부산대학교 도서관

Abstract Background Although the use of technology viz. mobile phones, personalised digital assistants, smartphones, notebook and tablets to monitor health and health care (mHealth) is mushrooming, only small, localised studies have described their use as a data collection tool. This paper describes the complexity, functionality and feasibility of mHealth for large scale surveillance at national and sub-national levels in South Africa, a high HIV-prevalence setting. Methods In 2010, 2011–12 and 2012–13 three nationally representative surveys were conducted amongst infants attending 580 facilities across all 51 districts, within all nine provinces of South Africa, to monitor the effectiveness of the programme to prevent mother-to-child transmission of HIV (PMTCT). In all three surveys a technical protocol and iterative system for mobile data collection was developed. In 2012–13 the system included automated folders to store information about upcoming interviews. Paper questionnaires were used as a back-up, in case of mHealth failure. These included written instructions per question on limits, skips and compulsory questions. Data collectors were trained on both systems. Results In the 2010, 2011–12 and 2012–2013 surveys respectively, data from 10,554, 10,071, and 10,536 interviews, and approximately 186 variables per survey were successfully uploaded to 151 mobile phones collecting data from 580 health facilities in 51 districts, across all nine provinces of South Africa. A technician, costing approximately U$D20 000 p.a. was appointed to support field-based staff. Two percent of data were gathered using paper- questionnaires. The time needed for mHealth interviews was approximately 1,5 times less than the time needed for paper questionnaires 30–45 min versus approximately 120 min (including 60–70 min for the interview with an additional 45 min for data capture). In 2012–13, 1172 data errors were identified via the web-based console. There was a four-week delay in resolving data errors from paper-based surveys compared with a 3-day turnaround time following direct capture on mobile phones. Conclusion Our experiences demonstrate the feasibility of using mHealth during large-scale national surveys, in the presence of a supportive data management team. mHealth systems reduced data collection time by almost 1.5 times, thus reduced data collector costs and time needed for data management.