Risk Adjustment from Early Telehealth Visits

Written by Ted Kyi

June 16, 2020

Because of social distancing concerns about the spread of COVID-19, and CMS’s acceptance of telehealth visits as risk adjusting encounters, most plans have started doing prospective telehealth visits.  With a couple months of experience completing prospective visits virtually, it’s natural to ask how the risk adjustment impact from telehealth visits compares to in-home visits.  I’ve been monitoring this situation, and wanted to share some early data.

First, let me say a word about some challenges to comparing diagnosing from different time periods.  The main challenge is that there is strong seasonality in the complexity of patients seen, and there can be significant variation from plan to plan or from year to year.  Below is a chart showing a high level snapshot of the seasonality in risk scores for a few sample plans:

Chart showing large month by month variability in patient complexity for four MA plans, which combined averaged over 15,000 visits per month in 2019

The overall increasing or decreasing trends and the sharp spikes both make it difficult to get a reasonable apples-to-apples comparison between any two time periods.  A simple solution to this problem is to compare charts for the same patient from this year and from last year.

A second challenge to measuring revenue management impact is that most plans use some form of net impact calculation, measuring the RAF impact from diagnoses that are unique to the prospective assessment, and not counting any conditions that come across claims from any other provider in that same calendar year, before or after the assessment.  Given that we are looking to analyze telehealth assessments from April and May, we will have only a tiny fraction of the claims run out needed to accurately measure net impact.  To address this challenge, we will use the All Lift (AL) risk score, which includes every diagnosis from the prospective assessment.  I have found that the ratio between the net impact and the AL risk score tends to stay relatively constant over time.  This means that if the AL score goes up then the net impact usually also goes up, and if the AL score goes down then the net impact usually also goes down.  By checking to see if there is any change in the AL risk score from last year to this year on same patient reassessments, we get a pretty good leading indicator whether there will be any change to the net revenue management impact.

With those preliminaries out of the way, let’s take a look at the results.  My analytics team has performed multiple reassessment analyses, and here are the results of the latest one.

Upon reviewing the data, it was immediately clear that risk scores are up overall in 2020.  I know that my organization implemented clinical guideline changes in January and early February, so we devised a way to control for that.  We started by looking to see how much telehealth diagnosis rates and risk scores from April and May reassessments changed from the prior year assessment.  Then we did the same thing for late February and early March in-home reassessments.  To control for the impact of the clinical guideline changes, we planned to subtract the increases we saw from the in-home reassessments from the increases we saw from the telehealth reassessments.  Subtracting should adjust for any underling increases in 2020 that weren’t due to telehealth.  When we pulled the data for Feb/Mar in-home and Apr/May telehealth reassessments, what we saw was they had identical percentage increases in both average AL risk scores and number of HCCs diagnosed per assessment.  Telehealth was right on parity with in-home, with both increasing over 2019 by the same amount.

In terms of individual HCCs, the top ten HCCs diagnosed were the also the same for both in-home and telehealth, although in a slightly different order:

Home Telehealth
1 HCC18 – Diabetes with Chronic Complications HCC18 – Diabetes with Chronic Complications
2 HCC108 – Vascular Disease HCC59 – Major Depressive, Bipolar, and Paranoid Disorders
3 HCC22 – Morbid Obesity HCC22 – Morbid Obesity
4 HCC111 – Chronic Obstructive Pulmonary Disease HCC19 – Diabetes without Complication
5 HCC96 – Specified Heart Arrhythmias HCC111 – Chronic Obstructive Pulmonary Disease
6 HCC59 – Major Depressive, Bipolar, and Paranoid Disorders HCC96 – Specified Heart Arrhythmias
7 HCC85 – Congestive Heart Failure HCC108 – Vascular Disease
8 HCC19 – Diabetes without Complication HCC85 – Congestive Heart Failure
9 HCC88 – Angina Pectoris HCC88 – Angina Pectoris
10 HCC40 – Rheumatoid Arthritis and Inflammatory Connective Tissue Disease HCC40 – Rheumatoid Arthritis and Inflammatory Connective Tissue Disease

Most commonly diagnosed HCCs in 2020 comprehensive health assessments

In the telehealth timeframe, HCC 59 (major depression) and HCC 19 (uncomplicated diabetes) have moved up, and HCC 108 (vascular disease) has moved down.  Note that while existing vascular disease conditions were largely being re-diagnosed in telehealth visits, no PAD diagnostic tests were being performed, eliminating that source of PAD diagnoses.

Looking at this study as well as an analysis of national payers, Medicare Advantage diagnosis rates for telehealth for major depression and uncomplicated diabetes seem to be about 1% higher than earlier 2020 in-home assessments.  On the other hand, vascular disease, morbid obesity, and congestive heart failure have decreased about 2%.

One question I’ve gotten multiple times is if there are HCCs we no longer capture in telehealth visits.  Our clinical and coding teams reviewed each of the body systems and reported that they expected decreased diagnosis rates for morbid obesity, malnutrition and chronic skin ulcers, but there were no HCCs which we diagnose in the home that would drop to zero.  This has been verified in the telehealth reassessment data.  Every single HCC category with diagnoses in 2019 also had diagnoses in 2020 telehealth assessments.

It should be noted that while the risk adjustment impact for telehealth assessments is close to that of in-home assessments, there are key limitations to the virtual encounter:

  • Most in-person diagnostic tests, such as PAD testing or spirometry, are not being performed
  • While some tests can be mailed to members, such as a FIT kit, the key metric of test completion rate remains unproven
  • Blood pressure, BMI, and other portions of the physical exam are infrequently documented
  • Environmental scans and other elements of the assessment are limited

While we are dealing with the COVID-19 pandemic, the early data shows that prospective telehealth visits are a viable alternative to prospective in home visits for overall proper documentation of each member’s disease burden.  For members who are comfortable with the virtual encounter, the risk adjustment impact is probably around 95% that of in-home assessments, and both the member and provider can safely stay in their homes.

Interesting recent articles about telehealth:

There are many articles about telehealth these days.  Here are just a couple articles I thought were worth mentioning.

Right in my backyard, in San Diego, Steven Green from Sharp Rees-Stealy commented on the need to allow audio-only telehealth visits to risk adjust.  There are several good quotes, but this is my favorite:

“The main value of having a video visit is more for the human connection and seeing their face. I agree that a video visit can help you get more of a connection to the patient while you’re talking with them,” he emphasized. “But I don’t think it contributes very much to what’s going on in their clinical care.”

You can read the full article and all of his arguments here: Audio-only Telehealth Coverage Essential During COVID-19 Outbreak.

I’m also paying attention to the Washington Post story about a Census Bureau survey about anxiety and depression, where they reported that a third of Americans are showing signs of clinical anxiety or depression.  It was nearly one half in Mississippi.  I’m interested to follow these kinds of reports, and to see if we can analyze the data from our assessments to see increases in depression diagnoses.  I invite readers to share their own experiences, or to suggest other ideas for research.

Have a question or want to learn more about Matrix's Telehealth offering?

Ted Kyi
SVP, Business Intelligence & Analytics at Matrix Medical Network

Ted Kyi is a leader in the Business Intelligence & Analytics group responsible for measurement and analysis of current and new products and services at Matrix.  Ted leads the healthcare analytics and data science teams, and is a subject matter expert on risk adjustment and government programs.  He has worked in healthcare analytics for over twenty years.  Prior to joining Matrix, Ted was president of the analytics vendor Ascender Software, and vice president of the consulting firm Infotech Systems Management.