Healthcare resource utilization and costs in patients with newly diagnosed acute myeloid leukemia

Abstract

Aim: Acute myeloid leukemia (AML) is associated with high disease burden. This analysis estimated HRU and costs among newly diagnosed AML patients in a US commercially insured population.

Materials and methods: This was a retrospective observational study using the IMS Health PharMetrics Plus and Hospital Charge Detail Master databases. Patients included adults who were newly diagnosed with AML between January 2007 and June 2016 (“study period”). Patients with <12 months of continuous enrollment prior to the index date were excluded, as were those whose first diagnosis was AML in remission/relapse, those diagnosed with acute promyelocytic leukemia, those on Medicare supplemental insurance, or those with a diagnosis of AML in remission/relapse without evidence of treatment during the study period. Patients were stratified by receipt of AML treatment (chemotherapy/hematopoietic cell transplantation [HCT]), and their follow-up was partitioned into initial, remission, and relapsed health states. Mean HRU and costs were tallied by treatment and, for treated patients, by health state and time since entry into health state (≤6 vs >6 months).

Results: A total of 9,455 patients met study criteria, including 6,415 (68%) treated and 3,040 (32%) untreated patients, with mean follow-up of 18.3 and 16.4 months, respectively. Mean age was 55 years in treated patients and 60 years in untreated patients. Mean total costs per patient were $386,077 in treated patients and $79,382 in untreated patients. For treated patients, 60% of total costs ($231,867 per patient) were incurred during the initial health state, representing time without remission/relapse. Mean monthly total healthcare costs were $21,055 and $4,854 among treated and untreated patients, respectively.

Limitations and conclusions: HRU and costs of managing AML patients are substantial. In treated patients, the majority of costs were incurred during the initial treatment period, without claims indicating remission/relapse.

Keywords:

• AML
• healthcare resource utilization
• costs
• newly diagnosed

JEL classification codes:

• I00
• I13

Introduction

Acute myeloid leukemia (AML) is an aggressive cancer in which the bone marrow makes abnormal myeloblasts (a type of white blood cell) that accumulate in the bone marrow and interfere with the production of normal white blood cells, red blood cells, and platelets, resulting in infection, anemia, and bleeding. The abnormal leukemic cells may spread outside the blood to other parts of the body, including the central nervous system (brain and spinal cord), skin, and gums¹. Untreated, the disease is progressive and rapidly fatal.

AML is the most common acute leukemia in adults, accounting for ∼25% of all adult leukemias worldwide, with the highest incidence rates occurring in the US, Europe, and Australia². It is estimated there were 21,380 new cases of AML and 10,590 deaths due to the disease in the US in 2017. Despite some improvements in treatment, 5-year relative survival in patients with AML in the US is only 27%³.

The standard of care for front-line treatment of AML includes remission induction and post-remission consolidation chemotherapy. Conventional induction therapy consists of cytarabine as a continuous infusion for 7 days with an anthracycline (e.g. daunorubicin or idarubicin) given as an intravenous push for 3 days (“7 + 3” regimen)¹. Because of the need for continuous infusion and patient monitoring, intensive induction chemotherapy is typically provided in an inpatient setting. Intensive therapy is generally limited to patients <60 years of age and older patients with few comorbidities⁴. Options for consolidation therapy include additional chemotherapy and/or allogeneic hematopoietic cell transplantation (HCT), depending upon factors such as resistance to induction therapy and genetic testing¹.

While several studies have examined healthcare resource utilization and costs among patients with newly diagnosed AML^5–12, these studies focused on either elderly patients^5–7,10,11 or treated AML patients^8–12. Additionally, data on utilization and costs for patients with treatment-related AML (tAML) or AML with myelodysplasia-related change (MRC), important sub-groups for whom novel treatments have recently been approved¹³, are unavailable. The objective of this study was to examine healthcare resource utilization and costs in a broadly defined, commercially insured population of AML patients in the US.

Methods

Study design

This was a retrospective cohort study of healthcare resource utilization and costs in patients with newly diagnosed AML in the US using the IMS Health PharMetrics Plus health insurance claims database and the IMS Health Hospital Charge Detail Master (CDM) database over the period from January 2007 through June 2016 (“study period”). The design of the study is depicted in Figure 1.

Figure 1. Study design. Abbreviations. AML, Acute myeloid leukemia; Dx, Diagnosis.

Data sources

A custom sub-set of data from the IMS Health PharMetrics Plus database, along with linked information from the IMS Health Hospital CDM data, were used in this study. The custom sub-set included all available data from both databases for all patients who had ≥1 claims with a diagnosis of AML in the PharMetrics Plus database during the study period.

IMS Health PharMetrics Plus database

The IMS Health PharMetrics Plus database is one of the largest US health plan claims databases, comprising adjudicated claims for ∼150 million unique patients across the US. Data are available from 2006 to the present, with a typical 3–4-month service date lag due to claims adjudication.

The data are longitudinal, with ∼20 million patients who have both medical and pharmacy coverage with ≥3 years of continuous enrollment. The PharMetrics Plus database includes representation from diverse geographic regions, employers, payers, providers, and therapeutic areas. Patients in each 3-digit zip code and every Metropolitan Statistical Area of the US are included, with data from 90% of US hospitals and 80% of US doctors¹⁴.

Data include inpatient and outpatient diagnoses and procedures, and retail and mail order prescription records, in addition to detailed information on the pharmacy and medical benefit, inpatient stay (admission type and source, discharge status), provider details (specialty, zip code, attending, referring, rendering, prescribing, primary care provider), amounts allowed and paid by health plans, and dates of service for all claims.

Information available for each person in the database also includes demographic variables (age, gender, and geographic region), product type (e.g. HMO, PPO), payer type (e.g. commercial or Medicare Advantage plan), and start and stop dates of health-plan enrollment. The database is fully de-identified and compliant with the Health Insurance Portability and Accountability Act of 1996 (HIPAA).

IMS Hospital CDM

Through a unique and proprietary algorithm, data from the PharMetrics Plus database can be linked to the IMS Hospital CDM database. The CDM database provides the most detailed level of hospital billing information for patient-level drug utilization by brand, medical devices, procedures, and diagnoses in both the inpatient and outpatient care settings from ∼450 hospitals. Hospital data are drawn from operational files and other reference sources. The database consists primarily of general medical/surgical hospitals, but also includes some specialty hospitals, such as children’s and rehabilitation hospitals. Each hospital’s data include all inpatient stays and outpatient encounters/visits. All procedures and therapies are available by the day of stay on which the item was delivered or performed. Diagnoses cover the entire length of stay.

Patient selection

The study sample consisted of all persons in the IMS Health PharMetrics Plus database with evidence of newly diagnosed AML during the study period. Because diagnosis codes for AML are distinguished by specification of remission and relapse, AML patients identified by diagnosis codes for AML (ICD-9-CM: 205.00–205.02; ICD-10-CM: C92.00, C92.01, C92.02, C92.50, C92.51, or C92.52) were considered newly diagnosed if their first AML diagnosis claim was for AML without mention of remission or relapse (ICD-9-CM: 205.00; ICD-10-CM: C92.00, C92.50).

For each newly diagnosed AML patient, the date of the first claim with a diagnosis of AML without remission or relapse was designated the “index date”. Patients were excluded if they had fewer than 12 months of continuous enrollment prior to their index date, had evidence of acute promyelocytic leukemia during the study period, were <18 years of age as of the index date, were enrolled in Medicare supplemental insurance plans^a, had evidence of hospitalizations during follow-up and zero total cost of hospitalization during follow-up, had a diagnosis of AML in remission or relapse without evidence of AML treatment, or had a first AML treatment associated with AML diagnosis in remission or relapse (ICD-9-CM: 205.01 or 205.02; ICD-10-CM: C92.01, C92.02, C92.51, or C92.52).

Baseline characteristics were assessed based on enrollment and claims information during the “pre-index period”, which was defined for each patient as the period beginning with the first date of the study period and ending with the index date.

Outcome measures of interest were examined over the “follow-up period”, defined as beginning with the index date and ending with the end of the study period or the end of post-index continuous enrollment, whichever occurred first (Figure 1).

For patients who had at least one CDM record with AML medications during the initial health state (defined below), healthcare resource utilization and costs were evaluated with respect to the initial inpatient AML treatment regimen received.

Patient characteristics at baseline

Information on age, region, and plan type were obtained from the enrollment file. Diagnosis codes on medical claims during the pre-index period were scanned to ascertain the presence of selected comorbidities¹⁵, including: chronic myelomonocytic leukemia (CMML); myelodysplastic syndromes (MDS); other hematologic malignancies excluding AML, CMML, and MDS; solid tumors (breast cancer, colorectal cancer, head and neck cancer, kidney cancer, liver cancer, lung cancer, ovarian cancer, prostate cancer, and other solid tumors); chronic obstructive pulmonary disease (COPD); chronic liver disease; chronic renal disease; coronary heart disease (CHD); dementia; diabetes; and hypertension. A comorbidity index (e.g. Charlson Comorbidity Index–National Cancer Institute Modification) was calculated for all patients based on comorbidities identified in the 12-month pre-index period¹⁶.

Other characteristics assessed at baseline included evidence of receipt of HCTs, radiotherapy, hypomethylating agents (HMAs), and cytotoxic treatments, and measures of healthcare resource utilization and costs, including the numbers of physicians’ office or outpatient hospital visits, outpatient pharmacy claims (i.e. prescriptions), emergency department (ED) visits, hospitalizations, and inpatient days, as well as outpatient costs, pharmacy costs, inpatient costs, and total costs. Costs were estimated based on reimbursed amounts.

Definition of health states

Patients were considered treated if there was any evidence of receipt of chemotherapy (either inpatient or outpatient) or HCT during follow-up. For each treated patient, the follow-up period was partitioned into health states based on diagnosis codes for AML as follows:

• Initial State: The period beginning with index date and ending with the earliest of the date of first diagnosis code for AML in remission (ICD-9-CM: 250.01; ICD-10-CM: C92.01, C92.51), AML in relapse (ICD-9-CM: 250.02; ICD-10-CM: C92.02, C92.52), or the end of follow-up;

• Remission State: The period beginning with the date of first diagnosis code for AML in remission and ending with the earliest of the date of first diagnosis code for AML in relapse or the end of follow-up; and

• Relapse State: The period beginning with the date of first diagnosis code for AML in relapse and continuing to the end of follow-up.

For patients with a first diagnosis code for AML in relapse prior to first diagnosis code for AML in remission, all time after the first diagnosis code for AML in relapse was considered to be in the relapse state. The time in each health state was partitioned further by time since entry into the state (≤6 vs >6 months).

Outcome measures

Healthcare resource utilization

For each patient in the study, a variety of measures of healthcare resource utilization were assessed during the follow-up period, including the number of physicians’ office or outpatient hospital visits, outpatient pharmacy claims (i.e. prescriptions), emergency department visits, hospitalizations, and inpatient days. Receipt of HCTs, red blood cell and platelet transfusions, and granulocyte colony-stimulating factors (GCSFs) was also identified. Evidence of anemia, bleeding, lymphopenia, thrombocytopenia, fever, treated infection, febrile neutropenia, or other neutropenia was also assessed based on diagnosis and procedure codes on claims (results of lab tests are not available in the IMS Health PharMetrics database). Treated infection was defined by the presence of diagnosis codes for infection and receipt of intravenous antimicrobials on the same day. Since ICD-9-CM or ICD-10-CM diagnosis codes for febrile neutropenia are unavailable, follow-up inpatient admissions and outpatient non-lab claims with diagnosis codes of neutropenia were evaluated, and the following cases were considered as febrile neutropenia: (1) inpatient admission with admission diagnosis of neutropenia; (2) inpatient admission with any diagnosis of neutropenia and any diagnosis of fever; (3) inpatient admission with any diagnosis of neutropenia and any diagnosis of infection; (4) non-lab outpatient claim with a diagnosis of neutropenia and outpatient claim with a diagnosis of fever on the same day; (5) non-lab outpatient claim with a diagnosis of neutropenia and outpatient claim with a diagnosis of infection on the same day; or (6) non-lab outpatient claim with a diagnosis of neutropenia and outpatient claim with receipts of GCSF on the same day.

Healthcare costs

For each patient, total healthcare costs, as well as the following healthcare cost components, were assessed: HCT, outpatient chemotherapy, inpatient chemotherapy, other AML-related outpatient visits, other AML-related inpatient visits, non–AML-related outpatient visits, non–AML-related inpatient visits, and outpatient pharmacy.

Time to disenrollment

Time to disenrollment was defined for all patients as the number of days from the index date to the end of the follow-up period. Because data on mortality were not consistently available in the PharMetrics Plus database and it was not possible to distinguish reliably between disenrollment and death, overall survival was not analyzed.

Statistical analyses

Patient characteristics at baseline

Descriptive statistics (means and standard deviations [SDs], or frequency and percentage) of patient characteristics were reported by receipt of treatment.

Healthcare resource utilization and costs

Descriptive statistics for measures of healthcare resource utilization and costs were reported by receipt of treatment. For treated patients, healthcare resource utilization and costs were examined by health state. Healthcare resource utilization and costs were evaluated using two different approaches. In the first set of analyses, descriptive statistics of healthcare resource utilization and costs were calculated with the denominator population composed of all treated or untreated patients. In the second set of analyses, measures of healthcare resource utilization and costs were reported on a person-month basis.

Time to disenrollment

Time to disenrollment was analyzed using Kaplan-Meier methods. Patients for whom the date of the end of follow-up was equal to the end of the study period were censored. For treated patients, time to disenrollment was partitioned by health state using the proportion of patients in each health state.

Factors predicting receipt of AML treatment

A logistic regression was conducted to determine factors that predict receipt of AML treatment as defined above. Starting with all the baseline characteristics included as independent variables, stepwise selection was used to identify predictive factors using p-values of 0.1 for entering and 0.05 for remaining in the regression model.

Analyses of healthcare resource utilization and costs by initial AML regimen

For AML patients who had one or more CDM records with AML medication during the initial state, the initial AML treatment regimen was identified using all AML medication claims in the IMS Health PharMetrics Plus database and all records in the CDM database. AML medications considered include azacitidine, cladribine, clofarabine, cytarabine, daunorubicin, decitabine, etoposide, fludarabine, gemtuzumab, idarubicin, mitoxantrone, sorafenib, and topotecan.

AML treatment episodes were identified by scanning the databases to identify gaps of 3 days or more in the receipt of these medications. Treatment regimens within an episode were classified into the following four groups: 7 + 3 regimen (cytarabine plus either daunorubicin or idarubicin, or both), cytarabine monotherapy regimen, regimen with HMA (decitabine or azacitidine), and other regimen.

Measures of healthcare resource utilization and costs were then reported by initial treatment regimen and health state using the same approach described above.

Sub-group analysis

A sub-group analysis was conducted for patients with tAML or AML with MRC. Patients with tAML or AML with MRC were identified based on evidence of MDS diagnosis during the pre-index period or receipt of HCT, radiotherapy, or cytotoxic treatment before the index date.

Results

Patient selection

There were 26,565 patients in the sample with the first claim with AML diagnosis on or after January 1, 2008 (Figure 2). Of these patients, 9,455 (36%) qualified for the study, including 6,415 (68%) who received AML treatment and 3,040 (32%) who did not.

Figure 2. Patient selection. Abbreviations. AML, Acute myeloid leukemia; APL, Acute promyelocytic leukemia.

Patient characteristics

Patient characteristics at the index date are reported by receipt of AML treatment in Table 1. Mean age at initiation of AML diagnosis was 55 and 60 years in treated and untreated patients, respectively; 57% of treated and 55% of untreated patients were male. Treated patients were more likely to be diagnosed from 2012 to 2016, have a history of hematologic malignancy other than MDS and CMML, and have received previous cytotoxic treatment. Treated patients also had higher mean (standard deviation [SD]) total healthcare costs during the 12-month pre-index period, $54,894 ($108,321) vs $50,127 ($101,405). Untreated patients were more likely to be diagnosed from 2008 to 2009, have a history of CMML, solid tumors, COPD, chronic renal disease, CHD, diabetes, and hypertension, and have received HCT. Mean (SD) Charlson comorbidity index for untreated and treated patients was 1.99 (2.12) and 1.20 (1.57) (p-value <0.001), respectively. Untreated patients had more frequent use of all the healthcare resources during the 12-month pre-index period.

Table 1. Patient characteristics at baseline by receipt of AML treatment.

Characteristic	Treated (n = 6,415)	Untreated (n = 3,040)	p-value	Standard difference
Age at index date, years, mean (SD)	55.3 (13.8)	60.3 (15.2)	<.001	0.3490
Age <60, n (%)	3,662 (57.1)	1,328 (43.7)	<.001	–0.2705
Male gender, n (%)	3,650 (56.9)	1,666 (54.8)	.055	–0.0422
Year of index date, n (%)
2008/2009	1,501 (23.4)	834 (27.4)	<.001	0.0928
2010/2011	1,716 (26.7)	849 (27.9)		0.0264
2012/2013	1,637 (25.5)	691 (22.7)		–0.0652
2014/2015	1,430 (22.3)	606 (19.9)		–0.0578
2016	131 (2.0)	60 (2.0)		–0.0049
Comorbidity during pre-index period, n (%)
Chronic myelomonocytic leukemia	63 (1.0)	46 (1.5)	.024	0.0479
MDS	1,147 (17.9)	530 (17.4)	.596	–0.0117
Other hematologic malignancy	2,502 (39.0)	1,066 (35.1)	<.001	–0.0816
Solid tumor	1,702 (26.5)	864 (28.4)	.054	0.0423
COPD	1,235 (19.3)	848 (27.9)	<.001	0.2047
Chronic liver disease	577 (9.0)	278 (9.1)	.812	0.0052
Chronic renal disease	872 (13.6)	717 (23.6)	<.001	0.2590
CHD	1,241 (19.3)	854 (28.1)	<.001	0.2067
Diabetes	1,470 (22.9)	1,093 (36.0)	<.001	0.2891
Hypertension	3,371 (52.5)	1,987 (65.4)	<.001	0.2627
Charlson comorbidity index during 12-month pre-index period (NCI modified version)
Mean (SD)	1.20 (1.57)	1.99 (2.12)	<.001	0.4268
HRU during pre-index period, n (%)
HCT	122 (1.9)	87 (2.9)	.003	0.0630
Radiotherapy	588 (9.2)	248 (8.2)	.107	–0.0358
Cytotoxic treatment	1,271 (19.8)	390 (12.8)	<.001	–0.1898
HMA therapy	485 (7.6)	184 (6.1)	.008	–0.0599
Other cytotoxic treatment	835 (13.0)	229 (7.5)	<.001	–0.1813
HRU during 12-month pre-index period, mean (SD)
Number of physicians’ office or outpatient visits	30.8 (35.0)	34.0 (36.8)	<.001	0.0870
Number of outpatient pharmacy claims	26.8 (29.8)	35.1 (36.1)	<.001	0.2494
Number of emergency department visits	0.78 (2.06)	0.98 (3.14)	.002	0.0734
Number of hospitalizations	0.66 (1.50)	0.74 (1.49)	.020	0.0512
Number of inpatient days	4.90 (13.4)	5.73 (14.4)	.007	0.0601
Healthcare costs during 12-month pre-index period, $, mean (SD)
Outpatient costs	29,738 (63,913)	23,344 (52,409)	<.001	–0.1094
Outpatient pharmacy costs	5,576 (16,119)	8,241 (21,351)	<.001	0.1408
Inpatient costs	19,579 (66,156)	18,543 (64,513)	.469	–0.0159
Total costs	54,894 (108,321)	50,127 (101,405)	.037	–0.0454

Abbreviations. CHD, Coronary heart disease; COPD, Chronic obstructive pulmonary disease; HCT, Hematopoietic cell transplantation; HMA, Hypomethylating agent; HRU, Healthcare resource utilization; MDS, Myelodysplastic syndromes; NCI, National Cancer Institute; SD, Standard deviation.

Healthcare resource utilization and costs during follow-up

The mean (standard error [SE]) duration of follow-up was 18.3 (0.23) months among treated patients and 16.4 (0.38) months among untreated patients. Among treated patients, the majority of follow-up was spent in the initial health state (mean 9.9 months per treated patient), followed by the remission and relapsed states (mean 6.1 and 2.4 months per treated patient, respectively). Mean (SE) total healthcare cost per patient was $386,077 ($4,745) per treated patient and $79,382 ($2,652) per untreated patient (Table 2). In treated patients, 60% of total costs (mean $231,867 per patient) were incurred during the initial health state. Most of these costs were incurred during the first 6 months in the state (mean $173,863 per patient, Figure 3). Mean (SE) total healthcare costs per person-month were $21,055 ($324) among treated patients and $4,854 ($255) among untreated patients (Table 3). For treated patients, monthly costs were greatest during the first 6 months in the initial state ($44,246 per person-month; Figure 3).

Figure 3. Healthcare costs among treated patients by health state and time since entry into health state, (A) per treated patient, $1,000, (B) per month, $1,000.

Table 2. Healthcare resource utilization and costs per newly diagnosed treated and untreated AML patients.

Outcome	Treated patients (n = 6,415)				Untreated patients (n = 3,040)
	Initial state	Remission state	Relapsed state	All
Duration of follow-up, months, mean (SE)	9.87 (0.18)	6.08 (0.17)	2.39 (0.09)	18.34 (0.23)	16.35 (0.38)
Utilization, mean (SE) or % (SE)
Hospitalizations, mean	2.12 (0.03)	0.87 (0.02)	0.54 (0.02)	3.52 (0.04)	1.14 (0.03)
Inpatient days, mean	30.4 (0.40)	8.7 (0.26)	9.0 (1.60)	48.2 (1.68)	11.3 (0.43)
Office or outpatient visits, mean	63.2 (0.95)	28.6 (0.73)	16.3 (0.62)	108.1 (1.24)	36.6 (1.08)
Outpatient pharmacy claims, mean	38.6 (0.92)	22.3 (0.75)	10.8 (0.54)	71.6 (1.25)	43.2 (1.62)
Emergency department visits, mean	1.02 (0.03)	0.41 (0.02)	0.23 (0.01)	1.66 (0.04)	1.09 (0.09)
Percent receiving
HCT	11.5 (0.40)	10.1 (0.38)	5.9 (0.29)	26.3 (0.55)	0
RBC transfusion	48.7 (0.62)	16.6 (0.46)	11.4 (0.40)	57.3 (0.62)	18.5 (0.70)
Platelet transfusion	30.0 (0.57)	10.0 (0.38)	8.6 (0.35)	38.8 (0.61)	7.8 (0.49)
GCSF	27.1 (0.55)	14.1 (0.43)	5.3 (0.28)	40.0 (0.61)	2.6 (0.29)
Percent with evidence of condition
Fever	67.5 (0.58)	21.5 (0.51)	16.6 (0.46)	75.9 (0.53)	28.6 (0.82)
Treated infection	66.4 (0.59)	22.1 (0.52)	16.3 (0.46)	76.3 (0.53)	45.2 (0.90)
Febrile neutropenia	16.7 (0.47)	6.8 (0.31)	4.0 (0.24)	24.5 (0.54)	1.9 (0.25)
Other neutropenia	12.4 (0.41)	5.3 (0.28)	2.4 (0.19)	18.2 (0.48)	1.4 (0.22)
Lymphopenia	1.9 (0.17)	0.9 (0.12)	0.7 (0.10)	3.2 (0.22)	0.4 (0.12)
Anemia	21.6 (0.51)	5.3 (0.28)	2.7 (0.20)	26.8 (0.55)	18.6 (0.71)
Bleeding	24.6 (0.54)	9.2 (0.36)	6.5 (0.31)	36.0 (0.60)	23.1 (0.76)
Thrombocytopenia	64.9 (0.60)	23.0 (0.53)	15.4 (0.45)	74.4 (0.54)	30.2 (0.83)
Total healthcare costs, $, mean (SE)	231,867 (3,366)	84,173 (2,346)	70,038 (2,544)	386,077 (4,745)	79,382 (2,652)
HCT	22,990 (1,217)	19,327 (1,024)	12,034 (836)	54,351 (1,746)	0
Chemotherapy total	105,565 (1,634)	16,222 (731)	22,236 (1,014)	144,023 (2,149)	0
Inpatient chemotherapy	78,154 (1,529)	10,501 (524)	15,186 (844)	103,840 (1,929)	0
Outpatient chemotherapy	27,411 (825)	5,721 (488)	7,050 (484)	40,183 (1,166)	0
Other inpatient	35,727 (1,140)	16,080 (764)	15,226 (898)	67,033 (1,621)	37,325 (1,772)
Other outpatient	54,441 (1,424)	27,755 (1,002)	18,005 (855)	100,200 (1,922)	20,474 (1,084)
Pharmacy	13,143 (533)	4,790 (203)	2,537 (160)	20,469 (590)	21,583 (1,354)

Abbreviations. AML, Acute myeloid leukemia; GCSF, Granulocyte colony-stimulating factors; HCT, Hematopoietic cell transplantation; RBC, Red blood cell; SE, Standard error.

Table 3. Healthcare resource utilization and costs per person-month among newly diagnosed AML patients.

Outcome	Treated patients				Untreated patients
	Initial state	Remission state	Relapsed state	All
No. of patients (%)	6,415 (100)	2,481 (39)	1,460 (23)	6,415 (100)	3,040 (47)
Duration of follow-up, months, mean (SE)	9.87 (0.18)	15.73 (0.36)	10.49 (0.34)	18.34 (0.23)	16.35 (0.38)
Total person-months	63,289	39,016	15,323	117,628	49,719
Utilization, mean (SE)
Hospitalizations	0.215 (0.004)	0.143 (0.003)	0.224 (0.003)	0.192 (0.003)	0.070 (0.004)
Inpatient days	3.08 (0.067)	1.43 (0.032)	3.79 (0.263)	2.63 (0.088)	0.69 (0.046)
Office or outpatient visits	6.41 (0.068)	4.70 (0.048)	6.84 (0.070)	5.89 (0.056)	2.24 (0.051)
Outpatient pharmacy claims	3.91 (0.039)	3.66 (0.040)	4.50 (0.043)	3.90 (0.038)	2.64 (0.052)
Emergency department visits	0.103 (0.003)	0.067 (0.002)	0.098 (0.002)	0.091 (0.002)	0.067 (0.004)
Total healthcare costs, $, mean (SE)	23,502 (433)	13,840 (268)	29,322 (439)	21,055 (324)	4,854 (255)
HCT	2,330 (136)	3,178 (147)	5,038 (190)	2,964 (122)	0
Chemotherapy total	10,700 (334)	2,667 (92)	9,309 (268)	7,854 (191)	0
Inpatient chemotherapy	7,922 (315)	1,726 (72)	6,358 (244)	5,663 (178)	0
Outpatient chemotherapy	2,778 (126)	941 (56)	2,952 (117)	2,191 (76)	0
Other inpatient	3,621 (133)	2,644 (113)	6,375 (190)	3,656 (110)	2,282 (233)
Other outpatient	5,518 (118)	4,563 (96)	7,538 (140)	5,465 (101)	1,252 (74)
Pharmacy	1,332 (26)	788 (16)	1,062 (18)	1,116 (21)	1,320 (47)

Abbreviations. AML, Acute myeloid leukemia; HCT, Hematopoietic cell transplantation; SE, Standard error.

Time to disenrollment

Figure 4 provides a graphical depiction of the time to disenrollment among treated (Figure 4(A)) and untreated (Figure 4(B)) patients. Time to disenrollment was partitioned by health state (initial, remission, and relapse). Median time to disenrollment was 13.1 months (95% confidence interval [CI] = 12.5–13.6 months) among treated patients and 7.0 months (95% CI = 6.1–8.0 months) among untreated patients.

Figure 4. Kaplan-Meier estimates of time to disenrollment, (A) treated patients, (B) untreated patients.

Factors predicting receipt of AML treatment

Patient characteristics associated with increased probability of receiving AML treatment were younger age, male gender, initial diagnosis in year 2012 or later (vs 2008–2011), history of MDS and other hematologic malignancy (excluding CMML), and previous receipt of cytotoxic treatment (excluding HMAs). Characteristics associated with lower probability of receiving AML treatment were history of CMML, COPD, chronic renal disease, dementia, and diabetes, higher Charlson comorbidity index, prior HCT, and higher number of outpatient pharmacy claims (Table 4).

Table 4. Logistic regression analysis of factors predicting receipt of AML treatment.

Characteristic	All covariates included		Stepwise selection of covariates
	Estimate (SE)	p-value	Estimate (SE)	p-value
Intercept	2.2881 (0.162)	<.001	2.2951 (0.134)	<.001
Age at index date, years	–0.0175 (0.003)	<.001	–0.0180 (0.002)	<.001
Age ≥60	0.0025 (0.072)	.972
Male gender	0.1744 (0.047)	<.001	0.1871 (0.047)	<.001
Year of index date (vs 2008/2009)
2010/2011	–0.0538 (0.039)	.164	–0.0557 (0.039)	.149
2012/2013	0.1025 (0.041)	.012	0.1044 (0.041)	.010
2014/2015/2016	0.1137 (0.042)	.006	0.1245 (0.041)	.003
Comorbidity during pre-index period
CMML	–0.5126 (0.206)	.013	–0.5038 (0.206)	.014
MDS	0.2293 (0.076)	.003	0.2231 (0.074)	.003
Other hematologic malignancy	0.1059 (0.053)	.045	0.1199 (0.051)	.019
Solid tumor	0.0356 (0.057)	.531
COPD	–0.1235 (0.058)	.033	–0.1262 (0.058)	.029
Chronic liver disease	0.0913 (0.082)	.268
Chronic renal disease	–0.2495 (0.068)	<.001	–0.2513 (0.067)	<.001
CHD	0.0591 (0.061)	.334
Dementia	–1.2142 (0.302)	<.001	–1.2376 (0.302)	<.001
Diabetes	–0.1598 (0.058)	.006	–0.1641 (0.057)	.004
Hypertension	–0.0361 (0.056)	.517
Charlson comorbidity index during 12-month pre-index period (NCI modified version)	–0.1336 (0.017)	<.001	–0.1333 (0.016)	<.001
HRU during pre-index period
HCT	–0.9173 (0.170)	<.001	–0.8359 (0.153)	<.001
Radiotherapy	–0.0483 (0.096)	.615
Cytotoxic treatment (vs none)
HMA therapy and other cytotoxic treatment	–0.1955 (0.203)	.334	–0.1931 (0.201)	.336
HMA therapy only	0.1062 (0.108)	.323	0.0921 (0.104)	.377
Other cytotoxic treatment only	0.4048 (0.101)	<.001	0.4194 (0.099)	<.001
HRU during 12-month pre-index period
Number of physicians’ office or outpatient visits	–0.0014 (0.001)	.158
Number of outpatient pharmacy claims	–0.0035 (0.001)	<.001	–0.0033 (0.001)	<.001
Number of emergency department visits	–0.0097 (0.010)	.330
Number of hospitalizations	0.0063 (0.026)	.808
Number of inpatient days	–0.0024 (0.003)	.412
Total healthcare costs during 12-month pre-index period, $1,000	0.0009 (0.000)	.022

Abbreviations. AML, Acute myeloid leukemia; CHD, Coronary heart disease; CMML, Chronic myelomonocytic leukemia; COPD, Chronic obstructive pulmonary disease; HCT, Hematopoietic cell transplantation; HMA, Hypomethylating agent; HRU, Healthcare resource utilization; MDS, Myelodysplastic syndromes; NCI, National Cancer Institute.

Total healthcare costs by initial AML regimen

Among the 9,455 patients in the study sample, 1,003 patients were captured in the CDM database. Of these patients, 311 patients had evidence of AML treatment in the CDM database during the initial health state and were included in the CDM database analyses. The 7 + 3 regimen was the most frequently utilized initial regimen (n = 145 [47%]) (Table 5). There were equal numbers of patients who received cytarabine monotherapy (n = 45 [14%]) and HMAs (n = 45 [14%]) as initial treatment. Seventy-six (24%) patients received other regimens. Mean age and Charlson comorbidity index were highest among those receiving HMAs. Mean total healthcare costs during the 12-month period before the index date were highest among patients receiving cytarabine monotherapy. Patients receiving cytarabine monotherapy had the longest mean (SD) duration of follow-up, 21.3 (16.1) months, and the highest mean (SE) total healthcare costs per patient, $496,123 ($47,701). For all regimens, the largest share of costs was incurred during the first 6 months after initial treatment.

Table 5. Total healthcare cost among newly diagnosed AML patients, by initial regimen and health state.

Outcome	Regimen	Initial state	Remission state	Relapsed state	All
No. of patients (%)	7 + 3	145 (100)	78 (54)	46 (32)	145 (100)
	Cytarabine monotherapy	45 (100)	10 (22)	9 (20)	45 (100)
	HMA	45 (100)	10 (22)	11 (24)	45 (100)
	Other	76 (100)	29 (38)	19 (25)	76 (100)
Duration of follow-up per treated patients, months, mean (SE)	7 + 3	4.7 (0.64)	9.2 (1.42)	2.9 (0.59)	16.7 (1.62)
	Cytarabine monotherapy	16.6 (2.33)	3.0 (1.08)	1.7 (0.63)	21.3 (2.40)
	HMA	10.3 (1.42)	1.6 (0.58)	2.0 (0.87)	13.9 (1.61)
	Other	9.4 (1.61)	8.0 (2.11)	3.5 (1.18)	20.9 (2.57)
Total person-months	7 + 3	675	1,328	419	2,422
	Cytarabine monotherapy	749	134	74	958
	HMA	464	70	91	625
	Other	717	610	262	1,589
Total healthcare costs, mean (SE), $
Per treated patient	7 + 3	183,337 (12,723)	117,292 (16,708)	113,310 (20,624)	413,939 (29,686)
	Cytarabine monotherapy	385,068 (41,746)	49,006 (18,528)	62,048 (24,204)	496,123 (47,701)
	HMA	219,527 (31,122)	63,268 (23,339)	55,457 (24,913)	338,252 (50,348)
	Other	240,995 (25,358)	77,114 (19,617)	86,155 (26,534)	404,264 (41,985)
Per person-month	7 + 3	39,363 (3,693)	12,809 (1,519)	39,247 (2,864)	24,784 (2,088)
	Cytarabine monotherapy	23,133 (3,659)	16,444 (2,332)	37,572 (6,382)	23,316 (3,317)
	HMA	21,280 (2,219)	40,438 (4,734)	27,521 (2,112)	24,342 (2,209)
	Other	25,546 (10,912)	9,604 (2,358)	24,962 (2,832)	19,330 (2,795)

Abbreviations. AML, Acute myeloid leukemia; HMA, Hypomethylating agent; SE, Standard error.

Sub-group analysis among patients with tAML or AML-MRC

A total of 2,901 patients were identified with tAML or AML-MRC, including 2,080 (72%) treated patients and 821 (28%) untreated patients. Compared with the overall population of AML patients, patients with tAML or AML-MRC were older (age 59.9 vs 56.9 years, respectively), more likely to be male (58.5% vs 56.2%), and had shorter duration of follow-up, with a mean (SD) of 14.5 (17.4) months vs 17.7 (19.5) months, respectively. Among treated patients, mean total healthcare costs during follow-up were similar among patients with tAML or AML-MRC vs among all AML patients ($352,606 vs $386,077, respectively). Among untreated patients, mean total healthcare costs during follow-up were also similar among patients with tAML or AML-MRC vs among all AML patients ($80,536 vs $79,382, respectively), reflecting higher monthly costs ($9,130 vs $4,854) but shorter follow-up (8.8 vs 16.4 months, respectively). Detailed results for patients with tAML or AML-MRC are reported in the Supplemental tables and figures.

Discussion

Summary

In this retrospective cohort study of patients with newly diagnosed AML in the US using a large health insurance claims database combined with hospital CDM data for a sub-set of patients, ∼ 60% of patients with newly diagnosed AML received treatment with chemotherapy or HCT. Treated patients were younger and had fewer comorbidities than untreated patients. Mean total and monthly healthcare costs were substantially greater in treated vs untreated patients ($386,077 vs $79,382 and $21,055 vs $4,854, respectively). For the treated patients, the largest share of total costs was incurred during the initial health state prior to diagnosis of remission or relapse. Among patients for whom initial therapy could be assessed using CDM data, almost half received 7 + 3 regimens. Monthly costs were similar for patients receiving 7 + 3, cytarabine monotherapy, or HMAs. Follow-up was shortest for patients receiving HMAs.

Comparisons with recent literature

In our study, conducted among commercially insured patients with AML, 68% of the overall AML population and 72% of patients in the tAML/AML-MRC sub-group received chemotherapy or HCT. In studies of elderly patients using Medicare data, Medeiros et al.⁷, Katz et al.⁵, and Meyers et al.⁶ reported receipt of chemotherapy among 40%, 42%, and 57% of AML patients, respectively. The higher treatment rate reported in our study likely reflects the younger age of patients in our sample due to our analysis’ focus on commercially insured patients rather than a population exclusively insured by Medicare.

Our study found that younger age, male gender, history of MDS and other hematologic malignancy (excluding CMML), and previous receipt of cytotoxic treatment (excluding HMAs) were associated with increased probability of receiving AML treatment. We also found that history of CMML, COPD, chronic renal disease, dementia, diabetes, higher Charlson comorbidity index, and prior HCT were associated with lower probabilities of receiving treatment. In a study of elderly AML patients, Medeiros et al.⁷ reported that treated patients were younger, more likely to be male, had better performance status and fewer comorbidities, and were less likely to have secondary AML (defined as prior MDS or myeloproliferative neoplasm) compared with untreated patients. These finding are similar to ours, with the exception of the observed association between prior MDS and receipt of AML treatment.

In our study, mean total healthcare costs were $287,467 among all patients (treated and untreated). In contrast, Meyers et al.⁶ reported mean total healthcare costs of $96,078 among 4,058 treated and untreated elderly AML patients using SEER Medicare data from 1997–2007. The higher costs in our study are likely related to differences in the patient population (commercially insured vs Medicare) and calendar years.

The mean monthly costs for treated patients in our study were $21,055. This result is similar to that reported by Bell et al.¹¹, who studied healthcare costs among AML patients aged 60 years or older using the Optum claims database from 2008–2015. They reported that, among 183 treated patients, the mean total cost during follow-up was $25,243 per person per month.

In the current study, mean total costs among treated patients during the initial/induction state were $231,867 over a mean duration of 9.9 months with estimated mean monthly costs of $23,502. Katz et al.⁵ reported the mean total costs of inpatient induction was $163,159 among 113 elderly AML patients from 2000–2003. Irish et al.⁸ reported that the mean cost of first-line induction therapy was $208,857 among 681 commercially insured and Medicare patients in the MarketScan database from 2009–2015. Bell et al.¹¹ reported the mean monthly cost of treatment during the first year of follow-up among 183 elderly AML patients was $27,756. Our estimates of the costs for the initial state representing the period during which patients are receiving induction therapy are, therefore, generally consistent with the range of costs reported previously.

Using IMS Health PharMetrics Plus and CDM databases from January 2006 to June 2016, Medeiros et al.⁹ identified AML treatment episodes defined as inpatient high-dose cytarabine ± anthracycline (high-intensity chemotherapy; HIC) induction, HIC consolidation, low-intensity chemotherapy (LIC), HCT, and relapsed AML after HIC/LIC/HCT (R/R). Mean cost of episode was $198,528, $73,304, $53,081, $329,621, and $145,634 among HIC induction, HIC consolidation, LIC, HCT, and R/R patients, respectively. The estimated mean cost of the HIC induction state reported by Medeiros et al. ($198,528) is less than the mean costs during the initial/induction state among treated patients in our study ($231,867). The mean total costs for the relapsed state for treated patients in our study is greater than the costs of an R/R episode reported by Medeiros et al. ($307,733 vs $145,634). On one hand, these differences may reflect differences in the algorithms used to define remission and relapse health states in our study (i.e. based exclusively on diagnosis codes for AML) and, on the other hand, they may represent episodes in the study by Medeiros et al. (i.e. based on both diagnosis codes and specific medications received).

We identified only one study that examined the specific induction treatments received among newly diagnosed AML patients. Bell et al.¹⁰ studied treatment characteristics and outcomes among newly diagnosed AML patients aged 60 years or older who were treated with either 7 + 3 or a HMA over the period of January 1, 2008 to July 31, 2015 using the Humedica database. Among the 274 patients who qualified for the study, the majority (60%) received 7 + 3 chemotherapy. In our analyses of AML patients by initial regimen received, we found 145 and 45 out of 311 patients received 7 + 3 and HMA regimen, respectively. Thus, among those receiving 7 + 3 or HMAs, 76% received 7 + 3 as the initial treatment, which is somewhat greater than that reported by Bell et al. This difference might relate to the different populations of the two studies, or that our study evaluated initial treatment, whereas Bell et al. examined any treatment during the follow-up period. In our study, patients who received HMAs as initial therapy were older and had more comorbidities than those receiving 7 + 3. These results are consistent with those reported by Bell et al.¹⁰.

Limitations

Our study is subject to several limitations. First, as with all studies using health insurance claims data, identification of study subjects, assessment of their baseline characteristics, and measurement of study outcomes are subject to possible errors due to miscoding of diagnoses and/or procedures on health insurance claims. Second, although patients in the PharMetrics Plus database are geographically diverse, they may not be representative of all commercially insured patients in the US. Third, while the PharMetrics Plus database includes patients over 65 years, these patients are largely limited to those enrolled in managed care plans. Accordingly, elderly patients are under-represented, and those that are included may not be representative of patients enrolled in fee-for-service Medicare plans. Fourth, the database does not include information on clinical outcomes such as response, relapse, or mortality. Instead, we identified health states using diagnosis codes for AML that differentiated between AML with and without remission or relapse, although the extent to which these codes corresponded to clinical assessments of remission and relapse was unknown. Although patients who died were disenrolled from their health plans, it was not possible to distinguish between disenrollment due to death and disenrollment for other reasons. Because patients with an incurable disease, such as AML, are probably less likely to disenroll for other reasons, time to disenrollment is likely to be highly correlated with overall survival. Fifth, no inflation adjustment was applied in the cost analyses, since published CPI data on medical care are not necessarily representative of inflation for medical care services for AML patients. Sixth, because the CDM database does not cover all patients in the PharMetrics Plus database and contributing hospitals in the CDM database and PharMetrics Plus database do not necessarily overlap, a relatively small sample size was available for analyses of total costs by initial regimen. Lastly, characterization of patients into initial, remission, and relapsed health states was based on diagnosis codes on health insurance claims. The extent to which this coding accurately corresponds to clinical assessments of remission and relapse has not been established.

Conclusions

Healthcare resource utilization and costs of managing AML patients are substantial. Patients who were male, of a younger age, had a history of MDS or other hematologic malignancy (excluding CMML), and had previously received cytotoxic treatment (excluding HMA) were more likely to be treated. In treated patients, most of the costs are incurred during the initial period before remission or relapse. Patients with tAML/AML-MRC have similar healthcare resource utilization and costs as the overall AML population and have shorter follow-up, possibly reflecting poorer survival.

Transparency

Declaration of funding

This study was sponsored by Jazz Pharmaceuticals, Inc.

Declaration of financial/other interests

MH, AS, and TED have received research funding from Jazz Pharmaceuticals, Inc. KCC was employed by Jazz Pharmaceuticals at the time of this research and is currently employed by Juno Therapeutics, Inc., a Celgene company, and has stock ownership in Amgen, Baxter, Bayer, Celgene, Gilead, Jazz Pharmaceuticals, and Shire. A peer reviewer on this manuscript discloses that they are employed by Pfizer, Inc.; the remaining peer reviewers have no conflicts of interest to disclose.

Acknowledgments

Editorial support was provided by Kimberly Brooks, PhD, CMPP of SciFluent Communications, and was financially supported by Jazz Pharmaceuticals, Inc.

Note

Notes

a In the PharMetrics Plus database, there are patients covered by the Medicare Cost plan, also known as Medicare Gap or Medicare Supplemental. It is an insurance purchased by individuals to cover services not covered by traditional Medicare insurance. Patients submit claims to either Medicare or the Medicare Cost insurer depending on the services consumed. Only the data submitted to the Medicare Cost insurer are captured in the PharMetrics Plus database. Thus, patients on Medicare Supplemental plan lack complete healthcare information and were excluded from the study.