Toxicidad tubular renal en pacientes vih positivos en tratamiento antirretroviralprevalencia, papel diagnóstico de los biomarcadores renales y riesgo de progresión a insuficiencia renal

Abstract

Introduction and objectives: Nowadays performing a blood culture in the emergency department depends on the ability of the clinician to suspect the presence of bacteraemia. The objectives of this thesis are several. The first objective is to assess the ability of clinicians to predict clinical bacteraemia. The second objective of this thesis is to create a clinical predictive model (CPM) to detect bacteraemia, to subsequently compare the predictive ability of the clinicians versus the predictive ability of this model. The third objective of this thesis is to validate the new model created in another hospital from our national health system. Material and methods: This is a cross-sectional study which (I) analyses the predictive factors of bacteraemia in an emergency department setting and (II) establishes a CPM of bacteraemia. Inclusion criteria includes patients over 13 years old who presented to the emergency department at “Hospital Universitario de Guadalajara” between the 1st of April 2005 and the 31st of March 2007, and who had a blood culture performed. Initially a univariate statistical analysis was done, using as a dependent variable the existence of bacteraemia. Comparison of categorical data was done using a chi-squared or a Fisher's exact test, and for quantitative variables a logistic regression model was used. A CPM that established the exact probability of bacteraemia using the logistic regression formula was developed. The significant variables in the univariate analysis were introduced in a multiple logistic regression model, with a stepwise regression system, using the existence of bacteraemia as a dependent variable. Independent predictor variables of bacteraemia were identified using logistic regression, and an equation was developed using these variables to estimate the specific risk of bacteraemia for each individual patient. We calculated the sensitivity, specificity, negative and positive predictive values, and ROC curve of the logistic regression model. The validation was performed using the variables included in the model and proving its association with the presence of bacteraemia in another hospital. Results: In total 693 patients were analysed, of which 149 had bacteraemia. The variables independently associated with the diagnosis of bacteraemia were: diabetes (OR 2.17; 95% CI 1.16 to 4.07; p=0.016), treatment with antibiotics prior to blood culture sampling (OR 0.16 ; 95% CI 0.08 to 0.33; p=0.00001), having more than 80% neutrophils (OR 3.4; 95% CI 1.88 to 6.28; p=0.001), having an onco-haematological disease (OR 1.72, 95% CI 0.98 to 3.02; p=0.058), age/10 (OR 1.16, 95% CI 1.02 to 1.31; p=0.023) and increased plasmatic fibrinogen/100 (OR 1.2; 95% CI 1.08 to 1.33; p=0.0005). Clinicians were able to predict bacteraemia with a sensitivity of 82.2%, a specificity of 28.9%, a positive predictive value of 0.155 and a negative predictive value of 0.911. The area under the ROC curve was 0.59 (95% CI 0.52 to 0.66). Using the logistic regression formula, and for a cut-off point of 0.1, the CPM would present a specificity of 27.5%, a sensitivity of 96.7%, a positive predictive value of 0.362 and a negative predictive value of 0.952. The area under the ROC curve would be 0.77 (95% CI 0.72-0.82). Using this CPM would prevent up to 20.3% of blood cultures performed in an emergency setting. Four hundred and eleven consecutive patients who presented to the emergency department of the “Hospital de Talavera” and who had a blood culture taken were used for external validation of the model. The 6 variables derived from the created CPM were collected prospectively. Only 3 of these 6 variables had a statistically significant association with the presence of bacteraemia: the presence of neutrophilia, with more than 80% neutrophils in the blood differential test, age and previous antibiotic treatment. These three variables, which had a confirmed association with bacteraemia in both populations, were used to create an CPM called parsimonious model. With this model, the area under the curve for the population of Guadalajara, with a cut-off point of 0.1, was 0.71 (95% CI 0.66 to 0.75), with a sensitivity of 97.2%, a specificity of 23.3%, a negative predictive value of 0.963 and a positive predictive value of 0.281. The area under the curve with this model for the town of Talavera de la Reina and for the same cut-off point was 0.73 (95% CI 0.67 to 0.79), with a sensitivity of 95.7%, a specificity of 36.7%, a positive predictive value of 0.24 and a negative predictive value of 0.98. Using this model 18.3% of blood cultures in the city of Guadalajara and 31.1% of blood cultures in the town of Talavera de la Reina would have been avoided. The models were validated in both populations, presenting similar ROC curves and a statistically nonsignificant p value. The model is especially useful in patients that are taking antibiotics and visit the emergency department with suspected bacteraemia. Conclusions: The models created to detect bacteraemia are superior to clinical intuition. Age, onco-haematological disease, diabetes, taking antibiotics prior to blood culture sampling, neutrophilia and increased fibrinogen behave as clinical and epidemiological variables that can predict independently the presence of bacteraemia. This models are useful to prevent blood culture sampling to people with low probability of bacteraemia who present to the emergency department. Both models have been validated in another hospital.