https://orcid.org/0000-0002-9568-3195
Abstract
Systematically review and critically appraise the evidence for the association between delirium and falls in community-dwelling adults aged ≥60 years.
We searched EMBASE, MEDLINE, PsycINFO, Cochrane Database of Systematic Reviews, CINAHL and Evidence-Based Medicine Reviews databases in April 2023. Standard methods were used to screen, extract data, assess risk of bias (using Newcastle–Ottawa scale), provide a narrative synthesis and, where appropriate, conduct meta-analysis.
We included 8 studies, with at least 3505 unique participants. Five found limited evidence for an association between delirium and subsequent falls: one adjusted study showed an increase in falls (risk ratio 6.66; 95% confidence interval (CI) 2.16–20.53), but the evidence was low certainty. Four non-adjusted studies found no clear effect. Three studies (one with two subgroups treated separately) found some evidence for an association between falls and subsequent delirium: meta-analysis of three adjusted studies showed an increase in delirium (pooled odds ratio 2.01; 95% CI 1.52–2.66); one subgroup of non-adjusted data found no clear effect. Number of falls and fallers were reported in the studies. Four studies and one subgroup were at high risk of bias and one study had some concerns.
We found limited evidence for the association between delirium and falls. More methodologically rigorous research is needed to understand the complex relationship and establish how and why this operates bidirectionally. Studies must consider confounding factors such as dementia, frailty and comorbidity in their design, to identify potential modifying factors involved. Clinicians should be aware of the potential relationship between these common presentations.
Key Points
This is the first systematic review of the association between delirium and falls in the wider community population.
There is limited but consistent evidence on the direction of effect for delirium preceding falls and falls preceding delirium.
More high-quality longitudinal work is needed to explore the nature of this potentially complex and bidirectional relationship.
History of falls and delirium should be considered when assessing patients with incidence/suspected incidence of falls/delirium.
Background
Falls, defined as ‘an unexpected event in which the participants come to rest on the ground, floor, or lower level’ [1], affect nearly one-third of community-dwelling adults aged ≥65 years each year. This rises to >50% for those aged ≥80 [2–4]. While most fall-related injuries are minor, in the UK >223 000 falls in people aged ≥65 resulted in hospital admissions between 2021 and 2022 [5]. This has a personal burden in terms of pain, injury, fear of falling, loss of confidence and independence, and higher mortality. Falls in the community are estimated to cost the National Health Service >£1.7 billion per year [6, 7].
Delirium is a condition of acute onset, causing altered attention and awareness with additional disturbances in cognition, which may fluctuate due to underlying medical causes [8]. The prevalence of delirium in the community is estimated at 1%–2%, increasing to 14% in people aged >85 years [9, 10]. In long-term care facilities, the prevalence of delirium among people aged ≥65 years is estimated at 10%–40% [11]. However, it is often under-detected and underdiagnosed in the community and sometimes misdiagnosed as other conditions including dementia, depression and psychosis [12]. Delirium causes considerable burdens in terms of functional or cognitive decline in individuals and economic burden to the healthcare system due to increased risk of hospitalisation, higher levels of care and institutionalisation [13–15].
Delirium and falls share common risk factors including older age, frailty, prior history of falls, impaired balance and gait, visual and auditory impairment, cognitive impairment and polypharmacy [16, 17]. The relationship between delirium and falls can be complex and bidirectional.
In hospital settings, there is an increased incidence of falls in patients with delirium and an increased risk of delirium in people who fall. A systematic review [18] reported a higher risk of falls for inpatients with delirium than those without delirium across 10 studies (median risk ratio (RR) 54.5, range 1.4–12.6). A recent cross-sectional study analysing the association between delirium and falls in a hospital screening programme with >29 000 patients [19] found that those who screened positive for delirium during admission had a significantly increased risk of falling while they were an inpatient (adjusted odds ratio (OR) 2.81 [95% confidence interval (CI): 2.12–3.70]). Delirium screening is recommended as a standard part of fall care pathways [18, 19].
However, little is known about the association between delirium and falls in community settings. Given that falls are a common reason for pre-hospital service use and hospital admission, this is important because there is the potential to reduce hospital admissions and healthcare costs [20]. There is no available systematic review considering the relationships between the incidence of falls and the occurrence of delirium in the community. Our objective was to conduct a rigorous systematic review of the association between delirium and falls in community settings.
Methods
We followed Cochrane methods for systematic reviews and report the review using the Preferred Reporting Items for a Systematic Review and Meta-analysis (PRISMA) guidelines [21]. The review protocol was prospectively registered on PROPSERO (http://www.crd.york.ac.uk/PROSPERO/ registration number: CRD42022309982).
Search strategy
A literature search, guided by an information specialist, was conducted in April 2023, using strategies developed around the facets of delirium, falls and older people (for full search strategy, see Supplementary material S1). Searches were completed using the following databases: EMBASE (Ovid), MEDLINE (Ovid), PsycINFO (Ovid), Cochrane Database of Systematic Reviews (Ovid), CINAHL (EBSCO), and Evidence-Based Medicine Reviews (EBMR) (Ovid). The search was limited to articles published from 1995 onwards when the delirium diagnosis in the Diagnostic and Statistical Manual of Mental Disorders-IV (DSM-IV) criteria was updated [22]. Web of Science and Google Scholar were also searched and restricted to the first 200 relevant records. Searches were limited to English-language publications. References and forward citations were checked for additional relevant publications.
Inclusion and exclusion criteria
We included studies that met the following criteria (full details in Supplementary material S2).
Population: Adults aged ≥60 years, living in community or supported living/residential care settings.
Variables: Studies reporting evidence on the association between delirium and falls, regardless of which preceded the other.
Study design: Observational studies, randomised and non-randomised controlled trials.
Study selection
Search records were imported into Rayyan [23] and duplicate records removed. Two reviewers (L.M. and B.P.) independently screened titles and abstracts and the full texts of potentially relevant studies. Disagreements were resolved through discussion with a third reviewer (C.E-T.).
Data extraction
A predefined data extraction form was used to extract study characteristics, see Box 1. Two reviewers independently extracted data from included studies (L.M. and Y.Y.) and disagreements were resolved through discussion with a third reviewer (C.E-T. or E.R.L.C.V.).
Data extraction items.
Basic characteristics of studies, including first author, study date, study location, study aim, study design (type, duration, setting), publication type, publication year
Characteristics of participants, including population type and place of residence, inclusion criteria, the number of participants, mean age of participants, gender of participants and ethnicity of participants
Variable measurements (assessment tools/method/scales, time points reported) for both delirium and falls
Temporal association between falls and delirium, i.e. whether falls or delirium were recorded first
Falls outcomes: number of falls, number of fallers, falls rate per person per year, time to first fall, number of injurious falls
Delirium outcomes: number of delirium cases detected, number of people with delirium, number of delirium episodes per person, duration, severity and type of delirium (hyperactive, hypoactive or mixed)
Statistical methods used, and handling of missing data
Statistical analysis of results
Quality assessment
Two reviewers (L.M. and Y.Y.) independently assessed the quality of included studies using the Newcastle–Ottawa Scale (NOS) [24, 25]. We considered using alternative tools, e.g. Quality In Prognosis Studies (QUIPS) [24] and Risk Of Bias In Non-randomised Studies of Exposures (ROBINS-E) [26], but they were less appropriate for this review due to the study designs of the included evidence, i.e. studies not set up specifically as prognostic studies or studies of exposure. Disagreements were resolved in consultation with a third reviewer (C.E-T.).
Scores were assigned to studies based on the quality of selection criteria, comparability of groups and outcome (for cohort and cross-sectional) or exposure (for case–control), with a maximum score of 9 for case–control and cohort studies and 8 for cross-sectional studies. The overall score was used to rate the risk of bias; however, studies were also given an overall high risk of bias if any single domain was rated as high risk.
We considered the certainty of the evidence, guided by the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach for prognostic reviews [27–29], although the nature of the evidence (i.e. not designed as prognostic studies) meant that we did not conduct a full GRADE assessment.
Risk of bias, inconsistency of the results, indirectness of the evidence, imprecision of the statistical analysis and publication bias for each outcome were considered. In assessing imprecision, we accounted for the number of studies contributing to evidence, the size of the studies and width of the confidence intervals in each study.
Data analysis and synthesis
Data analysis was structured by the temporal association between delirium and falls. We report separately on data from studies where delirium preceded the recorded falls (D-F) and studies where falls preceded a record of delirium (F-D). For the D-F studies, which were prospective in design, we presented RRs or incidence risk ratios and 95% CI. For the F-D studies, which included retrospective fall data, we presented the ORs and 95% CI.
We conducted narrative syntheses and present individual studies’ data for outcome measures on forest plots. Where appropriate, we conducted meta-analyses in RevMan5.4 [30] using random-effects models with the generic inverse variance method.
Results
Study selection
After screening 971 records on title and abstract and 21 on full text, eight studies met the inclusion criteria for this review; see Figure 1 for the PRISMA flowchart [21].
PRISMA flowchart of study selection process.
Study and participant characteristics
Eight studies including at least 3505 unique participants from five countries were included in this review. Five reported on evidence where recorded delirium preceded falls (D-F) [31–35]. Three studies, one of which stratified data into two subgroups according to type of residence (nursing home or community) and which will be treated as two separate data sets in this review, reported evidence when falls preceded recorded delirium (F-D) [36–38]. Settings included residential homes and individual homes, and follow-ups ranged from 1 to 15.5 months. The main study and participant characteristics are summarised in Table 1, and a full description of study characteristics and additional study outcomes can be found in Supplementary S3.
Study characteristics
| Delirium-falls studies | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Study location | Study aim | Study design | Variable measurement | ||||||||
| Type | Duration | Study setting | Inclusion criteria | Number of participants | Age (mean years ± SD) | Gender | Ethnicity | Delirium | Falls | |||
| Eriksson et al. (2007) [31] | Sweden | Identify risk factors for falls in older people with and without a diagnosis of dementia living in residential care facilities (we used only non-dementia subgroup) | Prospective cohort study | 6-month follow-up | 4 residential homes | Inclusion: • All residents in the care facilities >65 years Exclusion: • Unclear dementia diagnosis • Age < 65 years | Older adults without dementia n = 83 | 83.5 ± 7.1 | Male: n = 30 (36.1%) | Not reported | Nurse recorded the presence of delirium in past month | Record of falls incidents |
| Kallin et al. (2002) [33] | Sweden | Identification of predisposing and precipitating factors for falls and recurrent falls among frail older people living in residential care facilities | Cross-sectional with prospective follow-up (treated as cohort study) | 12-month follow-up | 1 residential home | Inclusion: • All residents in the care facility | Older adults n = 83 | 79.6 ± 8.6 | Male: n = 25 (30.1%) | Not reported | Medical records for past delirium | Record of falls incidents and recurring falls |
| Kallin et al. (2004) [32] | Sweden | Study precipitating factors for falls among older people living in residential care facilities | Prospective cohort study | 12-month follow-up | 5 residential care facilities | Inclusion: • All residents in the care facilities | Older adults n = 199 | 82.4 ± 6.8 | Males: n = 59 (29.6%) | Not reported | Diagnosed by physician using DSM-IV criteria | Record by staff of falls incidents and recurring falls |
| Mahoney et al. (2000) [34] | USA | Evaluate the rate of falls, and associated risk factors, for 90 days following hospital discharge | Prospective cohort study | 13 weeks | Patient’s home | Inclusion: • Adults ≥65 years discharged from hospital after medical illness, receiving home health care Exclusion: • In hospice care or terminal cancer • New cerebrovascular accident (CVA) or myocardial infarction (MI) in previous 2 months • Dementia and no caregiver in home • Between hospital discharge and home health care • Non-ambulatory • Other incl. no phone, unable to speak English | Older adults n = 311 | 80.0 ± 7.1 | Male = 116 (37.2%) | • White = 96.8% • African American = 2.2% | Confusion Assessment Method (CAM) | Self-reported record of falls on a calendar, with follow-up postcards and telephone interviews |
| von Heideken Wagert et al. (2009) [35] | Sweden | Describe the incidence of falls and fall-related injuries and identify predisposing factors for falls in very old people | Prospective cohort study | 6-month follow-up | Ordinary and institutional housing | Inclusion: • Residents aged ≥85 years (random sample from half of those >85 years and all residents >90 years) | Older adults n = 220 (n = 109 in ordinary housing, n = 111 institutional housing) | 90.3 ± 4.8 | Males n = 53 (24%) | Not reported | N/R | For those in ordinary housing self-report on a calendar and follow-up telephone calls For those in institutional settings, falls reported by staff were reviewed from records |
| Boorsma et al. (2012) (Nursing home subgroup) [36] | The Netherlands | Prevalence and incidence of delirium and its risk factors | Retrospective cohort | 4 years 7 months Observation time 10.8 months | 6 nursing homes | Inclusion: • All residents in the nursing home Exclusion: • If observations missed information about delirium | n = 828 | No mean age recorded Older age (>85) n = 257 (31.0%) | Males n = 274 (33.1%) | Not reported | The Nursing Home–Confusion Assessment Method (NH-CAM) [50] | Fall incidents (at least one fall incident in the last 90 days, yes/no) |
| Boorsma et al. (2012) (Residential home subgroup) [36] | The Netherlands | Prevalence and incidence of delirium and its risk factors | Retrospective cohort | 4 years 7 months Observation time 15.5 months | 23 residential homes | Inclusion: • All residents in the residential homes Exclusion: • If observations missed information about delirium | n = 1365 | No mean age recorded Older age (>85) n = 706 (51.7%) | Males n = 346 (25.3%) | Not reported | Positive score on the Nursing Home–Confusion Assessment Method (NH-CAM) [50] | At least one fall incident in the last 90 days yes/no |
| Perez-Ros et al. (2019) [37] | Spain | Identify delirium predisposing and triggering factors and develop a predictive model | Case–control study (relevant data for this review is treated as retrospective cohort) | 12 months | 6 nursing homes | Inclusion: • Residents ≥65 years, living in nursing home during study Exclusion: • Spent <1 year in the home • Spent >2 months out of the home during study • End-of-life status | n = 443 | Mean 85.76 ± 6.7 years | Males n = 96 (21.7%) | Not reported | Geriatrician diagnosed according to DSM-IV criteria and/or Confusion Assessment Method (CAM) | Number of falls during the study period |
| Sabbe et al. (2021) [38] | Belgium | To investigate (point) prevalence of and risk factors for delirium in nursing homes | Cross-sectional (relevant data for this review is treated as a retrospective cohort) | 1 month | 6 nursing homes | Inclusion: • Residents of nursing homes ≥65 years Exclusion: • In coma • Aphasia • End-of-life status • Specifically on dementia ward | n = 338 | Mean 84.7 ± 8.0 years | Males n = 110 (32.5%) | Not reported | 13-item Delirium Observation Screening Scale (DOSS) scale based on DSM-IV-TR criteria | Fall incident in previous 90 days—from records |
| Delirium-falls studies | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Study location | Study aim | Study design | Variable measurement | ||||||||
| Type | Duration | Study setting | Inclusion criteria | Number of participants | Age (mean years ± SD) | Gender | Ethnicity | Delirium | Falls | |||
| Eriksson et al. (2007) [31] | Sweden | Identify risk factors for falls in older people with and without a diagnosis of dementia living in residential care facilities (we used only non-dementia subgroup) | Prospective cohort study | 6-month follow-up | 4 residential homes | Inclusion: • All residents in the care facilities >65 years Exclusion: • Unclear dementia diagnosis • Age < 65 years | Older adults without dementia n = 83 | 83.5 ± 7.1 | Male: n = 30 (36.1%) | Not reported | Nurse recorded the presence of delirium in past month | Record of falls incidents |
| Kallin et al. (2002) [33] | Sweden | Identification of predisposing and precipitating factors for falls and recurrent falls among frail older people living in residential care facilities | Cross-sectional with prospective follow-up (treated as cohort study) | 12-month follow-up | 1 residential home | Inclusion: • All residents in the care facility | Older adults n = 83 | 79.6 ± 8.6 | Male: n = 25 (30.1%) | Not reported | Medical records for past delirium | Record of falls incidents and recurring falls |
| Kallin et al. (2004) [32] | Sweden | Study precipitating factors for falls among older people living in residential care facilities | Prospective cohort study | 12-month follow-up | 5 residential care facilities | Inclusion: • All residents in the care facilities | Older adults n = 199 | 82.4 ± 6.8 | Males: n = 59 (29.6%) | Not reported | Diagnosed by physician using DSM-IV criteria | Record by staff of falls incidents and recurring falls |
| Mahoney et al. (2000) [34] | USA | Evaluate the rate of falls, and associated risk factors, for 90 days following hospital discharge | Prospective cohort study | 13 weeks | Patient’s home | Inclusion: • Adults ≥65 years discharged from hospital after medical illness, receiving home health care Exclusion: • In hospice care or terminal cancer • New cerebrovascular accident (CVA) or myocardial infarction (MI) in previous 2 months • Dementia and no caregiver in home • Between hospital discharge and home health care • Non-ambulatory • Other incl. no phone, unable to speak English | Older adults n = 311 | 80.0 ± 7.1 | Male = 116 (37.2%) | • White = 96.8% • African American = 2.2% | Confusion Assessment Method (CAM) | Self-reported record of falls on a calendar, with follow-up postcards and telephone interviews |
| von Heideken Wagert et al. (2009) [35] | Sweden | Describe the incidence of falls and fall-related injuries and identify predisposing factors for falls in very old people | Prospective cohort study | 6-month follow-up | Ordinary and institutional housing | Inclusion: • Residents aged ≥85 years (random sample from half of those >85 years and all residents >90 years) | Older adults n = 220 (n = 109 in ordinary housing, n = 111 institutional housing) | 90.3 ± 4.8 | Males n = 53 (24%) | Not reported | N/R | For those in ordinary housing self-report on a calendar and follow-up telephone calls For those in institutional settings, falls reported by staff were reviewed from records |
| Boorsma et al. (2012) (Nursing home subgroup) [36] | The Netherlands | Prevalence and incidence of delirium and its risk factors | Retrospective cohort | 4 years 7 months Observation time 10.8 months | 6 nursing homes | Inclusion: • All residents in the nursing home Exclusion: • If observations missed information about delirium | n = 828 | No mean age recorded Older age (>85) n = 257 (31.0%) | Males n = 274 (33.1%) | Not reported | The Nursing Home–Confusion Assessment Method (NH-CAM) [50] | Fall incidents (at least one fall incident in the last 90 days, yes/no) |
| Boorsma et al. (2012) (Residential home subgroup) [36] | The Netherlands | Prevalence and incidence of delirium and its risk factors | Retrospective cohort | 4 years 7 months Observation time 15.5 months | 23 residential homes | Inclusion: • All residents in the residential homes Exclusion: • If observations missed information about delirium | n = 1365 | No mean age recorded Older age (>85) n = 706 (51.7%) | Males n = 346 (25.3%) | Not reported | Positive score on the Nursing Home–Confusion Assessment Method (NH-CAM) [50] | At least one fall incident in the last 90 days yes/no |
| Perez-Ros et al. (2019) [37] | Spain | Identify delirium predisposing and triggering factors and develop a predictive model | Case–control study (relevant data for this review is treated as retrospective cohort) | 12 months | 6 nursing homes | Inclusion: • Residents ≥65 years, living in nursing home during study Exclusion: • Spent <1 year in the home • Spent >2 months out of the home during study • End-of-life status | n = 443 | Mean 85.76 ± 6.7 years | Males n = 96 (21.7%) | Not reported | Geriatrician diagnosed according to DSM-IV criteria and/or Confusion Assessment Method (CAM) | Number of falls during the study period |
| Sabbe et al. (2021) [38] | Belgium | To investigate (point) prevalence of and risk factors for delirium in nursing homes | Cross-sectional (relevant data for this review is treated as a retrospective cohort) | 1 month | 6 nursing homes | Inclusion: • Residents of nursing homes ≥65 years Exclusion: • In coma • Aphasia • End-of-life status • Specifically on dementia ward | n = 338 | Mean 84.7 ± 8.0 years | Males n = 110 (32.5%) | Not reported | 13-item Delirium Observation Screening Scale (DOSS) scale based on DSM-IV-TR criteria | Fall incident in previous 90 days—from records |
Study characteristics
| Delirium-falls studies | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Study location | Study aim | Study design | Variable measurement | ||||||||
| Type | Duration | Study setting | Inclusion criteria | Number of participants | Age (mean years ± SD) | Gender | Ethnicity | Delirium | Falls | |||
| Eriksson et al. (2007) [31] | Sweden | Identify risk factors for falls in older people with and without a diagnosis of dementia living in residential care facilities (we used only non-dementia subgroup) | Prospective cohort study | 6-month follow-up | 4 residential homes | Inclusion: • All residents in the care facilities >65 years Exclusion: • Unclear dementia diagnosis • Age < 65 years | Older adults without dementia n = 83 | 83.5 ± 7.1 | Male: n = 30 (36.1%) | Not reported | Nurse recorded the presence of delirium in past month | Record of falls incidents |
| Kallin et al. (2002) [33] | Sweden | Identification of predisposing and precipitating factors for falls and recurrent falls among frail older people living in residential care facilities | Cross-sectional with prospective follow-up (treated as cohort study) | 12-month follow-up | 1 residential home | Inclusion: • All residents in the care facility | Older adults n = 83 | 79.6 ± 8.6 | Male: n = 25 (30.1%) | Not reported | Medical records for past delirium | Record of falls incidents and recurring falls |
| Kallin et al. (2004) [32] | Sweden | Study precipitating factors for falls among older people living in residential care facilities | Prospective cohort study | 12-month follow-up | 5 residential care facilities | Inclusion: • All residents in the care facilities | Older adults n = 199 | 82.4 ± 6.8 | Males: n = 59 (29.6%) | Not reported | Diagnosed by physician using DSM-IV criteria | Record by staff of falls incidents and recurring falls |
| Mahoney et al. (2000) [34] | USA | Evaluate the rate of falls, and associated risk factors, for 90 days following hospital discharge | Prospective cohort study | 13 weeks | Patient’s home | Inclusion: • Adults ≥65 years discharged from hospital after medical illness, receiving home health care Exclusion: • In hospice care or terminal cancer • New cerebrovascular accident (CVA) or myocardial infarction (MI) in previous 2 months • Dementia and no caregiver in home • Between hospital discharge and home health care • Non-ambulatory • Other incl. no phone, unable to speak English | Older adults n = 311 | 80.0 ± 7.1 | Male = 116 (37.2%) | • White = 96.8% • African American = 2.2% | Confusion Assessment Method (CAM) | Self-reported record of falls on a calendar, with follow-up postcards and telephone interviews |
| von Heideken Wagert et al. (2009) [35] | Sweden | Describe the incidence of falls and fall-related injuries and identify predisposing factors for falls in very old people | Prospective cohort study | 6-month follow-up | Ordinary and institutional housing | Inclusion: • Residents aged ≥85 years (random sample from half of those >85 years and all residents >90 years) | Older adults n = 220 (n = 109 in ordinary housing, n = 111 institutional housing) | 90.3 ± 4.8 | Males n = 53 (24%) | Not reported | N/R | For those in ordinary housing self-report on a calendar and follow-up telephone calls For those in institutional settings, falls reported by staff were reviewed from records |
| Boorsma et al. (2012) (Nursing home subgroup) [36] | The Netherlands | Prevalence and incidence of delirium and its risk factors | Retrospective cohort | 4 years 7 months Observation time 10.8 months | 6 nursing homes | Inclusion: • All residents in the nursing home Exclusion: • If observations missed information about delirium | n = 828 | No mean age recorded Older age (>85) n = 257 (31.0%) | Males n = 274 (33.1%) | Not reported | The Nursing Home–Confusion Assessment Method (NH-CAM) [50] | Fall incidents (at least one fall incident in the last 90 days, yes/no) |
| Boorsma et al. (2012) (Residential home subgroup) [36] | The Netherlands | Prevalence and incidence of delirium and its risk factors | Retrospective cohort | 4 years 7 months Observation time 15.5 months | 23 residential homes | Inclusion: • All residents in the residential homes Exclusion: • If observations missed information about delirium | n = 1365 | No mean age recorded Older age (>85) n = 706 (51.7%) | Males n = 346 (25.3%) | Not reported | Positive score on the Nursing Home–Confusion Assessment Method (NH-CAM) [50] | At least one fall incident in the last 90 days yes/no |
| Perez-Ros et al. (2019) [37] | Spain | Identify delirium predisposing and triggering factors and develop a predictive model | Case–control study (relevant data for this review is treated as retrospective cohort) | 12 months | 6 nursing homes | Inclusion: • Residents ≥65 years, living in nursing home during study Exclusion: • Spent <1 year in the home • Spent >2 months out of the home during study • End-of-life status | n = 443 | Mean 85.76 ± 6.7 years | Males n = 96 (21.7%) | Not reported | Geriatrician diagnosed according to DSM-IV criteria and/or Confusion Assessment Method (CAM) | Number of falls during the study period |
| Sabbe et al. (2021) [38] | Belgium | To investigate (point) prevalence of and risk factors for delirium in nursing homes | Cross-sectional (relevant data for this review is treated as a retrospective cohort) | 1 month | 6 nursing homes | Inclusion: • Residents of nursing homes ≥65 years Exclusion: • In coma • Aphasia • End-of-life status • Specifically on dementia ward | n = 338 | Mean 84.7 ± 8.0 years | Males n = 110 (32.5%) | Not reported | 13-item Delirium Observation Screening Scale (DOSS) scale based on DSM-IV-TR criteria | Fall incident in previous 90 days—from records |
| Delirium-falls studies | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Study location | Study aim | Study design | Variable measurement | ||||||||
| Type | Duration | Study setting | Inclusion criteria | Number of participants | Age (mean years ± SD) | Gender | Ethnicity | Delirium | Falls | |||
| Eriksson et al. (2007) [31] | Sweden | Identify risk factors for falls in older people with and without a diagnosis of dementia living in residential care facilities (we used only non-dementia subgroup) | Prospective cohort study | 6-month follow-up | 4 residential homes | Inclusion: • All residents in the care facilities >65 years Exclusion: • Unclear dementia diagnosis • Age < 65 years | Older adults without dementia n = 83 | 83.5 ± 7.1 | Male: n = 30 (36.1%) | Not reported | Nurse recorded the presence of delirium in past month | Record of falls incidents |
| Kallin et al. (2002) [33] | Sweden | Identification of predisposing and precipitating factors for falls and recurrent falls among frail older people living in residential care facilities | Cross-sectional with prospective follow-up (treated as cohort study) | 12-month follow-up | 1 residential home | Inclusion: • All residents in the care facility | Older adults n = 83 | 79.6 ± 8.6 | Male: n = 25 (30.1%) | Not reported | Medical records for past delirium | Record of falls incidents and recurring falls |
| Kallin et al. (2004) [32] | Sweden | Study precipitating factors for falls among older people living in residential care facilities | Prospective cohort study | 12-month follow-up | 5 residential care facilities | Inclusion: • All residents in the care facilities | Older adults n = 199 | 82.4 ± 6.8 | Males: n = 59 (29.6%) | Not reported | Diagnosed by physician using DSM-IV criteria | Record by staff of falls incidents and recurring falls |
| Mahoney et al. (2000) [34] | USA | Evaluate the rate of falls, and associated risk factors, for 90 days following hospital discharge | Prospective cohort study | 13 weeks | Patient’s home | Inclusion: • Adults ≥65 years discharged from hospital after medical illness, receiving home health care Exclusion: • In hospice care or terminal cancer • New cerebrovascular accident (CVA) or myocardial infarction (MI) in previous 2 months • Dementia and no caregiver in home • Between hospital discharge and home health care • Non-ambulatory • Other incl. no phone, unable to speak English | Older adults n = 311 | 80.0 ± 7.1 | Male = 116 (37.2%) | • White = 96.8% • African American = 2.2% | Confusion Assessment Method (CAM) | Self-reported record of falls on a calendar, with follow-up postcards and telephone interviews |
| von Heideken Wagert et al. (2009) [35] | Sweden | Describe the incidence of falls and fall-related injuries and identify predisposing factors for falls in very old people | Prospective cohort study | 6-month follow-up | Ordinary and institutional housing | Inclusion: • Residents aged ≥85 years (random sample from half of those >85 years and all residents >90 years) | Older adults n = 220 (n = 109 in ordinary housing, n = 111 institutional housing) | 90.3 ± 4.8 | Males n = 53 (24%) | Not reported | N/R | For those in ordinary housing self-report on a calendar and follow-up telephone calls For those in institutional settings, falls reported by staff were reviewed from records |
| Boorsma et al. (2012) (Nursing home subgroup) [36] | The Netherlands | Prevalence and incidence of delirium and its risk factors | Retrospective cohort | 4 years 7 months Observation time 10.8 months | 6 nursing homes | Inclusion: • All residents in the nursing home Exclusion: • If observations missed information about delirium | n = 828 | No mean age recorded Older age (>85) n = 257 (31.0%) | Males n = 274 (33.1%) | Not reported | The Nursing Home–Confusion Assessment Method (NH-CAM) [50] | Fall incidents (at least one fall incident in the last 90 days, yes/no) |
| Boorsma et al. (2012) (Residential home subgroup) [36] | The Netherlands | Prevalence and incidence of delirium and its risk factors | Retrospective cohort | 4 years 7 months Observation time 15.5 months | 23 residential homes | Inclusion: • All residents in the residential homes Exclusion: • If observations missed information about delirium | n = 1365 | No mean age recorded Older age (>85) n = 706 (51.7%) | Males n = 346 (25.3%) | Not reported | Positive score on the Nursing Home–Confusion Assessment Method (NH-CAM) [50] | At least one fall incident in the last 90 days yes/no |
| Perez-Ros et al. (2019) [37] | Spain | Identify delirium predisposing and triggering factors and develop a predictive model | Case–control study (relevant data for this review is treated as retrospective cohort) | 12 months | 6 nursing homes | Inclusion: • Residents ≥65 years, living in nursing home during study Exclusion: • Spent <1 year in the home • Spent >2 months out of the home during study • End-of-life status | n = 443 | Mean 85.76 ± 6.7 years | Males n = 96 (21.7%) | Not reported | Geriatrician diagnosed according to DSM-IV criteria and/or Confusion Assessment Method (CAM) | Number of falls during the study period |
| Sabbe et al. (2021) [38] | Belgium | To investigate (point) prevalence of and risk factors for delirium in nursing homes | Cross-sectional (relevant data for this review is treated as a retrospective cohort) | 1 month | 6 nursing homes | Inclusion: • Residents of nursing homes ≥65 years Exclusion: • In coma • Aphasia • End-of-life status • Specifically on dementia ward | n = 338 | Mean 84.7 ± 8.0 years | Males n = 110 (32.5%) | Not reported | 13-item Delirium Observation Screening Scale (DOSS) scale based on DSM-IV-TR criteria | Fall incident in previous 90 days—from records |
Quality appraisal
Delirium-falls studies
Using the appropriate NOS, four studies were rated as having a high risk of bias and one study had a low risk of bias (Figure 2 and Supplementary S4a). The highest risk arose because studies did not account for possible confounding factors in their analysis of the association between delirium and falls [31–33, 35].
Forest plot of the association between prior delirium and number of falls.
Falls-delirium studies
One study [37] and one study subgroup had a low risk of bias [36], the other study subgroup was at high risk of bias due to the lack of adjustment for confounding variables relating to falls and delirium [36]. The third study had some concerns over the lack of information provided on participant selection and outcome measures [38] (Figure 3 and Supplementary S4b).
Forest plot of the association between prior falls and delirium.
Evidence for associations
Delirium-falls studies
All five studies reported the number of people with delirium (five studies [31–35]); none reported the number of delirium episodes per person, duration, severity or type of delirium. Falls outcomes were reported as the number of falls (four studies [31–33, 35]), number of fallers (five studies [31–35]), falls rate (four studies [31, 33–35]), time to first fall (one study [35]) and the number of injurious falls (three studies [32, 34, 35]). Two studies also reported the prevalence of recurrent falls [33, 34].
One study (n = 311) presented data adjusted for confounders, including number and type of medication, cognitive impairment, balance score and needing help with activities of daily living (see Supplementary S5 for full details), for the association between delirium and falls [34]. The other four studies (total n = 585) presented unadjusted data [31–33, 35]. Meta-analysis was not considered appropriate for the outcomes in these studies because of high heterogeneity in study methodology and analysis, i.e. whether the data were adjusted for confounding variables. All studies are presented in Figure 2 to illustrate effect sizes only.
Association between delirium and falls
The study adjusted for confounding variables reported a significant association between delirium and an increase in falls in a home setting within 13 weeks of discharge from hospital (RR = 6.66 [95% CI: 2.16–20.53]), Figure 2 [34]. However, the evidence was considered low certainty due to high imprecision.
The remaining four studies reported unadjusted data on the number of falls, all undertaken within the same regional population. Data across all unadjusted studies was consistent in the direction of effect but not statistically significant. The largest and most recently published of these studies (n = 220) [35] showed no clear association between delirium and the number of falls recorded within 6 months (n = 220, RR = 1.75 [95% CI: 0.89–1.90]). Evidence from all these studies was very low certainty due to the high risk of bias and high imprecision. One non-adjusted study found that participants with a record of delirium in the previous month had a significantly higher risk of falling more than once compared to not falling or falling only once (RR 2.73 [95% CI: 1.05–7.07]; P = .039) [33]. However, this study was considered very low quality, and we have low confidence in the findings.
Falls-delirium studies
Three studies (one with two subgroups) reported on falls occurring prior to a recorded delirium episode. Falls outcomes reported the number of fallers (three studies [36–38]). No studies reported on whether fallers experienced single or recurrent falls, the fall rate, time to first fall or number of injurious falls. Delirium outcomes reported the number of delirium cases (two studies [36, 37]), number of people with delirium (three studies [36–38]) and the number of delirium episodes (one study [37]). No studies reported the duration, severity or type of delirium.
Association between falls and delirium
Two studies [30, 31] and one study subgroup [36] presented data adjusted for confounders on delirium risk (see Supplementary S5 for details). The second study subgroup presented unadjusted data [36]. A random-effects meta-analysis (Figure 3) indicated that the studies adjusted for confounding variables, two at low risk of bias and one with some concerns, showed a significant association between falls and increased risk of delirium (OR 2.01, 95% CI: 1.52–2.66). We chose the random-effects model due to variations in study design, participants and included confounders, suggesting the possibility of multiple true effects. However, we acknowledge the potential for small study bias with this method. While the evidence was graded as high quality, we recommend interpreting it with caution. Although no evidence of publication bias was found, and we did not downgrade for it, it is important to note that all the studies were retrospective in design, making them more vulnerable to publication bias [39]. Sensitivity analysis showed the significant effect remained after the study with some risk-of-bias concerns was removed (OR 1.95, 95% CI: 1.38–2.76). Evidence from the subgroup with unadjusted data was low certainty (due to a high risk of bias and imprecision) and resulted in inconclusive evidence.
Discussion
Summary of results
To our knowledge, this systematic review is the first to examine the evidence on the association between delirium and falls in community settings. Eight studies, encompassing at least 3505 unique participants from five high-income countries, were predominantly conducted in residential care settings.
Due to the challenges in establishing the temporal association between delirium and falls, we included evidence from studies where delirium preceded falls (D-F) and where falls preceded delirium (F-D). We suggest that in the D-F studies, the delirium may have presented too far in advance, up to 6–12 months in some studies, to impact the falls directly although there may be an ongoing longer-term effect on fall risk. Conversely, in the F-D studies it is possible that the delirium was present when the fall took place and was subsequently recorded during the follow-up period.
From the five D-F studies, only one study adjusted for confounders, including cognitive impairment, polypharmacy and balance, showed a significant increase in the number of falls associated with a prior diagnosis of delirium, albeit with low-certainty evidence. The four unadjusted studies showed no clear association although the directions of effect were consistent with the adjusted data.
Pooled evidence from the three F-D studies accounting for confounding factors (e.g. age, gender and cognitive impairment) showed a significant association between delirium and an increased falls risk. These studies were high quality, but with a caveat that they were retrospective and therefore highly susceptible to a number of biases, discussed below. The single subgroup presenting unadjusted data showed an inconclusive association between falls and delirium, but a consistent direction of effect. Our findings suggest that the association between delirium and falls, previously identified in hospital settings [18, 19], is also relevant in community settings.
Confounders such as cognitive impairment, polypharmacy, frailty and comorbidity can significantly influence both the incidence of falls and the occurrence of delirium. Not adjusting for these factors can lead to biassed results because the observed association may be attributed to the confounders rather than a direct link between falls and delirium. Adjusting for confounders enables a more accurate assessment of the true relationship, helping to isolate the specific impact of delirium on fall risk and vice versa.
There was considerable heterogeneity in the design of the included studies included, and also in the populations they included. Settings varied: six studies were based in care homes or nursing homes, two in both care homes and participants’ own homes, and one exclusively in participants’ own homes. This variation likely impacts the speed and effectiveness of detecting delirium and falls, potentially introducing bias in the measured outcomes. Additionally, participant populations differed; some studies included individuals with conditions such as Parkinson’s disease, dementia and a history of previous falls while others only included healthy older adults. This further complicates the comparison of findings across studies, potentially impacting the results and reducing generalisability.
Furthermore, we recognised that there was a possible overlap in participants between three of the D-F studies, which were based in the same location (Sweden) and used data from an ongoing longitudinal study [31–33]. Multiple studies drawn from the same participant pool increase the risk of data redundancy, inflated sample size and overclaiming the strength of association. However, since none of these studies accounted for confounders in their analysis, the data were deemed unsuitable for meta-analysis and were not included in our overall conclusions regarding the evidence for the D-F studies.
Strengths and limitations
This systematic review has a number of strengths. We followed transparent and robust methods to summarise the available evidence for the association between delirium and falls in community settings. We included studies where delirium preceded documented falls, and where falls occurred before a recorded episode of delirium, to ensure our review was comprehensive. Mirroring the approach for prognostic studies, we included observational studies: cohort, cross-sectional and case–control design. We assessed the risk of bias using the appropriate NOS, considering both the overall rating and individual domain ratings in our judgements. Evidence was assessed for certainty using an approach based on GRADE for prognostic studies as a general guide (although GRADE assessments were not specifically undertaken) [29]. This was a review of association and cannot be considered a prognostic review due to the limitations of the primary study evidence.
There are some limitations to this review and the included studies. We incorporated observational studies utilising retrospective data and relying on patient records. Clarity of the data collection and recording process was sometimes insufficient. Such studies are susceptible to recall bias, incomplete record-keeping and potential inaccuracies [40–42]. Furthermore, they may be influenced by selection biases including self-selection, non-response and attrition, areas largely overlooked in the discussions of included studies. Finally, studies reporting associations are highly vulnerable to publication bias, as findings may go unreported if no association between variables is observed [32].
We found a predominance of supported care settings in the studies. The definition of residential/nursing homes varies in different countries, and the level of support, health and care provision, and independent living varied across studies. The lack of studies reporting data from people’s own homes may be due to the lower incidence and prevalence of recorded delirium and falls in these settings [9, 11]. Very large studies with substantial datasets would be needed to accurately establish any association in the home setting, and many episodes of delirium and falls in the community are not accurately recorded in healthcare records.
Several factors make the comparison of findings between studies less robust. No study was designed to focus specifically on the link between delirium and falls, potentially missing key variables and limiting the evidence’s directness. A wide range of assessment tools were used to measure delirium and sometimes the methods for identifying and assessing delirium were not clearly specified. No study recorded the type of delirium participants experienced (i.e. hypoactive or hyperactive), and none included a measure of the severity of the delirium or explored how these aspects impacted falls. Definitions of delirium and falls were not universally presented in the papers, and we also included studies with author’s own definition or no definition.
Due to the complex relationship between delirium and falls, it is crucial to consider confounders when analysing data to ensure the validity and reliability of the findings. Of the three F-D studies presenting data adjusted for confounders, two included dementia in their multivariate models and one included cognitive impairment (but not dementia) (see Supplementary S5 for a full list of variables considered in the analysis). This is important because the coexistence of dementia is a potential confounding factor in the development of delirium, and delirium is a major factor in the development of dementia [43]. The studies had a follow-up period of between 1 and 16 months. However, there is a lack of data on the longer-term effects of the relationship between falls and delirium in a community setting.
None of the studies in this review made any consideration of equity factors, as identified in the PROGRESS Plus framework [44] in their analysis and interpretation of the data. This is crucial because environmental and equity factors may affect both falls and delirium, and some populations may be at greater risk; these factors were outside the scope of the current review.
Finally, we only searched for peer-reviewed publications written in English. We may have missed unpublished studies, pre-prints and studies in other languages. Together with the risk of publication bias, this increases the risk that relevant studies may not be included in this review, that the pattern of missingness is non-random and that the strength of the associations reported here may be an overestimate of the true effects.
Conclusions and further research
We found a limited amount of evidence for an association between a record of delirium and an increase in falls and some evidence for an association between falls and an increased risk of a recorded delirium episode. However, these findings are based on predominantly retrospective studies with relatively small numbers of participants, and we note the potential for publication bias impacting these types of studies. The limitations of the studies and data do not allow a direct comparison between the relative risks of D-F and F-D, but the evidence in this review identifies a potential association in both adjusted and unadjusted data. Nevertheless, the true effect may differ considerably from the estimates presented here as further evidence is highly likely to substantially alter the effect estimate. This is due to issues with study power, design and potential underreporting in the literature of data showing no associations.
Our review highlights the need for more methodologically rigorous research to quantify the relationship between falls and delirium, accounting for important potential confounders or mediators such as dementia, frailty and polypharmacy. Prospective studies including an adequate number of people regularly monitored for delirium and ensuring all falls are recorded would allow for a more robust understanding of the temporal sequence of falls and delirium. In line with good practice as recommended by the COMET initiative [45] which promotes the use of standardised outcome measures for clinical trials, we recommend the use of standardised definitions and assessment measures for delirium and falls, e.g. CAM [46] or 4AT [47] for delirium and the ProFaNE core outcome dataset for falls research [1]. More studies focusing explicitly on the association between delirium and falls, adjusting for important confounding factors, would strengthen the body of high-certainty evidence.
Further research is needed to understand the potentially complex relationship between delirium and falls, aiming to establish whether, how and in what way this may operate bidirectionally and identifying potential modifying factors. Studies should assess the potential contribution of frailty, cognitive impairment and polypharmacy to any relationship. Recognition and focus on equity factors (e.g. socioeconomic, gender, cultural aspects, healthcare access and environmental factors) in studies may also help to identify groups who are more at risk of delirium and falls and allow for tailored interventions to meet the needs of underserved populations. Lastly, it is also imperative that studies finding null results are fully reported and published.
This work may have implications for clinical practice and policy including the implementation of routine screening for both delirium and falls to identify and treat preventable risks. In hospital settings, it is recommended that patients aged ≥65 years are screened on admission using 4AT for delirium and assessed for fall risk [48]. Community pathways for screening and diagnosis of delirium or falls are less well established [20, 49]. It is essential that the delirium–falls link is considered, and clinicians should be alert to the potential relationship between the two.
Declaration of Conflicts of Interest:
None declared.
Declaration of Sources of Funding:
This research was funded by the National Institute for Health and Care Research through the Applied Research Collaboration Greater Manchester (NIHR ARC-GM) (NIHR200174) and a Senior Investigator Award to C.T. (NIHR200299). C.E-T., Y.Y., C.S., S.A., A.M., C.T. and E.R.L.C.V. are funded/part-funded by the NIHR ARC-GM. The views expressed in this publication are those of the authors and not necessarily those of the National Health Service, the National Institute for Health and Care Research, the Department of Health and Social Care, or its partner organisations. The funders played no role in the design, execution, analysis or interpretation of data or writing the study.
S.S. is funded by The Advanced Care Research Centre (ACRC). The ACRC is funded by Legal and General PLC as part of their corporate social responsibility (CSR) programme. The funders had no role in the design, conduct or interpretation of the study and the views expressed are those of the authors.
Research data transparency and availability:
This is a review of previously published studies; therefore, all the data used in the review are already in the public domain. All data relevant to the included studies are included in the article or uploaded as supplementary information.
Comments