Skip to main content

Table 5 Comparison among unadjusted, adjusted and multilevel regression models for estimating the impact of carbapenem resistance

From: Economic and clinical burden from carbapenem-resistant bacterial infections and factors contributing: a retrospective study using electronic medical records in Japan

  Logistic regression   
  OR (95% CI) P-value AIC
Death    
 Unadjusted 1.23 (0.76–2.00) 0.407 10,007
 Adjusteda 1.18 (0.70–1.99) 0.543 10,007
 Multilevela,b 1.24 (0.72–2.11) 0.441 8827
  Log-linear regression†‡
  Percentage change (95% CI) P-value AIC
Total LOS    
 Unadjusted 38.5% (23.0–53.9%)  < 0.001 106,802
 Adjusteda 25.4% (10.0–40.8%)  < 0.001 106,214
 Multilevela,b 42.1% (29.1–55.2%)  < 0.001 104,230
Total Cost    
 Unadjusted 50.1% (37.8–62.5%)  < 0.001 266,426
 Adjusteda 44.7% (33.3–56.2%)  < 0.001 264,132
 Multilevela,b 50.4% (40.9–60.0%)  < 0.001 263,400
  1. The impacts of carbapenem resistance infections were calculated against carbapenem susceptible infections as reference
  2. Continuous outcomes of the LOS and cost were log-transformed. The estimated coefficient β of the carbapenem resistance was transformed as percent change per one-unit increase
  3. aAdjusted for age, sex, BMI, CCI, immunosuppressive drugs, antibiotic use before culture test, ICU admission, operation, disease type, and death (if outcome was death, it was excluded from the model)
  4. bHospital was treated as a random effect
  5. AIC Akaike information criterion, BMI body mass index, CCI Charlson Comorbidity Index, CI confidence interval, CR carbapenem-resistant; CS carbapenem-susceptible, LOS length of stay, OR odds ratio