This paper aimed to measure the prevalence and outcomes of delirium for patients over 70 admitted to a general hospital for acute medical care and to assess the validity of the Delirium Rating Scale-Revised-98 (DRS-R-98) in this setting.
This paper aimed to measure the prevalence and outcomes of delirium for patients over 70 admitted to a general hospital for acute medical care and to assess the validity of the Delirium Rating Scale-Revised-98 (DRS-R-98) in this setting.
Prospective study in a British acute general hospital providing sole emergency medical services for its locality. We screened consecutive patients over 70 with an unplanned emergency hospital admission and recruited a cohort of 249 patients likely to have mental health problems. They were assessed for health status at baseline and followed over 6 months. A sub-sample of 93 participants was assessed clinically for delirium.
27% (95% confidence interval (CI) 23–31) of all older medical patients admitted to hospital had DRS-diagnosed delirium, and 41% (95% CI 37–45) had dementia (including 19% with co-morbid delirium and dementia). Compared with clinician diagnosis, DRS-R-98 sensitivity was at least 0.75, specificity 0.71. Compared with reversible cognitive impairment, sensitivity was at least 0.50, specificity 0.67. DRS-diagnosed delirium was associated with cognitive impairment, mood, behavioural and psychological symptoms, activities of daily living, and number of drugs prescribed, supporting construct validity. Of those with DRS-diagnosed delirium, 37% died within 6 months (relative risk 1.4, 95% CI 0.97–2.2), 43% had reversible cognitive impairment, but only 25% had clinically important recovery in activities of daily living. Behavioural and psychological symptoms were common and mostly resolved, but new symptoms frequently developed.
Delirium is common. Some, but not all, features are reversible. DRS-R-98 has reasonable validity in populations where co-morbid dementia is prevalent. Copyright © 2013 John Wiley & Sons, Ltd.
Delirium is a syndrome of disordered cognition and attention or arousal caused by a physical illness, injury, drug or drug withdrawal (American Psychiatric Association, 2000; National Institute for Health and Clinical Excellence, 2010). Other features include fluctuation, disordered thinking, psychomotor agitation or withdrawal, emotional changes, delusions and hallucinations, and change in sleep–wake cycle. Delirium is common in ill older people, particularly among those with pre-existing dementia, in whom the risk is 6–10 times higher than in people who are cognitively intact (Siddiqi et al., 2006; Fick et al., 2002). However, delirium shares features with dementia, which is also common in patients admitted to hospitals (Mukadem and Sampson, 2011; Cole et al., 2002). Many clinicians find it hard to distinguish the two, and delirium often remains undiagnosed. Delirium is associated with poor outcomes such as increased length of hospital stay, institutionalisation and mortality (Siddiqi et al., 2006; Saravay et al., 2004; George et al., 1997; Andrew et al., 2005; McCusker et al., 2003), and does not recover in all cases (Cole et al., 2009).
Several diagnostic and severity measurement scales have been developed to facilitate greater recognition and further study of delirium, but a recent review concluded that ‘not all are properly evaluated in terms of psychometric properties, and there is no unanimity about which scale is best’ (Adamis et al., 2010). The Delirium Rating Scale-Revised-98 (DRS-R-98) (Trzepacz et al., 2001) is one of the better developed (Adamis et al., 2010), but there are limited data on its validity in populations with a high prevalence of dementia or other mental health problems. This paper describes the prevalence and outcome of delirium in older adults admitted for acute medical or trauma care to a British general hospital and assesses the validity of the DRS-R-98 in this setting.
Consecutive patients, over 70, with an unplanned admission to one of 12 wards in a general hospital (three acute geriatric medical, two trauma orthopaedic and seven general medical) were screened between days two and five of admission by using brief tests of cognition, mood and prior diagnosed psychiatric illness (Goldberg et al., 2012). These comprised the abbreviated mental test score (Hodkinson, 1972), four-item Geriatric Depression Scale (Shah et al., 1997), CAGE alcoholism questions (Ewing, 1984) and a question to identify patients with a prior psychiatric diagnosis or behaviours suggesting a mental health problem. Two hundred and fifty patients who were likely to have a mental health problem on the basis of screening were recruited.
Two hundred and forty-nine participants completed a battery of health status measures including history of diagnosed dementia, DRS-R-98, cognition (Mini Mental State Examination (MMSE) (Folstein et al., 1975)), behavioural and psychological symptoms (Neuropsychiatric Inventory (Cummings et al., 1994)), mood (Cornell Scale for Depression in Dementia (Alexopoulos et al., 1988)), acute illness severity (Modified Early Warning Score (Subbe et al., 2001), chronic co-morbidities (Charlson Scale (Charlson et al., 1987)), a frailty index (Study of Osteoporotic Fractures 3-point Scale (Ensrud et al., 2008)) and dependency in activities of daily living (Barthel Index, scored out of 20 (Mahoney and Barthel, 1965)) at the time of admission and prior to the acute illness.
A sub-sample was assessed by a clinician, including 49 psychiatric and 53 geriatric medical evaluations, nine of which were conducted on the same patient. Patient selection for the sub-sample depended on the availability of assessors and the participant still being in hospital and so excluded those with very short lengths of stay. Selection was not on the basis of scale scores, and assessment was blind to other research data.
If patients had mental capacity, they were asked to give written informed consent. For patients assessed as lacking capacity, consultee agreement was sought from a family member or carer.
Baseline information was collected by interview with the participant, carer informants, observation and scrutiny of clinical records by research assistants who were nurses or psychology graduates. Delirium was identified using the standard cut-off of 17.75 or above on DRS-R-98 (‘DRS-diagnosed delirium’).
Clinical diagnostic assessments for delirium were carried out during the same hospital admission, but not on the same day. Assessors were consultants (senior specialists) or senior trainees and were asked to make a thorough clinical assessment, using case notes, questioning of the patient or carers, examination and review of investigations already performed (but did not order any further investigations). Clinicians' diagnoses of delirium (‘clinician-diagnosed delirium’) were made according to DSM-IV criteria (American Psychiatric Association, 2000).
Participants were followed up 180 days after recruitment. Mortality and hospital length of stay were ascertained from hospital administrative systems, general practitioners, care homes or family informants. Those surviving were interviewed at home (or in a care home), and cognitive function, activities of daily living and behavioural and psychological symptoms were reassessed. Those with an improvement of ≥3 on the MMSE compared with the score at admission were considered to have reversible cognitive impairment (Inouye et al., 2006). The study flow diagram is given in Figure 1.
The study was approved by a Research Ethics Committee.
Prevalence of delirium was calculated taking account of the two-stage sampling procedure. We estimated the proportions with delirium alone, delirium superimposed on dementia and dementia alone by considering the difference between the prevalence of cognitive impairment (MMSE < 25) and delirium (DRS-R-98 > 17.75) to estimate the prevalence of dementia alone and used the proportion of general hospital patients with delirium reported in the literature to have prior dementia (67%) (Siddiqi et al., 2006; Fick et al., 2002; Mukadem and Sampson, 2011) to estimate the proportions with delirium alone and delirium superimposed on dementia. Criterion validity of the DRS-R-98 was investigated by comparing DRS-diagnosed delirium with two comparators: the ‘gold standard’ of the independent psychiatric or medical assessment and reversible cognitive impairment on follow-up. Sensitivity and specificity of the DRS-R-98 were calculated, and receiver operating characteristic (ROC) curves drawn. Construct validity was tested using Cronbach's alpha for internal consistency and associations with baseline cognition, behavioural and psychological symptoms, mood, illness severity score, number of drugs taken, admission activities of daily living score, co-morbidity score and frailty score using Spearman's rank correlation coefficients. Predictive validity was tested by comparing outcomes for those with and without DRS-diagnosed delirium, with respect to mortality, recovery in function and behavioural and psychological symptoms, length of hospital stay and readmission.
Of the 1004 patients screened, 361 (36%, 95% confidence interval (CI) 33–39) had no evidence of mental health problems. Two hundred and fifty of the remaining 643 were recruited. One participant withdrew before baseline data were collected. Reasons for non-recruitment were patient with capacity declined consent (23%), unable to contact carer before discharge (17%), no carer identified for patient without capacity (9%), carers declined consultee agreement for patient without capacity (7%), patient too ill (1%) and other (3%). No systematic differences at baseline were identified between those recruited and those not recruited. Those not recruited were similar to those who took part with respect to mean age (84.1 vs 84.1 years), gender (63% vs 66% female), cognition (abbreviated mental test score ≤7 70% vs 71%) and depression (four-item Geriatric Depression Scale ≥1 57% vs 63%).
The median age of patients recruited was 84 years (inter-quartile range: 79–89), 66% were female, 79% were admitted from their own home and 21% from a care home. One hundred and eighteen (47%) participants were recruited from acute geriatric medical wards, 85 (34%) from acute general (internal) medical wards and 46 (18%) from trauma orthopaedic wards. Characteristics of participants from these three ward types were similar.
One hundred and seven (43%) participants had DRS-diagnosed delirium, representing 27% (95% CI 24–31) of all acutely admitted patients over age 70, taking account of the two-stage sampling procedure. We estimate that 9% (of all admissions) had DRS-diagnosed delirium alone, 19% had DRS-diagnosed delirium complicating dementia and 23% had dementia alone. One hundred and six (43%) of the participants (or 27% of all admissions) had a prior diagnosis of dementia, of whom 72 (68%) had DRS-diagnosed delirium (Table 1).
|Delirium (n = 107)||No delirium (n = 142)||p-value|
|Age, years||Median (IQR)||86 (80–89)||83 (78–87)||0.4|
|Female||63 (68)||93 (66)||0.8|
|Accommodation||Care home||43 (40)||9 (6)||<0.001|
|Own home alone||34 (31)||84 (59)|
|Own home with other||30 (28)||49 (35)|
|Barthel Index prior to acute illness||0–5||9 (9)||5 (4)||<0.001|
|6–10||23 (22)||10 (7)|
|11–15||32 (31)||26 (19)|
|16–20||40 (38)||99 (70)|
|Barthel Index at admission||Median (IQR)||6 (3, 11)||12 (7, 15)||<0.001|
|Charlson co-morbidity score||0–1||31 (29)||45 (32)||0.4|
|2–3||50 (47)||54 (38|
|4+||26 (24)||43 (30)|
|SOF frailty index||0–2||78 (74)||116 (82)||0.05|
|3||28 (26)||26 (18)|
|Number of medications||Median (IQR)||6 (5–9)||7 (5–10)||0.2|
|Illness severity score (MEWS)||Median (IQR)||1 (1–2)||1 (1–2)||0.7|
|Diagnosed dementia||72 (68)||34 (32)||<0.001|
|Cognition (MMSE)||Median (IQR)||8 (1, 14)||21 (18, 25)||<0.001|
|Depression (CSDD)||Possible (11–18)||44 (41)||55 (39)||0.003|
|Probable (≥19)||23 (22)||11 (8)|
|Behavioural and psychological symptoms (NPI, severity moderate or marked)||Delusions||27 (25)||4 (3)||<0.001|
|Hallucinations||17 (16)||8 (6)||<0.001|
|Agitation and aggression||32 (29)||3 (2)||<0.001|
|Depression||45 (42)||40 (28)||0.02|
|Anxiety||44 (41)||41 (29)||0.05|
|Elation||2 (2)||1 (1)||0.4|
|Apathy||54 (50)||29 (20)||<0.001|
|Disinhibition||16 (15)||3 (2)||<0.001|
|Irritability||29 (27)||15 (11)||<0.001|
|Psychomotor activity||31 (29)||11 (8)||<0.001|
|Sleep problems||39 (36)||44 (31)||0.3|
|Appetite or eating problems||55 (51)||58 (41)||0.08|
|NPI total||Median (IQR)||33 (20–52)||20 (11–27)||0.02|
Participants with DRS-diagnosed delirium had more severe cognitive impairment, more severe mood disturbance, more delusions, hallucinations, agitation or aggression, apathy and more behavioural and psychological symptoms (Table 1) than the remaining participants (who had all screened positive for a mental health disorder).
Ninety-three participants were assessed for delirium clinically, a mean of 8 days after admission (range 2–17 days). Twenty (22%) had clinician-diagnosed delirium, and 36 (39%) had DRS-diagnosed delirium. The sensitivity of the DRS-R-98 was 0.75 (95% CI 0.56–0.93) and the specificity 0.71 (95% CI 0.66–0.76). ROC analysis indicated that the diagnostic cut-off with the greatest likelihood ratio was 18. The area under the curve was 0.76 (95% CI 0.63–0.88), indicating moderate discriminating ability (Figure 2).
One hundred and twenty-one participants had repeat cognitive function assessment at follow-up, and 40 (33%) demonstrated reversible cognitive impairment (an improvement of ≥3 points on the MMSE). Forty-seven (39%) of those with repeat measurements had DRS-diagnosed delirium at baseline. The sensitivity of the DRS-R-98 to predict reversible cognitive impairment was 0.50 (95% CI 0.28–0.72) and the specificity 0.67 (95% CI 0.46–0.88). ROC analysis identified the cut-off with the greatest validity was between 17 and 18. The area under the curve was 0.65 (95% CI 0.56–0.75), indicating moderate discriminating ability (Figure 1).
Data were available for 249 participants. Cronbach's alpha for the DRS-R-98 was 0.88 for the 13 severity items and 0.87 for the 16-item diagnostic scale, indicating good internal consistency. Correlations with related health status measures demonstrated the following: very strong association between DRS-R-98 scores and cognitive function (MMSE); moderate associations with behavioural and psychological symptoms, admission activities of daily living score and mood; and a weak association with number of drugs taken prior to admission (Table 2). Age, sex, Charlson co-morbidity, frailty index and illness severity score were not associated with DRS-R-98 (Tables 1 and 2).
|Spearman's rank correlation coefficient, rho||p-value|
|Cognitive function (MMSE)||0.80||<0.001|
|Behavioural and psychological symptoms (NPI)||0.45||<0.001|
|Mood (Cornell Scale for Depression in Dementia)||0.25||<0.001|
|Illness severity score (MEWS)||0.0||1.0|
|Number of drugs taken||−0.13||0.04|
|Admission activities of daily living (Barthel Index)||−0.49||0.001|
|Charlson co-morbidity score||−0.02||0.7|
|SOF frailty score||0.09||0.1|
At 180 days, 78 (31%) of the participants had died: 40 (38%) of those with DRS-diagnosed delirium at baseline and 38 (27%) of those without (relative risk 1.4, 95% CI 0.96–2.0, p = 0.09). 20/47 (43%) of survivors with DRS-diagnosed delirium improved by three or more MMSE points at follow-up, compared with 20/74 (27%) of those without (relative risk 1.6, 95%CI 0.95–2.6, p = 0.08). Those with DRS-diagnosed delirium were less likely to improve in activities of daily living (25% vs 36%, relative risk 0.6, 95% CI 0.3–1.1, p = 0.08). Hospital length of stay (median 14 vs 16 days), readmission (42% vs 42%), and proportions moving to a care home (32% vs 20%, p = 0.07) were not statistically significantly different between groups (Table 3).
|Delirium n = 107||No delirium n = 140||p-value|
|Mortality||n (%)||40 (37)||38 (27)||0.09|
|Length of stay in hospital (days)||Median (IQR)||14 (7–32)||16 (6–39)||0.56|
|Hospital re-admissions||n (%)||45 (42)||59 (42)||0.81|
|Moved to care home from community||n (%)||20/63 (32)||26/132 (20)||0.07|
|Barthel Index||Median (IQR)||8 (2, 11)||14 (8, 18)||<0.001|
|Increase in Barthel Index ≥ 3||n (%)||14/ 55 (25)||31/86 (36)||0.08|
|Cognitive function (MMSE)||Median (IQR)||12 (2, 20)||23 (15, 27)||<0.001|
|Increase in MMSE ≥ 3||n (%)||20/47 (43)||20/74 (27)||0.11|
|Behavioural and psychological symptoms (NPI total)||Median (IQR)||13 (7–29)||9.5 (4–22)||0.09|
Behavioural and psychological symptoms were more common in the DRS-diagnosed delirium group at both baseline and follow-up, although individual symptoms, and total Neuropsychiatric Inventory scores, reduced over time in both groups (follow-up NPI total median 13; inter-quartile range 7–29 vs 9.5, 4–22; p = 0.09). For delusions, hallucinations, agitation and aggression, depression, disinhibition, irritability and psychomotor behaviour, this comprised resolution of symptoms in about 80% of cases, persistence (or recurrence) in the remainder and the new appearance of symptoms in some participants who did not have them at baseline. Apathy, sleep and appetite problems improved less often (Table 4). Although symptoms were much more common at baseline in the group with DRS-diagnosed delirium, the pattern of change was similar in those both with and without DRS-diagnosed delirium, and differences in proportional change between the groups were small and not statistically significant.
|Behavioural and psychological symptoms (NPI, severity moderate or marked)||Delirium at baseline n = 57||No delirium at baseline n = 64|
|Baseline||Follow-up||Change in proportion (95%CI)||Improved||Developed||Baseline||Follow-up||Change in proportion (95%CI)||Improved||Developed|
|Delusions||17 (30)||11 (19)||−11 (−28, 7)||14||8||3 (5)||4 (6)||2 (−7, 10)||2||3|
|Hallucinations||8 (14)||7 (12)||−2 (−16, 13)||7||6||4 (6)||1 (2)||−5 (−13, 4)||4||1|
|Agitation and aggression||16 (29)||12 (21)||−7 (−24, 10)||12||8||2 (3)||7 (11)||8 (−2, 17)||1||6|
|Depression||26 (46)||10 (17)||−28 (−45, −10)||21||5||15 (23)||7 (11)||−13 (−26, 0)||12||4|
|Anxiety||25 (44)||13 (23)||−21 (−38, 4)||17||5||17 (27)||8 (13)||−15 (−29, 0)||13||4|
|Elation||2 (4)||2 (4)||0 (−9, 9)||2||2||1 (2)||0 (0)||−2 (−6, 3)||1||0|
|Apathy||25 (44)||20 (36)||−7 (−23, 9)||10||6||15 (23)||17 (27)||3 (−10, 17)||7||9|
|Disinhibition||10 (18)||3 (5)||−13 (−24, 0)||8||1||1 (2)||2 (3)||2 (−5, 8)||1||2|
|Irritability||18 (32)||10 (18)||−13 (−27, 1)||10||3||7 (11)||4 (6)||−5 (−14, 5)||5||2|
|Psychomotor activity||17 (30)||10 (18)||−13 (−27, 2)||11||4||9 (14)||3 (5)||−10 (−19, 0)||7||1|
|Sleep problems||16 (28)||12 (21)||−7 (−23, 9)||10||6||17 (27)||7 (11)||−15 (−29, 0)||14||5|
|Appetite or eating problems||28 (49)||11 (20)||−29 (−46, 12)||19||3||25 (39)||16 (25)||−13 (−28, 2)||14||6|
Of the older patients admitted to a general hospital as an emergency admission, 27% had delirium. Mortality was high, and at follow-up, cognition and activities of daily living often did not improve, even on the modest criteria we set for defining improvement. Behavioural and psychological symptoms were common amongst patients with delirium at the time of admission. Most resolved over the next 6 months, but symptoms also recurred or appeared for the first time, regardless of the presence of delirium at baseline. The DRS-R-98 had reasonable levels of sensitivity and specificity in detecting delirium in a population with co-morbid mental health problems, including dementia. It was associated with measures of cognitive function, behavioural and psychological symptoms, admission activities of daily living, mood and number of drugs taken prior to admission and formed a unidimensional scale. The ROC analyses supported the recommended cut-off of 17.75.
A strength of this study was systematic recruitment and data collection. The population studied was one in which diagnosis of delirium is both important and difficult and in which use of a valid rating scale would be particularly useful. This was a difficult population to recruit and study, however, on account of the fast pace of acute medical admission systems in the UK, the need to identify and recruit a family carer and acute physical illness and cognitive impairment itself. Given the reasons for non-recruitment, more frail and vulnerable patients may have been underrepresented, which could have biased (underestimated) prevalence rates. The DRS-R-98 was chosen over the widely used Confusion Assessment Method (Inouye, 2003), as the Confusion Assessment Method proved difficult to operationalise in this research setting, with multiple researchers and varying levels of prior psychiatric expertise. Two comparators were used for assessing the validity of the DRS-R-98, clinical diagnoses and reversible cognitive impairment. Our results almost certainly underestimate scale validity. Clinical assessments were not made on the same day as the DRS-R-98, so transient cases may have been missed (Cole et al., 2009), and assessments were made on a single occasion, on wards where staff may not have been skilled in detecting features of delirium. The clinical comparator is therefore likely to have been an imperfect ‘gold standard’. Fluctuation and incident delirium may also have compromised validity assessment. Reversible cognitive impairment could only be assessed in survivors, and it is known that not all cases resolve (Siddiqi et al., 2006; McCusker et al., 2003; Cole et al., 2009). Baseline cognition was measured in a hospital ward, which was often busy and noisy, whereas follow-up was at home, where the environment was calmer and quieter, accounting for some improved cognitive scores in the absence of delirium. Inouye also described ‘recoverable cognitive dysfunction’ in acutely ill older patients who did not meet diagnostic criteria for delirium or dementia (Inouye et al., 2006). Both contribute to understating scale validity. In describing prevalence and outcomes, the imperfect validity of the DRS-R-98 will have diminished the precision of estimates of associations and may also have accounted for some of the reversibility seen in the group without DRS-diagnosed delirium. Illness severity was measured using the Modified Early Warning Score, which was designed to be a screening test for severe medical illness and may have inadequately quantified physiological severity. We measured frailty by using the simple 3-point Study of Osteoporotic Fractures Index, which includes weight loss, muscle weakness and fatigue and is an approximation to the Fried frailty phenotype. This may have been too crude to show an association with delirium, although it was strongly associated with mortality and other adverse outcomes in this dataset (unpublished).The study lacked statistical power for some comparisons.
Previous studies have noted poor cognitive and functional recovery after delirium, especially when complicating dementia (Siddiqi et al., 2006; George et al., 1997; Andrew et al., 2005; McCusker et al., 2003; Marcantonio et al., 2003), and it is uncertain whether different or better management can improve this (Pitkälä et al., 2006; Cole et al., 1994; Cole et al., 2002). Much prevalent delirium (present at hospital admission) is unavoidable, but there is a clear imperative to prevent incident delirium (appearing during a hospital stay) where possible (Inouye et al., 1999; Marcantonio et al., 2001). In this study, the sensitivity and specificity of the DRS-R-98 were lower than previously reported in validation studies (Trzepacz et al., 2001; Kato et al., 2010; Huang et al., 2009) but similar to those reported in a study of a similar population (Andrew et al., 2009). This illustrates the need to validate diagnostic scales in the population in which they are used.
The sensitivity and specificity mean that delirium could be missed or misdiagnosed if the DRS-R-98 alone were to be used for detecting delirium in older patients on general hospital wards. Therefore, in clinical practice, the DRS-R-98 therefore can augment, but not replace, a clinical assessment. However, the DRS-R-98 is sufficiently valid to be used in epidemiological studies, where non-differential misclassification causes a loss of statistical power but does not change the direction of association (Hennekens et al., 1987).
Delirium and dementia are closely related among older general hospital patients. In our study, patients with delirium did appear to form a distinct group, on the basis of symptom patterns, suggesting that diagnostic criteria are reasonable. Some reversibility of cognitive impairment and other symptoms was observed, but mortality was high, and symptoms and function often did not reverse. Possible mechanisms include delirium damaging the brain, delirium unmasking previously unrecognised dementia or progression of vascular dementia being mistaken for delirium. The limitations of the diagnostic ability of the DRS-R-98 are likely to be those that also make it difficult for clinicians to diagnose delirium, because of the overlap of features between delirium and dementia. In the case of vascular dementia, there are sudden ‘stepwise’ changes. Dementia with Lewy bodies is characterised by fluctuation, drowsiness and psychosis. Alzheimer's disease is also sometimes associated with psychotic features. In severe dementia of all types, agitation may arise as a consequence of an unfamiliar surroundings, and poor attention, drowsiness and sleep reversal are common. The longer-term prognosis of behavioural and psychological symptoms in delirium has not been previously described. The prevalence of behavioural and psychological symptoms at 180-day follow-up in those with delirium at baseline suggests that this population may suffer recurrent episodes of delirium, with repeated onset and resolution of symptoms, or that symptoms may be related to the type or manifestation of dementia, prior personality, life events or some other enduring feature of their disease, rather than the onset of delirium. This emphasises the difficulty facing practitioners assessing a confused older person in a crisis at home or in a care home, where the need to exclude delirium drives many acute hospital admissions.
Delirium research remains somewhat neglected, despite its high prevalence and associations with poor outcomes. Future research should attempt to characterise or define delirium more closely. New diagnostic tests for delirium are needed. Rigorous evaluations of interventions to improve outcome are also required.
This paper presents independent research funded by the UK National Institute for Health Research (NIHR) under its Programme Grants for Applied Research funding scheme (RP-PG-0407-10147). The views expressed are those of the authors and not necessarily those of the National Health Service, the NIHR or the Department of Health. The funders and study sponsors had no role in study design, collection, analysis and interpretation of data; writing the manuscript; or the decision to submit for publication. The authors thank the NIHR Mental Health Research Network, Trent Dementia Research Network and the Trent Comprehensive Local Research Network for support with recruitment. Clinical assessments were made by Drs Emily Laithwaite, Nicola Watson, Sujata Das, Mitra Raisi and Bipin Ravindran. The Medical Crises in Older People study group also included Justine Schneider, Simon Conroy, Anthony Avery, Judi Edmans, Adam Gordon, Bella Robbins, Jane Dyas, Pip Logan, Rachel Elliott and Matt Franklin.