Mean total health care costs (reported in 2005 dollars) by month, adjusted for all of the variants in the regression models, specifically, index hospitalization; patient age, race, and sex; whether the patient received the delirium prevention intervention; Charlson comorbidity score; whether the patient had dementia; the number of impairments in activities of daily living; whether the patient died during follow-up; and the interaction of the Charlson comorbidity score with whether the patient died.
Leslie DL, Marcantonio ER, Zhang Y, Leo-Summers L, Inouye SK. One-Year Health Care Costs Associated With Delirium in the Elderly Population. Arch Intern Med. 2008;168(1):27-32. doi:10.1001/archinternmed.2007.4
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
While delirium has been increasingly recognized as a serious and potentially preventable condition, its long-term implications are not well understood. This study determined the total 1-year health care costs associated with delirium.
Hospitalized patients aged 70 years and older who participated in a previous controlled clinical trial of a delirium prevention intervention at an academic medical center between 1995 and 1998 were followed up for 1 year after discharge. Total inflation-adjusted health care costs, calculated as either reimbursed amounts or hospital charges converted to costs, were computed by means of data from Medicare administrative files, hospital billing records, and the Connecticut Long-term Care Registry. Regression models were used to determine costs associated with delirium after adjusting for patient sociodemographic and clinical characteristics.
During the index hospitalization, 109 patients (13.0%) developed delirium while 732 did not. Patients with delirium had significantly higher unadjusted health care costs and survived fewer days. After adjusting for pertinent demographic and clinical characteristics, average costs per day survived among patients with delirium were more than 2½ times the costs among patients without delirium. Total cost estimates attributable to delirium ranged from $16 303 to $64 421 per patient, implying that the national burden of delirium on the health care system ranges from $38 billion to $152 billion each year.
The economic impact of delirium is substantial, rivaling the health care costs of falls and diabetes mellitus. These results highlight the need for increased efforts to mitigate this clinically significant and costly disorder.
Delirium, characterized as an acute decline in cognition and attention, represents a common and severe problem for hospitalized older patients, with occurrence rates from 14% to 56% and hospital mortality rates from 25% to 33%.1,2 The development of delirium has been associated with increased morbidity, persistent functional decline, increased nursing time per patient, higher per-day hospital costs, increased length of hospital stay, higher rates of nursing home placement, and increased mortality.3- 6 Delirium often initiates a cascade of events that can include functional decline, caregiver burden, increased morbidity and mortality, and higher health care costs.3- 5,7- 10 The problem of delirium in older hospitalized patients has assumed particular importance because patients 65 years and older currently account for more than 48% of all days of hospital care.11
Although the short-term implications of delirium have been well documented, recent evidence2- 6,8,10,12- 17 suggests that delirium also has substantial long-term sequelae with significant implications for health care utilization and costs. However, previous studies of health care costs related to delirium have been limited to specific services (ie, hospital length of stay, intensive care unit stay, or nursing home care). To document the broader economic and health care burden of delirium, we determined the long-term direct health care costs associated with delirium. The present study provides a comprehensive cost estimate for all direct health care services from the index hospitalization through 1 year after discharge.
The study sample consisted of 841 individuals who participated in a controlled trial of a delirium prevention intervention at Yale–New Haven Hospital between 1995 and 1998. Details of the study are described elsewhere.18 Briefly, patients meeting the following criteria were enrolled: consecutive admissions to 3 non–intensive care general medical units, aged 70 years or older, no evidence of delirium at admission, and intermediate or high risk for delirium based on a previously developed risk model.19 Patients who could not participate in interviews (eg, profound dementia, language barrier, profound aphasia, intubation, coma, or respiratory isolation), who had a terminal illness, who had a hospital stay of 48 hours or less, or who had prior enrollment in the study were excluded. Informed consent for participation and permission to acquire subsequent follow-up data were obtained from the patients, or from a proxy for those with substantial cognitive impairment, according to procedures approved by the institutional review board of the Yale University School of Medicine.
Delirium was ascertained daily during hospitalization by the Confusion Assessment Method,20,21 with delirium defined by the presence of acute onset and fluctuating course, inattention, and either disorganized thinking or altered level of consciousness. Patients who developed delirium while hospitalized were identified, and all patients were followed up for up to 1 year after discharge to determine health care service use and costs. Of the 919 subjects enrolled in the original trial,18 25 were excluded because they could not be linked to the Medicare files, 50 were excluded because they were enrolled in a Medicare managed care health maintenance organization and hence did not have detailed cost data, and 3 were excluded because they were missing cost data from the index hospitalization. Thus, the final study sample, which included both intervention and control subjects, consisted of 841 individuals.
Information on patient demographic characteristics, comorbidities, and functional status were obtained from primary data collected during the controlled trial. Data on health care service use and costs, including inpatient, outpatient, nursing home, home health, rehabilitation, and other services, were obtained from Medicare Parts A and B administrative claims files for these patients. Additional service use and cost data were obtained from Yale Medical Information Systems for the index hospitalization and subsequent readmissions to Yale–New Haven Hospital. Because Medicare nursing home coverage is limited to 100 days of care and information on stays beyond this limit may be inaccurate or missing, the Connecticut Long-term Care Registry was used to supplement the Medicare files. The Long-term Care Registry is a longitudinal database containing demographic, health status, and nursing home length of stay information (including dates of all nursing home admissions and discharges) for all Connecticut nursing facility resident stays.
Patient deaths were identified by telephone follow-up contacts at 1-, 6-, and 12-month periods; by daily obituary review; and by the Social Security Death Index. All deaths and dates of death were confirmed by at least 2 sources: review of medical records, death certificates, systematic obituary review, Medicare Enrollment and Claims files, and/or National Death Index or Social Security databases.
Total health care costs for patients in the controlled trial were computed during the index hospitalization and through 1 year after discharge. For costs incurred during the index hospitalization, hospital charges were converted to costs by means of the hospital-specific cost to charge ratio. For all other services, costs were calculated with the use of Medicare reimbursed amounts rather than charges because reimbursed amounts are payments actually received by providers for their services and hence are a better measure of transaction prices than billed charges.22- 24 For patients with unqualified nursing home days (ie, days not reimbursed by Medicare because they exceed the 100-day limit), the number of additional days of care for these patients was determined from the Medicare records or Long-term Care Registry, and costs for these days were imputed by means of the average daily cost of care associated with the nursing home in which the patient was admitted. Costs were adjusted for inflation by means of the medical care component of the consumer price index and are reported in 2005 dollars.
We used SAS statistical software, version 9.1 (SAS Institute Inc, Cary, North Carolina) for all analyses. We first compared unadjusted mean total costs across the delirium and nondelirium groups by means of a Wilcoxon test. Next, we calculated adjusted mean total costs by linear regression models. Independent variables in the model included whether the patient had delirium during the index hospitalization, patient age, race, sex, whether the patient received the delirium prevention intervention, Charlson comorbidity score, whether the patient had dementia, the number of impairments in activities of daily living, whether the patient died during the study period, and an interaction term of the Charlson comorbidity score with whether the patient died during the study period. We explored other interaction terms as well, but the interaction of the Charlson score and whether the patient died was the only interaction term that significantly improved the fit of the model. Because traditional ordinary least-squares regression is not appropriate for skewed data, costs were log-transformed before running the regression model, and adjusted average total costs were retransformed to the nonlog scale by means of the smearing estimator,25 after ascertaining that the log-scale residuals were homoscedastic.26
Because some patients died during the study period, costs may be right-censored. Moreover, if more patients with delirium than patients without delirium died before the end of the study period, the costs associated with delirium may be underestimated. To account for this potential bias, total direct health care costs were also modeled in 2 additional ways. First, total costs were divided by total days survived to derive an average cost per day survived. Adjusted costs per day survived were computed for patients with delirium and for those without delirium by the same regression model techniques described in the preceding paragraph, with average cost per day survived used as the dependent variable. These adjusted average costs per day survived were then multiplied by the average number of days survived in each group to derive a total cost for each group. Standard errors of these total cost estimates were calculated by bootstrapping methods,27 and an unpaired, 2-tailed t test was used to compare costs across the delirium and nondelirium groups.
The second approach was to use a partitioned estimator to model total costs based on methods developed by Lin et al28 and Bang and Tsiatis.29 The study period was divided into 1-month intervals, and average total direct health care costs for patients with and without delirium were computed in each month among individuals who survived to the end of that month. A Cox proportional hazards regression model was used to estimate fitted Kaplan-Meier estimators for surviving to the end of each month, and costs were summed across months with the Kaplan-Meier estimators used as inverse weights. Bootstrapping methods27 were again used to compute standard errors for the cost estimates, and an unpaired, 2-tailed t test was used to compare costs across the delirium and nondelirium groups.
Characteristics of the sample are presented in Table 1. Of the 841 individuals included in the study sample, 109 (13.0%) developed delirium during the index hospitalization. A higher proportion of patients with delirium were admitted from a nursing home, had comorbid dementia, or died during the study period (Table 2) compared with patients who did not develop delirium. Patients with delirium also had more impairments in activities of daily living, higher Charlson and Acute Physiology and Chronic Health Evaluation II scores, and lower Mini-Mental State Examination scores. A smaller proportion of patients who received the delirium prevention intervention developed delirium compared with patients who did not receive the intervention.
As shown in Table 2, patients with delirium survived an average of 256 days during the 1-year follow-up period, compared with 322 days for patients without delirium, although this difference was not statistically significant (P = .89). Despite the shorter survival time, total unadjusted health care costs were significantly higher for patients who developed delirium during the index hospitalization than for those without delirium (mean [SD],
$69 498 [$59 120] vs $47 958 [$45 640], respectively; P < .001). Total costs per day survived were also higher for patients with delirium than for those without, both among patients who died during the study period and among those who survived.
Results from the regression models showed that patients with delirium had significantly higher costs than patients without delirium even after adjusting for relevant demographic and clinical characteristics. As expected, patients with higher Charlson scores, who had dementia, or who died during the follow-up period also had significantly higher total health care costs. Receipt of the delirium prevention intervention did not significantly affect costs (data not shown). Adjusted total health care costs by month for the delirium and nondelirium groups based on the regression models are illustrated in the Figure. Adjusted costs were higher for the delirium group in each month. The difference in adjusted total costs between the delirium and nondelirium groups was initially relatively large ($6613 in the first month), then declined over time until about month 5, and then generally increased again through month 9.
As shown in Table 3, adjusted total costs were significantly higher for the delirium group than for the nondelirium group. Total costs per day survived were more than 2½ times higher for patients with delirium than for patients without delirium. In the model that ignores the right-censoring problem (method 1), costs for patients with delirium were $16 303 higher than for those without delirium. Costs attributable to delirium were higher in the 2 models that accounted for the fact that the data were right-censored (methods 2 and 3), ranging from $60 516 to $64 421. Ninety-five percent of the difference in costs was due to inpatient and nursing home care.
This study documents the considerable direct health care costs associated with delirium in the United States. We estimate that delirium is responsible for between $60 516 and $64 421 in additional health care costs per delirious patient per year. Following Inouye et al2 and assuming that delirium complicates hospital stays for 20% of the 11.8 million persons 65 years and older who are hospitalized each year, our results imply that total direct 1-year health care costs attributable to delirium range from $143 billion to $152 billion nationally. These estimates are adjusted for the difference in survival time. Even when we use our most conservative estimate, which ignores the right-censoring problem, costs associated with delirium exceed $38 billion per year. Given that a number of effective interventions have been developed to prevent or treat delirium,18,30- 35 at least some of these costs may be avoidable.
We took great care not to underestimate costs associated with delirium due to more patients with delirium dying before the end of the study period than patients without delirium. However, costs may also be underestimated if patients with delirium die quietly, ie, without additional diagnostic or therapeutic intervention. To explore this possibility, we compared average daily costs for patients with and without delirium stratified by whether they survived the entire study period. Average daily costs were significantly higher for the patients with delirium regardless of whether they died during the study period (Table 2). Although in our secondary data analysis we did not demonstrate the cost savings for delirious patients who die quietly, this remains a possibility for a subset of patients, which may bias our results toward underestimating the costs associated with delirium.
National annual health care costs have been estimated for a number of conditions, including hip fracture ($7 billion),36 nonfatal falls ($19 billion),37 diabetes mellitus ($91.8 billion),38 and cardiovascular disease ($257.6 billion).39 While we acknowledge the difficulty and limitations in comparing across conditions owing to differences in study methods, diagnostic overlap, and shared comorbidities, our results suggest that the economic burden of delirium is substantial, even relative to other conditions.
The pattern of costs over time is interesting. As previous studies have shown,8,10,40- 42 delirium increases hospital length of stay and costs, so the large initial costs associated with delirium are not surprising. The increased costs later in the period may be due to recurrence of delirium or terminal care costs, although more research is needed to explore the sources of these costs.
We included patients in the study sample who had received the delirium prevention intervention to have the largest possible sample size. Although these patients had lower rates of delirium than patients in the control group, receipt of the delirium prevention intervention did not significantly affect costs in the multivariate models. To the extent that including these patients biases our results, we would argue that the bias would be conservative, because, if anything, delirium in the intervention group would have been anticipated to be less costly. Moreover, as a sensitivity analysis, when the sample was limited to just the usual-care patients who did not receive the intervention, the costs associated with delirium were not substantially different (data not shown).
Although previous studies have demonstrated the increased hospital and nursing home costs associated with delirium,5,41,42 this study is the first, to our knowledge, to document the costs associated with delirium across such a wide range of services (inpatient, intensive care unit, emergency department, outpatient, nursing home, home health, rehabilitation, and other services) and during such a long period. While the study has a number of strengths, such as the availability of detailed clinical information and comprehensive service use and cost data from multiple sources, some limitations of the analysis deserve comment. First, although our cost estimates are adjusted for a number of patient sociodemographic and clinical characteristics, there may be residual confounding due to inherent differences between the delirium and nondelirium groups that might affect costs. However, we believe that any bias introduced by such residual confounding would be small because we are able to include a number of detailed clinical measures in our models. Second, cost estimates are derived from a single site only, and hence the generalizability of the results may be limited. In addition, cost estimates include direct health care costs only and do not take into account important indirect costs associated with caregiver burden or reduced quality of life. Finally, follow-up was truncated at 1 year; therefore, any costs associated with delirium that accrue more than 1 year after discharge are not included.
Despite these limitations, it is clear that the economic burden of delirium is substantial. It is our hope that these results draw attention to delirium as a serious condition with significant long-term clinical and economic implications. Future research will need to focus on the specific sources of the increased health care costs associated with delirium. Given that the condition is costly, increasing in magnitude with the aging population, and potentially preventable, increased efforts to prevent, detect, and treat delirium are urgently needed.
Correspondence: Douglas L. Leslie, PhD, Department of Health Administration and Policy, Medical University of South Carolina, 151 Rutledge Ave, Bldg B, PO Box 250961, Charleston, SC 29425 (firstname.lastname@example.org).
Accepted for Publication: August 12, 2007.
Author Contributions: Drs Leslie, Marcantonio, Zhang, and Inouye and Leo-Summers had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Leslie and Inouye. Acquisition of data: Leo-Summers and Inouye. Analysis and interpretation of data: Leslie, Marcantonio, Zhang, and Inouye. Drafting of the manuscript: Leslie and Inouye. Critical revision of the manuscript for important intellectual content: Leslie, Marcantonio, Zhang, and Inouye. Statistical analysis: Leslie and Zhang. Obtained funding: Inouye. Administrative, technical, and material support: Leslie, Leo-Summers, and Inouye. Study supervision: Leslie and Inouye.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by grants R01AG12551, R21AG025193, and K24AG00949 (Dr Inouye) and R21AG026566 (Dr Marcantonio) from the National Institute on Aging.
Role of the Sponsor: The sponsor did not participate in the design and conduct of the study; in the collection, management, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.