Calculating inequality metrics for CCGs

There are a number of ways of measuring health inequality. The commonly used ones are:

• Simple absolute or relative differences in health status between groups
• Univariate summary measures of health variation e.g. Gini coefficient
• Multivariate summary measures of health variation e.g. slope index of inequality

For this analysis we are using data from general practice profiles.

This contains for most practices:

• IMD2015 scores
• Practice level estimates of life expectancy in men and women
• List size

We extracted data from the profiles into a table for analysis

The dataset contains data for 7758 practices

'data.frame':   7758 obs. of  6 variables:
 $ Code: chr  "A81001" "A81002" "A81003" "A81004" ...
 $ ccg : chr  "NHS Hartlepool And Stockton-On-Tees CCG" "NHS Hartlepool And Stockton-On-Tees CCG" "NHS Hartlepool And Stockton-On-Tees CCG" "NHS South Tees CCG" ...
 $ IMD : num  28.5 29.4 41.4 34.4 15.6 ...
 $ Lem : num  76.2 77.1 75.9 76.7 82 ...
 $ Lef : num  82 81.8 80.9 81.4 83 ...
 $ pop : int  4188 19603 3446 9006 7889 12165 9525 4088 9265 11285 ...

Observations: 7758
Variables:
$ Code   (chr) "B83019", "B83624", "B83002", "B83620", "B82099", "B820...
$ ccg    (chr) "NHS Airedale, Wharfdale And Craven CCG", "NHS Airedale...
$ IMD    (dbl) 5.560, 6.300, 6.356, 7.469, 10.120, 11.235, 11.815, 12....
$ Lem    (dbl) 80.844, 81.099, 81.092, 81.751, 82.538, 80.903, 79.448,...
$ Lef    (dbl) 85.004, 84.493, 84.514, 84.711, 85.025, 84.056, 82.360,...
$ pop    (int) 6627, 13663, 4386, 2902, 4044, 12157, 11156, 14152, 938...
$ IMD_wt (dbl) 19.89082, 19.89082, 19.89082, 19.89082, 19.89082, 19.89...
$ a      (dbl) 0.04206444, 0.08672498, 0.02783984, 0.01842025, 0.02566...
$ b      (dbl) 0.02103222, 0.08542693, 0.14270934, 0.16583938, 0.18788...
$ y      (dbl) 16.58079, 23.88293, 13.53042, 11.09535, 13.22387, 22.47...
$ a1     (dbl) 0.2050962, 0.2944910, 0.1668528, 0.1357212, 0.1602155, ...
$ b1     (dbl) 0.004313628, 0.025157466, 0.023811446, 0.022507924, 0.0...

[1] 209   7

• interdecile difference is a better predictor of slope index of inequality than range
• correlation coefficients

• there are a number of ways of measuring health inequality in CCGs
• a widely used measure, the slope index of inequality, can be calculated from practice level data for CCGs, given estimates of life expectancy and practice level deprivation scores: these data are available from GP practice profiles (for life expectancy 2008-12, and IMD 2015)
• the slope index can be appoximated by simpler measures such as the range (highest-lowest), or inter-decile range (90th centile - 10th centile)
• of these the IDR is more stable and a better predictor of SII and is preferred