Using an interactive short message service (SMS) data collection system in an HIV pre-exposure prophylaxis (PrEP) trial in Kenya and Uganda
F. Kibengo1, D. Mark2, J. Haberer3, E.J. Sanders4, G. Mutua5, A. Kamali1, P. Mugo4, O. Anzala5, H. Grosskurth1, B. Barin6, D. Bangsberg3, J. Rooney7, R. Lima2, P. Fast2, S. Berkley2, F. Priddy2
1MRC Entebbe, Entebbe, Uganda, 2International AIDS Vaccine Initiative, New York, United States, 3Harvard Medical School/ Harvard Initiative for Global Health-HIGH, Boston, United States, 4Kenya Medical Research Institute, Ctr for Geographic Medicine Research, Coast, Kilifi, Kenya, 5Kenya AIDS Vaccine Initiative, University of Nairobi, Nairobi, Kenya, 6EMMES, Rockville, United States, 7Gilead Sciences, Foster City, United States
Background: Collecting behavioral data via interactive SMS is innovative but little is known about feasibility and response rates in low income settings. We present preliminary results using SMS to interactively collect sexual activity data on a daily basis from HIV at-risk populations in Kenya and Uganda as part of a PrEP safety and adherence trial.
Methods: 144 men who have sex with men (MSM), female sex workers (FSW), and HIV-discordant couples (DC) in Kenya and Uganda were randomized to daily or intermittent emtricitabine/tenofovir or placebo to assess safety and adherence to PrEP. Each volunteer received a mobile phone, SIM card, and individual instruction on SMS messaging. Volunteers received daily SMS during 4 months follow-up with one question about sexual activity and one question about condom use in their language of choice, and provided numerical responses (e.g. 1=Yes, 2=No) at no cost. Queries and responses were password protected.
Results: Data on 100 volunteers comprising 37% of study follow-up time is presented. [Complete data will be available July 2010]. The median response rate for MSM/FSW (Kenya) was 29%, increasing to 42% when days with major server outages were excluded. Median response rate for DCs (Uganda) was 67%, increasing to 76% when excluding major network outage days. Network outages lasting >2 hours per day occurred on 17/95 days in Kenya and 8/95 days in Uganda. Problems with SMS gateway providers required changing mobile networks and associated SIM cards at two sites. Lost mobile phones were not common in either country.
Conclusions: Using SMS to collect clinical trial data in low-income countries is technically possible, but major challenges of server outages and network issues need to be addressed and field-tested in advance. Volunteer technical or acceptability factors may also contribute to suboptimal response rates and warrant further study.
Download the e-Poster (pdf)