Hydroxyurea Adherence for Personal Best in Sickle Cell Treatment (HABIT) Efficacy trial: Community health worker support may increase hydroxyurea adherence of youth with sickle cell disease

Abstract

Despite disease-modifying effects of hydroxyurea on sickle cell disease (SCD), poor adherence among affected youth commonly impedes treatment impact. Following our prior feasibility trial, the HABIT multi-site randomized controlled efficacy trial aimed to increase hydroxyurea adherence for youth with SCD ages 10–18 years. Impaired adherence was identified primarily through flagging hydroxyurea-induced fetal hemoglobin (HbF) levels compared to prior highest treatment-related HbF. Eligible youth were enrolled as dyads with their primary caregivers for the one-year trial of a novel semi-structured supportive, multi-dimensional dyad intervention led by community health workers (CHW), augmented by daily tailored text message reminders, compared to standard care during a 6-month intervention phase, followed by a 6-month sustainability phase. Primary outcomes from the intervention phase were improved HbF levels compared to enrollment and proportion of days covered (PDC) for hydroxyurea versus the pre-trial year. The secondary outcome was sustainability of changes up to month 12. The 2020–2021 peak COVID-19 pandemic disrupted enrollment and clinic-based procedures; CHW in-person visits shifted to virtual scheduled interactions. We enrolled 50 dyads, missing target enrollment. Compared to enrollment levels, both HbF level and PDC significantly - but not sustainably - improved within the intervention group (p=0.03 and 0.01, respectfully) with parallel increased MCV (p=.05), but not within controls. No significant between-group differences were found at months 6 or 12. These findings suggest that our community-based, multi-modal support for youth-caregiver dyads had temporarily improved hydroxyurea usage. Durability of impact should be tested in a trial with longer duration of CHW-led and mobile health support.

Introduction:

Medication adherence can be challenging for people with chronic health conditions, especially among underserved populations. Barriers to medication adherence include logistical impediments to prescription refill, complexity of medication regimens, incomplete knowledge of drug benefit and/or risk.(1–7) Cultural misalignment between medical staff and families may also lead to communication gaps.(3) Barriers to adherence are common in youth with chronic illness and a source of health disparities from differential health outcomes in under-resourced communities.(8–11) Poor adherence is also associated with inadequate integration of adherence into a daily medication habit.(12, 13) For youth with sickle cell disease (SCD) treated with hydroxyurea (HU), poor adherence increases risk of acute and life-threatening complications, e.g. pain-associated hospital admissions.(9, 14, 15)

Systematic reviews and meta-analyses of randomized controlled trials to improve adherence among youth with chronic illness favor multi-modal approaches, some of which include community-based support and/or mobile application reminders.(16, 17) Community health workers (CHW) are a well-established bridge between health care services for chronic health conditions, especially in under-resourced communities.(17–19) Using HU-induced fetal hemoglobin as the primary biomarker of adherence, our prior 2-site feasibility trial to increase HU adherence demonstrated feasibility and acceptability of a novel CHW-led intervention, augmented by text messaging, “Hydroxyurea Adherence for Personal Best in Sickle Cell Treatment (HABIT)” Efficacy trial and a trend towards improved adherence.(20)

Conceptually, the Self and Family Management Framework (SFMF) guided study design focused on key youth, family and community factors for risk and protection, and included engagement of community-based resources.(21) Congruent with this model, adherence issues included psychosocial needs and developmentally influenced challenges to youth-caregiver partnership in HU use. Behavioral aspects of daily medication use included the importance of developing a daily medication habit linked to an established daily behavior (e.g. tooth-brushing or mealtime).

Here we describe the primary outcomes of our 4-site randomized controlled efficacy trial from the intervention phase (between months 0–6) and the secondary outcome from the sustainability phase (between months 6–12) to improve HU use in youth ages 10–18 years with SCD with poor adherence. Our hypotheses were that: 1) the intervention group would exhibit better adherence to HU than the control group between months 0 and 6; 2) the intervention group would sustain improved HU adherence from months 6 to month 12.

Methods:

Details of the trial design and its organization have been previously reported.(9) The institutional review board approved the study at each participating site. An independent multi-disciplinary Data Safety and Monitoring Committee (DSMC) regularly reviewed study progress and any trial adverse effects. Participating sites were academic medical centers with established comprehensive pediatric SCD centers serving medium or large patient populations in New York City (or environs) and Philadelphia. Three sites had previously collaborated with CHWs to support SCD needs; the other site worked closely with an established regional SCD advocacy support group.

Trial participants:

Eligibility and exclusion criteria were specified for youth with SCD (HbSS or HbS-B⁰ thalassemia) and their primary caregiver (Supplemental Table 1). Participants were between 10–18 years and prescribed HU for at least 18 months. Participants were potentially eligible if their preferred language was English or Spanish to enable communication with their site’s CHWs and work with study documents and surveys. Prior individual maximum HU-induced HbF was the personalized target, deemed as “personal best HbF” (PB HbF). (9, 20, 22) PB HbF was assessed at a healthy state and at >4 years of age. Evidence of poor HU adherence was a decline from PB HbF. Eligibility metric for HbF was a decline of ≥15% from PB HbF using the mean value of the two most recent HbF levels over the preceding 12 months (assessed at clinically steady state), despite HU dosing within 5% of dose at PB. At enrollment, the final qualifying criteria was HbF remained ≥15% of PB.

The Social Vulnerability Index (SVI), a federally established method to estimate U.S. household neighborhood resources by census tract, was calculated for the sample by home address at enrollment. SVI is scored between 0–1, with 1 representing the highest vulnerability. Nationally, communities with low SVI experienced the highest risk of COVID-19 infection.(23)

Trial structure:

Participant youth-caregiver dyads were randomized to trial group with centralized 1:1 allocation via computer-generated random numbers stratified by study site with random block sizes of eight. Due to the type of intervention, blinding of study hypotheses but not group assignment was possible for participants, but not for study staff.

Primary objective outcome measures of improved adherence were: 1) change in HU-induced blood fetal hemoglobin (HbF) level from enrollment to month 6. (24, 25) To tailor HbF to individualized HU response, data were also examined by comparison to (“deviation from”) their PB HbF. (For example, if PB HbF was 20% with current HbF at 10% at trial entry, deviation would be 50%.); and 2) Proportion of days covered (PDC) by HU, determined using pharmacy refill data for the 12 study months and 12-month period preceding enrollment. Secondary outcomes were: 1) HbF and deviation from PB HbF from month 6 to month 12; 2) change in PDC from month 0–6 to months 6–12.

Both study groups received the same educational materials on SCD and HU developed for 7^th grade reading levels and available in English or Spanish (per participant choice), and were similarly scheduled for clinical study visits and laboratory assessments.(9) The intervention group received a 4-month semi-structured intervention of community health worker (CHW) visits to the enrolled youth-caregiver dyads, augmented by 2 additional months of tailored daily text message reminders to each dyad member. CHWs also reviewed HbF levels and both education handouts with intervention dyads.

The behavioral intervention target was a daily HU habit, strengthened through an improved dyad partnership for adherence.(9) Interventions were provided to both dyad members to foster their partnership. Topic development for CHW support was previously described, and included extensive CHW input and dyad-driven identification of adherence barriers.(9) (10, 17, 20, 21) Content and schedule of CHW visits are depicted in Figure 1, with content listed in each visit guide (Supplemental Figure). CHWs also used the guide to document topics covered at each visit, lasting approximately one-hour. To avoid interference with dyad-CHW relationships, no other study staff were present at CHW visits.

Figure 1. HABIT Efficacy Trial Intervention Scheme.

CHW collective training and site-specific supervision occurred as described, including review of the SFMF, study topics and materials, role-playing and on-going training.(9) In brief, a semi-structured CHW-led intervention consisted of 5 visits at home (or other location convenient for dyads).

During the semi-structured visits, CHWs and dyads jointly discussed social determinants of health needs, e.g, housing or food insecurity, and made community-based and/or health care referrals (Supplemental figure).(20) CHWs provided social support; addressed informational and logistical barriers to SCD care and HU adherence; developed a habit-based plan for HU adherence; and encouraged communication with SCD providers. The intervention also included individualized daily text message reminders for HU use during intervention months 5–6. Automated messages, designed by dyad members with CHW guidance, were implemented via a HIPAA-compliant system (Mir3^™, OnSolve, Atlanta).

Data collection and analyses:

A REDCap database (http://project-redcap.org) sponsored by Columbia University was used by study staff at each site to collect participant historical and trial data. Clinical data were identified from participant electronic health records. Each site had access only to its own data, with de-identified data across sites available only to the lead site. HbF levels were tracked from pre-HU use (if available), at PB HbF, enrollment and each study visit. Trial-related HbF levels were not collected where clinical circumstances could affect HbF levels: 1) an acute clinical change (e.g. pain, fever, worsened anemia, urgent hospital-based care) occurred during the prior two weeks; 2) blood transfusion over the preceding 3 months; or 3) dyad withdrawal. HU refill data were obtained from participants’ pharmacies by study staff by FAX, email or in person. CHWs informed the intervention dyads of the youth’s HbF values at PB, enrollment and at study visits.

Analytic methods

Based on estimated effect size from our prior feasibility trial, the target enrollment was 87 dyads plus additional 20% to account for either attrition before month 6 or blood transfusions or acute illness rendering HbF inaccurate.(20)

Data were analyzed using descriptive statistics; groups were compared using two sample t-tests, paired t-tests, and chi-square. For the two primary outcomes, HbF and HbF deviation (percentage decrease from PB HbF), we used an individual growth model to incorporate group assignment and study month for examining the difference in each outcome.(9) In the model, the independent variables were time, study group; and their interactions. The models included a person-level random intercept to incorporate clustering due to repeated measured data. Statistical significance was defined as p<0.05. We obtained model-based estimation of outcome changes from baseline to 6 months and from 6 to 12 months for within-group and between group comparisons. Data were analyzed using an intent-to-treat approach, with no imputed data for missing values. No interim analyses were planned or performed.

Results

In all, 477 youth were screened for eligibility (Figure 2). The most common reasons for ineligibility were: HbF within 15% of PB (n=85), not prescribed HU (n=54), lacking sufficient recent HbF levels (n=42), out of age range (n=34), prescribed HU for less than 18 months (n=34), not on a stable HU dose (n=26). Of the 83 eligible patients, 33 caregivers and/or youth declined due to insufficient interest or time. Screening and enrollment were administratively suspended in March 2020 at two study sites due to institutional pandemic-related safety measures.(26) Target enrollment was not met, as 50 dyads enrolled between August 2018 and December 2020. Forty-three (86%) dyads enrolled prior to March 2020. The trial closed in December 31, 2021 based on our funding cycle. No trial-related adverse outcomes were reported.

Figure 2. CONSORT Diagram of the HABIT Efficacy Trial.

Trial fidelity:

Trial conduct conformed to the protocol.(9) CHWs reported that supportive relationships were established and all topics in each visit guide were covered. Prior to the 2020 pandemic, CHW visits primarily were held in the dyad home, with a few dyads having visits held in a coffee shop near their home or SCD clinic. After March 2020, CHW visits continued virtually, mostly via video mobile phone applications. All CHW visits were completed, although with some pandemic-related delays. The automated text messaging system provided reports of daily messages delivered, with successful delivery to each participant in the intervention group. Some HbF and other laboratory data were missed due to pandemic-related restrictions.

Sample characteristics at enrollment:

Of 50 dyads enrolled, 24 dyads were allocated to the intervention group and 26 to the control group (Figure 2). SVI of the sample was 0.84±019, indicating high social vulnerability, with no significant difference by group.(27) Five (10%) dyads chose early trial termination, with 45 (90%) completing the trial. Four dyads elected study termination prior to month 6 (2 from each group), including 1 who chose to undergo curative therapy. One additional dyad (control) relocated to another state after month 10.

At enrollment, neither youth nor caregivers significantly differed by allocation group in any key demographic or clinical feature (Table 1). Three caregivers also were affected by SCD (two in the intervention group and one control).

Table 1. Sample characteristics at enrollment, by study group.

Mean values and standard deviations (SD) or numbers (percentages) are shown. Data were not adjusted for missing values.

Table 1a. Youth Demographic and Clinical Data at Enrollment
	Intervention Group N=24	Control Group N=26	p-value
Age (years) (mean, SD)	14.1±1.9	13.0±1.7	0.051
Youth sex (% female; n)	12 (50.0%)	11± (42.3%)	0.59
Black race (n)	21 (87.5%)	20 (76.9%)	0.33
Ethnicity (Latino/a) (n)	3 (12.5%)	5 (19.2%)	0.52
Born in United States (n)	20 (83.3%)	18 (69.2%)	0.24
Grade level (mean, SD)	8.3±1.8	7.3±2.1	0.11
Age started HU (in years) (mean, SD)	8.3±3.2	7.9±2.4	0.61
Age of personal best (in years) (mean, SD)	10.9±2.5	10.0±2.5	0.23
No additional chronic conditions* (n)	12 (50.0%)	13 (50.0%)	1.00
HU dose at personal best HbF (mean, SD)	25.1±4.0	23.6±4.0	0.20
HU dose at enrollment (mean, SD)	26.6±4.3	25.7±4.7	0.50
Number of daily medications beyond hydroxyurea (mean, SD)	2.7±0.9	2.7±0.8	0.84
Participants taking medications ≥2 times daily (n)	15 (62.5%)	10 (38.5%)	0.09
Number of ED visits during the prior year for SCD (mean, SD)**	1.0±1.5)	0.8±1.4	0.64
Participants with any ED visits during the prior year (n)*	11± (45.8%)	10 (40.0%)	0.68
Number of hospitalizations during prior year (mean, SD)	1.0±1.9	1.0±1.7	0.94
Participants with any hospitalizations in pre-trial year (n)	11 (45.8%)	11 (44.0%)	0.90

Table 1b. Caregiver Demographics at Enrollment
	Intervention Group Caregivers N=24	Control Group Caregivers N=26	p-value
Age (years; mean, SD	42.9±7.5	42.8±10.3	0.97
Caregiver sex (female; n)	21 (87.5%)	21 (80.8%)	0.52
Black race (n)	21 (87.5%)	21 (80.8%)	0.52
Latino/a ethnicity (n)	4 (16.7%)	6 (23.1%)	0.57
English preferred (n)	21 (87.5%)	23 (88.5%)	0.92
Born in U.S. (n)	20 (83.3%)	18 (69.2%)	0.24
Marital status			0.71
Married and living with spouse (n)	8 (33.3%)	10 (38.5%)
Single, separated or divorced (n)	16 (66.7%)	16 (62.5%)
Education			0.18
Elementary or some high school (n)	5 (20.8%)	8 (30.7%)
High school graduate (n)	2 (8.3%)	6 (23.1%)
Some college, college graduate or graduate school (n)	17 (70.8%)	12 (46.2%)
Parental work status*			0.36
Full Time (n)	16 (66.7%)	16 (61.5%)
Part Time (n)	3 (12.5%)	1 (3.9%)
Laid off, unemployed, disabled, retired and/or attending school (n)	5 (20.8%)	9 (34.6%)
Caregiver affected by SCD	2 (8.3%)	1 (3.9%)	0.50
Other people living in home affected by SCD	4 (16.7%)	6 (23.1%)	0.57
Social Vulnerability Index (mean, SD)	0.84±0.19

^*Mostly commonly asthma

^**Emergency department

^*No caregiver was disabled, retired or attending school during the trial.

Historical pre-HU HbF values were available for 22 participants (10 intervention, 12 control), and did not significantly differ by group (p=0.25) (data not shown). PB HbF values were available for all participants: 22.9±11.3% for intervention group and 19.2%±7.4 for controls (p=0.18). At PB and enrollment, no significant differences were found by study group in multiple hematologic parameters, including CBC parameters, HbF or lactate dehydrogenase (LDH) (Supplemental Tables 2a, 2b).

Primary outcomes:

At trial enrollment (month 0), mean HbF was 12.2±5.9% (N=21, 88%) and 12.4±5.3% (N=23, 88%) (p=0.94) for the intervention and control groups, respectively. (Table 2 presents the unadjusted analyses at each time point.) Deviation of HbF from PB was also similar between the two groups (p=0.70). At month 6, mean HbF levels for the 36 youth were 16.1±6.3% versus 11.8±7.3% (p=0.067) for the intervention and control groups, respectively. For youth with HbF data at both time points, no significant changes were seen in either group.

Table 2. Adherence measures: Fetal Hemoglobin (%) and Percent Deviation from Personal Best HbF (%) for Between- and Within-group Comparisons.

Months 6 and 12 completed the intervention and sustainability phases, respectively. Deviation data are expressed as negative values (i.e. below PB HbF). Data were not adjusted for missing values.

	HbF (%)*			Percentage Deviation from Personal Best HbF (%)
	Intervention Group (N)	Control Group (N)	p-value	Intervention Group (N)	Control Group (N)	p-value
Between-group Comparisons
Historical Personal Best HbF (mean, SD)	N=24 22.9±11.3	N=26 19.2±7.4	0.18
Enrollment month 0	N=21 12.2±5.9	N=23 12.4±5.3	0.94	N=21 39.6±17.7	N=23 37.8±12.4	0.70
Month 2	N=19 13.9±7.4	N=18 14.1±5.2	0.92	N=19 -7.8±19.2	N=18 30.2±14.5	0.67
Month 4	N=19 15.7±5.5	N=18 15.1±6.7	0.79	N=19 26.6±16.2	N=18 27.12±20.8	0.93
Month 6*	N=18 16.1±6.3	N=18 11.8±7.3	0.07	N=18 25.0±26.1	N=18 -37.0±24.0	0.16
Month 9	N=16 16.0±6.4	N=15 14.7±9.0	0.65	N=16 24.4±31.1	N=15 -24.2±30.2	0.98
Month 12**	N=22 14.2±5.0	N=18 14.1±9.3	0.98	N=22 33.4±25.2	N=18 28.2±31.8	0.57
Within-group Comparisons
	Intervention Group: Change (%)	Control Group Change (%)		Intervention Group: Change (%)	Control Group Change (%)
Change from month 0 to month 6*	2.7±1.3 p=.03	0.3±1.3 p=0.84		15.7±5.7 p=.006	1.4±5.6 p=0.81
Change from month 6 to month 12**	−1.7±1.2 p=0.17	1.8±1.3 p=0.16		−7.6±5.6 p=0.17	8.7±5.9 p=0.14

Significance defined as p<.05

^*Defines primary outcome; N as shown for month 6

^**Defines secondary outcomes; N are as shown for month 12

By individual growth model analysis and adjusting for time, our primary outcome was that mean HbF significantly improved from months 0 to 6 for the intervention group (2.7±1.3%; p=0.03) but not for controls (0.3±1.3%; p=0.84).(9) However, between-group changes did not meet statistical significance for either change in HbF (p=0.067) (Figure 3a). Consistent with the HbF findings, mean corpuscular volume (MCV) at month 6 was higher for the intervention group (n=20) compared to controls (n=22): 95.7±8.2 ×10⁹/L versus 88.7±13.7 ×10⁹/L (p=0.05) (Supplemental Table 2c). Similarly, within-group MCV increased for the intervention group (p=0.05) but not controls. However, between-group changes did not differ when assessed by individual growth model analysis (p=0.17) (Figure). No group, on average, attained their mean PB HbF.

Figure 3. Primary outcome data for HbF (panel A) and HbF% deviation from personal best HbF (panel B) using an individual growth model analysis at enrollment, month 6 (end of Intervention Phase) and month 12 (end of sustainability phase).

*Within-group change (p=.03).

Similarly, by individual growth model analysis and adjusting for time, deviation from PB HbF at month 6 improved for the intervention group (15.7±5.7%; p=0.006) but not for controls (1.4±5.6%; p=0.81) (Figure 3b). Between-group changes did not reach significance (p=0.07) (Figure 3b).

Pharmacy data were available for 43 participants (86%), 21 intervention and 22 control. Table 3 provides detail regarding the number of days represented and PDC for each period by group assignment. Compared to the year prior to enrollment, PDC in the intervention group improved significantly during the intervention phase (p=0.01) but not in the control group (p=0.06). However, no between-group differences were found in PDC from the year pre-trial to months 0–6 (p=0.60). A higher proportion of intervention youth reached a PDC threshold of >80%: 12 versus 9 controls (Table 3), but was not statistically significant (p=0.35).

Table 3: Proportion of Days Covered (PDC) for Hydroxyurea by Trial Group and Phase.

Pharmacy data are expressed in days and percentage of time for 3 time periods: the 12 months preceding study enrollment (baseline), enrollment to 6 months (Intervention Phase) and 6 to 12 months (Sustainability Phase). Within-group PDC improvement from baseline to the intervention phase was seen only the intervention group.^a No within-group changes were seen between the intervention and sustainability phases.^b

	Intervention Group (N=21)		Control Group (N=22c)
PDC for Hydroxyurea	Mean	SD	Mean	SD	p-value
Pharmacy collection for year prior to entry (days)	341.3	56.3	350.4	62.7	0.62
PDC (%) for the 1 year prior to entry (baseline)	55.2	26.7	55.9	26.1	0.93
Pharmacy collection for trial months 0 to 6 (days)	182.9	30.7	182.6	37.1	0.98
PDC (%) Months 0 to 6 vs. baselinea	75.0	22.2	67.5	27.1	0.32
Pharmacy collection for months 7 to 12 (days)	186.4	47.3	169.8	58.7	0.32
PDC (%) months 7 to 12 vs. intervention phase b	73.5	25.5	67.3	29.8	0.47
	N	%	N	%
PDC ≥80% months 0 to 6	9	42.9	10	45.5	1.0
PDC ≥80% months 7 to 12	12	57.1	8	38.1	0.35
Change in PDC for Hydroxyurea	Mean	SD	Mean	SD	p-value
Change in PDC between the year prior to entry and the intervention phase (%)b	17.9	27.3	13.1	29.7	0.60
Change in PDC between the intervention phase and sustainability phase (%)c	−1.6	22.8	−1.9	33.7	0.97

^aCompared to baseline, PDC during the intervention phase improved within the intervention group (p=.01), but not within the controls (p=.06).

^bCompared to the intervention phase, PDC during the intervention phase did not change for the intervention group (p=0.76) or controls (p=.80).

^cN=21 during month 7 to 12 collection.

Significance defined as p <0.05.

Secondary outcomes:

Compared to month 6, we found no HbF differences between the intervention and control groups at month 12 (14.2±5.0% versus 14.1±9.3%; p=0.97) (Table 2). Within-group changes in HbF from month 6 to month 12 trended in opposite directions: lower HbF in the intervention group (−1.7±1.2%; p=0.17) and higher among controls (1.8±1.3% (p=0.16). individual growth model analysis, a significant difference from month 6 to month 12 was found between groups (p=0.0496) (Figure 3.) However, these changes occurred in the direction opposite to our hypothesis.

There were no between-group differences in deviation from PB HbF at 12 months (33.4±25.2% versus 28.2±31.8%, p=0.57). No significant within-group changes were found in either the intervention (−7.6±5.6% (p=0.17)) or the control groups (8.7±5.9% (p=0.14)). No other laboratory values differed between group at month 12 (Supplemental table 2d). For PDC, on average neither group changed significantly from 6 months to 12 months for the or control groups (intervention group: −1.6±22.8%, p=0.76; control group: −1.9±33.7%, p=0.80). No between-group differences were seen (p=0.97).

We also assessed transfusions and urgent care (ED visits and/or hospitalizations) needed during the trial. Both groups had similar numbers of participants requiring transfusions (2 versus 3). During months 0–6, 7 intervention participants required urgent care compared to 12 control participants. In contrast, during months 7–12 two participants in each group received urgent care. These small numbers and potential pandemic-related effects precluded formal comparisons. Patient-reported outcomes and dyad evaluations were not primary outcomes, and will be reported separately.(9)

Discussion

Our multi-site RCT testing a novel intervention, combining community-based CHW support with text messaging, to improve daily HU adherence in youth with SCD was conducted at four participating sites in NYC and Philadelphia. As we have previously reported from the HABIT feasibility trial, both HbF and deviation from PB HbF were used to quantify trial impact, the latter rooted in each study group’s individualized historical maximum HU-induced HbF.(20) Use of PDC data, a measure of medication acquisition by patients with chronic health conditions, support the validity of the HbF findings for assessing adherence.(28) The trial did not meet target enrollment, in part due to pandemic-associated interruptions in 2020 and 2021. Trial sites also experienced disrupted patient attendance at out-patient clinical visits over that time.(29) Nonetheless, the CHW-led intervention was successfully delivered to all intended dyads.

Our primary findings were: 1) The CHW-led intervention, supplemented by daily text message reminders, improved HbF levels and deviations from PB levels from baseline. Those improvements were not seen in the control group; 2) Similarly, the HU PDC increased only for the intervention group. Concordance of results from the intervention using both HbF levels and HU PDC, as well as MCV, reinforced the validity of having improved HU adherence through the intervention. 3) The trial failed to demonstrate between-group differences, to which inadequate enrollment likely contributed; 4) Impact on adherence was not sustainable when assessed for the 6 months post-intervention.

Self-management of chronic health conditions requires coordination with family, health care providers and health care systems for supporting the complex, dynamic processes needed for optimizing health.(21) CHW support can impact healthy behaviors beyond traditional clinical services and promote health equity for youth with SCA and other chronic health conditions.(19, 30) Developmentally appropriate approaches to enhance self-management by youth with SCD and their caregivers, including daily HU use, are important for successful intervention. The HABIT intervention trial appears to have supported those needs through CHWs and individualized text message reinforcement. Our finding that the intervention was delivered both in person and virtually with good fidelity across study sites could stimulate future intervention trials. Interventions of longer duration may better address persistent social and developmental barriers to adherence.

Findings from recent HU adherence intervention trials utilizing community-based patient navigators or direct observation by mobile phone also reported transient increases among adults or adolescents with SCD, respectively.(18, 31) Another mobile health approach showed promise, but lacked a control group and assessment of long-term sustainability.(32) Barriers to improving self-management for people living with SCD can be challenging to overcome, such as family dynamics, social stressors, barriers within health care systems and disease-associated fatigue and/or cognitive symptoms.(33, 34) Better HU adherence - and the health of youth with SCD - may be improved through continuing CHW-led support for youth-caregiver dyads for developmentally appropriate self-management. Durability of impact should be evaluated.(35, 36)

Limitations included modest enrollment and missing HbF data, which collectively contributed to reduced statistical power for primary outcomes and analyses of potential moderators. For PDC analyses, large standard deviation and low power likely contributed to lack of between-group significances following the intervention phase. Contrary to our prior feasibility trial, a relatively low proportion of youth screened met eligibility criteria.(20) This finding and the geographic restriction to two states suggest that the trial sample may not be representative of the U.S. population of youth with SCD. Trial design did not permit separating effects from direct CHW support from text message reminders. An unblinded trial may have affected outcomes, especially among the control group. However, no evidence of group contamination was found. (37) Low participant numbers needing acute health care intervention during the trial precluded meaningful comparison of clinical effects by group. While we were unable to assess pandemic-related effects on HU adherence, a separate U.S.-based SCD trial demonstrated reduced hydroxyurea adherence during the public health emergency.(38) Nationally, trial interruption in non-COVID-19 studies substantially increased during that period.(39)

Results from the intervention group suggest a positive impact – albeit temporary - from multi-modal, community-based, socially supportive interventions on improving HU adherence in youth with SCD. However, effects were not sustained in the post-intervention period. In our feasibility trial, dyad participants who received the intervention expressed positive perspectives on the CHW-led intervention.(20) The HABIT intervention may be transferable to the use of additional medications for SCD and for adherence among underserved youth with other chronic conditions. Our findings suggest that long-term community-based CHW support with text message reminders should be tested for sustaining medication adherence in youth with SCD and possibly other chronic health conditions. In clinical settings, the HABIT intervention may provide a useful model for supporting youth and family, especially in under-resourced communities.

Abbreviations Table:

SCD

Sickle Cell Disease

HbF

Fetal hemoglobin

CHW

Community health worker

PDC

Proportion of days covered (by medication)

COVID-19

Coronavirus Disease 2019

HU

Hydroxyurea

HABIT

Hydroxyurea Adherence for Personal Best in Sickle Cell Treatment

PB HbF

Personal best HbF

SFMF

Self and Family Management Framework

DMSC

Data Safety and Monitoring Committee

SVI

Social Vulnerability Index

MCV

Mean corpuscular volume