Suboptimal adherence to prescribed daily growth hormone regimen among medicaid beneficiaries in the United States

Abstract

Objective

The objective of this retrospective cohort study was to describe the adherence and discontinuation patterns of somatropin over 3 years among children with pGHD insured by Medicaid across the United States.

Methods

Eligible children were aged ≥3 and <16 years with Medicaid coverage, diagnosed with pGHD, and had ≥2 new prescriptions for somatropin between 1 July 2014 and 31 December 2018. Four non-exclusive patient cohorts were constructed (≥3, 12, 24, and 36 months of continuous enrollment after initial prescription). Suboptimal adherence was defined as medication possession ratio <0.80, and discontinuation as a gap of >60 days between somatropin fills. Logistic and proportional hazards regression methods were used to estimate odds of suboptimal adherence and time to discontinuation, respectively.

Results

In the 12-month cohort (n = 3623), mean age was 10.5 ± 3.2 years, 70.8% were male, 44.4% White, 29.1% Hispanic, 7.1% Black, and 1.7% Asian. At months 12, 24, and 36, the proportion with suboptimal adherence was 40.9, 50.4, 54.4%, respectively, and 49.2% of patients with ≥3 months of follow-up discontinued therapy. At 12 months, lower age and race/ethnicity (Black vs. White referent) had greater odds of suboptimal adherence. Discontinuation was associated with Black (vs. White referent) race and geographic region.

Conclusions

Sociodemographic characteristics may be risk factors for suboptimal adherence and/or discontinuation of prescribed somatropin therapy. Improving GH regimen adherence among this at-risk population, and specifically among subgroups at highest risk, is warranted to improve clinical outcomes.

Keywords:

• Adherence
• daily growth hormone
• real world
• discontinuation

Introduction

Pediatric growth hormone deficiency (pGHD) is treated with daily subcutaneous injections of recombinant human growth hormone (r-hGH), known as somatropin. The goal of somatropin therapy is to improve growth in childhood, to achieve final adult height to within 2 standard deviations of the population mean, and to prevent or improve metabolic impairments seen with GH deficiency. Worldwide, the prevalence of pGHD in different regions has been estimated at approximately 1 in 3500–10,000 children, and the Utah Growth Study estimated the prevalence of pGHD in the US as at least 1 in 3480^1–4. Some estimates of pGHD prevalence may be influenced by differential treatment access, particularly among subgroups that have barriers to care, such as those with lower socioeconomic status or racial/ethnic minorities. Thus, there is limited data related to pGHD prevalence among racial/ethnic subgroups, and those with barriers to healthcare may have historically resulted in an underestimated prevalence of pGHD. Data from the National Health and Nutrition Examination Survey (NHANES data) has suggested different growth trajectories for non-Hispanic White and non-Hispanic Black children; Black children’s growth outpaced that of White children during childhood, but growth of White children exceeded the growth of Black children during adolescence⁵. White children have been overrepresented among pGHD registries relative to the population, and Black children have been underrepresented^6,7. Other research has suggested that racial/ethnic biases may play a role in the underdiagnosis of GHD and less frequent referral to treatment among Black as compared to White children⁷.

Adherence to somatropin has been demonstrated to be suboptimal in real-world studies^8–10. The requirement for daily injections is burdensome and may contribute to suboptimal adherence. An estimated 2 in 3 children/adolescents with pGHD may miss more than one injection per week⁸. Suboptimal adherence to somatropin can be associated with poorer growth^8–10. Factors influencing adherence may include the following: (1) therapy related barriers (e.g. lack of disease awareness and benefits of treatment, improper use of device, injection pain and schedule burden), (2) payer hurdles (e.g. frequent health plan-required authorizations), (3) emotional or social barriers (e.g. peer pressure, forgetting doses), (4) the physician/patient relationship, and (5) socioeconomic barriers (e.g. chaotic households)¹¹. Strategies to improve adherence to prescribed somatropin therapy include injection devices which can track adherence, effective partnership between patient and physician, and education regarding the benefits of treatment. New long-acting growth hormones (LAGHs) are in development; a less frequent injection schedule may be less burdensome to patients and caregivers, facilitate better adherence, and long-term, may facilitate improved growth and improved quality of life¹². Recent data indicates that both children and adults, treated with daily r-hGH injections, indicated a preference for a less frequent injection regimen¹³.

Evaluations of real-world GH adherence in pGHD are needed, particularly among populations of lower socioeconomic status or racial/ethnic subgroups in the US, where barriers to care continue to persist. Data from a large dataset, such as that contained within a US Medicaid Claims Database, offers an opportunity to understand current rates of suboptimal adherence with current therapy (i.e. daily injections of somatropin) among Medicaid recipients. The objectives of this study were to describe medication adherence and discontinuation among children diagnosed with pGHD, who were treated with daily somatropin injections over 3 years, from a US Medicaid population. Secondary objectives were to explore barriers to somatropin adherence, by characterizing the demographic and clinical profiles of a US Medicaid population with pGHD who have good somatropin adherence compared to those with suboptimal adherence.

Methods

Data source

This retrospective cohort study employed the Centers for Medicare and Medicaid Services (CMS) Medicaid database accessed by Clarify Health through the Virtual Research Data Center. The CMS Medicaid database contains 100% of the state-level Medicaid claims submitted to CMS. This database includes information on patient demographic and enrollment data, as well as inpatient and outpatient services and prescription drug claims, for more than 60 million people covered by Medicaid fee-for-service (FFS) or Managed Medicaid insurance. Approximately 24% of the US population is covered by Medicaid, and of this, Medicaid FFS and Managed Medicaid represent roughly 17% and 83%, respectively (approximately 4% and 20% of the US population, respectively). Individual states submit Medicaid claims to the Transformed Medical Statistical Information System (T-MSIS). CMS aggregates the T-MSIS data into T-MSIS Analytic Files (TAF), which are structured specifically for research and analytical purposes. Since 2016, the CMS Medicaid database represents over 90% of the Medicaid FFS and Managed Medicaid market in the US. However, data in the T-MSIS vary regionally due to differences in the Medicaid program at the state level. Data coverage is lower in 2014 and 2015 as states transitioned to the T-MSIS file structure at different times; during this time, only 18 states submitted T-MSIS data in 2014 and 25 sites in 2015, respectively. Since 2016, all 50 states have been participating in TAF. CMS provides assessments of TAF claims at the individual state level to ensure that the data are usable for research and analysis. In the CMS Medicaid database, diagnoses, procedures and prescription drugs are coded using the International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification (ICD-9/10-CM), the Current Procedural Terminology, Fourth Edition (CPT-4) and National Drug Code (NDC), respectively.

Study population

Patients were between 3 and 16 years of age who had at least 2 prescription claims for somatropin between 1 July 2014 and 31 December 2018. This was the most recent data available from the CMS Medicaid database at the time of analysis. The first somatropin claim during this study identification window denotes each individual’s “index date.” Eligible patients were further required to have at least 1 diagnosis code for GHD [ICD-9 CM (253.3, 253.4) or ICD-10 CM (E23.0, E23.1)] in the period 6 months prior to or on the index date (“baseline period”). Patients were further required to have continuous medical and pharmacy eligibility during the baseline period and during the variable study follow-up period. Patients were excluded for 1 or more of the following reasons: (1) prescription claims for somatropin during the baseline period (prior to initiation of new somatropin therapy); (2) diagnostic claim of psychosocial dwarfism (ICD-9-CM 259.4 or ICD-10-CM E34.3); (3) other identified causes of short stature, such as celiac disease ((ICD-9-CM 579.0 or ICD-10-CM K90.0), uncontrolled primary hypothyroidism (ICD-9-CM 244.9 or ICD-10-CM E03.9), or rickets (ICD-9-CM 268.0 or ICD-10-CM E55.0).

Four non-exclusive cohorts were constructed based on available follow-up time after somatropin initiation. Patients were selected based on follow-up time of up to 48 months, but insufficient sample size was obtained for the 48-month follow-up patient group (n < 11). Thus, cohorts were followed for up to 36 months post-index date and defined by follow-up ≥3 months, ≥12 months, ≥24 months, and ≥36 months. Discontinuation of somatropin therapy was evaluated in the 3-month follow-up cohort, which included patients with at least 3 months and up to 36 months of follow-up, while adherence was evaluated in the 12, 24 and 36-month follow-up cohorts.

Study measures

Primary outcome measures for this study were adherence to, and discontinuation of, somatropin therapy. Adherence was measured using the medication possession ratio (MPR), which was calculated for cohorts with 12, 24, and 36 months of follow up based on continuous eligibility during each follow-up period. MPR was calculated as total days’ supply of somatropin during the specific (12, 24, 36 months) follow-up period, divided by the date of the last somatropin prescription during the specific follow-up period plus days’ supply of that prescription, minus the date of the index somatropin prescription. Patients were classified as having good adherence (MPR ≥0.80) or suboptimal adherence (MPR <0.80), commonly used as a threshold for adherence studies, which equates to about 1 missed injection/week; there is clinical rationale for this threshold, as prior research has supported that this approximate adherence threshold is associated with poorer growth outcomes^8,14. In addition, sensitivity analyses were performed among cohorts of patients with 24 and 36 months of eligibility after the index date. Discontinuation was defined as the first observation of a gap of >60 days between prescription fills for somatropin during the follow-up period. The date of the first observed gap was defined as discontinuation date plus days’ supply of the prescription preceding the gap, and sensitivity analyses were conducted using alternate gap criteria (15-, 30-, 45-, and 90-day gaps). Time to discontinuation was defined as the number of days from the index date to the discontinuation date.

Other measures included patient demographic and clinical characteristics, including age (as of index date), US geographic region (Northeast, South, North Central, West, and Other), sex, and race/ethnicity (Black, White, Hispanic, Asian, and Other). The Quan–Charlson comorbidity index (CCI) was calculated based on diagnosis codes associated with medical claims during the study baseline period^15,16. Finally, the total number of concomitant prescription medications filled for each patient during the study baseline period was calculated using pharmacy claims data.

Statistical analysis

Descriptive analyses were performed for all study variables, including means and standard deviations for continuous variables, and frequency counts and percentages for categorical variables. To evaluate adherence, unadjusted as well as adjusted logistic regression models were fitted for baseline characteristics associated with having suboptimal adherence to somatropin (MPR <0.80) for 12, 24, and 36 months of continuous follow-up. Odds ratios (OR) with 95% confidence intervals (CI) and p-values were calculated for each of the characteristics. Sensitivity analyses were performed which evaluated suboptimal adherence at 12 and 24 months among the 24-month follow-up cohort, and at 12, 24, and 36 months among the 36-month follow-up cohort. To evaluate discontinuation, unadjusted as well as adjusted Cox regression models were fitted for baseline characteristics associated with time to discontinuation of somatropin, and hazard ratios (HR) with associated 95% CIs and p values were generated for patient baseline characteristics. Variables included for adjustment in multivariable models were age, sex, geographic region, race/ethnicity, CCI, and number of concomitant medications at baseline.

Results

Study population

Overall, 40,064 patients were identified in the CMS Medicaid database with medical and prescription drug benefit eligibility and at least 2 new somatropin prescriptions filled between July 2014 and December 2018 (Figure 1). After application of age and diagnosis criteria, and excluding other causes of short stature, 8450 individuals (21%) remained. After determination of available follow-up based on continuous eligibility post-index date, 4 non-exclusive cohorts were constructed: 3 months follow-up (n = 5064), 12 months follow-up (n = 3623), 24 months follow-up (n = 1004), and 36 months follow-up (n = 90). Demographic and clinical characteristics of the study population are included in Table 1. Mean patient age for all cohorts was 10–11 years, and about 7 in 10 patients were male. Overall, about 40% were White non-Hispanic, 30% were Hispanic, 7% Black, and 2% Asian. Almost half of patients were prescribed 3 or more concomitant medications at baseline; only about 1 in 4 were not prescribed other medications at baseline.

Figure 1. Sample attrition for Medicaid pediatric growth hormone deficiency cohort. ¹Eligible patients were grouped into non-exclusive cohorts based on length of available follow-up, determined by index date and continuous enrollment. For example, all patients in the 36-month continuous enrollment cohort, by definition, are also included in the 3-, 12-, and 24-month continuous enrollment cohorts.

Table 1. Medicaid pGHD patient demographic and baseline clinical characteristics.

	Months of continuous follow-up
	3 Months	12 Months	24 Months	36 Months
	N = 5064	N = 3623	N = 1004	N = 90
Age, mean (SD)	10.5 (3.2)	10.5 (3.2)	10.4 (3.1)	10.2 (3.4)
Age 3–9	1756 (34.7%)	1259 (34.8%)	356 (35.5%)	37 (41.1%)
Age 10–16	3308 (65.3%)	2364 (65.2%)	648 (64.5%)	53 (58.9%)
Sex, n (%)
Male	3613 (71.3%)	2564 (70.8%)	682 (67.9%)	57 (63.3%)
Female	1451 (28.7%)	1059 (29.2%)	322 (32.1%)	33 (36.7%)
Census region, n (%)
Midwest	790 (15.6%)	514 (14.2%)	83 (8.3%)	16 (17.8%)
Northeast	614 (12.1%)	469 (12.9%)	135 (13.4%)	17 (18.9%)
South	1242 (24.5%)	925 (25.5%)	255 (25.4%)	21 (23.3%)
West	1565 (30.9%)	1108 (30.6%)	381 (37.9%)	23 (25.6%)
Other	853 (16.8%)	607 (16.8%)	150 (14.9%)	13 (14.4%)
Race/ethnicity, n (%)
White	2197 (43.4%)	1610 (44.4%)	435 (43.3%)	44 (48.9%)
Black	337 (6.7%)	257 (7.1%)	81 (8.1%)	<11
Hispanic	1480 (29.2%)	1053 (29.1%)	337 (33.6%)	30 (33.3%)
Asian	93 (1.8%)	61 (1.7%)	13 (1.3%)	<11
Other	85 (1.7%)	67 (1.8%)	25 (2.5%)	<11
Null/Unknown	872 (17.2%)	575 (15.9%)	113 (11.3%)	<11
Index year, n (%)
2014	276 (5.5%)	257 (7.1%)	219 (21.8%)	39 (43.3%)
2015	784 (15.5%)	738 (20.4%)	658 (65.5%)	45 (50%)
2016	1433 (28.3%)	1324 (36.5%)	116 (11.6%)	<11
2017	1441 (28.5%)	1302 (35.9%)	11 (1.1%)	<11
2018	1130 (22.3%)	–	–	<11
Cohort clinical characteristics (Charlson Comorbidity Index), n (%)
CCI 0	4222 (83.2%)	3000 (82.8%)	825 (82.2%)	79 (87.8%)
CCI 1	545 (10.8%)	407 (11.2%)	116 (11.6%)	<11
CCI 2	167 (3.3%)	120 (3.3%)	35 (3.5%)	<11
CCI 3	73 (1.4%)	52 (1.4%)	13 (1.3%)	<11
CCI 4	23 (0.5%)	19 (0.5%)	<11	<11
CCI 5 +	34 (0.7%)	25 (0.7%)	<11	<11
Baseline obesity, n (%)	2.4%	2.5%	2.9%	<11
Concomitant medications, mean (SD)	3.2 (3.6)	3.4 (3.7)	3.2 (3.5)	2.7 (3.3)
0	1139 (22.5%)	769 (21.2%)	227 (22.6%)	20 (22.2%)
1	890 (17.6%)	621 (17.1%)	182 (18.1%)	16 (17.8%)
2	694 (13.7%)	501 (13.8%)	135 (13.4%)	20 (22.2%)
> =3	2341 (46.2%)	1732 (47.8%)	460 (45.8%)	34 (37.8%)

**Cells with n < 11 denote inability to provide full descriptive contents due to CMS cell suppression policy.

Abbreviations pGHD, pediatric growth hormone deficiency; SD, standard deviation; CCI, Charlson Comorbidity Index.

Adherence and discontinuation

Mean (95% CI) adherence, as measured using medication possession ratio (MPR), among those with 12, 24, 36 months of follow-up was 0.79 (0.78–0.80), 0.75 (0.73–0.76), and 0.74 (0.70–0.79), respectively (data not shown). Overall, among the 12-, 24-, and 36-month follow-up cohorts, 40.9%, 50.4%, and 54.4% of pGHD patients, respectively, had suboptimal somatropin adherence (Figure 2(a)). Suboptimal adherence was higher among children aged 3-9 years compared to older children, and this difference was more pronounced among cohorts with longer follow-up. Suboptimal adherence was most pronounced among Black children. At 12 months, suboptimal adherence was observed among 52.5% of Black, 40.8% of Hispanic, 40.1% of White, and 32.8% of Asian children; at 24 months, suboptimal adherence was observed among 64.2% of Black, 51.0% of Hispanic, and 48.7% of White children. Insufficient sample size prohibited estimation of 24-month adherence among Asian children. Sensitivity analysis demonstrated that among the 24-month follow-up cohort, overall, 28.0% had suboptimal adherence at 12 months and 50.4% at 24 months (Figure 2(b)); among the 36-month follow-up cohort, 33.3% had suboptimal adherence at 12 months, 43.3% at 24 months, and 54.4% at 36 months (Figure 2(c)).

Figure 2. Suboptimal somatropin adherence (MPR <0.80) over 3 years among a Medicaid population diagnoses with pGHD. (a) Proportion of patients* with suboptimal somatropin adherence (MPR <0.80) among 3 pGHD cohorts with 1, 2, and 3 years of follow-up. ( b) Sensitivity analysis proportion of patients* with suboptimal somatropin adherence (MPR <0.80) at 1 and 2 years of follow-up among children with pGHD and 2 years of continuous follow-up. (c) Sensitivity analysis proportion of patients* with suboptimal somatropin adherence (MPR <0.80) at 1, 2, and 3 years of follow-up among children with pGHD and 3 years of continuous follow-up. *For categories with n < 11, descriptive statistics were unable to be presented due to Medicaid privacy regulations. Abbreviations. MPR, medication possession ratio; pGHD, pediatric growth hormone deficiency.

Unadjusted logistic regression results are included in Table 2. Within the 12- and 24-month follow-up cohorts, significant predictors of suboptimal adherence included age, census region, and race/ethnicity. In the 12-month follow-up cohort, adjusted odds of suboptimal adherence for age 10-16 years (vs. 3–9 referent) were 0.83 (95% CI 0.72–0.96), and Black (vs White ref) were 1.67 (95% CI 1.27–2.16). Among the 24-month follow-up cohort, the adjusted odds of suboptimal adherence for age 10–16 (vs. 3–9) were 0.70 (95% CI 0.54–0.91); Black (vs. White ref) 1.89 (95% CI 1.15–3.08). Due to small sample size and insufficient power, no significant predictors were identified in the 36-month follow-up cohort. Significant predictors identified in the unadjusted logistic regressions persisted after adjustment for covariates.

Table 2. Unadjusted logistic regression for baseline characteristics associated with suboptimal adherence to somatropin (MPR <0.80) among Medicaid pGHD cohorts with 1–3 years of follow-up.

	12 Months continuous follow-up (n = 3623)	24 Months continuous follow-up (n = 1004)	36 Months continuous follow-up(n = 90)
	OR (95%CI)	OR (95%CI)	OR (95%CI)
Demographic characteristics
Age group	p = .0092	p = .0067	p = .0994
Age 3–9	Ref	Ref	Ref
Age 10–16	0.834 (0.724 − 0.956)	0.698 (0.538 − 0.905)	0.484 (0.204 – 1.147)
Sex	p = .3558	p = .6153	p = .3728
Male	Ref	Ref	Ref
Female	1.071 (0.926 − 1.238)	1.070 (0.821 − 1.395)	1.485 (0.622 − 3.546)
Census region	p < .0001	p = .0026	p = .9117
Midwest	Ref	Ref	Ref
Northeast	1.122 (0.872 − 1.444)	0.711 (0.408 − 1.237)	0.857 (0.212 − 3.473)
South	0.896 (0.720 − 1.115)	0.697 (0.421 − 1.154)	0.660 (0.175 − 2.485)
West	1.028 (0.833 − 1.270)	0.737 (0.454 − 1.195)	0.550 (0.150 − 2.021)
Other	0.59 (0.461 − 0.754)	0.371 (0.214 − 0.645)	0.700 (0.158 − 3.099)
Race/ethnicity	p = .0020	p = .1572	p = .7096
White	Ref	Ref	Ref
Black	1.656 (1.271 − 2.156)	1.886 (1.154 − 3.084)	3.286 (0.317 − 34.083)
Hispanic	1.033 (0.882 − 1.210)	1.097 (0.825 − 1.457)	1.892 (0.732 − 4.889)
Asian	0.73 (0.424 − 1.257)	0.657 (0.212 − 2.041)	2.190 (0.185 − 25.956)
Other	1.288 (0.789 − 2.104)	0.826 (0.357 − 1.861)	1.095 (0.064 − 18.638)
Null	0.92 (0.757 − 1.119)	0.963 (0.636 − 1.457)	0.821 (0.164 − 4.108)
Clinical characteristics
CCI	p = .1041	p = .6557	p = .9553
0	Ref	Ref	Ref
1	0.808 (0.652 − 1.002)	0.828 (0.561 − 1.222)	1.545 (0.345 − 6.908)
2	1.093 (0.756 − 1.58)	0.898 (0.456 − 1.766)	>999.999
> =3	1.315 (0.875 − 1.975)	0.713 (0.333 − 1.525)	>999.999
Number of concomitant medications	p = .1265	p = .4888	p = .2821
0	Ref	Ref	Ref
1	1.036 (0.836 − 1.282)	1.147 (0.775 − 1.695)	1.111 (0.288 − 4.289)
2	0.926 (0.737 − 1.164)	0.873 (0.570 − 1.337)	0.359 (0.100 − 1.294)
> =3	0.854 (0.719 − 1.015)	0.892 (0.649 − 1.227)	0.952 (0.309 − 2.935)

Unadjusted logisitic regression for demographic and clinical characteristics associated with being non-adherent to somatropin (MPR <0.80). Abbreviations. OR, odds ratio; pGHD, pediatric growth hormone deficiency; CI, confidence interval; CCI, Charlson Comorbidity Index; MPR, medication possession ratio; ref, referent category. Bold font indicates statistical significance at p < .05. “Other” geographic region includes American Samoa, Guam, Puerto Rico, US Virgin Islands, and Medicaid beneficiaries with unpopulated or null values for state of residence.

Discontinuation was evaluated among all study patients with at least 3 months of continuous follow-up; among these individuals (n = 5064), 49.2% discontinued somatropin therapy. Among those who discontinued therapy, mean time to discontinuation was 263.3 days (SD 213.1). In unadjusted Cox regression models, Black race was associated with increased risk of discontinuation, while Northeast census region (Maine, New Hampshire, Vermont, Massachusetts, Rhode Island, Connecticut, New York, New Jersey, and Pennsylvania) was associated with decreased risk (Table 3). After adjustment, both factors remained significant predictors of time to discontinuation.

Table 3. Unadjusted and adjusted Cox regression model results for baseline characteristics associated with time to discontinuation of somatropin among children with pGHD and ≥3 months of follow-up.

	Unadjusted cox regression model	Adjusted cox regression model
	3 Month follow-up cohort (N = 5064)	3 Month follow-up cohort (N = 5064)
	HR (95%CI)	HR (95%CI)
Demographic characteristics
Age	p = .8800	p = .9530
Age 3–9	Ref	Ref
Age 10–16	1.006 (0.927 − 1.092)	1.003 (0.923 − 1.089)
Sex	p = .6435	p = .5436
Male	Ref	Ref
Female	1.021 (0.936 − 1.113)	1.028 (0.941 − 1.122)
Census region	p = .0327	p = .0339
Midwest	Ref	Ref
Northeast	0.844 (0.727 − 0.981)	0.853 (0.733 − 0.992)
South	0.925 (0.810 − 1.056)	0.933 (0.816 − 1.066)
West	0.989 (0.870 − 1.124)	1.028 (0.901 − 1.173)
Other	0.859 (0.743 − 0.993)	0.906 (0.779 − 1.054)
Race/ethnicity	p = .0004	p = .0004
White	Ref	Ref
Black	1.349 (1.154 − 1.577)	1.341 (1.147 − 1.568)
Hispanic	0.979 (0.892 − 1.075)	0.954 (0.861 − 1.057)
Asian	1.075 (0.797 − 1.448)	1.044 (0.773 − 1.409)
Other	1.259 (0.953 − 1.663)	1.272 (0.963 − 1.681)
Null	1.149 (1.024 − 1.291)	1.126 (1.001 − 1.266)
Clinical characteristics
CCI	p = .1610
0	Ref
1	0.920 (0.806 − 1.050)
2	0.932 (0.756 − 1.149)
> =3	1.226 (0.972 − 1.546)
Number of concomitant medications	p = .3553
0	Ref
1	0.954 (0.844 − 1.079)
2	1.013 (0.887 − 1.157)
> =3	0.928 (0.839 − 1.027)

Abbreviations. HR, hazards ratio; pGHD, pediatric growth hormone deficiency; CI, confidence interval; CCI, Quan–Charlson Comorbidity Index; ref, referent category. Bold font indicates statistical significance at p < .05. “Other” geographic region includes American Samoa, Guam, Puerto Rico, US Virgin Islands, and Medicaid beneficiaries with unpopulated or null values for state of residence.

Discussion

In this Medicaid population, a substantial proportion of children with pGHD had suboptimal adherence to prescribed daily somatropin injections. The proportion of patients with suboptimal adherence to somatropin was related to cohort length of follow-up: the proportion with suboptimal adherence in the Medicaid database was 40.9%, 50.4%, 54.4% at months 12, 24, and 36, respectively. Furthermore, almost 1 in 2 (49.2%) patients followed for ≥3 months discontinued prescribed somatropin therapy. We also found that lower age, race/ethnicity (Black as compared to non-Hispanic White), and geographic region were associated with greater odds of suboptimal adherence, and the risk of discontinuation was associated with Black race and geographic region.

Our finding of suboptimal somatropin adherence in a Medicaid population was more than twice that of what has been reported for commercially insured children (19.6%) at 12 months, and 60% higher at 36 months. A recent retrospective claims analysis estimated that about 20 and 34% of commercially insured children with pGHD were suboptimally adherent at 12 and 36 months, respectively¹⁷. In addition to higher rates of suboptimal adherence in our Medicaid population, discontinuation rates were higher (49.2%) than what was previously reported for commercially insured children (42.2%) followed for at least 3 months, and median time to discontinuation was 206 days in the current study compared to 442 days among commercially insured children. A recent study of the economic burden of pGHD, found suboptimal adherence (proportion of days covered <0.80) for 81.6% of Medicaid patients and 67.7% of commercially insured patients (mean follow-up 3–4 years)¹⁸. This study also found high levels of discontinuation with 49.1% of Medicaid patients and 24.3% of commercially insured patients discontinuing GH treatment before the age of 13¹⁸. A survey of patients treated with GH found the main factors associated with poor treatment adherence were discomfort with injections, dissatisfaction with treatment outcomes, inadequate contact with healthcare providers and misperceptions about the potential consequences of missed GH doses¹⁹.

In the current study, among a Medicaid population, younger children (3–9 years) had greater odds of suboptimal adherence than older children (11–16 years). One potential explanation may be increased adolescent self-consciousness about height, which may lead to greater motivation to adhere to their treatment regimen. For example, a retrospective study of Medicaid enrollees with acne vulgaris found that adolescents may be more likely than children to be adherent to medication for acne, although the difference was not significant²⁰.

Black children were more likely to have suboptimal adherence and to discontinue therapy than White children. Prior research has indicated lower quality of health care for children and adolescents enrolled in Medicaid compared to their commercial counterparts^21,22. Landon et al.²¹ found that based on Health Plan Employer and Data Information Set (HEDIS) quality of care indicators, Medicaid enrollees had lower rates for clinical access indicators than members of commercial plans¹⁵. While both studies found worse adherence among Black children as compared to White, overall higher rates of suboptimal adherence in the Medicaid population are not solely explainable based on measured patient demographic characteristics. In the current study of Medicaid enrollees, suboptimal adherence was observed among 40.1% of White, 52.5% of Black, and 40.8% of Hispanic children, compared to 16.8% of White, 18.9% of Black, and 20.7% of commercially insured children¹⁷. We also found that children from the Northeastern US were less likely to discontinue GH treatment; this may reflect regional differences in access to health care or be related to state-level differences in Medicaid health plans. Additional research is warranted to understand structural, policy, and other barriers to GH treatment specific to the Medicaid population.

Several other studies have evaluated the effects of socioeconomic status on adherence to prescribed GH therapy outside of the US. In a study of GH treatment in New Zealand, Cutfield et al.⁸ found that Maori ethnicity was associated with lower GH adherence. Aydin et al.¹⁰ studied 217 pediatric GH patients in Turkey, and while the authors found that male gender was associated with better adherence, socioeconomic status was not associated with adherence. Farfel et al.²³ studied 2263 pediatric GH patients from an Israeli health maintenance organization (HMO), and observed poor adherence associated with younger age of treatment initiation, but no effect of socioeconomic status or gender. Hughes et al.²⁴, in an Australian GH program evaluation, found that among patients who ended GH therapy, patients from a lower socioeconomic background more likely to have early cessation of GH therapy. De Pedro et al.²⁵, in a study of 158 children treated with GH in Argentina, found that maternal educational level predicted adherence, with a higher proportion of children with “good adherence” (≥92%) among mothers with a college education as compared to those with moderate to poor adherence (<92%). Our study provides a valuable addition to this literature by evaluating GH adherence in a socioeconomically disadvantaged population within the US healthcare system. In a 2011 literature review, key drivers of poor GH adherence identified both psychological/emotional and social problems and stressed the interconnected nature of these factors to socioeconomic issues such as poverty, low education levels, and lack of social support²⁶. While socioeconomic status may impact GH adherence across study settings, structural and policy related barriers to adherence may differ substantially based on characteristics unique to each setting and its healthcare system. Since poor adherence to prescribed GH regimens is associated with decreased final height, our study’s findings suggest that children with pGHD who are socioeconomically disadvantaged may be less likely to achieve maximum adult height potential, possibly impacting quality of life in the longer term. While there is a lack of long-term data on such outcomes in pGHD, short-term treatment with GH has been associated with improved social, emotional, and physical quality of life²⁷. Provider-targeted initiatives to recognize at-risk children and parents for outreach and education related to the implications of GH adherence on clinical outcomes may be warranted. Furthermore, GH regimens with fewer doses and more convenient dosing requirements, such as LAGH, could help to improve adherence and outcomes among socioeconomically disadvantaged children.

While our study has several strengths, including the size of the database used and geographic representation across the US, several limitations are important to consider. This study utilized administrative claims data for patient encounters, and since claims data incorporates diagnosis information using ICD9/ICD10 codes for billing purposes, it does not measure validated clinical diagnoses. However, to minimize misclassification, we limited our study to patients with ≥2 somatropin prescriptions and a GHD diagnosis, and we excluded patients who may be prescribed somatropin for other indications. Claims data are collected for service reimbursement, and as such, use of administrative data for research purposes may be subject to errors or omissions. While the data from the TAF are adjudicated and adjusted to control for severe data quality issues, remaining issues may impact this study. Firstly, data in the T-MSIS/TAF vary significantly due to state-level differences in administration of the Medicaid program. Additionally, data coverage varies year over year, due to states transitioning to the T-MSIS file structure at different times. In 2014 and 2015, only a subset of states provided T-MSIS data to CMS, and 2016 was the first year that all 50 states began participating in the T-MSIS/TAF. Overall data quality and usability across states improved from 2016 to 2018, especially in the more populous states. We suspect that having fewer states participating in 2014 and 2015 is one of the primary causes for the low sample sizes observed in the 36-month cohort, and these factors may impact ability to follow-up patients longitudinally over several years in the Medicaid data. Specifically, we hypothesize that the relatively poor 2016 data quality is one of the causes for the notable drop in cohort size from 12 months to 24 months. Patients may have had changes to insurance (i.e. from Medicaid to Commercial coverage) which may have contributed to this drop in cohort size over time since we required continuous enrollment; disruptions in Medicaid coverage are common, as patients may drop in and out of Medicaid coverage²⁸. Another limitation of the Medicaid data is that only two diagnosis codes are available per claim for non-inpatient claims. While we cannot be sure of the impact of this on sample size, we suspect that this may have impacted the initial step of identifying patients with documented GHD diagnosis in claims data. Finally, our findings suggest that almost half (49.2%) of children in this study discontinued therapy with somatropin. While we conducted sensitivity analyses using different thresholds for a gap which qualified as a discontinuation, some children may have restarted somatropin therapy after a prolonged gap in therapy, and our analysis did not assess this possibility.

Conclusions

Within a Medicaid population of children with GHD, the proportion with suboptimal adherence to somatropin grew over time, and the risk of suboptimal adherence was associated with younger age and Black race/ethnicity. Risk of discontinuation of somatropin varied by geographic region but was consistently higher among Black children (as compared to White children). Strategies that facilitate adherence to somatropin may support improved clinical outcomes in these vulnerable groups and could support the achievement of final height outcomes that are more consistent with those seen among White and Asian cohorts. Improving adherence to prescribed GH regimen among this at-risk population, and specifically among sociodemographic subgroups at highest risk, is warranted to improve clinical outcomes.

Transparency

Declaration of funding

This study was sponsored by Pfizer.

Declaration of financial/other relationships

JL, YC, JMJA, AG and MPW are employees of Pfizer and may hold stock/stock options in Pfizer. PJ is an employee of Genesis Research and was a paid consultant by Pfizer in connection with this research and development of this manuscript. SP and MK are employees of Clarify Health Solutions and were paid consultants by Pfizer in connection with this research and development of this manuscript. LS has served as a consultant to Novo Nordisk, Opko Health Inc., and Pfizer, and served as an investigator in clinical trials funded by Novo Nordisk, Opko Health Inc., and Lumos Diagnostics. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgements

Writing and editorial assistance was provided by Jenifer Wogen and Amy Glenwright of Genesis Research, LLC, Hoboken, NJ.