METHODS: Data were collected by 24 community health workers (CHWs) in 2 Rwandan communities, 1 urban and 1 rural.
METHODS: The same CHWs collected paper and electronic records. Paper data contain weight and age for 320 boys and 380 girls. Electronic data contain weight and age for 922 girls and 886 boys. Electronic data were collected over 9 months; most of the data is cross-sectional, with about 330 children with time-series data. Both data sets are compared with the international standard provided by the WHO growth chart.
METHODS: The plan was to collect 2000 individual records for the electronic data set--we finally collected 1878 records. Paper data were collected by the same CHWs, but most data were fragmented and hard to read. We transcribed data only from children for whom we were able to obtain the date of birth, to determine the exact age at the time of measurement.
RESULTS: Mean absolute error (MAE) and mean absolute percentage error (MAPE) provide a way to quantify the magnitude of the error in using a given model. Comparing a model, log(weight)=a+b log(age), shows that electronic records provide considerable improvements over paper records, with 40% reduction in both performance metrics. Electronic data improve performance over the WHO model by 10% in MAPE and 7% in MAE. Results are statistically significant using the Kolmogorov-Smirnov test at p<0.01.
CONCLUSIONS: This study demonstrates that using modern electronic tools for health data collection is allowing better tracking of health indicators. We have demonstrated that electronic records facilitate development of a country-specific model that is more accurate than the international standard provided by the WHO growth chart.