Modelo predictivo de bacteriemia en los pacientes atendidos en el servicio de urgencias por infeccion

Zusammenfassung

SUMMARY TITLE: PREDICTIVE MODEL OF BACTERAEMIA IN PATIENTS ATTENDED IN THE EMERGENCY SERVICE FOR INFECTION AIM AND RATIONALE Nowadays, an infectious process is finally diagnosed in 15% of the patients attended in hospital Emergency Departments (HED). In 43% of these cases, samples are obtained for microbiological studies. Blood cultures (BC) are the most frequent samples collected and are obtained in 15% of the patients with infection attended in the HED. Bacteraemia is defined as the presence of bacteria in the blood, which is evidenced by the isolation of these in BCs. Despite the availability of new quick detection techniques, BCs continue to be the principal tool used for the aetiological diagnosis of bacteraemia, providing information about the sensitivity of the microorganism isolated and favoring the optimization of antimicrobial treatment. At present, the incidence of community bacteraemia has risen to 1-2/1,000 healthcare visits in the HED and to 10 episodes / 1000 hospital admissions from these devices. In relation to the source of true bacteraemia (TB), urinary tract and respiratory track infections are the most frequent with 45-55% and 10-25%, respectively, while bacteraemia of unknown focus is around 10%. The aetiology of these cases is due to Gram-negative bacteria in 65-70% (Escherichia coli,) to Gram-negative 30-35% of the cases (Sthapylococcus aureus and Streptococcus pneumoniae) and anaerobic bacteria in around 1 % of the cases. The 30 day-crude mortality of patients with positive BCs obtained in the HED has been estimated to be from 10-25%, in relation to the severity of the clinical status (presence of sepsis-septic shock), the primary source of infection (urinary, respiratory, abdominal, nervous system, unknown), and the characteristics of the patients (age, comorbidity, particular situations). One of the controversies that arise when indicating the extraction of BCs in HED is its diagnostic profitability since it is very variable (2-20%). On the other hand, it is considered optimal when the frequency of “contaminated BCs” is less than 3%. But, in reality, they can reach much higher percentages. In addition, BCs with significant isolation in patients discharged from the emergency department may represent 3-5% of those extracted in the HED. These facts represent real problems, as they lead to an increase in the diagnostic tests performed, hospital stay, costs and the administration of unnecessary antibiotic treatments or, high discharges in cases of patients with BCs positives. For all these reasons, the suspicion and confirmation of true bacteraemia has a relevant diagnostic, prognostic significance and forces us to change some of the most important decision to be taken in the HED. Among others, indicate discharge or admission, extract HC, and administer the appropriate and early antimicrobial treatment. In this scenario, the objective of many authors is to find a useful predictive model of bacteraemia combining different clinical, epidemiological and analytical data which can be applied in the HED in order to avoid inadequate discharges and unnecessary admissions and their consequences. Biomarkers of inflammatory response and infections (BMIRI) are increasingly included among the analytical variables since they significantly increase the predictive power of these models. Taking this into account, the aim of this study is developing a predictive model of bacteraemia, which is useful and can be applied in the HED. WORKING HYPOTHESIS As mentioned in the previous section (introduction), the suspicion and detection of significant bacteraemia is very important, because of its relevant diagnostic significance, prognosis and because it forces us to change some of the most important decisions to be taken in HED (discharge or admission, indication of extracting HC, administering the appropriate and early antimicrobial, etc.), therefore, our work is based on the hypothesis of knowing the identifiable factors of bacteraemia in HED, studying different clinical, epidemiological and analytical variables , among them the BMRII, to be able to raise a simple and applicable predictive model of bacteraemia in all HED and, after that, to ensure that their use improves suspicion of bacteraemia and, in this way, the quality of patient care. OBJECTIVES The working hypothesis and the general line of research taken for the three research projects comprised In the present dissertation led to the following main objectives: S1: To analyze and compare the ability of procalcitonin (PCT), C-reactive protein (CRP) and leukocytes to differentiate true bacteremia from culture contamination in patients seen in the emergency department (ED) for an episode of infectious disease. S2: To analyze predictive factors of bacteraemia in patients seen in the emergency department (ED) for an episode of infectious disease. S3: To develop a simple risk score to predict bacteremia in patients in our hospital emergency department for infection. METHODS Three studies (S1 to S3) were designed in order to achieve the above objectives at Complejo hospitalario de Toledo. S1: Observational, retrospective and descriptive analytical study of all blood cultures with positive growth extracted in an ED in adult patients (18 years) during the years 2016 and 2017. The follow-up was carried out during 30 days to calculate the predictive power and the prognostic performance for the true bacteremia. S2:Observational,retrospective and descriptive analytical study of all blood cultures extracted in an ED in adult patients (18 years) seen in ED due to infectious disease from 1-1-2019 to 1-7-2019. The follow-up was carried out during 30 days. Thirty-eight variables for predicting bacteraemia were assessed. They covered epidemiological, comorbidity, functional, clinical and analytical factors. Univariate and multivariate logistic regression analysis was performed. S3: Retrospective observational cohort study of all blood cultures ordered in the emergency department for adults (aged 18 or older) from July 1, 2018, to March 31, 2019. We gathered data on 38 independent variables (demographic, comorbidity, functional status, and laboratory findings) that might predict bacteremia. Univariate and multiple logistic regression analyses were applied to the data and a risk scale was developed. RESULTS S1: A total of 266 blood cultures with positive growth were finally enrolled in the study. Of those were considered true bacteremia 154 (57,9 %) and as contaminants blood cultures 112 (42,1 %). The area under the Receiver Operating Characteristic curve (AUC- ROC) for PCT to predict true bacteremia was 0,983 (95% CI: 0,972-0,994; P< .001) and, considering a cut-off value 0,43 ng/ml, PCT achieved 94% sensivity, specificity, positive predictive value of 94%, and negative predictive value of 92%. The AUC-ROC for CRP was 0,639 (95% CI: 0,572-0,707, P< .001), for leukocytes of 0,693 (95% CI: 0,630-0,756, P< .001) and for immature leukocytes (>10% bands) of 0,614 (95% CI: 0,547-0,682, P< .001). The mean levels for PCT were 3,44 (6,30) ng/ml in true bacteremiavs.0,16 (DE 0,18) ng/ml in contaminants blood cultures (P <.001). S2: A total of 1.425 blood cultures were finally enrolled in the study. Of those were considered true bacteraemia 179 (12,6 %) and as negative blood cultures 1.246 (87,4 %). Amongst negatives, 1.130 (79,3%) without growth and 116 (8,1%) as contaminants blood cultures. Five variables were significantly associated with true bacteraemia: serum procalcitonin (PCT) 0,51 ng/ml [odds ratio (OR): 4.52; 95% confidence interval (CI): 4.20- 4.84, P<.001], temperature > 38,3°C [OR:1.60; 95% CI:1.29-1.90, P<.001], systolic blood presure (SBP) < 100 mmHg [OR:3.68; 95% CI:2.78-4.58, P<.001], septic shock [OR:2.96; 95% CI:1.78-4.13, P<.001] and malignancy [OR:1.73; 95% CI:1.27-2.20, P<.001]. S3: A total of 2181 blood samples were cultured. True cases of bacteremia were confirmed in 262 (12%). The remaining 1919 cultures (88%) were negative. No growth was observed in 1755 (80.5%) of the negative cultures, and 164 (7.5%) were judged to be contaminated. The 5MPB-Toledo model identified 5 predictors of bacteremia: temperature higher than 38.3°C (1 point), a Charlson comorbidity index of 3 or more (1 point), respiratory frequency of at least 22 breaths/min (1 point), leukocyte count greater than 12 000/mm3 (1 point), and procalcitonin concentration of 0.51 ng/mL or higher (4 points). Low risk for bacteremia was indicated by a score of 0 to 2 points, intermediate risk by 3 to 5 points, and high risk by 6 to 8 points. Bacteremia in these 3 risk groups was predicted for 1.1%, 10.5%, and 77%, respectively. The model’s area under the receiver operating characteristic curve was 0.946 (95% CI, 0.922–0.969). CONCLUSIONS S1: In blood cultures with positive growth extracted in an ED, PCT has got the best prognostic performance of true bacteremia vs. contaminants blood cultures, better than CRP and leukocytes S2: Several factors evaluated in an initial assessment in the ED, including serum PCT, temperature, hypotension (with/without septic shock) and being malignancy, were found to predict true bacteraemia. S3: The 5MPB-Toledo score could be useful for predicting bacteremia in patients attended in hospital emergency departments for infection. KEYWORDS Emergency Department. Bacteraemia. Prognostic scale. Blood cultures. Procalcitonin. C-reactive protein. Predictive factors. Predictive model. Biomarkers.