Mortality and hallucinations in nursing home patients with advanced Parkinson's disease

Placement in a nursing home or other chronic-care facility is considered a significant and often irreversible medical and social decision ^[1]. In Parkinson's disease (PD), the factors associated with nursing home placement have not been studied in detail. In a prior case-control study of 11 patients with PD who were admitted to nursing homes and 22 matched controls who remained in the community, ^[2] we found that hallucinations were a significant risk factor for nursing home placement. In contrast, dementia and motor disability, while present in both groups, did not significantly differentiate those requiring nursing home placement from those who could remain in the community.

In this follow-up study, we examined the consequences of nursing home placement in these patients and monitored the hallucinatory status over an additional 2 years from the original study's close. Specifically, we examined (1) the comparative mortality outcome of nursing home and community-dwelling patients, (2) the stability of hallucinatory behaviors, and (3) within the original control population of community-dwelling PD patients, the association between hallucinations and later nursing home placement.

Methods. Subjects. For the patient sample, described previously, ^[2] we identified all PD patients from our tertiary care university practice who were admitted to nursing homes between 1987 and 1991. The diagnosis of PD was based on the presence of at least three of the following: rest tremor, bradykinesia, rigidity, postural reflex impairment, and no evidence of secondary parkinsonism.

To perform a case-control study, we identified two control patients for each case; controls were patients with PD who were living in the community and who matched the cases for gender, age within 2 years, and duration of PD symptoms within 2 years. The mean age of the 11 cases was 75.6 +-\4.9 years (SD) and 75.8 +-\4.9 years for controls (NS). Mean disease duration was 14.4 +-\5.4 years for cases and 14.5 +-\4.7 years for controls (NS). Fifty-five percent of each group were men.

Follow-up interviews. This study was conducted in February 1994, 2 years after the close of the prior study and a mean of 50.6 months (range, 24 to 80 months) from time of nursing home admission or comparable data collection for community-dwelling controls. All community-dwelling patients were seen at least once yearly for neurologic care. At these visits, patients received ratings on the Unified Parkinson's Disease Rating Scale (UPDRS) ^[3] as part of their regular office care. Patients were scored during an "on" phase, when they were experiencing medication effects. We used the total Motor subsection score of the UPDRS to assess parkinsonism, the Intellectual Impairment subscore to assess dementia, and the Thought Disorder subscore to assess hallucinations. We defined hallucinators as patients with a score of >=3 on the Thought Disorder subscale of the UPDRS (occasional to frequent hallucinations or florid psychosis) and patients with >=3 (severe memory loss or inability to make judgments) on the Intellectual Impairment subscale as having significant memory problems.

We contacted nursing home subjects or their primary family contact to determine whether patients were alive or dead and whether they were still living in a nursing home. To assess hallucinations, we obtained the UPDRS Thought Disorder subscale score by performing the test by telephone or we gathered the data by reading the scale to the primary family contact. In all cases, this person was the spouse or a child who visited or had visited the patient at least twice monthly throughout the follow-up period or until death. The score was based on current status or the patient's state 1 month prior to any acute medical complication that led to death.

Stability of hallucinations. To determine the stability of hallucinations and the likelihood of hallucination presence or absence predicting the same status at follow-up, we calculated positive and negative predictive values using Bayes' theorem ^[4]. We developed 2 x 2 tables listing the number of patients without or with hallucinations at each of the two assessment points. The classification resulted in the following four groupings: (NH NH) number of nonhallucinators at both assessments; (NH H) number of original nonhallucinators who were hallucinators at follow-up; (H NH) number of original hallucinators who were nonhallucinators at follow-up; and (H H) number of hallucinators at both assessments.

We calculated positive and negative predictive values (PVs) with confidence intervals based on study-based probabilities of hallucinatory status ^[4]. In calculating positive and negative PVs, we assumed equal prior probability of the presence or absence of hallucinations since there was no reason to expect a differential effect of prior hallucinatory status on stability of hallucinations. As such, the positive predictive value (+PV) determined the likelihood that a nonhallucinator at the first evaluation would show stability and remain a nonhallucinator on the second evaluation and was calculated as:

1PV = (NH NH)/((NH NH) + (H NH)).

Likewise, the negative predictive value (-PV) determined the likelihood that a hallucinator at the first evaluation would show stability and remain a hallucinator on the second evaluation and was calculated as:

2PV = (H H)/((H H) + (NH H)).

Data analysis. We used a chi-square test to compare the mortality rate and the stability of hallucinations between nursing home residents and community-dwelling controls. To assess differences in motor disability, dementia, and hallucinations in community-dwelling controls who remained in the community versus those who were admitted to a nursing home, we used a chi-square test. A Mann-Whitney U test was used to assess differences in severity of motor disability, intellectual impairment, and thought disorder between these two groups. Ninety-ninth percent confidence intervals (99% CI) were calculated for the positive and negative PVs.

Results. Mortality. Mortality rate for the original nursing home patients was 100% compared with 32% for the original control patients (chi squared (chi²) = 11.14, p < 0.001). Five of these latter patients died at home and two after admission to a nursing home. For all patients admitted to a nursing home, the mean survival time after admission was 15.6 months (range, 3 to 50).

Prevalence of hallucinators and stability of hallucinations. The number of hallucinators on follow-up was greater, but not significantly so, than in the original survey (52% versus 33%, chi squared (chi²) = 1.55, p > 0.05). Hallucinators were significantly more numerous among the nursing home population than in the community dwellers at follow-up (100% versus 23%, chi squared (chi²) = 17.53, p < 0.001).

At follow-up assessment, 82% of patients retained their original status as a hallucinator or nonhallucinator. All hallucinators from the first study continued to be hallucinators at the follow-up assessment regardless of nursing home or community residence (negative PV, 1.00; 99% CI = 0.283, 0.692; p = 0.01). Similarly, 76% of the original nonhallucinators remained without hallucinations (positive PV, 0.73; 99% CI = 0.164, 0.553; p = 0.01). Of the six new hallucinators, one was originally a nursing home case and five belonged to the original community control group.

Risk factors for nursing home admission. Four subjects from the original community-dwelling control group entered nursing homes during the follow-up period. Comparison of these subjects' last recorded disability scores prior to nursing home placement with the last outpatient visit in community-dwelling patients confirmed our original observation. The two groups scored similarly for motor function (median UPDRS motor score = 27.50 nursing home versus 27.00 community dwellers, U = 30.0, p > 0.05) and dementia (median intellectual impairment score = 2.50 nursing home versus 1.00 community dwellers, U = 16.5, p > 0.05), but significantly differently for thought disorder (median thought disorder score = 3.50 nursing home versus 0.00 community dwellers, U = 1.0, p < 0.001). The presence of hallucinations was significantly more frequent among patients transferred to nursing homes (chi squared (chi²) = 11.68, p < 0.001).

Discussion. PD represents an important primary diagnosis in nursing homes. Based on utilization statistics and a survey study of 40 nursing homes, 2 to 5% of the total nursing home population have PD ^[5-7]. Among 267 PD patients assessed in Aberdeen, Scotland, 23.8% lived in a hospital setting or in specialized residences for the elderly ^[5]. These data demonstrate the magnitude of impact posed by PD on supervised chronic care facilities.

In all instances, our patients entered nursing homes as a last residence and never left other than for day trips, holidays, or weekends. This finding contrasts with studies of nursing home admissions in general elderly subjects where transient stays account for approximately 30% of admissions ^[8]. However, in general elderly nursing home admissions, factors associated with permanent nursing home admissions included combined cognitive and functional impairment, both prominent features in our patient group. Likewise, our mortality rate was higher (100%) than that reported in studies of general elderly subjects admitted to nursing homes ^[9]. In those latter studies, however, poor survival correlated directly with low activity levels, a clinical feature of all our subjects ^[10]. Furthermore, PD itself is associated with an increase ^[11] in relative mortality rate as high as 2.35, with higher rates among patients with more motor and cognitive impairment. No study of general elderly subjects or PD patients has otherwise focused on hallucinations as a risk factor for death.

In our sample, once patients became hallucinatory, there was a strong tendency for them to remain in this category on subsequent evaluation. Goetz et al ^[12] observed that drug-induced hallucinations developed and worsened with increased or stable doses of dopaminergic or anticholinergic drugs. We did not have reliable medication information on the nursing home subjects in this study and cannot make conclusions regarding medication correlation with hallucinations. In this analysis, we considered hallucinations, motor function, and dementia as independent variables with full recognition that antiparkinson medications can have an influence on all three ^[13,14]. Negative findings of associations between nursing home admissions and motor disability or dementia could relate to lack of sufficient statistical power. The thought disorder inventory on the UPDRS is a single assessment that encompasses dreams, hallucinations, and psychotic behavior, and we had only two time points to test stability of hallucinations. We are in the process of assessing a larger sample of patients with different intensities of hallucinations using more extensive assessments over several time points to determine if our findings can be extended beyond the nursing home population and advanced PD patients.

Our finding that hallucinations were a significant risk factor for nursing home placement reinforces our original observation ^[2]. We emphasize that patients with advanced PD remaining in the community as well as those in nursing homes were motorically very disabled and demented, but only the presence of hallucinations in the nursing home patients differentiated the groups. Since mortality was associated with nursing home placement, which itself was linked to hallucinations, we conclude that assertive treatment interventions of hallucinations are important in advanced PD. We recognize that the sample size was small and that our tertiary care practice continued for the community-based patients and stopped for the nursing home patients. Whereas hallucinations were the focus of this study, other elements, both medical and psychosocial, may potentially contribute to the different survival outcomes of the two groups.

Treatment for hallucinations in PD is limited to modalities that are regularly problematic. Reductions of antiparkinsonian medications cause expected motor decline, drug-free periods or "holidays" carry inherent medical risks, neuroleptic drugs exacerbate parkinsonism, and clozapine is associated with high cost and the potential lethal risk of bone marrow dyscrasia. Because of these risks, clinicians often manage hallucinations with no or only minor medication adjustments, and do not energetically aim to stop them. With knowledge of the serious risks of untreated hallucinations in terms of nursing home placement and associated mortality, physicians may reconsider more assertive attention and experimental interventions as reasonable choices ^[12].