Does Delirium Increase Hospital Stay?

Does Delirium Increase Hospital Stay?

Jane McCusker, MD, DrPH,�z Martin G. Cole, MD,w§ Nandini Dendukuri, PhD,�z

and Eric Belzile, MSc�

OBJECTIVES: To determine the effects of prevalent andincident delirium on length of hospital stay.

DESIGN: Prospective cohort study, comparing (1) lengthof stay after admission in cases of prevalent delirium versuscontrols without prevalent delirium with (2) length of stayafter diagnosis in cases of incident delirium versus controlsmatched by day of diagnosis.

SETTING: The medical services of a primary, acute carehospital.

PARTICIPANTS: Medical admissions of patients aged 65and older from the emergency department with deliriumdiagnosed during the first week in hospital. Patientsadmitted to intensive care or oncology and those with aprimary diagnosis of stroke were excluded. A sample ofthose without delirium was also enrolled.

MEASUREMENTS: Delirium was diagnosed using theConfusion Assessment Method. Data on length of stay anddiagnosis-related groups (DRGs) were abstracted fromadministrative data. Measures of covariates included theInformant Questionnaire on Cognitive Decline in theElderly, the Delirium Index, the instrumental activities ofdaily living questionnaire from the Older AmericanResources and Services project, the Charlson ComorbidityIndex, the Clinical Severity Scale, and the Acute PhysiologyScore.

RESULTS: The study sample comprised 359 patients: 204with prevalent delirium, 37 with incident delirium, and 118without delirium. After controlling for covariates, prevalentdelirium was not associated with a significantly longerhospital stay, but incident delirium was associated with anexcess stay after diagnosis of 7.78 days (95% confidence

interval53.07, 12.48). Similar results were obtained usinglog-transformed or DRG-adjusted estimates of length ofstay.

CONCLUSION: In older medical inpatients, incident butnot prevalent delirium is an important predictor of longerhospital stay. Interventions to prevent incident deliriummay reduce length of stay. J AmGeriatr Soc 51:1539–1546,2003.

Key words: aged; delirium; inpatients; length of stay;cohort study

Delirium is common in older hospital inpatients and isassociated with various adverse outcomes, including

increased mortality and poor physical function and cogni-tive status,1–3 but previous research has found inconsistentevidence for an effect of delirium on length of stay. Twostudies that controlled for dementia, increased severity ofillness, and comorbidity found a significant effect ofdelirium1,4 but one did not.5 One study that controlledfor severity of illness but excluded patients with dementiafound that delirium was associated with longer stays.6

Another study found that patients with hypoactive deliriumhad longer hospital stays than those with hyperactive ormixed subtypes.7 It has been hypothesized that hypoactivedelirium is related to longer hospital stays for two reasons.First, these patients may be more likely to develop pressureulcers or nosocomial infections.7 Second, hospital staff areless likely to detect and treat hypoactive delirium thanhyperactive delirium. However, none of these studiesdistinguished between the effects of delirium present atadmission (prevalent delirium) versus after admission(incident delirium).8 Incident delirium may have a greatereffect than prevalent delirium on length of stay because itmay result from intercurrent illnesses (or complications)that lead to additional investigations and treatments andcause longer hospital stays.

The authors conducted a study of the 12-monthprognosis of delirium in older medical inpatients andreported the adverse effects of delirium on physicalfunction, cognitive status, and mortality.2,3 In this paper,the effects of prevalent and incident delirium, hypoactivity,and severity of delirium on length of stay were investigated,

This research was supported by grants from the Medical Research Councilof Canada (MA14709), the Fonds de la Recherche en Sante du Quebec(980892), and St. Mary’s Hospital Center (SMH9627).

Address correspondence to Jane McCusker, MD, DrPH, Department ofClinical Epidemiology and Community Studies, St. Mary’s Hospital,3830 Lacombe Avenue, Montreal, Quebec H3T 1M5.E-mail: [email protected]

From the Departments of �Clinical Epidemiology and Community Studiesand wPsychiatry, St. Mary’s Hospital, Montreal, Quebec; and Departmentsof zEpidemiology and Biostatistics and §Psychiatry, McGill University,Montreal, Quebec.

JAGS 51:1539–1546, 2003r 2003 by the American Geriatrics Society 0002-8614/03/$15.00

controlling for dementia, comorbidity, severity of illness,physical function, and sociodemographic variables.

METHODS

The study was conducted at a 400-bed, university-affiliated,primary acute care hospital in Montreal. The study designwas a prospective, observational, cohort study comprisingtwo cohorts of older medical inpatients, with and withoutprevalent delirium (Figure 1). The methods of recruitmentof the cohorts have been described previously.2,3 In brief, astudy nurse screened consecutive admissions of patientsaged 65 and older from the emergency department to themedical or geriatric services. Prevalent delirium cases wereidentified using a two-stage screening process, first using theShort Portable Mental Status Questionnaire (SPMSQ)9 andreview of nursing notes. Patients with an SPMSQ score ofthree or more errors or symptoms of delirium in the nursingnotes were assessed with the Confusion AssessmentMethod(CAM).10,11 Of 206 patients with prevalent deliriumenrolled in the study, 204 were matched against thehospital’s administrative database.

For the control cohort of patients without prevalentdelirium, a stratified sampling was used to reduce con-founding by dementia, selecting all patients with an SPMSQscore of three or more errors and a systematic sample ofpatients with an SPMSQ score of less than three. Thesampling ratio was one in 10 for patients admitted to thegeneral medical service and one in three for those admittedto the geriatric service, to take into account the higherprevalence of dementia in the latter. The sample of patients

without prevalent delirium, used as controls for theanalyses of prevalent delirium, consisted of 136 patients(Figure 1).

Patients without prevalent delirium (including thosescreened but not selected for the control sample) wererescreened for incident delirium during the first week ofhospitalization. One patient with incident delirium whoselength of stay was unusually high (156 days) was removedfrom these cases, leaving 36 cases of incident delirium in theanalyses (Figure 1). For the analysis of incident delirium,controls were restricted to patients without prevalent orincident delirium. Because patients with incident deliriumcould have been hospitalized for up to a week beforedelirium was diagnosed, at least three controls werematched to each case by day of diagnosis in the index case(i.e., the number of days between admission to hospital anddiagnosis of delirium in the index case). Thus, controls wereselected randomly from patients who had been hospitalizedat least as long as the predelirium length of stay of the indexcase. This day was used as the ‘‘zero time’’ for measuringlength of stay. For example, a patient whose delirium wasdiagnosed on Day 3 would be matched to controls whowere still hospitalized on Day 3, and the length of staymeasured from Day 3. After each incident case wasmatched to three controls, the 10 remaining controls wererandomly matched to 10 incident cases.

The study was conducted in conjunction with arandomized trial of the detection and treatment of deliriumand a subgroup of 227 patients (183 with prevalent and 44with incident delirium) were also participants in the trial.12

These 227 patients were randomly allocated to receiveeither an intervention comprising systematic detection andmultidisciplinary care or usual care. The intervention didnot significantly improve time to improvement in cognitivestatus or other study outcomes.12 Both studies used thesame methods of data collection, measures, and researchstaff.

Data Collection

A research assistant, blind to study group, assessed patientsat enrollment and interviewed a family member.

Severity of Delirium Symptoms

The severity of delirium symptoms was assessed using theDelirium Index (DI), which has satisfactory interraterreliability and concurrent criterion validity.13

Delirium Subtypes

Based on data from the CAM and the first DI assessment,patients were classified into the following four categories:no hyperactive or hypoactive symptoms, hyperactive but nohypoactive symptoms, hypoactive but no hyperactivesymptoms, and both hyperactive and hypoactive symp-toms.

Dementia

The presence of dementia was assessed using a 16-itemInformant Questionnaire on Cognitive Decline in theElderly (IQCODE),14–16 using a cutoff of 3.5 to definedementia.

Figure 1. Flow chart for study. �Includes 283 patients with ShortPortable Mental Status Questionnaire (SPMSQ) score of 3 ormore and 17 with SPMSQ score of less than 3. wIncludes 140patients with SPMSQ score of 3 or more and a random sample of70 patients with SPMSQ score of less than 3. zAll patients haveSPMSQ score of less than 3.

1540 MCCUSKER ET AL. NOVEMBER 2003–VOL. 51, NO. 11 JAGS

Instrumental Activities of Daily Living

The instrumental activities of daily living (IADL) ques-tionnaire from the Older American Resources and Servicesproject, reported by an informant, was used to assesspremorbid function (before the current illness but not morethan 1 month before hospital admission). The validity ofinformant reports of IADLs is satisfactory, particularly forobservable behaviors.17,18 Although nondemented subjectswould be capable of reporting this information, the authorspreferred to use informant reports for all subjects, to reducesystematic differences in reporting.

The Charlson Comorbidity Index

The Charlson Comorbidity Index (CCI)19 was computedfrom data abstracted by a research nurse from the patients’charts on medical problems present at admission or atprevious admissions.

Clinical Severity Scale

The research nurse rated the Clinical Severity Scale20,21 atthe time of diagnosis of delirium or at study enrollment forpatients without delirium, based on clinical impression andreview of the patient’s chart.

The Acute Physiology Score

The Acute Physiology Score (APS), derived from the AcutePhysiology, Age, and Chronic Health Evaluation scale,22

was coded from measures made on or before the date ofdiagnosis or study enrollment. Other baseline variablesincluded age, sex, marital status (married vs single,divorced, or widowed), residence (home, seniors’ residenceor foster home, or nursing home), and admitting service(medicine or geriatrics).

To assess whether in-hospital death or discharge to adifferent level of care mediated the effects of the indepen-dent variables, a three-category discharge outcome variablewas created: dead; alive, discharged to the same or a less-dependent level of care than preadmission; and alive,discharged to a more-dependent level of care.

Administrative Data

Study data were linked with the hospital’s administrativedatabase, using the chart number. The accuracy of thematch was checked using the patient’s age and sex. Thelength of stay (acute care days) was calculated, andoutcome of hospitalization (dead, alive, and dischargelocation) was determined. One subject with prevalentdelirium was excluded because there was a 2-yeardifference in the date of the closest admission in thedatabase, and another because the patient died during thehospital admission according to study data but notaccording to the database. The diagnosis-related group(DRG) codes for study subjects were abstracted from thehospital’s administrative database. The average lengths ofstay associated with each DRG were obtained from theprovincial ministry of health for the group of Quebechospitals similar to the one in the study. For each patient, aDRG-adjusted excess length of stay was computed as thedifference between the observed length of stay and theaverage length of stay for that DRG in similar Quebechospitals.

Statistical Methods

The effects of prevalent and incident delirium on length ofstay were evaluated in separate models using multivariatelinear regression. Three sets of models were fitted for eachdelirium measure: first using untransformed length of stay,second using a logarithmic transformation of length of stayto account for its nonnormal distribution, and third usingthe DRG-adjusted excess length of stay. In each model,those covariates that could be associated with the indepen-dent variable of interest (prevalent or incident delirium) orwith increased length of stay were included. Covariates inthese analyses were age, sex, marital status, residence,admitting service, dementia (present or missing vs absent),comorbidity (CCI), physiological severity (APS), andclinical severity. Analyses were conducted including andexcluding premorbid IADLs, to elucidate the effect ofdementia. The authors also adjusted for outcome ofhospitalization to determine whether death or dischargelocation mediated the effect of delirium. Interactionsbetween the primary independent variable and residence,admission service, dementia, comorbidity, and clinicalseverity were evaluated for all models. A treatment groupcovariate was added to the multivariate regression modelsto adjust for the possible effect of the intervention on any ofthe outcomes. This variable was found not to be astatistically or clinically significant predictor of any of theoutcomes and, hence, was not included in the final analyses.

To evaluate the effect of delirium severity and deliriumsubtypes on length of stay, a multivariate linear regressionmodel was used for patients with prevalent or incidentdelirium. Length of stay was measured from admission forprevalent delirium patients and from time of diagnosis forincident delirium patients. This model was adjusted for thecovariates and interactions listed above in addition toincident/prevalent delirium and its interaction with delir-ium severity and subtypes. Coefficients in the linearregression models were considered statistically significantwhen their confidence interval did not include zero.

RESULTS

Characteristics at enrollment of the three patient groups(prevalent, incident, and no delirium) are shown in Table 1.In spite of the stratified sampling by SPMSQ status, higherproportions of patients with prevalent delirium weredemented and resident in nursing homes. Both deliriumgroups were more severely impaired than patients withoutdelirium on measures of disease severity and comorbidityand experienced higher in-hospital mortality.

Prevalent Delirium

The means7standard deviations for lengths of stay for the204 prevalent delirium cases and the 136 controls in thecontrol sample without prevalent deliriumwere 16.2713.2days and 12.6711.8 days, respectively. The medians (andinterquartile ranges) of the length of stay in the two groupswere 13.0 (7.0–21.0) and 8.0 (5.0–17.0), respectively.The results of multivariate regression models are shown inTable 2. A positive regression coefficient for prevalentdelirium in these models indicates that the length of stay ofpatients with prevalent delirium, unexplained by thecovariates, was greater than that of patients without

DOES DELIRIUM INCREASE HOSPITAL STAY? 1541JAGS NOVEMBER 2003–VOL. 51, NO. 11

prevalent delirium. The magnitude of this coefficient is thedifference in the mean value of the length of stay betweenthe prevalent delirium group and control group afteradjusting for other covariates. The results of the threemodels are consistent and show that patients with prevalentdelirium did not have a significantly longer average staythan controls, after controlling for dementia and othercovariates. Dementia was also not associated with length ofstay in any of the models. Patients residing at a nursinghome at baseline had a significantly shorter length of stay.Covariates that were significantly associated with a longerstay included higher clinical severity or comorbidity scores.The addition of the discharge outcome variable orpremorbid IADLs to these models did not change thedirection or the statistical significance of the coefficients for

prevalent delirium or dementia. There was no significantinteraction between prevalent delirium and residence,admission service, dementia, clinical severity, comorbidity,or outcome of hospitalization.

Incident Delirium

The mean lengths of stay for the remaining 36 cases ofincident delirium and the 118 matched controls were20.2714.2 and 10.779.8, respectively. The medians (andinterquartile ranges) of the length of stay were 16.5 (11.0–23.0) and 7.5 (4.0–15.0), respectively, for incident deliriumcases and the matched control group. In the multivariateregression model, after controlling for dementia and othercovariates, a diagnosis of delirium was associated with a

Table 1. Characteristics of Prevalent Incident and Control Cohorts at Enrollment, In-Hospital Death, and DischargeDisposition

Prevalent Incident ControlCharacteristic (n5 204) (n5 37) (n5 118)

Categorical variables, n (%)SexMale 82 (40.2) 13 (35.1) 32 (27.1)Female 122 (59.8) 24 (64.9) 86 (72.9)

Marital statusMarried 66 (32.4) 13 (35.1) 26 (22.0)Single, divorced, widowed 138 (67.7) 24 (64.9) 92 (78.0)

Admission serviceMedicine 142 (69.6) 28 (75.7) 68 (57.6)Geriatrics 62 (30.4) 9 (24.3) 50 (42.4)

ResidenceHome 144 (70.9) 33 (89.2) 80 (67.8)Senior residence/foster home 33 (16.3) 3 (8.1) 30 (25.4)Nursing home or long-term care 26 (12.8) 1 (2.7) 8 (6.8)Missing (1) (0) (0)

DementiaNo 42 (22.1) 16 (50.0) 42 (42.9)Yes 148 (77.9) 16 (50.0) 56 (57.1)Missing (14) (5) (20)

In hospital death 33 (16.2) 9 (24.3) 6 (5.1)Discharge dispositionMore dependent level of care 64 (37.4) 9 (32.1) 31 (27.7)Same or less dependent level of care 107 (62.6) 19 (67.9) 81 (72.3)

Days from admission to enrollment1 193 (94.6) 0 (0.0) 112 (94.9)2–3 7 (3.4) 12 (32.4) 6 (5.1)4–5 3 (1.5) 14 (37.8) 0 (0.0)6–8 1 (0.5) 11 (29.7) 0 (0.0)

Continuous variables, mean7standard deviationAge 83.6177.40 82.3076.28 83.6476.58Delirium severity� 9.0773.80 6.9273.78 4.3472.79Clinical severityw 5.2871.43 5.1171.39 3.8671.25Acute physiology scorez 5.0773.49 4.9774.11 2.9272.68Charlson Comorbidity Index§ 2.6671.97 2.9572.20 2.1071.76Instrumental activities of daily living8 6.5673.68 8.4973.40 7.7773.38

�Range 0 (no symptoms) to 21 (most severe).wRange 1 (minimum) to 9 (most severe).zRange 0 (no impairment) to 44 (maximum impairment).§Range 0 (no comorbidity) to 12 (maximum in our study).8Range 0 (completely dependent) to 14 (completely independent).


significantly longer stay after diagnosis of approximately 8days, with and without adjustment for DRGs (Table 3).Residence in a nursing home was again associated withshorter stays. None of the other covariates, includingdementia, clinical severity, and comorbidity, were signifi-cantly associated with length of stay. The addition of thedischarge outcome variable or premorbid IADLs to thesemodels did not change the direction or the statisticalsignificance of the coefficients for incident delirium ordementia. The only significant interaction was betweendelirium and dementia for both length of stay and excesslength of stay models, indicating that the effect of incidentdelirium on length of stay was limited to patients withdementia.

Severity and Hypoactivity

Among patients with prevalent or incident delirium(n5240), the mean and median lengths of stay (from

admission for prevalent cases and from diagnosis forincident cases) were longer for patients with hypoactivesymptoms only (mean516.2 days, median514.0 days,n555) or with both hypo- and hyperactive symptoms(mean518.2 days, median5 14.0 days, n5134) thanfor those with hyperactive symptoms only (mean510.9 days, median59.0 days, n541) or neitherhypo- nor hyperactive symptoms (mean59.6 days,median5 7.0 days, n5 10). This difference remainedsignificant in multivariate analyses, after excluding the lastgroup because of the small sample (data not shown).Severity of delirium symptoms was not associatedwith length of stay. There were no significant inter-actions between hypoactivity or severity and type ofdelirium (prevalent vs incident). The only significantinteraction between delirium subtype and comorbiditywas that there was a greater effect of delirium subtypes inpatients with low comorbidity scores than in those withhigh scores.

Table 2.Multiple Linear Regression Analysis for Length of Stay Comparing Patients with Prevalent Delirium (n5204) withControls (n5136)

Length of Stay (From Admission)�

Without LogTransformation

With LogTransformation

DRG-Adjusted ExcessLength of Stayw

Independent variable Parameter Estimatez (95% Confidence Interval)

Prevalent deliriumNo (reference) 0.00 F 0.00 F 0.00 FYes 1.07 (� 2.02–4.16) 0.15 (� 0.06–0.36) 0.32 (� 2.66–3.31)

DementiaAbsent (reference) 0.00 F 0.00 F 0.00 FPresent � 0.21 (� 3.36–2.95) 0.02 (� 0.19–0.23) � 0.25 (� 3.30–2.80)Missing � 3.87 (� 8.62–0.89) � 0.29 (� 0.61–0.02) � 2.26 (� 6.85–2.33)

Age � 0.04 (� 0.24–0.17) � 0.003 (� 0.02–0.01) 0.02 (� 0.17–0.22)Sex

Female (reference) 0.00 F 0.00 F 0.00 FMale 0.23 (� 2.89–3.35) 0.01 (� 0.20–0.22) 1.03 (� 1.98–4.05)

Marital statusSingle, divorced, widowed (reference) 0.00 F 0.00 F 0.00 FMarried 1.23 (� 2.12–4.58) 0.07 (� 0.16–0.29) 0.20 (� 3.03–3.43)

ResidenceHome (reference) 0.00 F 0.00 F 0.00 FSenior residence, foster home � 0.65 (� 4.13–2.83) � 0.04 (� 0.28–0.19) � 1.08 (� 4.44–2.29)Nursing home � 8.19 (� 12.88–3.50) � 0.63 (� 0.94–0.31) � 7.06 (� 11.59–2.54)

Admitting serviceGeriatrics (reference) 0.00 F 0.00 F 0.00 FMedicine � 0.83 (� 3.92–2.26) � 0.18 (� 0.39–0.03) 1.55 (� 1.43–4.53)

Clinical severity§ 1.74 (0.65–2.82) 0.11 (0.04–0.18) 1.48 (0.44–2.52)Comorbidity8 0.87 (0.16–1.59) 0.05 (� 0.00–0.10) 0.78 (� 0.09–1.47)Acute physiology scorez � 0.02 (� 0.47–0.42) � 0.003 (� 0.03–0.03) � 0.04 (� 0.47–0.39)

�The dependent variable is observed length of stay or log (observed length of stay).wThe dependent variable is observed length of stay – expected length of stay for each diagnosis-related group.zFor a continuous covariate, parameter estimates represent the change in the dependent variable corresponding to a unit change in the covariate. For categoricalcovariates, parameter estimates represent the estimated mean difference in the dependent variable between each level of the covariate and the reference category. Forexample, with every unit increase in clinical severity, the length of stay increased by 1.74 and the excess length of stay increased by 1.48. Patients with prevalentdelirium had a length of stay of 1.07 days and an excess length of stay of 0.32 days more than those without prevalent delirium.§Range 1 (minimum) to 9 (most severe).8Range 0 (no comorbidity) to 12 (maximum in our study).zRange 0 (no impairment) to 44 (maximum impairment).


DISCUSSION

The results of this study indicate that the effect of deliriumon length of stay may depend on the type of delirium,incident or prevalent. Although prevalent delirium does notappear to contribute independently to length of stay, theseresults suggest that incident delirium may increase thelength of hospital stay, even after adjusting for comorbidity,severity of illness, and other confounding variables. Thehigher death rate found in these patients did not appear tomediate this increased length of stay. Previous research onthe relationship between delirium and length of stay has notdistinguished between incident and prevalent delirium; thisstudy’s outcome may explain some discrepancies betweenthe results. Those studies that included cases of incidentdelirium (in proportions varying between 28% and 67%)found that delirium increased length of stay.1,4,6 One studythat found no association included only prevalent delir-ium.5 There are three potential explanations for this

finding. First, in some cases, incident delirium may resultfrom intercurrent illnesses or complications that are theunderlying cause of the longer stay, but the associationbetween incident delirium and length of stay persisted evenafter adjustment for three different clinical measures ofseverity of illness and comorbidity and for DRGs. Second, adeterioration in physical function may accompany incidentdelirium, rendering discharge inappropriate until thepatient’s clinical condition improves. Third, a diagnosis ofincident delirium may prompt further evaluation and tests,which require longer stays.

This study confirms previous results indicating that,when the effects of delirium and dementia are investigated,dementia is not associated with longer stays.1 Thisconclusion is not in keeping with two studies that reportedlonger hospital stays in patients with dementia, but deliriumwas not assessed in either of these studies.23,24 Similarly,several studies have reported that cognitive impairment isassociated with longer hospital stays, but these studies did

Table 3. Multiple Linear Regression Analysis for Length of Stay fromDiagnosis Comparing Patients with Incident Delirium(n5 36) with Controls (n5 112)

Length of Stay (From Diagnosis)�

Without LogTransformation

With LogTransformation

DRG-Adjusted ExcessLength of Stayw

Independent Variable Parameter Estimatez (95% Confidence Interval)

Incident deliriumNo (reference) 0.00 F 0.00 F 0.00 FYes 7.78 (3.07–12.48) 0.96 (0.49–1.43) 8.05 (3.59–12.51)

DementiaAbsent 0.00 F 0.00 F 0.00 FPresent 1.77 (� 2.20–5.75) 0.14 (� 0.26–0.53) 1.09 (� 2.68–4.86)Missing � 2.62 (� 8.49–3.26) � 0.20 (� 0.79–0.39) � 1.56 (� 7.13–4.01)

Age 0.02 (� 0.28–0.32) 0.004 (� 0.03–0.03) 0.07 (� 0.22–0.35)Sex

Female 0.00 F 0.00 F 0.00 FMale 0.10 (� 4.25–4.44) � 0.01 (� 0.44–0.42) 1.26 (� 2.86–5.39)

Marital statusSingle, divorced, widowed 0.00 F 0.00 F 0.00 FMarried � 2.16 (� 7.15–2.84) � 0.04 (� 0.53–0.46) � 2.86 (� 7.60–1.88)

ResidenceHome 0.00 F 0.00 F 0.00 FSenior residence, foster home 1.14 (� 3.59–5.86) 0.43 (� 0.05–0.90) 0.50 (� 3.98–4.98)Nursing home � 6.86 (� 15.17–1.45) � 0.85 (� 1.68–0.02) � 7.43 (� 15.32–0.45)

Admitting serviceGeriatrics 0.00 F 0.00 F 0.00 FMedicine � 0.61 (� 4.60–3.37) � 0.32 (� 0.72–0.08) 1.10 (� 2.68–4.88)

Clinical severity§ 1.09 (� 0.43–2.61) 0.14 (� 0.01–0.29) 0.75 (� 0.69–2.19)Comorbidity8 0.06 (� 0.93–1.05) � 0.02 (� 0.12–0.07) 0.22 (� 0.71–1.16)Acute physiology scorez 0.04 (� 0.56–0.65) 0.01 (� 0.05–0.07) � 0.001 (� 0.57–0.57)

�The dependent variable is length of stay after delirium diagnosis for subjects with delirium and matched controls.wThe dependent variable is observed length of stay – expected length of stay for each diagnosis-related group.zFor a continuous covariate, parameter estimates represent the change in the dependent variable corresponding to a unit change in the covariate. For categoricalcovariates, parameter estimates represent the estimated mean difference in the dependent variable between each level of the covariate and the reference category. Forexample, with every unit increase in clinical severity the length of stay increased by 1.09 and the excess length of stay increased by 0.75. Patients with incident deliriumhad a length of stay of 7.78 days and an excess length of stay of 8.05 days more than those with no delirium.§Range 1 (minimum) to 9 (most severe).8Range 0 (no comorbidity) to 12 (maximum in our study).zRange 0 (no impairment) to 44 (maximum impairment).


not distinguish between delirium and dementia.25–28 Thepresent study’s results suggest that associations betweendementia or cognitive impairment and length of stay shouldbe interpreted with caution, because delirium rather thandementia may explain them.

The study found that the presence of hypoactivedelirium symptoms was associated with significantly longerhospital stays in patients with prevalent or incidentdelirium, even after adjusting for covariates. As previouslysuggested, patients with hypoactive delirium may havelonger hospital stays because of their increased risk ofpressure ulcers, nosocomial infections, or other complica-tions or because hospital staff are less likely to detect andtreat them than those with hyperactive delirium,7 butgreater severity of delirium symptoms was not associatedwith an increased length of stay. The authors havepreviously reported, from the same study, that increasedseverity of deliriumwas a predictor of 12-monthmortality,2

but greater mortality did not appear to explain the lack ofassociation of severity with length of stay in the currentstudy.

The method of selection of the control sample ofpatients without prevalent delirium has some strengths andlimitations. The primary objective of this investigation wasto determine the effect of delirium on length of stay andother outcomes, independent of the effects of otherpotential determinants of these outcomes. Because demen-tia was considered the most important of these determi-nants, and because the authors could use one of thescreening tests, the SPMSQ, to identify those with probabledementia, stratified sampling based on the SPMSQ scorewas used in an attempt to reduce confounding by dementia.Although reduction of confounding by other variables(severity of illness, comorbidity, sociodemographic vari-ables) through stratification or matching might have beendesirable, it was not considered feasible. Residual con-founding by dementia and other variables was controlledthrough the use of multivariate analyses. Even though thiswas a secondary analysis of this data set, it had a sufficientlylarge sample size. With the type I error fixed at 0.05, thestudy had 87% power to detect whether the observeddifference in the unadjusted log-transformed length of staybetween prevalent delirium cases and controls was sig-nificantly different from zero; similarly, the study had 99%power to detect the observed difference in length of staybetween incident delirium cases and controls.

This study has several limitations. First, certain groupsof patients (those who died or were discharged ortransferred on the day of admission, those with acommunication problem or diagnosis of stroke, and thoseadmitted to intensive care or oncology) were not assessed.Thus, the results cannot be generalized to all medicalinpatients aged 65 and older. Second, the sample of incidentdelirium was small (n536), limiting the precision of theestimate of effect. Third, the cause of longer stays could notbe determined (e.g., medical problems vs need for place-ment). Fourth, the measure of dementia, the IQCODE, hasnot been validated in patients with delirium, but the IADLscore, known to be associated with dementia,29 was alsounrelated to length of stay.

There are several implications of the results of thisinvestigation for patient care and health policy. First, the

prevention of incident delirium through special careprotocols30 has the potential to reduce length of stay.Second, hospital staff need to be alert to the development ofincident delirium in patients who were not delirious atadmission; the cause should be investigated promptly andappropriate treatment initiated,31 but previous studies havedocumented that delirium is detected only in a minority ofcases.32,33 Thus, interventions that improve the detectionand diagnosis of delirium need to be developed andevaluated. Third, it is important that delirium be detectedand coded to optimize predictions of length of stay andreimbursement of hospitals for the care of patients withdelirium.

REFERENCES

1. O’Keeffe S, Lavan J. The prognostic significance of delirium in older hospital

patients. J Am Geriatr Soc 1997;45:174–178.

2. McCusker J, Cole M, Dendukuri N et al. Delirium in older medical inpatients.

Marker for subsequent cognitive and functional status. Can Med Assoc J

2001;165:575–583.

3. McCusker J, Cole M, Abrahamowicz M et al. Delirium predicts 12-month

mortality. Arch Intern Med 2002;162:457–463.

4. Francis J, Martin D, Kapoor WN. A prospective study of delirium in

hospitalized elderly. JAMA 1990;263:1097–1101.

5. Inouye S, Rushing J, Foreman M et al. Does delirium contribute to poor

hospital outcomes? A three-site epidemiologic study. J Gen Intern Med

1998;13:234–242.

6. Pompei P, Foreman M, Rudberg MA et al. Delirium in hospitalized

older persons. Outcomes and predictors. J Am Geriatr Soc 1994;42:

809–815.

7. O’Keeffe S, Lavan J. Clinical significance of delirium subtypes in older people.

Age Ageing 1999;28:115–119.

8. Inouye S. The dilemma of delirium: Clinical and research controversies

regarding diagnosis and evaluation of delirium in hospitalized elderly medical

patients. Am J Med 1994;97:278–288.

9. Pfeiffer E. A short portable mental status questionnaire for the assessment

of organic brain deficit in elderly patients. J Am Geriatr Soc 1975;23:

433–441.

10. Inouye S, VanDyck C, Alessi C et al. Clarifying confusion: The confusion

assessment method: A new method for detection of delirium. Ann Intern Med

1990;113:941–948.

11. Zou Y, Cole M, Primeau F et al. Detection and diagnosis of delirium in the

elderly: Psychiatrist diagnosis, Confusion Assessment Method or consensus

diagnosis? Int Psychogeriatr 1998;10:303–308.

12. Cole M, McCusker J, Bellavance F et al. Systematic detection and multi-

disciplinary care of delirium in older medical inpatients: A randomized trial.

Can Med Assoc J 2002;167:753–759.

13. McCusker J, Cole M, Bellavance F et al. Reliability and validity of a new

measure of severity of delirium. Int Psychogeriatr 1998;10:421–433.

14. Jorm A. A short form of the Informant Questionnaire on Cognitive Decline in

the Elderly (IQCODE). Development and cross-validation. Psychol Med

1994;24:145–153.

15. Jorm FA, Broe AG, Creasey H et al. Further data on the validity of the

informant questionnaire on cognitive decline in the elderly (IQCODE). Int J

Geriatr Psychiatry 1996;11:131–139.

16. Law S, Wolfson C. Validation of a French version of an informant-based

questionnaire as a screening test for Alzheimer’s disease. Br J Psychiatr

1995;167:541–544.

17. Magaziner J, Zimmerman S, Gruber-Baldini A et al. Proxy reporting in five

areas of functional status: Comparison with self-reports and observations of

performance. Am J Epidemiol 1997;146:418–428.

18. Ostbye T, Tyas S, McDowell I et al. Reported activities of daily living:

Agreement between elderly subjects with and without dementia and their

caregivers. Age Ageing 1997;26:99–106.

19. Charlson M, Pompei P, Ales K et al. A new method of classifying prognostic

comorbidity in longitudinal studies: Development and validation. J Chronic

Dis 1987;40:373–383.

20. Charlson M, Sax F, MacKenzie R et al. Assessing illness severity: Does clinical

judgment work? J Chronic Dis 1986;39:439–452.

21. Inouye S, Viscoli C, Horwitz R et al. A predictive model for delirium in

hospitalized elderly medical patients based on admission characteristics. Ann

Intern Med 1993;119:474–481.


22. Knaus W, Draper E, Wagner D et al. APACHE II. A severity of disease

classification system. Crit Care Med 1985;13:818–829.

23. Erkinjuntti T, Wikstrom J, Palo J et al. Dementia among medical inpatients

Evaluation of 2000 consecutive admissions. Arch Intern Med 1986;146:1923–

1926.

24. Lyketsos C, Sheppard JRP. Dementia in elderly persons in a general hospital.

Am J Psychiatry 2000;157:704–707.

25. Glass R, Weiner M. Seeking a social disposition for the medical

patient: CAAST, a simple and objective clinical index. Med Care 1976;14:

637–641.

26. Glass R, Mulvihill M, Smith H et al. The 4 score: An index for predicting a

patient’s non-medical hospital days. Am J Public Health 1977;67:751–755.

27. Evans R, Hendricks R, Lawrence K et al. Identifying factors associated with

health care use: A hospital-based risk screening index. Soc Sci Med

1988;27:947–954.

28. Satish S, Winograd CH, Chavez C et al. Geriatric targeting criteria as

predictors of survival and health care utilization. J Am Geriatr Soc

1996;44:914–921.

29. Tombaugh TN, McIntyre NJ. The Mini-Mental State Examination: A

comprehensive review. J Am Geriatr Soc 1992;40:922–935.

30. Inouye S, Bogardus S, Charpentier P et al. A multicomponent intervention to

prevent delirium in hospitalized older patients. N Engl JMed 1999;340:669–720.

31. Inouye S, Schlesinger M, Lydon T. Delirium. A symptom of how hospital care

is failing older persons and a window to improve quality hospital care. Am J

Med 1999;106:565–573.

32. Levkoff S, Besdine R, Wetle T. Acute confusional states (delirium) in the

hospitalized elderly. Ann Rev Gerontol Geriatr 1986;6:1–26.

33. Elie M, Rousseau F, Cole M et al. Prevalence and detection of delirium

in elderly emergency department patients. Can Med Assoc J 2000;163:

977–981.


Documents

Does Delirium Increase Hospital Stay?