Caveats for interpreting data

We believe that MEDMI can be a useful tool to explore the possible interconnections between climate, weather, environment and human health and wellbeing by linking data from these different areas. Nevertheless when linking such diverse data, it is important to be careful about a number of issues:

  • Exploring these data and data linkages can only be considered “hypothesis generating,” which means that the associations discovered need to be explored further in greater detail and with other data to confirm the findings
  • The data need to be linked appropriately in both space and time; and it is important to understand the data being linked in terms of what is actually being measured
  • The linkage must often take into account lag periods, i.e. the time from exposure to effect
  • It is easy to over-interpret the linkage results, seeing causal associations (i.e. that a particular exposure definitely leads to or causes a specific health outcome) where none exist; it is therefore essential to design robust statistical analyses to confirm or reject the findings
  • It is not possible to take all of the different factors into account that are associated with a truly causal relationship in a single analysis

When analysing data there are a number of statistical/analysis issues that need to be taken into account with the data interpretation of any of the MEDMI tools and databases:

  • Seasonality
  • Multiple drivers
  • Geographic/temporal data linkage
  • Autocorrelation
  • Random noise
  • Similar responses in different pathogens
  • Choice in method
  • Measuring the impact of between year changes
  • Trawling Surveillance dataset