We performed a retrospective analysis on all patients reviewed concurrently by ID physicians and the ASP between 1^st August 2011 and 30^th November 2012 at our institute – a 994-bed tertiary university hospital. ID physicians provide a standard referral service whereas the prospective audit and feedback system of our ASP is also standard and has been described elsewhere[4, 5]. Clinical and demographic data were extracted from our ASP database as well as the hospital electronic records. The Charlson comorbidity index was used as an aggregate measure for patients’ comorbidity prognostication[6].

The unit of measurement was a patient that had been reviewed within one week by both an ID physician and the ASP. Patients where differing therapy-modifying recommendations were made by the ID physician and the ASP were selected for more in-depth analysis of outcomes. Outcome measures include 30-day mortality post-ASP review, unplanned re-admission within 30 days post-discharge from hospital, and clinical deterioration at 7 days post-ASP review – a composite measure comprising persistence of fever (if febrile), failure of microbiological clearance, and lack of general physiological improvement defined by the primary physicians’ records.

Intercooled Stata 12.1 (StataCorp, College Station, USA) was used for all statistical calculations, with the level of significance set at 5%. For determination of covariates associated with the various outcomes, we performed logistic regression, with a multivariate model based on covariates identified in the univariate analysis (p-value < 0.200) using the likelihood ratio test, with probability of removal set at 0.05.

The NUH Medical Board approved this work. Informed consent was not obtained from individual patients as both the operations of ID physicians and the ASP constituted routine clinical practice, and only anonymized data were analyzed.

The National University Health System funded this work internally and had no role in study design, analysis or write-up of the results.

The majority of patients were from the surgical disciplines (54.5%) or the hematology-oncology department (33.6%). There was no significant difference in terms of demographic or clinical characteristics for patients where the ASP recommendations had been accepted or rejected in favor of ID physicians’ recommendations, although physicians of patients with bone and joint infections had higher rates of acceptance of ID physicians’ recommendations, with the converse being true for physicians of patients with undifferentiated fevers or where no infection was present.

There was no difference in 30-day mortality or 30-day re-admission rates between the two groups, although clinical deterioration rates were lower in patients where the ASP recommendations had been accepted. Six of 10 deaths occurred in patients with advanced cancers who were on palliative care.

The most effective way to implement antimicrobial stewardship is not known at this point in time[7], but is probably strongly influenced by local practices, history and culture. Both ID physicians (not involved in antimicrobial stewardship) and ASPs should ideally support the concept of antimicrobial stewardship and appropriate prescription of antimicrobial agents[1, 7]. However, there may be differences in the perceived importance of various aspects of clinical care, leading at times to radically different antimicrobial prescribing recommendations. As an example, institutional antimicrobial resistance rates and the differential costs of antibiotics may matter far less to an ID physician than an ASP, resulting in a more conservative approach to de-escalating or stopping antibiotics in patients that have more complications or a slower clinical response, regardless of culture results.

In an institution with both ID physicians and an ASP, there may also be professional friction if one is seen to encroach on the clinical domain of the other. Patients’ primary physicians may also be confused if conflicting recommendations were received from the ID physicians and the ASP. In the majority of local institutions, the ASPs do not audit patients who are under the active review of an ID physician (data not shown). Nonetheless, there may be benefits for patients – particularly in large clinical institutions with high patient workload – if ASPs and ID physicians provide independent reviews and cross-checks, providing an answer to the perennial question of "quis custodiet ipsos custodes" (who shall watch the watchmen) in the context of antimicrobial prescribing and stewardship. Our results suggest that such an approach can be complementary. Concordant recommendations were made for the vast majority of patients reviewed by both ASP and ID physicians. Where recommendations have differed, patients did not have worse outcomes when the ASP’s recommendations were preferred. Although patients where the ASP recommendations were accepted appeared to have better clinical response at Day 7 post-recommendation, it is plausible that this is confounded by primary physicians preferring ID physicians’ recommendations over ASP recommendations in the setting where patients had been more ill or were clinically deteriorating. Unfortunately there were insufficient data and patients to clarify this. Anecdotally, primary physician confusion was minimal, with most teams recognizing the different roles and review processes of the ID physicians and the ASP.

A compromise approach would be to initiate discussions between ID physicians and the ASP each time there is a disagreement about antimicrobial prescribing recommendations. This will reduce although probably not eliminate conflicting recommendations to the patients’ primary physicians, and may be ideal in most institutional settings where the clinical workload is not overwhelming.

Our other results are unsurprising in finding that higher age and having a hematological or oncological (the patients mostly had hematological malignancies) condition were associated with subsequent clinical deterioration, and most unplanned admissions for patients from this department were for the development of febrile neutropenia post-chemotherapy. Because data for calculating acute physiological scores were not collected as part of the study, and the number of deaths was relatively small, there were no significant co-variates associated with 30-day mortality found.

Our audit is primarily limited by the relatively small patient numbers and the methodology. The definition of clinical deterioration was also very broad and data collection for this variable was performed as a routine function of the ASP (and therefore not blinded with respect to whether ASP recommendations were accepted). Nonetheless, the results appear robust and it is difficult if not impossible to run a prospective study within a single institution – perhaps cluster randomization will be required to determine if an approach of independent reviews and recommendations by both ID physicians and ASPs will positively benefit clinical care. We have also not rigorously audited the results in terms of the cost-savings of ASP recommendations vis-à-vis ID physicians’ recommendations. One final major limitation of this audit is that the findings may not be applicable outside of our institution or country, as cultural differences may significantly affect the success or failure of such practices.

In conclusion, we have demonstrated that independent reviews by both ID physicians and ASPs can be compatible within a large tertiary university hospital, providing primary physicians even in situations of conflicting recommendations viable alternative antimicrobial prescribing recommendations. This approach will also not result in social or professional conflict between ID physicians and ASPs provided that communication is good. Initiating early discussions between ID physicians and ASPs in situations where antimicrobial prescribing advice differs may significantly reduce the number of conflicting recommendations.