Measuring the Effect of Social Determinants on Patient Outcomes:A Systematic Literature Review
Given the movement towards value-based purchasing in the United States, health care leaders need methods to characterize and address the complex effect that social determinants have on health care outcomes. This systematic literature review was specifically designed to understand current research on the effect that patient material and social deprivation has on health care delivery outcomes and the potential benefit of clinical interventions designed to mediate this effect. A total of 310 studies were identified for review with 80 studies included in the final synthesis. Results highlight significant variation in the methods used to measure the effect of social determinants on health care outcomes and the need for common measurement standards. More robust identification of deprivation-sensitive diseases or conditions is needed to channel scarce program resources to effected conditions. Finally, further research is needed to evaluate the benefits of data-driven, tailored clinical interventions designed to serve the needs of materially-deprived patient populations.
Social determinants of health, population health, health care organizations, material deprivation, delivery systems, socioeconomic status, patient outcomes, quality improvement, evaluation
Population health follows a social gradient.1 People who are economically and/or socially less deprived have better health outcomes.1,2 Social determinants of health ("social determinants") are defined by the World Health Organization as "the circumstances in which people are born, grow up, live, work and age, and the systems put in place to deal with illness"3[p.1] and describe the social, economic and political processes and relationships that can influence key health outcomes.3 People higher on the social gradient generally have more favorable social determinants that affect health.1 Payer demand for more value-based purchasing of health care that holds health care organizations responsible for health care outcomes increases the need for these organizations to understand how social determinants affect outcomes and population health in the context of health care delivery.
The term deprivation is used in the literature to describe the "disadvantaged position [End Page 81] of an individual, family or group relative to the society in which they belong"  and has economic and social dimensions.1 Material deprivation includes the lack of basic resources for living and is closely related to measures of socio-economic status and poverty.4 Social deprivation describes the lack of support provided by other persons.5 Material and social deprivation are correlated.6–7 Both forms of deprivation can be measured at either the person or ecologic level and are shown to have independent effects on health.1 Research suggests that patients who are more materially deprived or that come from more materially deprived neighborhoods have poorer health care outcomes, including increased mortality,8 higher emergency department (ED) utilization,9 increased readmission risk,10 delays in time to diagnosis and treatment,11–12 poorer medication adherence,13 and less effective engagement in shared decision making.14
A proposed causal pathway between lower socioeconomic status and poor health care outcomes includes more limited access to care (result: inadequate treatment and increased risk of complications), lower quality of care and poorer self-care behaviors (including diet and exercise).15 More recent theories point to related issues including patient health literacy16 or patient activation and engagement17 as contributing factors. As a result, barriers to receiving equitable care may include limited resources to obtain care, communication difficulty between the patient and the provider and challenges navigating the health care delivery system.18 The Institute for Healthcare Improvement (IHI),19 the Institutes of Medicine (IOM),20 the National Academies of Science, Engineering and Medicine (NASEM)4 and the Agency for Healthcare Research and Quality (AHRQ)21 have each proposed frameworks to characterize these relationships.
The purpose of this study was to characterize the results of recent published research on two critical questions facing health care organizations. First, how does patient or area deprivation modify the effect of standard of care interventions? Second, what targeted or design interventions modify the effect of patient or area deprivation on health care outcomes? Understanding the relationship between material deprivation and health care outcomes can assist health care organizations in designing effective interventions that address the potentially distinct needs of these more vulnerable populations, reduce health care disparities and lower the cost of care delivery.
A systematic review of the peer-reviewed literature was conducted. Research into health and health care disparities is an extremely broad topic. The Conceptual Framework of Social Risk Factors for Healthcare Use, Outcomes and Cost ("the Framework") developed by NASEM was the basis for development of the analytic framework for the review as noted in Figure 1.4
Evidence supports utilization of the PICO framework to improve searching PubMed and other data bases for clinical questions.23 Following the PICO model, the following terms were set: [End Page 82]
[End Page 83]
• Population = All U.S. or Canadian patients receiving health care at a delivery system for any disease or condition
• Intervention or exposure = Deprivation or poverty and related clinical interventions
• Comparator = Delivery system interventions designed to mitigate the effect of deprivation on health care outcomes
• Outcomes = Health care outcomes including mortality, morbidity, utilization, cost, patient/clinician behaviors
Based upon this model, the following initial query was developed and run in PubMed in August 2015. A subsequent update to the query was run in May 2016 to capture more recent studies.
[("patient deprivation" OR "area deprivation" OR "neighborhood deprivation" OR "community deprivation" OR "social deprivation" OR "deprivation index" OR "social determinants" OR "socio-economic status" OR "socioeconomic status" OR "poverty" OR "high school education" OR "household income"] AND [("United States" OR "Canada")] AND [("health system" OR "health care "OR "integrated health system" OR "delivery system")]
Despite differences in payment systems, Canadian studies were included given Canada's close geographic proximity to the United States, its similar standard of medical care and the similar challenges faced by both healthcare systems in addressing the needs of underserved populations. Study criteria were limited to systematic reviews, observational [End Page 84] and experimental studies, case reports, and evaluation studies performed in the past 10 years and reported in English in PubMed. Title and abstract screening criteria was used to exclude studies that met the criteria listed in Box 1. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) model was used to report results as noted in Figure 3.24 Once the final set of peer-reviewed articles was identified, studies were further classified based upon study design type (observational versus experimental), deprivation measure type (by characteristic measured and measurement method used), disease or condition classification, primary study outcomes (classified using the NASEM Framework (Figure 1)), and health care delivery system intervention components.
Consistent with the analytic framework presented in Figure 2, two sets of studies were identified. The first set of studies included observational studies (n=66) that specifically examined the effect modification of deprivation on existing standard of care interventions and related health care outcomes. Patient-level studies assessed the underlying risk factors associated with deprivation, patient behaviors and health care outcomes. Clinician-level studies included understanding the association between patient deprivation and disparities in encounter-level clinician behavior. Characteristics of health care delivery system performance associated with disparities in health care outcomes were also included. Given the frequency of use, studies using insurance status as a proxy for deprivation were included despite limitations using this approach as noted in the Discussion section.
The second set included experimental and quasi-experimental studies (n=14) that identify targeted or design interventions that may mediate the impact of patient or area deprivation on health care outcomes. Health care intervention components included enhanced clinical content, workflow redesign, additional care support, and/or the [End Page 85] introduction of new technologies to enable non-traditional patient and clinician interaction. Within this set, studies were further classified to identify those studies that were performed in a health care setting or with a cohort of patients selected from a health care setting from those performed in the general population. Interventions were also classified based upon phase of treatment (screening/prevention, diagnosis, treatment, monitoring/follow up), location of care (inpatient, ED, primary care) and intervention components (such as evaluation and counseling).
A total of 310 studies were identified for review with 80 studies included in the final synthesis as described in Figure 3. The initial and subsequent queries identified a set of 307 articles for analysis. An additional 3 articles were identified from reviewing references included in the article set. Primary reasons for exclusion of studies after the title/abstract screening included population age (n=48), studies limited to evaluating [End Page 86] race/ethnicity as the primary measure of social determinants (n=38), studies outside the US and Canada (n=35), and studies limited to examining the association between deprivation and the general incidence or prevalence of disease in the general population (n=31). A total of 93 full-text articles were assessed for eligibility. Primary additional exclusions included studies that were limited to broad structural issues with the national or regional system or that evaluated structural disparities across payer type that focused specifically on insurance plan design and access (n=11).
Deprivation measures used in the studies were focused almost exclusively on material deprivation. Only a limited number of studies included elements of social deprivation.25–27 Given this, the term material deprivation will be used in describing review results. Meaningful variation existed in the characteristics measured to determine material deprivation status (individual versus neighborhood) and the measurement methods used (individual or person-level, neighborhood-compositional, neighborhood-contextual) noted in Table 1.
Summarizing overall results by frequency of measures used (n=154), 86% of the material deprivation measures used were based upon person-level characteristics involving three distinct methods. Sixty-eight percent (68%) of person-level measures were patient self-reported measures of individual characteristics. The most common individual characteristics included a combination of income, education, payer status, employment status or Federal Poverty Level status of the individual. Twenty-seven percent (27%) of person-level measures involved a neighborhood compositional estimate used to estimate individual deprivation. The most common measures were similar combinations to direct capture of individual characteristics but were averaged at a small-area level and then assigned to patients based upon address of residence. The majority of these small-area methods used measures of the neighborhood boundaries based upon ZIP code followed by census tract. The remaining 5% of person-level measures used neighborhood contextual measures to describe deprivation characteristics of individuals. The most common of these was to use patient admission to a specific hospital located in a particular deprived neighborhood to define the deprivation status of the patient (i.e., the patient was considered materially deprived because they were admitted to a hospital that resides in a neighborhood defined as deprived). The use of neighborhood characteristics to define material deprivation status were used less frequently (14% of total measures used). Three similar measurement methods were used with the frequency more evenly distributed by method as noted in Table 1.
Summarizing results by study (n=80), 50% of studies (39/80) used a single measure of material deprivation—generally income or payer status. The remaining studies used multiple measures with five studies (6%) using some form of composite measure.
Contextually, studies of the impact of material deprivation on patient outcomes varied by disease type as noted in Table 2. Summarizing results by disease or condition studied, cancer was most frequent (36%), followed by cardiovascular (16%), all-cause disease studies (15%), and diseases of the endocrine system (primarily diabetes) (9%).
Classifying observational studies using the NASEM Conceptual Framework, most studies noted significant variation in health care outcomes as material deprivation increased. Material deprivation was associated with access to care/treatment received.18,28–30 Material deprivation was negatively associated with patient behavior [End Page 87] risk factors that affect health care outcomes including use of preventive care,31–37 timing of diagnosis and resolution,13,27,34,37–41 demonstration of self-care behaviors (including treatment adherence),25,42–43 and disease control.44–45
Studies capturing the effect of material deprivation on the clinical care process including provider or health system behaviors suggest variability in treatment given;14,46–58 undertreatment;59–62 delayed treatment;63 higher failure to rescue rates;64 effectiveness of clinician communication;18 and other effects.26,65–68 Material deprivation was associated with increased inpatient length of stay;54,69–70 increased hospitalizations,71–72 higher [End Page 88] inpatient readmissions;73–74 higher hospital transfer rates;75 and other effects.76–77 Study findings included both increases in Emergency Department (ED) visits44,78 and no effect.79 The effect of deprivation on primary care visits varied by age.44,79 All three studies that included costs included lower inpatient costs for materially deprived patients, generally associated with fewer procedures performed and possible undertreatment.54,63,66
Other health care outcomes associated with material deprivation included higher rates of mortality,46,50,55,58,63–64,70,80–86 increased complication rates,44,64,69 and lower quality of life following care.44,87–88
Examining experimental studies, 14 studies had some form of intervention that either directly or indirectly addressed the potential impact of patient or area material deprivation on health care outcomes. Of these, 13 were conducted within a health care organization or using patient data directly from a health care organization and are listed in Table 3. Nine of 13 studies used person-level measures of material deprivation [End Page 89]
[End Page 91]
(income, education, employment, insurance status). The remaining studies used either neighborhood compositional,89 neighborhood contextual factors,90–91 or a combination of individual and neighborhood composite factors92 to estimate person-level material deprivation status. The most common diseases/conditions studied included diabetes (4),91,93–95 cancer (4),90,95–97 and depression (3).89,96,98 The studies covered most aspects of the health care process including disease prevention,91diagnosis,90,96,98 treatment,89,93,95,97,99 and patient monitoring and follow up.92,94,100,101 The most common enhanced service elements included patient education, evaluation and counseling as noted in Table 4. Services were delivered on-line, telephonically or in-person.
Of the 13 intervention studies, seven included interventions with specifically adapted intervention components directly designed to address the needs of more materially deprived patients. These interventions considered assumptions regarding the underlying characteristics of materially deprived patients and how they would interact with the health care delivery system including access to telephonic/data resources,94 design and content of education materials,93 and tailored counseling96 that addressed specific needs of more materially deprived patients. Six of these studies found improvements in patient outcomes following the introduction of an intervention including increased patient engagement,98 improved regimen adherence,91,96 improved disease control,94 improved patient outcomes93 and increased patient satisfaction.101 The remaining intervention study found no improvement following intervention in timely diagnostic resolution among indigent women.90 No interventions were designed to provide information on patient material deprivation status that could inform clinicians directly at the point of care. [End Page 92]
A growing body of evidence points to a correlation between patient material deprivation and health care disparities. However, questions remain. Under what conditions does material deprivation affect the health care experience and health care outcomes? Does it vary by disease? How do patient versus neighborhood-level measurement methods influence results? When should health care delivery systems tailor care to the needs of materially deprived patients? What interventions are most effective in reducing disparities in care? Studies on interventions specifically designed to mediate the impact of material deprivation were limited overall and by specific contexts, including patient-or neighborhood-level characteristics, disease or condition type, phase of the care process, place of service and the intervention components.
Measuring patient material deprivation
Proper identification of materially deprived patients is an important first step in efforts to measure the true scope of disparities in health care outcomes and to evaluate delivery interventions. Two distinct groups of material deprivation measures emerged from this review—person-and neighborhood-level measures—that were often used interchangeably. More precisely, measurements of individual deprivation status are designed to measure an individual patient's circumstances using patient self-reported measures including income, education level and employment status. Such measures are linked to health and health care outcomes by measuring patient capacity. Neighborhood or small-area measures are designed to characterize at a more macro-level the circumstances in which the patient lives, including contextual factors that influence health and health care outcomes. Local environmental economic and social conditions have been linked to general health outcomes through the interaction of individuals with their local community.102–107
Traditional measures of person-level material deprivation including race and insurance status remain common due to data availability but are increasingly problematic as measures of material deprivation. Race is a complex construct with potential to characterize both genetic and social elements.108 While race has been historically linked to material deprivation in certain populations, evidence of health disparities in poorer white populations is increasing.109 Asian Americans males now have the highest median income of any racial group.110 As a measure of deprivation, insurance status is transitory in nature. Eligibility requirements for Medicaid patients are also highly specific, limiting identification of deprivation within certain populations, including adult males, non-child bearing women and the elderly.111 Other common measures including education, income and occupational status have strengths and weaknesss.112–113
More recent developments in the United States include the introduction of deprivation indices common in Western European countries, designed to provide a geographic based view of material and social deprivation experiences by neighborhood. These composite measures include a combination of several risk factors associated with population characteristics such as mortality or morbidity.114 When used, considerable variation exists in the geographic breakdown of these index measures. Neighborhood units of measure in these studies included counties, ZIP codes, U.S. census tracts and block groups as well as urban/rural designations. The most common measure used was ZIP code, which is an artificial construct developed by the U.S. Postal System [End Page 93] to efficiently deliver mail and has little association with actual neighborhood-level interaction. Census tract small areas approximate neighborhood constructs better than ZIP code.115–116
Local and national standardization of a common set of measures and measurement methods designed to identify both materially and socially deprived patients for planning, research or clinical care within a health care setting is needed. Some researchers have argued for the use of multiple measures of socio-economic status in research with the selection of specific measures linked to the appropriate health care outcomes.112–113 Multi-level measures of deprivation status that incorporate both person-level and neighborhood-level characteristics into a single, two-dimensional bundled measure should be evaluated. Capturing the bundled components separately would support analysis into the relative weighting of person-level and neighborhood-level deprivation status on health care outcomes. A more robust set of characteristics that include social support elements as well would improve understanding of the underlying mechanisms promoting disparities in care.107
Material deprivation and health care outcomes by disease
A proposed causal pathway between lower socioeconomic status and poor health care outcomes includes poorer access to care (result: inadequate treatment and increased risk of complications), lower quality of care and poorer self-care behaviors (including diet, exercise).15 This is reflected in the more recent NASEM Framework included in Figure 1.
More recent theories are examining inequities in patient outcomes through understanding underlying variation in the burden of disease on similar patients. Patient burden of disease varies by individual patient characteristics including patient living circumstances, capacity and resilience.117 As a result, health care outcomes for a similar disease of similar severity may vary by person. Using this theory, high-burden diseases requiring regular access to care or a high level of self-care, for example, to maintain disease control, may disproportionately affect deprived patients leading to poorer health care outcomes for these groups. The concentration of studies in this systematic review in patients with cancer, heart disease, diabetes, mental health and other chronic conditions suggests that material deprivation may produce greater disparity in health care outcomes in diseases with a high burden that require regular care access, self-efficacy, activation and engagement.
Identification of potentially deprivation-sensitive diseases or conditions, including the relative impact of specific diseases or conditions on health care outcomes, could assist delivery systems in the design and development of disease-specific pathways that address the potentially distinct health care needs and available social supports of deprived patients with specific diseases. Other contextual factors that may influence the effect of patient material deprivation on health care outcomes should be evaluated.
Mediating the effect of material deprivation on health care outcomes
The initial studies identified through this systematic review highlight the potential positive impact that certain interventions can have on health care outcomes for materially deprived patients by addressing patient resource, communication and navigation barriers. Research into the effect of material deprivation on health care outcomes is highly contextual. Future studies should examine intervention effects in light of disease or condition type, the phase of the care process and the accessibility of technology [End Page 94] and so forth. Potentially generalizable elements of any intervention including patient counseling, education, alert and monitoring, and communication, should be tailored by contextual factors.
Introduction of lay support resources present one interesting and potentially cost-effective approach to addressing the needs of materially deprived patients.90,118 Evidence suggests that material deprivation is linked to increased social isolation that can have negative effects on health outcomes.7,119 Interventions designed to mitigate the effects of isolation have potential to improve health care outcomes.120 The presence of informal social supports including extended family, neighborhood and community resources are positively associated with better health care outcomes and present an important avenue for further study.121
It is worth noting the role of technology as a delivery mechanism for addressing shortcomings in more traditional in-person interactions between clinicians and patients. The underlying enabler for patient-clinician interaction in the majority of these solutions is communications technology which, given its relative ease of use, low cost and ubiquitous nature, has the potential to deliver health care solutions that transcend socio-economic class.122 Communications technology also has the potential to reduce social isolation.
Conclusion and future research
Health services research regarding the impact of deprivation on health care outcomes is fragmented with limited interventions in place. Identifying a measure of social determinants that applies across diverse settlement patterns and is readily available holds promise to address unmet measurement needs in evaluating impact of social determinants on effective treatment, quality improvement and value-based purchasing. There is a need to expand studies beyond select chronic conditions and to establish clear associations between deprivation and patient outcomes by disease type or condition, perhaps leading to the identification of deprivation-sensitive diseases most affected by patient deprivation. Similar work is needed to examine other contextual effects. More research is needed to examine the effect of deprivation in the context of care delivery including understanding and testing interventions specifically designed to mediate the impact of deprivation on health care outcomes for these more vulnerable populations.
ANDREW J KNIGHTON, BRAD STEPHENSON, and LUCY A. SAVITZ are all affiliated with the Intermountain Institute for Healthcare Delivery Research, Intermountain Health Care in Salt Lake City, UT.