Recognizing that low-value care is shaped by the broader healthcare environment and influenced by socioeconomic factors impacting care, Milliman MedInsight developed a model to explore the connections between low-value care and social determinants of health. The goal is to offer the healthcare industry actionable insights to enhance quality, efficiency, equity, and outcomes. In this blog post, we present the context, key features of the model, and illustrate its application through real-world examples.
Background
Social Determinants of Health (SDOH) refer to the non-medical factors that influence health outcomes, including conditions of birth, growth, living, working, and aging, as well as broader economic, social, and environmental system1. Key elements include education, economic stability, healthcare access, neighborhood environment, and social support.
Low-Value Care (LVC) refers to medical services that offer little to no benefit, potentially causing unnecessary costs and harm. A common example2 is preoperative EKGs for low-risk patients undergoing minor procedures like cataract surgery. Such tests can lead to false-positives, and further unnecessary services, escalating healthcare costs without improving outcomes. Accumulation of LVC diverts resources from essential care, worsening health outcomes and disparities, and weakening healthcare system efficiency.
Over the past decade, efforts like the Choosing Wisely campaign by the American Board of Internal Medicine Foundation have identified over 500 low-value tests and procedures, promoting clinician-patient discussions to reduce LVC. A 2021 review found that interventions based on these recommendations can effectively change practice patterns.
LVC is driven by both providers and consumers. Providers may follow established practices or respond to financial incentives and perceived patient expectations. Patients may demand unnecessary services due to limited health literacy or cultural beliefs. These behaviors are influenced by the broader healthcare environment, including systemic incentives and SDOH. For instance, a “more is better” mentality and insurance coverage can encourage LVC, while education and language barriers can lead to patient consent to unnecessary services.
Examining LVC through the lens of SDOH
Empirically, is there data that supports the association between LVC and SDOH? To explore this, we leverage the Social Vulnerability Index (SVI) and the Milliman MedInsight Health Waste Calculator (HWC).
Data and methodology
Social Vulnerability Index
The SVI3, developed by the CDC, measures how socioeconomic factors such as poverty, access to transportation, and household composition contribute to community vulnerability. The index includes themes such as socioeconomic status, household composition, minority status, race, ethnicity, language, and housing type. While the SVI provides a good representation of community vulnerability, it was not specifically developed for measuring low-value care. Nonetheless, it offers valuable geographic granularity that can match up well with healthcare claims data for an exploratory analysis such as ours.
MedInsight Health Waste Calculator
The MedInsight Health Waste Calculator (HWC)4 is a collaborative effort between Milliman MedInsight5 and VBID Health6, leveraging USPSTF D rated services and several Choosing Wisely7® recommendations. These recommendations, provided by national medical organizations, help identify medical tests and procedures that are often unnecessary. Using claims data, the Health Waste Calculator quantifies healthcare services that lack value based on clinical context, offering a comprehensive tool for addressing LVC. A key metric from the HWC is the Waste Index, which is percent of LVC utilization as a share of total qualified services.
Model overview
Combining the SVI and the Waste Index in modeling, we developed the Health Waste Disparity Index, a measure that associates the impact of SDOH on the administration of wasteful healthcare services. With this approach, we hope to foster a deeper understanding of key questions, such as: Does LVC exhibit the same geographic variation as overall healthcare utilization? Do factors that influence overall healthcare utilization impact LVC in a similar way? For instance, while we know that health literacy affects overall healthcare utilization and outcomes, how does it impact the prevalence of LVC? And more importantly, how can the industry leverage this knowledge to reduce LVC effectively?
Case study – Tale of two counties
Let’s take a closer look at the wasteful use of electrocardiograms (EKGs), other cardiac screenings, and opioid prescriptions in two California counties: San Bernardino and Santa Clara. San Bernardino County, with an SVI of 0.825, indicates higher social vulnerability compared to Santa Clara County, which has a significantly lower SVI of 0.1938. The bar charts below illustrate the percentage of wasteful utilization by county and by insurance coverage type.
Low value EKG and other cardiac screenings are used less in San Bernadino relative to Santa Clara. This may be because communities with lower socioeconomic levels such as San Bernardino County have less availability of these services. Providers in such regions may also prioritize more essential services given the limited healthcare resources, which could explain the lower rates of wasteful EKG utilization. It may also be the case that patients in higher SVI areas like San Bernardino get fewer wasteful screenings because they get less of all types of services (high and low value).
In contrast, the wasteful prescription of opioids shows a different pattern. San Bernardino County has notably higher rates, with 94% and 99% wasteful use for Medicare and commercial insurance, respectively, compared to Santa Clara’s significantly lower rates. This higher opioid waste in San Bernardino may be influenced by several factors tied to its higher SVI, such as a greater prevalence of chronic pain, fewer alternatives for pain management, and socioeconomic stressors like poverty and unemployment, which can exacerbate reliance on opioids. Additionally, historical trends of overprescription in high-SVI areas may further contribute to these elevated opioid waste rates.
Although we cannot pinpoint the exact causes, the variations in Low-Value Care (LVC) between these counties are closely tied to local socioeconomic conditions, which affect healthcare access, resource allocation, and provider behaviors. In higher vulnerability counties like San Bernardino, addressing opioid overprescription could be supported through targeted education, updated prescribing guidelines, and improved access to alternative pain management options. In contrast, lower vulnerability areas like Santa Clara could benefit from reducing the overuse of diagnostic tests, such as EKGs, by revisiting clinical practice guidelines and promoting value-based care incentives to encourage more efficient use of healthcare resources.
Discussion
Addressing low-value care requires a nuanced approach that accounts for geographic and socioeconomic disparities. Applying indices like the SVI and tools like the MedInsight Health Waste Calculator can provide crucial insights into how social vulnerability impacts healthcare quality. Collaborative efforts among healthcare providers, policymakers, and data scientists are essential to reduce wasteful healthcare practices, improve health equity, enhance overall population health and increase efficiency of health care delivery.
Limitations
Please note the following limitations. The analysis was based on the 2022 Milliman MedInsight Emerging Experience data using the Milliman MedInsight Health Waste Calculator. While the dataset includes over 80 million covered lives across various health insurance types in the U.S., it may not be fully representative of all geographic areas or specific subpopulations, limiting the generalizability of the results. High-level data quality checks were performed, but potential errors or inconsistencies might still impact the validity of the findings. Other SDOH indices might lead to different conclusions. Furthermore, potential confounding variables and limited granularity in modeling may affect the observed outcomes. Future research should address these limitations by using more granular, representative, and up-to-date data, and improved methodologies to enhance the robustness and generalizability of the findings.
References:
1. See the World Health Organization’s framework on SDOH, “https://www.who.int/health-topics/social-determinants-of-health#tab=tab_1”
2. “Stopping the Flood: Reducing Harmful Cascades of Care”, by Chandrashekar, P, Fendrick, AM, and Ganguli, I. The American Journal of Managed Care, 27(5), May 2021
3. See https://www.atsdr.cdc.gov/placeandhealth/svi/data_documentation_download.html
5. See https://medinsight.com/
6. See https://vbidhealth.com/
7. See https://www.choosingwisely.org/
8. Refer to the 2022 SVI, retrievable at https://www.atsdr.cdc.gov/placeandhealth/svi/data_documentation_download.html