How to Lose an App in 10 Days: Exploring Digital Health Engagement in Sickle Cell Disease

Digital health technologies, such as mobile health applications (mHealth) and wearable smart devices, have proven to be feasible and useful for individuals living with sickle cell disease (SCD). Even with the clear benefits of digital health tools, patient engagement with digital health is low. The possible reasons for low engagement include socioeconomic status, race, severity of disease, and digital literacy. Ultimately, digital health interventions struggle with clinical implementation due to a lack of understanding of how to keep users engaged. We aimed to identify patterns in user engagement to better understand why individuals with SCD may or may not engage with digital technology. After approval from the Institutional Review Board, individuals with SCD, ages 18 and older, were approached. Following consent, participants were instructed to use the Nanbar Health App for 6 months, report symptoms in the app at least once a day, and wear an Apple Watch. Data was combined and analyzed using R, with engagement calculated on a per-user level and a per-user-per-variable level in two different ways; lapses in engagement in days and an engagement rate. The lapse in engagement was calculated by determining the chronological gaps in patient reporting. The engagement rate (per-user-per-variable) was calculated by grouping the data by user and dividing the number of reporting days for a variable by the total days in the study. The engagement rate (per-user) was calculated by grouping the data by user, calculating the variable reporting per day, summing the number of reporting days, and dividing by total days in the study. Our team enrolled 12 users to our study (ages 18-43). Average total engagement for mobile app and smartwatch data combined was 57% (SD=18%). Average engagement with the Apple watch was 42% (SD=17%), while average engagement with symptom reporting alone was 31% (SD=21%). The correlation between mobile app engagement and watch engagement was 0.31. Within the Nanbar Health app symptom log page, the most commonly reported items were: 1) mood reported by emoji; 2) symptoms of tiredness; and 3) pain. Lapses in reporting were 3.8 days on average for mood reporting, with tiredness and pain average reporting lapse of 23.8 and 18.9 days respectively. There was no significant association between demographics and engagement. Lapses in reporting pain were most correlated with age and type of insurance at 0.4 and 0.45 respectively. Lapses in tiredness reports were correlated with sex most notably, at 0.39, with age and genotype at -0.31 and -0.32 respectively. Demographics related to lack of combined engagement, however, found no demographics were significantly influential. The number of VOCs was most correlated at -0.44 to reporting within the mHealth app and using the Apple watch. Of note, no significant correlation between engagement with the Nanbar Health app and the Apple Watch. Even though digital technology is frequently considered a single technology, full engagement requires using multiple technologies, which could reduce engagement. Within the Nanbar Health app, reporting mood was the most engaging feature, with a lapse time difference of about 15 days shorter than reporting symptoms. To report mood, a user only needed to hit a singular button and select an emoji. When reporting a symptom, users had to select a minimum of 3 buttons, and then numerically rate the intensity. The observed difference in engagement may be due to perception of non-significant symptoms, however, since mood is consistent, it was more consistently reported. Another consideration for differences in time lapse is the burden of interacting with the app (3 versus 1 button) and the increased demands of needing to numerically rate symptoms versus selecting an emoji. Finally, no strong demographic features correlated with app engagement, however, disease severity was correlated with poor engagement. Number of previous-year VOCs was negatively correlated with engagement, which is consistent with other studies citing disease severity as a negative predictor of engagement. Overall, our findings indicate potential reasons for varied engagement with digital health tools by patients with SCD. Engagement with SCD based digital health tools shows some correlation with disease severity and the design of the tool itself, as increased burden of reporting seems to correlate with decreased engagement with digital health.