Delirium upon admission to Swiss nursing homes : a cross-sectional study

Questions under study:Wewished to investigate the prevalence of delirium in patients upon admission to nursing homes and whether or not the previous place of residence predicts delirium. Methods: The Resident Assessment Instrument Minimum Data Set (RAI-MDS) and the Nursing Home Confusion Assessment Method (NHCAM) were used to determine whether the previous place of residence (community, nursing home, acute care, psychiatric, rehabilitation hospital) predicted the prevalence of sub-syndromal or full delirium in nursing home residents in three Swiss cantons (n = 11745). Results: 39.7% had sub-syndromal and 6.5% had full delirium. Lower cognitive performance and increased depressive symptoms were significant predictors of higher NHCAM values independent of previous residence. Age, civil status, continence, newly introduced drugs, and basic activities of daily living were predictors in some resident groups. The NHCAM scores explained varied between 25.1% and 32.3% depending on previous residence. Conclusions: Sub-syndromal and full delirium are common upon nursing home admission. Increased dependence and depression are consistently associated with higher NHCAM scores. Patients from psychiatric settings have an increased risk of delirium. Although factors associated with delirium depend on a patient’s previous residence, all patients must be carefully screened for sub-syndromal and full delirium.


Introduction
Delirium is a clinical syndrome characterised by acute changes in cognition, altered consciousness and impaired and fluctuating attention [1,2].Delirium is a protean syndrome and it often occurs in sub-syndromal expression with only some of the clinical features present [3][4][5].It occurs in more than a third of hospitalised people and its incidence during hospital stays ranges from 3 to 56% [6,7].Delirium is also frequent among nursing home residents in particular in the first weeks after admission.Nearly a quarter of all residents have symptoms of delirium after admission [8], most of them having sub-syndromal delirium [3].Delirium and signs and symptoms of delirium are often under-recognised [9].Correct diagnosis is, however,important because untreated delirium is associated with increased morbidity,i ncreased mortality,p oor functional recovery as well as increased health services utilisation including rehospitalisation [3, [10][11][12][13][14][15].Sub-syndromal delirium bears similar risks.It favours new complications in nursing home residents, re-hospitalisation, or six month mortality,and reduces the likelihood for a hospital patient being discharged into the community.T he extent of the risk usually lies somewhereb etween patients without delirium and those with full delirium [8,13].
We hypothesised that the place of living before nursing home admission -i.e.admission from home, acute medical care, mental health care, rehabilitation unit or by transfer from another nursing home -predicts delirium or sub-syndromal delirium in the first weeks upon admission to a nursing home.In Switzerland as in other western European countries residents are becoming older and frailer when entering a nursing home and the duration of stay in hospitals is getting shorter due to a lack of acute care beds and for costing reasons [13].The previous place of origin before admission may therefore affect the prevalence of delirium in the initial phase after nursing home admis-sion.To our knowledge, this hypothesis is uninvestigated.If true, it may have implications for health care provision, i.e. delirium prevention and treatment after nursing home admission.

Subjects
Allnursing home residents in the three Swiss cantons Aargau, Basel-City,a nd Solothurn (corresponding to 13.5% of the total Swiss population) [16] received a Resident Assessment Instrument Minimum Data Set (RAI-MDS) assessment within the first two weeks upon entry [17].We obtained the authorisation of the Qualitäts-Systeme Aktiengesellschaft (Q-Sys AG; Systeme zur Qualitäts-und Kostensteuerung im Gesundheitswesen) that pools all RAI-data, as well as of all directors of the nursing homes involved in this study to use this data for anonymous research purposes.Ninety of 160 nursing homes invited agreed to participate in the study.This yielded a total sample of 21821 nursing home residents.The residents' assessment took place between 1997 and 2007.Only residents for whom data on their previous residence was recorded were included in the study (n = 11745).Previous residence included 1) home care (n = 4569), 2) another nursing home (n = 1092), 3) acute care hospital (n = 4472), 4) psychiatric hospital (n =568), or 5) rehabilitation clinic (n = 1044).

Measures
Tr ained professionals assessed each resident upon admissiontothe nursinghome with the Swiss version [17] of the RAI-MDS for nursing homes, Ve rsion 2.0 1996 [18].The Nursing Home Confusion Assessment Method (NHCAM) [3] was used to screen for symptoms of delirium.This scale has been derived from the RAI-MDS through rearrangement of its variables to mimic the wellvalidated Confusion Assessment Method [4].Thus, the NHCAM defines four features including 1) acute onset and fluctuating course, 2) inattention, 3) disorganised thought or speech, and 4) altered level of consciousness.A further algorithm allows the separation of three delirium levels (level 1: sub-syndromal delirium [SS-1], i.e.only one of the four features is present; level 2: sub-syndromal delirium [SS-2], i.e. two of the four features are present; level 3: full delirium, i.e. features 1 and 2 are present in conjunction with either feature 3 and/or 4).
Independent variables considered a priori in this study were: 1) age, 4 categories (≤64 years, 64-79, 80-90, and >90 years) 2) gender, 2 categories (female, male) 3) civil status, 2 categories (single vs. married) 4) smoking, 2 categories (smokers vs. non-smokers) 5) special food requirements or habits, 2 categories (the group with special eating habits includes vegetarians, patients who eat kosher or who have a specific food intolerance etc. vs. "traditional Swiss cuisine") 6) alcohol use before nursing home admission, 2 categories (alcohol consumption at least once a week vs. less than once a week) 7) hearing, 4 categories ranging from 0 = good to 3 = very bad 8) eye sight, 4 categories ranging from 0 = good to 3 = very bad  [23] The data included in this analysis were anonymised, i.e. they did not allow any connection to a particular person in a specific nursing home and therefore submission to a research ethics committee was not required.

Analyses
The Statistical Package for Social Sciences (Version 15.0) was used.Frequencies, means and standard deviations were calculated where appropriate.Exploratory frequency analyses were carried out to determine subjects' characteristics and prevalence data of NHCAM levels across residence groups.Non-parametric ANOVA (Kruskal-Wallis test) was used to determine possible differences regarding prevalence of NHCAM levels as well as the different variables across residence groups.
Aiming at data reduction the factor structure of all variables was analysed using a principal components factor analysis with Va rimax rotation.This procedure created 19 factors with Eigenvalues varying between 0.44 and 2.29 with no clear-cut bend on the rather shallow screeplot slope.The principal components with eigenvalues above 1explained only 54.3% of the total variance observed.For these reasons, it was decided to use all the initial variables for the analyses.
As eries of ordinal regression analyses were then carried out to look for associations between the variables and NHCAM levels.CPS, BADL, and MDS-depression were enteredascovariates.First, all residence groups were pooled and, in as econd step, differential associations with NH-CAM were looked for separately in each residence group.Given the large sample size and the parallel analyses of the variables in five samples we corrected for multiple comparisons to retain aminimal significance level of p≤0.01.

Subjects' characteristics
The subjects' characteristics per residence group are shown in table 1.
The non-parametric ANOVA for all variables grouped according to previous residence showed highly significant differences across most variables (df = 4; p <0.01) except for eye sight (df = 4; p = 0.056).
NHCAM levels are not equally distributed across residence groups (Pearson's X 2 = 61.1;df = 15; p <0.001).In particular,subjects coming from psychiatric hospitals may have slightly more SS-1

Ta ble 1
Characteristics of study participants depending on their previous residence.

Home
Nursing Aseries of analogous ordinal regression analyses were then carried out for each residence group independently (cf.table 3).
In summary,h igher CPS and MDS-depression scores were highly significant predictors of high NHCAM values in all five residence groups.Age, civil status, continence, newly introduced drugs, and BADL were differential predictors depending on the residence group.

Ta ble 2
Prevalence in %of sub-syndromal and full-blown delirium in nursing homes stratified according to previous residence.Va riables significantly (p ≤0.01) associated with NHCAM (odds ratios, with 95% confidence intervals; only significant results are shown).Predictors of NHCAM across residence groups were CPS as well as MDS-depression.Age, civil status, continence, newly introduced drugs, and BADL were differential predictors.Note that gender fell short off the retained significance level (p = 0.016) in the group admitted from home, newly administered drugs in those from psychiatric clinics (p = 0.015), and the number of diseases (p = 0.016) in those coming from rehab units.

Home
Both sub-syndromala nd full delirium are prevalent in subjects within the first two weeks upon their entering a Swiss nursing home with an average of 39.7% of all patients showing at least some signs of delirium and 6.5% presenting with full delirium.These figures are within the range of those reported in the literature that vary,however, between nearly a quarter to over half of all nursing home residents exhibiting delirium symptoms on post-acute admission [8,13,24].Differences between findings may be due to the fact that some of the studies assessed admissions to specific postacute units in nursing facilities only,whereas the present study included all admissions to a nursing home.Although there seems to be a "monotonous greater risk of mortality with each level of delirium severity" [3], re-hospitalisation or non-discharge to the community is increased in delirious patients even for those with sub-syndromal features [13].
Despite the high prevalence of delirium and negative prognostic implications, delirium or its signs are commonly under-recognised [9].To increase diagnostic accuracy identifying groups of patients with an increased risk of delirium at admissionmay be useful.Although patients admitted from rehabilitation hospitals and skilled nursing facilities had a similar prevalence of delirium after admission [8], this is to our knowledge the first study assessing whether the place of residence before nursing home admission is associated with delirium after admission.The prevalence of delirium in our study was high for all residence groups; however,itwas even higher for those coming from psychiatric settings.T he association of a higher BADL score with higher NHCAM scores is in keeping with the association between increased dependency and delirium.However,a s the average BADL score in the psychiatric group is lower than in most other residence groups, one possible explanation is that patients with mental illness are particularly vulnerable to incident delirium, a hypothesis that our data do not allow us to test.Our data suggest that sub-syndromal and full delirium may be more frequent in nursing homes than in acute care hospitals [6,7]; thus, delirium may neither be sufficiently treated in hospitals nor perceived as a reason to stay in the hospital despite its negative prognostic meaning.Furthermore, admission to a nursing home may be a precipitating factor for delirium.
All residence groups had similar prediction patterns of NHCAM determinants.Increased dependence including more marked cognitive impairment and incontinence in some residence groups are known predictors for delirium as in our study.M ore pronounced depression was a clear predictor for delirium in our as well as a previous study [25], although this is less established in the literature.In this study,MDS-pain, the number of drugs and the number of diseases were not associated with NHCAM.This seems to contradict other studies where drugs are most often considered as either precipitating or predisposing factors [26][27][28][29].However,the number of drugs administered within 90 days prior to nursing home admission was associated with NHCAM in three residence groups.Thus, the result is only in apparent contradiction with the literature.However,these parameters may not be pivotal in nursing home cohorts with notoriously high amounts of drugs taken.
Why the married had more full (9.1%) and SS-1 (35.8%) delirium than the single (5.3% and 27.3%) in those coming from rehab clinics is open to speculation.The married may be more prone to delirium as they have to abandon their partners as opposed to the single.Premorbid habits are not important predictors of delirium.Interestingly,we have found no studies investigating the issues of premorbid habits and civil status as possible determinants of delirium although psychosocial variables may be stress factors and as such precipitators of delirium.
Despite the large sample size, which is a strength of this study,anumber of shortcomings must be mentioned.Firstly,d ata collection was done for quality control in nursing homes and this may have introduced some unidentified bias as assessment of delirium by clinical staff may be less accurate than assessment done by trained research personnel [30].We believe, however,that this bias did not influence our main findings.Furthermore, the consistency of the data had been regularly monitored as part of the RAI-MDS programme.The independent variables in this study predicted approximately 30% of the variance of the observed NHCAM scores suggesting that other variables substantially co-determine delirium in new nursing home residents within the first two weeks after admission.The data available for this study leave the question whether or not the variables identified are precipitating or predisposing factors unanswered.It would be most important to know whether delirium is new upon admission or persistent from before admission.The important question whether delirium newly evolves within the first weeks after nursing home admission or persists from before admission can only be addressed in prospective longitudinal studies.Both are, of course, possible.Nursing home was another variable influencing the occurrence of delirium suggesting that factors not assessed in this study such as staff number,training level, type and quality of medical and nursing activities as well as others significantly affect either detection or occurrence of delirium upon admission to a nursing home.Clearly,f urther studies are needed to improve the detection of delirium, to identify residents at risk and to reduce its occurrence.

Conclusions
Sub-syndromal and full delirium are astonishinglyf requent upon entry into Swiss nursing homes.Patients from psychiatric settings may have an increased risk of delirium when entering a nursing home.Increased dependence including more pronounced cognitive impairment and depression are consistently associated with higher NHCAM scores.Although factors associated with delirium differ somewhat as a function of a patient's previous residence, all patients, whatever their previous residence, must be equally carefully screened for delirium, especially for sub-syndromal delirium due to adverse outcome of these patients.