Costs and health care resource utilization among Medicare beneficiaries diagnosed with chronic lymphocytic leukemia

Abstract

BACKGROUND:

Chronic lymphocytic leukemia (CLL) is the most common type of leukemia. However, published studies of CLL have either only focused on costs among individuals diagnosed with CLL without a non-CLL comparator group or focused on costs associated with specific CLL treatments. An examination of utilization and costs across different care settings provides a holistic view of utilization associated with CLL.

OBJECTIVE:

To quantify the health care costs and resource utilization types attributable to CLL among Medicare beneficiaries and identify predictors associated with each of the economic outcomes among beneficiaries diagnosed with CLL.

METHODS:

This retrospective study used a random 20% sample of the Medicare Chronic Conditions Data Warehouse (CCW) database covering the 2017-2019 period. The study population consisted of individuals with and without CLL. The CLL cohort and non-CLL cohort were matched using a 1:5 hard match based on baseline categorical variables. We characterized economic outcomes over 360 days across cost categories and places of services. We estimated average marginal effects using multivariable generalized linear regression models of total costs and across type of services. Total cost was compared between CLL and non-CLL cohorts using the matched sample. We used generalized linear models appropriate for the count or binary outcome to identify factors associated with various categories of health care resource utilization, such as inpatient admissions, emergency department (ED) visits, and oncologist/hematologist visits.

RESULTS:

A total of 2,736 beneficiaries in the CLL cohort and 13,571 beneficiaries in the non-CLL matched cohort were identified. Compared with the non-CLL cohort, the annual cost for the CLL cohort was higher (CLL vs non-CLL, mean [SD]: $22,781 [$37,592] vs $13,901 [$24,725]), mainly driven by health care provider costs ($6,535 vs $3,915) and Part D prescription drug costs ($5,916 vs $2,556). The main categories of health care resource utilization were physician evaluation/management visits, oncologist/hematologist visits, and laboratory services. Compared with beneficiaries aged 65-74 years, beneficiaries aged 85 years or older had lower use and cost in maintenance services (ie, oncologist visits, hospital outpatient costs, and prescription drug cost) but higher use and cost in acute services (ie, ED). Compared with residency in a metropolitan area, living in a nonmetropolitan area was associated with fewer physician visits but higher ED visits and hospitalizations.

CONCLUSIONS:

The cooccurrence of lower utilization of routine care services, along with higher utilization of acute care services among some individuals, has implications for patient burden and warrants further study.

Plain language summary

We quantify the costs and utilization attributable to CLL and identify factors associated with these outcomes. Relative to beneficiaries without CLL, the mean total cost among beneficiaries diagnosed with CLL is 63% higher (or an increased cost of $8,880), driven by health care provider visits, outpatient hospital costs, and prescription drug costs. Older age and residence in a nonmetropolitan area was associated with lower utilization of routine care services and higher utilization of acute care services.

Implications for managed care pharmacy

This article adopts a payer perspective and provides novel information about the relative contribution to CLL-related health care utilization and costs across points of care. Our identified predictors point to different factors that should be considered. Our results related to nonmetropolitan residence may indicate that attention to regional variations in access to health care providers is warranted, whereas the disparity in specialist visits among older beneficiaries may relate to health care–seeking behavior.

Chronic lymphocytic leukemia (CLL)/small lymphocytic lymphoma (SLL) is the most common type of leukemia/lymphoma among US adults.¹ In the United States in 2023, approximately 18,740 persons were diagnosed with CLL and an estimated 4,490 deaths resulted from CLL.¹ Given the associated cost burden, several studies have examined the costs of care and health care resource utilization (HCRU) among individuals diagnosed with CLL.^2-4 The availability of novel and costly oral CLL therapies has warranted an examination of CLL-attributable costs. Available evidence indicates that costs can vary based on the treatment received and because of the care setting (eg, pharmacy, outpatient) in which individuals received treatment. Emond et al compared the HCRU and health care costs of ibrutinib monotherapy, chemoimmunotherapy regimens, and bendamustine/rituximab treatment using private payer and Medicare Advantage claims data from 2014 to 2017 and found lower overall costs in the ibrutinib group compared with the chemoimmunotherapy group.⁴ Ibrutinib, given orally, is generally associated with higher pharmacy costs, which are offset by the lower medical costs, whereas chemoimmunotherapy is associated with higher medical costs and lower pharmacy costs.⁴ Using a private payer claims database covering the 2016-2017 time period, Huang et al compared costs among individuals receiving ibrutinib or chemoimmunotherapy in the first-line setting.⁵ They found similar results, including lower monthly overall and medical costs among the ibrutinib group compared with the chemoimmunotherapy group. Although comparatively fewer studies have focused on the care setting as an independent factor, there is some evidence that costs can vary because of the care setting. One study identified higher (total; infusion day drug and administration) costs among individuals receiving rituximab in the hospital outpatient setting compared with those receiving rituximab in the physician office/community clinic setting.⁶

Most previous cost studies of CLL have either only focused on costs among individuals diagnosed with CLL without a non-CLL comparator group^2,3 or only examined costs associated with specific CLL treatments (eg, ibrutinib, chemoimmunotherapy).^4,5,7 To the best of our knowledge, one study has quantified the incremental costs of CLL (ie, comparing costs among individuals with and without CLL).⁸ Comparing beneficiaries diagnosed with CLL to a 1:3 matched cohort of individuals without a CLL diagnosis, using Medicare fee-for-service (FFS) claims data from 1999 to 2007, Lafeuille et al reported that individuals diagnosed with CLL generated average costs of $87,151 compared with $47,642 for matched controls (P < 0.001).⁸ They also found that compared with matched controls, beneficiaries diagnosed with CLL generated significantly higher monthly costs during their continuing and terminal phases. Reporting on the costs associated with the most frequently prescribed CLL treatments among beneficiaries diagnosed with CLL, they found that the average costs per patient were $5,140 for rituximab and $953 for radiation therapy.⁸

In recent years, with the approval of several new therapeutic agents (eg, duvelisib for relapsed or refractory CLL in 2018, venetoclax and acalabrutinib for untreated CLL in 2019)^9-11 and new combination therapies for the treatment of CLL,¹² it is important to quantify the costs and HCRU among individuals diagnosed with CLL using real-world data covering this time period. In addition, there is no prior study that examines costs and resource use across all points of care, including inpatient, emergency department (ED), skilled nursing facility, physician visits, and general outpatient care. This examination is necessary to better understand how the costs of CLL distribute across the various points of care in which individuals with CLL receive health care services.

To address these evidence gaps, this study characterized the costs and HCRU among individuals diagnosed with CLL. The study aims were to (1) quantify overall costs across points of care and quantify predictors of higher costs among individuals diagnosed with CLL; (2) quantify the incremental cost of CLL by comparing beneficiaries diagnosed with CLL to matched beneficiaries without CLL; and (3) quantify HCRU across points of care and identify predictors of increased HCRU among individuals diagnosed with CLL.

Methods

This retrospective study used a random 20% sample of the Medicare Chronic Conditions Data Warehouse (CCW) database covering the 2017-2019 period. The study population consisted of individuals with and without CLL. Individuals with a CLL diagnosis were defined as having at least 2 medical (ie, Parts A or B) claims with diagnosis codes for CLL or SLL (International Classification of Diseases, Tenth Revision, Clinical Modification codes: C91.1x, C83.0x) on separate dates that were at least 30 days apart during the study period and having their first claim with a diagnosis code for CLL or SLL during the cohort identification period (ie, July 1, 2017, to June 30, 2019). The index date was defined as the date of the first claim with a CLL diagnosis code during the cohort identification period. Beneficiaries were followed until disenrollment from FFS Medicare (ie, Parts A, B, or D), enrollment in a health maintenance organization (HMO), death, or the end of the study period (ie, December 31, 2019), whichever came first.

The following study inclusion criteria were applied: age above 65 years, at least 180-day pre-index (baseline period) continuous enrollment and 360-day follow-up period following the index date, and no HMO enrollment during the baseline period (to ensure data completeness). Individuals were excluded if they had any malignancy, including CLL, or any evidence of CLL-directed treatment (ie, systemic therapy, stem cell transplantation, and splenectomy) during the baseline period.

Individuals in the non-CLL matched cohort were identified as having at least one medical claim during the cohort identification period (ie, July 1, 2017, to June 30, 2019) and no claim with a diagnosis of CLL during the study period (ie, January 1, 2017, to December 31, 2019). The index date for the non-CLL group was defined as the date of the first claim during the cohort identification period. The CLL cohort and non-CLL cohort were matched using a 1:5 hard match based on 5 baseline categorical variables and on a categorical measure of follow-up time as a proxy for baseline disease severity. The matching variables were age, sex, count of Centers for Medicare & Medicaid Services (CMS) hierarchical condition categories (HCCs), receipt of the low-income subsidy (LIS), urban/rural location, and length of follow-up time (in days).

For the matching variables, the following categories were used. The individual’s age at index was grouped into 4 categories as follows: (1) 65-69 years, (2) 70-79 years, (3) 80-89 years, and (4) 90 years or older. The count of HCCs was grouped in categories at its tail values to facilitate the matching process. Otherwise, the HCC variable was matched based on its actual value. Specifically, HCC < 11 was left as a single category, whereas the remaining values were grouped as follows: (1) 11-14, (2) 15-19, (3) 20-24, and (4) 25 and above. Urban-rural categories were defined as follows: (1) metropolitan, (2) nonmetropolitan (urban population of ≥2,500 people), and (3) nonmetropolitan (completely rural or <2,500 urban population). The length of follow-up was grouped into quartile values as follows: (1) less than 375 days, (2) 375-600 days, (3) 601-794 days, and (4) 795 days or more. A report of the distribution of baseline covariates between CLL and non-CLL cohorts is displayed in tabular format (Table 1). All CLL patients who met the study criteria were included even after matching.

TABLE 1.

Baseline Characteristics Comparing Beneficiaries Diagnosed With CLL vs Non-CLL Cohort After Matching Among a Total of 16,307 Individuals

	Non-CLL		CLL		P a	Standardized difference
	n	%	n	%
Sample size	13,571	100	2,736	100
Age, years					0.05	0.08
65-74	6,190	46	1,155	42
75-84	4,847	36	1,078	39
85+	2,534	19	503	18
Sex					0.98	<0.01
Female	6,792	50	1,370	50
Male	6,779	50	1,366	50
Race and ethnicity					<0.01	0.15
White non-Hispanic	11,971	88	2,520	92
Non-White and/or Hispanicb	1,364	10	163	6
Unknown	236	2	53	2
Census region					0.30	0.08
South and other	4,910	36	931	34
Midwest	3,158	23	672	25
Northeast	3,017	22	647	24
West	2,486	18	486	18
Urban/rural residence					0.85	<0.01
Metropolitan	10,845	80	2,182	80
Nonmetropolitan	2,726	20	554	20
Part D low-income subsidy status ever					0.98	<0.01
No	11,813	87	2,381	87
Yes	1,758	13	355	13
Use of preventive services					<0.01	0.12
No	1,290	10	173	6
Yes	12,281	90	2,563	94
Poor performance status and frailty proxyc					0.89	<0.01
No	10,052	74	2,030	74
Yes	3,519	26	706	26
Charlson Comorbidity Index score					<0.01	0.06
0	4,897	36	1,064	39
1-2	4,745	35	939	34
3+	3,929	29	733	27
CMS-HCC condition counts					0.89	<0.01
0	4,293	32	863	32
1	3,103	23	624	23
2-3	3,523	26	712	26
4 +	2,652	20	537	20
Received any treatment during follow-up
No	NA	NA	2,298	84
Yes	NA	NA	438	16
Type of treatment received as first COT
No treatment	NA	NA	2,298	84
IB	NA	NA	147	5
RM	NA	NA	157	6
RM-BM	NA	NA	55	2
OB	NA	NA	20	1
Other	NA	NA	59	2

Variables that are used in the 1:5 hard matching process include the following: age, sex, CMS-HCC counts, Part D low-income subsidy status, urban/rural residence, and length of follow-up time.

^a Student’s t-test was used for continuous variables, chi-square test was used for categorical variables, and Cochran-Mantel-Haenszel test was used for multilevel nominal variables.

^b Non-White and/or Hispanic includes Black, Asian, Hispanic, Native American, and other as defined in the CMS Master Beneficiary Summary File.

^c Baseline poor performance status and frailty proxy was measured as a flag based on any of the following events during baseline period: ED visits, hospitalizations, skilled nursing facility stays, home health services, hospice services, and medical equipment uses of hospital bed, oxygen, wheelchair, and walking aids.

BM = bendamustine; CLL = chronic lymphocytic leukemia; CMS = Centers for Medicare & Medicaid Services; COT = course of therapy; HCC = hierarchical condition category; IB = ibrutinib; NA = not applicable; OB = obinutuzumab; RM = rituximab.

We characterized CLL treatments using the National Comprehensive Cancer Network guidelines for the relevant study years and assigned individuals to mutually exclusive groups based on the first treatment received. The CLL treatments included: rituximab monotherapy, ibrutinib monotherapy, bendamustine-rituximab (BM-RM), obinutuzumab monotherapy, and other treatments (Supplementary Table 1 ^{(184.1KB, pdf)} , available in online article). Study outcomes were (1) the cost of CLL overall and across points of care and (2) the count of encounters across points of care among individuals with CLL.

Individual-level variables included the following baseline measures: age, sex, race, ethnicity, count of CMS-HCC, receipt of the Part D LIS, urban/rural location, use of preventive services, and proxy indicators for poor performance status. The receipt of preventive services served as a proxy for an individual’s health-seeking behavior. We created a composite measure based on any utilization of the following preventive services during the baseline period: at least one claim for influenza vaccination, colorectal cancer screening (fecal occult blood test), prostate cancer screening (prostate-specific antigen blood test), mammography screening, or cervical cancer screening (Pap test). A baseline poor performance status and frailty proxy indicator was measured as a flag based on any of the following events during baseline period: emergency visits, hospitalizations, skilled nursing facility stays, home health services, hospice services, and medical equipment uses of hospital bed, oxygen, wheelchair, and/or walking aids.

Measures of interest were reported using summary statistics that were tailored to the variable type (ie, categorical, semicontinuous, count). For categorical variables, we reported the number and percentage. For semicontinuous and count variables, we reported the median, interquartile range, mean, and SD. All baseline characteristics were reported as categorical variables. We evaluated group differences by calculating standardized differences and tested for group differences using chi-square tests.

ANALYSIS OF COST VARIABLES: AVERAGE COSTS OF CARE

We summed costs over the period of follow-up to calculate total overall costs per CLL patient and across points of care. We reported costs per patient, mean, SD, median, and interquartile ranges. We calculated the number and percentage of individuals with zero costs for all cost variables. The multivariable generalized linear regression models of total costs and across type of services were used to identify factors associated with the costs of care among beneficiaries diagnosed with CLL while considering demographic, clinical, and contextual factors. For all included covariates, we reported average marginal effects, along with CIs, calculated using the Delta method.

ANALYSIS OF COST VARIABLES: INCREMENTAL COST OF CLL

Total cost of CLL was compared between CLL and non-CLL cohorts using the matched sample. Incremental costs between the two groups were calculated, focusing on the total costs of care.

ANALYSIS OF COUNT VARIABLES

HCRU was defined based on inpatient admissions, ED visits, skilled nursing facility stays, health care provider encounters (ie, carrier claims), physician office visits, oncologist/hematologist visits, nonphysician outpatient encounters (eg, laboratory and radiology services), home health care encounters, hospice encounters, and prescription drug fills. Binary variables (yes, no) were created for hospitalizations and ED visits, whereas the remaining HCRU categories were analyzed as count variables. For model specification, we used a likelihood-based model-selection approach (ie, the Akaike information criterion and the Bayesian information criterion to compare the performance of alternative models). Count models with the best fitting distribution were used to report effect sizes.

Results

The final sample included 2,736 beneficiaries diagnosed with CLL who met our inclusion criteria, of whom 16% (n = 438) received CLL treatment during the 1-year follow-up period. The cohort flow diagram is displayed in Figure 1. Among the 438 beneficiaries who received CLL treatment, the distribution of the first course of treatment was 34% ibrutinib, 36% rituximab, 13% BM-RM, 5% obinutuzumab, and 13% “other treatment,” of which the latter is a group composed of treatments with counts that are too small to report (per the data use agreement requirements). After matching, 13,571 individuals were identified in the non-CLL cohort. Beneficiaries’ characteristics are shown in Table 1. Among beneficiaries diagnosed with CLL, 50% were female, 92% were white/non-Hispanic, and 42% were aged 65-74 years.

FIGURE 1.

Sample Selection Flowchart for the Study Population of the CLL Cohort

The distribution of annual health care costs per person for CLL and non-CLL cohorts is presented in Supplementary Table 2 ^{(184.1KB, pdf)} . The mean (SD) annual total costs were $22,781 ($37,592) and $13,901 ($24,725) for CLL and matched non-CLL cohorts, respectively. After adjusting for unbalanced baseline variables (ie, race and ethnicity and preventive services) in the regression model, the annual total cost for the CLL cohort was 62% higher than that for the non-CLL cohort (adjusted cost ratio = 1.62; 95% CI = 1.54-1.71; P < 0.01), and the adjusted incremental cost was $8,922 (95% CI = $7,821-$10,024). This difference was mainly driven by hospital outpatient, health care provider, and Part D prescription drug costs. Based on descriptive statistics, the hospital outpatient costs were higher among those who received infused therapies (eg, RM: $17,982, RM-BM: $27,327, obinutuzumab: $20,433), as compared with the oral treatment group (ie, ibrutinib: $5,696), the no-treatment group ($2,472), or the non-CLL control group ($2,320). Descriptive results for health care costs per person per year for CLL patients overall and by type of initial treatment are available from the authors on request. Figure 2 compares cost components across initial treatment groups among beneficiaries diagnosed with CLL. Beneficiaries who received infusion medications (eg, rituximab, BM-RM) as initial treatment post-index date had a relatively higher proportion of costs associated with health care provider and hospital outpatient services, whereas beneficiaries who were prescribed with ibrutinib had a relatively higher proportion of Part D prescription cost.

FIGURE 2.

Proportion of Average Health Care Cost Across Places of Services per Person Over a 1-Year Period for the Matched Non-CLL and CLL Cohorts, Including CLL Subgroups Defined by Type of Initial Course of Treatment

The main service categories that differed between CLL cohort and non-CLL matched cohort regarding HCRU were physician evaluation/management visits (CLL vs non-CLL, mean [SD]: 21 [24] vs 16 [21]), oncologist/hematologist visits (4 [7] vs 1 [4]), and laboratory claim counts (4 [6] vs 2 [4]), as presented in Supplementary Table 3 ^{(184.1KB, pdf)} . Among the CLL cohort, Supplementary Figure 1 ^{(184.1KB, pdf)} illustrates the distribution of HCRU counts across initial treatment groups. Beneficiaries treated initially with BM-RM had the highest counts of health care provider claims and oncologist/hematologist visits and laboratory claims.

Table 2 presents the identified statistically significant predicters for health care costs and their average marginal effects among beneficiaries diagnosed with CLL. Some baseline factors were positively associated with both cost and use across places of services. Specifically, male sex, the Part D LIS indicator, the poor performance status/frailty proxy indicator, and higher counts of CMS-HCC comorbidities were associated with higher total costs and costs across various points of care. Similar results were found in the count models for male sex, Part D LIS, poor performance status proxy, and counts of CMS-HCC comorbidities (Table 3). The results related to age and urban-rural residency varied across the places of service. Older age and nonmetropolitan residency were associated with lower utilization and cost of maintenance care and with more utilization and cost of acute care services. Compared with beneficiaries aged 65-74 years, beneficiaries aged 85 years or older had fewer oncologist/hematologist visits, lower hospital outpatient cost, and lower prescription drug cost but more physician evaluation and management sits, more ED visits, and higher ED costs. Compared with those living in a metropolitan area, beneficiaries diagnosed with CLL living in a nonmetropolitan area had fewer physician visits, fewer oncologist/hematologist visits, and lower health care provider costs but higher hospital outpatient costs, higher odds of any inpatient cost, and more ED visits.

TABLE 2.

Average Marginal Effect of Covariates on Health Care Costs Across Types of Service Among Beneficiaries Diagnosed With Chronic Lymphocytic Leukemia

Variable	Total cost, $	Health care provider cost a , $	Hospital outpatient cost, $	Part D prescription drug cost, $	ED cost, $	Any inpatient cost, %	Any SNF cost, %
Age, years
65-74	ref	ref	ref	ref	ref	ref	ref
75-84	−2,094	−504	−1,108c	−828	213c	−0.003	0.025d
85+	−2,238	−746	−2,100e	−2,446c	893e	0.000	0.109e
Sex
Female	ref	ref	ref	ref	ref	ref	ref
Male	4,715e	489	1,278e	1,536	125	0.029c	−0.015
Race and ethnicity
White non-Hispanic	ref	ref	ref	ref	ref	ref	ref
Non-White (including Hispanic) or unknownb	−494	−172	−266	−498	304	−0.030	0.008
Census region
South and other	ref	ref	ref	ref	ref	ref	ref
Midwest	−2,037	−1,476d	934c	−1,524	61	0.010	−0.001
Northeast	−191	−1,190c	1,234c	−733	255c	0.004	0.004
West	2,785	−806	248	3,423	446d	−0.022	−0.021
Urban-rural residency
Metropolitan	ref	ref	ref	ref	ref	ref	ref
Nonmetropolitan	1,323	−2,118e	1,356c	−76	−127	0.062e	0.014
Part D low-income subsidy indicator
No	ref	ref	ref	ref	ref	ref	ref
Yes	7,804d	−315	−25	7154d	728e	0.000	0.050e
Preventive service use
No	ref	ref	ref	ref	ref	ref	ref
Yes	4,286	1,648d	1,779e	2,563c	−405	−0.005	−0.032
Poor performance status/frailty proxy indicator
No	ref	ref	ref	ref	ref	ref	ref
Yes	3,483c	721	−213	−691	823e	0.059e	0.030d
CMS-HCC index score
0	ref	ref	ref	ref	ref	ref	ref
1	2,802	1,622e	−7	−1,586	52	0.024	−0.002
2-3	7,274e	2,576e	1,177c	325	296c	0.038c	0.017
4+	16,944e	4,423e	2,926e	3,924c	1,037e	0.083e	0.065e

^a Health care provider total claim cost is derived from the carrier claims data file in the Chronic Conditions Data Warehouse database.

^b Non-White (including Hispanic) includes Black, Asian, Hispanic, Native American, and other as defined in the CMS Master Beneficiary Summary File.

^cP < 0.05.

^dP < 0.01.

^eP < 0.001.

CMS = Centers for Medicare & Medicaid Services; ED = emergency department; HCC = hierarchical condition category; SNF = skilled nursing facility; ref = reference.

TABLE 3.

Average Marginal Effect of Covariates on Quantities of Health Care Utilization Across Types of Service Among Beneficiaries Diagnosed With Chronic Lymphocytic Leukemia

Variable	Physician E/M visits	Oncologist/hematologist visits	ED visits	Laboratory claims	Radiology claims
Age, years
65-74	ref	ref	ref	ref	ref
75-84	0.309	−0.432b	0.114b	−0.320	0.100
85+	3.879c	−1.560d	0.482d	−0.670c	0.203c
Sex
Female	ref	ref	ref	ref	ref
Male	1.184	0.348b	0.043	0.073	0.083
Race and ethnicity
White non-Hispanic	ref	ref	ref	ref	ref
Non-White (including Hispanic) or unknowna	1.145	0.704b	0.076	0.014	0.064
Census region
South and other	ref	ref	ref	ref	ref
Midwest	−3.470d	−0.808d	0.011	0.871d	0.084
Northeast	−0.084	−0.829d	0.145b	1.025d	0.065
West	−2.715b	−0.337	0.050	0.090	−0.014
Urban-rural residency
Metropolitan	ref	ref	ref	ref	ref
Nonmetropolitan	−3.977d	−0.552c	0.241c	0.534b	0.176b
Part D low-income subsidy indicator
No	ref	ref	ref	ref	ref
Yes	5.015d	−0.183	0.418d	0.734c	0.406d
Preventive service use
No	ref	ref	ref	ref	ref
Yes	2.053	2.002d	0.001	1.117d	0.184
Poor performance status/frailty proxy indicator
No	ref	ref	ref	ref	ref
Yes	5.256d	0.018	0.507d	0.732d	0.458d
CMS-HCC index score
0	ref	ref	ref	ref	ref
1	2.150b	0.100	0.038	0.193	0.055
2-3	5.440d	0.370	0.159b	0.592c	0.187c
4+	15.689d	1.020d	0.463d	1.685d	0.523d

^a Non-White (including Hispanic) includes Black, Asian, Hispanic, Native American, and other as defined in the CMS Master Beneficiary Summary File.

^bP < 0.05.

^cP < 0.01.

^dP < 0.001.

CMS = Centers for Medicare & Medicaid Services; ED = emergency department; E/M = evaluation and management; HCC = hierarchical condition category; ref = reference.

Discussion

This retrospective cohort study used Medicare claims data to quantify health care costs and utilization among beneficiaries diagnosed with CLL in comparison with a matched cohort without CLL. To improve our understanding of the economic burden in Medicare beneficiaries diagnosed with CLL, we reported descriptive statistics and covariate-adjusted marginal effects of predictors across all care settings. Overall, relative to the mean total cost among beneficiaries without CLL, the mean total cost among beneficiaries diagnosed with CLL is 63% higher (or an increased cost of $8,880), driven by increased health care provider visits, outpatient hospital costs, and prescription drug costs. Notably, among beneficiaries diagnosed with CLL, and compared with adults aged 65-74 years, adults aged 85 years or older had lower hospital outpatient and prescription drug cost but incurred higher ED cost. We identified a similar pattern among nonmetropolitan residents. Nonmetropolitan residents had fewer physician visits but higher odds of inpatient costs and more ED visits.

The cooccurrence of higher costs in one care setting and lower costs in another has been described as offsetting costs in prior studies focused on specific treatments.⁴ Defining offsetting costs as higher in one category and lower in another category is appropriate when comparing costs of the same type, such as acute care costs. Our findings suggest that lower routine care costs and higher acute care costs are more common for a subset of the population. Rather than offsetting each other, lower routine care costs and higher acute care costs may be mutually reinforcing and signal greater patient burden. Additional research is needed to better understand the predictors and implications of the cooccurrence of lower routine care utilization and higher acute care utilization. If this pattern reflects barriers to care or similar factors, that will be important to study given that this pattern was associated with a vulnerable subgroup (ie, the oldest individual in the study sample).

A common focus of prior cost studies has been to quantify and compare the cost of CLL treatments among beneficiaries diagnosed with CLL and across points of care. Goyal et al studied the economic burden of CLL in beneficiaries using Medicare claims data from 2012 to 2016 and evaluated the costs across various points of care, treatment regimens, and the observed line of therapy.² Kabadi et al quantified costs among beneficiaries with employer-sponsored health insurance using private payer claims data from 2012 to 2015 and calculated the costs based on the care setting and specific CLL treatment.³ Dashputre et al reported on the relationship between costs of care and adherence to oral oncolytic therapy (ie, ibrutinib or idelalisib) using claims data of commercial and Medicare supplemental insurance from 2013 to 2016.⁷ Emond et al,⁴ using commercial insurance claims from 2014 to 2017, and Huang et al,⁵ using Medicare FFS and Medicare Advantage claims from 2016 to 2017, both compared ibrutinib with chemoimmunotherapy as first-line treatment and reported their association with cost and utilization across care settings. To our knowledge, only one study has quantified the attributable cost of CLL by comparing with a non-CLL control group.⁸ Lafeuille et al, using 1999-2007 Medicare data, showed that the average lifetime cost to Medicare for beneficiaries diagnosed with CLL was $39,509 higher than that for matched controls without cancer.⁸ Our results are not directly comparable with this previous study, which reported lifetime costs. Notably, our findings highlight differences in the magnitude and direction of cost accumulation across acute care and non–acute care settings.

This study has several strengths. First, we used recent data for a large, nationally representative sample of Medicare enrollees. We examined distributions of cost and HCRU for the full sample and across treatment groups. Second, instead of using a comorbidity index validated for postdischarge mortality, such as the Charlson Comorbidity Index,¹³ we used the HCC index developed by CMS to adjust for the impact of comorbidity on health care costs. Third, we conducted comprehensive analyses for predictors of cost and HCRU across multiple care settings. This allowed us to consider cost and HCRU estimates for the same sample of beneficiaries across multiple care settings (eg, ED, hospitalization, physician evaluation and management office visits) and identify subgroups who may be experiencing excess cost burdens.

LIMITATIONS

This study also has several important limitations. First, since we used claims data, we did not have confirmed diagnosis dates with which to establish the CLL treatment as first-line therapy. We also did not have clinical measures to further stratify patients by severity, whereas baseline clinical factors can determine the type of treatment received. Treatment is usually indicated based on criteria defined by the International Workshop on Chronic Lymphocytic Leukemia,¹⁴ such as massive adenopathy/splenomegaly, marrow failure, extranodal involvement, constitutional symptoms, and progressive lymphocytosis. The choice among available regimens is usually made based on patient-level factors (eg, fitness, risk of bleeding), tumor characteristics (eg, TP53 mutation, del17p, immunoglobulin heavy chain variable region gene mutation), and the treatment goals. It will be important to further examine the patterns of routine and urgent care reported above among individuals who may be asymptomatic or otherwise ineligible for the receipt of systemic therapy.

Second, we did not have long-term follow-up data to examine cost and HCRU over a longer follow-up period. The costs associated with receipt of CLL treatment (and thus overall CLL-related costs) are likely underestimated because many individuals diagnosed with CLL do not require treatment initially. Survival rates can vary widely, and many patients who do not receive treatment will have survival rates that are like those of the non-CLL population.^15,16 A longer follow-up period will be needed to capture the full range of utilization and costs associated with CLL treatment options. In the present study, we desired to strike a balance between maximizing the length of the follow-up period and preserving our ability to study treatments received in the postibrutinib period. Our study findings suggest that, even with a longer follow-up period, it will be important to consider costs across different care settings. Third, we did not have access to many potential confounders, such as fitness or general health status. We were, thus, unable to match on these and other variables, which can influence our cost and HCRU outcomes. Fourth, the study uses data for Medicare FFS beneficiaries and may not generalize to commercially insured, Medicare Advantage, or Medicaid populations.

Conclusions

The results of this study provide several directions for future research. First, future research should build on the quantification of cost and HCRU in this study to identify disease management strategies for sample subgroups who experience highest cost burden. Second, future study should use data sources that allow for a longer follow-up to examine changes in use patterns over time and across acute care and non–acute care settings. Third, the findings related to reduced use of routine care and increased use of acute care deserve further study as it relates to the older and nonmetropolitan residents diagnosed with CLL. Additional data to identify potential barriers (eg, limited transportation access, low socioeconomic status, other social determinants of health) will be important to identify avenues for intervention among individuals diagnosed with CLL.

This study represents a comprehensive characterization of cost and HCRU associated with CLL among beneficiaries enrolled in Medicare. The information presented can be used to guide clinicians, payers, and policymakers to outline programming strategies designed to lower the cost burden among vulnerable subgroups of beneficiaries.

Funding Statement

This study was supported by BeiGene, Ltd. Drs Onukwugha, Cooke, and Yared report research funding from BeiGene, Ltd related to the current work. Dr Yang is an employee of BeiGene, Ltd. Drs Liu and Tang were employees of BeiGene, Ltd at the time the study was conducted.