Occurrence and outcome of delirium in medical in-patients: a systematic literature review

Abstract

Background: Despite the acknowledged clinical importance of delirium, research evidence for measures to improve its management is sparse. A necessary first step to devising appropriate strategies is to understand how common it is and what its outcomes are in any particular setting.

Objective: To determine the occurrence of delirium and its outcomes in medical in-patients, through a systematic review of the literature.

Method: We searched electronic medical databases, the Consultation-Liaison Literature Database and reference lists and bibliographies for potentially relevant studies. Studies were selected, quality assessed and data extracted according to preset protocols.

Results: Results for the occurrence of delirium in medical in-patients were available for 42 cohorts. Prevalence of delirium at admission ranged from 10 to 31%, incidence of new delirium per admission ranged from 3 to 29% and occurrence rate per admission varied between 11 and 42%. Results for outcomes were available for 19 study cohorts. Delirium was associated with increased mortality at discharge and at 12 months, increased length of hospital stay (LOS) and institutionalisation. A significant proportion of patients had persistent symptoms of delirium at discharge and at 6 and 12 months.

Conclusion: Delirium is common in medical in-patients and has serious adverse effects on mortality, functional outcomes, LOS and institutionalisation. The development of appropriate strategies to improve its management should be a clinical and research priority. As delirium prevalent at hospital admission is a significant problem, research is also needed into preventative measures that could be applied in community settings.

Received 16 November 2005; accepted in revised form 19 February 2006

Introduction

Delirium is said to be common in most hospital settings [1–4]. It is associated with significant adverse physical, cognitive and psychological outcomes [1, 5, 6] and increased costs to healthcare services [4, 7]. It is often seen as a complication of hospital care and a marker of the quality of in-patient care [8]. Despite its clinical importance, surprisingly little is known about its epidemiology, outcomes, prevention or management.

Delirium has been recognised as a mental disorder for thousands of years, with fairly consistent clinical descriptions since the second century CE [9]. There is now agreement about its core features: disturbance of consciousness, disturbance of cognition, rapid onset, fluctuating course and external causation [10] (the syndrome can be attributed to an independently diagnosable cerebral or systemic disease or disorder). Diagnostic criteria for delirium have been formulated in the Diagnostic and Statistical Manual of Mental Disorders [11–13] (DSM-III, DSM-III-R and DSM-IV) and in the tenth edition of the International Classification of Diseases [14] (ICD-10). Use of such operationalised diagnostic criteria should improve comparability of studies [15].

Delirium is undetected and misdiagnosed in the clinical setting [16–18]. The transient and fluctuating nature of symptoms and the heterogeneity of presentations associated with several different delirium subtypes contribute [15, 19]. Standardised research instruments have improved diagnostic consistency [15], but under-recognition remains a problem.

Evidence for effectiveness of measures to detect, prevent or manage delirium is sparse [20]. The wide range of potential aetiological factors suggests that to be effective, interventions will need to address not only the specifics of direct care but also service delivery issues [8, 21]. There may also be setting-specific factors to be considered. Measures to improve delirium management may have benefits in terms of improving healthcare for in-patients generally [8].

A necessary first step to devising appropriate strategies to prevent and manage delirium is to determine its occurrence and outcomes in a particular setting; these will have implications in the planning and evaluation of any intervention. The cost per case and predictive value of screening will depend on how common it is in that setting. Outcomes of delirium, including its economic implications, will determine feasibility of screening and intervention strategies. The type of service offered will also be influenced by how common the disorder is.

Medical in-patient settings have patients with a wide range of conditions and include a large proportion of older patients—a known risk factor for delirium [22]. Investigating delirium in medical in-patients would, therefore, have advantages in terms of wider relevance and generalisability of findings.

Objectives

To determine the occurrence and outcomes of delirium in medical in-patients in hospital through a systematic review of the literature.

Criteria for selecting studies for this review

Types of study

For occurrence, we included prospective cohort and cross-sectional studies.

For outcomes, we included prospective cohort studies, case–control studies and controlled trials.

Studies in hospital general medical in-patient settings were included, as were studies in settings or population groups where patients were judged to be similar to those found in general medical in-patients. Studies in community or hospice settings, psychiatric, surgical, accident and emergency and intensive care units were excluded. Studies solely of patients referred to liaison psychiatry services were also excluded. Studies in mixed populations were only included if data for general medical in-patients were reported separately, and results for this subset only were included in the analyses.

We included studies using a case definition consistent with current consensus criteria for delirium and all its subtypes but excluded studies of delirium tremens.

Outcome measures

We included studies with preset, clearly defined important outcomes. A preliminary review of the literature suggested no single widely accepted primary outcome. We, therefore, examined immediate short-term and long-term outcomes as follows:

Up to discharge: reversibility of delirium, duration of delirium episode, number of episodes, persistence of delirium symptoms, complications (e.g. falls, infections), mortality, cognitive function, physical function, length of admission, cost of admission, requirement for institutional care, psychological distress, carer distress and impact on staff.

At 6, 12 and 24 months: mortality, presence of delirium symptoms, physical and cognitive function, psychological distress, institutionalisation and carer distress.

For details of the quality criteria and scoring and search strategy used in this review, please see Appendix 1 in the supplementary data on the journal website (http://ageing.oxfordjournals.org/).

Methods

Systematic data extraction and assessments of quality were carried out using a data extraction tool by one reviewer. A 10% sample of studies considered for inclusion was also examined by a second reviewer independently, and good agreement was found.

There is often confusion about the distinction between the statistical terms incidence and prevalence. Incidence rates represent new events, noted in the follow-up of a cohort. Prevalence represents existing events, noted at a single point in time for the state of the group under study [23]. In clinical practice, the distinction may be problematic, particularly in transient or fluctuating conditions, where the frequency of examination will have a major impact on reported rates. Feinstein [23] suggests the term ‘occurrence rate’ to avoid some of these ambiguities. We use this term wherever incidence or prevalence has not been clearly determined, and we give a description of the measure actually used.

Results

Prevalence, incidence and occurrence studies

The initial search produced 1,052 citations of potential relevance, and following examination of titles and abstracts, 116 full-text articles were retrieved for further consideration. Sixty-five were excluded; 26 included surgical, nursing home or liaison psychiatry settings [24–49]; 10 used inappropriate definitions for delirium [50–59]; 7 were retrospective studies [60–66]; 4 were reviews [5, 7, 67, 68]; 3 included only male patients [69–71] and in 15, the incidence, prevalence or occurrence of delirium was not determined or reported [72–86].

Fifty papers met our inclusion criteria [6, 16, 17, 87–133], but several of these reported data from the same original study population (17 reports from 7 cohorts) [6, 16, 17, 89, 90, 98, 103, 105, 106, 108, 114, 117, 118, 121–123, 132]. In these, we took the earliest paper reporting relevant data as the index study. Additional information available in related subsequent papers was also extracted and presented alongside the index study. In two studies, distinct cohorts were examined and reported separately [109, 110]. Results for the occurrence of delirium in hospitalised general medical in-patients were, therefore, available for 42 cohorts reported in 40 studies (Table 1).

Sample

All studies were carried out in general medical or elderly care units, mainly sampling consecutive admissions. Two studies used a census of in-patients, over 1 week [96] and over 6 months [117]. Inclusion and exclusion criteria were broadly similar, with most studies excluding subjects with communication difficulties. Six studies [91, 92, 104, 109, 111, 119, 124] excluded subjects with dementia, either explicitly or by virtue of exclusion criteria such as pre-existing confusion, difficulty completing interviews or cognitive impairment. Thirty-five studies were carried out in older populations.

Methods to obtain consent and reporting of response rates also varied considerably. The number of exclusions was particularly high in controlled trials [99, 100, 112], although participants in these studies were reported to be largely similar to those excluded.

Results for delirium prevalence, incidence and occurrence

Twenty-one studies reported delirium prevalent at admission; only eight of these indicated delirium assessment had been undertaken within 24 hours of admission (Table 2).

In 13 studies, the incidence of new delirium occurring at any time during admission was determined (Table 3). Four further studies described delirium incidence rates in varying time frames [17, 92, 104, 110].

Occurrence rates for delirium per admission were given or could be derived from presented data in 13 studies (Table 4). A further seven studies reported various other measures of delirium occurrence [88, 96, 116, 117, 123, 129, 130].

Prevalence of delirium at admission ranged from 10 to 31% (limiting results to studies in which patients were examined within 24 hours of admission). Incidence of new delirium per admission ranged from 3 to 29%. Occurrence rate per admission varied between 11 and 42%.

Methodological differences

Delirium screening and diagnostic methodology differed, and there was marked heterogeneity in the measures used to describe delirium occurrence. Procedures to obtain consent also differed, and there was some variability in the methodological quality of studies. The presence of co-morbid conditions was not reported in most studies. The sensitivity and specificity of diagnostic instruments has been shown to vary depending on the training and professional background of the administrator [30, 134, 135]. Again, studies differed in the use of researchers and clinicians employed in case ascertainment. Distinction between delirium and dementia cases was also problematic in some studies.

In determining delirium incidence and occurrence rates, the frequency of assessments would be expected to influence results. However, we were not able to find any clear association between examination frequency and reported rates.

Many studies reported delirium in terms of incidence or occurrence rate per admission; clearly the length of admission would affect results. This information was not available in most studies and reported variously as median, mean or range of duration of stay in others. Where mean duration was given, it ranged from 8 to 30 days.

We were unable to pool results from studies due to methodological heterogeneity.

Outcomes studies

For the outcomes review, we examined 93 full-text articles. Of these, 65 reports were excluded; in 29, the settings or populations were not equivalent to general medical in-patients [25, 27, 29, 31, 34, 35, 37, 40–42, 44, 45, 53, 70, 72, 136–150]; 5 used retrospective methods [63, 66, 151–153]; in 10, the diagnosis used did not approximate to currently accepted criteria for delirium [50, 59, 76, 154–160]; 17 did not examine outcomes [21, 69, 78, 86, 98, 109, 110, 112, 132, 161–168] and 3 included less than 20 subjects [102, 169, 170].

Twenty-eight reports were included in the outcomes review [6, 16, 79–82, 89, 90, 95, 99, 100, 103, 105–107, 111, 113, 115, 122, 123, 125–127, 130, 131, 133, 171, 172]. Of these, 15 reported outcomes from 6 original study cohorts, giving results at different time intervals [1, 6, 79–82, 89, 90, 99, 103, 106, 122, 123, 125, 172] (Table 5).

Sample

Outcomes results were available for 19 study cohorts. Most employed a prospective cohort design except for one nested case–control study [131], two randomised controlled trials [99, 100] and one controlled trial [89]. Two studies included outcomes for incident cases only, 5 for admission prevalent cases only and 12 for both incident and prevalent delirium. Reporting of co-morbidity including the presence of dementia was variable, as was reporting of and methodological or statistical adjustments for relevant confounders (Table 5).

Outcomes for delirium

Outcomes for delirium are summarised in Table 6.

Fifteen studies reported death rates at discharge. We found a wide range from 6.1 to 62%, which precluded pooling of results. The lowest values were obtained from one study that included only incident cases [131] and two studies that excluded large numbers of potential subjects [105, 111]. The two studies reporting the highest death rates were limited by small numbers. Excluding these five studies, death rate at discharge was reported to be 14.5–37%. In comparisons with controls, there were mixed results with some studies reporting no significant difference, but several reporting a significant increase. In studies which examined for the independent effect of delirium, adjusting for important confounders, two reported an increase in death rate at discharge, whilst three found no significant difference. The small numbers of cases or outcome events in most of these studies raise the possibility of both type I and type II errors. The study with the highest score for quality reported increased mortality at discharge [79]. This study also described a 2-fold independent increase in mortality at 12 months (please see Table 7 in Appendix 2 in the supplementary data on the journal website, http://www.ageing.oxfordjournals.org/).

The mean length of hospital stay (LOS) was reported in 11 studies and ranged from 9 to 32 days; again, the results varied with three studies [16, 105, 173] showing a significant increase in LOS, but seven other studies [111, 113, 115, 123, 126, 130, 131] showing no significant difference in comparison with controls. One study [79] showed an independent excess LOS of 8.05 days (95% CI 3.59–12.51), attributable to incident delirium, but no significant increase with prevalent delirium.

Four studies examined institutionalisation at discharge. Of these, two [111, 115] reported no difference in rates, one [105] showed a significant increase in adjusted institutionalisation rates and another [95] reported a significant increase only for prevalent delirium. At 6 months, one study [6] showed delirium independently increased institutionalisation, odds ratio 2.8 (95% CI 1.3–6.1); and two studies [105, 125] reported no independent effect but did not adjust for potential confounders. At 12 months, one study [79] showed increased institutionalisation in patients with delirium and dementia.

In describing the clinical course of delirium, McCusker and colleagues [79] found 39% had transient symptoms (recovery within 24 hours), 29% recovered and 32% had persistent symptoms at discharge. Two other studies reported persistence of delirium symptoms at discharge to be high. In one [107], 23% subjects had no resolution of symptoms, with partial resolution in 17%, and in the other [126] complete resolution of symptoms occurred in only 40%. McCusker and colleagues [79] also reported persistence of delirium symptoms at 6 and 12 months in 32 and 41% patients, respectively. Clearly, a large proportion of patients with delirium are discharged from hospital with on-going delirium symptoms.

In functional outcomes, one study [109] found a significant association between delirium and decline in activities of daily living (ADL) scores at discharge. Francis and Kapoor [105] reported no difference in ADL or mini-mental state examination (MMSE) scores at 6 months, but McCusker and colleagues [79] showed delirium resulted in worse physical and cognitive status at 12 months.

Again, comparison of results across studies was problematic as study methodology, outcomes measurement and reporting varied so greatly. Surprisingly, many important outcomes such as psychological morbidity in patients, carers or staff, and economic costs to healthcare services were not reported for this population.

Discussion

It is clear from our review that delirium is common in general medical in-patients and has serious adverse outcomes, including increased mortality, LOS and institutionalisation. Even in highly selected groups, we found minimum occurrence rates per admission of 11% and more typical rates of 20–30%. Our findings are comparable with previous reviews [7, 174] but provide more robust evidence of how common delirium is and how poor its outcomes are. The results are even more striking given that they are likely to be an underestimate [175], not least because we excluded delirium tremens.

Clinical implications

With typical non-detection rates of 33–66% [175], strategies to improve delirium management must include measures to improve its detection.

As at least 20–30% of admissions will be affected, we cannot rely on referral to psychiatry services but must rather skill up the whole team, and in particular nursing staff, to screen, detect and manage delirium. Liaison psychiatry services may still have a role in this by offering education, training and advice to staff, as well as consultation for more complex management problems.

We know that a range of aetiologies and maintaining factors are implicated in delirium requiring a broad multi-factorial and multi-disciplinary approach [4, 19, 112]. However, given the scale of the problem, interventions also need to be simple and quick. The balance between a necessarily comprehensive and yet practicable intervention is difficult but must be achieved with particular attention to addressing issues of implementation and adherence [165].

We found delirium already present at admission to be more common than new delirium occurring during admission. Recent studies have shown that delirium is common in nursing homes [176, 177]; moreover, admission from an institution rather than the community is a risk for delirium in hospital [35]. Intervening in these settings could have the potential to deliver important benefits, including reducing hospital admissions, and therefore needs evaluation.

Research implications

We found considerable heterogeneity in case-finding and ascertainment methods. Despite the consensus in diagnostic criteria, Laurila et al. [116] have shown how much variability is introduced simply by applying different DSM and ICD10 criteria to the same data set. Clearly there needs to be greater standardisation of delirium screening and diagnostic methods.

We found a range of measures used to describe delirium frequency. The denominator used to calculate rates is integral to the results obtained, but most studies gave incidence or occurrence rates ‘per admission’; as length of admission will inevitably vary, this again limits comparisons between studies and generalisability of findings. A more useful measure may be to describe rates of delirium per in-patient day.

A common difficulty was the exclusion of some of the target population, because exclusion criteria often included properties of the index condition. Delirium may affect people’s ability to consent, communicate or complete interviews, and selection criteria requiring these conditions will obviously differentially exclude more subjects with delirium. This raises ethical considerations, including issues of conducting research in unrepresentative study populations [178]. The high mortality associated with delirium means that any prognostic or intervention studies need to take account of attrition rates of around 20–30%. In other outcomes, as delirium increases LOS and the number of people discharged to nursing or residential institutions, it is important to include a robust economic analysis.

There is surprisingly little known about the psychological impact of delirium on patients, staff and carers in this population; future outcomes studies should include measures of psychological morbidity.

Limitations of the review

We excluded delirium tremens from this review; although this is an important cause of delirium, we judged it to be a sufficiently distinct condition to warrant a separate review. We used a broad search strategy and imposed no language restrictions for included studies but confined our search to English-language databases. Resource limitations also meant that we were unable to independently review all citations or abstracts identified by the original search. We did not contact authors for information additional to that published. Nevertheless, we believe the review was sufficiently comprehensive to identify most important findings in this area.

Conclusion

Delirium is a significant problem associated with considerable adverse outcomes including increased mortality in general medical inpatients. There are many methodological and ethical concerns which have impeded delirium research. However, given the scale of the problem, addressing the problem of delirium should be a priority for clinicians and researchers.

Key points

• Delirium is common in medical in-patients and has serious outcomes including increased mortality, length of hospital admission and institutionlisation.

• Given the scale of the problem, developing interventions to prevent and improve management of delirium should be priority for clinical services.

• We cannot rely on referral to psychiatry services alone, but must improve the skills of the whole team in detection and management of delirium in these settings.

• Delirium research is sparse, and has been impeded by methodological and ethical difficulties.

• Further research with greater standardisation of delirium screening and diagnostic methods is required.

Funding

None.

Conflicts of interest

None.

References