Both HEDIS and Stars quality measures are used to determine the quality of care provided by health plans and providers. Higher performance on these measures can lead to higher payments and bonuses, influencing the financial stability and reputation of health plans and providers. By using advanced analytics payers and providers can predict low quality measures and target them and affected patients can administer care that will both improve patient outcomes AND get those health plans and providers paid more.
Due to these direct incentives for improved quality, we are seeing GenHealth.ai users interact with our G-Mode Generative AI Analytics application to predict their performance on these quality measures in the future.
Background
HEDIS (Healthcare Effectiveness Data and Information Set) provides standardized performance measures to help purchasers and consumers compare health plans, covering issues like preventive care, treatment, and outcomes. The Stars rating system, developed by CMS, rates Medicare Advantage and Part D plans from 1 to 5 stars based on performance measures like care effectiveness, patient experience, and plan administration.
Some specific HEDIS measures include:
- Breast Cancer Screening
- Comprehensive Diabetes Care
- Childhood Immunization Status
- Colorectal Cancer Screening
Some specific Stars quality measures include:
- Annual Flu Vaccine
- Monitoring Physical Activity
- Medication Reconciliation Post-Discharge
- Improving or Maintaining Physical Health
Enter G-Mode
The GenHealth.ai team recently launched G-Mode which allows you to chat with all of your past, present, AND FUTURE data. This works by running our large medical model against a population to predict the futures of all the patients in the population.
One of the first use cases we noticed from early users were queries around HEDIS and Stars Measures. For example one of the HEDIS measures published by NCQA is for screening of osteoporosis in women. Specifically, "Osteoporosis Screening in Older Women assesses the percentage of women 65–75 years of age who receive osteoporosis screening.”
With G-Mode a user can simply ask “whats the percent of women age 65-75 who may receive osteoporosis screening in 2024” and get the results in seconds to identify the expected outcome this coming year. Users can further explore their data via queries like “which women 65-75 are least likely to receive the screening and are most likely to be at risk for osteoporosis” and get a resulting list with the patient ids.
The query above took about 10 seconds to run. It did not require weeks of data analysts to create a specific model to predict any of these HCPCS codes. It did not require any temporal analytics by data science teams or any SQL writing. All it took was a natural language question and a few seconds of compute. Additionally if the resulting query is not exactly what is intended or the specific rules vary by population or geography, users can ask the model to update the query with elements like exclusion criteria or other attributes that matter to them.
How this works
G-Mode is not simply regurgitating saved queries. In fact we are not storing every single HEDIS or Stars measure query, and we do not have specific models for each of these quality measures. Instead G-Mode uses two distinct generative AI models. Our own Large Medical Model (LMM), which can predict any condition, procedure, medication, etc. into every patient’s future, and traditional off the shelf Large Language Models (LLMs) which are great for SQL query generation.
By combining these models, G-Mode can dynamically generate and run queries based on user input, providing real-time insights and predictions. This innovative approach allows healthcare providers to proactively manage patient care and improve their HEDIS and Stars quality measures more efficiently.
This real-time capability not only aids in current performance tracking but also empowers healthcare providers to make data-driven decisions proactively. As a result, they can allocate resources more effectively to areas that need improvement and ensure compliance with quality measures, ultimately leading to better patient outcomes and increased revenue for healthcare organizations.
Note: GenHealth.ai is not a HEDIS or quality measures submission vendor. However our solution can be used to aid organizations in improving their scores and outcomes when subject to quality measure incentives.