prod | ops | dev

Author
Affiliation

David L Smith

University of Washington

High quality, robust analytics for malaria in Uganda have the potential to leverage information and transform malaria control. To manage the diversity and complexity of the tasks, workflows are organized into production pipeline (prod), operations (ops), and a development pipeline (dev).

The need for stable, high quality information creates a tension between having stable code that produces reliable outputs and the need for critical evaluation of methods, ad hoc analytics, and innovation. To address these needs, malaria analytics in Uganda are organized around a production pipeline (prod), operations (ops), and a development pipeline (dev).

dev and ops work in a Sandbox that maintains and updates prod.

prod

prod – the production pipeline – provides consistent, high quality, routine analytics.

To address the need for stable code that produces reliable outputs in a timely way, malaria analytics establishes a stable production pipeline (hereafter, called prod). It is a set of algorithms, protocols, and procedures that are run on schedule to update data assets, situational analysis, or statistical estimates.

The prod algorithms:

  • have well defined protocols and procedures;

  • have undergone extensive testing;

  • are well-documented; and

  • produce standard outputs.

The outputs of prod ought to become familiar to most members of NMED and the MoH.

ops

ops – operations – ad hoc analysis to address urgent programmatic needs

There is a constant need for new analyses in response to changing malaria conditions, changing budgets, and changing programmatic needs. The primary responsibility of malaria analysts is to respond to those needs. Malaria analysts are thus:

  • participating in policy discussions and identifying the need for new analysis

  • developing new analyses to fill those needs

dev

dev – the development pipeline – was developed for ad hoc analytics and to improve prod

There are open questions about prod. The first release of prod was extensively tested, but there were some difficult issues about how to resolve inconsistencies in the facility data; how to handle missing data; the right way to compute some of the key metrics; etc. These had been resolved well enough to be useful, but there was room for improvement. A key goal for dev is to critically evaluate and improve prod: revisit consistency checks for facility data; review facilities; develop of methods to validate the predictions;

Some important elements are currently missing from prod. NMED ought to have an operating map of vector species and insecticide resistance, and a better understanding of care seeking and drug taking. New elements in prod are created by dev.

As part of the dev pipeline, analysts have several tasks:

  • transform some of the ad hoc analysis done by ops into new pipelines and added to prod

  • test and critically evaluate existing algorithms or data inputs looking for ways to improve the accuracy and utility of their analysis

  • review existing data sources to assess their accuracy;

  • expand the list of available data assets to fill critical gaps; and

  • monitor peer-reviewed literature and adapt or assimilate new ideas and analysis, perhaps in collaboration with academics

  • update prod whenever the underlying data changes (e.g. a new instance of DHIS2, changes to existing location hierarchies, …)