Dimagi

D-Tree, Dimagi and other partners evaluated whether the use of an electronic job aid on a mobile device improves adherence to the Integrated Management of childhood Illness (IMCI) protocols in rural Tanzania. D-Tree, along with Brian DeRenzi from the University of Washington, lead the implementation and research while Dimagi lead the programming.

The team developed and piloted e-IMCI, a program that runs on a mobile device and guides a health worker step-by-step through the IMCI treatment algorithm. The current system covers only first visits for children 2 months to five years old, and does not cover immunizations or malnutrition. We field tested our prototype system at a dispensary in Mtwara, Tanzania. During early pre-testing the team extended e-IMCI to allow the clinicians more freedom to choose drugs and use approximate measures for certain investigations, to allow the use common sense to interpret the protocols when necessary. Our formal investigation consisted of structured interviews with clinicians, and observation of patient encounters with and without e-IMCI.

The team observed 24 patient encounters without e-IMCI, and 28 with e-IMCI. Using the e-IMCI prototype, clinicians performed 84.7% of investigations required by IMCI, a significant improvement over the 61% of investigations observed during conventional practice.(p < 0.01). Our current prototype is almost as fast as the current practice, where the book is rarely referenced. The team analyzed 18 trials comparing the time by the same clinician in a traditional IMCI session to one using e-IMCI; the average for both was about 12.5 minutes. The training time for e-IMCI was less than 20 minutes, after which clinicians were easily able to train each other. The four clinicians unanimously preferred e-IMCI, citing it as faster and easier to use than the chart book. These results and our experience suggest that e-IMCI is fast, improves adherence, and thus the quality of care, and also affords the health worker enough flexibility to apply their skills on a case-by-case basis.

Research on this project was presented by Brian DeRenzi at the Computer Human Interaction conference in 2008.

A second pilot is currently under way by D-Tree in Arusha, Tanzania, led by John Wesonga. This pilot will leverage a new platform technology build on JavaRosa.