Sweet Talk is a novel intervention for supporting young people with type 1 diabetes through text messaging. The intervention is informed by social cognitive theory, which states that health behaviors are influenced by self-efficacy, or the belief in one’s ability to perform actions that will influence outcomes [11], which, in turn, is influenced by goal setting and social support [12,13]. The system was designed to deliver a unique form of push support [14], in contrast to conventional support groups and websites where users have to actively access a site to read messages [15], thus favoring motivated patients and potentially enhancing health inequalities [9,16,17]. The system contains a database of text messages, including information, tips, and reminders categorized according to the main diabetes self-management tasks of insulin injections, blood glucose testing, healthy eating, and exercise. Messages are automatically scheduled based on patient profiles (age, gender, and treatment regimen) and personal diabetes self-management goals created at each clinic visit (healthy eating, exercise, insulin injections, and blood glucose testing). Such personalization appears fundamental to behavioral support interventions [18-20]. Patients receive a weekly text message reminder of their personal goal and a daily text message from the database, thus receiving either one or two messages daily. In addition, patients receive occasional text “newsletters” relating to topical issues and asking questions about their own diabetes self-management routine. They are encouraged to send in messages containing information or questions related to their diabetes self-management. Patients’ ideas and responses that are felt likely to be of general interest are forwarded anonymously to the whole group in order to develop a sense of community among the participants while avoiding the risks of unmoderated peer-to-peer networks, such as the sharing of health-harming practices and text bullying [7,21,22]. The different message types are explained further in a previous paper describing the development of the system [14]. Since Sweet Talk aims to motivate effective self-care, it may also be regarded as a type of “persuasive technology” [23]. Figure 1 represents the theoretical model of the intervention, and a screenshot of the Web-based interface is shown in Figure 2.

Figure 1

Theoretical basis of the Sweet Talk intervention

The subjects were 64 boys and girls aged 8-18 years with type 1 diabetes participating in the intervention arms (Sweet Talk plus conventional therapy n = 33; Sweet Talk plus intensive therapy n = 31) of a three-arm clinical trial during a 12-month period between October 2002 and March 2004.

Written informed consent was obtained from patients and their families, and the study was approved by The Tayside Committee on Medical Research Ethics. Participating patients received a pay-as-you-go mobile phone and a £10 phone card, and incoming text messages to the Sweet Talk system were free of charge. Mobile phones could also be used for personal use. Patients allocated to Sweet Talk were given an information card highlighting the messages that they could expect to receive and suggestions of how they could use the system (Figure 3).

Figure 3

Sweet Talk information card

All of the text messages sent to and from the Sweet Talk system were recorded over the 12 months of the study. This produced observational data on messaging patterns, which could be triangulated with patient clinical and demographic data, as well as message transcripts. Post-hoc analyses for associations between message content and demographic variables were undertaken using chi-squaretests for categorical variables and analysis of variance (ANOVA) for continuous variables.

Message transcripts were analyzed by VF using the constant comparative method in order to generate descriptive themes [24]. This process was facilitated by Nvivo textual analysis software (QSR International, Doncaster, Australia). AG independently analyzed 10% of the messages in order to validate the themes identified by the first rater, and inconsistencies were resolved through discussion. Further verification was achieved through team-based review of identified themes and representative raw data by VF, AG, and CP. Transcripts were then content-analyzed by VF according to these thematic categories [25].

All but 4 of the 64 patients allocated to the Sweet Talk intervention submitted one or more text messages during the 12 months of the study. A total of 1180 messages were submitted, representing an average of 18.4 messages per patient. However, total messaging varied widely between individuals, from 0 to 240 (median 6), and the distribution was skewed by 5 patients who contributed 52% (614/1180) of the messages (Figure 4). A significant proportion of these messages were from 2 boys who sent in very regular blood glucose readings, comprising 338 of the total 1180 messages received (29%).

No participants took the opportunity to use the birthday reminder service or to nominate family or friends to receive patients’ goal reminders so that they could act as personal supporters—two options offered on the information card (see Figure 3).

Figure 4

Number of messages sent to Sweet Talk during the 12-month study

The content of the text messages that patients sent to Sweet Talk fell into 8 broad thematic categories covering blood glucose readings, diabetes questions, diabetes information, personal health administration, social aspects, technical messages, message errors, and message responses. Illustrative text messages are shown in Table 1. The total exceeds 1180 because 77 messages were coded into more than one category.

Messages containing blood glucose values accounted for 35% of all messages (418/1180). Of these, 56% (232) followed the advice to submit blood glucose values alone (see Figure 3), while the remainder incorporated these values within text. Two boys contributed to 81% of the total blood glucose text messages sent.

Messages containing questions related to some aspect of diabetes self-management made up 6% (74/1180) of all messages. Sweet Talk appeared to provide an opportunity for obtaining information between clinic visits and to send questions that patients may have found difficult to ask in a clinical setting (eg, “Cld DiaBT’s get their belly pierced”). Text messages containing diabetes questions were further categorized into topic themes, as illustrated in Table 2.

Messages containing information about a patient’s own diabetes self-management or health status accounted for 4% (50/1180) of all messages (eg, “I hav managed 2 change my injection site 4 a few days now! =)”). Sweet Talk also provided an outlet for expressing frustration with their diabetes. One “emergency” message was received: “I got ketones bloods r up HELP.” This message was sent despite clear instructions on the information card that the Sweet Talk system was not intended for this use and that patients should continue to use our emergency help line. Telephone follow-up revealed that the patient knew this but simply wanted to know what would happen if he sent a message of this kind.

Patients were encouraged to use the Sweet Talk system as an easy method of contacting the diabetes team with any requests. Of the total messages, 5% (59/1180) contained requests for supplies such as insulin pump consumables, blood glucose meters, and insulin travel authorization letters (eg, “Could I have new meter because it went through the wash on holiday”) and requests for information about clinic appointments (eg, “Hi, can you please tell me when my next clinic appointment is. Thank you”).

Messages of a social nature made up 6% (75/1180) of patients’ incoming messages. Although not directly related to diabetes, these messages provide insight into how patients integrated the system into their daily lives and its value as a source of social support (eg, “Just ate an ice-cream and done a dual wave. Off to colosseum!” and “Happy xmas 2 every1 at 9wels ”). Post-hoc analyses for associations between message content and demographic variables identified gender differences: females sent significantly more messages containing information and questions unrelated to diabetes (females: mean 1.53 ± 2.51; males: 0.09 ± 0.30; P = .002).

Messages about technical aspects of the Sweet Talk system accounted for 7% (86/1180). Of these, most were related to difficulties with message transmission and cost of the messages (n = 55). A further 18 messages indicated problems with the content of the Sweet Talk messages, highlighting failures in message personalization or not understanding the messages (eg, “I keep getting messages about injections but I’m on the pump”).

Of the total messages, 19 (2%) appeared to have been sent to Sweet Talk in error (eg, “Nyt nyt Dad”).

Messages that were sent by patients in direct response to a Sweet Talk text message made up 40% (472/1180, Table 3). Of these, the sporadic text message newsletters generated the most responses (40%, 190/472). For example, one message that asked patients what symptoms they got when their blood sugars were running high provoked responses such as “I get thirsty and a dry throat when I’m high. I also can get a bit moody.” The four newsletters that triggered the greatest flurry of responses had updates on diabetes research, raised the issue of a chocolate manufacturer offering tokens for sports equipment, and reported about a film star and a soap opera character with diabetes. The remaining messages were in response to the daily scheduled messages (30%, 142/472), personal messages (25%, 118/472), and the weekly goal reminder (5%, 22/472). There was a significant correlation between the number of messages patients sent to Sweet Talk and the number of individual response messages they received (r = .521, P = .01).

While the primary intended function of Sweet Talk was to deliver passive support to patients, most participants in this study took the opportunity to submit messages to the system. Analysis of these messages has provided insight into the ways users may adapt text messaging interventions to best serve their needs. Although average messaging frequency was low, there was wide variation among participants, with most messages submitted by 5 power users. No associations were found between total messaging frequency and clinical or psychosocial measures. The content of patients’ messages fell into 8 broad categories covering submission of blood glucose readings, questions about diabetes treatment or lifestyle, information about diabetes self-management, personal health administration such as supply re-ordering, social messages, technical messages, messages sent in error, and responses triggered by a scheduled Sweet Talk message. Unprompted submission of blood glucose readings was most common, followed by messages submitted in response to a system-generated message. Of the latter, those suggesting that patients share tips and frustrations about diabetes self-management generated the most responses. Diabetes news items also stimulated many responses. Females sent significantly more text messages of a social nature, unrelated to diabetes, than did males. No participants took the opportunity to nominate family or peer supporters to receive their goal messages or used the birthday reminder function.

While the generalizability of the results is limited by the fact that only 5 users accounted for the majority of the messages, most participants interacted with the system during the study. This is consistent with observations of diabetes chat rooms, where only a minority of users post messages but the remaining lurkers read and benefit from other peoples’ messages [17,26]. In a user satisfaction questionnaire, reported separately, most participants indicated that Sweet Talk had helped them to look after their diabetes and that they wished to continue receiving messages at the end of the study period [10].

The lack of association between messaging frequency and clinical or psychosocial measures may reflect our choice of scoring systems, which were largely diabetes centered. Assessing personality measures such as neuroticism, extraversion, openness, agreeableness, and conscientiousness [27] may have revealed important associations and should be considered in future studies assessing the uptake and use of such interventions.

The formative data generated by this study have helped to further our understanding of the fit of this technology with users’ needs, to challenge our pre-existing ideas about how it might support young people, and to generate ideas for refining the service.

Patients’ interactions with the system suggest that many valued the opportunity to engage in reciprocal communication, although not all participants chose to take advantage of this. Patients who more frequently submitted questions to the system inevitably received more individual responses, illustrating how motivated patients may obtain more personalized services despite efforts to design equitable technology-based support interventions.

Submission of blood glucose readings was the most common type of message, supporting results of studies indicating the potential for remote disease monitoring [28]. Using Sweet Talk to request supplies and enquire about appointments also has potential to increase efficiency through avoiding the telephone tag that can occur when health professionals and patients try to communicate between clinic visits [29]. The volume of text messages sent by patients over the year of the study was low, and minimal health professional time was required for correspondence with patients. This is consistent with studies of email consulting, which did not show the expected unmanageable burden of correspondence [30].

Newsletters containing information about topical aspects of diabetes or reports about public figures with the condition may have stimulated responses through reinforcing the sense of community and boosting self-esteem through identification with respected role models. Studies of disease-specific websites and chat rooms indicate that it is the information and companionship components that are most valued by patients [17,26].

Sweet Talk appears to have provided a forum for patients to ask personal or embarrassing questions that they may have felt unable to ask at a clinic visit. This is compatible with research demonstrating the value of computer-based interventions for encouraging disclosure of sensitive information, such as mental health problems [31].

Females’ greater use of the system for social messaging accords with studies of mobile phone and chat room use [6,17,32]. However, social messaging represented far fewer interactions in this study (4%) than in a previous telephone support study by our group, in which it accounted for the majority of talk time [33], possibly reflecting teenagers’ preference for verbal over written communication for social interaction [8].

While ongoing parental and peer support for diabetes self-management is important for optimizing glycemic control in adolescence [34-36], none of the participants in this study took the opportunity to nominate family or friends to receive messages related to their diabetes goals, which contrasts with results in other areas such as smoking cessation [37]. This may have been due to inadequate advertisement or explanation of the concept, lack of interest, or a concern that it would place patients under unhelpful pressure. A previous study in which children’s blood glucose readings were sent to a parent’s mobile phone showed that while this promoted a sense of reassurance in some children, in others it generated a feeling of surveillance and reduced their sense of personal control [38].

Our qualitative and usage data illustrate both individual differences in patients’ propensity to interact with the system and the multiple potential utilities that such complex interventions may provide. Further research to explore patterns of use in different age and clinical groups would be valuable, as would studies of the personal and contextual factors influencing the adoption of such technology.

Inviting patients to join the editorial board of similar text messaging interventions will help to ensure the appropriateness of message content and delivery schedules as well as identify users’ expectations for reciprocal messaging and the likely implications for practitioner time.