NAs in Cost related attributes.

[eashwarsiddharth]

@Rajhan Any suggestions on how to substitute NAs in the following attributes:

• brand_drug_cost
• generic_drug_cost
• other_drug_cost
• total_drug_cost_ge65
• total_day_supply_ge65

NAs correspond to the suppress_flag. But, the description of the flag only talks about the claim_count and not the cost/supply related attributes.

bene_count_ge65_suppress_flag from prescriber_summary, gives a clue for claim_count:

bene_count_ge65 = 5, iff, bene_count_ge65_suppress_flag = '*' Thus, for all ge65_suppress_flag = '#', total_claim_count_ge65 = total_claim_count - 5

How do I go about substituting NAs in the cost/supply related attributes ?

• Should I treat them as actual missing values and try imputing them with packages like MICE ?

In spite of this roadblock, I have 75% complete cases (2430 records) for the (Avonex vs Copaxone) scenario.

[eashwarsiddharth]

@Rajhan here is a look at the pattern in missing values.

ggplot_missing <- function(x){
                    x %>% is.na %>% melt %>%
                      ggplot(data = ., aes(x = X1, y = X2)) +
                      geom_raster(aes(fill = value)) +
                      scale_fill_grey(name = "", labels = c("Present","Missing")) +
                      labs(y = "Variables in Dataset", x = "Rows / observations")
                  }
ggplot_missing(avocop)

[Rajhan]

2430 is a good number. Go with it.