Lets make up some data:
df = DataFrame(val=rand(8), injury = [rep("shoulder",4), rep("leg",4)])
This version does the same as the Python version:
injuries = [injury=>df[df[:injury] .== injury,:] for injury in unique(df[:injury])]
Now injuries
is a dictionary of DataFrames, one DataFrame for each injury. You can then just do:
injuries["shoulder"]
and it returns a DataFrame with only the shoulder injuries.