Leveling Up

It is by now well-established that both health and health disparities are profoundly social phenomena. However, the majority of consumer health informatics interventions to date have focused solely on individuals, with particular attention accorded to influencing individual-level psychosocial characteristics and health behaviors. For example, a 2018 systematic review of mobile health intervention (mHealth) research for racial/ethnic minorities and/or those of low-socioeconomic status found that most interventions targeted individuals and were most commonly informed by psychosocial theories.1 This is despite the fact that interventions which rely primarily on individual effort, behavior, and choice tend to be less effective for marginalized groups—groups that experience socially stratifying processes of marginalization or exclusion from mainstream social, economic, cultural, or political life2—than those that target the context in which behavior occurs.3,4 Perhaps related to this individual focus, the impact of consumer health informatics interventions on marginalized groups is often limited. For example, the aforementioned systematic review revealed few significant impacts of mHealth interventions on health outcomes for these groups.1 In addition, as we have argued elsewhere, health informatics interventions are at particular risk of reinforcing health disparities by disproportionately benefiting nonmarginalized groups that already possess health-related advantages.5 The result can be intervention-generated inequality, in part from differential effectiveness of prevailing intervention models for marginalized groups.5 We contend that health informatics, including its subfields of consumer, population, and clinical informatics, will be more effective for marginalized groups (and less likely to produce intervention-generated inequality) if we better apply our understandings of the societal origins of health and health disparities to intervention design. We argue for greater emphasis on “upstream” interventions that focus on the social, political, economic, and physical contexts in which health is (re-)produced.6–8 We define these interventions, also called “structural” and “environmental,” as meso-level and macro-level interventions. In addition to the potential effectiveness improvements mentioned previously, these interventions may have multiple efficiency-related advantages through targeting multiple pathways to multiple health outcomes,8 and through reaching large numbers of people due to their lack of dependence upon individual patient/consumer agency for uptake.8 Despite the potential advantages of upstream interventions described above, their broad nature means that there are conceptual difficulties in applying this approach in health informatics. Accordingly, there is a need to systematically identify potential targets for interventions, to think critically about technological capabilities, and to consider the ways in which prior noninformatics interventions might better incorporate information and communication technologies (ICTs). To these ends, we introduce a model of health disparities that identifies potential targets for intervention at different levels of social organization. Next, we discuss types of noninformatics interventions that have addressed these targets. Finally, we present a typology of ICT capabilities, mapping the potential roles that ICT can play in influencing these intervention targets. We conclude by outlining the implications of upstream interventions for downstream intervention effectiveness, and consider their possible synergies. A MODEL OF HEALTH DISPARITIES: TARGETS FOR INTERVENTION Health disparities are persistent differences in the incidence and prevalence of disease, morbidity, mortality, and survival rates in one group compared with the general population. By contrast, health equity, which is the goal of upstream interventions, occurs when everyone has an equal opportunity to be healthy. Figure 1 shows an extended version of the World Health Organization (WHO) Committee on the Social Determinants of Health empirically derived conceptual framework. The framework maps pathways between health outcomes and social conditions, so as to highlight approaches to reduce health disparities and enhance health equity.9 As shown, the social determinants of health are found at the meso level and micro levels of social organization. However, the social determinants of health disparities are macro-level social, political, and economic mechanisms that generate socioeconomic hierarchies and a range of possible socioeconomic positions that individuals can occupy.9 These structural determinants then shape health through intermediary social determinants of health that are unevenly distributed, primarily due to macro-level factors.9FIGURE 1: Extended World Health Organization model of health disparities. LGBT indicates lesbian, gay, bisexual, and transgender.To increase the utility of the model for health informatics, we extend the original WHO model by mapping the role of ICTs at each level, as well as incorporating the contribution of “flexible resources” to health technology uptake in disparities.10 Furthermore, we incorporate insights from other models that have better detailed meso-level factors11–15 and environmental health-related determinants.16–18 In addition, because the WHO model primarily focused on socioeconomic status, we expanded this model to better represent health disparities rooted in social stratification based on sex,19–23 race/ethnicity,22–29 lesbian, gay, bisexual, and transgender identity,30–36 disability,30,37–41 and place of residence (including rural residence).25,42–44 Targets for Intervention in the Model The WHO model identifies specific pathways by which social stratification and marginalization influence socioeconomic position.9 Our expanded model suggests 4 broad types of intervention strategies to confront disparities: Influencing social hierarchies, which may involve disrupting processes of stratification and marginalization; Reducing exposures, which may be environmental, psychosocial, biological, or combined, as in physical stress responses; Decreasing vulnerability, both vulnerability to health conditions and vulnerabilities related to lower resource availability; Preventing unequal consequences of ill health, which may be related to health or socioeconomic position.45 If we apply this framework, we see that many informatics interventions are designed to reduce vulnerability to disease (eg, enhancing the psychological resource of self-efficacy, increasing physical activity, encouraging screening for early disease detection) or reducing negative consequences once disease has developed (eg, developing disease management or coping strategies, monitoring biometrics such as blood glucose at home, text-based medication, or health management reminders). We also note that many prior interventions do not specifically address differences in vulnerability or consequences for marginalized groups, but rather focus on consumer/patient groups in general. Targets of Upstream Interventions Outside of Health Informatics Outside of informatics, meso-level approaches have focused on living and working conditions (typically by reducing exposures) and social and community networks (typically by reducing vulnerabilities, though sometimes by reducing exposures to behaviors such as bullying). Meso-level interventions have included strategies such as redesign of spaces; changing social norms; making resources more or less available in a given area (eg, condom distribution programs, zoning to reduce density of alcohol retailers); and collective mobilization around unequal negative exposures such as environmental hazards or stigma.6,8,46,47 Health care system-led strategies have also included quality improvement initiatives designed to benefit all patients of the health care organization, as well as service reorganization and/or integration.48 Macro-level approaches focus on larger-scale phenomena such as the economy, political system, legislation, and culture. Macro-level interventions typically focus on influencing social hierarchies or reducing exposures or vulnerabilities. Interventions have included policy change; pricing through mechanisms such as subsidies, incentives and taxes; product modification (eg, food safety standards, requiring seatbelts in cars); and resource redistribution (eg, municipal minimum wage laws).8 At the macro level, redesigning health insurance programs (Medicaid expansion and insurance subsidies)49 and subsidizing health care through tax-supported community health centers50 have also been designed to make persons with lower income less likely to experience negative consequences from illness and accidents. Interventions which use the above approaches tend to organize efforts around: (1) specific settings (eg, schools, workplaces); (2) specific populations (eg, African Americans, low-income groups or neighborhoods); (3) specific behaviors/risks (eg, physical activity, smoking, motor vehicle accidents); or (4) enhancing resources and power for marginalized groups (eg, microfinancing programs to empower women economically, stable housing).8,46,51 Using the above noninformatics examples, we next identify potential roles for ICT below. POTENTIAL ROLES FOR ICT ICTs may have a negligible role to play in confronting some of the social determinants of health disparities. For example, ICT may not have much ability to ameliorate childhood poverty. Nevertheless, we hold that ICT can play a much greater role in upstream interventions than it has to date. To facilitate this role, we must systematically consider what ICT can do well. Therefore, as Table 1 shows, we outline key ICT capabilities, which include: (1) collection and display of information; (2) customization of information; (3) education; (4) mediation of communication; (5) support of workflows and activities; (6) framing and support of decisions; (7) social coordination; (8) optimization of resource allocation; (9) identification of patterns and anomalies; and (10) prediction of outcomes.TABLE 1: Potential Macro-level and Meso-level Interventions in Health InformaticsTable 1 presents a nonexhaustive list of potential ICT intervention types that can target health at macro and meso levels. These strategies address a mix of pathways, including influencing social hierarchies (eg, decision support for policy-makers, diversifying social networks beyond segregated schools and neighborhoods); decreasing exposures (eg, preventing injuries with smart vehicles and home appliances); reducing vulnerability (eg, enabling resource pooling); and preventing unequal consequences (eg, facilitating coordination and referrals between health systems and the nonprofit sector). A key, deliberate, aspect of the proposed intervention types (eg, collective action) is that they do not necessarily require significant agency on behalf of an individual patient or person at risk of a health condition to have an effect. Rather, this agency can be located in policy-makers, community activists, health care providers, public health officials, or nonprofit organization staff. Alternatively, the interventions reduce the effort involved in the activity, as with the approach of reducing complexity for patients by redesigning workflows. Although there are some prior examples of the informatics interventions in Table 1 (see references in the table), most have not been widely implemented—and others have yet to be realized at all. MULTI-LEVEL INTERVENTIONS: RECONCILING THE UPSTREAM AND THE DOWNSTREAM Although we have argued for a greater focus on upstream interventions in this paper, we acknowledge the complimentary need for downstream interventions that support marginalized individuals who want to change their health behaviors, but face multi-level barriers to doing so. To support these individuals, we should recognize that the effectiveness of individual-level interventions is sensitive not only to psychosocial, behavioral, and biological factors, but also to contextual factors beyond individual control. For instance, a recent meta-analysis of human immunodeficiency virus prevention interventions for African Americans found that condom use effect sizes were moderated by local levels of racism and racial residential segregation.90 Accordingly, building on previous noninformatics research, we also argue for the potential value of multi-level interventions, which typically target both individual-level and meso-level or macro-level factors (eg, school-based interventions that focus on both food supply and nutritional education91). Any of the intervention types outlined in Table 1 could be implemented as part of a multi-level intervention, or on its own. CONCLUSIONS The model and typology provided in this paper offer conceptual resources for upstream interventions in informatics. Specifically, they can enable informatics researchers and practitioners in informatics to “level up” by more systematically focusing their efforts on health equity, using an expanded set of intervention targets and strategies. We also hope to inspire more comprehensive thinking about the disparity-related mechanisms by which informatics interventions may operate. Although the approaches outlined in this paper may require some refocusing on our parts, we argue that leveling up offers unparalleled opportunities to enhance health equity with informatics.