P79 Automated cell count for the diagnosis of spontaneous bacterial peritonitis: is it useful in clinical practice?

IntroductionSpontaneous bacterial peritonitis (SBP) is both a common and severe complication of ascites. It carries a mortality rate of 11–19.1%,1–3 thus early diagnosis and treatment is imperative in this vulnerable group.The incidence of SBP ranges from 10–30% in hospital in-patients with cirrhotic liver disease.4 However, asymptomatic outpatients carry a much lower rate.5 A recent UK-wide report observing both inpatients and outpatients found a total SBP rate of 3.13%,6 though underreporting may have affected this.The gold standard for diagnosing SBP is an ascitic fluid manual cell count (>250 mm3 polymorphonuclear leukocytes).7 Our trust does not have access to same day manual counts and therefore relies on automated cell count for initial diagnosis. Our trust was identified to have a higher than expected rate of SBP compared to the UK average (11.01% vs. 3.13%).6 Further to this, a local audit of ascitic samples identified 18.9% were positive for SBP, a significant outlier in the national trends. We reviewed our practice to establish the validity of automated cell count as a diagnostic method and establish its usefulness in the diagnosis of SBP.MethodWe obtained a list of patients who had a fluid sample analysis between April 2018-April 2019 (n=300). Non-ascitic or non-processed samples were excluded. Samples were included for analysis if both an automated and manual cell count (gold standard) were sent. 211 patients met the inclusion criteria and results were reviewed using the electronic patient record. 103 (48.9%) were excluded for having automated count only and a further 10 (4.7%) for having one sample not suitable for analysis. 98-paired samples (46.4%) met inclusion criteria for analysis.Results20 automated samples were positive for SBP, of which 3 were positive on the corresponding manual count (positive predictive value (PPV) 15%). It must be noted that the negative predictive value (NPV) was 100% (n=78). Of 103 automated only samples, there were 37 positive results. With a PPV of 15% we would expect a further 5.5 cases. Therefore, potentially 31.5 cases of SBP were over diagnosed due to our reliance on the automated result.DiscussionA PPV of 15% suggests the automated count has little value in clinical practice. Its benefit lies in its strong NPV to rule out SBP, but reliance on this method results in inflated SBP rates and overtreatment with potentially harmful antibiotics.ReferenceFernández J, Acevedo J, Prado V, et al. Clinical course and