly, as falls incidence is too rare an event to present over smaller time periods (i.e., days or weeks). For analysis, all fall count data were converted to monthly rates per 1,000 residents to account for monthly fluctuations in cohort size.

We applied Generalised Estimating Equations (GEE) to model the association between COVID-19 lockdowns and each outcome measure (all falls, injurious falls, falls requiring hospitalisation) while adjusting for confounders. Several time-invariant and time-variant variables were considered in the analysis. The time-invariant variables included age, gender, medical conditions, and length of stay in residential aged care. Four binary COVID-19 indicator variables were created and entered as time-variant variables in the analysis. The GEE model was clustered by resident. We adjusted for seasonality using Fourier terms [27]. We applied robust standard errors to accommodate for the panel nature of the data and negative binomial log link function to handle the over-dispersed data. We used an unstructured working correlation matrix to allow for unconstrained correlations between measures. The GEE calculated incident rate ratio (IRR) and 95% confidence intervals (CI). Statistical significance was assessed at P < 0.05. We applied established model selection metrics, quasilikelihood under the independence model criterion (QIC), to select the best working correlation structure and subset of explanatory variables [28]. All analyses were conducted using Stata Version 18 (StataCorp LP, College Station, TX).

Results

The study sample included 2,996 residents over the two-year period. Table 1 presents cohort characteristics. Approximately three quarters of participants were female, and half had a history of falls and dementia. During the study period residents frequently experienced one fall (median: 1, IQR: 0–4). No statistical differences were observed in time-invariant demographics when comparing the cohort admitted during the pandemic compared to the cohort before the pandemic (Appendix 1).

Table 1 Cohort demographics (n = 2,996)*

Incidence of falls before and during the pandemic

Over two years the cohort experienced a total of 11,658 falls. The number of falls each month ranged from 401 (December 2020) to 589 (March 2020). The proportion of residents who experienced a fall each month during the study period ranged from 13.5% (April 2019) to 17.8% (May 2020) (Table 2).

Over the study period, 35.0% (n = 4,076) of all falls resulted in injury (injurious fall). The total number of injurious falls per month ranged from 135 (August 2018) to 206 (March 2020). The proportion of residents who experienced an injurious fall each month ranged from 5.9% (June 2019) to 8.5% (May 2020).

During the two-year period, 11.3% of all falls (n = 1,313) were assessed as requiring hospitalisation. The number of falls requiring hospitalisation per month ranged from 31 (June 2019) to 81 (March 2021). The proportion of residents who experienced falls requiring hospitalisation per month ranged from 1.7% (April 2019) to 3.9% (March 2021).

Table 2 Number of fall related incidents per month and number of residents who experienced the incident

The rate of fall-related outcomes during COVID-19

During Lockdown 1 there was a 32% increase in the rate of falls compared to pre-pandemic (IRR 1.32, 95% CI 1.19–1.46, p < 0.01) (Table 3). The increase in all falls was also reflected in the rate of injurious falls which increased by 28% compared to pre-pandemic (IRR 1.28, 95% CI 1.12–1.46, p < 0.01). Despite this, the rate of falls assessed as requiring hospitalisation did not significantly change by the first COVID-19 lockdown (IRR 1.07, 95% CI 0.86–1.32, p = 0.519).

The rate of all falls remained high following the end of Lockdown 1 compared to pre-pandemic (IRR 1.15, 95% CI 1.04–1.28, p < 0.01). The rate of injurious falls also remained slightly elevated. However, this was not statistically significant (IRR 1.12 0.98–1.27, p = 0.09). The rate of falls assessed as requiring hospitalisation remained unchanged following the end of the first lockdown (IRR 1.05, 95% CI 0.85–1.29, p = 0.68).

In Lockdown 2 the fall rate returned to pre-pandemic levels (IRR: 0.95, 95% CI 0.83–1.09, p = 0.44), including for falls resulting in injury (IRR: 0.95 95% CI 0.81–1.13, p = 0.58). This trend continued into the period following Lockdown 2. Following the end of the Lockdown 2, the rate of falls assessed as requiring hospitalisation increased by 32% compared to pre-pandemic (IRR 1.32 95% CI 1.02–1.70, p = 0.03).

Table 3 GEE results for all falls, injurious falls, and falls assessed as requiring hospitalisation for COVID-19 indicator variables (n = 2,996).^A

Discussion

Lockdown 1 was associated with 32% increase in the rate of all falls and a 28% increase in injurious falls compared to the 12 months prior to COVID-19. Despite the significant increase in falls, there was not increase in the rate of residents assessed as requiring hospitalisation during Lockdown 1 The rate of all falls, injurious falls, and falls assessed as requiring hospitalisation remained unchanged compared to pre-pandemic rates in Lockdown 2. However, interestingly, after Lockdown 2 there was a 32% increase in falls assessed as requiring hospitalisation without an increase in the rate of all falls.

International community-based research has attributed the increase in falls during waves of COVID-19 to physical deconditioning caused by either lower limb weakness post COVID-19 infection or physical inactivity due to lockdowns or self-imposed isolation [18, 29, 30]. Physical deconditioning may have also played a role in the increase in all falls observed in our RAC cohort. A Canadian study of almost 200,000 older adults in RAC found that residents were 20% more likely to experience functional decline during the first lockdown of the pandemic compared to pre-pandemic [31]. However, in our RAC cohort, the factors causing physical deconditioning are likely to vary from the community-setting and some of the experiences of residents in RAC in other countries. First, the incidence of COVID-19 infection was low in our cohort and therefore side effects from COVID-19 infection are not likely to have significantly impacted the overall rate of falls. Second, RAC residents may have limited ability to self-impose an increase or decrease in physical activity as they often require assistance to facilitate or participate in activity. Instead, in RAC settings, residents may have physically deconditioned as they received less opportunity to participate in physical activity as routine exercise classes and falls prevention activities performed by allied health professionals were paused and because nursing staff had less time to assist with mobility [12, 13]. An analysis of over 8,000 RAC facilities in the UK found that staff shortages during the pandemic were associated with greater loss of mobility among residents compared to those who in facilities which maintained pre-pandemic staffing levels [32]. Finally, in RAC settings, compared to the community, there are likely to be a number of factors, beyond physical deconditioning, that may have contributed to an increase in falls during Lockdown 1. Factors such as disrupted routines and separation from friends and family during the pandemic increased RAC resident agitation and delirium, particularly for residents with dementia, and may have heightened the risk of falls [33]. Additionally, prescribing patterns of fall-risk increasing drugs such as psychotropics and opioids prescribing increased during lockdowns in healthcare settings, if the prescribing patterns also changed in RAC facilities this may have further increased the risk of falls [34, 35].

In our analysis, the rate of falls assessed as requiring hospitalisation did not increase during Lockdown 1 despite an overall increase in all falls and injurious falls during the same period. Interestingly, after Lockdown 2 the rate of falls assessed as requiring hospitalisation increased 32% compared to pre-pandemic despite no increase in the rate of all falls. Using routinely collected data, other Australia researchers have noted a reduction in hospital transfers from Australian RAC facilities during COVID-19 outbreaks. The researchers suggested that RAC facility imposed hospital-in-the-home policies, established during the initial lockdowns, for COVID-19 positive cases may have influenced staff members assessment to refer the resident to hospital for all other conditions [36]. International evidence also suggests that staff hesitation to recommend transfer to hospital may also reflect resident preferences. A study from the United States reported that almost half residents updated their advanced care plan during the pandemic to avoid hospitalisation [37]. Irrespective of the factors driving lower rates of falls assessed as requiring hospitalisation, the reduction is potentially concerning as residents may not have received necessary medical care post-fall. Further qualitative and quantitative research, such as analysis of linked hospital data and staff and resident interviews, is required to understand how COVID-19 lockdowns impacted resident transfer to hospital and health outcomes. This research is important to understand how to manage and monitor hospital transfers in future facility lockdowns and infectious disease outbreaks.

In our analysis, COVID-19 lockdowns were not associated with consistent changes in the rate of all falls in RAC. Falls increased during Lockdown 1 but not in Lockdown 2, a finding that was also observed among older people in the community [38]. As other authors have suggested, the variability in the relationship between falls and COVID-19 restrictions may be attributed to shifts in policies, behaviours, and perceptions of COVID-19. In our cohort, Lockdown 2 may not have impacted falls because the lockdown was only implemented for a short period of time and staff and residents were acclimatised to infection control practices. Additionally, public health orders in Lockdown 2 placed an emphasis on continuing care as normal in RAC facilities, including the rostering of additional lifestyle and allied health staff [25]. The inconsistency in the relationship between lockdowns and the rate of falls suggests that not all lockdowns have a negative impact on resident health outcomes. In the future more measured approaches such as short lockdown periods, visitor screening, and support to continue routine activities with heightened infection control standards may balance the impact of the disease outbreaks and the negative impacts of service disruption and isolation.

The strengths of our study include its longitudinal design, use of person level data, sample size, and division of COVID-19 into significant periods rather than a pre-post cross-sectional study. However, some weaknesses exist. Our study was not able to gather linked hospital data, therefore in this analysis falls are only assessed as requiring hospitalisation and the exact rate at which residents attended hospital is unknown. Additionally, person level data such as past medical history is only entered at admission into RAC and is not updated throughout the stay. Therefore, past medical history entered in the analysis is time-invariant and likely underestimates the true number of people with any of the medical conditions included in our analyses. The analysis also relies on incident reports entered by aged care staff and therefore it may misrepresent the true number of falls occurring or be influenced by trends in reporting practices. Lastly, we were not able to collect and analyse data on contextual factors, such as staffing levels, that may have affected the fall rates. Australian RAC facilities had limited mandatory data collection standards during the study period. As a result, we can only speculate, drawing on available qualitative studies, about other factors could have influenced the rate of falls in our cohort. RAC data collection is improving under new mandates and more detailed modelling could be possible in the future.

Conclusion

Our analysis revealed distinct patterns in falls within RAC settings during COVID-19 lockdowns. During Lockdown 1 all falls increased 32% and injurious falls increased 28% compared to pre-pandemic, however, falls assessed as requiring hospitalisation did not change. During Lockdown 2 fall rates were not significantly different from pre-pandemic rates. The patterns in this analysis indicate the need for further research into, and careful consideration of specific areas of success and concern. Namely, how the extent and nature of lockdowns can be implemented with minimal impact on other resident outcomes in future infectious disease outbreaks.