We used an interrupted time-series design to examine the impact of the text message intervention, deployed in an adolescent medicine clinic serving youth aged 12 to 24 years. Baseline service levels were documented during a 35-week lead-in period (weeks 1-35). This was followed by a 20-week period when patients were asked to send text messages to 5 friends who they believed were sexually active and lived in the clinic’s service area (weeks 36-55). Following the intervention was a 16-week period of continued clinic observation (weeks 56-72). Chart abstraction from intake forms continued during the entire study (weeks 1-72).

The study was approved by the University of California, San Francisco, Institutional Review Board (approval #12-08516).

Our target was 100 participants to ascertain feasibility and acceptability. Of the 153 patients approached by research staff, 28 (18%) refused, 25 (16%) did not have a mobile phone with them, and 100 (65%) provided informed consent and sent text messages. Participants were told that the goal of the text messages was to encourage their friends to visit the clinic and receive STI screening. Participants were provided a text messaging guide that reviewed considerations for developing messages (eg, let your friend know you care about them). Participants developed their own message. Subsequent to sending the messages, participants were provided US $10 for completing a short posttexting interview. The text message and responses from friends while the participant was still in the clinic were recorded by research staff. Staff recontacted participants 1 week after their clinic visit to ascertain issues with friends following the intervention (n=93).

Analyses compared the 35-week “pretexting” data to the 36-week postintervention period. Additional comparisons accounted for seasonal fluctuations in service levels by comparison to the same dates in the preceding year. We examined changes in the number of new patients seen (including patients not seen for at least 12 months), the number of STI tests conducted and positive results, the number of HIV tests conducted and positive results, and the number of at-risk patients seen at the clinic. Patients were characterized as at risk if they reported multiple partners, infrequent condom use, or had an STI at intake. Counts for each outcome were aggregated into average monthly totals, and analyses were performed using a generalized linear mixed model. Outcomes were modeled as overdispersed Poisson distribution variables. For clarity, only the mean of the counts during each time period was provided.

All participants sent the required 5 messages (see Textbox 1 for examples of messages sent and received), and 18% (n=18) of participants sent more than 5 messages. In the brief posttexting interview, most patients (n=55, 55%) reported no concerns with sending the text message, 18% (n=18) expressed worry about their friends’ response, and 27% (n=27) expressed concern that their friend would not heed their advice (eg, their friend would not call the clinic or would not listen). In the 1-week postmessaging interview, no youth reported adverse events or negative outcomes from sending the message.

A total of 430 new patients were seen at the clinic during the 72 weeks of data collection, of whom 338 (79%) were female. New patients were 61% (n=262) African American and 17% (n=73) Hispanic, had a mean age of 19.6 (SD 2.6, range 13-25) years, and primarily identified as heterosexual (n=375, 87%). Almost half of the patients reported they had no regular doctor (n=185, 43%) or health care (n=194, 45%).

In a brief posttexting interview with the 100 participants, most patients (n=55, 55%) reported no concerns with sending the text message, 18% (n=18) expressed worry about their friends’ response, and 27% (n=27) expressed a concern that their friend would not heed their advice. In a follow-up interview 1 week after messaging, no youth reported adverse events or negative outcomes from sending the message.

We examined changes in the number of new patients seen, the number of STI tests conducted and positive results, the number of HIV tests conducted and positive results, and the number of at-risk patients seen at the clinic. Patients were characterized as at risk if they reported multiple partners, infrequent condom use, or any STI infection. All clinic services outcomes increased during the posttexting period compared to the pretexting period (Table 1), including weekly STI tests (4.11 vs 5.18, P=.58), weekly number of youths testing positive for STI (0.60 vs 0.76, P=.17), weekly number of youths tested for HIV (0.91 vs 1.35, P=.48), and weekly number of patients reporting high-risk sexual behaviors (4.89 vs 5.29, P=.55). However, these findings did not achieve statistical significance. In order to account for possible seasonal and period effects, we also examined the means for a subset of the preintervention phase that matched the calendar months covered in the postintervention phase but occurred 1 year earlier. Findings were similar, suggesting that the variation in average patient counts between the pre- and postintervention periods was not due to seasonal variation.

Given the low number of youths accessing health care services and STI/HIV screening, new strategies are needed to address the barriers that exist. We found that text messaging between peers is a feasible and acceptable strategy with potential for increasing STI/HIV testing. We did not provide standard messages; rather, we were interested in whether youth were able to develop their own persuasive messages. Most youths agreed to send a message and did not express concern or experience complaints from peers. One in 5 participants sent more than the required number of messages. The approach appears feasible and acceptable. The simplicity and low-resource requirements of this approach also make it more sustainable and significantly less costly than recent efforts to build smartphone apps, social media campaigns, or other technology-based interventions to increase testing [18].

We were unable to examine if messages were read or to directly ask recipients about their reactions. Given that over 97% of text messages are opened and 90% are read within 3 minutes of their delivery [22], we did record the immediate response from message recipients. There were few negative text responses, and no negative consequences or adverse events reported by participants during the follow-up interviews. This suggests that the intervention may also be acceptable to the recipients of the message.

This study was a proof of concept, and, thus, has limitations in interpretation. The study had a small sample size and was conducted at a single adolescent medical clinic. There may have been selection bias where youths who were more comfortable discussing sexual health with their friends may be overrepresented. However, it is likely that a formal implementation of this intervention would also be on an opt-in basis, strengthening the external validity of our pilot.

While this pilot study was not powered to detect significant differences over time, we did find encouraging increases in the number of weekly STI tests conducted, the number of youths testing positive monthly for STIs, the number of youths tested monthly for HIV, and the number of patients reporting sexual risk behaviors. Thus, text messaging may have increased the number of youths accessing services and STI/HIV screening and reached those engaging in risky sexual behavior who are most in need of screening. This is particularly noteworthy given the clinic primarily serves a low-income population of Black and Hispanic adolescents, a population disproportionately impacted by STI and HIV [23].