RESUMO
BACKGROUND: Obesity is a global epidemic affecting all age groups, populations, and income levels across continents, though is known to disproportionately affect socioeconomically disadvantaged populations. The causes of obesity are complex, informed by diet and weight practices, but shaped by social, commercial, and environmental factors and government policy. Consequently, a Whole System Approach (WSA)-which considers the many causes of obesity and shifts the focus away from individuals as points of intervention and puts an emphasis on understanding and improving the system in which people live-is required. This scoping review of reviews aims to: determine how WSAs to diet and healthy weight have been implemented and evaluated nationally and internationally; to determine what models or theories have been used to implement WSAs; describe how WSAs have been evaluated; determine if WSAs are effective; and to identify the contribution of the public and/or service users in the development of WSAs. METHOD: Systematic searches were carried out using CINAHL, Scopus, PsycINFO (ProQuest), the Cochrane Library, and MEDLINE. Included review papers were those that focused on the application of a whole system approach to diet and/or healthy weight, and/or reported the theory/model used to implement or simulate this approach. Databases were searched from 1995 to March 2022 using a combination of text and Medical Subject Headings (MeSH terms). In addition, reference sections of identified articles were examined for additional relevant articles. Covidence software was used to screen titles and abstracts from the electronic databases and resolve conflicts. RESULTS: A total of 20,308 articles were initially retrieved; after duplicate removal 7,690 unique title and abstracts were reviewed, and 110 articles were selected for full text review. On completion of full text review, 8 review articles were included for data extraction. These included: one umbrella review, four systematic reviews, a rapid review, and two literature reviews (one of which was on strategic reports written for government and public health policy). Evaluations of WSA were mainly process evaluations although health outcomes were assessed in some studies. Several conceptual frameworks or mathematical modelling approaches have been applied to WSAs for diet, healthy weight, and obesity to inform their planning or delivery, and to understand/map the associated systems. Common mathematical approaches include agent based or System Dynamic Modelling. Underlying both conceptual and mathematical models is an understanding how the elements of the complex systems impact each other to affect diet, healthy weight, and obesity. WSA implementations have reported some success in positively impacting health outcomes including reducing Body Mass Index, reducing sugary food intake, and increasing physical activity. Public and user involvement in WSA was not widely reported. CONCLUSION: The application of WSA to diet and healthy weight shows promise, yet the research is lagging behind their implementation. Further robust evidence for using WSA to address diet and healthy weight are required, including incorporating process and outcome evaluations (perhaps using established approaches such as Systems Dynamic Modelling). Furthermore, the analysis of epidemiological data alongside longitudinal process and outcome evaluation regarding the implementation of a WSA is required.
Assuntos
Peso Corporal , Obesidade , Humanos , Obesidade/prevenção & controle , DietaRESUMO
Uncertainty quantification in machine learning can provide powerful insight into a model's capabilities and enhance human trust in opaque models. Well-calibrated uncertainty quantification reveals a connection between high uncertainty and an increased likelihood of an incorrect classification. We hypothesize that if we are able to explain the model's uncertainty by generating rules that define subgroups of data with high and low levels of classification uncertainty, then those same rules will identify subgroups of data on which the model performs well and subgroups on which the model does not perform well. If true, then the utility of uncertainty quantification is not limited to understanding the certainty of individual predictions; it can also be used to provide a more global understanding of the model's understanding of patient subpopulations. We evaluate our proposed technique and hypotheses on deep neural networks and tree-based gradient boosting ensemble across benchmark and real-world medical datasets.