The Outcomes and Costs of Acute Myeloid Leukemia Among the Elderly

Background  The incidence of acute myeloid leukemia (AML) among the elderly can be expected to grow as the population continues to age. However, data on current treatment practices and costs for this form of cancer are sparse.

Methods  We used a retrospective inception cohort design and data from a linkage between 11 Surveillance, Epidemiology, and End Results cancer registries and Medicare administrative claims. We evaluated survival, use of health care resources, use of chemotherapy, and Medicare payments among adults 65 years and older with an initial diagnosis of AML between January 1, 1991, and December 31, 1996.

Results  A total of 2657 elderly patients with AML and complete Medicare claims data were identified. The prognosis for these patients was poor, with median survival estimated to be 2 months and a 2-year survival rate of 6%. Mean ± SE total Medicare payments were $41 594 ± $870 (in 1998 US dollars), 84% of which was attributed to inpatient payments. In the 2 years after the AML diagnosis, 790 patients (30%) underwent chemotherapy treatment. These patients had costs almost 3 times higher than those of other patients, and their median survival was 6 months longer. The use of hospice care was rare (17% of patients).

Conclusions  Among the elderly, AML is associated with a poor prognosis and substantial costs during the relatively few remaining months of life. Moreover, most patients do not receive active treatment with chemotherapy or hospice services. Further work is needed to characterize this disease and the patient-related factors that influence treatment decisions and associated health outcomes.

ACUTE MYELOID leukemia (AML) is the most common type of acute leukemia among US adults, with about 10 000 new cases reported annually.¹ The incidence of AML increases with age and is almost 10 times greater among persons 65 years and older than among those younger than 65 years (12.2 vs 1.3 per 100 000).² If left untreated, AML usually results in death within a few months of diagnosis.^1,2 Conventional cytotoxic chemotherapy for AML can be associated with serious adverse effects and, as a result, often cannot be tolerated by older patients.² Consequently, 5-year survival is only 2% among elderly patients and has not improved appreciably in the past 2 decades.^3,4

Limited data are available on treatment practices or the associated outcomes and costs after a diagnosis of AML in older patients. Most previous studies characterize prognosis after the initial AML diagnosis in terms of cytogenetics, hematologic factors, performance status, organ function, multidrug resistance gene expression, and age, but these studies are based on data from relatively small numbers of patients from clinical trials or individual study centers whose experience may not be applicable to patients with AML treated in typical practice settings.^5,6

This study therefore used linked data from the Surveillance, Epidemiology, and End Results (SEER) registries and Medicare claims to address the following questions in a national-level cohort of elderly patients: (1) How do survival and use of health care services vary by age within the elderly population with AML? (2) What is the prevalence of the use of chemotherapy among these patients? and (3) What factors influence the receipt of chemotherapy and associated outcomes of AML treatment? The answers to these questions can provide an important baseline against which the benefits of emerging therapies can be compared.

This study used clinical data from the SEER registries linked to Medicare claims data. The SEER-Medicare database is a collaborative effort of the National Cancer Institute (Bethesda, Md), the SEER registries, and the Centers for Medicare and Medicaid Services (formerly, the Health Care Financing Administration) (Baltimore, Md). The SEER registries comprise an epidemiological surveillance system consisting of 11 population-based tumor registries designed to track cancer incidence and survival in the United States. The registries routinely collect information on patients with newly diagnosed cancer in geographically defined areas that represent approximately 14% of the US population.

The SEER-Medicare linkage includes data for all persons with cancer reported by the SEER registries between January 1, 1991, and December 31, 1996, who could be identified as Medicare enrollees. The SEER data items include demographics (eg, age, sex, and race), and diagnostic information for up to 10 different incident cancer cases for each person, including month and year of cancer diagnosis, cancer site and type (eg, AML), and vital status. The Medicare details include monthly enrollment data (ie, Part A or B), a monthly indicator for health maintenance organization (HMO) enrollment, and hospital, outpatient, physician, home health care, hospice, and skilled nursing facility (SNF) claims containing dates of service, a primary and up to 9 secondary diagnoses and procedure codes from the International Classification of Diseases, Ninth Revision, Clinical Modification, a diagnosis-related group code (in the inpatient file), a health care procedure code (in the outpatient, home health care, and hospice claims), and Medicare payments from 1991 through 1998 for all SEER patients eligible for Medicare sometime during this period.

With these data, we created a patient-level analytic file containing the date of AML diagnosis (assumed to be the first day of the month of diagnosis because only month and year of diagnosis were recorded in the SEER data, which may slightly overestimate survival), date of death (if applicable), and a summary of the use of Medicare services by category of service for each health care encounter during the first 2 years after the diagnosis of AML. A 2-year time frame was chosen for the analysis because it was the maximum amount of time for which all patients could be followed up, and it would be expected to capture the remaining lifetime costs of most patients with AML, given the high rate of short-term mortality associated with this disease.

All patients 65 years and older with a diagnosis of AML in a SEER registry between January 1, 1991, and December 31, 1996, were candidates for inclusion in the analysis. We defined AML as coded in the SEER data file (ie, cancer site recode 77). We excluded patients whose month of diagnosis was unknown or who were younger than 65 years at diagnosis. In addition, persons who were enrolled in HMOs or who were not entitled to Medicare Parts A and B benefits during the study period were excluded from the analyses because complete claims were not available for them.

Survival was evaluated in terms of the percentage of patients alive each month after the AML diagnosis and the median number of months after the AML diagnosis that patients survived. Use of health care resources for all causes was evaluated in terms of the percentages of patients who were hospitalized and who received SNF care, hospice care, and home health care and the total numbers of days in the hospital and in SNFs in the 2 years after the diagnosis. In addition, for patients who died during follow-up, the percentage of remaining days spent in a hospital or an SNF was calculated. Medicare payments were evaluated overall (calculated as the sum of payments for hospital, SNF, home health care, hospice, outpatient, and physician services) and by individual category of care. All payments were converted to 1998 US dollars using the medical care component of the Consumer Price Index.

The receipt of chemotherapy was evaluated as the presence of an inpatient claim in the 2 years after the AML diagnosis indicating chemotherapy, as follows: diagnosis code v58.1 (chemotherapy) or procedure code 99.25 (injection or infusion of chemotherapy) from the International Classification of Diseases, Ninth Revision, Clinical Modification or diagnosis-related group codes 410 (chemotherapy without acute leukemia as a secondary diagnosis) or 492 (chemotherapy with acute leukemia as a secondary diagnosis). The proportion of patients who received chemotherapy was compared across patient groups stratified on the basis of age, sex, and comorbidities in the year before the AML diagnosis. Retreatment (for relapse of AML) during the study period was defined as the presence of a claim for chemotherapy after at least 120 days with no chemotherapy claims. This method of identifying patients treated for relapse of AML has been validated in an independent sample of patients with AML.⁷

Study patients were described in terms of baseline characteristics, including age at diagnosis, sex, race, year of AML diagnosis, comorbidities in the previous year, and residence in a metropolitan area, to identify characteristics that might influence study outcomes. Survival, receipt of chemotherapy, use of health care resources, and costs were analyzed descriptively overall and stratified by age at diagnosis, sex, race, presence of comorbidities, and year of diagnosis. In addition to overall and stratified descriptive analyses, multivariate analyses were undertaken to estimate the impact on study outcomes of potentially confounding influences, including age at diagnosis, sex, race, year of diagnosis, and residence in a metropolitan county. Models including comorbidities in the year before diagnosis also were estimated (for the subset of patients whose condition was diagnosed between January 1, 1992, and December 31, 1996, only, since those whose condition was diagnosed in 1991 did not have a 1-year history available), but are not reported herein, since findings were similar. Comorbidity was assessed using the Charlson Comorbidity Index as adapted by Deyo et al.⁸

Least squares regression analyses were performed for inpatient days and the logarithm of Medicare payments (a semilogarithmic model was estimated because of nonnormal payments), and a Cox proportional hazards model was estimated for the number of months of survival, with censoring of those still alive 24 months after diagnosis. In addition, a logistic model was estimated, including the same covariates listed above to identify the characteristics that were potentially associated with receipt of chemotherapy. Finally, characteristics and outcomes among the subset of patients who received chemotherapy were evaluated, and study measures were assessed by length of survival (ie, among patients who survived longer vs shorter than the median survival). In all multivariate analyses, P values of less than .05 were deemed statistically significant. We used SAS, Version 8.0 (SAS Institute Inc, Cary, NC), in all analyses.

From an initial sample of 3738 Medicare beneficiaries with a diagnosis of AML between 1991 and 1996, 1081 patients (29%) were excluded. The reasons included unknown month of diagnosis (n = 29), AML diagnosis at younger than 65 years (n = 509), and enrollment in an HMO (n = 495), nonentitlement to Medicare benefits (n = 44), or both (n = 4). This left a total of 2657 Medicare beneficiaries 65 years and older with an initial diagnosis of AML in a SEER registry between 1991 and 1996 for inclusion in the analyses. The average age at diagnosis was 77 years, and 51% of the patients were male (Table 1). Most patients were white (87%), and most resided in metropolitan counties (84%). Based on information available for the year before the AML diagnosis, congestive heart failure (7%) and chronic obstructive pulmonary disease (7%) were the most common preexisting comorbidities.

In the first year after the diagnosis, 2273 (86%) of 2657 elderly patients with AML died, and 94% died within 2 years (Figure 1). Median survival was 2 months overall, ranging from 1 month among patients 85 years and older to 3 months among those aged 65 to 74 years. Differences in survival based on age were statistically significant on the basis of results of the log-rank test (P<.001). Survival did not differ by sex, race, year of diagnosis, or residence in a metropolitan county in the stratified analyses. Findings were similar based on a multivariate Cox proportional hazards model, which indicated that the risk for death was 50% greater among patients aged 75 to 84 years (P<.001) and 150% greater among patients 85 years and older (P<.001), relative to those aged 65 to 74 years.

In the first 2 years after the AML diagnosis, 89% of study patients were hospitalized (Table 2). Patients had a mean of 2.3 hospitalizations, for a mean total of 28.1 hospital days, 77% of which occurred in the first 2 months after the diagnosis. Twelve percent of patients received care in SNFs for a mean of 2.4 days. Thirty-nine percent of patients received home health care, and 17% received hospice care. Among the patients who died during follow-up (94% of the sample), an average of 31% of their remaining days were spent in inpatient facilities. Total Medicare payments (mean ± SE) among study patients in the 2 years after the AML diagnosis were $41 590 ± $870 (Table 3), with 84% attributable to hospitalization costs. Physician costs constituted 7% of total costs, whereas outpatient, SNF, home health care, and hospice costs constituted 4%, 2%, 2%, and 1%, respectively. Median payments were $25 705 (interquartile range, $14 970-$54 386).

In stratified analyses, only small differences in use and costs of health care resources were observed by sex, race, previous comorbidities, and year of AML diagnosis. However, sizable differences were detected by age and residence in a metropolitan county. On average, patients 85 years and older used fewer hospital days (10.8 days) than those aged 75 to 84 years (23.5 days) and those aged 65 to 74 years (39.3 days) (Table 2), consistent with the shorter survival times among the oldest patients. The oldest patients also were somewhat less likely to use home health care services. Mean ± SE total costs for all services were lowest among the oldest patients ($21 771 ± $1180) compared with patients aged 75 to 84 years ($35 978 ± $1050) and 65 to 74 years ($54 720 ± $1625) (Table 3). A similar pattern was observed in terms of median total costs ($19 265 among patients ≥85 years, $24 829 among patients aged 75-84 years, and $36 962 among patients aged 65-74 years). The distribution of costs by category of service was similar across the age groups, with most costs attributable to hospitalization. Finally, compared with patients living in nonmetropolitan areas, those living in metropolitan counties had more hospital days (mean, 29 vs 23 days; median, 15 vs 11 days) and higher costs (mean, $43 580 vs $31 540; median, $26 680 vs $19 760).

Multivariate models of inpatient days and total Medicare payments confirmed the stratified results, indicating more than a $20 000 mean total cost difference between the youngest and oldest age groups and close to a $10 000 difference between metropolitan and nonmetropolitan patients (P<.01 for both).

In the 2 years after the diagnosis of AML, 790 study patients (30%) received chemotherapy. Treated patients tended to be younger than untreated patients (average age, 73 vs 78 years), slightly more likely to live in metropolitan counties (88% vs 82%), and more likely to have been diagnosed as having congestive heart failure (8% vs 3%) or chronic obstructive pulmonary disease (8% vs 4%). Sex and race did not significantly influence the likelihood of receiving chemotherapy treatment. In multivariate analyses (logistic model) of chemotherapy use, findings (given as odds ratios [ORs] and 95% confidence interval [CIs]) were consistent, indicating that age (ORs, 0.39 [95% CI, 0.32-0.48] for 75-84 years of age and 0.08 [95% CI, 0.05-0.12] for ≥85 years vs 65-74 years of age), the presence of congestive heart failure (OR, 0.45 [95% CI, 0.27-0.76]), and the presence of chronic obstructive pulmonary disease (OR, 0.53 [95% CI, 0.34-0.85]) were negatively associated with chemotherapy use, whereas residence in a metropolitan county was positively associated with treatment (OR, 1.83 [95% CI, 1.39-2.44]).

On an overall basis, median survival was 6 months longer among treated patients (7 vs 1 month). Longer survival was observed among treated patients regardless of age at diagnosis (ie, differences in median survival of 6, 3, and 1 month among treated vs untreated patients aged 65-74, 75-84, and ≥85 years, respectively). Consistent with these longer survival times, patients who received chemotherapy used more health care resources than those who did not receive chemotherapy and had higher total costs (mean, $77 769 vs $26 287) (Table 4). This difference in total costs based on receipt of chemotherapy was consistent across age groups, with treated patients having total costs 2 to 3 times higher than those of untreated patients. Patients undergoing retreatment for relapsed AML (n = 124) had the highest level of resource use, with an average of 105 total inpatient days and mean total costs of $128 630 during the 2-year follow-up.

Patterns of resource use and Medicare costs differed substantially between patients who survived shorter vs longer than the median 2-month length of survival after the diagnosis of AML. Patients who survived more than 2 months were more likely to use home health care and hospice services (55% and 20%, respectively) than those who survived less than 2 months (20% and 14%, respectively). Those who died rapidly spent an average of 37% of their remaining days in institutions, whereas those who lived longer than the median spent an average of 24% of their remaining days in institutions. Finally, patients who survived more than 2 months were 3 times more likely to be treated with chemotherapy than those who survived less (43% vs 15%, respectively), but had mean costs that were 3 times higher than those who survived less than 2 months ($60 330 vs $19 950, respectively).

We found the prognosis among elderly patients with AML to be extremely poor, with a median survival of 2 months and a 2-year survival rate of 6%. The prognosis has not changed during the past decade, with similar survival regardless of the year of diagnosis. Our findings are consistent with other studies that have documented a poor prognosis among elderly patients with AML.^2,4,9 To our knowledge, however, our investigation is the first comprehensive evaluation of the outcomes and costs associated with AML.

Costs for the use of all health care resources after diagnosis were substantial, exceeding $41 000 per patient on average, most of which was incurred in the first few months. Costs were dominated by hospitalization, as has been found in studies of patients with solid tumors.^10,11 Costs varied only slightly by race, sex, comorbidities, and year of diagnosis, but varied greatly by age at diagnosis, with the oldest patients having the lowest overall costs. The lower overall costs among the oldest patients were due to their shorter survival times, because the average cost per day alive was similar for all 3 age groups (ranging from about $340-$370 per day) and did not vary by receipt of chemotherapy. It should be noted that this analysis focused on the use of all health care resources and not only those specifically attributable to AML. Given the methodological issues involved in attributing costs to AML, and given that more than 80% of hospitalizations had primary diagnoses directly related to AML or complications of its treatment, we believed that this approach was appropriate.

The infrequent use of chemotherapy in our sample may reflect the fact that elderly patients cannot tolerate standard cytotoxic treatments for AML, as noted in the literature.^2,12 This finding is consistent with data from the United Kingdom Medical Research Council study, which reported that as many as 28% of all study patients and 38% of study patients older than 60 years received only palliative treatment.¹³ Furthermore, similar to our study in which patients who received chemotherapy lived longer, patients who received only palliative treatment in the UK Medical Research Council study were found to have a 1-year survival of 11% (50/475 patients) relative to 32% (391/1221 patients) among those who received chemotherapy. In our study, age, residence in a metropolitan area, and the presence of congestive heart failure or chronic obstructive pulmonary disease in the year before the AML diagnosis were the only factors that appeared to influence chemotherapy use, perhaps reflecting the better prognosis among the youngest and healthiest patients and more access to treatment in metropolitan areas. It is encouraging to note that emerging treatment strategies for AML are expected to be more effective for and better tolerated by the elderly.^14-23 The results from this study may provide a baseline for assessing the potential benefits of these interventions, especially if they offer less unfavorable risk-benefit choices between tolerability and survival.

Although Medicare hospice program costs have more than doubled in the past decade,²⁴ there was very little use of hospice care in our sample, even among those who survived the longest. Instead, the patients who died spent an average of 31% of their remaining days in hospitals or nursing facilities rather than at home, which may have had a negative impact on the health-related quality of life among these patients.²⁵ Further research that assesses potential barriers to use of hospice services would be beneficial.

Our study is subject to several limitations. We relied on administrative claims data, which were not collected specifically for research purposes.²⁶ Consequently, we were unable to address other cancer outcomes, such as remission rates, disease-free survival, and quality of life. In addition, patients from the SEER registries may not be representative of all US patients with AML. Although the SEER data cover about 14% of all cancer cases, certain groups are underrepresented or overrepresented (eg, blacks and other races, respectively).^27,28 Nonetheless, the linked SEER-Medicare database has proven extremely useful for case selection and for profiling oncology treatment patterns and survival for many different cancer types, including cancers of the breast, prostate, lung, and ovaries.^29-34 Costs were evaluated from a limited perspective (ie, patients eligible for Medicare and not enrolled in HMOs). Because Medicare is a primary payer for this age group, we expect that the bulk of costs were captured in our analysis. The fraction of the sample excluded because of HMO enrollment (13%) is consistent with HMO penetration in the United States (14.5%).³⁵ Since the excluded HMO patients were similar to our study population in terms of demographics and survival, their exclusion should not have greatly influenced our results.

This study indicates that AML among the elderly is associated with a poor prognosis and with substantial costs during the relatively few remaining months of life. Moreover, most patients do not receive active treatment with chemotherapy or hospice services. Further work is needed to characterize this disease and the complex comorbidity- and patient preference–related factors that influence treatment decisions and associated health outcomes.

Accepted for publication November 20, 2001.

This study was funded in part by Wyeth-Ayerst Research, Radnor, Pa.

We thank Rick deFriesse, MA, for his expert programming assistance, and the National Cancer Institute and the Centers for Medicare and Medicaid Services for allowing us to use the linked SEER-Medicare data.

Wyeth-Ayerst Research has no direct financial interest in the findings reported herein.

Corresponding author and reprints: Joseph Menzin, PhD, Boston Health Economics, Inc, 20 Fox Rd, Waltham, MA 02451 (e-mail: jmenzin@bhei.com).