RESEARCH ARTICLE

Prognosis of delirium in hospitalized elderly: worse than

we thought
Monidipa Dasgupta1,2 and Chris Brymer1
1
Division of Geriatric Medicine, Department of Medicine, Shulich School of Medicine, University of Western Ontario, London, ON, Canada
2
Lawson Health Research Institute, London, ON, Canada
Correspondence to: Monidipa Dasgupta, MDCM, E-mail: [email protected]

Background: Despite treatment of the associated condition, delirious persons do not always recover for
unknown reasons. We sought to determine early prognostic indicators of poor recovery following an
episode of delirium in older medical in-patients.
Methods: Between October 2009 and July 2011, consecutively admitted older (≥70 years old) medical
in-patients at the London Health Sciences Centre (Ontario) were screened for delirium. Delirious
patients were followed. The primary outcome was poor recovery, in delirious patients, defined by
death, long-term institutionalization, or functional decline (decreased activities of daily living), at
discharge or 3 months after discharge, elicited from the medical chart or post-discharge caregiver
telephone interviews.
Results: One thousand two hundred thirty-five in-patients (mean age 82.6 years, 42% men) were
screened, delirium occurred in 355 (29%). Follow-up data was known on 342 (96%), and 237 (69%)
had poor recovery: 55 died (54 in hospital and one after discharge), 136 were permanently institutionalized
(86 directly from hospital and 50 after discharge), and 46 had functional decline (at a median of 103 days
after discharge). Poor recovery was associated in the derivation sample with advanced age, lower baseline
function, hypoxia, higher delirium severity scores, and acute renal failure; this was predictive of poor
recovery in the validation sample (receiver operating characteristic area 0.68, 95% confidence interval:
0.57–0.79); however, even individuals with “low” risk had high (50%) poor recovery rates.
Interpretation: Poor recovery after delirium is common and associated with certain characteristics.
However, even “lower risk” delirious individuals do poorly. More research is needed to understand
prognostic factors in delirium. Copyright # 2013 John Wiley & Sons, Ltd.
Key words: delirium; older; hospitalization; risk factors; prognosis
History: Received 23 May 2013; Accepted 6 September 2013; Published online 3 October 2013 in Wiley Online Library
(wileyonlinelibrary.com)
DOI: 10.1002/gps.4032

Introduction 1992; Rockwood et al., 1999; Marcantonio et al.,

2000; McCusker et al., 2002).
Delirium is a commonly occurring acute disorder of Although it is well accepted that delirious individ-
attention and cognition associated with many illnesses uals do worse than non-delirious individuals, it is
(Lipowski, 1983; Francis and Kapoor, 1992; Inouye not understood why some individuals recover after
et al., 1993; Rockwood, 1993). It dramatically increases delirium, whereas others do not (Andrew et al.,
the length and cost of hospitalization (cost estimated in 2005). Delirium was originally described as a poten-
the USA in 2004 at $6.9bn, Inouye, 2006). Independent tially reversible condition with cognitive and func-
of confounding factors, delirium has been shown to tional recovery expected as the underlying medical
increase functional decline, institutionalization, and condition was appropriately treated (Lipowski, 1983;
mortality (Francis and Kapoor, 1992; Levkoff et al., Rockwood, 1989). However, recent studies have also

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505
498 M. Dasgupta and C. Brymer

highlighted that delirium symptoms can persist both hearing impairment or communication difficulties;
at discharge and beyond in 0–78% of individuals (vii) pre-hospitalization residence in a nursing home
(Levkoff et al., 1992; Kiely et al., 2004; Siddiqui et al., (NH) or complete dependence for ADLs; (viii) direct
2006; Inouye et al., 2007; Cole, 2010; Dasgupta and transfers from other in-patient units; and (ix) enrol-
Hillier, 2010) and that persistent delirium symptoms ment in other interventional studies. Occasionally,
are associated with functional impairments, such as a individuals were admitted to the medicine teams and
loss in an individual’s baseline abilities to perform enrolled but later in their stay (after 10 days) were
their activities of daily living (ADLs, e.g., toileting, taken to the intensive care unit or surgery. These
grooming, ambulating, and transferring) and instru- individuals were included, as they were still medical
mental ADLs (IADLs, e.g., cooking and shopping) in-patients for the first 10 days of their hospitalization.
(Marcantonio et al., 2000; McCusker et al., 2003; Kiely As the primary purpose of this study was to develop
et al., 2006; Speciale et al., 2007). In today’s medical a prognostic model for delirious patients, participants
environment, there is often insufficient time to could be approached more than once for the study if
observe delirious individuals to see if their delirium both of the following occurred: (i) a participant was
and functional disabilities resolve. Clinicians often admitted to hospital and discharged within 72 h; and
have to advise family members regarding the potential (ii) if the participant was not delirious during the
for recovery and make decisions about discharge des- initial admission. If this occurred, data from only
tinations. Yet, little is known about predictors of poor one of the admissions were included in the dataset
recovery after delirium, and management remains (the admission during which the participant was
largely empiric (Andrew et al., 2005; Britton and delirious, or the longer admission). Occasionally,
Russell, 2006; Siddiqui et al., 2006; Pitkala et al., 2011). patients were discharged and re-admitted within
We conducted a prospective cohort study to examine 1 week of discharge to the same medical team, with
prognostic factors, specifically in delirium, associated the same problem(s) that prompted the initial admis-
with acute medical hospitalization in older patients. sion (re-admission may have occurred because of
We hypothesized that there are factors, identifiable early inability to manage at home or because the patient
in the course of hospitalization, associated with poor initially left against medical advice). In this case, data
recovery after an episode of delirium. from the second admission was also collected (as a
continuation of the first hospitalization) and included
as part of the participant’s hospital data.
Methods

Consecutive patients 70 years of age or older, admitted Systematic screening and assessing for delirium
to the general medicine in-patient teaching units at
London Health Sciences Centre (LHSC), in London, Consenting patients were screened for delirium every
Ontario, were approached within 3 days of admission, 2 days (excepting weekends and holidays) three times
by the investigators for possible study inclusion. LHSC after enrollment, or until hospital discharge, by a
is a tertiary care 800-bed acute care hospital divided trained research assistant (RA). Demographics, admis-
between two sites (University Hospital and Victoria sion diagnoses, and past medical history were collected
Hospital). Recruitment occurred at either University from the chart. Delirium screening comprised a chart
Hospital, between 6 October 2009 and 29 July 2011, audit tool, used in previous studies (Dasgupta et al.,
or Victoria Hospital, between 18 April 18 and 5 May 2009), assessing for documentation of key delirium
2011. Consenting patients were included in the study symptoms (e.g., “confusion”, “drowsy”, and “agitated”)
and screened for delirium. In all cases of questionable and brief mental status screening, using the Short
ability to consent, and for all delirious patients, con- Portable Mental Status Questionnaire (SPMSQ: Pfeiffer,
sent from the caregiver was required for inclusion. 1975; Erkinjuntti et al., 1987). Similar to other studies
Exclusion criteria were the following: (i) lack of a (McCusker et al., 2003), individuals were assessed
willing caregiver or substitute decision maker; (ii) trans- further for delirium if during screening any of the
fer to another non-medical service (e.g., intensive care following occurred: (i) delirium symptoms were noted
unit or surgical service) within 7 days of admission; on chart audit; (ii) there were at least three errors on
(iii) admission for palliative or long-term institutionali- the SPMSQ assessment; and (iii) new errors occurred
zation purposes only; (iv) inability to speak English; on the SPMSQ. Delirium was assessed as performed
(v) presence of a known pre-terminal medical condi- in a prior study (Al-Aama et al., 2011). This in-
tion (expected life expectancy <6 months); (vi) severe volved the RA asking family members, caregivers,

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505
Delirium prognosis in medically ill adults 499

or staff whether they noted any new changes in the any one of the following at discharge or at follow-
participant’s level of confusion or behavior, or if up: (i) death; (ii) long-term institutionalization in
this was recorded in the chart. In addition, the a NH, providing 24-h nursing care; and (iii) func-
RA conducted a structured interview with the par- tional decline (refer to the succeeding texts).
ticipant comprising two general questions, ten
questions on orientation, 3-object registration with
Follow-up and functional decline
5-min recall, digit span testing (Strub and Black,
1985), and ability to spell WORLD backwards.
The duration of delirium varies with reported medians
Patients were diagnosed with delirium if they met
or means of 9–20 days or up to 60 days in demented
the Confusion Assessment Method (CAM) criteria,
individuals (Koponen et al., 1989; Rockwood, 1993;
a valid bedside test for delirium (Inouye et al.,
Manos and Wu, 1997). It was therefore believed that
1990) that requires the following: (i) acute onset
3 months would be a sufficient amount of time to
of fluctuating cognitive impairment; (ii) deficits of
allow for recovery and short enough to minimize
attention; and (iii) either altered level of conscious-
new inter-current illnesses, which are common in frail
ness or disorganized thought processes. Individuals
older people (Marcantonio et al., 2005).
who met CAM criteria for delirium were followed
At least 3 months after discharge, follow-up tele-
throughout their hospital stay (non-delirious patients
phone interviews were conducted with caregivers of
were not followed further after screening).
delirious participants who were discharged back to
the community or a temporary alternate facility, such
Monitoring of delirious individuals as a rehabilitation hospital. Telephone follow-up
interviews enquired about the study participant’s
After the screening period, delirious individuals were present ability to perform his or her ADLs (Older
followed by the RA (Table 1), in hospital biweekly for American Resources and Services Questionnaire),
2 weeks, weekly thereafter for 2 months, and monthly which was compared with his or her pre-hospital
thereafter until study participants were discharged from ability to perform ADLs. Ability to perform ADLs
hospital. Delirious individuals had the CAM and the were rated on a scale of 0–2 (0 being most dependent).
Memorial Delirium Assessment Scale (MDAS, Breitbart Functional decline was defined as an individual
et al., 1997) administered by the RA. The MDAS is a experiencing either a full decline in the ability to
reliable and valid scale for delirium severity that assesses perform an ADL (score decreasing from 2 to 0) or a
and quantifies all ten symptoms of delirium, with higher partial decline in at least two ADLs (score decreasing
scores indicating more severe delirium; it also measures by 1 point in at least two ADLs). Measuring functional
whether hypoactive symptoms are present. status using a proxy interview has been validated and
The delirious individual’s caregiver was interviewed performed previously (Magaziner et al., 1997, Ostbye
using the following questionnaires: (i) the Informant et al., 1997; Marcantonio et al., 2002; Andrew et al.,
Questionnaire of Cognitive Decline in the Elderly 2005; Kiely et al., 2006). This study was approved by
(Jorm, 2004), as a measure of baseline cognitive status; the Health Sciences Research Ethics Board of the
and (ii) the ADL and IADL questionnaire from the University of Western Ontario.
Older American Resources and Services Project
(Fillenbaum, 1978) to measure baseline function.
Statistical analyses and model development
Caregivers were also asked when they first noticed
the increasing confusion (duration of new confusion).
Variables found to be associated with prolonged
A RA was trained by the principal investigator in the
delirium in a previously published systematic review
use of the SPMSQ, CAM, and MDAS. Reliability was
(Dasgupta and Hillier, 2010) were collected, including
tested initially and at 4-month intervals (weighted
a history of dementia or other neurodegenerative
kappa exceeded 0.85 each time). Chart audits were also
disorder, increasing co-morbidity, initial delirium
checked on a sub-sample of eight charts of delirious
severity, hypoactive symptoms, and hypoxic illnesses
patients, every 6 months throughout the study period.
(Table 1). Other collected variables included basic demo-
graphics (age and gender) and factors felt to be clinically
Outcomes relevant, including baseline functional dependence,
severity of illness (using the Acute Physiology Score,
The primary outcome was poor recovery, a composite Knaus et al., 1985), and duration of new confusion.
categorical outcome defined by the occurrence of In addition, the three most common admitting

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505
500 M. Dasgupta and C. Brymer

Table 1 Variables considered for prognostic delirium model development and definitions (collected in all delirious participants)a

Variable(s) and definition(s) Source of information

Age and gender Medical chart

Acute Physiology Score/derived from APACHE Scale Medical chart and interview (Glasgow Coma Scale)
Comorbidity with the Cumulative Illness Rating Scalea Medical chart
Average number of medications (daily administered Documented in medical chart
medications or treatments on admission to hospital
including inhalers, home oxygen, and dialysis)—average
number of medications
Neurodegenerative illness (any kind of dementia, Documented in medical chart
Parkinson’s disease, known memory impairment or
mild cognitive impairment, or being on a cholinesterase inhibitor)
Informant Questionnaire Cognitive Decline in Elderly; higher Caregiver interview
scores indicate worse cognitive function
Baseline ADL and IADL scores, OARS project questionnaireb Caregiver interview
Duration of confusion (days) (incident delirium was defined as Medical chart and caregiver interview
confusion duration of zero days) PTA
Hypoxia (present if on home oxygen or if oxygen saturation Documented in medical chart
≤90% on room air in hospital)
Delirium characteristics
Memorial Delirium Assessment Scale—average MDAS scores Patient interview
of first two assessments (once delirious) or first recorded MDAS
(if two not available)
Hypoactive: hypoactive only, or both hypoactive and hyperactive Patient interview
symptoms present on first two (or one if two unavailable)
MDAS assessments
Three most common admitting diagnoses
Pneumonia Documented in medical chart
Acute renal failure Documented in medical chart
Urinary tract infection Documented in medical chart

APACHE, acute Physiology and Chronic Health Evaluation; ADL, activities of daily living; IADL, instrumental activities of daily living; OARS,
Older Americans Resources and Services; PTA, prior to admission.
a
CIRS Scale, score range 0–52 (higher scores indicating greater comorbidity), from Parmelee PA, Thuras PD, Katz IR, Lawton MP. 1995. Valida-
tion of the Cumulative Illness Rating Scale in a geriatric residential population. J Am. Geriatr Soc 43 (2): 130–37.
b
This questionnaire records how much assistance is required for an individual to perform his or her basic ADLs (eating, dressing, grooming, walk-
ing, transferring, bathing, toileting, and incontinence) and IADLs (using the phone, traveling, shopping, meal preparation, doing housework,
administering medications, and managing money), with lower scores indicating greater functional dependence.
c
Variables not included in the final model because of collinearity concerns with at least two variables included IADL score (correlated with age and
ADL status), number of medications (correlated with comorbidity (CIRS) and IADL scores), and IQCODE score (correlated with age and history
of neurodegenerative disorder).

diagnoses, as listed in the medical record, were also con- Receiver operating characteristic curves were used to
sidered for inclusion in the model. evaluate the model and its application in the validation
Similar to what was performed in a prior study data set. In addition, a shrinkage factor (Steyerberg
(Marcantonio et al., 1994), 2/3 of the total study et al., 2004) is presented for application of the model
population was used for the derivation (exploratory) to other populations. Analyses were performed using
sample and 1/3 used for the validation cohort SAS 9.3.
(randomly selected). To avoid collinearity, the correla- Using an estimate of a 1 : 10 ratio, of outcome to
tion amongst variables measuring similar characteris- variable, and assuming seven clinical variables would
tics (e.g., ADL and IADL) were examined and be tested in the multivariable derivation sample, at
inclusion was based on the strength of its association least 70 delirious individuals in both comparator
with the primary outcome. Univariable associations groups (with and without poor recovery) would be
(logistic regression) with the primary outcome in the required. Assuming a rate of poor recovery of
derivation cohort were evaluated, and those character- 30–70%, a resultant sample size of 233 delirious
istics significant at the 0.20 level were considered for individuals would be required in the derivation cohort
step-wise removal at the p = 0.05 level. For the final (233 × 0.3 = 70). An additional 117 (233 : 2) delirious
model, the Hosmer–Lemeshow statistic was reported. individuals would be required for the validation

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505
Delirium prognosis in medically ill adults 501

resulting in a total sample size of 350 (233 + 117) Of the 355 delirious patients, recovery status was
delirious individuals. known on 342 (96%). At discharge, 54 (15%) had
Assuming a 25% delirium rate in medical in-patients, deceased and 86 (24%) were discharged to a residen-
1400 potentially eligible in-patients would need to be tial NH. Of the remaining, 215 discharged to the
screened (1400 × 0.25 = 350). Assuming an attrition, community or a temporary rehabilitation hospital,
loss to follow-up, or non-consent rate of 40%, approx- recovery status at follow-up, was known in 202 (94%),
imately 2333 (1400/0.4) individuals would need to be at a median/mean (standard deviation) of 103/121.4
approached in total. (standard deviation 81.7) days after discharge. At
follow-up, 97 (48%) people had poor recovery: one
deceased, 50 newly institutionalized, and 46 with ADL
Results decline of at least 2 points. Therefore, the overall rate
of poor recovery after delirium was 69% (or 237 out
Two thousand seven hundred nineteen patients at of 342 delirious patients with known follow-up status).
least 70 years of age were admitted during the study Length of follow-up at the time of telephone interview
period and 1235 were included in the study (refer to was different in individuals with (median 97 days)
Figure 1 for participant flow). Delirious patients and those without (median 104 days) poor recovery
(n = 355 or 28.7% of the sample) differed from non- (p < 0.001, Wilcoxin two-sample test).
delirious patients (n = 880) in some baseline charac- Results of the modeling are shown in Table 3. The
teristics (Table 2). As expected, delirious individuals final model, derived from the derivation sample
had longer lengths of stay and were more likely to (n = 237, recovery status known on 231), included
die or be discharged to a residential NH. the following variables: advanced age, lower baseline

Figure 1 Enrolment of patients.

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505
502 M. Dasgupta and C. Brymer

Table 2 Study sample baseline characteristics (entire sample, delirious and non-delirious participants)

Variable Entire sample (n = 1235) Non-delirious (n = 880) Delirious (n = 355) Univariable (p-value)*

Mean age 82.6 81.8 84.4 <0.001

APS 6.77 6.37 7.71 < 0.001
CIRS 9.70 9.69 9.74 0.396
Average number of medications 7.48 7.56 7.28 0.496
Male gender 530 (42.9%) 381 (43.3%) 149 (42.0%) 0.687
Neurodegenerative illness present 235 (19.2%) 92 (10.6%) 143 (40.4%) < 0.001
Hearing impaired 100 (8.2%) 48 (5.5%) 52 (14.7%) < 0.001
Visual impairment 73 (6.0%) 44 (5.1%) 29 (8.22%) 0.034
Died in hospital 91 (7.4%) 37 (4.2%) 54 (15.2%) < 0.001
Discharge to NH 139 (11.3%) 53 (6.0%) 86 (24.2%) < 0.001
Median LOS 7.0 6.0 15.0 <0.001
Common diagnoses and geriatric syndromes noted on admission in the medical recorda
Pain/soft tissue injury 130 (10.5%) 91 (10.3%) 39 (11.0%) 0.738
Pneumonia 281 (22.8%) 202 (23%) 79 (22.2%) 0.790
Acute renal failure 178 (14.4%) 113 (12.8%) 65 (18.3%) 0.013
Urinary tract infection or urosepsis 163 (13.2%) 104 (11.82%) 59 (16.62%) 0.024
Heart failure 151 (12.2%) 119 (13.5%) 32 (9.0%) 0.029
Unwell/weak 132 (10.7%) 88 (10.0%) 44 (12.4%) 0.218
Falls/poor mobility 419 (33.9%) 260 (29.6%) 159 (44.8%) < 0.001
Failure to cope or Failure to thrive 81 (6.6%) 40 (4.6%) 41 (11.6%) < 0.001
Psychotropic drugs on admissionbc
Psychotropics PTA 704 (57%) 462 (52.2%) 242 (68.2%) < 0.001
SSRIs PTA 200 (16.2%) 130 (14.8%) 70 (19.7%) 0.033
Bdzs PTA 155 (12.6%) 102 (11.6%) 53 (14.9%) 0.109
Narcotics PTA 124 (10%) 82 (9.3%) 42 (11.8%) 0.184

APS, Acute Physiology Score (score range 0–56); CIRS, Cumulative Illness Rating Scale (score range 0–52); PTA, prior to admission; SSRIs,
selective serotonin reuptake inhibitors; Bdzs, benzodiazepines; NH, nursing home; LOS, length of stay in days.
*Characteristics of delirious and non-delirious patients were compared using chi-square tests for categorical variables and unpaired t-tests for
continuous variables.
a
The five most common admitting diagnoses (within the first 3 days of admission, excluding confusion) and common geriatric diagnoses (falls or mobility
difficulties, “failure to cope or thrive”, and feeling “unwell”) are listed. There was no attempt to diagnose specific disorders by study investigators (all
diagnoses were listed in the medical chart), as study investigators were not necessarily the most responsible physician, for study participants.
b
Psychotropics noted, on admission, as per the admission note, including narcotics, antipsychotics, lithium, antidepressant classes, steroids, benzo-
diazepines, antihistamines, dopaminergic agents, anti-seizure medications, muscle relaxants (e.g., cyclobenzaprine), anti-emetics (e.g., dimenhydri-
nate, metochlopramide, and prochlorperazine), hypnotics (zoplicone), anticholinergic urinary agents, cholinesterase inhibitors, and memantine.
c
The three most commonly prescribed psychotropic drugs (included on admission are listed (SSRIs, benzodiazepines, and narcotics).

ADL ability, having higher initial MDAS scores, being Delirium was present in 29% of participants, similar
hypoxic, and an admission diagnosis including acute to rates reported in previous medical studies (11–42%,
renal failure (ARF). These were associated with poor Siddiqui et al., 2006). The rate of poor recovery was very
recovery. The area under the curve for the derivation high, exceeding two-thirds, showing the often irrevers-
model was 0.80 (95% confidence interval: 0.74–0.85) ible nature of delirium. Also noteworthy was the high
and in the validation cohort was 0.68 (95% confidence rate (close to 50%) of poor recovery even after hospital
interval: 0.57–0.79). The model had some discrimina- discharge, highlighting the continued vulnerability of
tory ability beyond chance alone (Table 4). However, delirious individuals and demonstrating that the cost
in both cohorts, even in lower risk quartiles, there was of delirium exceeds estimates on the basis of hospital
still a high chance of poor recovery (approximately data alone. This study highlights the importance of
50%), although this was still lower than the average risk
delirium as a prognostic factor, further emphasizing
in the highest risk quartile (greater than 90%).
the need to incorporate active multi-disciplinary
delirium prevention programs into routine hospital
Discussion care, which can prevent delirium (Holroyd-Leduc
et al., 2010). This study also showed that certain charac-
This is the largest outcome study of delirium teristics, present early in the course of hospitalization,
conducted in acutely hospitalized, medical in-patients. were more predictive of poor recovery.

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505
Delirium prognosis in medically ill adults 503

Table 3 Associations with poor recovery (functional decline, institutionalization, or death) from delirium

Functional recovery Poor recovery Multivariable model Multivariable model

(n = 78), derivation (n = 153), derivation Univariable OR (95% CI), OR (CI), derivation OR (CI), validation
Variable* sample sample p-value derivation sample sampleab sample

Age (SD) 81.40 (6.00) 86.62 (6.94) 1.13 (1.07,1.18) p < 0.001 1.14 (1.01,1.20) 1.12(1.04, 1.21)
ADL 13.44 (2.51) 12.35 (2.91) 0.86 (0.77, 0.96) p < 0.006 0.88 (0.78, 0.99) 0.97 (0.83, 1.15)
score
(SD)
Hypoxia 21 (26.92) 61 (39.87) 1.80 (0.99, 3.26) p = 0.053 2.28 (1.12, 4.64) 0.85 (0.31, 2.36)
(%)
MDAS 11.35 (3.56) 13.33 (4.57) 1.12 (1.05, 1.21) p = 0.001 1.16 (1.06, 1.26) 1.03 (0.92. 1.14)
median
(SD)
ARF (%) 10 (12.82) 31 (20.26) 1.73 (0.80, 3.74) p = 0.165 2.69 (1.10, 6.58) 2.64 (0.67, 10.42)

OR, odds ratio; CI, confidence interval; SD, standard deviation; ADL, activities of daily living; MDAS, Memorial Delirium Assessment Scale
(possible score range 0–30); ARF, acute renal failure.
*Variables included in final multivariable model.
a
Variables not entered into multivariable model because of collinearity with at least two other variables included: instrumental activities of daily
living score, average number of medications on admission, and Informant Questionnaire of Cognitive Decline in the Elderly.
Variables eliminated during multivariable backwards elimination: presence of hypoactive symptoms (univariate association: OR 1.82, 95%
CI: 0.88–3.78, p = 0.11; multivariate p = 0.58).
Variables not included in the multivariable model because of their univariate associations (derivation cohort) were male sex (OR 0.73, p = 0.26),
Cumulative Illness Rating Scale score (OR 0.99, p = 0.86), prevalent versus incident delirium (OR 1.17, p = 0.68), duration of confusion (OR 1.03,
p = 0.25), history of neurodegenerative illness (OR 0.93, p = 0.81), Acute Physiology Score (OR 1.0, p = 0.7), admitting diagnosis of pneumonia
(OR 1.4, p = 0.34), admitting diagnosis of urinary tract infection or urosepsis (OR 0.7, p = 0.31).
Data missing in variable (number): prevalent versus incident delirium and duration of confusion (5) and history of neurodegenerative disorder (1).
b
Final model: intercept value = 11.533.
2
Hosmer–Lemeshow Goodness of Fit statistic X8 = 3.694, p = 0.884.
Shrinkage: likelihood ratio chi-square for model = 235.614. With five factors, the shrinkage factor is (235.614–5)/235.614 = 0.979.
Final model was not altered when duration of follow-up was included in the model (except for hypoxia, which was no longer significant; hypoxia was
also highly correlated with length of follow-up, thus introducing collinearity, for unclear reasons).

Table 4 Predicted probabilities at differing risk score levels post-acute care patients (Kiely et al., 2006) and a
35% combined rate of institutionalization or death
Predicted probabilities at quartile cut-offs of poor recovery in delirious hip fracture patients (Marcantonio et al.,
Derivation data Validation data
2002). Differences in reported rates compared with
set set this study may be related to differences in the
populations and outcomes studied. This is the first
<Q1* 24/57 (42.1%) 18/32 (56.3%) study to look primarily at multi-variable predictors of
Q1-median* 30/58 (51.7%) 24/31 (77.4%)
Median-Q3 43/58 (74.1%) 19/24 (79.2%) poor recovery in undifferentiated medical in-patients.
>Q3* 56/58 (96.6%) 22/24 (91.7%) Older age, ADL dependence, the presence of ARF,
Chi-square p-value <0.001 0.020 hypoxia, and delirium severity (as measured by the
Mantel–Haenszel Test <0.001 0.003
(linear trend) average of the first two MDAS measures) were
independently associated with increased risk of poor
*Cut-off probabilities were Q1 = 0.472, median = 0.720, Q3 = 0.865. recovery. Study results are consistent with other
studies demonstrating an association between increasing
Individuals with poor recovery had a slightly delirium severity, older age, functional dependence,
shorter (median of 7 days) length of follow-up. This and hypoxia with persistent symptoms, institutionali-
is likely not clinically significant given that it is only zation, or death (Kelly et al., 2001; Marcantonio et al.,
a 7 day difference (out of 3 months) and because it 2002; Kiely et al., 2006; Dasgupta and Hillier, 2010).
reflects when individual caregivers were available for Results suggest that delirium severity may be a mea-
follow-up. A similar high rate of poor recovery has sure of the extent of the acute brain malfunction and
been reported in one study (Andrew et al., 2005) in that the delirious brain may be particularly vulnerable
which 68% of 77 delirious patients referred to geriatric to hypoxia and ARF. Although these variables were
consult services died or had functional decline. Other more predictive of poor outcomes than chance alone,
studies have reported a 25% ADL decline rate in even individuals without these factors did poorly

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505
504 M. Dasgupta and C. Brymer

indicating that, in delirium, there is no low risk group. decrease its burden. Management trials should include
Other factors such as in-hospital, post-discharge, and active and early detection for delirium with other pos-
other, as of yet, unknown factors (such as possibly sible early interventions. Potential interventions may
baseline frailty) may also impact on recovery, as has include non-pharmacologic (e.g., aggressively treating
been suggested in another study (Eeles et al., 2012). hypoxia and ARF) or pharmacologic options (e.g., by
The present study did not assess for baseline frailty treating severe delirium, on the basis of early MDAS
and its impact on recovery after delirium. A compre- scores, with neuroleptic medications early on) to see
hensive measure such as frailty may have avoided how these impact recovery. Future studies specifically
some possible collinearity issues. This could be enrolling large numbers of delirious patients are also
performed in future studies although assessing for needed to further understand prognostic factors to
some aspects of baseline frailty (e.g., grip strength, help to design trials to decrease adverse consequences
baseline mood, or self-rated fatigue) may be difficult associated with delirium.
to do in actively delirious hospitalized individuals.
Strengths of this study include the enrolment of a
large number of acutely medically ill patients, deriva- Conflict of interest
tion and validation of a multivariable model, system-
atic active screening for delirium, regular None declared.
methodologic checks of reliability, and the low loss
to follow-up rate. The study was conducted indepen-
dently of clinical care and involved a separate study Key points
team with no role in clinical care or education of care
providers. Although clinical providers were aware of • Delirium in older medical in-patients is
associated with poor recovery (death,
an ongoing study, it is not believed that the study af- institutionalization, or functional decline) in
fected care pathways. more than two-thirds of affected individuals
Weaknesses include the following: (i) lack of daily 3 months after discharge. Delirium therefore
screening (possibly missing briefly delirious people is associated with irreversible consequences.
who may have had more favorable outcomes); (ii) sin-
gle-center enrolment (LHSC, with two hospitals); (iii) • Old age, functional dependence, hypoxia,
and acute renal failure are risk factors for
persistent delirium symptoms per se were not assessed poor recovery.
at 3-month follow-up (this would have required a
face-to-face interview); and (iv) the effect of events oc- • Even individuals without these risk factors
have a high rate of poor recovery,
curring during hospitalization, which may impact on highlighting that there is no low risk group,
outcomes were not included in the model. In addition, in medically hospitalized, older adults with
a significant percentage of people were excluded for delirium. Understanding prognostic factors
lack of consent or because their length of stay was in delirium is important to design trials to
shorter than 72 h. It is difficult to know the potential improve outcomes after delirium.
biases that may have been introduced because of these
exclusions as baseline or outcome data were not col-
lected in non-consented individuals. It was assumed
that delirious patients would have longer lengths of Acknowledgements
hospital stay and that very few delirious people would
have been missed by excluding those with shorter stays The authors gratefully acknowledge the assistance of
although this is difficult to know with certainty. The Dr. Tareef Al-Aama (patient recruitment), Dr. SheriLynn
delirium rate in this study approximates rates reported Kane (patient recruitment), Ms. Cindy Madigan (admin-
in other medical settings, suggesting that few cases of istrative assistance), and Dr. Iris Gutmanis (admin-
delirium were missed. istrative assistance and original grant review)
Future studies should be conducted on other delir- Both authors work at the University of Western
ious patient populations (present results are only ap- Ontario. The primary author obtained funding from
plicable to medical in-patients) and to assess whether the Physicians’ Services Incorporated for this research
other in-hospital and post-discharge factors are associ- project. The study has not been published previously
ated with functional recovery. Given the very poor nor has it been submitted elsewhere (a poster was
prognosis of delirium, this study highlights the need presented at the Canadian Geriatrics Society annual
for ongoing management trials for active delirium to scientific meeting in April 2013).

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505
Delirium prognosis in medically ill adults 505

Author contributions Kelly KG, Zisselman M, Cutillo-Schmitter T, et al. 2001. Severity and course of delir-
ium in medically hospitalized nursing facility residents. Am J Geriatr Psych 9(1):
72–77.
Monidipa Dasgupta: project conception, funding sup- Kiely DK, Bergmann MA, Jones RN, et al. 2004. Characteristics associated with
delirium persistence among newly admitted post-acute facility patients. J Gerontol
port, recruiting patients, training RA, acquisition of Med Sci 59(4): 344–349.
and analyzing data, access to and responsibility for Kiely DK, Jones RK, Bergmann MA, et al. 2006. Association between delirium reso-
data, drafting and final approval of manuscript lution and functional recovery among newly admitted post acute facility patients.
J Gerontol Med Sci 61(2): 204–208.
Chris Brymer- recruiting patients, acquisition of Knaus WA, Draper EA, Wagner DP, Zimmerman JE. 1985. APACHE II. A severity of
data, drafting and final approval of manuscript disease classification system. Crit Care Med 13: 818–829.
Koponen H, Stenback U, Mattila E, et al. 1989. Delirium among elderly persons
admitted to a psychiatric hospital: clinical course during the acute stage and one
year follow-up. Acta Psychiatr Scand 79: 579–585.
Levkoff SE, Evans DA, Liptzin B, et al. 1992. Delirium: the occurrence and persistence
References of symptoms among elderly hospitalized patients. Arch Intern Med 152: 334–40.
Lipowski ZJ. 1983. Transient cognitive disorders (delirium, acute confusional states)
in the elderly. Am J Psych 140: 1426–1436.
Al-Aama T, Brymer C, Gutmanis I, et al. 2011. Melatonin decreases delirium in Magaziner J, Zimmerman SI, Gruber-Baldini AL, Hebel JR, Fox KM. 1997. Proxy
elderly patients: a randomized, placebo-controlled trial. Int J Geriatr Psychiatry reporting in five areas of functional status. Comparisons with self-reports and
26: 687–694. observations of performance. Am J Epidemiol 146(5): 418–428.
Andrew MK, Freter SH, Rockwood K. 2005. Incomplete functional recovery after Manos PJ, Wu R. 1997.The duration of delirium in medical and postoperative
delirium in elderly people: a prospective cohort study. BMC Geriatr 5: 5. patients referred for psychiatric consultation. Ann Clin Psych 9(4): 219–226.
Breitbart W, Rosenfeld B, Roth A, et al. 1997. The Memorial Delirium Assessment Marcantonio ER, Goldman L, Mangione CM, et al. 1994. A clinical prediction rule for
Scale. J Pain Symptom Manag 13(3): 128–137. delirium after elective noncardiac surgery. JAMA 271(2): 134–139.
Britton A, Russell R. 2006. Multidisciplinary team interventions for delirium in Marcantonio ER, Flacker JM, Michaels M, Resnick NM. 2000. Delirium is indepen-
patients with chronic cognitive impairment. The Cochrane Database of Systematic dently associated with poor functional recovery after hip fracture. J Am Geriatr
reviews (2): CD000395. DOI: 10.1002/14651858. Soc 48(6): 618–624.
Cole MG. 2010. Persistent delirium in older hospital patients. Curr Opin Psychiatry 23(3): Marcantonio ER, Ta T, Duthie E, Resnick NM. 2002. Delirium severity and psycho-
250–254. motor types: their relationship to outcomes after hip fracture repair. J Am Geriatr
Dasgupta M, Hillier LM. 2010. Factors associated with prolonged delirium: a system- Soc 50(5): 850–857.
atic review. Int Psychogeriatr 22(3): 373–394. Marcantonio ER, Kiely DK, Simon SE, et al. 2005. Outcomes of older people admitted
Dasgupta M, Rolfson DB, Stolee P, Borrie MJ, Speechley M. 2009. Frailty is associated to postacute facilities with delirium. J Am Geriatr Soc 53(6): 963–969.
with postoperative complications in older adults with medical problems. Arch McCusker J, Cole M, Abrahamowicz M, Primeau F, Belzile E. 2002. Delirium predicts
Gerontol Geriatr 48: 78–83. 12-month mortality. Arch Intern Med 162(4): 457–463.
Eeles EMP. White SV, O’Mahony SM, Bayer AJ, Hubbard RE. 2012. The impact of McCusker J, Cole M, Dendukuri N, Han L, Belzile E. 2003. The course of delirium in
frailty and delirium on mortality in older inpatients. Age Ageing 41: 412–416. older medical inpatients. A prospective study. J Gen Intern Med 18(9): 696–704.
Erkinjuntti T, Sulkava R, Wikstrom J, Autio L. 1987. Short Portable Mental Status Ostbye T, Tyas S, McDowell I, Koval J. 1997. Reported activities of daily living:
Questionnaire as a screening test for dementia and delirium among the elderly. J agreement between elderly subjects with and without dementia and their
Am Geriatr Soc 35(5): 412–416. caregivers. Age Ageing 26(2): 99–106.
Fillenbaum G. 1978. Multidimensional Functional Assessment: The OARS Pfeiffer E. 1975. A Short Portable Mental Status Questionnaire for the assessment of
Methodology—a Manual (2nd edn). Center for the study of aging and human organic brain deficit in elderly patients. J Am Geriatr Soc 23(10): 433–441.
development, Duke University: Durham, NC. Pitkala KH, Strandberg TE, Tilvus RS, Laurila JV. 2011. Effective treatment of
Francis J, Kapoor WN. 1992. Prognosis after hospital discharge of older medical delirium is difficult but not impossible. J Am Geriatr Soc 59(1): 167–168.
patients with delirium. J Am Geriatr Soc 40(6): 601–606. Rockwood K. 1989. Acute confusion in elderly medical patients. J Am Geriatr Soc 37(2):
Holroyd-Leduc JM, Khandwala F, Sink KM. 2010. How can delirium best be 150–154.
prevented and managed in older patients in hospital? Can Med Assoc J 182(5): Rockwood K. 1993. The occurrence and duration of symptoms in elderly patients
465–470. with delirium. J Gerontol 48(4): M162–M166.
Inouye SK. 2006. Delirium in older persons. N Engl J Med 354(11): 1157–1165. Rockwood K, Cosway S, Carver D, et al. 1999. The risk of dementia and death after
Inouye SK, van Dyck CH, Alessi CA, et al. 1990. Clarifying confusion: the Confusion delirium. Age Ageing 28(6): 551–556.
Assessment Method. Ann Intern Med 113: 941–948. Siddiqui N, House AO, Holmes JD. 2006. Occurrence and outcome of delirium in
Inouye SK, Viscoli CM, Horwitz RI, Hurst LD, Tinetti ME. 1993. A predictive model medical in-patients: a systematic literature review. Age Ageing 35: 350–364.
for delirium in hospitalized elderly medical patients based on admission character- Speciale S, Bellelli G, Lucchi E, Trabucchi M. 2007. Delirium and functional recovery
istics. Ann Intern Med 119: 474–481. in elderly patients. J Gerontol Med Sci 62(1): 107–108.
Inouye SK, Zhang Y, Jones RN, et al. 2007. Risk factors for delirium at dis- Steyerberg EW, Borsboom GJ, van Houwelingen HC, Eijkemans MJ, Habbema JD.
charge: development and validation of a predictive model. Arch Intern Med 2004. Validation and updating of predictive logistic regression models: a study
167(13): 1406–1413. on sample size and shrinkage. Stat Med 23: 2567–2586.
Jorm AF. 2004. The Informant Questionnaire on Cognitive Decline in the Elderly Strub RL, Black FW. 1985. The Mental Status Evaluation in Neurology (2nd edn). FA
(IQCODE): a review. Int Psychogeriatr 16(3): 275–293. Davis: Philadelphia, PA; 42–43.

Copyright # 2013 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2014; 29: 497–505