These data were collected as part of an institutional review board–approved study. The details of the Mobile DOT intervention and the protocol for this study were previously published [19]. Briefly, this was a single-arm, cross-over study performed at Nationwide Children’s Hospital (NCH), a large, comprehensive pediatric hospital in Ohio. English-speaking patients with SCD (any genotype) who were aged ≤19 years, prescribed hydroxyurea for at least the previous 180 days, not receiving chronic transfusion therapy, planned to receive care at NCH for the following year, and had (or their legal guardian had) personal daily access to an electronic device (eg, smartphone or tablet) that could receive SMS text message alerts and record and send daily videos were eligible. Electronic device access was confirmed, before enrollment, when the patient successfully submitted a video to the secure Mobile DOT website.

Patients were sequentially approached to participate when they presented for clinical care, and follow-up study visits were completed during participants’ standard of care hematology visits. Participants’ electronic medical records (EMRs) were used to collect demographic information, laboratory and clinical data, hydroxyurea dispensing pharmacies, and duration of hydroxyurea treatment. Hydroxyurea dosing and monitoring were left to clinicians’ discretion and tracked using the EMR, but institutional standards, derived from published hydroxyurea protocols [20], were available to guide this management.

There were 3 sequential study periods: baseline, Mobile DOT, and postintervention that were each approximately 180 days. The baseline period was defined as the 180 days before enrollment, the Mobile DOT period started the day following enrollment, and the postintervention period started immediately after the Mobile DOT period. The Mobile DOT and postintervention periods were designed to be 180 days but could vary slightly in length to coordinate participants’ study visits with their standard of care hematology visits. In addition, because hydroxyurea administration that occurs in the hospital setting is highly monitored and includes interventions that are separate from Mobile DOT (eg, all hydroxyurea dose administrations are observed and tracked by a nurse), only participants who completed at least 160 days of the Mobile DOT period and who were hospitalized for fewer than 20 days during this time were considered to be engaged participants. Participants who completed at least 160 days of the postintervention period before their final study visit and were hospitalized for fewer than 20 days during this period were considered to have completed the study.

A custom and secure Mobile DOT website to send the alerts and store participants’ videos was built by the research technology team at NCH and run under an Apache Web server (version 2.2.15; Apache Software Foundation), was written in the Perl and JavaScript languages, and was run on a 64-bit CentOS 6.5 Linux operating system. Participants were informed before consenting of the potential risk that videos could potentially be intercepted during transmission to the secure website, but their videos were secure once they were received by the website.

Participants who discontinued hydroxyurea moved out of the area or informed the research team that they did not want to continue to participate were immediately withdrawn from the study. The research team also stopped sending SMS text message alerts and attempting to contact participants who did not submit a video or respond to research staff communications for more than 30 consecutive days during the Mobile DOT period. These participants were withdrawn from the study at their next clinic visit.

Descriptive statistics were used to describe sample characteristics and report the clinical outcomes, frequency and percentages for qualitative variables, and median and range or mean and standard error for quantitative variables. Nonparametric tests were used to compare baseline characteristics between engaged and nonengaged participants. McNemar test was used to compare proportion of patients who were adherent at baseline and with Mobile DOT. To compare adherence and hematologic measures across the 3 study periods, linear mixed models with Tukey-adjusted pairwise comparisons were used. All P values were 2-sided, and P<.05 was considered statistically significant. Statistical analyses were performed using SAS software, version 9.4 (SAS Institute).

Between July 2014 and January 2017, 91 unique prospective participants were approached, 55 enrolled (76% of the planned sample size), 34 engaged with the Mobile DOT intervention, and 29 completed the study (Figure 1). All the participants planned to use a smartphone as their electronic device. None of the patients who declined participation reported privacy concerns as a reason for study refusal. Recruitment ended before enrolling the planned sample size when 97% of the prospective patients had been approached because recruitment would have needed to be extended for multiple months to approach the remaining patients at their standard hematology visits.

Dispensing records were received for 54 (33 engaged) participants, and these participants were included in the MPR analyses. At baseline, 14 (25%, 14/54) enrolled participants were adherent (MPR ≥80%) to hydroxyurea, and 17 (31%, 17/54) had a hydroxyurea MPR <50%.

Table 1 shows a comparison of the baseline characteristics between engaged and nonengaged participants. Of the 17 engaged participants who were aged<10 years, 8 (47%) were adherent to hydroxyurea at baseline and 3 (27%) of the 11 nonengaged participants who were aged<10 years were adherent at baseline. In addition, 2 (13%) of the 17 engaged participants who were aged ≥10 years were adherent to hydroxyurea at baseline, and one (10%) of the 10 nonengaged participants who were aged ≥10 years was adherent to hydroxyurea at baseline. Engaged participants received Mobile DOT for a median of 191 days, and those who completed the study were followed for a median of 181 days during the postintervention period. Engaged participants’ median prescribed hydroxyurea dose did not significantly change from baseline to the end of the Mobile DOT period (25.9 mg/kg/day vs 25.8 mg/kg/day; P>.99). Of those who completed the study, 21 (72%) were prescribed a dose that was within 5 mg/kg/day of their baseline dose, but their median prescribed hydroxyurea dose was significantly lower at the end of the postintervention period compared with baseline (23.5 mg/kg/day; P=.004).

Participants’ median hydroxyurea MPR was significantly higher with Mobile DOT compared with baseline, but median postintervention MPR was not significantly different compared with baseline (Figure 2). The proportion of engaged participants that were adherent (MPR ≥80%) to hydroxyurea significantly increased from 30% at baseline to 67% with Mobile DOT (P=.002). In addition, 20 (61%) of the engaged participants increased their MPR by >15% with Mobile DOT compared with their baseline. Median VDOT adherence (88.1%) and MPR adherence (87.8%) for the engaged participants during the Mobile DOT period were not significantly different (P=.99).

Engaged participants’ mean MCV and HgbF levels at the end of Mobile DOT were significantly higher than baseline but were not significantly different at the end of postintervention period compared with baseline. Engaged participants’ mean Hgb and ANC did not significantly change throughout the study (Figure 3).

During the baseline period, 8 participants received at least one erythrocyte transfusion (6 nonengaged vs 2 engaged, P=.04), 11 had at least one ACS episode (8 nonengaged vs 3 engaged, P=.01), and 12 had at least one VOC hospitalization (7 nonengaged vs 5 engaged, P=.18). The cohort experienced a total of 13 ACS episodes and 26 VOC hospitalizations during the baseline period, but 1 participant accounted for 9 of these VOC hospitalizations and was nonengaged during the Mobile DOT period because of continued recurrent VOC hospitalizations. During the Mobile DOT period, 3 engaged participants received at least one erythrocyte transfusion, 2 had at least one ACS episode, and 6 had at least one VOC hospitalization. During the postintervention period, 1 participant required at least one transfusion, 3 had at least one ACS episode, and 3 had at least one VOC hospitalization.

All the engaged participants and 16 (76%) of the nonengaged participants completed the satisfaction survey. Of the engaged participants, 31 (92%) agreed that they took hydroxyurea more often with Mobile DOT, 34 (100%) agreed that Mobile DOT was easy to use, 30 (88%) agreed that they would be willing to continue Mobile DOT, 19 (56%) agreed that they would be willing to continue Mobile DOT without pay, and 30 (88%) reported they could complete Mobile DOT in less than 3 min each day. Of the nonengaged participants, 9 (56%) agreed that they took hydroxyurea more often with Mobile DOT, 8 (50%) agreed that Mobile DOT was easy to use, 7 (44%) agreed they would be willing to continue Mobile DOT, 5 (31%) agreed that they would be willing to continue Mobile DOT without pay, and 8 (50%) reported that they were able to complete Mobile DOT in less than 3 min each day.

Hydroxyurea nonadherence is common among children with SCD [7,24], and an effective hydroxyurea adherence intervention is needed to improve health outcomes. Overall, our results suggest that Mobile DOT improves hydroxyurea adherence, at least temporarily, in children who engage with the intervention. Of the engaged children, two-thirds achieved 80% or higher MPR adherence with Mobile DOT, which is noteworthy considering many of these children had appreciably lower baseline hydroxyurea adherence. Also, median MPR significantly improved with Mobile DOT, despite including some subjects with less potential to improve their MPR because they had 80% or higher adherence at baseline. Finally, although not all the engaged participants achieved 80% or higher adherence with Mobile DOT, many still incrementally increased their adherence (eg, >15% increase or an additional hydroxyurea dose per week), suggesting that Mobile DOT has the potential to improve adherence for some of the most nonadherent children.

Our results also showed that engaged participants’ mean MCV and HgbF levels increased during the Mobile DOT period. Given that MCV and HgbF level changes from hydroxyurea exposure typically occur before clinical improvements manifest [30], this could be an early indication that with continued use, Mobile DOT may have potential hematologic and clinical benefits. Although significant Hgb changes were not appreciated, this may be related to the relatively short duration of time that Mobile DOT was provided compared with the time required to see improved Hgb with hydroxyurea [3] and our participants’ relatively high baseline Hgb levels. In addition, 15% of our engaged patients had either Hemoglobin SC, Sβ⁺, or SE, and past studies suggest that other SCD genotypes may not have the same degree of hematologic response with hydroxyurea compared with those with hemoglobin SS [31,32].

Similar to other medication adherence interventions [33], participants’ adherence and laboratory improvements declined once the intervention was discontinued. Survey results showed that many engaged participants were willing to continue Mobile DOT beyond 6 months, suggesting that for some, Mobile DOT may be able to be used longer term to sustain adherence. However, short booster sessions or additional interventions may be needed for others when their adherence declines. Laboratory improvements likely returned to baseline because adherence declined but also the median prescribed hydroxyurea dose decreased by the end of the study. Mean ANC did not change during the study to suggest that severe myelosuppression frequently occurred that would have necessitated this decrease. Most participants were also prescribed doses at the end of the study that were similar to their baseline doses. We, therefore, suspect this decrease was because many participants were growing children and the hydroxyurea protocol [20] and institutional guidelines do not provide guidance about how frequently to increase hydroxyurea to account for weight gain from growth.

Future studies are needed to definitively determine if Mobile DOT reduces erythrocyte transfusions, hospitalizations, and ACS and VOC episodes because these events occurred infrequently at baseline and participants were followed for a relatively short period of time. To limit underestimating the benefit of effective adherence interventions, future studies should also measure their impact on preventing other patient-reported and costly outcomes, such as VOC episodes, that do not result in hospitalization [34] and stroke [35] that may be ameliorated or prevented with hydroxyurea.

Several limitations must be considered when interpreting our results. First, despite approaching almost all the eligible patients at NCH, we were unable to enroll our projected sample size [19]. This occurred because we anticipated that more patients at NCH were going to initiate hydroxyurea and become eligible than actually did and because some patients perceived the intervention as too burdensome. These findings underscore the need for interventions, such as shared decision-making tools [36], to increase hydroxyurea acceptance and add to the existing literature that suggests that the time and burden to participate in clinical studies for patients with SCD are prohibitive [37]. In the future, we plan to recruit patients outside of the clinical care setting to avoid approaching patients when they may be overwhelmed by other aspects of their SCD care. This strategy could also increase the number of nonadherent patients that are recruited because clinic nonadherence is associated with hydroxyurea nonadherence [38].

Second, only two-thirds of the enrolled participants remained engaged with Mobile DOT. As only those who engaged improved their adherence, it is possible that the factors that promoted engagement may have contributed to the observed adherence gains. Therefore, identifying these mediating factors is paramount to increasing the impact that this and other adherence interventions may have on children with SCD. For example, our results suggest that nonengaged participants were more likely to be female and have experienced a recent SCD complication (eg, erythrocyte transfusion and/or ACS episode) compared with engaged participants. The small sample size and small number of clinical complications that occurred during the study limit the ability to definitively determine if these factors influence engagement, if they were related to an unmeasured moderating factor, or if they were merely a random finding observed in this sample. In addition, consistent with the literature that suggests widespread electronic device ownership among African Americans [39], including children with SCD [13], only 3 patients reported that not owning a device prevented them from participating or led to their study withdrawal. However, increasing literature suggests that having an insufficient data plan or inconsistent access to Wi-Fi are common among minority populations [40] and could explain why multiple participants became nonengaged and stopped submitting videos or responding to the research team shortly after they enrolled. Although providing devices and cellular data could potentially increase enrollment and engagement with electronic health interventions and ensure consistent intervention delivery, this could also add significant costs that may limit intervention sustainability and use in clinical practice. Finally, although our study suggests Mobile DOT increases hydroxyurea adherence, it is possible that the Hawthorne effect may have also contributed because participants were aware their adherence was going to be closely monitored.

Participants’ feedback from this study provides insights into modifications that may make this and other adherence interventions more engaging, reduce attrition, and increase adherence in future studies. First, changing the frequency that medication reminders alerts are provided could reduce alert fatigue. In addition, providing feedback that notifies participants how their adherence compares with their previous adherence and with others’ could influence participants’ perception of the intervention. Incentives may be needed to reward both high adherence and incremental adherence improvements to encourage the most nonadherent participants to continue to engage with the intervention. Finally, for multidimensional interventions, quantifying each component’s impact on adherence could identify if components need to be eliminated or optimized or if additional components, such as gamification, may be needed because they might promote engagement with the entire intervention [41].

MPR was used to measure hydroxyurea adherence in this study. Although the medication event monitor system caps are considered the gold standard adherence measure [42], these devices were not used in this study because they are not compatible with liquid hydroxyurea formulations, which are commonly used in young children. It is important to note, however, that MPR only measures access to medication and not actual medication administration and can also be unreliable if dispensing reports are inaccurate or incomplete [21]. Engaged participants’ hydroxyurea adherence by MPR and VDOT were similar in this study, but for patients who may have accumulated a supply of hydroxyurea from past prescriptions, MPR may underestimate adherence. Ultimately, studies that validate hydroxyurea adherence measures are needed to accurately identify nonadherent children who could benefit from an adherence intervention. Furthermore, other hydroxyurea adherence interventions are under study, including an electronic alert intervention [43] and a community health care worker intervention [44], and show feasibility, and valid measures will be needed to determine which intervention is most effective.

Finally, systematic reviews suggest there are insufficient data to determine if mobile health interventions, such as Mobile DOT, are cost-effective [45,46]. Although an increasing number of insurers are already providing financial incentives to their members to encourage healthy behaviors [47], to employ insurers to provide this health behavior intervention in clinical practice, future studies will need to evaluate if it is economically feasible and sustainable. We acknowledge that the cost to improve hydroxyurea adherence may be significant, considering poor adherence is difficult to improve and sustain. However, the cost to provide this intervention could be reduced if technology improvements reduce the time and resources required for implementation and if it is reserved for patients with adherence challenges. Ultimately, we suspect that because SCD is a chronic disease that results in costly complications that are reduced with effective hydroxyurea use [48], even relatively costly adherence interventions may still be implementable, and perhaps cost-effective, in clinical practice.

In conclusion, we demonstrate improved 6-month hydroxyurea adherence, MCV, and HgbF level among children with SCD who engaged with an electronic, multidimensional hydroxyurea adherence intervention. Future studies that determine how to increase patient engagement with hydroxyurea adherence interventions, how to sustain improved adherence, and how these interventions impact the outcomes of children with SCD are warranted. If successful, they may be generalizable to improve medication adherence and outcomes of other chronically ill populations.