HealthProSight

Background & Motivation

Within the evolving landscape of value-based healthcare payment models, the role of Hierarchical Condition Categories (HCC) and the Medicare Risk Adjustment model is becoming paramount. This project seeks to develop a unified solution that not only automates HCC coding but also accurately predicts Medicare Risk Adjustment scores. By doing so, we aim to bring clarity, efficiency, and scalability to the healthcare cost estimation process. The target users are primarily for healthcare service providers.

Risk Adjustment is a process for quantifying an individual's health (or sickness) into a Risk Adjustment Factor or risk score. Risk scores are calculated using demographic factors (gender, age, disability status, etc.) as well as medical status and history (specifically chronic illnesses such as cancer, diabetes, heart failure, etc). A numeric value is assigned to these various factors using a risk model. Various risk models exist (e.g., Rx and ESRD) to address the needs of various payment systems. A risk score can be used in many applications, such as calculating capitated payment, or normalizing hospital performance scores by accounting for the general sickness of the population they treated.

Hierarchical Condition Categories (HCCs) are an implementation of Risk Adjustment and are used to capture medical status and history in many risk models (including the current risk models used by CMS and ACA requirements). In HCC methodology, certain diagnoses (i.e., ICD-10-CM codes) are assigned an HCC according to the nature and severity of the diagnosis. These HCCs in turn are also assigned a risk factor. A patient's risk score is generated by adding together the demographic risk factors with the risk factors for the various HCCs they qualify for (with hierarchies preventing multiple diagnoses in the same disease group from inappropriately increasing the risk score).

Problem Statement

Currently, healthcare providers have to manually map vast numbers of ICD codes to their corresponding HCC categories, a procedure that is both tedious and prone to inaccuracies. Once mapped, determining a patient's Risk Adjustment Factor (RAF) score, which is pivotal in predicting future healthcare costs, is another elaborate task. This manual, multi-step process often leads to inefficiencies, inaccuracies, and potential revenue leakage for healthcare providers.

Our Solutions

Our application HealthProSight streamlines healthcare processes through a single application, powered by two innovative models for health Providers and individual Patients. We’re using the BERT（NLP classification model) to automatically map the disease diagnosis to HCC code and use XG Boost to predict patients’ risk score and premium. For providers, our product can accelerate monthly premium calculations with efficiency and accuracy in their operations. For patients, our product can let them get timely access to preventive care opportunities, which can enhance their overall well-being.