Non-obvious correlations to disease management unraveled by Bayesian artificial intelligence analyses of CMS data

Non-obvious correlations to disease management unraveled by Bayesian artificial intelligence analyses of CMS data

Authors: Vijetha Vemulapalli, Jiaqi Qu, Jeonifer M. Garren, Leonardo O. Rodrigues, Michael A. Kiebish, Rangaprasad Sarangarajan, Niven R. Narain, Viatcheslav R. Akmaev

Artificial Intelligence in Medicine 2016, Vemulapalli et al. – bAIcis™, the Interrogative Biology® AI engine discovers a novel link between use of asthma medications and renal insufficiency

Bayesian networks are powerful mathematical frameworks for learning probable cause and effect relationships from large amounts of data.  Publicly available summarized healthcare data was analyzed with Berg’s bAIcis™, Bayesian network technology to build a cause and effect network of diagnostic codes.  From this network, subnetworks of heart failure and kidney failure were selected, as both of these health issues are associated with significant mortality in the US population.  In these subnetworks, we found known connections as shown by previous scientific studies, thus validating our investigative approach.  For example, within the heart failure subnetwork, associations were made between heart failure and chronic obstructive pulmonary disease (COPD), heart failure and cardiac arrhythmia and conduction disorders, and heart failure and large volume blood loss.  Within the kidney failure network, the known interactions of kidney failure with high blood pressure, recurrent kidney infection and fluid, metabolism, and nutrition disorders were identified.  A novel link was discovered between kidney failure and asthma/bronchitis, which lead to a novel hypothesis on whether some asthma medications can have a negative effect on kidney function. We proposed a detailed molecular mechanism linking a certain class of asthma medications with renal insufficiency. We found that strong support exists in the literature for each intermediary of the molecular hypothesis but the big picture interaction was missed by previous studies. This novel finding, if validated by further epidemiological and clinical studies may have a significant impact on clinical management of the disease and lead to improvements in patient outcomes. Our results show that analysis of large data sets using artificial intelligence and powerful mathematical tools can identify non-obvious relationships between clinical events and therefore provide patients with improved care and fewer complications.

https://www.journals.elsevier.com/artificial-intelligence-in-medicine/open-access-articles