We ran an MNCH household survey for our project in Nepal. The last interview was finished on Saturday, and the data were analyzed and shared with partners yesterday – about 48 hours after the final person was interviewed.
As I have blogged about a few times (here, here and here), data collection and management continues to be a challenge as we and other NGOs work towards timely and rigorous project evaluation. HealthBridge has worked with direct data capture over the past few years, using handheld computers (PDAs, remember them?) and the software Visual CE so that household interviews are collected and immediately stored on an electronic database, and, with good Visual CE programming, the data are clean and immediately ready for analysis. (see here, here, and here for some HealthBridge experiences with PDAs).
While highly effective, there were a couple of drawbacks. First Visual CE is pretty difficult to learn and so we were not successful in passing on the programming skills to many partners. And we had to lug 50 or more PDAs (and batteries, protective cases and other gear) all over creation and show the interview teams how to use them, because PDA ownership was uncommon.
These two disadvantages are addressed through a newish direct data capture technology called ODK. Questionnaires are developed in Excel, with which most computer users have some skill, and then are converted to XML format using an online tool. The XML file is downloaded to an Android device (tablet or phone) where the interview data can be directly captured. Android devices are common and seem to be available for fairly cheap purchase (maybe $200 apiece) almost everywhere.
And finally, after the data are collected on the device, when the interviewer can access the internet (through cell, wifi or whatever) the data are uploaded to a password protected server.
This was how we did it for our survey in Nepal. Some of the Android devices had problems with their GPS, but everything else went as planned. The survey took place over 5 days last week. At the end of each day, the data were uploaded to our server. I checked the data daily for any strange entries and advised the survey supervisor by email about the few issues I saw, and he communicated with the interviewers. The last data were uploaded Saturday in Nepal and I downloaded the data Monday morning in Ottawa.
I had drafted the SAS code for data analysis, so I was able to complete preliminary analysis within about 8 hours, and shared the results with our partners in Nepal by the end of the day. This is how it is supposed to work. It isn’t news to people who have worked with ODK, but this is the first one for me and it is encouraging that it worked so well.
Thanks to Tim Roberton at Johns Hopkins for helping me learn how to do all of this.