Individual project applications were associated with one to five HRCS health category codes. If multiple codes were assigned to a project, they were equally apportioned. For example, if four codes were assigned to one project, each was recorded as 25% irrespective of the ‘actual’ relevance for the project.6 Furthermore, prior to 2018, health categories were coded manually by the funders. Since the 2018 HRCS submission, coding is automated based on publicly available titles and abstracts (‘auto-coding’). Variations in coding between the manual and automated HRCS coding approaches may have occurred.
The analysis of HRCS codes gave equal weight to each association, irrespective of whether a project was assigned one or more codes, on the assumption that the project is equally relevant to each assigned health area. A caveat to this approach is that HRCS coding rules may assign multiple, predetermined codes to certain health topics, which vary in their level of relevance. Therefore, for codes that are ‘split’ in this way more often than others, the degree to which projects address the corresponding health area may be overstated. The analysis approach used reflects the number of projects that address each health area to at least some degree, but may overstate the level to which health areas commonly assigned alongside other codes are addressed.
Categories that were most frequently assigned to EME projects alongside other codes were ‘metabolic and endocrine’ and ‘infection’. The categories ‘mental health’, ‘musculoskeletal’, ‘respiratory’, ‘reproductive health and childbirth’ and ‘eye’ were rarely assigned alongside other codes ().
Health Research Classification System health codes: level of association with multiple codes per project
The findings of the approach used in the evaluation (giving equal weight to each association) were compared with those of an analysis based on fractional assignment of categories. Following the latter approach (i.e. focusing on the level to which a health category was addressed rather than the number of projects) changes the order of health categories in , as ‘mental health’ and ‘respiratory’ were more strongly represented (moving up from rank 9 to rank 5, and from rank 11 to rank 7, respectively). Conversely, the health category ‘metabolic and endocrine’ moved down from rank 5 to rank 12. All other codes remained within one rank of their position.
Using apportioned counts also resulted in a reduction in the difference in shares between the EME portfolio and UK DALYs for ‘reproductive health’, ‘metabolic and endocrine’ and ‘oral and gastrointestinal’ (). However, the difference remains highest for these health areas (i.e. the evaluation’s conclusions are unchanged). The shares of HRCS health categories ‘blood/cardiovascular/stroke’ and ‘cancer and neoplasms’ are reduced, but remain highest overall. Shares for the areas ‘injuries and accidents’, ‘musculoskeletal’ and ‘mental health’ are nearly unchanged when analysed as fractional counts.
Share of EME projects by HRCS codes compared with shares of UK DALYs for 2012 and 2016. Source: Technopolis analysis of EME portfolio and data obtained from UK Health Research Analysis 2018.
