Comparison of the Andersen–Gill model with poisson and negative binomial regression on recurrent event data

Statistical methods for the analysis of relapse data in MS clinical trials

“…On the contrary, the quasi-Poisson method does not correspond to a probability distribution. Although in our example, the negative binomial model yields a more significant result, the simulation study by Jahn-Eimermacher [23] indicates that the quasi-Poisson method may wrongly reject the null hypothesis of no treatment effect slightly more often than the negative binomial distribution.…”

“…The comparability between the NB and AG models was further demonstrated by means of simulations [22,23]. Under the proportional rate assumption (the relapse rate may not be constant), the counts contain most information about treatment difference [26].…”

Predictors of 10-year hospital use in a community-dwelling population of Brazilian elderly: the Bambuí cohort study of aging

“…We assessed two different approaches to modeling such data. The first, known as the Anderson-Gill model 13 for repeated events includes all subjects, handles dependence of events through clustering among individuals (using robust standard errors), assumes a common exposure effect is sustained over time, and also assumes a common baseline hazard functionthat is, that prior hospitalization will not influence one's chances of a subsequent hospitalization 13,14,15 . The second approach (proposed by Prentice, Williams, and Peterson) is known as a gap time model.…”

Eplerenone in Patients With Systolic Heart Failure and Mild Symptoms

“…10 Alternatively, a survival-based technique is the Andersen-Gill extension of the Cox proportional hazards model. 12 The Poisson and Andersen-Gill regression models, however, both assume independence of events within individuals, an assumption that is clearly violated because recurrent hospitalizations within individuals will be dependent. 10 The negative binomial is considered an attractive distribution to use, because it naturally accommodates the different probabilities for events across members of the population.…”

“…Simulation studies have also shown that the negative binomial produces results that are similar to the Andersen-Gill approach. 12 Rate ratios, 95% confidence intervals, and probability values were calculated with the use of models adjusted for the following prespecified baseline covariates: sex, age, estimated glomerular filtration rate, ejection fraction, body mass index, hemoglobin value, heart rate, systolic blood pressure, diabetes mellitus, and a history of hypertension, myocardial infarction, atrial fibrillation, and a left bundle-branch block or QRS duration Ͼ130 ms. Sensitivity analyses were performed by means of unadjusted models.…”