Data Could Now Guide Population-Based Improvements in Local Eldercare

Abstract:

Medicare and Medicaid data, augmented with additional datasets, can now help communities improve eldercare by revealing local priorities and monitoring improvements. With initial partner communities, we developed and monitored dozens of performance measures, such as inappropriate medication use, the prevalence of severe pressure ulcers, and fall rates, all at a local level. Nursing home and home care assessments, together with Medicare claims, enabled measurement for cohorts with functional or cognitive disability or persons in their last year of life. These data could support efficient eldercare improvement activities in all U.S. communities.

Key Words:

data, population-based, eldercare, performance measures, disability, fall rates, pressure ulcers, Medicare, Medicaid

Most older Americans will live for many months with serious disabilities that medical and supportive care can modify but can neither prevent nor cure. Living with disabilities ties us to our geographic community. But does the community have enough affordable and disability-adapted housing? Can you get prepared food delivered? Do employers support family caregivers and is there an adequate supply of skilled caregivers for hire? Do the medical care system and long-term care facilities serve us well?

Shortcomings in these types of service and support arrangements cannot be remedied for one person or by one provider or care coordinator. Instead, remedies require concerted actions on behalf of the affected local population. This population approach, or public health perspective, tied to the locality has rarely been implemented in U.S. eldercare (that is, long-term services and supports for older adults living with disabilities).

The two of us have worked with a dozen projects anchored by local leaders who aimed to improve eldercare for geographic communities. Each community made some gains, but each encountered substantial frustrations with the lack of useful public data and tried to make do with ad hoc indicators of uncertain merit. That situation is no longer necessary. Every geopolitical area in the United States could have useful data concerning its performance regarding eldercare, if only some entity had the will and the financing to develop and sustain the analyses. The anchoring data would be Medicare and Medicaid claims and assessment data, linked to census and other existing survey data. Targeted local data collection could supplement those readily available data, when needed to monitor the performance of eldercare arrangements at baseline and with efforts at improvement.

The Medicare and Medicaid Data

The administration of Medicare and Medicaid generates data concerning beneficiaries’ medical diagnoses, procedures, and services, including assessments of functional and cognitive performance for individuals in a nursing facility (the Nursing Home Minimum Data Set, or MDS) or in a home health episode (the Outcomes Assessment Information Set, or OASIS).

In the past, these data generated aggregate tables and graphs published by the Centers for Medicare & Medicaid Services (CMS), and little more. Explicitly with the aim of improving healthcare payment and policy, CMS established the Virtual Research Data Center (VRDC) and now licenses analysts to work on the complete Medicare and Medicaid data sets. Their work must pass privacy review, and access costs licensee organizations at least $35,000 annually (CMS, 2023). This access opens remarkable new opportunities, including insights into the experiences of a geographically defined population of Medicare beneficiaries (including persons also enrolled in Medicaid).

The most complete and timely data describe the Medicare beneficiary population and include all claims incurred under the original fee-for-service Medicare system. Complete but less timely files contain nurses’ assessments from Medicare home health episodes and all-payer nursing facility stays, as well as Part D prescription drugs.

While quality and completeness issues now limit the utility of Medicare Advantage and Medicaid data, CMS recently has sought to improve data from both programs. Analysts can merge these data at the small-area level with aggregate data points from census surveys and with information supplied locally. Data providers also can link these data to neighborhood deprivation indices and other existing surveys. OASIS and MDS assessments can establish a cohort of older adults living with dependencies in activities of daily living or with cognitive failure. Claims data can contribute to cohort identification as well, for example with diagnoses of cognitive impairments or with claims-based frailty indices (Heins et al., 2023). Data drawing upon an array of sources can help to profile quality and access issues in each community.

The time has come to consider making the resulting data available to communities that develop strategies to improve eldercare for their geographic populations. Using these data, a community could see that its older adults face too much use of contraindicated drugs, or excessive pressure ulcers, or too little use of hospice services—and a host of other elements that Medicare and Medicaid document. They could test hunches that certain sub-populations experience much better or worse care than the overall eldercare population. Leaders can see what is happening, for example, to persons living with serious disabilities, persons coming to the end of life, or persons with specific demographic characteristics or living in disadvantaged neighborhoods. Communities can readily compare their metrics to other communities, including to peer communities with similar demographics (Robert Wood Johnson Foundation (RWJF) & University of Wisconsin Population Health Institute (UWPHI), 2017).

Analysts can aggregate these data to provide insights that guide community leaders and affected organizations. Our experience confirms the usefulness of these data to identify and then address problems in care delivery, social needs, and policy. Better data on eldercare could generate commitments to invest in underfunded nonmedical services and workforce development and to enhance the performance of the medical care arrangements for all local older adults living with disabilities.

Synopsis of Our Methods

In 2020–21, we had the opportunity to work with two organizations with access to CMS’s Virtual Research Data Center (CareJourney and the Institute for Accountable Care at Brandeis University). We had recruited a dozen communities where some entity was trying to improve some aspects of eldercare, and all partnering entities were hamstrung by the lack of data to guide and evaluate improvement work.

We developed geographic delineations of each community and calculated some well-known metrics among traditional Medicare beneficiaries for that population, such as rehospitalization rates, hospitalization rates, potentially avoidable use of the emergency room, and use of hospice. Additionally, we constructed these metrics by provider, to identify providers whose patients faced the greatest risks of adverse events or had the greatest rates of receiving eldercare. We incorporated census data, provider resources, and area deprivation indices. In each county, using OASIS, MDS, and claims data, we estimated the cohort that probably had serious cognitive failure or dependency in two or more activities of daily living (2+ ADL), or both. We could also describe the cohort that was in the last 12 months of life.

Using a published algorithm, we could see the rate of falls among persons ages 65 and older, by county and by quarter. A summary and national map of the resulting falls data are available publicly (Hoffman et al., 2022). We analyzed Part D prescription medications for the use of drugs commonly inappropriate for use in older adults (The 2023 American Geriatrics Society Beers Criteria Update Expert Panel, 2023).

Any analysis that could be done for any one community could be done for all communities—and indeed comparisons to the national distribution or to peer counties required this breadth of data. With the cohort analyses, we ended up with more than 400 data points for every county’s Traditional Medicare population. And for any interested entity, the analysts could home in on the providers in that county.

Examples Using These Data

County performance and adverse events can be measured using a variety of data, for example, by examining benzodiazepine and severe pressure ulcers in cohorts with varying disabilities, dual Medicare-Medicaid enrollment, and proximity to end of life. Another example is to plot the adverse event of falls over time with comparisons to the experience of other counties.

1. Benzodiazepine use among people with self-care disability. This density plot shows the distribution of all U.S. counties’ share of older adults with self-care disability who filled one or more benzodiazepine prescription during 2018. The national average is shown in a vertical blue line, and Miami-Dade’s rates are shown in red, indicating strong opportunity for improvement in that county’s benzodiazepine rates. Three subcohorts are shown. For Miami-Dade, subcohort comparisons demonstrate that the issue is worst among dual Medicaid enrollees and is not primarily driven by patterns of care in the last year of life.

Figure 1. U.S. County Distribution of Benzodiazepine Prescribing Prevalence Among Four Cohorts of Older Adults with Self-Care Disability in 2018: Miami-Dade in Focus

Adapted from author’s work, published online (Franco, 2021). The analyzed population is traditional Medicare beneficiaries ages 65 or older continuously enrolled in Parts A & B in 2018.

2. Severe pressure ulcers among people with self-care disability. Here, a density plot shows the distribution of counties’ shares of older adults with self-care disability who experienced severe pressure ulcers (stage 3, 4, or unstageable) during the year. Miami-Dade has more than twice the measured rate of ulcers as the national average. Subcohort comparisons suggest the Medicare-only (not dual Medicaid enrolled) population may be a strong focus for improvement efforts. Remarkably, one in four older adults who died in Miami in 2018 suffered with severe pressure ulcers that year.

Figure 2. U.S. County Distribution of Severe Pressure Ulcer Prevalence Among Four Cohorts of Older Adults with Self-Care Disability in 2018: Miami-Dade in Focus

Adapted from author’s work, published online (Franco, 2021). The analyzed population is traditional Medicare beneficiaries ages 65 or older continuously enrolled in Parts A & B in 2018.

3. County-level fall-related injuries among older adults. This visual provides falls rates over time, allowing monitoring of fall risk in the county and comparing with the average and low rates across U.S. counties. The general trend toward higher fall rates and the seasonal pattern are apparent, as well as the observation that this county is doing better than the national average on this measure, but not as well as 10% of counties.

Source: Traditional Medicare claims and beneficiary data analyzed by the Institute for Accountable Care, using methods described in authors’ published work (Hoffman et al., 2022). The analyzed population is Traditional Medicare beneficiaries ages 65 or older continuously enrolled in Parts A & B.

Our partnering community entities usually wanted to understand how their metrics compared to the national distribution, or to peer counties (RWJF & UWPHI, 2017), and to scan for metrics that were concerning or favorable. Some counties had low rates of pressure ulcers, or high rates of benzodiazepines; others had comparatively high emergency room rates among people with cognitive impairment, and not among people with functional impairment only. A few communities (such as Miami-Dade County) had high rates of nearly every medical misadventure—a medical care system with high use and avoidable acute-care events, high use of potentially inappropriate medications, high rates of fall-related injuries, and high costs. For these communities, the data would suggest an urgent need for reinvestment toward eldercare improvement.

Leadership teams often wanted to review provider-level data. Several teams identified which providers served a large share of community residents with functional or cognitive disabilities or who were in the last year of life. One county team identified a handful of providers whose patients ages 65 and older faced particularly high risks of adverse events and belonged to a language-isolated subcommunity, meaning equitable outreach to the subcommunity could be more effective by providing more support to and working with these providers. In all cases, the nature of our collaborators’ interest in provider-level data was collaborative and supportive. Among those interested in provider-level data were county public health departments and accountable care organizations.

The projects we supported ranged from a falls prevention program to a dyadic support program for people with dementia and their caregivers. Across every project, the data found an immediate use by informing leaders to target a specific area, population, or provider. The data also motivated financial discussions: in one case, the data bolstered an Area Agency on Aging’s value story for its post-hospital transitional services program with a local hospital by showing the post-acute care experiences among people living with disabilities. We shared the data with a variety of state and regional agencies concerned with aging, and each found insights that helped to guide targeting of topics and resources.

Limitations

This “test of concept” work had important constraints, and other factors may constrain wider data access in the future. A first constraint limited our impact: our data came to be available during the pandemic, when many of our partners faced overwhelming demands. Most could not continue their level of effort on eldercare improvement. Another constraint relates to the data itself: available data does not include direct insights into caregiver experience, family and financial impacts of eldercare, or workforce challenges. These shortcomings probably require local data collection at this point.

Except for falls, our data was limited to a year or two before the pandemic, which is now too outdated to be directly useful. One of us (Franco) intends to update these data and to encourage other analysts to develop similar resources. While creating the datasets and data dashboards proved intensive at first, the maintenance cost of using and updating the data would be small, especially when running the numbers for many communities.

We worked with Traditional Medicare beneficiaries ages 65 and older. Analysts now can work with more timely data and some data from Medicaid, Medicare Advantage, the Program of All-Inclusive Care for the Elderly (PACE), and perhaps even the Veterans Health Care System. Future work should test and likely improve on this use of MDS and OASIS assessments. Future work also should respond to the felt needs and observed utilization of these data, provided they are made broadly available.

More broadly, CMS policies on research data access could hinder some of the eldercare improvement analytics discussed here. Logistical and financial hurdles for gaining access to CMS data are considerable, and barriers that data providers face could translate to higher prices for data analytics, making those analytics less accessible to under-resourced communities. These hurdles may rise soon, as CMS plans to eliminate its lowest-cost pathways for research data access in favor of more secure but costlier pathways (CMS, 2024). Additionally, gaining access to CMS research data requires that analysts request the minimum necessary data for the minimum necessary population under health information privacy laws.

Generating nationwide but locally specific metrics on a wide range of topics in eldercare requires a large scope of CMS data files, and analysts must make a compelling case for accessing each data file. Our experience analyzing CMS research data to inform a wide variety of eldercare improvement projects suggests that these constraints are navigable, but they may limit the supply of data providers available to enable our vision.

Discussion

Good eldercare depends upon the reliable availability of adequate supply and quality of a variety of services and supports to all older adults who need them. Remarkably, the data to guide the assessment of local performance and of the effects of eldercare improvement projects could be made available in a timely manner, affordably, and repeatedly. Unlike most countries, the United States would need to develop entities capable of managing aspects of the local population’s experience of living with disabilities in advanced age. This responsibility and opportunity might build on the Area Agencies on Aging as well as other governmental or voluntary entities. But the challenge of having no useful and reliable data about the local performance of eldercare arrangements, as our partnering communities faced, is now surmountable.

This intersection of eldercare improvement and data provision will be a fertile arena for research, including how best to develop the metrics and which improvement activities work well. While some issues will be best addressed with randomized controlled trials and similar formal studies, most initiatives may well require a form of realist evaluation, with awareness of the role of the specific context in a particular community and how the proposed mechanism of improvement interacts to achieve the outcomes (Pawson & Tilley, 1997). Aggregating these experiences thoughtfully will yield wisdom as to what works in which contexts and with what likely effects.

This sort of societal intervention research is powerful and critically important but has been downplayed for the past half-century in pursuit of generalizable research findings. Of course, some immediately generalizable research will arise and will be valuable, but much of what we need is the aggregate wisdom that comes from a large body of work that arises from trying things out with good monitoring and evaluation.

Providing services to persons with substantial disabilities living in a defined geographic community is the standard in many other countries. The United States is hamstrung by a belief in competition and by anti-trust traditions. But having dozens of providers caring for patients in residences in any one community guarantees inefficiencies—minimum times on site, transportation and parking burdens, inept coordination with and development of other local services, and so on. The United States cannot enable development of local monitoring and management of the arrangements for eldercare (and disability generally) without overcoming those inhibitions and without reliable, timely, and useful data.

In our work, we have seen communities cooperate meaningfully to address and monitor shared problems. The United States could now develop the needed data, using the approaches outlined above. Next steps include generating, funding, and staffing the coalitions (which may include governmental authorities) set up to develop improvement activities guided by these data. With such an approach, we could see substantial gains in the efficiency and reliability of care for older adults living with disabilities in each geographic area.

Many more eldercare improvement projects will arise in the coming years to support people with disabilities, as needs increase alongside trends in population aging. The value and scale of eldercare improvements will depend upon setting the right aims, targeting the right bottlenecks, supporting individuals with the most need, and collaborating across traditional silos.

These data can help—by identifying new organizations and providers that share an interest in these improvements, by promoting a shared focus on crucial problems, and by fostering payment arrangements that reflect the interdependence of community-based organizations with the medical entities that have financial incentives to manage population healthcare use. Multisector collaborations can level financial imbalances between sectors by using these data to design shared-payment arrangements, like those piloted in the Collaborative Approach to Public Good Investments (CAPGI) program (Nichols et al., 2020).

Our work also establishes how the data can help shape government programs and policies. The continued monitoring of key indicators of quality and access will provide a reality check for the improvement projects undertaken by each community. The targets and improvement mechanisms may well vary by community, and local entities should bring their own expertise and data to bear on the problems. However, meaningful use of these data, anchored in Medicare and Medicaid, could be pivotal in coordinating eldercare improvements in every community.

Nils Franco is a senior analyst at ATI Advisory in Washington, DC. He may be contacted at nils@atiadvisory.com. Joanne Lynn, MD, MA, MS, is a clinical professor of Geriatrics and Palliative Care at The George Washington University, in Washington, DC. She may be contacted at DrJoanneLynn@gmail.com.

Photo credit: Shutterstock/Master1305

References

The 2023 American Geriatrics Society Beers Criteria Update Expert Panel. (2023). American Geriatrics Society 2023 updated AGS Beers Criteria for potentially inappropriate medication use in older adults. Journal of the American Geriatrics Society, 71(7), 2052–2081. https://doi.org/10.1111/jgs.18372

Centers for Medicare & Medicaid Services. (2023). CMS fee list for CCW VRDC cloud environment. United States Department of Health and Human Services. https://resdac.org/sites/datadocumentation.resdac.org/files/2023-07/CMS%20Fee%20List%20for%20CCW%20VRDC%20Cloud%20Environment.pdf

Centers for Medicare & Medicaid Services. (2024, April 15). Important research data request & access policy changes. United States Department of Health and Human Services. https://www.cms.gov/data-research/files-order/data-disclosures-and-data-use-agreements-duas/important-research-data-request-access-policy-changes-0

Franco, N. (2021, January 7). Medicare Data Reveal Actionable Quality–Cost Gaps in Care for Elders with Disability. CareJourney. https://carejourney.com/medicare-data-reveal-actionable-quality-cost-gaps-in-care-for-elders-with-disability/

Heins, S. E., Agniel, D., Mann, J., & Sorbero, M. E. (2023). Reviewing, refining, and validating claims-based algorithms of frailty and functional impairment: Final report. RAND. https://doi.org/10.7249/RRA1871-1

Hoffman, G., Franco, N., Perloff, J., Lynn, J., Okoye, S., & Min, L. (2022). Incidence of and county variation in fall injuries in US residents aged 65 years or older, 2016–2019. JAMA Network Open, 5(2), e2148007. https://doi.org/10.1001/jamanetworkopen.2021.48007

Nichols, L., Taylor, L. A., Hughes-Cromwick, P., Miller, G., Turner, A., Rhyan, C., & Hamrick, R. (2020, August 13). Collaborative approach to public goods investments (CAPGI): Lessons learned from a feasibility study. Health Affairs Forefront. https://www.healthaffairs.org/do/10.1377/forefront.20200811.667525/full/

Pawson, R., & Tilley, N. (1997). Realistic evaluation. SAGE Publications.

Robert Wood Johnson Foundation & University of Wisconsin Population Health Institute. (2017, August 30). Peer counties tool. https://www.countyhealthrankings.org/resources/peer-counties-tool