Developing a length of stay prediction model for newborns, achieving better accuracy with greater usability.

BACKGROUND: One in ten newborn children is born prematurely. The elongated length of stay (LOS) of these children in the Neonatal Intensive Care Unit (NICU) has important implications on hospital occupancy figures, healthcare and management costs, as well as the psychology of parents. In order to allow accurate planning and resource allocation, this study aims to create a generalizable and robust model to predict the NICU LOS of preterm newborns. METHODS: Data were collected from a large tertiary center NICU between 2011 and 2018 and relates to 5,362 newborns. The selected model was externally validated using a data set of 8,768 newborns from another tertiary center NICU. This report compares several models, such as Random Forest (RF), quantile RF, and other feature selection methods, including LASSO and AIC step-forward selection. In addition, a novel step-forward selection based on False Discovery Rate (FDR) for quantile regression is presented and evaluated. RESULTS: A high-orderquantile regression model for predicting preterm newborns' LOS that uses only four features available at birth had more attractive properties than other richer ones. The model achieved a Mean Absolute Error (MAE) of 6.26 days on the internal validation set (average LOS 27.04) and an MAE of 6.04 days on the external validation set (average LOS 29.32). The suggested model surpassed the accuracy obtained by models in the literature. It is shown empirically that the FDR-based selection has better properties than the AIC-based step-forward selection approach. CONCLUSION: This paper demonstrates a process to create a predictive model for NICU LOS in preterm newborns, where each step is reasoned. We obtain a simple and robust model for NICU LOS prediction, which achieves far better results than the current model used for financing NICUs. Utilizing this model, we have created an easy-to-use online web application to ease parents' worries and to assist NICU management: https://tzviel.shinyapps.io/calcuLOS.

Deep-Learning-Based Screening and Ancillary Testing for Thyroid Cytopathology.

Thyroid cancer is the most common malignant endocrine tumor. The key test to assess preoperative risk of malignancy is cytologic evaluation of fine-needle aspiration biopsies (FNABs). The evaluation findings can often be indeterminate, leading to unnecessary surgery for benign post-surgical diagnoses. We have developed a deep-learning algorithm to analyze thyroid FNAB whole-slide images (WSIs). We show, on the largest reported data set of thyroid FNAB WSIs, clinical-grade performance in the screening of determinate cases and indications for its use as an ancillary test to disambiguate indeterminate cases. The algorithm screened and definitively classified 45.1% (130/288) of the WSIs as either benign or malignant with risk of malignancy rates of 2.7% and 94.7%, respectively. It reduced the number of indeterminate cases (N = 108) by reclassifying 21.3% (N = 23) as benign with a resultant risk of malignancy rate of 1.8%. Similar results were reproduced using a data set of consecutive FNABs collected during an entire calendar year, achieving clinically acceptable margins of error for thyroid FNAB classification.

Early signaling of bacteremia in patients who present to the department of emergency medicine with relatively low C-reactive protein (CRP) concentrations.

OBJECTIVES: Examiningthe usefulness of C-reactive protein velocity (CRPv) as an early biomarker for the presence of bacteraemia in patients presenting to the Department of Emergency Medicine with acute infection/inflammation and suspected bacteraemia. METHODS: A retrospective study examining a cohort of patients who presented to the E.R and in whom blood cultures were taken. CRPv was calculated as the difference in mg/hour/litter between two consecutive CRP tests performed within 12 h. RESULTS: 256 patients were included in the cohort. Using CRPv in patients who at first presented with a relatively low (17.9 ≤ mg/L 1stquartile) CRP concentration, we found an AUC of 0.808 ± 0.038 (p < 0.001) for the presence of positive versus negative blood cultures (what is AUC?). This was better than the AUC that was obtained when the WBC for the same purpose. CONCLUSIONS: CRPv may be a useful biomarker in the identification of patients with suspected bacteremiaand a low CRP-a challenging situation for clinicians who may underestimate the severity of illness in this patient group.

The Liver Can Deliver: Utility of Hepatic Function Tests as Predictors of Outcome in COVID-19, Influenza and RSV Infections.

BACKGROUND: liver test abnormalities have been described in patients with Coronavirus-2019 (COVID-19), and hepatic involvement may correlate with disease severity. With the relaxing of COVID-19 restrictions, seasonal respiratory viruses now circulate alongside SARS-CoV-2. AIMS: we aimed to compare patterns of abnormal liver function tests in patients suffering from COVID-19 infection and seasonal respiratory viruses: respiratory syncytial virus (RSV) and influenza (A and B). METHODS: a retrospective cohort study was performed including 4140 patients admitted to a tertiary medical center between 2010-2020. Liver test abnormalities were classified as hepatocellular, cholestatic or mixed type. Clinical outcomes were defined as 30-day mortality and mechanical ventilation. RESULTS: liver function abnormalities were mild to moderate in most patients, and mainly cholestatic. Hepatocellular injury was far less frequent but had a strong association with adverse clinical outcome in RSV, COVID-19 and influenza (odds ratio 5.29 (CI 1.2-22), 3.45 (CI 1.7-7), 3.1 (CI 1.7-6), respectively) COVID-19 and influenza patients whose liver functions did not improve or alternatively worsened after 48 h had a significantly higher risk of death or ventilation. CONCLUSION: liver function test abnormalities are frequent among patients with COVID-19 and seasonal respiratory viruses, and are associated with poor clinical outcome. The late liver tests' peak had a twofold risk for adverse outcome. Though cholestatic injury was more common, hepatocellular injury had the greatest prognostic significance 48 h after admission. Our study may provide a viral specific auxiliary prognostic tool for clinicians facing patients with a respiratory virus.

Neutrophil-to-lymphocyte ratio trend at admission predicts adverse outcome in hospitalized respiratory syncytial virus patients.

Background and aims: Severe cases of respiratory syncytial virus (RSV) infection are relatively rare but may lead to serious clinical outcomes, including respiratory failure and death. These infections were shown to be accompanied by immune dysregulation. We aimed to test whether the admission neutrophil-to-leukocyte ratio, a marker of an aberrant immune response, can predict adverse outcome. Methods: We retrospectively analyzed a cohort of RSV patients admitted to the Tel Aviv Medical Center from January 2010 to October 2020d. Laboratory, demographic and clinical parameters were collected. Two-way analysis of variance was used to test the association between neutrophil-lymphocyte ratio (NLR) values and poor outcomes. Receiver operating characteristic (ROC) curve analysis was applied to test the discrimination ability of NLR. Results: In total, 482 RSV patients (median age 79 years, 248 [51%] females) were enrolled. There was a significant interaction between a poor clinical outcome and a sequential rise in NLR levels (positive delta NLR). The ROC curve analysis revealed an area under curve (AUC) of poor outcomes for delta NLR of (0.58). Using a cut-off of delta = 0 (the second NLR is equal to the first NLR value), multivariate logistic regression identified a rise in NLR (delta NLR>0) as being a prognostic factor for poor clinical outcome, after adjusting for age, sex and Charlson comorbidity score, with an odds ratio of 1.914 (P = 0.014) and a total AUC of 0.63. Conclusions: A rise in NLR levels within the first 48 h of hospital admission can serve as a prognostic marker for adverse outcome.

Exposure to Non-Antimicrobial Drugs and Risk of Infection with Antibiotic-Resistant Enterobacteriaceae.

Antimicrobial resistance (AMR) has consistently been linked to antibiotic use. However, the roles of commonly prescribed non-antimicrobial drugs as drivers of AMR may be under-appreciated. Here, we studied a cohort of patients with community-acquired pyelonephritis and assessed the association of exposure to non-antimicrobial drugs at the time of hospital admission with infection with drug-resistant organisms (DRO). Associations identified on bivariate analyses were tested using a treatment effects estimator that models both outcome and treatment probability. Exposure to proton-pump inhibitors, beta-blockers, and antimetabolites was significantly associated with multiple resistance phenotypes. Clopidogrel, selective serotonin reuptake inhibitors, and anti-Xa agents were associated with single-drug resistance phenotypes. Antibiotic exposure and indwelling urinary catheters were covariates associated with AMR. Exposure to non-antimicrobial drugs significantly increased the probability of AMR in patients with no other risk factors for resistance. Non-antimicrobial drugs may affect the risk of infection with DRO through multiple mechanisms. If corroborated using additional datasets, these findings offer novel directions for predicting and mitigating AMR.

Risk factors, clinical characteristics and prognostic value of acute kidney injury in COVID-19 compared with influenza virus and respiratory syncytial virus.

BACKGROUND: Acute Kidney Injury (AKI) complicates a substantial part of patients with COVID-19. Direct viral penetration of renal cells through the Angiotensin Converting Enzyme 2 receptor, and indirect damage by the aberrant inflammatory response characteristic of COVID-19 are likely mechanisms. Nevertheless, other common respiratory viruses such as Influenza and Respiratory Syncytial Virus (RSV) are also associated with AKI. METHODS: We retrospectively compared the incidence, risk factors and outcomes of AKI among patients who were admitted to a tertiary hospital because of infection with COVID-19, influenza (A + B) or RSV. RESULTS: We collected data of 2593 patients hospitalized with COVID-19, 2041 patients with influenza and 429 with RSV. Patients affected by RSV were older, had more comorbidities and presented with higher rates of AKI at admission and within 7 days (11.7% vs. 13.3% vs. 18% for COVID-19, influenza and RSV, respectively p = 0.001). Nevertheless, patients hospitalized with COVID-19 had higher mortality (18% with COVID-19 vs. 8.6% and 13.5% for influenza and RSV, respectively P < 0.001) and higher need of mechanical ventilation (12.4% vs. 6.5% vs.8.2% for COVID-19, influenza and RSV, respectively, P = 0.002). High ferritin levels and low oxygen saturation were independent risk factors for severe AKI only in the COVID-19 group. AKI in the first 48 h of admission and in the first 7 days of hospitalization were strong independent risk factors for adverse outcome in all groups. CONCLUSION: Despite many reports of direct kidney injury by SARS-COV-2, AKI was less in patients with COVID-19 compared to influenza and RSV patients. AKI was a prognostic marker for adverse outcome across all viruses.

Improving Hospital Outpatient Clinics Appointment Schedules by Prediction Models.

Patient no-shows and suboptimal patient appointment length scheduling reduce clinical efficiency and impair the clinic's quality of service. The main objective of this study is to improve appointment scheduling in hospital outpatient clinics. We developed generic supervised machine learning models to predict patient no-shows and patient's length of appointment (LOA). We performed a retrospective study using more than 100,000 records of patient appointments in a hospital outpatient clinic. Several machine learning algorithms were used for the development of our prediction models. We trained our models on a dataset that contained patients', physicians', and appointments' characteristics. Our feature set combines both unstudied features and features adopted from previous studies. In addition, we identified the influential features for predicting LOA and no-show. Our LOA model's performance was 6.92 in terms of MAE, and our no-show model's performance was 92.1% in terms of F-score. We compared our models' performance to the performance of previous research models by applying their methods to our dataset; our models demonstrated better performance. We show that the major effector of such differences is the use of our novel features. To evaluate the effect of our prediction results on the quality of schedules produced by appointment systems (AS), we developed an interface layer between our prediction models and the AS, where prediction results comprise the AS input. Using our prediction models, there was an 80% improvement in the daily cumulative patient waiting time and a 33% reduction in the daily cumulative physician idle time.

Surgery duration: Optimized prediction and causality analysis.

Accurate estimation of duration of surgery (DOS) can lead to cost-effective utilization of surgical staff and operating rooms and decrease patients' waiting time. In this study, we present a supervised DOS nonlinear regression prediction model whose accuracy outperforms earlier results. In addition, unlike previous studies, we identify the features that influence DOS prediction. Further, in difference from others, we study the causal relationship between the feature set and DOS. The feature sets used in prior studies included a subset of the features presented in this study. This study aimed to derive influential effectors of duration of surgery via optimized prediction and causality analysis. We implemented an array of machine learning algorithms and trained them on datasets comprising surgery-related data, to derive DOS prediction models. The datasets we acquired contain patient, surgical staff, and surgery features. The datasets comprised 23,293 surgery records of eight surgery types performed over a 10-year period in a public hospital. We have introduced new, unstudied features and combined them with features adopted from previous studies to generate a comprehensive feature set. We utilized feature importance methods to identify the influential features, and causal inference methods to identify the causal features. Our model demonstrates superior performance in comparison to DOS prediction models in the art. The performance of our DOS model in terms of the mean absolute error (MAE) was 14.9 minutes. The algorithm that derived the model with the best performance was the gradient boosted trees (GBT). We identified the 10 most influential features and the 10 most causal features. In addition, we showed that 40% of the influential features have a significant (p-value = 0.05) causal relationship with DOS. We developed a DOS prediction model whose accuracy is higher than that of prior models. This improvement is achieved via the introduction of a novel feature set on which the model was trained. Utilizing our prediction model, hospitals can improve the efficiency of surgery schedules, and by exploiting the identified causal relationship, can influence the DOS. Further, the feature importance methods we used can help explain the model's predictions.

Sepsis Related Mortality Associated with an Inflammatory Burst in Patients Admitting to the Department of Internal Medicine with Apparently Normal C-Reactive Protein Concentration.

Background: Patients who are admitted to the Department of Internal Medicine with apparently normal C-reactive protein (CRP) concentration impose a special challenge due the assumption that they might not harbor a severe and potentially lethal medical condition. Methods: A retrospective cohort of all patients who were admitted to the Department of Internal Medicine with a CRP concentration of ≤31.9 mg/L and had a second CRP test obtained within the next 24 h. Seven day mortality data were analyzed. Results: Overall, 3504 patients were analyzed with a mean first and second CRP of 8.8 (8.5) and 14.6 (21.6) mg/L, respectively. The seven day mortality increased from 1.8% in the first quartile of the first CRP to 7.5% in the fourth quartile of the first CRP (p < 0.0001) and from 0.6% in the first quartile of the second CRP to 9.5% in the fourth quartile of the second CRP test (p < 0.0001), suggesting a clear relation between the admission CRP and in hospital seven day mortality. Conclusions: An association exists between the quartiles of CRP and 7-day mortality as well as sepsis related cause of death. Furthermore, the CRP values 24 h after hospital admission improved the discrimination.

Diagnostic approaches to syncope in Internal Medicine Departments and their effect on mortality.

Most data on mortality and investigational approaches to syncope comes from patients presented to emergency departments (ED). The aim of this study is to report intermediate term mortality in syncope patients admitted to Internal Medicine Departments and whether different diagnostic approaches to syncope affect mortality. Methods and results A single-center retrospective-observational study conducted at the Tel Aviv "Sourasky" Medical Center. Data was collected from electronic medical records (EMRs), from January 2010 to December 2020. We identified 24,021 patients, using ICD-9-CM codes. Only 7967 syncope patients were admitted to Internal Medicine Departments and evaluated. Logistic regression models were used to determine the effects of diagnostic testing per patient in each department on 30-day mortality and readmission rates. All-cause 30-day mortality rate was 4.1%. There was a significant difference in the number of diagnostic tests performed per patient between the different departments, without affecting 30-day mortality. The 30-day readmission rate was 11.4%, of which 4.4% were a result of syncope. Conclusion Syncope patients admitted to Internal Medicine Departments show a 30-day all-cause mortality rate of ∼4%. Despite the heterogeneity in the approach to the diagnosis of syncope, mortality is not affected. This novel information about syncope patients in large Internal Medicine Departments is further proof that the diagnosis of syncope requires a logic, personalized approach that focuses on medical history and a few tailored, diagnostic tests.

Serum Hypoalbuminemia Is a Long-Term Prognostic Marker in Medical Hospitalized Patients, Irrespective of the Underlying Disease.

Hypoalbuminemia is common in hypoalbuminemia-associated disorders (HAD), e.g., liver and kidney disease. We hypothesize that hospitalized patients with hypoalbuminemia have poor prognosis irrespective of their underlying disease. Records of patients admitted to Medicine (2010−2018), with and without HAD were analyzed, comparing low (<35 g/L) to normal serum albumin. Mann−Whitney and Chi-squared tests were used, and a logistic regression model was applied. Patients: 14,640 were admitted; 9759 were analyzed (2278 hypoalbuminemia: 736 HAD, 1542 non-HAD). All patients, and the subgroups with (as expected) and without HAD had worse outcomes. Specifically, in patients without HAD, those with hypoalbuminemia (n = 1542) vs. normal albumin (n = 6216) were older, had a higher Charlson Comorbidity Index (CCI, 5 vs. 4), longer median hospital stay (5 vs. 4), higher one year re-admission rate (49.9% vs. 39.8%), and one year mortality (48.9% vs. 15.3%, p < 0.001 for all). LR model predicting 3 month, 1 year and 5 year mortality confirmed the predictive power of albumin (1 year: OR = 4.49 for hypoalbuminema, p < 0.01). Hypoalbuminemia portends poor long-term prognosis in hospitalized patients regardless of the underlying disease and could be added to prognostic predictive models.

Emergency department electric scooter injuries after the introduction of shared e-scooter services: A retrospective review of 3,331 cases.

BACKGROUND: To determine the frequency, characteristics, and use of resources related to electric scooter (e-scooter) injuries in the emergency department (ED) of a major metropolitan area hospital. METHODS: We performed a retrospective review of all ED presentations related to e-scooter injuries at a level I trauma center between May 2017 and February 2020. We identified ED presentation data, injury-related data, patients' clinical course after evaluation, injury diagnosis, surgical procedures, and ED readmissions. RESULTS: A total of 3,331 patients with e-scooter injuries presented to the ED over a 34-month period. There was a 6-fold increase in e-scooter-related injuries presenting to the ED, from an average of 26.9 injuries per month before the introduction of shared e-scooter services in August 2018 to an average of 152.6 injuries per month after its introduction. The average injury rate during weekdays was 3.27 per day, with the majority of injuries occurring in the afternoon. The most common mechanism of injury was rider fall (79.1%). There were a total of 2,637 orthopedic injuries, of which 599 (22.7%) were fractures. A total of 296 (8.9%) patients were hospitalized following the initial ED admission, and 462 surgeries were performed within 7 days of ED arrival. CONCLUSIONS: The introduction of the shared e-scooter services is associated with a dramatic increase in e-scooter injuries presenting to the ED. E-scooter use carries considerably underestimated injury risks of high-energy trauma and misunderstood mechanisms of injuries. These injuries challenge the healthcare system, with a major impact on both EDs and surgical departments.

Prognostic value of neutrophil-to-lymphocyte ratio in COVID-19 compared with Influenza and respiratory syncytial virus infection.

A high neutrophil to lymphocyte ratio (NLR) is considered an unfavorable prognostic factor in various diseases, including COVID-19. The prognostic value of NLR in other respiratory viral infections, such as Influenza, has not hitherto been extensively studied. We aimed to compare the prognostic value of NLR in COVID-19, Influenza and Respiratory Syncytial Virus infection (RSV). A retrospective cohort of COVID-19, Influenza and RSV patients admitted to the Tel Aviv Medical Center from January 2010 to October 2020 was analyzed. Laboratory, demographic, and clinical parameters were collected. Two way analyses of variance (ANOVA) was used to compare the association between NLR values and poor outcomes among the three groups. ROC curve analyses for each virus was applied to test the discrimination ability of NLR. 722 COVID-19, 2213 influenza and 482 RSV patients were included. Above the age of 50, NLR at admission was significantly lower among COVID-19 patients (P < 0.001). NLR was associated with poor clinical outcome only in the COVID-19 group. ROC curve analysis was performed; the area under curve of poor outcomes for COVID-19 was 0.68, compared with 0.57 and 0.58 for Influenza and RSV respectively. In the COVID-19 group, multivariate logistic regression identified a high NLR (defined as a value above 6.82) to be a prognostic factor for poor clinical outcome, after adjusting for age, sex and Charlson comorbidity score (odds ratio of 2.9, P < 0.001). NLR at admission is lower and has more prognostic value in COVID-19 patients, when compared to Influenza and RSV.

Predicting bloodstream infection outcome using machine learning.

Bloodstream infections (BSI) are a main cause of infectious disease morbidity and mortality worldwide. Early prediction of BSI patients at high risk of poor outcomes is important for earlier decision making and effective patient stratification. We developed electronic medical record-based machine learning models that predict patient outcomes of BSI. The area under the receiver-operating characteristics curve was 0.82 for a full featured inclusive model, and 0.81 for a compact model using only 25 features. Our models were trained using electronic medical records that include demographics, blood tests, and the medical and diagnosis history of 7889 hospitalized patients diagnosed with BSI. Among the implications of this work is implementation of the models as a basis for selective rapid microbiological identification, toward earlier administration of appropriate antibiotic therapy. Additionally, our models may help reduce the development of BSI and its associated adverse health outcomes and complications.

Marginal Contribution Feature Importance - an Axiomatic Approach for Explaining Data.

In recent years, methods were proposed for assigning feature importance scores to measure the contribution of individual features. While in some cases the goal is to understand a specific model, in many cases the goal is to understand the contribution of certain properties (features) to a real-world phenomenon. Thus, a distinction has been made between feature importance scores that explain a model and scores that explain the data. When explaining the data, machine learning models are used as proxies in settings where conducting many real-world experiments is expensive or prohibited. While existing feature importance scores show great success in explaining models, we demonstrate their limitations when explaining the data, especially in the presence of correlations between features. Therefore, we develop a set of axioms to capture properties expected from a feature importance score when explaining data and prove that there exists only one score that satisfies all of them, the Marginal Contribution Feature Importance (MCI). We analyze the theoretical properties of this score function and demonstrate its merits empirically.

Antibiotic Use during the COVID-19 Pandemic in a Tertiary Hospital with an Ongoing Antibiotic Stewardship Program.

During the recent pandemic, the fact that the clinical manifestation of COVID-19 may be indistinguishable from bacterial infection, as well as concerns of bacterial co-infection, have been associated with an increased use of antibiotics. The objective of this study was to assess the effect of targeted antibiotic stewardship programs (ASP) on the use of antibiotics in designated COVID-19 departments and to compare it to the antibiotic use in the equivalent departments in the same periods of 2018 and 2019. Antibiotic consumption was assessed as days of treatment (DOT) per 1000 patient days (PDs). The COVID-19 pandemic was divided into three periods (waves) according to the pandemic dynamics. The proportion of patients who received at least one antibiotic was significantly lower in COVID-19 departments compared to equivalent departments in 2018 and 2019 (Wave 2: 30.2% vs. 45.6% and 44.9%, respectively; Wave 3: 30.5% vs. 47.8% and 50.1%, respectively, p < 0.001). The DOT/1000PDs in every COVID-19 wave was lower than during similar periods in 2018 and 2019 (179-282 DOT/1000PDs vs. 452-470 DOT/1000PDs vs. 426-479 DOT/1000PDs, respectively). Moreover, antibiotic consumption decreased over time during the pandemic. In conclusion, a strong ASP is effective in restricting antibiotic consumption, particularly for COVID-19 which is a viral disease that may mimic bacterial sepsis but has a low rate of concurrent bacterial infection.

The impact of tumor detection method on genomic and clinical risk and chemotherapy recommendation in early hormone receptor positive breast cancer.

BACKGROUND: Symptomatic breast cancers share aggressive clinico-pathological characteristics compared to screen-detected breast cancers. We assessed the association between the method of cancer detection and genomic and clinical risk, and its effect on adjuvant chemotherapy recommendations. PATIENTS AND METHODS: Patients with early hormone receptor positive (HR+) HER2neu-negative (HER2-) breast cancer, and known OncotypeDX Breast Recurrence Score test were included. A natural language processing (NLP) algorithm was used to identify the method of cancer detection. The clinical and genomic risks of symptomatic and screen-detected tumors were compared. RESULTS: The NLP algorithm identified the method of detection of 401 patients, with 216 (54%) diagnosed by routine screening, and the remainder secondary to symptoms. The distribution of OncotypeDX recurrence score (RS) varied between the groups. In the symptomatic group there were lower proportions of low RS (13% vs 23%) and higher proportions of high RS (24% vs. 13%) compared to the screen-detected group. Symptomatic tumors were significantly more likely to have a high clinical risk (59% vs 40%). Based on genomic and clinical risk and current guidelines, we found that women aged 50 and under, with a symptomatic cancer, had an increased probability of receiving adjuvant chemotherapy recommendation compared to women with screen-detected cancers (60% vs. 37%). CONCLUSIONS: We demonstrated an association between the method of cancer detection and both genomic and clinical risk. Symptomatic breast cancer, especially in young women, remains a poor prognostic factor that should be taken into account when evaluating patient prognosis and determining adjuvant treatment plans.

The superiority of 72 h leukocyte descent over CRP for mortality prediction in patients with sepsis.

BACKGROUND: Detection of an eventful course in the early days of sepsis treatment is clinically relevant. The white blood cell count (WBCC) and C-reactive protein (CRP) are used in daily practice to monitor the intensity of the inflammatory response associated with sepsis. It is not entirely clear which of the two might better discriminate the outcomes of patients with sepsis. METHODS: 30-day mortality was assessed in a cohort of patients who were hospitalized with sepsis in the departments of Internal Medicine in a tertiary medical center. Admission and 72-hour time points were analyzed to discriminate between patients with increased versus decreased 30 days mortality risk. RESULTS: The study included 195 patients. Higher 72 h CRP, WBCC, neutrophil counts and neutrophils to lymphocyte ratio were associated with increased mortality (p < 0.02). Baseline WBCC and CRP failed to discriminate between patients who died and those who survived (AUC = 0.551, 0.479). In multivariate analysis of the 72 h tests, higher WBCC count (OR = 1.12, 95%CI 1.05-1.20, p = 0.001), was associated with increased mortality whereas CRP was not (OR = 1.004, 95%CI 0.998-1.01, p = 0.146). CONCLUSION: Patients who presented a 72-hour leukocyte descent had a better outcome and in this regard, WBCC was superior to 72-hour CRP in predicting 30 days mortality.