Data to support health management of maricultured salmonids in Scotland.

Marine salmonid farms are recording more information than ever before. Some information relevant to health management is generated continuously in near real time (e.g. water temperature), others daily (e.g. phytoplankton counts), weekly (e.g. lice counts), or as opportunities arise (e.g. diagnostic results). This hight level of data recording provides new opportunities to process data that can lead to data-driven insights beyond what would have been possible with intuition alone. For example, developing a better understanding of risk factors for disease, seasonal trends, predicting future occurrence of disease, and developing early warning systems for disease. Even though all this data provide many possibilities, it is unclear if and how health managers utilize, select and digest this information in the day-to-day health management on farms in Scotland. This study aimed at better understanding data utilization by health managers with a goal to identify needs that can be met by data-scientists and epidemiologists. We held four workshops with in each 2-4 salmonid health experts to discuss data usage and needs. Salmonid health experts agreed that data are useful for health management. Use of data vary between companies due to different data-storage platforms and protocols. Whereas some platforms had the ability to compile, process and visualize data to some extent, others did not or to a lesser extent. The level of technical support available to health managers differed between companies and this affected data usage. Sought after information could vary day-to-day and differed per expert based on their professional experience. For example, the sought after information could be a time series looking back a week, several months, several years, or aggregating different data sources temporally or spatially. There was wide variation in terms of eagerness to use data to support decision making. For example, some participants were keen to evaluate as much data as possible, whereas others preferred to rely mostly on professional experience. Participants expressed the advantages of a preventive health management approach and identified the need to early detect abnormalities. However, although sometimes available to them through their management software platforms, predictive models were not routinely included in their decision-making processes, and there was a distrust towards them. Evaluating the routine use of data for day-to-day decision making by health managers of salmonid in Scotland revealed variability and opportunities. There is scope to improve software platforms and for companies to provide technical support to help with data needs without having to share data with externals. Applied data-skill and visualization training for salmonid health managers may be beneficial for individuals with little technical support, but most participants demonstrated an underlying discomfort towards relying too much on data to make decisions. Perhaps, data support tools need to be simple, integrated into existing software so avoiding the need for copying and pasting, save time, provide visual dashboards including only sought after information that can be used as one of the pieces of evidence to support decision-making. There is an ongoing need for engagement between data-scientists, epidemiologists, and industry to better understand which information is relevant under which circumstance