Clarity Healthcare Intelligence: population health management based on artificial intelligence - optimization in risk prediction and management of coronary artery disease and post-operative results

Abstract

Faced with limited resource and the brutal dissipation of health-related expenditures (> 20 per cent of the volume of resources), population health management methods have become worldwide popular and are at the forefront of sustainable health care. Using predictive models for future costs based on clinical data, it is possible to identify the highest risk individuals and trigger intensive measures of prevention and health promotion, with the aim of reducing the chances of hospitalization, reducing costs and increasing longevity. More than 20 companies in the world have actions dedicated to the management of population health (Population Health Management [PHM]); however, its presence is very tenuous in Brazil. Despite the variety of predictive models of individual risks and future costs, the established models have low success rates (not higher than 78-82%). That is, ~ 20% of patients are inappropriately engaged in more intensive approaches or are not identified for such approaches, which reduces the efficiency of PHM actions. For preventive health strategies to produce cost-effective results, it is essential that they be sustained under three pillars: (i) segmentation / stratification; (ii) coordination of care; and (iii) analytical intelligence. Failures in any of the 3 pillars generate inefficiency and waste. Previous data from our group suggest that machine learning techniques not only stratify subpopulations of higher risk of becoming ill, but also point who are the patients with a high load of modifiable and uncontrolled risk factors. In order to anticipate the future costs and care needs of the most vulnerable individuals, we propose to study health determinants in two high-cost contexts: rehospitalization, dialysis and death in patients with (i) coronary artery disease (CAD); and in (ii) pre-operative care (PrOp). In patients with CAD we will analyze 4 Brazilian databases and for PrOp we will have access to a large national database. Using state-of-the-art tools in predictive statistics: (i) we will select predictors by supervised learning techniques; (ii) we will construct prediction models for risk of hospitalization, renal failure and death, (iii) we will establish predictive models for future costs. With this, we will produce the computer system prototype (Clarity HI Soft) that: (a) estimates risks in health in real time; (b) estimate future costs; (c) include visual analytics platforms for patients and managers. This system will be configured for use in other databases, such as community data from the Unified Health System (SUS) and data from private health plans. The proposed analyzes will contribute to a more precise diagnosis of population health, allowing to identify the individuals of greater health risk and those with greater chance of implying in increased costs in the future years. These estimates are the key link to delivering Population Health Management (PHM) services, Clarity Healthcare Intelligence's niche market. We also expect to produce a patent for Clarity HI Soft, establish governance strategies around the use of analytical data and results, as well as operationalize analyzes and incorporate lessons from the real world. Impacts on the company's business and the market. Innovative Research will enable Clarity Healthcare Intelligence to introduce a PHM model with high predictive accuracy, raising utility data on public health and chronic disease epidemiology. Optimizing the data analysis process will make the company highly competitive in the PHM market. The interaction with Brazilian data (an unprecedented fact) will put the company ahead of competitors in negotiations with health plans and SUS managers. (AU)