oa How Does Health Data Collected Electronically Compare To The Data From The Standard Paper System?

We develop a mobile application for Community Health Workers (CHWs) to collect health data for monitoring children's growth and development with a custom mobile application. The application is designed to be delay tolerant and optimized for low-resource settings. The evaluation is carried out in an urban and a rural location in Rwanda. The preliminary results show that CHWs are very successful in electronic data collection for tasks they already routinely perform, such as measuring weight gain and Middle Upper Arm Circumference (MUAC), which is a UNICEF protocol for detecting malnutrition. They had no problem adjusting to the electronic system since we followed the same format as the paper forms. We consider how mobile data collection improves both data availability and accuracy. The considerations that can relate to the advantages of electronic data versus the present paper-based system are: timeliness, accuracy and consistency. There is no doubt that electronic data is timelier than paper reports since the present system are submitted once a month. For data accuracy, we compare the accuracy of the present paper system with the Electronic Health Record approach. Our application has a built in error correction because it issues a warning when data is outside of the normal range of ±2sd. It does not prevent CHW to enter the data but it warns them of a possibility of an error, and improves the internal validity of data. By comparing the values that are input with a model based on previous measurements, it is possible to ensure self-consistency and detect errors. We have data from the paper records for the six months before we started the study at the same two locations. We were not able to locate and follow the same children using both systems because the paper data is fragmented and incomplete. However, we have many records available and the quantity of data will facilitate comparison of the two systems and estimates of statistical significance of the results. In addition we build a distribution model based on a large data set from the Rwandan Ministry of Health concerning the weight of children under five. We test how the electronic and paper data is described by this distribution over successive measurements and use the results to develop improved methods for checking the accuracy of data input by CHW in real-time.