Field capture on a bad connection

Health surveillance assumes the data arrives. In a remote clinic on an intermittent 2G connection it frequently does not, and the failure is quiet: a form is filled in, the request times out, and the record is gone. Nobody upstream sees a gap, because a gap and a quiet region look identical.

The engagement was contracted through AMREF Health Africa with Global Fund financing, covering the field data entry interface and the sync layer for a DHIS2 integration. DHIS2 is the open-source platform much of this reporting runs on, and the work sat entirely on the collection side of it.

Entry is local first. Records persist to IndexedDB in the browser the moment they are entered, so the clinic is never waiting on a network round trip to continue working, and closing the tab does not lose the morning.

Sync is store-and-forward. A queue trickles compressed payloads out over whatever connection exists, retrying without duplicating and without dropping fields on a partial send. During a high-volume event the compression and the queueing both matter, and so does staying off the main browser thread: an interface that locks up while flushing a backlog is an interface that gets abandoned mid-outbreak.

Payloads are compressed for the connection rather than for the server. On a metered 2G link the size of a submission is a cost somebody is paying.