NCQA ECDS

For those who want the three original NCQA reports you can find them here:

Before we get started, it’s worth clarifying a few things that can be confusing:

1. ECDS is a reporting system. ECDS enables the use of clinical data for quality measurement. Permissible ECDS data sources include claims, HIE/Registry, case management and EHR data.
2. ECDS measures can be contrasted with traditional measures which combine claims data and hybrid medical record review.
3. The number of ECDS measures has been increasing over the past four years. These measures are often released initially in parallel with the traditional measure, and over time the measure become “ECDS-only”, meaning that only ECDS reporting is available. With time, we can expect all measures will transition to “ECDS-only”.
4. Just because ECDS reporting is available, doesn’t mean that health plans are actually reporting with substantial non-claims data. They could still be reporting ECDS but primarily with claims data. This will be especially true for those plans who are not putting substantial effort into obtaining clinical data.
5. Currently, non-claims data sources arrive in a wide variety of formats, ranging from DAV-certified CCDs (a standard source) to unstructured data that must be abstracted across a population (a non-standard source). The challenge of having so many different styles of data is that someone must ultimately convert all of this data into a common format that can be ingested by a HEDIS engine. This is costly and time-intensive.
6. Digital measurement is a powerful new technology that you can use when you have collected clinical data (such as for ECDS) and you are representing it using a standard such as FHIR. Among other benefits, digital measurement can greatly reduce the effort required to aggregate data from so many different sources. One way to think about ECDS is that it is a first step towards real digital measurement.

With all of that in mind, where are we today? One thing that these publications tell us is that we are making some progress in moving to ECDS reporting. Let’s start by looking at the percentage of plans reporting ECDS from 2021-2024. I’ve picked 6 measures that I think are fairly representative – 2 each from immunization, cancer screening and behavioral health for which we have data in since 2021 and then linked them up so that you can see a little longer time horizon.

This data tells us that health plans are responding to the NCQA change in reporting requirements. This should not be too surprising because once these measures become “ECDS-only”, there really is no alternative. But remember (see point 5 above) that just because the plan is reporting using the ECDS method, it doesn’t necessarily mean that they are using substantial clinical data yet.

The 2022 publication also provided some excellent details on performance rates for 11 ECDS versus traditional measures, and for data sources used. As you would expect, plans that use clinical data for reporting see higher rates, often significantly higher, on many measures. For example, PRS-E Tdap indicator was 46.4% for plans that used only claims data versus 53.7% for plans that used any non-claims data source (2022 results, Figure 2). But this trend was not consistently observed. For example, among the cancer screening measures, rates for BCS-E were similar to traditional reporting. And at least for COL-E, traditional reporting rates were higher than ECDS rates by 6.2 percentage points for commercial and 9.3 percentage points for Medicare. Obviously, as of 2022 substantial compliance is still the result of hybrid lift for COL. More hopeful is the data on source, which suggests that over time we are seeing more and more EHR data being used, for example in calculating BCS-E. It is also apparent that ECDS will be particularly helpful in achieving more complete data for behavioral health measures, because claims data has been poor in capturing compliance for these measures. Finally, ECDS can provide better longitudinal data (especially important on measures like COL), which is another shortcoming of claims-based measurement.

Fast forward two years, and let’s dig into the results that NCQA published this month to see if these trends continue as the number of ECDS measures has grown to 16. Taking the PRS-E Tdap indicator as our example again, the PRS-E Tdap indicator was 50.3 % for plans that used only claims data versus 55% for plans that used any non-claims data source (2024 results, Figure 1). More good news is that the percentage of submissions using any non-claims data continues to increase, with almost all measures (commercial, Medicaid and Medicare) now showing >90% of submissions that include some non-claims data. But the news is not all good. The 2024 publication reports that hybrid rates were higher than ECDS rates by an average of 5.2 % for commercial and 6.3% for Medicare plans in MY23. Apparently the MRR lift persists, and now that COL is no longer hybrid we know that this is resulting in missed compliance for many health plans. A similar pattern was observed for CCS. Meanwhile for the measures that remain in the hybrid set (such as LCS and CBP) substantial MRR lift is shown (Figures 7 and 8).

One thing that is missing from the 2024 publication is a detailed table for data sources (EHR. Case Management, HIE/Registry, Claims) that we saw in the previous publication (2022 results, Table 4). This data would be especially helpful to understand how we are evolving in terms of access to data.

What can we learn from all of this data?

I think the big lesson from all of this is that health plans are trying hard to move their quality reporting to use clinical data. And many cases they are seeing more complete, more accurate and higher rates when they use clinical data. However, it’s also clear that there remains a significant data gap as we anticipate the retirement of the hybrid methodology in 2029. And it may be helpful to review the challenges in the 2021 Issues brief (above). I think you will find that these same challenges persist.

What to do? I’ll offer three action items on the data front that may be helpful:

1. While it would be great if all sources were DAV-certified SSD, the volume of data coming through DAV remains exceedingly small. Don’t be too choosey. Go broad and leverage both standard and non-standard supplemental data sources.
2. Excellent health plans are also bridging the data gap by using unstructured data (for example as PDF charts or CCDAs) with population-scale prospective MRR, often with natural language processing (NLP). This can be particularly valuable right now and before we see increased data availability through the interoperability initiatives.
3. Your provider network is your greatest asset. Can you leverage existing relationships to develop direct data feeds? This may be your most economical option at the moment.

What comes next?

One of the most challenging parts of digital transformation is that as soon as you think you’ve mastered one step forward, you find yourself up against another. The next step in the digital journey is already starting with use of the new CQL-based digital measures. And unfortunately, you won’t have the luxury of waiting until you’ve sorted out all of your data issues before you’ll need to take the leap. Instead, consider the next phase to be iterative and align your efforts so that you make progress calculating digital measures, even as you grow your data strategy.

Thank you to NCQA for making this data available. The more transparency we have about this transition, the easier it will be for all health plans to plan their next move.

We hope you’ll consider Astrata technology for NLP-powered prospective HEDIS and native FHIR and CQL digital measurement as you make your plans for 2025 and beyond.