I have evaluated the impact of the covid19 outbreak in the codification of multiple codes in the health services.

I have summarized the number of codifications of each disease-code in two equal periods of time for two populations. The number of codifications of each code is normalized by the total number of codifications within the population and the period of time. Afterwards, I calculate the impact by: ( %_impact_pandemy - %_pandemy) , leading into a number that represents the percentage of increase or decrease in the mean codification of the code after the cov19.

I did the same for the two populations.The data has the following format (with more that 1500 codes):

code %_imact_pop1 %_impact_pop2

A00 -0.074 -0.054

... .... ....

My hypothesis is that the difference in diagnostic coding between the two populations should not be significant since the impact of the outbreak produced a similar effect in both populations. I want to know if the covid19 outbreak has had a similar inpact in the two territories.

---------------------------------------------------------------------------------

After researching, I tried to perform an unpaired two-way anova controlling for the code and the population. Nevertheless, it seems that the data does not cope with the assumption that the data is normally distributed.

Anyone knows if this is a good approach or knows how to overcome the problems with data homogeniety?

Thanks in advance