Authors

Inspiration

Having the ability and also the obligation to inform public health officials and providers with a validated approach to risk adjusting the population for Covid-19 adverse outcomes.

What it does

It maps the universe of ICD9 CM and ICD10 CM diagnosis codes to the 90 Centers for Medicare and Medicare Services (CMS) Hierarchical Condition Categories (HCCs).  Thus, enabling PurpleLab to observe which of the anonymized patients in our claims repository have one of more CMS HCCs.  Population cohorts are subsequently categorized by levels of risk related to adverse outcomes of a Covid-19 infection, including:

(1) Low Risk = If a patient appears in the Claims Repository but does not have ICD9 or ICD10 CM codes that map to a CMS HCC AND their age is less than 60 years old, then the patient is classified as having Low Risk.

(2) Moderate Risk = If a patient appears in the Claims Repository but does not have ICD9 or ICD10 CM codes that map to a CMS HCC AND their age is greater than 60 years old, then the patient is classified as having Moderate Risk. or If a patient appears in the Claims Repository and has an ICD9 or ICD10 CM code that maps to one CMS HCC that is NOT indicative of either (i) immuno-compromise; and/or (ii) cardio-respiratory compromise. Then the patient is classified as having Moderate Risk.

(3) High Risk = If a patient appears in the Claims Repository and has an ICD9 or ICD10 CM code that maps to more than one CMS HCC that is NOT indicative of either (i) immuno-compromise; and/or (ii) cardio-respiratory compromise. Then the patient is classified as having High Risk. or If a patient appears in the Claims Repository and has an ICD9 or ICD10 CM code that maps to a single CMS HCC indicative of either: (i) immuno-compromise; and/or (ii) cardio-respiratory compromise. Then the patient is classified as High Risk.

(4) Severe Risk =  If a patient appears in the Claims Repository and has an ICD9 or ICD10 CM code that maps to a more than one CMS HCC indicative of either: (i) immuno-compromise; and/or (ii) cardio-respiratory compromise. Then, the patient is classified as having Severe Risk.

These population cohorts are visible by 886 3-digit zip code ZIP3 regions covering the U.S. population.  The numbers have been projected to 2020 local population levels taking into account age and gender cohorts.  These risk cohorts provide levels of "demand" for treatments related to Covid-19.

This risk adjustment of the population is subsequently indexed by three capacities-to-treat, including:

(1) Total Hospital Beds (2) ICU Beds (3) Physician Intensivists = Physicians with the necessary experience in managing ventilator-dependent patients (which include Critical Care Medicine, Pulmonary Medicine, Emergency Medicine, Anesthesiology and Thoracic Surgery)

We produce Risk Cohorts by 886 ZIP3 regions with ratios of three capacities-to-treat.  This analysis provides critical insights into the geographic regions with the highest at-risk populations with the least amounts of capacity.

How we built it

We built groups of codes representing the co-morbidities we were modeling in our tools ( Health Nexus ) and measured multiple aspects of risk in the US population against our warehouse of claims data ( we have 65-70% of all commercial and medicare claims in our data warehouse going back 5-9 years )   We then took these indicators of health and stratified them into 4 risk groups:

Low Medium High Severe

The Severe Risk group represents populations that have multiple co-morbidities as well as specific co-morbidities that affect the respiratory system.

We then compared these risk groups nationally broken down by 3-digit zipcode - and compared the at-risk population vs Bed, ICU Beds and Physician Intesivisits and mapped the results.

Challenges we ran into

Getting good data on the existing health infrastructure of beds and ICU beds was challenging.  We found a starting point with the AHA Hospital Directory.  We then averages of ICU beds to total beds for null values absed on mean value for hospital classes.

Accomplishments that we're proud of

We developed what we believe is the only validated way to risk adjust the population for Covid-19 disease.  We hope that we can disseminate the findings and ultimately save lives.

What we learned

This is a very transparent demonstration of a timely and important use of Real World Data that was fully anonymized using the Datavant software.  We also demonstrated that combining real world data with our dynamic HealthNexus medical terminology master data management platform gave PurpleLab the advantage of accuracy and speed.

What's next for Risk Adjusting the US Population for Covid-19

Run scenarios for surging Capacities-to-Treat.  For example, what are the hotspots in the U.S. that remain if Total Hospital Beds are increased by 50%, ICU Beds are increased (via ventilator sharing) by 100% (2 patients per vent) or 300% (4 patients per vent) or Physician Intensivists (bringing back retired physicians) by 15%.

We are presenting our results to public health officials in our home state to help direct resources to areas that will be severely impacted based on the combination of populations with skewed risk of adverse events relative to existing medical capacity for treatment.

Try It out

Hackathons

Technologies

google-bigquery, mapbox, particle, spotfire

Devpost Software Identifier

256587