Abstract
The aims of this study are to: (1) examine the preliminary utility of the Self-Management and Research Technology (SMART) pilot project, (2) identify which adolescents were most likely to benefit from participation, and (3) examine interview feedback to inform future program iterations. Twenty-three adolescents (M age = 15.13 years) were enrolled in the six-week text message pilot program consisting of daily interactive blood glucose (BG) prompts and type 1 diabetes-related educational text messages. Medical charts were reviewed for hemoglobin A1c and to corroborate medical record and demographic data. Glucometer data were downloaded to calculate an average monthly BG level and daily BG monitoring frequency. No statistically significant improvements were observed pre-intervention to post-intervention in glycemic outcomes. Participants with a high text message response rate were more likely to demonstrate improvement in average monthly BG levels and daily BG monitoring frequency than those with a low text message response rate. Participants reported satisfaction with the text message program. The text message-based SMART pilot project demonstrated preliminary efficacy for a targeted subset of adolescents who were engaged with the program. Continued research with a larger sample and longer trial duration is warranted to evaluate the potential utility of text message-based interventions.
Keywords
Introduction
It has been estimated that less than 25% of adolescents with type 1 diabetes (T1D) meet the International Society for Pediatric and Adolescent Diabetes as well as the recently revised American Diabetes Association’s recommendation for target glycemic control: hemoglobin A1c (HbA1c) under 7.5% (Wood et al., 2013). Interventions utilizing text messages sent directly to personal cell phones may be an ideal method to reach such otherwise hard-to-reach adolescents with T1D and assess and facilitate appropriate diabetes self-care. Text messages are a useful tool for assessing general adolescent health, and the use of targeted text messages has improved adherence in adolescent populations with varying chronic illnesses (Miloh et al., 2009; Rhee et al., 2014).
Text message-based interventions for adolescents with T1D are feasible, yet their impact on glycemic control is unclear (Herbert et al., 2013). Intervention targets are varied, and research has not found consistent gains in HbA1c, blood glucose (BG) monitoring, or insulin injection rates (Bin-Abbas et al., 2014; Franklin et al., 2006; Froisland et al., 2012; Mulvaney et al., 2012; Rami et al., 2006) That being said, many of these studies were preliminary trials with limited information regarding participant engagement with the program, such as receipt of text messages and/or participants’ response rates, some studies included non-text message intervention components that complicated the evaluation of the impact of the text message component on health behavior change, and other studies provided cell phones as an incentive to participate, which may have influenced adolescents to participate and inflated participant engagement with the program. Further systematic evaluation of adolescent engagement and key components associated with health behavior change is needed.
The Self-Management and Research Technology (SMART) project was developed to provide adolescents with T1D self-care reminders and education via daily text messages. Pilot program development and initial feasibility has been reported previously (Herbert et al., 2014). On average, adolescents responded to 78% of program text messages, and girls responded to more text messages than boys. The present study aimed to: (1) evaluate the impact of the SMART project on T1D self-care behaviors, (2) identify adolescent characteristics associated with program engagement and health behavior change, and (3) evaluate participant experience with the SMART project.
Methods
Participants
Participants were adolescents aged 13–17 years who were diagnosed with T1D for at least one year followed by the endocrinology department at a Mid-Atlantic children’s hospital. Inclusion criteria were as follows: multiple daily injections or insulin pump; cell phone with an unlimited text message plan; English fluency; adolescent assent and parent consent; and the absence of severe psychopathology, developmental, and/or physical disabilities.
Procedure
Research procedures were approved by the institutional review board. Potentially eligible participants were mailed a recruitment letter with an opt out postcard. Those not returning the postcard were contacted by phone one to two weeks later to assess eligibility and interest. Interested participants met with a research assistant at a routine T1D clinic appointment to consent to participate and complete baseline procedures. Participants were then registered in the SMART project, a two-way text message software program contracted through Reify Health (Baltimore, Maryland, USA), and instructed to choose morning, afternoon, evening, and weekend midmorning times to receive SMART project text messages.
During weeks 1 and 6 (assessment weeks), the participants were prompted to text message their BG levels three times/day (morning, afternoon, and evening). During weeks 2 to 5, adolescents received two text messages/day regarding BG monitoring (week 2), nutrition (week 3), physical activity (week 4), and sleep/mood (week 5). Each text message during weeks 2 to 5 included T1D-related information and a specific question that necessitated the participant reply via text message.
Following the six-week text message portion of the SMART project, adolescents completed follow-up questionnaires via REDCap (Harris et al., 2009) and a 30-minute phone interview. HbA1c and glucometer data were collected from the medical chart at baseline and the diabetes clinic appointment immediately following SMART project program completion. Participants received modest incentives (US$25) for baseline and follow-up data collection.
Measures
Sociodemographic and medical questionnaires
Demographics-General Information and Medical Information Questionnaires were completed by adolescents.
Glycemic control
HbA1c is the most widely accepted measure of average BG over the preceding two to three months (American Diabetes Association, 1994). All assays were conducted with the DCA 2000 analyzer (Siemens/Bayer, Munich, Germany), using high-performance liquid chromatography to assure comparability between subjects (Tamborlane et al., 2005). Baseline HbA1c was obtained for each participant at the clinic appointment that was closest in time to their baseline, reflecting participants’ glycemic control during the two to three months immediately prior to participation (American Diabetes Association, 2014). Follow-up HbA1c was obtained for each participant at the clinic appointment following intervention completion.
BG data
Participants’ glucometers were downloaded at the clinic appointment during which they were oriented to the SMART project in order to obtain baseline BG data regarding the 30 days prior to their enrollment. Participants’ glucometers were downloaded again at the clinic appointment following intervention completion. These data were used to calculate average daily BG level and average daily BG monitoring frequency.
Response frequency
Responses to the SMART project text messages were collected via the Reify Health software program. Overall and weekly response rates were calculated as the percentage of text messages to which participants responded. During weeks 1 and 6, adolescents were asked to respond to a total of 21 texts each week. During weeks 2–5, adolescents were asked to respond to 14 texts each week. Overall, adolescents were asked to respond to 98 texts.
Qualitative interviews
Adolescents completed a 30-minute qualitative interview via phone with a member of the study team. Participants provided feedback regarding their experiences with the text message program, including intervention impact, content, and design. Questions were open ended, such as ‘Tell me about your experiences with this study’ and ‘Did participation in the study affect how you manage your diabetes?’
Data analyses
Statistical analyses were conducted using SPSS 22 edition. Descriptive statistics were generated for adolescent demographic and medical characteristics, text message response frequency, average BG level, and average daily BG monitoring frequency at baseline and follow-up. Change scores were calculated for HbA1c, average monthly BG levels, and average daily BG monitoring frequency. HbA1c change was categorized as improved (≥.5% decrease), stable (≤.5% change), or worsened (≥.5% increase). Average monthly BG change was categorized as improved (≥15 mg/dL decrease), stable (≤15 mg/dL change), or worsened (≥15 mg/dL increase). Average daily BG monitoring frequency change was categorized as improved (≥1 check/day increase), stable (≤1 check/day change), or worsened (≥1 check/day decrease). Text message response frequency was dichotomized into high responders (≥85%) and low responders (<85%).
Correlational and χ 2 analyses were conducted to assess whether demographic and medical characteristics (age, sex, race, and insulin regimen) were related to changes in glycemic control and/or BG monitoring. Independent samples t-tests were conducted to evaluate text message response rate in relation to changes in glycemic control and/or BG checks. Interview content was transcribed and reviewed, and frequency counts were calculated regarding specific questions about intervention impact, content, and design.
Results
Participants
Ninety-seven patients were initially identified as eligible for participation via clinic list review and received information about the SMART project by mail. Fifty-three patients (55%) were able to be contacted and assessed for eligibility, while 17 declined, 10 were ineligible, and 3 provided verbal consent but did not complete an orientation session. Reasons for declining to participate in the project included time commitment concerns, residence in a location with inconsistent cell phone service, disinterest in text messages, and desire for a larger incentive. Thus, the total sample included 23 participants or 54% of the patients who were eligible and contacted. Participants’ mean age was 15.13 years (standard deviation = 1.14). The sample comprised 75% Caucasian; 13% of participants identified as African-American, 4% as Hispanic, 4% as Caribbean-American, and 4% as East African, which is more racially diverse than most diabetes clinic populations (SEARCH for Diabetes in Youth Study Group et al., 2006). Sixty-one percent were female, and 61% of participants used an insulin pump diabetes regimen.
Follow-up questionnaire data and HbA1c data were available for all participants; however, glucometer data were available only for 18 participants. There were two reasons that glucometer data were unavailable for the other five participants: (1) the glucometer had not been downloaded during the participant’s endocrinology clinic appointment and the study team could not reach the family and (2) the date on the glucometer was incorrect so it was not possible to accurately assess the BG levels during the post-intervention time frame. Baseline HbA1c was lower (i.e. better glycemic control) for participants who had follow-up glucometer data than participants who did not have follow-up glucometer data, t(22) = 4.23, p < .001. Other demographic and baseline medical characteristics were not related to the availability of follow-up glucometer data. All 23 participants were included in HbA1c analyses, but only the 18 participants who had glucometer data for baseline and follow-up were used in average monthly BG level and BG monitoring frequency analyses.
Glycemic control
Table 1 presents descriptive statistics of medical characteristics and T1D self-care. From baseline to follow-up, mean HbA1c decreased (improved) by .06% (N = 23; 6 participants improved, 11 were stable, and 6 worsened). The mean average monthly BG level decreased (improved) by 4.95 mg/dL (n = 18; 7 improved, 7 were stable, and 4 worsened). Average daily BG monitoring frequency decreased (worsened) by .07 checks per day (n = 18; 4 improved, 11 were stable, and 3 worsened).
Baseline and follow-up medical characteristics.
Note: M: mean; SD: standard deviation; HbA1c: hemoglobin A1c; BG: blood glucose. Glucometer data were only available for 18 participants at follow-up.
Factors associated with change in glycemic control
There was a significant difference in whether adolescents’ average daily BG monitoring frequency improved, remained stable, or worsened by sex, χ 2(n = 18) = .18; p < .05. Girls were more likely to maintain their average daily BG monitoring frequency, whereas boys were more likely to improve or worsen their average daily BG monitoring frequency. No other analyses regarding demographic and medical characteristics (age, race, and insulin regimen) were significant.
Adolescents were classified as high responders (≥85%; n = 13; mean response rate = 95.47% ± 3.83) or low responders (<85%; n = 10; mean response rate = 54.82% ± 17.44). There was a trend for high responders to exhibit better glycemic control at baseline than low responders, t(21) = 1.98, p = .06; mean HbA1c of 8.12% versus 9.15%. High responders were more likely to improve their average monthly BG level than low responders, 27.15 mg/dL decrease versus 14.3 mg/dL increase; t(16) = −2.50, p = .02. There was a trend for high responders to improve average daily BG monitoring frequency as compared to low responders, .32 check/day increase versus .61 check/day decrease; t(16) = 1.73, p = .10. HbA1c change was not related to text response rate.
Interviews
Twenty-one participants completed an open-ended follow-up interview to further evaluate their experiences with the SMART project and their perceptions regarding its impact on their T1D management. Responses were allocated into three categories: intervention content, impact, and design. With respect to intervention content, half of the participants preferred information-prompting text messages, whereas a third preferred educational tips. Four adolescents reported they liked both types of text messages equally. Opinions regarding T1D topics varied. Half of the participants indicated that the physical activity text messages were most enjoyable and helpful. Adolescents were relatively less interested in text messages about mood and sleep.
Regarding intervention impact, most adolescents (79%) expressed interest in receiving text messages after the SMART project was completed. Many (70%) reported that the text messages positively changed their T1D management behaviors. For example, one participant stated, ‘I noticed that my blood sugars were a lot better; they were definitely more in range than they were before I started’. Another teen cited the educational aspect as valuable, ‘They taught me how much exercise I was supposed to get, the importance of counting carbs, and … the amount of times I’m supposed to check’. Many teens also reported that the SMART project gave them an opportunity to reflect on their BG patterns throughout the day and made them more aware of the way their BG levels were affected by diet and physical activity.
Finally, regarding intervention design, adolescents liked that they were able to tailor the timing of their text messages because it was convenient for their schedule. Many participants believed that parent and/or peer involvement could be helpful and several endorsed the use of text messages as a communication method with their endocrinology clinic for appointment reminders, prescription refills, and/or sending BG levels to parents. Over 80% indicated that they wanted parents to be included in future SMART project iterations in order to receive educational T1D tips and T1D management reminders. Finally, over half of the participants (57%) wanted their close friends involved in the SMART project so that they could learn about T1D.
Discussion
The purpose of this study was to evaluate the utility of the SMART project, a brief text message-based pilot intervention for adolescents with T1D. There were no statistically significant improvements in any of the glycemic control measures from baseline to follow-up. Adolescents who were more engaged were more likely to evidence T1D behavior change.
The lack of statistically significant changes from pre-intervention to post-intervention may be due, in part, to the intervention’s limited time frame and small sample size. HbA1c represents average glycemic control over the past 8–12 weeks, so significant HbA1c changes may not be evident immediately after a six-week intervention. Previous descriptive research among adolescents with T1D also has demonstrated a decline in diabetes self-care, so the overall stable glycemic control observed in this study suggests that the SMART project may have helped to maintain adolescents’ T1D self-care and related glycemic control. Other behavioral interventions among adolescent T1D samples have noted similar stabilization of glycemic control (Holmes et al., 2014). Text messages may be a translatable and cost-effective method of intervention, particularly as compared to more intensive behavioral interventions that require frequent in-person visits.
Adolescents who engaged more with the program, as measured by text message response rate, showed a greater improvement in glycemic control indicators than those who engaged less with the program. There was also evidence to suggest that adolescents with better glycemic control at baseline were more likely to engage in the program. Adolescents who are in adequate glycemic control may be more responsive to T1D-related text messages and adolescents who are more engaged with the program may be more motivated to change their health behavior.
Participant feedback was reviewed to inform intervention refinement. In general, participants enjoyed the SMART project, thought it was beneficial for their T1D self-care, appreciated tailored text message scheduling, and were interested in expanding the program to parents. There were varying preferences regarding text message content. Continued customization of text message format and topics will likely optimize participants’ experiences.
One of the primary unique features of the SMART project is that it focused solely on the use of text messages as a health behavior change agent for adolescents with T1D, rather than including text messages as one component of a broader intervention. Thus, the results provide insight regarding the potential impact of text message inclusion in broader intervention trials. Another strength is that the SMART project did not provide cell phones to participants as an incentive, suggesting that adolescents participated because they were interested in the intervention rather than as a means to access a cell phone.
This pilot project has a number of novel features that can be incorporated into future text message-based interventions as well, including the ability to tailor text messages to match participant schedules and analyses of participant- and time-specific text message response rates. There was high satisfaction in a sample that was more racially diverse than typically observed in T1D clinics (SEARCH for Diabetes in Youth Study Group et al., 2006). However, the project was a pilot study with a relatively small sample of adolescents, and follow-up glucometer data were only available for 18 participants, so the results must be interpreted with caution. The intervention occurred over a six-week period, which may not have been long enough to elicit significant changes in adolescent T1D behavior. Follow-up data were collected at one time point, so long-term and/or delayed program impacts may have been missed.
This study was a pilot, and replication among a larger, more diverse population is warranted to more thoroughly examine generalizability and subgroup differences. Expansion of the six-week program with longer term follow-up is also indicated. Future iterations of text message interventions for this population would likely benefit from inclusion of a parent and/or peer component, continued customization of the text message schedule, and increased engagement through the inclusion of fun facts and trivia questions. These simple measures may improve participant participation and the impact of the intervention on T1D self-care behaviors. A control or comparison group is also important to include in future investigations.
The SMART project was enjoyable and effective for adolescents with T1D who were interested in health behavior change. Continued research with a larger sample and longer trial duration can further examine the long-term utility of text message-based interventions for this population. It is likely that text message-based interventions will be most useful and effective when used as a reinforcing or augmenting component of a broader intervention that includes in-person or phone intervention sessions with a trained health professional.
Footnotes
Acknowledgment
The authors wish to thank Catherine Gillespie, PhD, for assistance with data analyses.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.