Skip to main content

Table 3 Conditional logistic regression analysis

From: Why do hospital prescribers continue antibiotics when it is safe to stop? Results of a choice experiment survey

Attribute2

Coefficient

SE

Lower CI

Upper CI

Average marginal effect (AME)1

p

SYMPTOMS

 UTI symptoms with kidney pain

1.132

0.155

0.829

1.436

0.173

< 0.001

 Fever, cough and possible pulmonary infiltrates on chest X-ray

0.246

0.108

0.034

0.459

0.038

0.023

 Unclear symptoms3

− 1.379

–

–

–

–

–

CONFLICT WITH GUIDELINES

 Strongly conflict

1.275

0.131

1.018

1.532

0.194

< 0.001

 Somewhat conflict

0.073

0.089

− 0.101

0.248

0.011

0.411

 No conflict3

− 1.348

–

–

–

–

–

CONTINUE RISK

− 0.085

0.012

− 0.108

− 0.061

− 0.013

< 0.001

STOP RISK

0.172

0.017

0.139

0.205

0.026

< 0.001

PREMORBID CONDITION

 Severe frailty and comorbidities

0.660

0.138

0.390

0.930

0.101

< 0.001

 Moderate frailty and comorbidities

0.300

0.098

0.108

0.492

0.046

0.002

 Fit and well3

− 0.960

–

–

–

–

–

EXTERNAL PRESSURE

 No pressure

− 0.660

0.097

− 0.851

− 0.469

− 0.101

< 0.001

 Some pressure

− 0.104

0.094

− 0.289

0.080

− 0.016

0.268

 Heavy pressure3

0.764

–

–

–

 

–

 McFadden’s R2 (pseudo-R2)

0.320

     

 AIC/BIC

976.767/1029.697

     

 Log-likelihood

− 478.384

     

N

1470

     
  1. 1AME is the average marginal effect of each factor level on the probability of choosing to continue. For the categorical factor levels, this indicates how much higher/lower the probability of continuing was at this attribute level than the probability at the factor’s base level. For the continuous variables (risk of continuing/risk of discontinuing), the AME indicates how much higher/lower the probability of continuing was for a 1% higher risk. AIC Akaike Information Criterion, BIC Bayesian Information Criterion, CI 95% confidence interval. p p value of coefficient, SE standard error clustered at the respondent level
  2. 2Attribute descriptions: SYMPTOMS = patient’s presenting symptoms (1 = UTI and kidney, 2 = fever cough and funny X-ray, 3 = unclear [base level]); CONFLICT WITH GUIDELINES = whether early discontinuation of antibiotic treatment within 72 h of treatment initiation would be in conflict with local antibiotic guidelines (1 = strongly conflict, 2 = somewhat conflict, 3 = not conflict [base level]); CONTINUE RISK = risk of significant harm arising from continued antibiotic treatment, expressed as a percentage; STOP RISK = risk of significant harm arising from discontinuing antibiotic treatment, expressed as a percentage; PREMORBID CONDITION = premorbid condition of the patient (1 = severe frailty and comorbidities, 2 = moderate frailty and comorbidities, 3 = fit and well [base level]); and EXTERNAL PRESSURE = level of external pressure to continue antibiotic treatment (1 = no pressure, 2 = some pressure, 3 = heavy pressure [base level]). The sample size arises from 98 respondents times 15 choice questions. The 2 respondents who chose ‘continue’ in all 15 choice questions were omitted because of the lack of variation in their responses
  3. 3These attributes were effects-coded. The coefficients of the base levels (unclear symptoms, no conflict, fit and well and heavy pressure) were calculated as the negative sum of the coefficients of the other levels
Back to article page

BMC Medicine

ISSN: 1741-7015

Contact us