Analyst: Valerie Garrison, MES, Evaluation Analyst, CIVHC

Project Name: Telehealth vs. In-Person Services Payment Parity in Colorado, 2020-2021: Analysis of Telehealth Denied Claims

Project Purpose (in one sentence, what high-level question was this analysis trying to answer?):

Colorado’s Office of eHealth Innovation (OeHI) wanted to know whether claims for telehealth services are approved at a comparable rate to in-person services. They also wanted to have a better understanding of any challenges with claim submission or reasons for denial.

What were your first steps when beginning this analysis?

The Colorado All Payer Claims Database (CO APCD) does not contain any information for claims that were fully denied. OeHI coordinated with the CO Division of Insurance (DOI) on a one-time special request for all claims, including those that were denied, from five insurers.

Were there specific considerations you needed to consider based on the data?

First, because we were working with external data, we had to spend extra time combining the responses from all five insurers into a clean data set. Each payer used slightly different methods for reporting, and we had to make some assumptions about this unfamiliar data. We had asked a few clarifying questions as part of the request, and the responses were very helpful for interpreting each payer’s unique submission.

Another consideration was to be cautious with our conclusions. The CO APCD contains fully adjudicated claims, which means that any contested claims were evaluated and a final decision for approval or denial was made. The data for this analysis contained all claims, including ones that did not have a final decision. This helped us better understand the submission and approval process for both telehealth and in-person claims but prevented us from making conclusions about the final denied claims rate. In other words, we found that submitted telehealth claims were more likely to be denied than in-person claims. However, we were unable to determine whether this finding still applies after dispute resolution and final decision-making.

What challenges did you encounter while performing the analysis? How did you overcome them?

We had several challenges throughout this project that included; the data request and transfer process, determining the key question of interest, and data cleaning. However, our biggest challenge was that we did not have a claim status field in the final data request. As a result, we had to determine whether each claim was denied or approved. First, we determined that any claim with a “Reason for Denial” was denied. For all others, we studied the cost data and grouped claims into unique cost scenarios and determined for each one whether the data indicated an approval or a denial. Insurer responses to clarifying questions helped us interpret some of the more complicated scenarios. We reviewed the final determinations with CIVHC’s Data Research & Client Solutions Analyst to confirm that these were reasonable conclusions.

Without delving into results, did anything surprise you about this analysis or the process of executing it?

A few things surprised me in this analysis. For example, this was my first time working with data submitted directly from payers. While I expected we would need to standardize the data we received from payers, I was surprised by just how different some of the reporting is in terms of what might be included in a certain field and completion rate.

I was also surprised – and perhaps excited – by the results of the project. This small analysis including only five CO insurers supported the provider survey results and highlighted potential directions for more investigation or targeted improvements.

What did you learn while performing this analysis?

I was reminded throughout this project that clear and concise communication is essential for evaluation projects. There are many approaches one could take and nearly infinite questions to answer, so it is critical to clearly define 1-2 evaluation questions, remain focused, and carefully select methods that align with your questions.

This analysis was fulfilled by Evaluation Analysts Valerie Garrison (methodology) and Rachel Jardim (statistical analysis)