Chasing the ghost of infection past: identifying thresholds of change during the COVID-19 infection in Spain

Luis Santamaria; Joaquin Hortal

The analyses described above are highly consistent, and they were not strongly affected by changes in the extension of the data series (as more data became available) or in the fitting procedures (untransformed vs log-transformed data). However, their validity is potentially constrained by a number of caveats. We used data on the number of reported cases -i.e. severe cases subject to testing. In general, tests were restricted to those requiring hospitalization or belonging to risk groups, thus underestimating total infection numbers. Similarly, while at the beginning the number of deaths recorded in Spain included all deaths of patients who tested positive or showed a compatible symptomatology, during the peak of the pandemic several regional governments failed to count all deaths outside hospitals if they were not tested either pre-or post-mortem. Given the significant numbers of deaths happening in nurse homes and particular residences, this unevenness in the measurement of COVID-related fatalities could have biased results, eventually flattening the curve. If such effect was significant, however, it should have resulted in changes in the intercept of the number of deaths' curve. Given the lack of evidence for such changes, we believe that our results will stand out if new data coming from autopsies or re-evaluations is available in the future. At any rate, these limitations emphasize that, at present, we can only use the available data as proxies for the actual rates of infection spread and lethality. Therefore, our results must be taken with due caution until more detailed studies surveying seroprevalence in the general populations (including a broad study launched by the Spanish government as we write; Mitchell 2020) are available.