Author Affiliations: Department of Biostatistics, Mailman School of Public Health, Columbia University, New York, New York (Mr Goldfeld); and Department of Health Care Policy, Harvard Medical School (Dr Stevenson), Department of Medicine, Beth Israel Deaconess Medical Center (Drs Hamel and Mitchell), and Hebrew SeniorLife Institute for Aging Research (Dr Mitchell), Boston, Massachusetts.
Dementia is a leading cause of death among older Americans.1 In 2000, 5 million adults in the United States had dementia, and projections predict 13 million by 2050.2 Health care expenditures for dementia are estimated to total $172 billion in 2010.2 These costs will rise as the number of persons living to experience the end stage of this disease increases.
Although 70% of persons with dementia die in nursing homes (NHs),3 most research describing health care spending for these individuals has been conducted in the community setting, and almost none has focused on end-stage disease. For Americans older than 65 years, Medicare covers the costs of acute care, subacute care, physician and other provider services, hospice, prescription drugs, and diagnostic tests. Medicare does not cover NH care, which is generally paid for by Medicaid after individuals exhaust their own resources. The total Medicare and Medicaid payments for patients with dementia are roughly 3 times higher than for age-matched controls.2,4 Total direct costs increase substantially over the course of the disease, with much of the increase in the later stages attributable to spending for NH care.5 Although per diem Medicaid payments account for most public spending within the NH setting, limited data suggest that Medicare expenditures explain most of the spending variation among dying residents with dementia.6 How this spending breaks down across the various Medicare services has not been well described.
Nursing home residents with advanced dementia commonly experience potentially burdensome and costly interventions (eg, hospital transfer, tube feeding) that may be of limited clinical benefit.7 Among patients with cancer, advance care planning has been shown to be associated with lower costs in the last week of life, and lower expenditures are associated with a higher-quality dying experience.8 We are not aware of prior research identifying factors that influence health care costs in end-stage dementia.
Therefore, the objectives of this study were to describe and examine factors associated with Medicare expenditures in advanced dementia. We used data from a multisite cohort study that prospectively followed up 323 NH residents with advanced dementia for 18 months.7,9
The study included participants in the Choices, Attitudes, and Strategies for Care of Advanced Dementia at the End-of-Life (CASCADE) study,9 a prospective cohort study conducted between February 2003 and February 2009 that described the experience of NH residents with advanced dementia and their families. The detailed methods of this study are provided elsewhere.7,9
Residents were recruited from 22 NHs with more than 60 beds and within 60 miles of Boston, Massachusetts. Resident eligibility criteria included (1) age older than 60 years, (2) dementia (any type) (determined from medical record), and (3) Global Deterioration Scale score of 7 (ascertained by nurse interview).10 A Global Deterioration Scale score of 7 is characterized by profound memory deficits (unable to recognize family), limited verbal communication (<5 words), incontinence, and inability to ambulate. Residents had to have English-speaking health care proxies (HCPs) who provided informed consent for their participation and for the residents' participation. The institutional review board of Hebrew SeniorLife Institute for Aging Research, Boston, approved the study's conduct.
Resident assessments were conducted at baseline and quarterly for up to 18 months through chart reviews, nurse interviews, and clinical assessments. If the resident died, a chart review and a nurse interview were conducted within 14 days of death. Telephone interviews with HCPs were conducted at baseline and quarterly for up to 18 months.
Medicare health services used during the intervals between assessments were collected from residents' charts. They included hospital admissions, emergency department (ED) visits, primary care provider visits in NH, and hospice enrollment. Hospitalization data included the admission date, length of stay, and reasons for admission (abstracted from the hospital discharge summary by research assistants). Emergency department data included dates of service and primary reason for the visit. Dates of enrollment into the Medicare hospice benefit and the number of physician and nurse practitioner visits (recorded separately) were also abstracted from the medical records.
The use of skilled nursing facility (SNF) services was not collected as part of the CASCADE study. Therefore, we identified all hospital admissions lasting at least 3 nights (required to qualify for an SNF stay). Senior NH administrators were contacted retrospectively to determine whether the resident received SNF care after hospitalization and, if so, the length of the SNF stay and the Resource Utilization Group codes used to bill Medicare.
Expenditures for Medicare services were estimated using publically available sources and based on nationally representative rates from 2007 in US dollars. Long-term care expenditures, mostly paid by Medicaid, were not examined.
Hospital admission expenditures were estimated using inpatient data from the 2007 Nationwide Inpatient Sample (NIS), Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality.11 The NIS database represents an approximate 20% stratified sample of in-patient stays from more than 1000 hospitals located in 40 states. Matching with the NIS records was conducted as follows: (1) the diagnoses identified in the CASCADE study as the primary reason for admission were matched with the closest principal clinical classifications software or International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes; (2) secondary diagnoses were matched with the 15 secondary clinical classifications software or International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes; and (3) dementia was considered as a secondary diagnosis for all admissions. For each hospitalization, the estimated average charge was converted to an average cost using the charges-to-cost ratios provided by the Healthcare Cost and Utilization Project.
Expenditures for ED visits not resulting in a hospital admission were calculated from Medicare reimbursement rates for facility and physician services. Facility costs were based on the 2007 National Centers for Medicare and Medicaid Services Hospital Outpatient Prospective Payment System, which uses ambulatory classification codes to determine reimbursements.12 Physician costs were based on Current Procedural Technology (CPT) codes in the National 2007 Centers for Medicare and Medicaid Services physician fee schedule.12 In consultation with an ED physician with specific expertise in ED billing, ambulatory classification and CPT codes for evaluation and management services (ranging from 1 [lowest complexity] to 5 [highest complexity]) were assigned for each ED visit based on typical billing practices for specific diagnoses. Expenditures for laboratory services and other minor tests were not considered. However, estimates for the following investigations and procedures were included in ED expenditures: (1) unenhanced computed tomographic scan of the head for all head injuries; (2) bone x-ray films for fracture evaluations (eg, knee x-ray films for the evaluation of knee fractures); (3) chest x-ray films for questionable pneumonia; (4) suture repair for lacerations; and (5) replacement of a percutaneous endoscopic gastrostomy tube for blockages or dislodgments.
Skilled nursing facility payment rates were estimated from the published 2006-2007 Medicare per diem fee schedule based on 53 Resource Utilization Group III codes.13 Information was available for 53 of 58 SNF-eligible hospitalizations in our cohort. Estimates for the remaining 5 cases were imputed using the average cost of the nonmissing data based on admissions that shared the same primary diagnosis.
Primary care provider unit costs were estimated using 2007 Medicare physician reimbursement rates obtained from the US Department of Health's Centers for Medicare and Medicaid Services.14 Physician and nurse practitioner data from the CASCADE study included the number of visits but not the CPT codes. Therefore, the distribution of the CPT codes was estimated from a 2002 database of a nationally representative sample of NHs (n = 202),15 which included all Medicare Part B claims submitted by physicians and nurse practitioners for residents with cognitive performance scores of 5 or 6 (severe to very severe cognitive impairment) (N = 18 243 claims).16 The average expenditures per visit for each provider type reflected the average reimbursement for services weighted by the frequency of CPT codes for each of the provider types. Hospice costs were based on the published 2007 Medicare per diem reimbursement ($130.79) for the “routine home care” category of hospice, a category that accounts for the vast majority of hospice days.17,18
To examine factors associated with expenditures, variables were selected a priori from the CASCADE data set based on the literature.4,6- 8,19,20 Baseline resident characteristics obtained from the chart included demographics (sex, race [white vs other], and age); whether or not the resident lived in a special care dementia unit; NH length of stay (<2 years, 2-5 years, and >5 years); and comorbidities (eg, congestive heart failure, active cancer, chronic obstructive lung disease); and presence of a percutaneous endoscopic gastrostomy tube. Additional resident variables were collected at baseline and at each follow-up assessment. Cognitive status was measured by direct examination using the Test for Severe Impairment (TSI) score (range, 0-24; higher scores indicate better cognitive function).21 The TSI was categorized as either 0 or greater than 0 in these analyses. Functional status was quantified by the residents' nurses using the Bedford Alzheimer Nursing Severity Scale (range 7-28; higher scores signify greater disability).22 The presence of a do-not-hospitalize (DNH) order was recorded. Finally, the determination was made as to whether the resident experienced at least 1 acute major illness (febrile episode, pneumonia, or other [eg, hip fracture, stroke, myocardial infarction]) in the interval between assessments.
Baseline HCP characteristics included sex, relationship to the resident (child vs other), and whether the HCP claimed to understand the health problems that are common in advanced dementia. At baseline and at every follow-up interview, it was also determined whether the HCP thought the resident had 6 months or less to live.
Means and proportions were calculated for all continuous and categorical variables, respectively. Health service use and expenditures were calculated for all residents (N = 323) as well as separately for the subset of residents who died (n = 177). The total number of hospital admissions, ED visits, primary care provider visits, and hospice admissions and the mean total number of days spent in the hospital and hospice were also determined. The total and mean expenditures were calculated for each Medicare service and for all services combined. Finally, mean expenditures for each 90-day period of follow-up were calculated as well.
Among residents who died, expenditures were allocated to each 90-day interval in the year before death (0-90 days, 91-180 days, 181-270 days, and 271-360 days). Mean expenditures during each interval were determined for (1) hospice services, (2) acute care services (inpatient and ED visits), (3) SNF services, and (4) provider services (physician and nurse practitioner primary care visits). Trends for intervals approaching death were estimated using generalized estimating equations, which were based on a gamma distribution and a log-link function for the mean.
Analyses were conducted to identify resident and HCP characteristics (independent variables) associated with total 90-day expenditures for 2 outcomes: (1) all Medicare services and (2) all Medicare services excluding hospice. We hypothesized that different factors would be associated with these 2 outcomes, as hospice represents a palliative approach to care, while other Medicare services (eg, hospitalizations) generally reflect a more aggressive approach. In both models, the level of analysis was the 90-day interval of observation corresponding to quarterly follow-up assessments. Static independent variables (eg, sex) came from the baseline assessments, while dynamic variables (eg, acute illness) were from the assessment closest to the 90-day expenditure interval. Because of the positive skewness, models assumed a gamma distribution, and mean 90-day expenditures were related to the covariates through a log link. Clustering at the resident level was accounted for using an exchangeable covariance structure. The measure of association for a characteristic was expressed as the natural log of the ratio of mean expenditures for observations with that characteristic relative to mean expenditures for observations without that characteristic. Ninety-five percent confidence intervals were calculated around this measure. Unadjusted associations between independent variables and 90-day Medicare expenditures were estimated. Independent variables associated with the outcome (P ≤ .10) were included in the multivariate model. All statistical analyses were conducted using SAS version 9.2 (SAS Institute Inc, Cary, North Carolina).
The mean age of the 323 residents was 85.3 years; 85.5% were female; and 89.5% were white (Table 1). The mean age of HCPs was 59.9 years; 63.8% were female; and 67.5% were children of the resident (spouses, 10.2%; nieces or nephews, 8.7%; siblings, 5.0%; and grandchildren, legal guardians or friends, 8.6%). A total of 177 residents (54.8%) died during the study.
Table 2 presents use and expenditures for Medicare-covered services utilized by the entire cohort of residents (N = 323) over the course of the study. Similar data are presented separately for the decedent subgroup (n = 177). For the entire cohort, the total mean expenditures were $8522 per resident and $2303 per 90 days. Mean expenditures for hospitalizations were $2570 per resident and $625 per 90 days, accounting for 30.2% of total Medicare expenditures. Visits to the ED contributed only 1.1% to total expenditures, the lowest proportion of all services. Among the 58 hospitalizations eligible for follow-up, 31 (53.4%) resulted in an SNF stay, accounting for 11.3% of total expenditures (mean, $963 per resident and $234 per 90 days). There were 4826 recorded primary care visits, which accounted for 11.9% of total Medicare expenditures (mean, $1014 per resident and $247 per 90 days). Although only 22.0% of the cohort received hospice services, hospice payments accounted for 45.6% of total Medicare expenditures, the largest proportion of services examined. Mean hospice length of stay was 122.0 days, and mean spending was $3885 per resident and $1050 per 90 days.
Medicare services provided to the subgroup of residents who died accounted for 45.2% of total expenditures. Mean expenditures per decedent were lower compared with all residents ($7026), but expenditures were higher on a per 90-day basis ($2797). The distribution of expenditures across services was comparable between the decedent group and the entire cohort.
Figure 1 displays the distribution of total Medicare expenditures per 90-day interval for the entire cohort based on 1394 assessments, each representing 90 days of follow-up. Total expenditures were less than $500 for 77.1% of these 90-day intervals, between $500 and $12 000 for 17.4%, and exceeded $12 000 for only 5.5% of the 90-day intervals. A similarly skewed distribution of Medicare expenditures was seen among decedents.
Distribution of 90-day Medicare expenditures among nursing home residents with advanced dementia (N = 1394, 90-day assessments). Expenditures were less than $500 for 77.1% of the 90-day assessment periods and exceeded $12 000 for 5.5% of assessment periods.
Figure 2 presents expenditures for each service incurred by decedents for each 90-day interval during the last year of life. Total mean Medicare expenditures increased as death approached as follows: 0 to 90 days before death, $3877 (n = 177); 91 to 120 days before death, $2297 (n = 128); 121 to 270 days before death, $1605 (n = 96); and 365 to 271 days before death, $1061 (n = 68). Mean total expenditures increased on average by 65% each period (P < .001). Mean expenditures for each individual service category also increased, with SNF and acute care showing the greatest proportional increases as death approached.
Mean Medicare expenditures for nursing home residents with advanced dementia in the last year of life during the following 90-day intervals: 0 to 90 days before death (n = 177); 91 to 120 days before death (n = 128); 121 to 270 days before death (n = 96); and 365 to 271 days before death (n = 68).
Table 3 presents the multivariate analyses examining factors associated with 90-day expenditures in the total cohort. After clustering at the individual level was adjusted for, the 6 factors independently associated with higher total Medicare expenditures were younger age, not living in a special care dementia unit, a TSI score of 0, chronic obstructive lung disease, acute illness in the prior 90 days, and a belief by the HCP that the resident has less than 6 months to live.
When hospice was excluded from total Medicare expenditures, factors independently associated with higher 90-day expenditures were younger age, not living in a special care dementia unit, a TSI score of 0, chronic obstructive lung disease, the presence of a percutaneous endoscopic gastrostomy tube, recent acute illness, and lack of a DNH order. Active cancer at baseline was associated with lower nonhospice Medicare expenditures.
This study provides a detailed report describing Medicare expenditures among NH residents with advanced dementia. Throughout the residents' clinical course, hospitalizations and hospice care accounted for the greatest proportion of expenditures. Expenditures were highly skewed: spending was less than $500 in 77.1% of 90-day assessments and more than $12 000 in 5.5% of assessments. In the last year of life, Medicare expenditures increased as residents approached death largely because of increasing use of acute care and hospice services. Not living in a special care unit was associated with higher total 90-day Medicare expenditures, regardless of whether or not hospice was included as part of these expenditures. The lack of a DNH order and the presence of a feeding tube were additional modifiable factors associated with higher 90-day Medicare expenditures when hospice was excluded from these expenditures.
Inconsistent methods hinder direct comparisons between the total expenditures that we observed in advanced dementia and those reported for other terminal illnesses. Nonetheless, limited available data suggest that our estimates of Medicare expenditures in this cohort of NH residents with end-stage dementia were relatively low compared with other Medicare patients at the end of life.8,19,20 This finding is likely explained, in part, by the fact that in the months before death most residents received only NH care, a costly non-Medicare service that is most often paid for by Medicaid; relatively few residents were hospitalized, in an SNF, or enrolled in hospice. In contrast, most Americans who are older than 65 years and dying of cancer or other terminal conditions are not cared for in NHs3 and therefore rely on Medicare services (eg, hospice or hospital) for their end-of-life care.
The increase in Medicare spending that we observed as patients with advanced dementia approached death is consistent with prior research.23 Also, we provide novel findings regarding the sources of those expenditures in end-stage dementia. Roughly one-third of all Medicare expenditures were for hospitalizations. Hospital transfers are potentially burdensome in advanced dementia. As we previously reported,7 most hospitalizations in this cohort were for conditions that were potentially treatable with the same efficacy and at reduced costs in the NH compared with the hospital setting (eg, pneumonia, 68%).24 Moreover, approximately 10% of Medicare expenditures were for SNF care after hospitalization; more than half of residents with a qualifying hospital stay transitioned to SNF status post discharge. Given that these residents were totally functionally and cognitively impaired, their ability to benefit from skilled nursing or intense rehabilitative therapies is questionable. Taken together, our observations support the notion that acute and subacute care in this population may be shaped not only by clinical need alone but also by financial incentives created by Medicare and Medicaid policies. For example, an SNF stay represents substantial financial benefit to NHs because daily Medicare SNF reimbursement is higher than Medicaid NH payments.
Hospice payments accounted for the largest proportion of all Medicare expenditures in our cohort. Hospice has been shown to benefit residents dying with dementia,25,26 although patients with dementia are relatively underserved by hospice.27 While only 22% of residents in our cohort received hospice care, hospice services accounted for close to half of the Medicare expenditures. Whether hospice services lower end-of-life expenditures remains unclear and appears to depend on the terminal diagnosis and the length of hospice stay.6,18 Total Medicare and Medicaid costs in the last month of life for NH residents with dementia are reported to be the same or slightly higher with hospice than without hospice.6
The rich CASCADE data set provided a unique opportunity to identify factors associated with higher Medicare expenditures in advanced dementia. The strong association between the lack of a DNH order and higher acute care expenditures supports the notion that advance care planning may be a key step toward preventing aggressive end-of-life care, while reducing costs.8 Tube feeding, a potentially burdensome intervention with no demonstrable benefits in advanced dementia,28 was also independently associated with higher nonhospice expenditures.
This study has several limitations. First, Medicare expenditures were not derived from claims data but were estimated from publicly available fee schedules. This commonly used approach likely underestimated Medicare expenditures in our cohort.5,8 Second, there may be inaccuracies in the utilization data obtained from chart reviews because of errors in both documentation and abstraction. Third, we did not examine Medicaid expenditures, which are expected to be less variable than Medicare expenditures.6 Fourth, we described expenditures only during a snapshot of the course of advanced dementia and the period leading up to death. However, it would be challenging to conduct a prospective study describing expenditures from the moment patients first meet the criteria for advanced dementia. Fifth, we are not able to make causal inferences between factors shown to be associated with expenditures. Finally, the generalizability of our findings outside the greater Boston area is uncertain in terms of both Medicare expenditures and clinical factors (eg, the CASCADE cohort had a relatively high use of DNH orders). Nonetheless, Medicare expenditures were based on national data, and our analyses focused on the association between clinical factors and expenditures. It is also notable that expenditures varied considerably among residents, even within the narrow region of this study.
Dementia is a terminal illness, yet prior work suggests that persons dying with this disease receive suboptimal end-of-life care.27 Medicare policies play a key role in shaping that care. This study demonstrates that a large proportion of Medicare care expenditures in advanced dementia are attributable to acute and subacute services that may be avoidable and may not improve clinical outcomes. Strategies that promote palliation in advanced dementia may shift expenditures away from these aggressive treatments in advanced dementia toward a more comfort care approach (eg, hospice); however, the net effect on total Medicare expenditures remains unresolved. Finally, a better understanding of fiscal incentives that may be driving the pattern of health care expenditures among NH residents with advanced dementia (eg, temporary cost shifting from Medicaid to Medicare) is needed.
Correspondence: Susan L. Mitchell, MD, MPH, Hebrew SeniorLife Institute for Aging Research, 1200 Centre St, Boston, MA 02131-1097 (firstname.lastname@example.org).
Accepted for Publication: October 3, 2010.
Published Online: January 10, 2011. doi:10.1001/archinternmed.2010.478
Author Contributions: Mr Goldfeld and Dr Mitchell had full access to all 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: Goldfeld, Stevenson, Hamel, and Mitchell. Acquisition of data: Goldfeld and Mitchell. Analysis and interpretation of data: Goldfeld, Stevenson, Hamel, and Mitchell. Drafting of the manuscript: Goldfeld, Stevenson, Hamel, and Mitchell. Critical revision of the manuscript for important intellectual content: Goldfeld, Stevenson, Hamel, and Mitchell. Statistical analysis: Goldfeld, Hamel, and Mitchell. Obtained funding: Mitchell. Administrative, technical, and material support: Stevenson and Mitchell. Study supervision: Mitchell.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by grants R01AG024091 and K24AG033640 (Dr Mitchell) from the National Institute on Aging.
Role of the Sponsors: The funding sources had no role in the design and conduct of the study; in the collection, management, analysis, and interpretation of the data; or in the preparation of the manuscript.
Thank you for submitting a comment on this article. It will be reviewed by JAMA Internal Medicine editors. You will be notified when your comment has been published. Comments should not exceed 500 words of text and 10 references.
Do not submit personal medical questions or information that could identify a specific patient, questions about a particular case, or general inquiries to an author. Only content that has not been published, posted, or submitted elsewhere should be submitted. By submitting this Comment, you and any coauthors transfer copyright to the journal if your Comment is posted.
* = Required Field
Disclosure of Any Conflicts of Interest*
Indicate all relevant conflicts of interest of each author below, including all relevant financial interests, activities, and relationships within the past 3 years including, but not limited to, employment, affiliation, grants or funding, consultancies, honoraria or payment, speakers’ bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued. If all authors have none, check "No potential conflicts or relevant financial interests" in the box below. Please also indicate any funding received in support of this work. The information will be posted with your response.
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 11
Customize your page view by dragging & repositioning the boxes below.
Users' Guides to the Medical Literature
Users' Guides to the Medical Literature
All results at
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.