The answer of comment 1 :
Denmark has a free healthcare system, and as shown in this study, they have comprehensive access to nationwide medical databases. Therefore, the issue you raised should be avoidable in their case.
The answer of comment 2 :
THIS is exactly the reason why we chose to split it up in 3 groups. Other studies have
compared positive tested with not-positive tested. However, the “not-positive” group
consists of both those only tested negative with a presumably normal healthcare contact
- and those never tested with a presumably lower health care contact.
This way, we showed that behavior of being tested has a big impact on whether you
receive a psychiatric diagnosis or not. This behavior of individuals, could evidently lead
to unmeasured confounding. To address age, we both adjusted for age in our cox
regression, and we did stratified analyses to isolate age-groups, as depicted in figure 4.
As to selection bias and surveillance bias, this could be present if we only compared
“not tested” with “positive tested”. The strength of the study is therefore to compare
with “only negative tested” and “positive tested”. These individuals are those who go to
the doctor and get tested, and therefore have a more normal contact with the healthcare
system, compared to those not tested.
Therefore, I believe that our method more strongly isolating the effect of the exposure
(COVID-19), from all the other competing societal, biological and demographic effects.
Denmark has a free healthcare system, and as shown in this study, they have comprehensive access to nationwide medical databases. Therefore, the issue you raised should be avoidable in their case.
The answer of comment 2 :
THIS is exactly the reason why we chose to split it up in 3 groups. Other studies have
compared positive tested with not-positive tested. However, the “not-positive” group
consists of both those only tested negative with a presumably normal healthcare contact
- and those never tested with a presumably lower health care contact.
This way, we showed that behavior of being tested has a big impact on whether you
receive a psychiatric diagnosis or not. This behavior of individuals, could evidently lead
to unmeasured confounding. To address age, we both adjusted for age in our cox
regression, and we did stratified analyses to isolate age-groups, as depicted in figure 4.
As to selection bias and surveillance bias, this could be present if we only compared
“not tested” with “positive tested”. The strength of the study is therefore to compare
with “only negative tested” and “positive tested”. These individuals are those who go to
the doctor and get tested, and therefore have a more normal contact with the healthcare
system, compared to those not tested.
Therefore, I believe that our method more strongly isolating the effect of the exposure
(COVID-19), from all the other competing societal, biological and demographic effects.