Skip to main content

Table 3 MAPE (Mean absolute prediction error) expressed as a percentage of the average cost of different models based on case-mix systems, using different regression models

From: Predictive risk modelling in the Spanish population: a cross-sectional study

   Linear regression (OLS) Two-part Hierarchical models
   ACG-PM CRG DCG-HCC ACG-PM CRG DCG-HCC ACG-PM CRG DCG-HCC
Recalibration Variables          
  Age and Sex (A&S)   106.80%    109.09%    106.52%  
  A&S + Only Dx 92.36% 96.36% 91.26% 114.59% 104.74% 109.74% 95.27% 95.97% 91.52%
  A&S + Only Rx 93.57%   91.87% 113.84%   115.74% 93.74%   91.98%
  A&S + Dx + Rx 88.41%   87.02% 114.96%   114.44% 89.87%   87.26%
  A&S + Dx + Rx + cost percentile 85.99% 87.70% 85.05% 98.52% 96.11% 97.92% 86.97% 87.72% 85.37%
  A&S + Dx + Rx + cost percentile + DI 86.02% 87.70% 85.06% 98.49% 96.09% 97.89% 86.97% 87.72% 85.37%
  1. Dx diagnoses, Rx prescriptions, DI Deprivation index.