A Novel Clinical Decision Support System for Diagnosis and Treatment of Heart Failure: Concordance with Expert Decision

Abstract

Objectives Conundrum health care providers and institutions face is how to deliver evidence-based medical care at lower cost to an ever expanding portion of society in need. Translation of clinical studies to daily practice and adoption of national and international guidelines have not met expectations. This disappointing shortfall stems primarily from lack of interface between ever growing and changing vast medical literature and individual patient profile. Concurrently, threats of penalties for underperformance and poor reimbursement related to improper documentation are increasing. Hence, transition to comprehensive clinical decision support systems (CDSS) will not just be inevitable but also rewarding. Purpose Examine the level of correlation between the novel CDSS and health care providers' clinical decisions. Methods CDSS for diagnosis and treatment of HF (Cardiologist Evaluator) was developed using a USPO patented methodology with artificial intelligence to function as an assistant to health care provider. It serves as a search engine to select evidence-based diagnostic and therapeutic options for an individual patient. A total of 389 patients with suspected HF were available between 2017 and 2018 at a suburban tertiary care hospital. Both Cardiologist Evaluator and health care providers (mid-level practitioner or junior faculty) gave recommendations regarding diagnosis and treatment for each patient independently. Concordance rate between Cardiologist Evaluator recommendations and actual health care providers’ decisions was evaluated. Furthermore, ICD-10 coding provided by Cardiologist Evaluator was compared to ICD-10 coding of health care provider for immediate cost effectiveness evaluation. Recommendations given by Cardiologist Evaluator were later presented to board certified HF cardiologists who were then asked if they find them acceptable or unacceptable for care of each individual patient. New concordance rate, this time between Cardiologist Evaluator and HF specialist was evaluated. Results Overall concordance rate between Cardiologist Evaluator recommendations and actual healthcare providers’ decisions was 69.7%. Most discrepancies were related to adoption of new therapies, dose advancements, workup for new onset heart failure, and/or referral for LVAD. HF specialists found recommendations given by Cardiologist Evaluator acceptable 99.3% of the time. Overall concordance rate between Cardiologist Evaluator and HF specialists was 98.9%. Cardiologist evaluator provided superior, automated ICD-10 coding in order of relevance. Conclusion Cardiologist Evaluator shows high accuracy in evidence-based diagnosis and treatment of HF. It may be useful for consistent implementation of clinical guidelines and more reliable coding, especially in centers without HF specialists.