Adherence to iron chelation therapy and associated healthcare resource utilization and costs in Medicaid patients with sickle cell disease and thalassemia

Abstract

Background:

Sub-optimal patient adherence to iron chelation therapy (ICT) may impact patient outcomes and increase cost of care. This study evaluated the economic burden of ICT non-adherence in patients with sickle cell disease (SCD) or thalassemia.

Methods:

Patients with SCD or thalassemia were identified from six state Medicaid programs (1997–2013). Adherence was estimated using the medication possession ratio (MPR) of ≥0.80. All-cause and disease-specific resource utilization per-patient-per-month (PPPM) was assessed and compared between adherent and non-adherent patients using adjusted incidence rate ratios (aIRR). All-cause and disease-specific healthcare costs were computed using mean cost PPPM. Regression models adjusting for baseline characteristics were used to compare adherent and non-adherent patients.

Results:

A total of 728 eligible patients treated with ICT in the SCD cohort, 461 (63%) adherent, and 218 in the thalassemia cohort, 137 (63%) adherent, were included in this study. In SCD patients, the adjusted rate of all-cause outpatient visits PPPM was higher in adherent patients vs non-adherent patients (aIRR [95% CI]: 1.05 [1.01–1.08], p < 0.0001). Conversely, adherent patients incurred fewer all-cause inpatients visits (0.87 [0.81–0.94], p < 0.001) and ER visits (0.86 [0.78–0.93], p < 0.001). Similar trends were observed in SCD-related resource utilization rates and in thalassemia patients. Total all-cause costs were similar between adherent and non-adherent patients, but inpatient costs (adjusted cost difference = −$1530 PPPM, p = 0.0360) were lower in adherent patients.

Conclusion:

Patients adherent to ICT had less acute care need and lower inpatient costs than non-adherent patients, although they had more outpatient visits. Improved adherence may be linked to better disease monitoring and has the potential to avoid important downstream costs associated with acute care visits and reduce the financial burden on health programs and managed care plans treating SCD and thalassemia patients.

Keywords::

• Iron chelation therapy
• Sickle cell disease
• Thalassemia
• Deferoxamine
• Deferasirox
• Adherence
• Healthcare resource utilization
• Healthcare costs

Background

Patients with inherited red blood cell anemias such as sickle cell disease (SCD) and thalassemia require continuous and comprehensive patient care, including pain management, prophylactic antibiotics, hydroxyurea, and blood transfusions, depending on disease severity1–3. Frequent blood transfusions can lead to iron overload, which can cause significant impairment of vital organs such as the liver, the heart, and the endocrine glands4. Iron chelation therapy (ICT) can help remove excessive iron from the body and prevent cell damage, thereby reducing complications induced by iron overload5.

Two iron chelating agents, deferoxamine (DFO) and deferasirox (DFX), are indicated for the treatment of chronic iron overload in patients with SCD and thalassemia. For patients with iron overload due to thalassemia-related transfusions, deferiprone (Ferriprox; DFP) is also used as a second-line oral agent, whenever first-line chelation therapy is ineffective or untolerated6. Combination ICT is also used in clinical practice to optimize clinical outcomes and to reduce severe iron overload. A switch from one chelating agent to another may occur due to reasons including increasing iron overload, side-effects, and patient or clinician preference7.

While previous clinical trials have shown that DFX has similar efficacy in controlling iron overload compared to DFO in patients with SCD or thalassemia8–11, one major distinction between the two agents is their route of administration. Deferoxamine is usually administered by subcutaneous infusions for 8–12 h, 5–7 days per week, whereas DFX is an oral suspension administered once per day12^,13. Prior research has indicated that the regimen and practical aspects of drug administration, including long hours of subcutaneous infusions, a variety of equipment required, infusion-site pain, and adverse events, may be the most significant factors associated with non-adherence14. As such, it was hoped that DFX as an oral chelator would improve ICT adherence. In fact, recent data have shown that DFX has an estimated adherence rate, measured by a medication possession ratio [MPR] ≥0.80, that ranges from 44–76%, whereas that for DFO is ∼42%15^,16.

Studies on the impact of adherence to ICT on healthcare resource use (RU) and costs have been limited to date. Nonetheless, the lack of patient adherence to treatment in general has a major impact on both patient health and the healthcare system. Patients who under-use medication are more likely to experience worsening of their condition and emergence of new co-morbid conditions due to iron overload17. According to the New England Healthcare Institute, medication non-adherence costs the US healthcare system over $300 billion annually18^,19. Studies evaluating adherence in thalassemia patients have reported an inverse relationship between adherence and complications of iron overload, including cardiac disease, endocrinopathies, and death20–22. A study by Renfroe et al.23 indicated that thalassemia patients with complications of iron overload, such as heart disease or diabetes, incurred an increased annual cost of $20,000 per year. A study by Delea et al.24 computed the lifetime costs of inadequate adherence to DFO in patients with thalassemia based on a Markov model and estimated them to be $33,142 per patient.

The potential cost savings associated with ICT adherence may span beyond direct costs associated with reduced rates of complications. Since the introduction of the five-star quality rating system by the Affordable Care Act in 2012, the Centers for Medicare and Medicaid Services has been issuing financial bonuses to advantage organizations with high quality ratings25. Among the 53 measures used to calculate star quality ratings, the ability of health plans to manage chronic conditions is a key component. For this component, measures of adherence to medications are weighted heavily as they are considered intermediate outcomes, and have the potential impact to slow the growth of healthcare spending by lowering avoidable hospitalization rates. As a consequence, there is substantial incentive for payers to enhance medication adherence among patients with chronic conditions, including SCD and thalassemia.

Despite the potential impact of ICT adherence on cost savings, the literature on adherence to ICT and its association with healthcare RU and associated direct medical costs remains scarce. In order for policy-makers and payers to emphasize the need for enhanced adherence to ICT and potential cost savings, evidence on the current economic burden of ICT non-adherence is needed. This study aimed to evaluate the impact of ICT adherence on healthcare RU and costs among patients with SCD or thalassemia.

Methods

Data sources

Health insurance administrative claims data from Medicaid programs from six states were used: Florida (1997–2012), New Jersey (1997–2013), Missouri (1997–2013), Kansas (2001–2013), Iowa (199–2013), and Mississippi (2006–2013). The Medicaid claims database contained multi-year data elements to identify the insurance enrollment, eligibility information (demographic), medical outpatient, inpatient (e.g., service dates, diagnoses, and procedure dates), and pharmacy (e.g., drug costs, days of supply, and number of units) claims of over 22 million low income and disabled patients across the available states.

Study population

A longitudinal, retrospective cohort design, consisting of patients with at least one diagnosis of SCD (International Classification of Diseases, Ninth revision [ICD-9: 282.6]) or thalassemia (ICD-9: 282.4), and with at least two filled prescriptions for DFO or DFX, was used. Eligible patients were required to have at least 6 months of continuous insurance coverage prior to their index date, defined as the date of first ICT claim. The observation period spanned from the index date to the earliest of the following: (a) switch to another iron chelating agent, (b) recorded death, (c) end of Medicaid insurance eligibility, or (d) end of data availability. The treatment period of each patient was defined as the time from the patient’s index date to the earliest of: (a) date of the last ICT claim plus the days of supply of that last claim, (b) treatment discontinuation (i.e., the days of supply of the last refill before a gap of 6 or more months between two claims of ICT), or (c) end of data availability (i.e., recorded death or end of Medicaid insurance eligibility). Since patients treated with ICT were typically evaluated every 3–6 months for ferritin level, a patient with a gap of 6 months in his ICT refills was considered to have discontinued his ICT therapy.

Patient adherence to ICT was measured using the medication possession ratio (MPR), calculated as the sum of the days of supply of ICT divided by the number of days between the first fill and the last refill plus the days of supply of the last refill12^,13. Patients were stratified into two cohorts based on their adherence to ICT during their treatment period. Adherent patients were defined as patients with a MPR of at least 0.80, whereas non-adherent patients had a MPR of less than 0.80. While alternative adherence measures, including proportion of days covered and persistence, could be used, MPR was selected for this study because ICT is not often used as a lifelong therapy. Therefore, being able to account for the varying duration of therapy in the denominator for adherence assessment per the MPR definition would provide a more clinically meaningful and precise measure.

Outcome measures

All-cause, and disease-related (i.e., SCD- or thalassemia-related) health RU during the treatment period and the 1-year post-treatment period was assessed, and included inpatient stays, emergency room (ER) visits, and outpatient visits. Healthcare costs were similarly stratified into all-cause, and disease-related costs. Disease-related refers to events (i.e., RU and costs) that were associated with claims for SCD or thalassemia diagnosis, or SCD- or thalassemia-related complications. The list of SCD- or thalassemia-related complications and ICD-9-CM codes can be found in the Supplemental Table (S1). Resource utilization and costs were evaluated during the 1-year post-treatment period in addition to the treatment period in order to assess any changes in trend after treatment discontinuation. The post-treatment period was limited to 1 year as opposed to longer durations to reduce the impact of potential survival bias among patients with a longer observation period, and to limit the impact of a re-treatment with an ICT.

Statistical analysis

Descriptive analysis of the length of observation period, number of filled prescriptions for ICT, days of supply per ICT dispensing, days on therapy, and adherence status (MPR) were presented for the full ICT patient sample, and stratified by type of ICT received at index date (i.e., DFO or DFX). Statistical comparisons were conducted between patients who only used DFX vs DFO throughout their observation period (i.e., DFX-only users vs DFO-only users). Baseline demographic and clinical characteristics, including age, gender, race, state, type of ICT received at index date, Charlson Comorbidity Index (CCI), RU, and costs (outpatient, ER, inpatient, and pharmacy) were presented and compared between adherent patients and non-adherent patients. Mean, standard deviation (SD), and median were reported for continuous variables; proportions were reported for categorical variables. Statistical comparisons were conducted using t-tests or Wilcoxon-Mann-Whitney non-parametric tests for continuous variables, and Chi-square tests for categorical variables.

Healthcare RU was quantified in terms of the number of events per patient per month (PPPM). Incidence rates were calculated as the number of events (i.e., outpatient visits, ER visits, inpatient visits, and inpatient days) divided by patient-months of observation. Adjusted incidence rate ratios (aIRRs) were used to compare healthcare RU between adherent and non-adherent patients, adjusting for baseline characteristics that were statistically or clinically significant for SCD or thalassemia, specifically. For both of the analyses on SCD and thalassemia patients, the following baseline characteristics were controlled for: age, sex, state of residence, CCI, index year, type of ICT received at index date, number of inpatient visits, number of outpatient visits, hospitalizations costs, number of serum ferritin tests, and number of transfusions. Confidence intervals (95% CIs) of aIRRs were calculated based on the Poisson distribution.

Overall inpatient, outpatient, and pharmacy costs were estimated based on the paid amounts in the claims database. Cost data from the payer’s perspective were inflation-adjusted to 2013 US dollars (USD) based on the medical care component of the Consumer Price Index. All costs were computed using weighted PPPM costs, for which the observation period of each patient was used as the weight, such that normalized per-patient costs were calculated by dividing the costs incurred over the observation period by the person-time observed for each patient. Costs were compared between adherent and non-adherent patients using adjusted cost differences based on generalized linear models with Gamma distribution (log-link). Because healthcare cost data were right skewed with a substantial proportion of zero values, standard errors associated with parameter estimates from a linear regression may be biased. As such, statistical inference for adjusted cost differences in adherent and non-adherent patients was based on bootstrapped standard errors using 499 replications. Two-part multivariate regression models were performed to account for individuals with zero costs and to evaluate differences of costs between adherent and non-adherent patients while adjusting for a similar set of covariates as in the healthcare RU models.

Results

Baseline characteristics

A total of 728 patients with SCD and 218 patients with thalassemia meeting study eligibility criteria were identified. Table 1 shows the treatment patterns of all ICT users, stratified by type of ICT received at index date. Mean (SD) observation period was 43.8 (31.9) months for patients with SCD and 36.4 (27.4) months for patients with thalassemia. Based on a ≥0.80 threshold for MPR, 461 (63%) patients with SCD and 137 (63%) patients with thalassemia were considered adherent. Among the 728 patients with SCD, 67 (9%) were DFO-only users, 586 (80%) were DFX-only users, and 75 (11%) received both DFO and DFX. For patients with thalassemia, the drug utilization pattern is similar to that in patients with SCD. Among patients with SCD, DFX-only users were more likely to be adherent than DFO-only users (389 [66%] vs 29 [43%], p < 0.001). In the thalassemia population, the proportion of adherent patients among DFX-only users was not statistically different than that among DFO-only users (116 [63%] vs 12 [63%]; p = 0.9922). The mean (SD) overall duration of any ICT therapy was 16.8 (17.2) months for patients with SCD and 15.9 (16.9) months for patients with thalassemia.

Table 1. ICT dosing patterns and adherence to ICT among patients with SCD or thalassemia, stratified by treatment cohort.

	All ICT users	DFO users only [A]	DFX users only [B]	DFO and DFX switchers^a DFO	DFX	p-value [A] vs [B]
Patients with SCD, n	728	67	586	75	75
Observation period (months), Mean ± SD [Median]	43.8 ± 31.9 [37.5]	46.1 ± 40.9 [31.5]	38.1 ± 25.6 [35.2]	86.5 ± 35.1 [98.6]	86.5 ± 35.1 [98.6]	0.0245*
Days of supply per dispensing, Mean ± SD [Median]	26.6 ± 8.7 [30.0]	8.8 ± 7.3 [6.4]	30.2 ± 2.7 [30.0]	14.2 ± 9.9 [7.0]	30.2 ± 1.5 [30.0]	<0.001*
Months on therapy,^b Mean ± SD [Median]	16.8 ± 17.2 [11.5]	11.5 ± 12.2 [7.0]	16.2 ± 15.5 [11.4]	33.8 ± 25.0 [25.1]	39.9 ± 26.3 [35.6]	0.0173*
Medication possession ratio (MPR)
MPR,^c Mean ± SD [Median]	0.8 ± 0.2 [0.9]	0.7 ± 0.3 [0.8]	0.8 ± 0.2 [0.9]	0.5 ± 0.3 [0.6]	0.7 ± 0.2 [0.7]	<0.001*
Adherence status
Adherent patients (MPR ≥0.8)	461 (63.3)	29 (43.3)	389 (66.4)	24 (32.0)	29 (38.7)	<0.001*
Non-adherent patients	267 (36.7)	38 (56.7)	197 (33.6)	51 (68.0)	46 (61.3)
Patients with thalassemia, n	218	19	184	15	15
Observation period (months), Mean ± SD [Median]	36.4 ± 27.4 [31.5]	42.0 ± 38.5 [25.5]	33.5 ± 23.4 [29.6]	70.5 ± 38.1 [67.1]	70.5 ± 38.1 [67.1]	0.0243*
Days of supply per dispensing, Mean ± SD [Median]	28.1 ± 7.5 [30.0]	13.7 ± 11.2 [6.8]	30.3 ± 3.3 [30.0]	12.1 ± 11.4 [7.0]	30.9 ± 1.9 [30.0]	<0.001*
Months on therapy,^b Mean ± SD [Median]	15.9 ± 16.9 [10.0]	11.8 ± 9.3 [12.1]	15.1 ± 14.6 [8.9]	15.8 ± 16.0 [8.4]	39.2 ± 29.8 [34.6]	0.3450
Medication possession ratio (MPR)
MPR,^c Mean ± SD [Median]	0.8 ± 0.2 [0.9]	0.8 ± 0.2 [0.9]	0.8 ± 0.2 [0.9]	0.6 ± 0.3 [0.9]	0.8 ± 0.3 [0.9]	0.6938
Adherence status
Adherent patients (MPR ≥0.8)	137 (62.8)	12 (63.2)	116 (63.0)	3 (20.0)	10 (66.7)	0.9922
Non-adherent patients	81 (37.2)	7 (36.8)	68 (37.0)	12 (80.0)	5 (33.3)

DFO, Deferoxamine; DFX, Deferasirox.

^aDFO and DFX switchers were defined as patients who switched from DFO to DFX or vice versa. Medication possession ratio and MPR categories are presented for the period during which the patients used DFO or DFX. All patients contributed results to both the DFO and DFX columns.

^bDays on therapy was calculated as the difference between the day of the first ICT claim and the last day of supply from the last refill.

^cThe Medication Possession Ratio (MPR) was calculated as the sum of the days of supply divided by the number of days on therapy.

*Significant at the 5% 6level.

Table 2 presents the baseline demographic and clinical characteristics of adherent and non-adherent patients with SCD and thalassemia during the 180 day baseline period. Adherent and non-adherent SCD and thalassemia patients were similar at baseline. In patients with SCD, adherent patients were slightly younger (adherent patients: 18.1 years old vs non-adherent patients: 20.8 years old, p = 0.0106). About 55% of patients were female in both cohorts. The most common complications in both cohorts were pain (adherent patients: 42%; non-adherent patients: 48%), stroke (19% vs 17%), pneumonia (17% vs 18%), infection (18% vs 21%), and cardiomegaly (13 vs 14%). The CCI was comparable between the adherent and non-adherent cohorts (0.97 vs 1.06, p = 0.2698). Overall, adherent patients incurred significantly more transfusions (2.70 vs 2.30, p = 0.0041), serum ferritin tests (2.4 vs 1.8, p = 0.0017), and slightly, although not significantly, more outpatient visits PPPM (1.27 vs 1.11, p = 0.0976), but fewer inpatient visits PPPM (0.29 vs 0.38, p = 0.0047) and inpatient costs PPPM ($3025 vs $3339, p = 0.0350) at baseline.

Table 2. Baseline demographic and clinical characteristics of patients with SCD or thalassemia who are adherent vs non-adherent to ICT during the 180-day baseline period.

	Patients with SCD			Patients with thalassemia
	Adherent patients (n = 461)	Non-adherent patients (n = 267)	p-value^b	Adherent patients (n = 137)	Non-adherent patients (n = 81)	p-value^b
Demographic characteristics
Female, n (%)	253 (54.9)	147 (55.1)	0.9634	80 (58.4)	51 (63.0)	0.5056
Age at index date, mean ± SD [median]	18.1 ± 12.8 [14.0]	20.8 ± 12.7 [18.0]	0.0004*	20.9 ± 14.1 [19.0]	25.8 ± 14.4 [24.0]	0.0106*
Race, n (%)
White	7 (1.5)	8 (3.0)	0.1762	5 (3.6)	10 (12.3)	0.0142*
Black	284 (61.6)	198 (74.2)	0.0006*	81 (59.1)	53 (65.4)	0.3551
Hispanic	6 (1.3)	6 (2.2)	0.3342	2 (1.5)	1 (1.2)	0.8903
Other	20 (4.3)	17 (6.4)	0.2298	10 (7.3)	6 (7.4)	0.9764
Unknown	144 (31.2)	38 (14.2)	<0.0001*	39 (28.5)	11 (13.6)	0.0115*
State, n (%)
Florida	229 (49.7)	114 (42.7)	0.0691	58 (42.3)	30 (37.0)	0.4410
Iowa	15 (3.3)	3 (1.1)	0.0745	13 (9.5)	2 (2.5)	0.0479*
Kansas	15 (3.3)	2 (0.7)	0.0310*	10 (7.3)	2 (2.5)	0.1308
Missouri	39 (8.5)	38 (14.2)	0.0147*	11 (8.0)	12 (14.8)	0.1150
New Jersey	53 (11.5)	50 (18.7)	0.0070*	23 (16.8)	17 (21.0)	0.4389
Mississippi	110 (23.9)	60 (22.5)	0.6694	22 (16.1)	18 (22.2)	0.2559
Type of ICT at index date
Deferoxamine	66 (14.3)	68 (25.5)	0.0002*	19 (13.9)	12 (14.8)	0.8467
Deferasirox	395 (85.7)	199 (74.5)	0.0002*	118 (86.1)	69 (85.2)	0.8467
Number of serum ferritin tests, mean ± SD [median]	2.4 ± 2.3 [2.0]	1.8 ± 1.9 [1.0]	0.0017*	2.3 ± 2.4 [1.0]	2.3 ± 2.4 [2.0]	0.8244
Hydroxyurea used, n (%)	91 (19.7)	62 (23.2)	0.2665	37 (27.0)	26 (32.1)	0.4229
Number of transfusions during the baseline period	2.7 ± 2.8 [2.0]	2.3 ± 3.6 [1.0]	0.0041*	2.4 ± 3.0 [1.0]	1.6 ± 2.4 [0.0]	0.0195*
Resource utilization (PPPM),^c mean ± SD [median]
Outpatient visits	1.27 ± 1.23 [1.17]	1.11 ± 1.13 [0.83]	0.0976	1.31 ± 1.27 [1.17]	1.32 ± 1.67 [0.83]	0.4752
Emergency room visits	0.23 ± 0.57 [0.00]	0.26 ± 0.60 [0.00]	0.1495	0.31 ± 0.71 [0.00]	0.42 ± 0.93 [0.17]	0.3310
Inpatient visits	0.29 ± 0.42 [0.17]	0.38 ± 0.47 [0.17]	0.0047*	0.40 ± 0.52 [0.17]	0.42 ± 0.53 [0.33]	0.4543
Healthcare costs (PPPM),^c mean ± SD [median]
Outpatient	$601 ± $823 [$309]	$491 ± $795 [$237]	0.0972	$575 ± $889 [$264]	$387 ± $550 [$161]	0.2454
Emergency room	$23 ± $61 [$0]	$32 ± $79 [$0]	0.2504	$32 ± $86 [$0]	$46 ± $101 [$0]	0.2299
Inpatient	$3025 ± $7352 [$382]	$3339 ± $6527 [$935]	0.0350*	$4245 ± $9259 [$624]	$3832 ± $6389 [$1090]	0.3517
Pharmacy	$817 ± $857 [$623]	$922 ± $1112 [$781]	0.0709	$803 ± $573 [$766]	$1019 ± $1158 [$837]	0.1299
Total costs	$4467 ± $7664 [$2118]	$4784 ± $6811 [$2418]	0.1766	$5655 ± $9390 [$2375]	$5284 ± $7047 [$3130]	0.5258
Observation period (days), Mean ± SD [Median]	1240.6 ± 955.9 [1031.0]	1439.3 ± 949.4 [1303.0]	0.0019*	1114.1 ± 839.7 [925.0]	1247.4 ± 789.6 [1144.0]	0.1305

ICT, iron chelation therapy; SCD, sickle cell disease; SD, standard deviation; PPPM, per patient per month; AIDS, acquired immune deficiency syndrome.

^aAdherence was defined as a medication possession ratio ≥0.80.

^bp-value was estimated using Chi-square tests for categorical variables and Wilcoxon rank-sum tests for continuous variables.

^cPPPM stands for per patient per month. PPPM was calculated by dividing the costs incurred over the observation period by the person-time in months in baseline observed for each patient. The same approach was taken for resource utilization.

*Significant at the 5% level.

The mean age of adherent patients with thalassemia was also lower when compared to non-adherent patients (20.9 vs 25.8 years old, p = 0.0106). At baseline, while the distribution of complications was similar between adherent and non-adherent patients (results not shown), the most common thalassemia-related complications were anemia (97% and 99%), infections (23% and 22%), cardiomyopathy (15% and 22%), osteoporosis (15% and 22%), and stroke (10% and 9%) for adherent and non-adherent patients, respectively. Similarly, CCI was not significantly different between the two cohorts (0.88 vs 0.94, p = 0.6999). Adherent patients received more transfusions (2.45 vs 1.63, p = 0.0195), but there were no significant differences in resource used and healthcare costs between the two cohorts at baseline.

Resource utilization

Figure 1 shows incidence rates and aIRR for patients with SCD. Adherent patients had significantly more all-cause outpatient visits during both the treatment period and the 1-year post-treatment period. On average, adherent patients had 1.43 all-cause outpatient visits PPPM compared to 1.29 outpatient visits PPPM for non-adherent patients (aIRR [95% CI] = 1.05 [1.01–1.08], p < 0.0001) during the treatment period, and 1.23 outpatient visits PPPM compared to 1.04 outpatient visits PPPM (aIRR [95% CI] = 1.09 [1.04–1.15]) during the 1-year post-treatment period. During the treatment period, adherent patients incurred fewer all-cause and SCD-related inpatient visits (aIRR [95% CI]; all-cause = 0.87 [0.81–0.94], p < 0.001; SCD-related = 0.90 [0.83–0.97], p < 0.001) and ER visits (all-cause = 0.86 [0.78–0.93], p < 0.001; SCD-related: 0.83 [0.75–0.92], p < 0.001). Adherent patients experienced an average of 1.22 inpatient days PPPM during the treatment period compared to 2.16 days PPPM for non-adherent patients (aIRR [95% CI] = 0.71 [0.69–0.73], p < 0.001). During the 1-year post-treatment period, adherent patients continued to incur fewer inpatient days, but slightly more ER visits, 0.32 in adherent patients vs 0.28 visits in non-adherent patients (aIRR [95% CI] = 1.25 [1.13–1.38], p < 0.001).

Figure 1. Adjusted incidence rates of all-cause and SCD-related resource utilization in adherent vs non-adherent to ICT patients with SCD.

Trends were similar in patients with thalassemia (Figure 2). More outpatient visits and fewer ER visits, inpatient visits, and inpatients days were observed for adherent patients compared to non-adherent patients during the treatment period.

Figure 2. Adjusted incidence rates of all-cause and thalassemia-related resource utilization in adherent vs non-adherent to ICT patients with thalassemia.

Healthcare costs

Table 3 shows average PPPM healthcare costs and adjusted cost differences between adherent and non-adherent patients in the SCD and thalassemia cohorts. In patients with SCD, the mean (SD) all-cause cost PPPM for adherent patients was $6281 ($4749) compared to $6934 ($8431) for non-adherent patients during the treatment period. After adjusting for covariates, the adjusted cost difference was −$356 (p = 0.5680) between adherent patients and non-adherent patients. This difference was primarily driven by lower inpatient costs (adjusted cost difference = −$1530, p = 0.0360), but partially offset by higher pharmacy costs (adjusted cost difference = $863, p < 0.0001). A similar trend was observed among SCD-related costs. During the 1-year post-treatment period, adherent patients incurred significantly less all-cause inpatient costs (adjusted cost difference = −$945, p = 0.0440) compared to non-adherent patients, without any significant difference in pharmacy, outpatient, or ER costs. In SCD-related costs, a similar trend was observed for inpatient costs (adjusted cost difference = −$643, p = 0.1160), but significantly higher outpatient costs (adjusted cost difference = $197, p = 0.0440). For patients with thalassemia, similar trends were observed during the treatment period and during the 1-year post-treatment period. The overall medical costs were lower, driven by lower inpatient costs for adherent patients.

Table 3. Unadjusted and adjusted mean difference in all-cause and disease-related costs in adherent vs non-adherent to ICT patients with SCD or thalassemia.

	Adherent patients [A]		Non-adherent patients [B]		Unadjusted cost difference [95% CI] [A] − [B]	p-value	Adjusted cost difference^a,b [95% CI]	p-value
Patient with SCD, n	461		267
Treatment period^c
All-cause costs, (PPPM),^d USD, mean [SD]	$6281	[$4749]	$6934	[$8431]	−653 (−1872–311)	0.2440	−356 (−1208–623)	0.5680
Pharmacy	$3875	[$2065]	$3090	[$1908]	785 (389–1,170)	<0.001*	863 (519–1227)	<0.001*
Pharmacy without ICT	$191	[$548]	$339	[$1089]	−148 (−335–3)	0.0600	−1278 (−27,738–32)	0.0720
Outpatient	$707	[$844]	$583	[$816]	124 (−58–279)	0.1560	9 (−172–183)	0.9040
Emergency room	$18	[$43]	$27	[$69]	−9 (−17–−0)	0.0480*	−12 (−83–−2)	0.0360*
Inpatient	$1681	[$4009]	$3234	[$8069]	−1553 (−2725–−647)	0.0120*	−1530 (−10,426–−202)	0.0360*
SCD-related cost, (PPPM),^d,e USD, mean [SD]	$2106	[$4071]	$3397	[$7707]	−1292 (−2469–−401)	0.0120*	−1267 (−3,691–−38)	0.0520
Outpatient	$519	[$729]	$370	[$665]	149 (−13–288)	0.0760	89 (−96–267)	0.3000
Emergency room	$11	[$30]	$18	[$51]	−7 (−13–−1)	0.0360*	−9 (−52–2)	0.0240*
Inpatient	$1576	[$3944]	$3010	[$7690]	1433 (−2554–−566)	0.0120*	−1297 (−14,788–−157)	0.0360*
First 12 months after the end of ICT treatment^f
All-cause costs, (PPPM),^d USD, mean [SD]	$3683	[$7703]	$4697	[$7482]	−1014 (−2181–119)	0.0720	−1213 (−3181–130)	0.0840
Pharmacy	$717	[$2641]	$681	[$1076]	36 (−169–226)	0.6800	94 (−185–402)	0.4880
Pharmacy without ICT	$302	[$688]	$379	[$848]	−77 (−228–42)	0.2280	−24 (−218–157)	0.8920
Outpatient	$514	[$891]	$430	[$740]	83 (−41–224)	0.2080	106 (−59–310)	0.2040
Emergency room	$35	[$107]	$37	[$105]	−2 (−20–16)	0.9000	−5 (−30–8)	0.4480
Inpatient	$2417	[$6541]	$3549	[$7028]	−1132 (−2209–−100)	0.0360*	−945 (−2015–−64)	0.0440*
SCD-related cost, (PPPM),^d,e USD, mean [SD]	$2603	[$6385]	$3420	[$6425]	−817 (−1,759–189)	0.1120	−772 (−2136–275)	0.1880
Outpatient	$363	[$838]	$254	[$543]	109 (4–213)	0.0480*	197 (7–518)	0.0440*
Emergency room	$23	[$82]	$24	[$67]	−1 (−13–12)	0.9200	−3 (−16–7)	0.4920
Inpatient	$2217	[$6160]	$3142	[$6359]	−925 (−1851–42)	0.0680	–643 (–1559–165)	0.1160
Patient with thalassemia, n	137		81
Treatment period^c
All-cause costs, (PPPM),^d USD, mean [SD]	$6547	[$6051]	$7461	[$8942]	–914 (–3209–1064)	0.4600	–234 (–2259–1981)	0.8720
Pharmacy	$4019	[$2222]	$3405	[$2148]	614 (–246–1353)	0.1520	1506 (822–2348)	<0.001*
Pharmacy without ICT	$189	[$457]	$566	[$1705]	–377 (–1092–21)	0.1200	–234 (–1412–130)	0.2000
Outpatient	$622	[$749]	$400	[$477]	222 (–4–460)	0.0520	92 (–328–360)	0.6880
Emergency room	$17	[$39]	$31	[$85]	–14 (–35–2)	0.0880	–16 (–77–7)	0.2000
Inpatient	$1890	[$5211]	$3626	[$8534]	–1736 (–3553 –−200)	0.0400*	–2029 (–8531–485)	0.1200
Thalassemia-related cost, (PPPM),^d,i USD, mean [SD]	$2407	[$5337]	$3896	[$8526]	–1489 (–3264–79)	0.0760	–1922 (–6075 –−15)	0.0560
Outpatient	$572	[$721]	$350	[$418]	222 (8–432)	0.0400*	113 (–183–359)	0.5040
Emergency room	$10	[$31]	$22	[$64]	–12 (–28–0)	0.0640	–17 (–71–2)	0.0880
Inpatient	$1825	[$5207]	$3523	[$8505]	–1698 (–3488 –−164)	0.0440*	–2504 (–13606 –−12)	0.0520
First 12 months after the end of ICT treatment^f
All-cause costs, (PPPM),^d USD, mean [SD]	$3542	[$6123]	$4153	[$6898]	–611 (–2620–1152)	0.4920	–1353 (–6620–616)	0.1600
Pharmacy	$460	[$883]	$571	[$3065]	–112 (–386–173)	0.4120	–361 (–1352–79)	0.0960
Pharmacy without ICT	$276	[$682]	$315	[$467]	–38 (–199–137)	0.7200	–188 (–756–31)	0.0720
Outpatient	$481	[$792]	$282	[$524]	199 (4–423)	0.0480*	187 (–207–707)	0.3200
Emergency room	$48	[$155]	$57	[$160]	–9 (–62–44)	0.7800	7 (–76–141)	0.9440
Inpatient	$2553	[$5715]	$3243	[$6277]	–690 (–2713–1020)	0.4360	–992 (–4401–607)	0.2240
Thalassemia-related cost, (PPPM),^d, USD, mean [SD]	$2930	[$5934]	$3334	[$6287]	–404 (–2409–1351)	0.6640	–117 (–3879–2550)	0.8680
Outpatient	$439	[$764]	$219	[$397]	219 (47–428)	0.0120*	300 (–33–858)	0.0960
Emergency room	$37	[$140]	$36	[$105]	1 (–38–44)	0.9160	13 (–49–177)	0.7400
Inpatient	$2454	[$5706]	$3078	[$6292]	–625 (–2665–1098)	0.4680	–997 (–4969–732)	0.2880

ICT, iron chelation therapy; SCD, sickle cell disease.

^aAdjusted costs difference was estimated using a two-part regression model for hospitalization and emergency-room costs with GLM model with a Gamma distribution and log link and a GLM model with a Gamma distribution and log link for outpatient costs, pharmacy costs, and overall costs. Controlling for age, sex, state of residence, CCI, index year, agent received at baseline, number of hospitalizations, number of outpatient visits, hospitalization costs, number of serum ferritin tests, and number of transfusions in the SCD population and for age, sex, state of residence, CCI, index year, agent received at baseline, hospitalizations costs, and number of transfusions in the thalassemia population.

^bStatistical differences between groups, as well as 95% confidence intervals [95% CI], were calculated using a non-parametric permutation test with 499 replications.

^cFrom initiation of ICT therapy until end of therapy or discontinuation (defined as a 6-month gap).

^dPPPM stands for per patient per month. PPPM was calculated by dividing the costs incurred over the observation period by the person-time in months in baseline observed for each patient.

^eSCD-related events were identified based on claims with a diagnosis of SCD (ICD-9: 282.6x) or a diagnosis of SCD-related complication, including pain, infection, renal disease/failure, COPD, pneumonia, asthma, acute chest syndrome, congestive heart failure, cardiomegaly, arrhythmia, cardiomyopathy, pulmonary hypertension, stroke, leg ulcers, and avascular necrosis–hip and shoulder.

^fFrom end of therapy or discontinuation (defined as a 6-month gap) until 1 year after end of ICT therapy or end of follow-up. Thalassemia-related events were identified based on claims with a diagnosis of thalassemia (ICD-9: 282.4x) or with a diagnosis of thalassemia-related complication, including anemia, cardiomegaly, cholecystitis, cholelithiasis, cirrhosis, congestive heart failure, coronary heart disease, diabetes mellitus, growth retardation, hepatitis, hepatocellular carcinoma, hepatomegaly, hypothyroidism, infections, intracranial arterial stenosis, leg ulcers, myocardial infarction, osteoporosis, pericarditis, pregnancy-related complications, pulmonary hypertension, splenomegaly, stroke.

Discussion

Non-adherence to medication costs ∼300 billion US dollars in avoidable medical spending annually19, which result from costs associated with avoidable complications arising from inadequate medication, increased rates of subsequent hospitalization, progression of controllable disease resulting in increased use of outpatient visits as well as urgent care, avoidable pharmacy costs, and diagnostic testing and procedures26. In recognition of the economic burden associated with medication non-adherence, the star rating system, as introduced by the Affordable Care Act, places a heavy emphasis on medication adherence, as it is believed that enhanced adherence could lead to better long-term health outcomes, and subsequent cost savings. Our results provide evidence for the relationship of adherence and healthcare resource utilization and costs in ICT among patients with SCD and thalassemia.

In the current study, adjusting for baseline differences between the two cohorts, adherent patients had significantly higher all-cause and disease-related rates of outpatient visits during the treatment period. Conversely, inpatient visits, inpatient days, and ER visits were significantly lower in adherent patients compared to non-adherent patients. In particular, current results indicated that the association between adherence and reduced inpatient and ER visits was not restricted to the treatment period, but had a longer term effect into 1 year post-treatment. In a study on SCD patients also based on Medicaid data, Blinder et al.4, assessing resource utilization among patients with ICT vs no ICT, found similar results on resource utilization, noting that ICT patients had lower inpatient days and ER visits. Together, such evidence indicates that adherence to ICT among patients who receive frequent blood transfusions may have a substantial impact on reduced acute healthcare use.

Multiple factors could explain the lower need for acute care in adherent patients. One reason may be that, with the more frequent outpatient visits, adherent patients were more closely managed, thereby rendering them more likely to receive earlier or more regular interventions to reduce the occurrence and severity of complications. As a consequence, adherent patients were less likely to receive acute care. In fact, while the lower rate of inpatient visits among adherent patients in our study suggests fewer complications in this patient cohort, the lower mean number of inpatient days between adherent patients and non-adherent patients further supports the notion that adherent patients may be hospitalized for less severe conditions. When looking at diagnoses incurred during the hospitalizations, non-adherent patients were more likely to have diagnoses related to SCD and thalassemia complications (pain, infections) as well as ICT complications (e.g., iron overloads, cardiac disease). These results suggest that patients non-adherent to ICT were likely less closely monitored and adherent to other treatment.

During the 1-year post-treatment period, adherent patients continued to incur fewer inpatient days, but slightly more ER visits and outpatient visits. The higher rate of ER visits post-ICT in adherent patients is in line with previous studies that have shown that SCD patients transitioning to adulthood rely more on ED for their care27.

In terms of healthcare costs, the mean differences in all-cause and disease-specific overall costs were not significantly different between adherent patients compared to non-adherent patients, despite the fact that the overall pharmacy costs were higher among adherent patients. However, such an observation was expected due to the increased use of ICT and its associated cost in adherent patients. Previous studies have indicated that the costs of ICT were substantial, which often exceeded the costs of the ICT agents themselves28. It has been estimated that ICT-related charges among SCD patients ranged from $1342.50–$3207.91 PPPM, in the US, including $773.83 for infusion for DFO and $2033.66–$4424.58 for DFX29. In our study, when non-ICT pharmacy costs were assessed, the cost difference between adherent and non-adherent patients was found to be −$1278 in SCD patients and −$234 in thalassemia patients, both of which were non-significant, suggesting that medication costs not related to ICT for both patient cohorts were in fact similar. As for inpatient and ER costs, both were lower in adherent patients. For instance, the inpatient cost in SCD patients was significantly less in adherent patients compared to non-adherent patients. After adjusting for covariates, the PPPM mean difference was $1530 between the two patient cohorts, which further supports the notion that hospitalizations incurred by adherent patients may be less severe in nature. Together, results on the resource utilization and costs in adherent vs non-adherent patients suggest that, while adherence increases prescription costs due to the increased use of ICT, such costs were offset by the main cost drivers from inpatient and urgent care, as patients may be better monitored and have fewer complications and better health outcomes.

Among patients with SCD and thalassemia, the majority of patients using DFX were adherent. Our adherence estimates based on MPR are comparable with estimates reported in previous studies10^,16^,30^,31. For example, studies conducted on patients with thalassemia or SCD showed that the proportion of patients adherent to DFO ranged between 62–64%10^,30. Similarly, a claims data study on patients with SCD found that 57% of patients treated with DFX were adherent16. According to an international survey conducted among thalassemia patients using DFO, nearly half of the patients reported missing at least one dose in the last month. After patients’ beliefs and feelings about the treatment, adverse effects of infusing DFO was the most frequently reported reason for missing at least a dose32^,33. In other studies on thalassemia patients, the most commonly reported adverse events were related to the infusion itself. For instance, thalassemia patients often reported stopping DFO due to severe local reactions. As many as 85% and 74% of patients have indicated site soreness and site irritation34. Such findings suggest that the mode of ICT administration has an impact on patients’ adherence to ICT.

The present study was subject to a number of limitations inherent to claims data. First, the computation of MPR included days when patients were hospitalized. Given that, during hospitalization, patients would not have consumed any days of supply of ICT, patients who were ever hospitalized would have lower MPR by design. Patients who would have been adherent otherwise would have been misclassified as non-adherent in our analyses. Second, we were unable to clearly isolate the impact of ICT adherence on RU and costs. Adherence to IT may encompass overall adherence to all forms of therapy and monitoring. Hence, the outcomes associated with ICT adherence may be confounded by the impact of overall adherence. Third, severity of disease may vary across patients. The claims database consisted of only information on procedure codes or diagnostic codes, such that information on the severity of disease was not available. Similarly, we did not have available lab information, such as the number of red blood cell units transfused, and details on serum ferritin tests. As such, the severity of disease associated with RU and the iron overload requirement between adherent patients and non-adherent patients could not be assessed and adjusted for in our analyses. However, analyses on baseline clinical characteristics indicated that adherent and non-adherent patients were generally not significantly different. Some difference was seen in age at baseline, which can partially explain the difference in adherence and resource utilization, since older patients tend to incur more healthcare costs. For characteristics that were significant at baseline, our analyses accounting for their potential impacts on outcomes. Fourth, intrinsic to claims databases, our database may have medical coding errors or data omissions or other misclassifications; however, these errors may happen non-differentially to both the adherent and non-adherent patient cohorts, thus the misclassification was likely to bias our findings towards the null. Lastly, we were not able to capture the actual utilization of medication, as claims data only provide filled prescriptions, which were used to compute the MPR in our study. As such, the MPR as assessed may be an over-estimation of actual adherence, as patients may not actually use ICT after medication dispensing. This may be especially true for DFX, which is an oral chelating agent.

In summary, the current study analyzed a large cohort of patients with SCD or thalassemia, who were treated with either DFO or DFX; adherent patients to ICT treatment were found to have less acute care need and incurred lower healthcare costs than non-adherent patients, with the exception of outpatient costs. These findings support the importance of patients’ adherence to medication in the context of a changing healthcare management setting with an emphasis on medication adherence and quality of care under the Affordable Care Act. This study’s findings indicate that patient adherence to preventive measures and treatments such as ICT is critical, for improved adherence may be linked to better disease monitoring, and has the potential to avoid important downstream costs associated with acute care visits and reduce the financial burden on health programs and managed care plans treating SCD and thalassemia patients. Given that the current overall observed adherence to ICT is still less than 70%, enhanced adherence should be emphasized in disease management programs for patients with SCD and thalassemia.

Transparency

Declaration of funding

Funding for this study was provided by Novartis.

Declaration of financial/other relationships

CP is an employee of Novartis Oncology. MS is an employee and holds stock in Novartis and YQ also owns Novartis stock. AVR and PAG have received consultancy fees/speakers bureau from Novartis. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.