Validation of an improved area-based method of calculating general practice-level deprivation

Abstract

Objective: To compare the methods of calculating practice deprivation scores in the absence of patient-level data.

Study design and setting: Three methods of deriving general practice deprivation scores without patient-level data were compared against "gold standard" patient-level scores in 226 English practices. The three methods were lower super output area (LSOA), middle super output area (MSOA), and a geographical information systems (GIS) method. Working, if necessary, on the log scale, agreement between scores was assessed using Bland and Altman's method, Kappa statistics, and Pitman's test.

Results: Based on the antilog 95% limits of agreement from Bland-Altman plots, GIS methods showed least variation compared with gold standard (0.66-1.47), followed by MSOA (0.61-1.70) and LSOA (0.38-2.29) methods. The differences in variances between both GIS and MSOA, and LSOA and MSOA comparisons, were greater than would be expected by chance (Pitman's P<0.001). High levels of agreement (kappa: 0.93, 0.86, and 0.80) were observed between GIS, MSOA, and LSOA methods compared with the "gold standard."

Conclusion: In situations where patient postcodes are unavailable, the GIS method is superior to area-based methods. However, where the GIS method cannot readily be applied, the MSOA method should be used in preference to the LSOA method.