NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Clinical Guideline Centre (UK). Delirium: Diagnosis, Prevention and Management [Internet]. London: Royal College of Physicians (UK); 2010 Jul. (NICE Clinical Guidelines, No. 103.)

Cover of Delirium

Delirium: Diagnosis, Prevention and Management [Internet].

Show details



What is the prevalence of delirium in different hospital settings and in long-term care?

5.1. Introduction

Delirium is a common clinical syndrome that can be found throughout the healthcare system. In order to understand more fully the clinical burden and associated health economic implications of delirium, it is necessary to first understand the epidemiology in terms of the occurrence of delirium within individual healthcare settings.

Operationalised diagnostic criteria for delirium have been formulated in the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association 1980; American Psychiatric Association 1987; American Psychiatric Association 1994) (DSM III, DSM III-R and DSM-IV), and in the International Classification of Diseases (10th Edition) (World Health Organisation 1992) (ICD-10). There is good diagnostic agreement between DSM-IV and its predecessors, with DSM-IV identifying all patients diagnosed with delirium by DSM III and DSM III-R in one prospective cohort study of elderly hospital patients and nursing home residents (Laurila 2004, and section 6.6).

There is a notable disparity between the DSM and ICD-10 criteria for the diagnosis of delirium. The DSM-IV criteria are more inclusive in terms of diagnosis of delirium, with ICD-10 being relatively restrictive. In a cohort of elderly medical hospital patients and nursing home residents (mean age 88.4 years), 24.9% met the diagnostic criteria of DSM-IV, whilst only 10.1% of the same cohort were diagnosed with delirium when the diagnostic criteria of ICD-10 were applied (Laurila 2004). A comparison of the DSM-IV and ICD-10 criteria (table 5.1) reveals the ICD-10 criteria to include additional requirements for the diagnosis of delirium. In addition, the Laurila study (Laurila 2004) informs us that three cohorts were identified, those identified by DSM alone, ICD10 alone and both, and suggests that people who are identified using the ICD-10 criteria are different to the people identified using DSM. The stricter inclusion criteria and additional diagnostic requirements of ICD-10 have an associated impact on case detection and identify a cohort of patients who are more frequently dependent for care needs and more likely to be resident in the long-term care setting (Laurila 2004). Therefore we used the DSM-IV criteria as being the standard operational definition for delirium.

Table 5.1. DSM-IV and ICD-10 Diagnostic Criteria (American Psychiatric Association 1994; World Health Organisation 1992).

Table 5.1

DSM-IV and ICD-10 Diagnostic Criteria (American Psychiatric Association 1994; World Health Organisation 1992).

5.2. Terminology

Confusion can exist between the epidemiological terms prevalence and incidence. Prevalence represents the number of existing cases at a single point in time. Incidence represents the number of new cases that develop within a cohort over a defined period of time. The term ‘occurrence’ has been proposed as an alternative when there is ambiguity or overlap between the measurement of prevalence and incidence (Porta 2008).

Prevalent delirium in hospital therefore defines the presence of delirium at the point of admission to hospital. Incident delirium in hospital represents the development of delirium after hospital admission.

This is an important distinction to make as incident (new) cases of delirium are more likely to be amenable to strategies aimed at preventing the onset of delirium. It is therefore of key importance to provide a priori definitions of prevalence, incidence and occurrence rates with regard to delirium. Where it is not possible to use these definitions because of healthcare setting, alternatives will be considered, for example in the surgical setting, in which the concept of pre- and post-operative delirium is likely to hold importance.

As the emergency department represents a healthcare setting in which patients spend a short period of time prior to admission to the hospital bed base or discharge home, the concept of point prevalence is most relevant in this setting and incidence/occurrence rates will not be measured.

Long-term care represents the permanent residence of an individual, rather than respite care on a temporary basis. The concepts of point prevalence (prevalence at a single point in time) and period incidence (cumulative incidence over a defined period of time) are likely to be relevant in the long-term care setting.

5.2.1. A priori definitions

These a priori definitions form the basis for the review of study data and subsequent data categorisation:

Prevalent delirium: The presence of delirium within the first 24 hours of admission to a healthcare setting (or the duration of the preoperative period within the surgical cohort).

Incident delirium: The development of delirium subsequent to the first 24 hours of admission (24 hours postoperatively in surgical cohorts), measured at least daily, until discharge from hospital or death.

Occurrence rate: Where study data reveal overlap between the a priori definitions of prevalent and incident data, or where the a priori conditions are not met, the term ‘occurrence rate’ will be used.

Total Delirium: Where there is more than one measure of rate of delirium available (e.g. both prevalent and incident delirium), or where occurrence rate represents data collected from healthcare admission to discharge, a fourth term, total delirium, will be summated to reflect the occurrence of delirium throughout the duration of stay.

5.3. Selection criteria

Types of study

Prospective cohort and cross-sectional studies were to be included. Epidemiological data derived from the control arm of randomised clinical trials and case-control studies could be considered if there was evidence of reasonable representativeness of the sample. Retrospective studies were to be excluded.

Patient population & healthcare setting

Selection criteria for the patient population are defined in the methods section. Settings included are hospital and long-term care. In much of the guideline, the hospital patient population has been considered as a whole, but it is clear that this population is diverse and heterogeneous. For this epidemiological review, each healthcare setting was to be considered separately and data were to be grouped according to individual healthcare settings.

Studies were preferred if they were conducted in the UK. However, studies were to be included regardless of the country in which they were conducted, although the representativeness was to be taken into consideration in the analysis.

The DSM-IV criteria for delirium were to be the desired operational definition. As set out in the introduction, there is consistency between cases of delirium identified with DSM-IV versus DSM III-R and DSM III. Studies using a case definition based on the DSM-IV, DSM III-R or DSM III criteria [or a diagnostic tool validated against DSM-IV, DSM III-R or DSM III e.g. Confusion Assessment Method (CAM), DRS] were therefore to be included. As set out in the introduction (section 5.1), there is a notable disparity between cases of delirium that are identified with application of ICD-10 as compared with DSM-IV. Consequent to this, studies using the ICD-10 criteria for delirium were excluded from the epidemiological review.

Hospital Episode Statistics (HES)

Locally generated clinical coding data is collated nationally in the Hospital Episode Statistics (HES) database, the national statistical data warehouse for the NHS. Clinical coding of data is used for clinical research, epidemiological mapping and health resource allocation. A bespoke HES dataset was generated in order to assess the agreement between the epidemiological profile of delirium as determined by prospective cohort data and clinical coding data collated by the HES database.

5.4. Description of studies

Description of included and excluded papers together with study design are reported in table 5.2.

Table 5.2. study inclusion, exclusion and design.

Table 5.2

study inclusion, exclusion and design.

Eleven studies had fewer than 100 participants (Adamis 2005; Angles 2008; Edlund 2009; Goldenberg 2006; Koebrugge 2009; Milisen 2001; Robinson 2008; Rolfson 1999; Rudolph 2005; Rudolph 2006; Santana Santos 2005); 11 studies had more than 500 participants (Brauer 2000; Holmes 2000; Inouye 2008; Leslie 2005; Marcantonio 1994; McCusker 2003; Morrison 2003; Ouimet 2007; Peterson 2006; Rudolph 2007; Van Rompaey 2009) and the remaining 50 studies had between 100 and 500 participants.

The majority of included studies were of North American origin (figure 5.1), with only two studies based in the UK health service setting(Adamis 2005; Holmes 2000).

Figure 5.1. study by country of origin.

Figure 5.1

study by country of origin.

Thirty-eight studies selected adult patients with age cut-off points (Adamis 2005; Balas 2007; Bickel 2008; Brauer 2000; Breitbart 1996; Cole 1994; Edlund 2001; Edlund 2006; Elie 2000; Faezah 2008; Franco 2001; Furlaneto 2006; Galanakis 2001; Goldenberg 2006; Greene 2009; Han 2009; Henon 1999; Holden 2008; Holmes 2000; Inouye 1998; Inouye 1998; Inouye 1999; Jones 2006; Kagansky 2004; Koebrugge 2009; Leslie 2005; Lewis 1995; Marcantonio 1994; Martin 2000; McAlpine 2008; McNicoll 2003; Naughton 1995; Naughton 2005; Pisani 2006; Pitkala 2005; Rockwood 1999; Santos 2004; Santana Santos 2005). One study selected patients above the age of 40 years, three those above the 50 years, six selected patients above 60 years, 17 above 65 years, eight above 70 years and three studies selected patients above the age of 75 years.

Mean patient age varied between healthcare settings, with a higher mean age of study participants noted in the general medicine and long-term care cohorts (see Appendix D). A younger mean age of study participants was notable in the ICU, HIV/AIDS medicine and psychiatry settings.

Healthcare Setting

Studies were first assessed and grouped according to healthcare setting (Figure 5.2).

Figure 5.2. Hospital study populations grouped by healthcare setting.

Figure 5.2

Hospital study populations grouped by healthcare setting.

Where applicable, study populations were further categorised into, for example, acute and elective surgical patient groups. The long-term care setting was considered separately.

Both the ICU and acute stroke unit settings represent a form of enhanced specialist care within standard/usual care pathways. Thus, patients with ongoing delirium episodes may be admitted from the inpatient bed base to the ICU/acute stroke unit and therefore the occurrence rate can be a useful record of delirium rate for these specific healthcare settings. This model of ICU/acute stroke unit care is commonplace within the UK healthcare system.

5.5. Methodological quality of studies

The study cohort as a whole was assessed for representativeness on the grounds of the inclusion and exclusion criteria defined in each individual study. Inclusion and exclusion criteria were broadly similar between studies in each healthcare setting. Three studies (Andrew 2006; Edelstein 2004; Kakuma 2003) stated exclusion criteria showing that the study cohort was not representative of the population for that setting (see Appendix E). This is an important consideration for this epidemiology review, and these studies were therefore not analysed further.

One study (Andrew 2006) was in a long-term care setting whereby people with dementia were excluded from the cohort.

One study (Edelstein 2004) was in a hip fracture setting whereby only ambulatory home dwelling people were included in the cohort.

One study (Kakuma 2003) was in an emergency department setting whereby people presenting from long-term care were excluded from the participant cohort.

Fourteen studies listed dementia as an exclusion criterion (Andrew 2006; Bickel 2008; Contin 2005; Koebrugge 2009; Lin 2004; Roberts 2005; Rudolph 2007) or severe dementia (Franco 2001; Galanakis 2001; Han 2009; Kagansky 2004; Leslie 2005; Martin 2000; McNicoll 2003). However, as many of these studies were in the surgical and ICU setting, it was felt that the exclusion of people with dementia in these studies would not necessarily affect the representativeness of the study cohort.

As set out earlier, studies using the DSM-IV, DSM III-R or DSM III criteria (or a diagnostic tool validated against DSM-IV, DSM III-R or DSM III) were considered for inclusion. As delirium may often be present at admission and may be present for a short period of time with a tendency to fluctuate, included studies were appraised for quality on the basis of (1) an initial assessment for delirium within the first 24 hours of admission (post admission, preoperative period in the surgical studies) and (2) the frequency of subsequent assessments for delirium. Included studies were also appraised on the basis of sample size. These three criteria form the overall basis of the methodological quality assessment (Appendix E).

The relative importance of each quality criterion varies according to the type of epidemiological measurement. For example, prevalent delirium represents delirium within the first 24 hours of admission (preoperative period in the surgical cohort). With regard to this measure, the study size is therefore the key index. With regard to occurrence rate, the frequency of measurement of delirium and the study duration are potentially of greater importance.

Therefore, where studies recorded more than one measure of delirium (e.g. both prevalent delirium and occurrence rates), these were given separate quality assessments (Appendix E).

The studies were pragmatically and qualitatively grouped into high, medium and low quality on the basis of the quality criteria (Appendix E). Studies in which the sample size was small, in which the assessment of delirium was notably infrequent and/or the overall study length was short compared to the expected length of healthcare stay were considered to be at high risk of bias if a combination of these factors were present. Studies in which the methodology was unclear were also considered to lead to risk of bias. There was significant heterogeneity noted in frequency of assessment of delirium across all studies.

On the basis of these factors, four studies (Edlund 1999; Rudolph 2005; Santana Santos 2005; Van Rompaey 2009) were excluded from the overall results summary as they were felt to give potential for bias. These studies are highlighted in blue and given in italics in the study summary tables (Appendix D).

5.6. Results

Full data are given in Appendix D. Sixteen studies reported incidence or prevalence in different healthcare settings. Sixty-one studies report occurrence of delirium. The meaning of occurrence varied between studies and is shown in Appendix E under ‘frequency of assessment’ for each study. Three studies reported data for more than one setting:

  • Pitkala 2005: General medicine (prevalence 32.6%); long-term care (15.9%)
  • Bickel 2008: Orthopaedics acute hip fracture (occurrence 41%); orthopaedics elective surgery (12.5%)
  • Galanakis 2001: Orthopaedics acute hip fracture (occurrence 40.5%); orthopaedics elective surgery (14.7%)

Summary data are reported by healthcare setting (table 5.2); in many healthcare settings the number of studies available for inclusion was limited, and the number ranged from 1 to 17 across all settings. Where more than one study is included, the median and range are given.

5.6.1. Sensitivity analysis

A sensitivity analysis was performed whereby the studies qualitatively graded as low quality were excluded from the dataset (table 5.6 – end of chapter). Removal of low quality studies led to significant change in a small number of cumulative results. Where this was the case, the sensitivity analysis results are preferred and these are shown in table 5.3 with the full results in square brackets. Exclusion of one low quality study with a low occurrence rate in the medical ICU setting led to a significant increase in the median (range) values for the occurrence of delirium, from 70.9 (22.4 – 83.3) to 80 (48 – 83.3). Following the sensitivity analysis, there was a decrease in the median (range) occurrence rate of delirium in the cardiac surgery setting, from 32 (13.5 – 50) to 21 (13.5 – 33.6), and an increase for the acute hip fracture setting. There was no apparent change in the rates of delirium in other healthcare settings when low quality studies were excluded. Where the only studies in a particular healthcare setting were low quality, this is indicated in the table.

Table 5.6. Sensitivity analysis.

Table 5.6

Sensitivity analysis.

Table 5.3. summary data by healthcare setting.

Table 5.3

summary data by healthcare setting. (Full results are shown in red)

5.6.2. UK Data

Two included studies gave data on rates of delirium in the UK healthcare setting. The first, a prospective cohort study in a general medical setting with a sample size of 940 (Adamis 2005), recorded an occurrence rate of delirium of 37.3%. The second, a larger prospective cohort study in an orthopaedic setting with a sample size of 731 (Holmes 2000), recorded an occurrence rate of delirium of 14.8% (this study was considered to be of low quality). The limited number of studies available in UK healthcare settings leaves significant uncertainty as to the actual rates of delirium within the UK healthcare system.

5.6.3. Hospital Episode Statistics (HES)

In order to compare the epidemiological data with national clinical coding data, a bespoke dataset was requested from HES. The dataset provided information on the 2006 – 2007 total number of Finished Consultant Episodes (FCEs) of delirium (ICD code F05, delirium not induced by alcohol and other psychoactive) thus reflecting the scope of the guideline. The data were subcategorised by specialty and age (table 5.4).

Table 5.4. Delirium Finished Consultant Episodes and Total Episodes by Specialty.

Table 5.4

Delirium Finished Consultant Episodes and Total Episodes by Specialty.

Primary diagnoses represent the first of up to 14 diagnoses in the HES dataset and provide the main reason as to why the patient was in hospital. Subsequent to the primary diagnosis are up to 13 secondary diagnoses that record other diagnoses related to the episode. The bespoke delirium F05 dataset included both primary and secondary coded diagnoses of delirium, hence capturing all episodes of delirium in the UK healthcare setting in 2006 – 2007. It is likely that one episode of delirium corresponds to one patient having delirium. In order to calculate incidence of delirium as a percentage, the total number of FCEs in 2006 – 2007 (again split by specialty) was also requested. The latter is the record of the primary diagnoses only, which approximates to the number of admissions to each specialty. Therefore the HES delirium percentage is a reasonable reflection of the total delirium rate.

The dataset was split by age. The HES dataset captures episodes between the ages of 15 – 44 years followed by age 45 – 64 years. In order to provide a dataset that was representative of the mean age and inclusion criteria of the study cohort populations and in order that non-adult data was not introduced into the dataset, data were extracted from the HES dataset with a lower age limit of 45 years.

5.6.4. Epidemiology data compared with coded HES data

HES data are generated over the course of the hospital admission. As discussed above, the proportion of episodes of delirium is very similar to the total rate of delirium in the study summary tables (Appendix D). In order to assess the reliability of the HES data, table 5.5 shows both the HES data and the appropriate median total delirium rate (from the sensitivity analyses) as reported by the epidemiological research studies and where total delirium rate was available.

Table 5.5. Comparison of Median Total Delirium Rates with HES Total Delirium Episode Rates.

Table 5.5

Comparison of Median Total Delirium Rates with HES Total Delirium Episode Rates.

There is a clear and significant disparity between the expected total delirium rates from a prospective cohort of patients admitted to hospital or long-term care as compared to the rates of delirium extracted from HES coding data. Less than one percent of the expected cases of delirium are identified by the coding process. There are also differences in the relative numbers of patients in the various healthcare settings, e.g. trauma & orthopaedic surgery has a similar level of delirium compared with general medicine in the studies, but the HES data show a much lower level for orthopaedic surgery. We recognise that some of the people identified by DSMIV may have had vascular dementia or dementia Lewy bodies, but the proportion of these groups is likely to be small. Even if the pure delirium rate in the studies is only 10% of that reported, there would still be a considerable disparity between the delirium rates in the studies compared with the HES data.

5.7. Discussion

Accurate coding of clinical data relies on all of the following taking place: the recognition of the underlying diagnosis, recording of the diagnosis by a clinician in the medical notes and subsequent extraction of the correct diagnosis/diagnoses from the medical notes by clinical coders. It is possible that there is an attrition of delirium diagnoses at each of these three stages. Clinicians often fail to identify delirium in the hospital setting, with up to two thirds of cases of delirium remaining unrecognised (Inouye 1998). The ‘terminological chaos’ (Lindesay 1999) of delirium creates a situation in which a variety of terms are used to describe the diagnosis of delirium. If the correct diagnostic terminology for delirium is not used, clinical coders will be unable to extract accurate diagnostic data from the clinical record and hence there is the potential for considerable under-reporting of delirium at a national healthcare level.

Delirium is ubiquitous throughout the healthcare system, being particularly common in the critical care, hip fracture, vascular surgery, cardiac surgery and general medical patient populations. Delirium also appears to be common in the long-term care setting, with a point prevalence estimate of 15.9% when residents with dementia are included within the prospective cohort (Pitkala 2005). We note that this study was considered to be of low quality.

In many healthcare settings there are few studies and these studies are often of lower quality. There is therefore significant uncertainty present with regard to the true epidemiology of delirium in a large proportion of healthcare settings. In these healthcare settings further large prospective cohort studies of high methodological quality would help provide rigorous data informing the true epidemiology of delirium.

There is a paucity of prospective cohort studies of delirium in the UK healthcare environment, with the majority of epidemiological data originating from North America. There are potential differences between the structure and organisation of healthcare in the UK compared to North America that may limit between-system comparisons and there is consequent uncertainty regarding the true rates of delirium within the UK healthcare system.

There is a significant disparity between the expected rates of delirium from prospective epidemiological studies and the rates of delirium as recorded in the HES data set. National clinical coding is systematically failing to accurately record the considerable scale and consequent importance of delirium as a healthcare priority.

5.8. Health economic evidence

No relevant health economic papers were identified.

5.9. From evidence to recommendations

The GDG noted from the epidemiological review, that there is widespread occurrence of delirium throughout the healthcare system but it was poorly reported in the UK. The GDG wished to reinforce the importance of accurately recording delirium by making a recommendation on coding (recommendation 1.5.2). In addition, people recovering from delirium may not receive adequate follow up care because of poor communication between hospitals and GPs, and hospitals and long-term care facilities. The GDG emphasised in the recommendation that delirium should be recorded in both the hospital and primary care health records.

The GDG observed that healthcare professionals were often unaware of the possibility that delirium might or has occured. The GDG thought that the slogan, “Think delirium” summarised their rationale, and incorporated this into a prominent statement at the beginning of the list of recommendations (see chapter 4 and section 9.7 of the consequences review).

The GDG made a future research recommendation (FRR) about recording delirium. This was informed by the multicomponent review showing that staff education may increase the awareness of delirium. This future research recommendation can be found in section 10.25.3 and Appendix H.

5.10. Recommendations

Ensure that the diagnosis of delirium is documented both in the person’s hospital record and in their primary care health record. [1.5.2]

Copyright © 2010, National Clinical Guideline Centre - Acute and Chronic Conditions.

Apart from any fair dealing for the purposes of research or private study, criticism or review, as permitted under the Copyright, Designs and Patents Act, 1988, no part of this publication may be reproduced, stored or transmitted in any form or by any means, without the prior written permission of the publisher or, in the case of reprographic reproduction, in accordance with the terms of licences issued by the Copyright Licensing Agency in the UK. Enquiries concerning reproduction outside the terms stated here should be sent to the publisher at the UK address printed on this page.

The use of registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and therefore for general use.

The rights of National Clinical Guideline Centre to be identified as Author of this work have been asserted by them in accordance with the Copyright Designs and Patents Act, 1988.

Bookshelf ID: NBK65548


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...