Dementia is a serious, chronic, and costly public health problem (1). A substantial portion of the costs associated with dementia involve the medical care of patients with the disorder (1, 2). At times the disorder necessitates admission to the medical-surgical units of general hospitals. A recent study (2) reported that 63% of the Medicare costs for patients with dementia of the Alzheimer’s type consists of payment for inpatient hospital care.
Little is known about the care of patients with dementia in medical-surgical units of general hospitals. Several studies, mostly from outside the United States, have reported that the prevalence of dementia among persons discharged from general hospitals is 8%–27% (3–10). Others have reported that hospital length of stay is longer for patients with dementia (4, 8–14) and that this prolonged length of stay might be reduced when a "social model" is applied, including an intensive effort to identify dementia and work on discharge soon after admission (12, 15, 16). Only two studies have investigated the in-hospital mortality rates of patients with dementia, and they report contradictory results (11, 12).
The goals of this study were to investigate the frequency of dementia among patients admitted to the Johns Hopkins Hospital and to compare the costs, lengths of stay, diagnoses, and mortality rates for patients with dementia to those for patients without dementia.
The computerized records of Johns Hopkins Hospital, collected as part of routine clinical care after each patient discharge, were analyzed without the presence of identifying information for this study. The study was thus exempt from institutional review board review. Data from psychiatric units were excluded. In 1996 and 1997 there were 76,856 discharges from the medical-surgical units of the hospital. Analyses were initially conducted for 1996 and 1997 separately, but the findings were so similar that pooled results are reported. If patients were discharged more than once, each discharge contributed a different set of observations. In these records 21,251 (28%) of the discharges involved persons age 60 and older at the time of admission. The analyses were limited to this age group because of the low prevalence of dementia before age 60.
To estimate the frequency of dementia in the study group, the proportion of all discharge diagnoses for which the attending physician recorded an ICD-9-CM code indicative of any form of dementia (290 or 331 series) was calculated. Patients with a diagnosis of dementia were then compared to patients with no discharge diagnosis of dementia on age, race, sex, primary discharge diagnosis (an indicator of the reason for which they were admitted), length of stay, and total hospital cost billed to the patient. The rates of in-hospital mortality between the two groups were compared as well.
The differences in sociodemographic characteristics or mortality rates were assessed by using t tests (age) or chi-square tests (gender, sex, and mortality rates). Differences in lengths of stay, total costs, and inpatient mortality rates were assessed for statistical significance in univariate and multivariate logistic (mortality rates) or linear (lengths of stay and costs) regression models; the latter were adjusted for age, race, and sex.
The cumulative frequency of dementia in the study group was 3.9% (823 of 21,251) compared to 4.5% in the general population of east Baltimore, which surrounds the hospital (17). In patients aged 60–64 this frequency was 2.6% (128 of 4,935). No regional population figures were available for comparison with this group. In patients aged 65–74 the frequency was 3.2% (319 of 9,954) compared to 1.2% in the general population of east Baltimore (17). In those aged 75–84 the frequency was 5.3% (281 of 5,298) versus 5.3% in general population of east Baltimore. In those 85 years and older the frequency was 8.9% (95 of 1,064) versus 25.6% in the general population of east Baltimore.
Comparison of Patients With and Patients Without Dementia
Patients with dementia were significantly older (mean age=74 years, SD=8.2) than patients without dementia (mean age=71 years, SD=7.4) (t=12.4, df=21249, p<0.001). Patients with dementia were also significantly more likely to be nonwhite than patients without dementia (47% and 51%, respectively) (χ2=4.80, df=1, p=0.03). There was no significant difference between the proportions of women in the two groups (with dementia, 49%; without dementia, 51%) (χ=1.22, df=1, p=0.27).
t1 compares the two groups on the attending physician’s primary discharge diagnoses (presumably the principal reason for the hospitalization). Given the wide range of discharge diagnoses in this group, only the 20 most frequent, along with their frequency of occurrence, are shown in the table.
t1 shows several differences in the primary diagnoses between groups. In both groups chronic ischemic heart disease was the principal discharge diagnosis. However, it was the primary diagnosis for 9.2% of the patients without dementia compared to 4.9% of the patients with dementia. The dementia itself, its causes (e.g., Alzheimer’s or Parkinson’s disease), or its complications (e.g., drug psychoses or psychotic symptoms) were the principal reasons for admission in 12.2% (N=100) of the patients with dementia but in none of the patients without dementia. Dehydration was three times more common in the dementia group, and urinary tract infection accounted for 3% of the diagnoses. Alcoholic and drug psychoses affected 5.7% of the dementia group, as opposed to 0.2% (N=41) of the group without dementia.
The two groups were also compared on rates of in-hospital mortality. A total of 5.2% (N=43) of the patients with dementia died in the hospital compared to 3.8% (N=776) of the patients without dementia (odds ratio=1.36, 95% confidence interval=0.99–1.88) (χ2=3.23, df=1, p=0.07). In a logistic regression model, the difference between mortality rates remained nonsignificant (dementia versus no dementia: odds ratio=1.25) (χ2=1.89, df=1, p=0.17) after adjusting for age, race, and sex.
Relationship Between Dementia and Length of Stay or Per Capita Cost
The mean length of stay for patients with dementia was 10.4 days (SD=10.4) compared to 6.5 days (SD=6.4) for patients without dementia (t=16.6, df=21249, p<0.0001). The modal length of stay for patients without dementia was 1 day, whereas for those with dementia, it was 3 days. A total of 18% (N=148) of the patients with dementia had lengths of stay of more than 15 days compared to 6.5% (N=1,328) of the patients without dementia. The per capita hospital cost for patients with dementia was $17,542 compared to $13,552 for patients without dementia.
We also compared mean lengths of stay and per capita hospital costs for patients in eight common diagnostic subgroups (from t1): chronic ischemic heart disease, acute myocardial infarction, heart failure, disorder of fluid or electrolyte balance, all cancers, mechanical complications of graft or implant, occlusion of cerebral arteries, and aortic aneurysm. For all diagnoses patients with dementia had a length of stay that was approximately 3–4 days longer than that of patients without dementia. Per capita hospital costs were approximately $4,000 more for patients with dementia. Differences in lengths of stay and per capita costs were consistent across diagnostic subgroups.
To confirm that differences in lengths of stay and per capita costs between patients with and patients without dementia were not due to the effects of age and sex, a series of linear regression models was computed. In the first set of models length of stay was the dependent variable, and dementia status, age, race, and sex were the independent variables. These models were estimated for the entire group and for each of the eight diagnostic subgroups just noted. In all instances, and for all diagnostic subgroups, dementia was associated with significantly longer lengths of stay, after adjusting for age and sex (in all cases, p<0.05).
In the second set of regression models total hospital cost was the dependent variable, and dementia status, age, race, and sex were the independent variables. The statistically significant relationship between dementia and higher hospital cost was maintained after adjusting for age, race, and sex (in all cases, p<0.05).
Dementia afflicts a substantial portion of elderly patients on the medical-surgical units of general hospitals—in this study, 3.9% of the patients aged 60 and older discharged from Johns Hopkins Hospital. This estimate is comparable to the population prevalence of dementia in east Baltimore (17), the neighborhood in which the hospital is located. Similar rates have been reported in other studies at general hospitals (3–10). This appears to be the first estimate of the frequency of dementia in a group of this size. In addition, we found that dementia is not associated with a higher rate of in-hospital mortality.
These findings strongly support the view that patients with dementia are admitted to general hospitals for reasons that are different than those for patients without dementia. Many of the conditions that were more prevalent in patients with dementia—such as urinary tract infections, drug psychoses, senile and presenile organic psychotic conditions, and behavioral, functional, or social complications of dementia—could potentially be prevented, recognized earlier, or managed in other settings, reducing the need for acute hospitalization (6, 8, 13). This issue merits further investigation.
A discharge diagnosis of dementia was associated both with a longer length of stay and with higher per capita cost. Higher costs are likely driven directly by longer lengths of stay. The reasons for the longer stays of patients with dementia are uncertain and can only speculated on. Since patients with dementia are predisposed to developing delirium (8), it might take them longer to recover from the complications of illness and hospital treatment. Alternatively, many patients with dementia may be admitted to the hospital, or their discharge may be delayed, because of difficulties in securing appropriate placement.
Several limitations of this investigation must be considered. Because this study was conducted at a single tertiary-care hospital, generalizability may be limited. The diagnosis of dementia relied on clinicians’ using standardized criteria. These clinicians might not have identified the condition or might not have recorded it in the chart for some patients, particularly those with milder or less complex dementia. Thus, the frequency given for dementia is likely an underestimate. Additionally, if dementia were more likely recorded for patients with more complex presentations, a bias in favor of a false association between dementia and longer length of stay or higher per capita cost would be introduced. The study also suffered from its reliance on a discharge database. There was no estimate of how critically ill the patients were at admission, and there were no data on where patients were before admission or to where they were discharged.
In summary, patients with dementia are admitted to general medical hospitals for different diagnostic problems than are patients without dementia. They also have longer hospital stays and increased associated costs. Since European studies have reported that in hospitals employing an intensive program of early screening and intervention, the lengths of stay of patients with dementia are comparable to those of patients without dementia (12, 15, 16), the screening and early identification of dementia in general hospitals should be evaluated as a means of reducing lengths of stay and costs. This would also benefit patients with dementia because they would spend less time in the hospital, an environment that is at times confusing and hostile to them. If the lengths of stay for the patients with dementia in our study had been the same as those for patients without dementia, a cost savings of $3,292,000 would have been achieved over 2 years (a per capita difference of $4,000 over 823 cases).
Received Jan. 4, 1999; revisions received April 22 and July 27, 1999; accepted Aug. 24, 1999. From the Neuropsychiatry and Memory Group, Department of Psychiatry and Behavioral Sciences, School of Medicine, Johns Hopkins University. Address reprint requests to Dr. Lyketsos, Osler 320, Johns Hopkins Hospital, Baltimore, MD 21287; firstname.lastname@example.org (e-mail). Supported by a grant from the Alzheimer Association and NIMH grant MH-56511.