Skip to main content

Table 2 Results of multivariable models of risk factors for methicillin-resistant Staphylococcus aureus carriage on admission

From: Variable performance of models for predicting methicillin-resistant Staphylococcus aureus carriage in European surgical wards

Risk factor

Stepwise model

Best BMA model

BMA model

Simple model

OR

95% CI

P

OR

95% CI

P

Posterior Probability

OR

95% CI

P

OR

95% CI

P

Female sex

      

2.3

1.0

0.9-1.1

.894

   

Age (per 1-year increment)

1.02

1.00-1.03

.009

   

44.8

1.01

0.99-1.03

.421

1.02

1.00-1.03

.009

Diabetes

      

7.8

1.0

0.8-1.4

.796

   

Chronic skin disease

3.0

1.5-5.8

.002

   

49.3

1.7

0.5-5.3

.369

2.9

1.5-5.6

.002

Hospitalisation (<1 year)

2.2

1.5-3.3

<.001

2.3

1.5-3.4

<.001

100

2.2

1.5-3.4

<.001

2.2

1.5-3.3

<.001

Nursing home resident

3.4

1.6–6.8

.001

4.2

2.1-8.3

<.001

82.4

3.1

0.9-10.3

.070

3.4

1.7-6.9

.001

Skin wound/sore

2.7

1.7-4.4

<.001

3.0

1.8-4.8

<.001

100

2.9

1.7-4.7

<.001

2.8

1.7-4.6

<.001

Antibiotics (<6 months)

      

2.6

1.0

0.9-1.2

.890

   

Urinary catheter

4.5

2.0-10.3

.018

4.3

1.9-9.6

<.001

72.4

2.9

0.7-12.8

.152

   

Tracheostomy

      

3.1

1.1

0.6-2.0

.867

   

C-statistics for validation dataset

Cross-validation, mean (SD)a

0.643 (0.029)

0.663 (0.028)

 

0.653 (0.031)

0.687 (0.030)

Athens

0.598

  

0.610

   

0.606

  

0.640

  

Barcelona

0.762

  

0.797

   

0.798

  

0.797

  

Cremona

0.585

  

0.641

   

0.601

  

0.670

  

Paris

0.585

  

0.579

   

0.627

  

0.686

  
  1. Note. The Stepwise model used stepwise backward elimination; the Best BMA model was the model with the highest posterior probability with the Bayesian model averaging approach; the BMA model included all covariates with a posterior probability of greater than zero using Bayesian model averaging; the Simple model included variables selected at least half the time in both Stepwise and BMA models on repeated random sub-samples of 50% of the cohort. CI, confidence interval; OR, odds ratio; SD, standard deviation.
  2. aCross-validation by repeated random sub-sampling of 50% of the full cohort data for derivation and validation datasets.