Literature Update - April, 2007
Compiled by Dr Emma McBryde, VICNISS ID Physician
NNIS Score Performs Badly in Predicting Risk of Surgical Site Infection after Paediatric Cardiac Surgery
Studies to date have shown that the NNIS score does not discriminate well for adults undergoing cardiac
surgery. This is because the wounds are almost always clean and the ASA score is almost always 3 or more.
Thus discrimination relies on duration of surgery alone. The study by Kagen et al (ICHE 2007; 28(4): 398-405)
confirms the findings of others for the paediatric population and goes further to describe potential
predictors that perform better than the NNIS score for this group. The best model for predicting
risk of SSI is surgery duration >300mins, age <30 days and pre-operative use of inotrope.
Prospective validation of the new scoring system is required.
Doctors may be Stealth Dosing Restricted Antibiotics
A university in the US uses an antimicrobial stewardship program, similar to many in Australia.
Approval is required prior to antibiotic use by contacting either an ID physician or pharmacist.
This operates from 8am to 10pm. After 10pm prior approval is not required. The study by LaRosa et al
(ICHE May 2007 28(5): in press) showed that physicians are more likely to prescribe restricted
antibiotics for the hour after 10pm compared with 9-10pm, suggesting that physicians are waiting until
restrictions are lifted before prescribing antibiotics.
VICNISS Study on Type 2 Hospitals Shows High Specificity but Low Sensitivity in Finding
Methicillin-resistant Staphylococcus aureus and Bloodstream Infections
Bennett NJ et al (ICHE 2007 28:4; 486-8) assessed the quality of data reported to a smaller-hospital
pilot surveillance program for primary BSI and MRSA detection. The study found that the surveillance of
MRSA had a sensitivity of 40% a specificity of 99.9% and a positive predictive value of 33.3%. For BSI,
the sensitivity was 42.9%, the specificity was 99.8% and the positive predictive value was 37.5%. Three
MRSA infections and 4 primary BSIs were found on review that were not reported to VICNISS.
Additionally three MRSA infections reported were not regarded as infections on review. These include one
colonised not infected one previously reported infection and one sensitive Staphylococcus
aureus. Of the BSIs reported, four were regarded as secondary to infection at other sites
and one was thought to be a contaminant.
This suggests that there is both under and over-reporting of MRSA and primary BSIs. Reasons for this
include difficulty distinguishing primary and secondary BSIs distinguishing colonisation from infection
in the case of MRSA.
Guidelines for Reporting Outbreak and Intervention Studies of Nosocomial Infection
Several problems are faced in reporting nosocomial infections and these have been addressed by the ORION
statement (Stone SP et al Lancet Infectious Diseases 2007; 7: 282-88). The issues addressed include:
- Need to distinguish between planned interrupted time series and outbreak reports. This is necessary
because regression to the mean following outbreaks makes it likely that there will be decline in prevalence and
incidence over time regardless of measures instituted.
- Aim and rationale of study: particularly whether the data of the study prompted any decisions to
intervene, what type of study, e.g. prospective, retrospective.
- Summary table: detailing precise nature of patients, settings and interventions.
- Description of intervention: avoid using general terms, specify exact nature of intervention.
- Documentation of potential confounders and bias: identify bias, confounders. Attempts should be made
to minimise, and adjust for these threats to validity of inference.
- Appropriate statistical analysis: techniques should be detailed in the methods section.
In general techniques that assume serial independence of outcomes are invalid when dealing with infection
or colonisation. It is also recommended that at least 12 monthly data points be collected before and after
any intervention.
- Economic analysis: not essential, but if one is to occur it must be comprehensive and document all
resources including staff time.
- Harms: all potential harms need to be specified before the study takes place.
Perusing the list above, it is evident that very few infection control studies published to date
would achieve all of the ORIEN recommendations. Infection control interventions have specific
challenges. The first is that randomisation is often logistically and ethically impossible. Even when
randomisation is planned, mixing of patients in different randomised groups must be avoided to maintain
study validity. Secondly, interventions often occur in response to outbreaks and are performed to control
the problem, not for the sake of scientific discovery. In this context, it is often not reasonable to
restrict a control plan to a single intervention. Studies of this nature are subject to regression to
the mean so that anything that is put in place at the peak of an outbreak may appear to work.
The ORIEN statement does not suggest that such studies not be reported but that the study design
be carefully documented.
Planned interrupted time series are better designed to answer study questions regarding the affect of
interventions.
Because of the dependent nature of colonisation or infection events (i.e. each colonisation depends on
the number of people already colonised in the unit) it is recommended that statistical analysis take this
into account, usually requiring consultation with a statistician prior to study commencement.