Technical Review Article

Title: An algorithm to estimate the importance of bacterial acquisition routes in hospital settings.

Authors: M. C. J. Bootsma, M. J. M. Bonten, S. Nijssen, A. C. Fluit, and O. Diekmann.

Introduction

The significant increase in antibiotic resistance amongst pathogens is making it very difficult to successfully treat infections, especially in intensive care units (ICU’s). Prevention of the spread of infection among patients within the hospitals is fast becoming amongst the most important methods for controlling infections. This requires the identification of the different acquisition routes, that is, routes by which bacterial colonization occurs. In this article, the authors analyzed the relative importance of various bacterial acquisition routes that resulted in colonization of the bacteria using data from individual patients.[1] This article was chosen because of the impact it can have on the healthcare system if the knowledge obtained from the algorithm regarding the most prevalent colonization routes in hospitals can help in the preventing spread of infections. Also the ability of the algorithm to incorporate specific patient characteristics makes it both novel and appealing.

Core Evaluation

Hypothesis Development

In this article, the authors present an algorithm that can be used to quantitatively determine the likelihood of colonization occurring from different acquisition routes. This algorithm uses individual patient data regarding the duration of their stay in the ICU and specific characteristics which builds on previously proposed algorithms that only consider the number of patients colonized. The authors state that the algorithm is “a promising tool for disentangling the contributions of various acquisition routes on the basis of longitudinal data without requiring labor intensive and costly genotyping procedures”,[1] which shows that they believe that the algorithm can give reliable results regarding the contribution of each route for a specific patients in a cost effective and efficient manner. Thus in this study the authors build on the Markov model developed by Pelupessy et al,[2] by incorporating more information specific to individual patients without any added assumptions, thereby making it more reliable, closer to reality and more deterministic.

Data Review

In the experiments performed by the authors, the algorithm is tested by analyzing data obtained from two ICU’s, a medical ICU (ICU-1) and a neurosurgical ICU (ICU-2), regarding colonization by a third generation cephalosporin-resistant Enterobacteriacease (CRE) and comparing the conclusions drawn from the results obtained on using the algorithm to the conclusions obtained from genotyping data. The authors used a “mechanistic acquisition model” to determine the changes in the colonization status of the patients.[1] In order to use this model they had to make a set of six assumptions that mainly talked about the criteria based on which the algorithm differentiated between colonized and un-colonized.[1] These assumptions can to lead to errors in detecting certain cases that are exceptions to these criteria. However it the authors have designed the algorithm such that some of the exceptions that occur due to the assumptions can be incorporated into the result.

One major flaw in the paper was that the authors did not clearly define exogenous and endogenous transmission. Since these are core concepts behind the article they should have been better defined. However on reading the paper it appeared that endogenous acquisition is when the pathogen already exists in the body of the patient and grows to a detectable amount and exogenous acquisition is when the pathogen is transmitted from another patient through a number of different routes.[1]

The algorithm allows input of the individual admission, discharge and culture times of each patient and it outputs the maximum likelihood estimates (MLEs) and the confidence parameters of the endogenous and the cross-transmission acquisition parameters.[1] On studying this data, the most dominant route can be easily identified along with the probablility that a patient is colonized at any instant during their stay. In this study genotyping and epidemiological linkage were used as the reference standards to which the algorithm results were compared. It was seen that the endemic prevalence (mean daily prevalence of colonization) in ICU-1 and ICU-2 were 26.1% (standard deviation of 15.4) and 15.1% (standard deviation of 15.4) from genotyping and 27.6% and 17.6% from model predictions, respectively.[1] The algorithm concluded that less than 50 % of all acquisitions were a result of cross-transmission which was consistent with the conclusion drawn from genotyping data that showed that majority of acquisitions were endogenous in nature.[1] Thus the authors were able to prove their argument and show that the algorithm was able to determine the most important acquisition route consistent with genotyping. However, their presentation of the results were hard to interpret and how they derived their conclusion was difficult to understand from the data.

Conclusion

On analyzing the data the authors come to the same conclusion as me. We concluded that the algorithm and the genotyping process led to the same result that the endogenous route was the predominant acquisition route for colonization with CRE. The authors stated that this showed that the algorithm was a cost effective, efficient and reliable method of determining the most important acquisition route for individual