Skip to main content

Table 2 The six models fit for each procedure

From: Exploring variation in low-value care: a multilevel modelling study

 

Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Model purpose

Base for comparing effects of adding levels

Effect of adding hospital to M1

Effect of adding LHD to M1

Effect of adding SLA to M1

1. Additional effect of including all 3 levels

2. Contribution of each level to variation in odds of receiving low value care

Effect of adding level-specific variables

Assess using

Increase in c statistic vs M1

Increase in c statistic vs M1

Increase in c statistic vs M1

1. Increase in c statistic vs M2, M3, and M4;

2. VPC, MOR

IOR80

OR

Variables includedb

 Levels

  Hospital

 

X

  

X

X

  LHD

  

X

 

Xa

Xa

  SLA

   

X

X

X

 Episode variables

  Age

X

X

X

X

X

X

  Sex

X

X

X

X

X

X

  Charlson

X

X

X

X

X

X

  Private patient

X

X

X

X

X

X

  Emergency care

X

X

X

X

X

X

  Financial year

X

X

X

X

X

X

 Hospital variables

  Hospital peer group

     

X

  Procedures episodes as proportion total episodes

    

X

 LHD variables

  Rural/metropolitan

     

X

 SLA variables

  IRSAD quintile

     

X

  Remoteness category

     

X

  Population prevalence of low-value indication

     

X

  1. aLHD was not included in Models 5 and 6 for spinal fusion and EVAR because each LHD had only one hospital that performed the procedure. b Additional file 1 contains more details about the variables
  2. IRSAD Index of Relative Socioeconomic Advantage and Disadvantage, LHD Local Health District, MOR median odds ratio, SLA Statistical Local Area, VPC variance partition coefficient