There has been growing interest among health systems in population health (1, 2). Population health aims to improve the overall health of a population across the full continuum of care by more targeted, effective and coordinated health services (3). Given the rising trend of aging population and chronic disease burden, managing population health becomes more important for health systems trying to control cost (4). In order to improve outcomes and efficiency, health systems need to customize care and interventions based on identified risks and costs (5). One of the systematic approaches in the literature for targeting interventions to subgroups of patients with different needs is population segmentation or risk-stratification (referred as patient segmentation in the remainder of this paper). Population segmentation that divides a population into groups with related service needs is an important foundation for effective and sustainable care delivery (6–8). Segmentation divides patients into distinct groups with specific needs, characteristics or behaviors and allows for health services to be organized around patients with similar needs (7). Patient segmentation models are becoming essential element of healthcare management due to the increase in the number of programs that incentivize value-based care (9). Although patient segmentation models can help design interventions targeting subgroups of patients, they are often based on International Classification of Diseases (ICD) codes found in electronic health records (EHRs) and/or insurance claims data and lack important social risk factors that are essential for designing interventions. World Health Organization defines social determinants of health (SDOH) as the conditions in which people are born, grow, work, live, and age (10). These factors include economic policies and systems, development agendas, social norms, social policies and political systems (10). There are numerous studies demonstrating social factors acting as powerful determinants on multiple health outcomes including coronary heart disease (11), breast cancer (12), childhood obesity (13) and end-stage renal failure (14). Literature suggests that high utilizers of healthcare resources among Medicaid and uninsured population often have multiple chronic conditions (15, 16) and programs targeting this population collectively argue that social risk factors including but not limited to language, health literacy, unemployment, substance abuse and housing are important drivers of healthcare utilization (17, 18). Most of the current patient segmentation models use administrative billing data because insurance claims data provides a nearly complete view of patients' interactions with health care delivery system; therefore, it is a reliable source to extract utilization outcomes (19). Majority of the EHRs on the other hand contain data from clinical encounters occurring between individuals and providers within a single health system and hence miss out of network events (19). On the positive side of EHRs is that they offer more extensive data including family history, lab results, vital signs and symptoms which could help improve the population segmentation model (20). One drawback of reliance on insurance claims data and EHRs is that they miss social and behavioral factors that complicate care (21). Although, there is a subset of ICD-10-CM codes, the Z codes, for documenting SDOH in EHRs, these codes are underutilized (22, 23). As such, SDOH Z codes may not reflect the actual burden of social needs experienced by patients. To address this gap, this paper presents the complementary benefit of consumer data when it is linked to EHRs or insurance claims data. The consumer marketing data include individual-level SDOH (including income, education, lifestyle variables, language spoken, household size, smoking status, life events, shopping activity) that are not available in the insurance claims data or majority of EHR data. The combined data provides 360-degree view of patients and can help predict the risk of repeat emergency room visits or hospital admissions (24). Inclusion of SDOH is essential to improve population health as medical interventions without addressing social determinants are not sustainable and effective. This unprecedented view into the lives of patients has significant potential to improve upon segmentation approaches relying exclusively on health plan or EHR data that lack measures or even decent proxies for fitness, diet and other SDOH which can profoundly alter the course of chronic diseases. A number of commercial companies provide marketing data that is well-utilized by organizations that subscribe to their services. Experian's ConsumerViewSM U.S. database is one of the world's largest consumer database on more than 300 million individuals and 126 million households (25). ConsumerViewSM U.S. database is compiled from hundreds of resources. For example, property and mortgage data are compiled from public records and county deeds while lifestyle and interest data are compiled from consumers who have completed self-reported surveys (25). Marketing companies match and mange patient identity across the healthcare ecosystems enabling the linkage of datasets across channels and silos (26, 27). According to Acxiom, two-thirds of hospitals actively use or want third-party consumer and lifestyle data to improve patient care (24).
Read full abstract