Data mining of common laboratory tests can be used to identify patients at risk

Stroke remains one of the leading causes of death worldwide. It is usually associated with a build-up of fatty deposits inside the arteries which increases the risk of blood clotting. The unannounced nature of the disease when it strikes has posed a major challenge in the health sector. Poor medical facilities, insufficient information on how to accurately diagnose stroke, late identification of the disease by the patients due to being ignorant of the disease are some of the reasons for the increasing mortality rate due it. The application of data mining technique in the field of medicine has brought about positive development in the area of diagnosing, prediction and deeply understanding of healthcare data. This study considers some of the Predictive Models developed using some data mining approaches to predict patients at risk of developing stroke in order for other researchers to build on.

Download Full-text

Predicting Poor Outcome of COVID-19 Patients on the Day of Admission with the COVID-19 Score

Critical Care Research and Practice ◽

10.1155/2021/5585291 ◽

2021 ◽

Vol 2021 ◽

pp. 1-8

Author(s):

Luke Tseng ◽

Erin Hittesdorf ◽

Mitchell F. Berman ◽

Desmond A. Jordan ◽

Nina Yoh ◽

...

Keyword(s):

At Risk ◽

Primary Endpoint ◽

Laboratory Tests ◽

Medical Center ◽

Multiorgan Failure ◽

Academic Medical Center ◽

Univariate Analysis ◽

Binary Logistic Regression Model ◽

Poor Outcome ◽

Patients At Risk

Background. COVID-19 may result in multiorgan failure and death. Early detection of patients at risk may allow triage and more intense monitoring. The aim of this study was to develop a simple, objective admission score, based on laboratory tests, that identifies patients who are likely going to deteriorate. Methods. This is a retrospective cohort study of all COVID-19 patients admitted to a tertiary academic medical center in New York City during the COVID-19 crisis in spring 2020. The primary combined endpoint included intubation, stage 3 acute kidney injury (AKI), or death. Laboratory tests available on admission in at least 70% of patients (and age) were included for univariate analysis. Tests that were statistically or clinically significant were then included in a multivariate binary logistic regression model using stepwise exclusion. 70% of all patients were used to train the model, and 30% were used as an internal validation cohort. The aim of this study was to develop and validate a model for COVID-19 severity based on biomarkers. Results. Out of 2545 patients, 833 (32.7%) experienced the primary endpoint. 53 laboratory tests were analyzed, and of these, 47 tests (and age) were significantly different between patients with and without the endpoint. The final multivariate model included age, albumin, creatinine, C-reactive protein, and lactate dehydrogenase. The area under the ROC curve was 0.850 (CI [95%]: 0.813, 0.889), with a sensitivity of 0.800 and specificity of 0.761. The probability of experiencing the primary endpoint can be calculated as p = e − 2.4475 + 0.02492 age − 0.6503 albumin + 0.81926 creat + 0.00388 CRP + 0.00143 LDH / 1 + e − 2.4475 + 0.02492 age − 0.6503 albumin + 0.81926 creat + 0.00388 CRP + 0.00143 LDH . Conclusions. Our study demonstrated that poor outcome in COVID-19 patients can be predicted with good sensitivity and specificity using a few laboratory tests. This is useful for identifying patients at risk during admission.

Download Full-text