layout/contents of cabg.dat

  Item       Contents                                         
--------     ----------------------------------------------------
1 AnnVol_L   Lower limit of Volume Category (annualized pts/year)

2 AnnVol_U   Upper limit of Volume Category (annualized pts/year)

3 Patients   Number of Patients undergoing CABG

4 Deaths     Number of Deaths                                      

5 Pct_exp    Expected PERCENTAGE mortality,
             as predicted by STS risk stratification method.
            
Notes:

1: Text says practice included if 50 or more patients and 6
   months of SEQUENTIAL data in database. However, the categories
   used in the analysis are based on ANNUALIZED caseloads. The 
   lower limit for the lowest category " <= 100 " is not specified
   and so I arbitrarily set it to be 50. 

2  Highest category is " > 900", but Figures 1-3 enable one to
   deduce that the largest annualized volume is approximately 1720.
   
3  From second column of Tables 2 and 3 (or successive subtractions
   applied to second column of Table 1)
   
4  Not at all easy to arrive at the correct numbers for this.

   If sufficient decimal places were given for the "percentage of
   mortality" column in Table 1, one could deduce them exactly
   by calculating the numbers of deaths in each " <= " category
   and then applying successive subtractions.
   
   The percentage mortality in Table 2 is unfortunately an average
   of the percentage mortality for the groups in each volume 
   category: e.g. the 2.9% for the 701-800 category is an average 
   of maybe 2.8% and 3.0%, and as such is not exactly the same as 
   the overall percentage based on the 3253 patients in the category
   that we would like to have.
   
   A third estimate is the average reported in Table 3 (I don't
   understand what units go into the means and SD's in Table 3).
   
   Faced with these 3 possible estimates of the total number of 
   deaths in each volume category, I took a weighted average, giving
   weights of 25%, 50% and 25% to the estimated numbers of deaths I
   arrived at from Tables 1, 2 and 3 respectively (fortunately, the
   three different estimates are not all that different!) and then
   rounded it to the nearest integer.
   
5  Note that this is a PERCENTAGE. Again, since I had the option
   of deducing it from the expected mortality percentage in Tables
   1, 2 or 3, I used the same weighted average, and kept two decimal
   places.