The uncertainty and risk created by pandemic-related disruptions in care have health plans in uncharted water when it comes to relying on claims data and utilization patterns to inform their risk models. But while these disruptions weakened the reliability of traditional risk models, they also created opportunities to deepen and enrich member data by leveraging consumer and other data related to social determinants of health (SDoH) that can deliver significant performance and accuracy benefits.
Expanding data sources and enhancing analytics capabilities allows health plans to develop the capacity for more precise, holistic member engagement while improving the most impactful health outcomes in individuals and populations. Doing so will help lift Star Ratings, improve care compliance, close gaps, enhance member satisfaction, reduce costs and drive profitability gains.
Pandemic-Era Compliance Levels
When elective procedures were paused and care was delayed or diverted in early 2020 in response to the initial COVID-19 surge, it had an immediate impact on utilization and ancillary activities like testing and imaging, diagnoses, treatment plans, etc. By late summer, most provider organizations had reopened for procedures and care under strict social distancing measures. This subsequently generated sufficient claims data for Carrot Health to analyze and estimate the impact of disruption.
We opted to focus on compliance levels among Medicare Advantage populations for specific procedures, tests, and care interventions. This is because risk models—following Centers for Medicare and Medicaid Services (CMS) guidelines—emphasize appropriate care and member experience to meet quality standards central to Star Ratings. In that sense, key compliance levels serve as proxies that track changes in care volume.
We found that, while the overall volume of care showed signs of returning to normal levels by the end of 2020, the erratic cadence of care delivery througout the year would significantly impact compliance rates, engagement levels, and risk models going forward.
In the first two quarters of 2020, all plans experienced dips in compliance. Our data shows annual wellness visits dropped by 60%, and hemoglobin A1c testing dropped by 40% when compared with the previous year.
The drop was so dramatic that leading plans took urgent steps to improve compliance and care delivery, as reflected in the surge in telehealth earlier in the year. And while that surge receded over the course of the year, telehealth has shown staying power which may impact how care is delivered for the foreseeable future.
To determine the degree to which care utilization bounced back in the fourth quarter, we examined Medicare/Medicare Advantage data from October 2019 to October 2020. That data was collected from health plan customers across a number of states and serves as a nationally representative sample.
It showed that even with the third- and fourth-quarter recovery, 2020 compliance rates fell short across key quality measures as compared to 2019. Most notably, colorectal screening fell by 8% and annual wellness visits by 5%. Medication adherence across diabetes, renin-angiotensin system (RAS), and Statins either rose or were unaffected.
It’s possible to make inferences based on these numbers and the events of 2020. The decline in primary care visits was not as severe as other measures because primary care includes any consultation with a clinician, whether in-person or via telehealth, through hospitals, clinics or urgent care centers.
On the other hand, wellness visits fell significantly despite regulatory changes implemented by CMS allowing for these examinations to be conducted via phone. The persistence of the decline could be due to the perceived lack of urgency, given other concerns. It’s also possible providers were not sufficiently educated on the policy change to adjust their care practices accordingly.
Likewise, breast cancer and colorectal screenings fell, possibly because they require an in-person diagnosis. Some beneficiaries likely considered the examination to be worth delaying because of the risk of contracting COVID-19.
It is also likely that medication adherence stayed flat or rose because prescriptions can be automatically refilled without clinician intervention, and can be received by mail or through drive-through pharmacy services. Moreover, many of the illnesses for which those medications are prescribed may count as comorbidities with COVID, giving patients strong motivation to adhere to their care plan.
Regardless of the reasons for the decrease in compliance and appropriate care, it’s likely that they will contribute to worsening health outcomes in subsequent years. Patients who did not receive annual wellness checks or important screenings may develop conditions that could have been prevented or treated earlier. Similarly, fewer visits and interactions with care providers can diminish engagement and adherence to treatment plans, medications, and healthy lifestyles and behaviors.
The Cascading Impact on Risk Modeling
To predict future care demand costs, health plans turn to utilization risk. This is determined by assessing the condition and clinical risk based largely on retrospective claims data from the previous year combined with established risk stratification tools. When claims data is incomplete or inaccurate, however, risk scores and risk stratification will also be inaccurate.
Traditional methodologies also fall short when predicting total and rising risk. This is because health plans rely on “sweeps” to update scores and adjust payments based on diagnoses that trickle in. But that data does not fully or accurately reflect changes in health status or emerging care needs over the course of the year. As a result, health plans do a poor job identifying people who are at risk of becoming high utilizers.
These challenges were exacerbated by 2020. The disruption in utilization patterns affected the capture of conditions used to drive risk adjustment, which means models relying on inaccurate or suspect data will be distorted.
Examining compliance data, we observed the following trends in the second quarter of 2020:
– Average HCC-RAF down 23%
– Diabetes diagnoses down 21%
– CHF diagnoses down 16%
– Kidney disease diagnoses down 19%
These dramatic drops track with expectations, given the pandemic’s impact on care delivery. By the fourth quarter, however, the gap in risk adjustment had narrowed significantly. Compared to October 2019, a number of key conditions were down across the Medicare Advantage population in October 2020. These included:
– Substance abuse (-0.2%)
– Congestive heart failure (-0.3%)
– Diabetes (-0.7%)
– Morbidity obesity (-0.7%)
– Kidney disorders (-0.8%)
Notably, these changes from 2020 to 2019 do not necessarily indicate actual declines in the prevalence of these conditions among MA beneficiaries. Instead, they are likely related to declines in engagement, testing, treatment, physician visits, etc., which would have some indeterminant impact on the prevalence of those conditions.
While those numbers may not seem large given the extent of the challenges the health system has had to overcome, they will have an outsized impact on health plan finances. If the prevalence of conditions turns out to be higher than the 2020 data projects, health plans could experience a significant loss in revenue from risk adjustment reimbursement.
If we also assume that members will resume a normal pattern of care once the pandemic has subsided—or receive more care than usual since they avoided it in 2020—this could add to the financial squeeze on health plans.
In short, health plans are right to be concerned about the impact of unusual utilization patterns on risk models this year. They should also be concerned about the limitations of their data under normal circumstances.
Data Options to Improve Predictions
Risk modeling tools are difficult to adjust within the workflow, but health plans have several options for leveraging better data in 2021. These include expanding internal data sources and/or leveraging non-traditional data to better understand member behaviors and propensities and more precisely predict utilization demand.
To bolster in-house data, health plans can combine available claims and clinical/condition data with other sources, such as:
– Engagement and adherence propensity
– Past treatment adherence
– Participation in care management programs
– Responsiveness to marketing outreach
A more significant lift in prediction accuracy can be achieved by tapping outside sources. For example, member data can be dramatically enhanced by including survey and consumer data. Such data is useful because it reflects actual lifestyles and the financial and demographic circumstances that influence as much as 90% of a person’s overall health status.
When combined with clinical and claims data, this consumer and survey data presents a more robust and holistic view of the health plan member. For example, the pandemic exacerbated social and economic barriers to health across specific populations. If this impact is not factored into risk projections, it will increase the likelihood that health plans under-estimate member care needs and costs.
Advancing the Data Revolution
COVID-19 has forced healthcare organizations to adjust and respond rapidly to unprecedented challenges on multiple fronts. Yet, the impact of 2020 on utilization data has been insufficiently appreciated.
Unless health plans adapt their risk models, they will be unprepared to deal with their members’ emerging health needs—and face the potential of significant financial consequences in 2021 and beyond as a result.
Yet, the problems with utilization risk models run deeper. Even under the best of circumstances, claims and clinical/condition data are not robust, holistic, or dynamic enough to give health plans the capacity to accurately predict changes in health status or rising risk.
By incorporating consumer, survey, and social determinant data into their existing risk models, health plans can develop more accurate predictions that enhance health outcomes, improve satisfaction scores, deepen member engagement and drive profitability.
About Kurt Waltenbaugh
Kurt Waltenbaugh is Co-Founder & CEO of Carrot Health, a leading provider of healthcare solutions powered by consumer data.