Evidens-baserat folkhälsoarbete bortom randomiserade försök: Ny kunskap genom data mining för att kartlägga orsakskedjor och förutsäga effekter på barnadödlighet
Tidsperiod: 2015-01-01 till 2017-12-31
Projektledare: Katarina Ekholm Selling
Medarbetare: Lars-Åke Persson, Eva-Charlotte Ekström, Oleg Sysoev
Budget: 3 000 000 SEK
A third of the global disease burden involves children and falls almost entirely in low- and middle-income settings. Efficacious interventions for child survival are established, but difficulties arise when scaling up these interventions in new settings resulting in lower effectiveness. There is a need to advance data modelling strategies considering individual, population, environmental, and health system contextual factors for prediction. Knowledge discovery in databases, i.e. data mining (DM) strategies, is designed at finding contextual differences and is increasingly used in business and biomedicine, but rarely in public health. This project aims at employing DM strategies on well-characterized, longitudinal databases from low- and middle-income settings to explore the importance of contextual factors and create prediction models that can be applied and tested for its predictive capacity. These databases emanate from three successful interventions with child mortality outcomes in Bangladesh, Vietnam, and Nicaragua (n = 4,000 to 25,000). We will identify determinants of mortality and the role of contextual factors in mediating or modifying effects. This will provide answers not only to the question: what works? but also to: when, where and for whom?. The developed prediction models can be applied to national representative datasets. This project offers unique prospects of DM strategies applied to priority public health problems that could inform policy and practice.