Structural causal model with expert augmented knowledge to estimate the effect of oxygen therapy on mortality in the ICU.

Recent advances in causal inference techniques, more specifically, in the theory of structural causal models, provide the framework for identifying causal effects from observational data in cases where the causal graph is identifiable, i.e., the data generation mechanism can be recovered from the joint distribution. However, no such studies have been performed to demonstrate this concept with a clinical example. We present a complete framework to estimate the causal effects from observational data by augmenting expert knowledge in the model development phase and with a practical clinical application. Our clinical application entails a timely and essential research question, the effect of oxygen therapy intervention in the intensive care unit (ICU). The result of this project is helpful in a variety of disease conditions, including severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) patients in the ICU. We used data from the MIMIC-III database, a widely used health care database in the machine learning community with 58,976 admissions from an ICU in Boston, MA, to estimate the oxygen therapy effect on morality. We also identified the model's covariate-specific effect on oxygen therapy for more personalized intervention.

Bayesian estimation of true prevalence, sensitivity and specificity of indirect ELISA, Rose Bengal Test and Slow Agglutination Test for the diagnosis of brucellosis in sheep and goats in Bangladesh.

The true prevalence of brucellosis and diagnostic test characteristics of three conditionally dependent serological tests were estimated using the Bayesian approach in goats and sheep populations of Bangladesh. Serum samples from a random selection of 636 goats and 1044 sheep were tested in parallel by indirect ELISA (iELISA), Rose Bengal Test (RBT) and Slow Agglutination Test (SAT). The true prevalence of brucellosis in goats and sheep were estimated as 1% (95% credibility interval (CrI): 0.7-1.8) and 1.2% (95% CrI: 0.6-2.2) respectively. The sensitivity of iELISA was 92.9% in goats and 92.0% in sheep with corresponding specificities of 96.5% and 99.5% respectively. The sensitivity and specificity estimates of RBT were 80.2% and 99.6% in goats and 82.8% and 98.3% in sheep. The sensitivity and specificity of SAT were 57.1% and 99.3% in goats and 72.0% and 98.6% in sheep. In this study, three conditionally dependent serological tests for the diagnosis of small ruminant brucellosis in Bangladesh were validated. Considerable conditional dependence between IELISA and RBT and between RBT and SAT was observed among sheep. The influence of the priors on the model fit and estimated parameter values was checked using sensitivity analysis. In multiple test validation, conditional dependence should not be ignored when the tests are in fact conditionally dependent.