Investigation into the Use of Short Message Services to Expand Uptake of Human Immunodeficiency Virus Testing,and Whether Content and Dosage Have Impact

Introduction

South Africa has one of the highest HIV morbidity and mortality rates in the world. ¹ Around 5.7 million South Africans are living with HIV, ¹ and acquired immune deficiency syndrome (AIDS)–related illnesses account for over half of adult deaths. ² Numerous studies have demonstrated the potential benefits of HIV counseling and testing (HCT) in addressing both prevention and transmission, and timely access to HIV treatment, care, and support services. ^3,4 Unfortunately, however, many South Africans have not tested: in the 2008 South African national HIV prevalence survey, only half of respondents indicated that they had tested for HIV. ²

Realizing the potential of HCT, the South African government has shifted from voluntary counseling and testing (VCT) to HCT being part of procedure at all health facilities. Through its national HCT campaign, the government aimed to test 15 million South Africans by May 2011. ⁵ The shift from VCT to provider-initiated HCT is potentially significant as increases in patients tested per healthcare worker have been shown in comparison with VCT. ⁵ The challenge, however, is to get people who do not routinely engage with public healthcare services to present at HCT sites. Reluctance to test is associated with complex barriers ranging from AIDS-related stigmas, ^6,7 to lack of knowledge/misinformation about HIV and AIDS. ^{8 –10}

Innovative means of encouraging people to test for HIV could reach those not presenting for HCT. South Africa has a remarkably high proportion of its population with mobile cellular subscriptions, at 93%. ¹¹ This presents an opportunity to use mobiles to get messages to people to increase HCT uptake and impact on other key health decision making. The extensive use of short message services (SMSs) in South Africa, which are cost-effective to send and are free to receive whether the person has airtime or not, makes interventions via SMS even more appealing. For instance the country's largest network, Vodacom, reports 6 billion messages sent by 26 million subscribers in their 2009/10 financial year. ¹² However, to our knowledge, there is no published literature examining SMS interventions for increasing uptake of HCT in South Africa, and globally only one other study has demonstrated this. In that study, men who have sex with men who had attended a sexual health clinic in Australia were reminded via SMS to return for HIV testing. ¹³

The potential of SMSs for HIV-related communications in South Africa and other African countries has been demonstrated. For example, SMS-based interventions have been successfully tested for promoting antiretroviral drug adherence, ¹⁴ for reducing HIV-related risk behaviors ¹⁵ and for promoting uptake of HIV/AIDS telephonic counseling. ¹⁶ However, a major concern with regard to SMS-based interventions is the variability in the magnitude of study outcomes. ¹⁷ Excluding obvious methodological shortcomings such as statistical underpowering and ineffective outcome evaluation procedures, ¹⁸ studies reported thus far do not permit for inferences to be made regarding the exact components of SMS-based interventions that affect their efficacies. ^17,18

The aim of this study was to investigate the effectiveness of using SMSs to facilitate uptake of HCT in South Africa. We also interrogated the impact of using two forms of SMS content and varying the dosage levels (the number of SMSs) on the efficacy of our SMS-based intervention. The study followed an innovative methodology, where recruitment of study participants, consent, and post-intervention assessment were done via SMS.

Materials and Methods

A randomized control trial study design was used for the evaluation with four experimental options, including both informational (INFO) and motivational (MOTI) content for the messaging, and sets of either 3 or 10 messages being delivered. Two groups of intervention SMSs were created: INFO and MOTI. They were guided by the information-motivation-behavioral skills model of AIDS risk reduction, ¹⁹ which hypothesizes that behavioral skills, information (knowledge), and motivation (attitudes and beliefs) are critical in influencing behavioral changes. This model has been used to understand and influence various AIDS-related behaviors. ^20,21 A brief literature search of published barriers to uptake of HCT ^6,22 was incorporated in the drafting of the SMSs. See Figure 1 for the overall design.

OPEN IN VIEWER

Fig. 1.

Flow diagram of paticipants through the study.

Participants who opted into the study were allocated to five groups: the control, INFO-3, INFO-10, MOTI-3, and MOTI-10. The MOTI-3 and INFO-3 groups each received 3 SMSs encouraging them to test for HIV, while MOTI-10 and INFO-10 each received 10 such SMSs. The SMSs were sent using Mobilisr, a Web-based system that allows for prescheduling and sending of SMSs in bulk. ²³ They were sent 3 days apart; hence, the groups that got three SMSs received them over 9 days, and the groups that got 10 SMSs got them over a month. A week after completion of each SMS intervention, an SMS reminding the recipient to go for HCT was sent. The control group did not receive any intervention SMSs.

Three weeks after the intervention for groups that received 10 SMSs, an SMS was sent requesting all groups to indicate whether they tested or not (see Table 1 for details of the SMSs). Two separate “please-call-me” (PCM) lines were set up: one captured PCMs from participants indicating that they had tested since the start of the intervention, while the other captured PCMs from participants who had not tested since that date. All participants who responded with PCMs were compensated with R10 (approximately $1.40) worth of airtime.

Results

Overall, 30.2% of participants self-reported having tested since the start of the intervention SMSs, while 22.3% reported that they did not test. However, a small proportion of participants (2.0%) self-reported as having tested for HIV as well as not having tested. Counts of the testers, nontesters, and the erroneous testers are shown in Figure 1. Overall, there was a retention rate of 54.1% (i.e., 54.1% indicated whether they had tested).

Discussion

This study demonstrates the potential for SMSs to communicate motivational messages and information to influence people to test for HIV. The results are preliminary and further investigation is required for this to be included more broadly in campaigns. This study clearly indicated that the content of the messages is of key importance, and that there is a threshold number of SMSs needed for positive impact.

While there is a growing body of studies investigating the use of SMSs for health promotion and management, ^17,18 literature about factors that are critical to the efficacy of SMS-based interventions is relatively sparse. ¹⁷ This study evaluated both dosage and content of SMSs. Our finding that MOTI messages are more effective than INFO ones for increasing uptake of HCT is supported by literature on barriers to testing. ⁷

The threshold effect of 10 SMSs was effective while 3 were largely ineffective is puzzling. It may be that there is a resistance to testing and the consistency of the 10 SMSs arriving every 3 days did eventually provide the impetus to get the person to act. Three may have been easier to ignore, or there may have been specific issues in the content of the messages. It does imply that for an intervention such as SMSs to be useful in the context of a behavior, like testing, where there are considerable barriers, for example, a certain number is required for the intervention to be successful. Clearly, the knowledge that we do have that different behaviors require different messages and intensity applies to SMS-based approaches as to other intervention approaches. Further study of intervention content and the impact of information intensity is clearly required as part of SMS-based campaigns in the future.

The results of the telephone interviews provide support for the use of SMSs for getting information to respondents. As most of the respondents were very supportive of receiving the SMSs and indicated that these messages were crucial in assisting them to overcome their fears, it provides support to the analyzed outcome that it is a combination of the delivery system and the content of the messages that produced the desired outcome. The specific content of the additional 7 messages may also indicate why there was a threshold effect at 10 messages. This adds further support to the idea that both the method of delivery and the content of the messages require specific attention.

A major innovation of this study was the successful use of PCMs, which bear no cost to the sender. Previous text-based interventions in resource-limited regions purchased airtime for participants to encourage responses ¹⁴ ; however, this affects the feasibility of scale-up. A further innovation was that participants were asked via SMS whether they had tested, removing the potential logistical difficulties of contacting each participant in person or by telephone. A potential limitation is that one cannot verify whether participants followed instructions correctly. Further, study participants were explicitly asked whether they had tested after the start of the intervention. Some may have reported that they had tested, but had in fact done so before the start of the intervention. Failure to follow instructions could be due to the use of English only in the intervention SMSs, or due to a lack of familiarity with this method of providing feedback.

Individuals' self-reporting of testing is a limitation in that participants could have given a socially acceptable answer. ²⁷ It is hoped that the confidentiality kept potential self-reporting bias to a minimum. In addition, the use of random allocation and a control would alleviate this self-reporting bias. Mitigation of this potential limitation was also limited by the fact that HIV testing is confidential and it is not possible to independently validate whether someone has tested for HIV in South Africa, without drawing on significant resources (for instance, to verify testing in a manner that does not compromise confidentiality). The use of an airtime incentive could also have introduced a bias through people giving a socially acceptable answer. We tried to mitigate this through clear indication in the text of the SMS asking whether they had tested, that the incentive would be earned whether they had tested or not. However, this bias would be the same for all conditions so it should not have affected the overall study results. The potential for bias in the database (as the Soul City show deals with HIV) was mitigated by the fact that Soul City has such broad viewership and according to research, the vast majority of South African adults have been reached by one or more Soul City programming components (87% in 2006²⁵).

Several other factors are important for the interpretation of these findings. Lack of demographic information on the participants is a major concern for the internal validity of the study. For example at baseline, demographic information could not be used to assess the effectiveness of the randomization of the control and the intervention groups. The absence of baseline demographic variables makes it difficult to analyze for specific higher effectiveness or response in certain sectors.

Future Works

The retention rate of 58% is in line with a response rate of 55.6% to paper surveys, and surpasses an average response rate of 39.6% from 68 electronic (e-mail and Web) surveys reported in 49 studies. ²⁸ The authors know of no published literature or studies looking at response rates for questions or surveys administered via SMS. Administration of questionnaires via SMS coupled with the use of PCM lines for responses is an innovative methodology. Given the potential of reaching many people using mobile, especially in developing countries, further examination of SMS survey methodologies would be useful.

A future study could incorporate some important methodological changes: probing of reasons for attrition, inclusion of demographic information, and means of corroborating responses from participants could be included. The SMS content could be translated into local languages, and collection of baseline data including independent variables known to influence HCT could facilitate a more robust analysis. Further exploration of the nature of the threshold effect is needed, to understand whether it is consistent with other interventions and if so, how to ascertain at what level of intensity to pitch an intervention. It would also be important to know whether there is a dosage effect once the threshold is reached. It is possible that once the threshold has been reached, additional doses may actually have a negative effect and reduce the impact.

Conclusions

Our results demonstrate the strong potential of SMSs as a tool for influencing the uptake of HCT, and the importance of the content of the messages. Our results also suggest the importance of determining a threshold for SMS-based interventions, although this is likely to vary according to the content and target of the particular intervention.