Abstract
Background
Geographically based social determinants of health (SDoH) measures are useful in research and policy aimed at addressing health disparities. In the United States, the Area Deprivation Index (ADI), Neighborhood Stress Score (NSS), and Social Vulnerability Index (SVI) are frequently used, but often without a clear reason as to why one is chosen over another. There is limited evidence about how strongly correlated these geographically based SDoH measures are with one another. Further, there is a paucity of research examining their relationship with patient-reported outcome measures (PROMs) in orthopaedic patients. Such insights are important in order to determine whether comparisons of policies and care programs using different geographically based SDoH indices to address health disparities in orthopaedic surgery are appropriate.
Questions/purposes
Among new patients seeking care at an orthopaedic surgery clinic, (1) what is the correlation of the NSS, ADI, and SVI with one another? (2) What is the correlation of Patient-Reported Outcomes Measurement Information System (PROMIS) Global-10 physical and mental health scores and the NSS, ADI, and SVI? (3) Which geographically based SDoH index or indices are associated with presenting PROMIS Global-10 physical and mental health scores when accounting for common patient-level sociodemographic factors?
Methods
New adult orthopaedic patient encounters at clinic sites affiliated with a tertiary referral academic medical center between 2016 and 2021 were identified, and the ADI, NSS, and SVI were determined. Patients also completed the PROMIS Global-10 questionnaire as part of routine care. Overall, a total of 75,335 new patient visits were noted. Of these, 62% (46,966 of 75,335) of new patient visits were excluded because of missing PROMIS Global-10 physical and mental health scores. An additional 2.2% of patients (1685 of 75,335) were excluded because they were missing at least one SDoH index at the time of their visit (for example, if a patient only had a Post Office box listed, the SDoH index could not be determined). This left 35% of the eligible new patient visits (26,684 of 75,335) in our final sample. Though only 35% of possible new patient visits were included, the diversity of these individuals across numerous characteristics and the wide range of sociodemographic status—as measured by the SDoH indices—among included patients supports the generalizability of our sample. The mean age of patients in our sample was 55 ± 18 years and a slight majority were women (54% [14,366 of 26,684]). Among the sample, 16% (4381of 26,684) of patients were of non-White race. The mean PROMIS Global-10 physical and mental health scores were 43.4 ± 9.4 and 49.7 ± 10.1, respectively. Spearman correlation coefficients were calculated among the three SDoH indices and between each SDoH index and PROMIS Global-10 physical and mental health scores. In addition, regression analysis was used to assess the association of each SDoH index with presenting functional and mental health, accounting for key patient characteristics. The strength of the association between each SDoH index and PROMIS Global-10 physical and mental health scores was determined using partial r-squared values. Significance was set at p < 0.05.
Results
There was a poor correlation between the ADI and the NSS (ρ = 0.34; p < 0.001). There were good correlations between the ADI and SVI (ρ = 0.43; p < 0.001) and between the NSS and SVI (ρ = 0.59; p < 0.001). There was a poor correlation between the PROMIS Global-10 physical health and NSS (ρ = -0.14; p < 0.001), ADI (ρ = -0.24; p < 0.001), and SVI (ρ = -0.17; p < 0.001). There was a poor correlation between PROMIS Global-10 mental health and NSS (ρ = -0.13; p < 0.001), ADI (ρ = -0.22; p < 0.001), and SVI (ρ = -0.17; p < 0.001). When accounting for key sociodemographic factors, the ADI demonstrated the largest association with presenting physical health (regression coefficient: -0.13 [95% CI -0.14 to -0.12]; p < 0.001) and mental health (regression coefficient: -0.13 [95% CI -0.14 to -0.12]; p < 0.001), as confirmed by the partial r-squared values for each SDoH index (physical health: ADI 0.04 versus SVI 0.02 versus NSS 0.01; mental health: ADI 0.04 versus SVI 0.02 versus NSS 0.01). This finding means that as social deprivation increases, physical and mental health scores decrease, representing poorer health. For further context, an increase in ADI score by approximately 36 and 39 suggests a clinically meaningful (determined using distribution-based minimum clinically important difference estimates of one-half SD of each PROMIS score) worsening of physical and mental health, respectively.
Conclusion
Orthopaedic surgeons, policy makers, and other stakeholders looking to address SDoH factors to help alleviate disparities in musculoskeletal care should try to avoid interchanging the ADI, SVI, and NSS. Because the ADI has the largest association between any of the geographically based SDoH indices and presenting physical and mental health, it may allow for easier clinical and policy application.
Clinical Relevance
We suggest using the ADI as the geographically based SDoH index in orthopaedic surgery in the United States. Further, we caution against comparing findings in one study that use one geographically based SDoH index to another study’s findings that incorporates another geographically based SDoH index. Although the general findings may be the same, the strength of association and clinical relevance could differ and have policy ramifications that are not otherwise appreciated; however, the degree to which this may be true is an area for future inquiry.
Introduction
There is a growing appreciation of the presence of health disparities in the United States, which are often driven by social determinants of health (SDoH) [15]. As defined by the WHO, SDoHs are “social, physical, and economic conditions that impact upon health” [16]. In orthopaedic surgery, SDoHs play a central role in patient health, care access, symptom severity, and clinical outcomes before and after operative and nonoperative interventions [2, 5, 9, 21, 25, 27]. Thus, it is critical to have a thorough, quantitative means of measuring SDoH to ensure the outcome of any intervention or initiative aimed at addressing health disparities can be accurately assessed. Further, it is important to understand the level to which presenting symptoms (such as functional and pain levels) may be associated with SDoH markers to better understand potential policy impacts on musculoskeletal health.
Many SDoH factors are potentially modifiable, and modifying them may help patients of all backgrounds improve their health and well-being. This could play an important role in efforts to decrease health disparities. For example, transportation vouchers may assist those with minimal means of traveling to a clinic appointment, or prescription coupons can help patients afford recommended medications when they otherwise would be unable to pay for them. In addition, payers, such as government agencies or insurance companies, recognize the need to develop novel payment structures that risk adjust for SDoH to avoid situations in which only patients who are more likely to have good clinical outcomes are treated or patients who are more likely to have poor clinical outcomes are not treated, despite having similar injuries. This work has begun globally, including in New Zealand, the United Kingdom, and the United States [11]. For example, in the United States, Massachusetts redesigned part of its Medicaid (MassHealth) payment structure to incorporate the Neighborhood Stress Score (NSS) [3]. Additionally, the University of Pittsburgh Medical Center used the Area Deprivation Index (ADI) as an additional decision-making tool for coronavirus-19 vaccinations [17]. Further, the Social Vulnerability Index (SVI) has been proposed to identify populations that require additional resources or support in the setting of an emergency or a natural disaster [1]. All three of these SDoH estimates (NSS, ADI, and SVI) are based on variables calculated using United States Census tract data [1, 8, 23], making the scores directly related to where a patient resides. With the growing use of these three geographically based SDoH indices in research, policy, and payor initiatives, it is critical that we better understand whether they measure the same construct to ensure apples-to-apples comparisons of interventions. Further, if we are to use such measures in orthopaedic surgery, it would behoove us to appreciate which measure (or measures) most strongly correlate with the symptoms our patients care about most, such as physical function and pain. This will allow us to create initiatives that better target the components of the geographically based SDoH index that is most correlated to such symptoms.
In the present study, we asked: (1) What is the correlation of the NSS, ADI, and SVI with one another? (2) What is the correlation of Patient-Reported Outcomes Measurement Information System (PROMIS) Global-10 physical and mental health scores and the NSS, ADI, and SVI? (3) Which geographically based SDoH index or indices are associated with presenting PROMIS Global-10 physical and mental health scores when accounting for common patient-level sociodemographic factors?
Patients and Methods
Study Design and Setting
This was a retrospective, observational study drawn from a large, institutional database. Between June 2016 and December 2021, all new orthopaedic patients across all orthopaedic subspecialties 18 years or older presenting for a clinic visit across multiple sites associated with a tertiary referral academic medical center were identified. Given that our institutional database is diverse in many ways (that is, across patient sociodemographic factors, clinic sites, and orthopaedic subspecialties), we believe it is appropriate for use in this study.
Patients
Overall, 75,335 new patient visits were potentially eligible. Of these, 62% of patients (46,966 of 75,335) were excluded because of missing PROMIS Global-10 physical and mental health scores. An additional 2.2% of patients (1685 of 75,335) were excluded because of at least one missing SDoH index at the time of their visit (for example, if a patient only had a Post Office box listed, the SDoH index could not be determined). This left 35% of the total eligible patients (26,684 of 75,335) in our final sample (Fig. 1). Although only 35% of possible new patient visits were included, the diversity of these individuals across numerous characteristics and the wide range of sociodemographic status—as measured by the SDoH indices—among included patients supports the generalizability of our sample.
Fig. 1.
This STROBE diagram outlines the path to our final patient sample.
Descriptive Data
Baseline patient characteristics of age, gender, self-reported race, self-reported ethnicity, language, and marital status were recorded. Self-reported race was dichotomized into White and non-White, while self-reported ethnicity was dichotomized into Hispanic and non-Hispanic. We believe this approach is sufficient, given that health disparities are less often present in majority groups (that is, self-reported White race and self-reported non-Hispanic ethnicity). For each patient, geographic data were used to determine the following SDoH indices: ADI, NSS, and SVI (Table 1). The ADI is determined at both the state and national level; for this study, we used the national ADI because the other geographic indices were not state-specific. Patients who had also completed the PROMIS Global-10 questionnaire as part of routine care were identified.
Table 1.
Variables included in each geographically based social determinants of health index
| SDoH index | Number of variables | Variable |
| Neighborhood Stress Score [3, 8] | 7 | % of families with incomes less than 100% of federal poverty level |
| % less than 200% of federal poverty level | ||
| % of adults who are unemployed | ||
| % of households receiving public assistance | ||
| % of households with no car | ||
| % of households with children and a single parent | ||
| % of people aged 25 years or older who have no high school degree | ||
| Area Deprivation Index [23] | 17 | % of the block group’s population aged 25 years or older with less than 9 years of education |
| % aged 25 years or older with at least a high school diploma | ||
| % of employed persons 16 years of age or older in white-collar occupations | ||
| Median family income | ||
| Income disparitya | ||
| Median home value | ||
| Median gross rent | ||
| Median monthly mortgage | ||
| % owner-occupied housing units (home ownership rate) | ||
| % of civilian labor force population 16 years or older unemployed (unemployment rate) | ||
| % of families below the federal poverty level | ||
| % of population below 150% of the federal poverty level | ||
| % of single-parent households with children younger than 18 years | ||
| % of occupied housing units without a motor vehicle | ||
| % of occupied housing units without a telephone | ||
| % of occupied housing units without complete plumbing (log) | ||
| % of occupied housing units with more than one person per room (crowding) | ||
| Social Vulnerability Index [1] | 16 | % below 150% of the federal poverty level |
| % unemployed | ||
| Housing cost burden | ||
| % no high school diploma | ||
| % no health insurance | ||
| % aged 65 years and older | ||
| % aged 17 years and younger | ||
| % civilian with a disability | ||
| % single-parent households | ||
| % English-language proficiency | ||
| % racial and ethnic minority statusb | ||
| % multiunit structures | ||
| % mobile homes | ||
| % crowding | ||
| % no vehicle | ||
| % group quarters |
Defined as the log of 100*ratio of number of households with less than USD 10,000 income to number of households with USD 50,000 or more income.
Hispanic or Latino (of any race); Black and African American, not Hispanic or Latino; American Indian and Alaska Native, not Hispanic or Latino; Asian, not Hispanic or Latino; Native Hawaiian and other Pacific Islander, not Hispanic or Latino; two or more races, not Hispanic or Latino; other races, not Hispanic or Latino. SDoH = social determinants of health.
Of the 26,684 patients included in our study, a slight majority were women (54% [14,366]) and 16% (4381) were of non-White race (Table 2). The mean age was 55 ± 18 years. The mean PROMIS Global-10 physical and mental health scores were 43.4 ± 9.4 and 49.7 ± 10.1, respectively. Given our data, distribution-based minimum clinically important difference estimates of one-half SD were 4.7 and 5.1 for PROMIS Global-10 physical and mental health scores, respectively. ADI and SVI values in our sample ranged from 1 to 100 (possible range 1 to 100) and from 0.0004 to 0.998 (possible range 0 to 1), respectively. The NSS ranged from -2.3 to 7.0 (possible range has no strict bounds and scales proportionally to a reference dataset). Given the comprehensive range of scores in our patient sample across each of the geographically based SDoH indices, we believe their correlations with each other can be assessed appropriately.
Table 2.
Comparison of baseline sociodemographic characteristics between included and excluded patients
| Patient sociodemographic characteristic | Excluded patients (n = 48,651) | Included patients (n = 26,684) | p value |
| Age in years | 55 ± 17 | 55 ± 18 | 0.02 |
| Women | 53 (25,545) | 54 (14,366) | < 0.001 |
| Non-White racea | 21 (10,078) | 16 (4381) | < 0.001 |
| Hispanic ethnicitya | 7 (3622) | 6 (1576) | < 0.001 |
| Non-English languageb | 8 (3674) | 4 (1141) | < 0.001 |
| Not married | 48 (23,488) | 47 (12,474) | < 0.001 |
Data are reported as mean ± SD or % (n).
Race and ethnicity were self-reported or selected by patients.
Primary language.
Primary and Secondary Study Outcomes
Our primary study goal was to assess the correlation of the NSS, ADI, and SVI—three different geographically based SDoH indices—with one another. We used our large institutional database that captures patient sociodemographic information as part of routine clinical care. Patient ZIP Codes are available in this database, which allowed us to determine individual geographically based SDoH index values. Correlations were then calculated.
Our secondary study goals were to assess the correlation of the three geographically based SDoH indices with PROMIS Global-10 physical and mental health scores and determine which geographically based SDoH indices are associated with presenting PROMIS Global-10 physical and mental health scores when accounting for common sociodemographic factors. The secondary study goals used the same SDoH index values as the primary study goal, as well as the PROMIS Global-10 physical and mental health scores, which were collected as part of routine care and stored in our institutional database. Correlations between the measures were then calculated. Additionally, we recorded patient sociodemographic factors from the electronic medical record to calculate associations between each SDoH index and PROMIS Global-10 physical and mental health. Six multivariable linear regression models were developed. Three models had PROMIS Global-10 physical health scores as the dependent variable, while three models had PROMIS Global-10 mental health scores as the dependent variable. Each geographically based SDoH index was included in a single PROMIS Global-10 physical health and single PROMIS Global-10 mental health multivariable model.
Ethical Approval
This study was approved by our institutional review board (protocol number: 2019P003521).
Statistical Analysis
We calculated the Spearman correlation coefficients of the SDoH indices. Pairwise Spearman correlation coefficients were also computed between the SDoH indices and PROMIS Global-10 physical and mental health scores. In addition, we computed an ordinary least squares regression separately between each of the PROMIS Global-10 physical and mental health scores and each SDoH index, accounting for age, gender, non-White self-reported race, Hispanic ethnicity, non-English language, and not married status as covariates. We confirmed the strength of association by calculating the partial r-squared for each SDoH index in each regression analysis. The partial r-squared captures the proportion of variation in the dependent variable (in our case, PROMIS Global-10 physical health score or PROMIS Global-10 mental health score) explained by the independent variable of interest (that is, each of the geographically based SDoH indices). Based on prior research, correlations were categorized as excellent (≥ 0.7), excellent-good (0.61 to 0.69), good (0.4 to 0.6), and poor (< 0.4) [4]. For all analyses, p < 0.05 was considered significant.
Results
Correlations Among Geographically Based SDoH Metrics
As the level of social deprivation, as measured by the ADI, grew more severe, neighborhood stress, as represented by the NSS, worsened. However, the strength with which the two measures moved in relationship to each other—which is the correlation—was poor (ρ = 0.34; p < 0.001). As the level of social deprivation, as measured by the ADI, grew more severe, social vulnerability, as measured by the SVI, worsened (ρ = 0.43; p < 0.001). Further, as the level of neighborhood stress, represented by the NSS, worsened, social vulnerability, as measured by the SVI, worsened as well (ρ = 0.59; p < 0.001). These relationships among geographically based SDoH indices demonstrate that the two measures moved in relation to each other, which defines correlation; however, although they were categorized as having good correlations based on cutoffs in prior research [5], the correlations are insufficient (that is, excellent correlations) to suggest the same construct is fully being captured by all indices.
Correlations Between Geographically Based SDoH Metrics and Physical and Mental Health
There was a poor correlation between the PROMIS Global-10 physical health and NSS (ρ = -0.14; p < 0.001), ADI (ρ = -0.24; p < 0.001), and SVI (ρ = -0.17; p < 0.001) (Table 3). There was a poor correlation between PROMIS Global-10 mental health and NSS (ρ = -0.13; p < 0.001), ADI (ρ = -0.22; p < 0.001), and SVI (ρ = -0.17; p < 0.001) (Table 3). In all analyses, as a geographically based SDoH index worsened (that is, the numeric value increased), physical and mental health, as measured by PROMIS Global-10 physical and mental health scores, also worsened (that is, the numeric value decreased); however, the strength with which these instruments moved with one another, which is correlation, was poor. The negative correlation coefficients captured the fact that one set of measures increased when it worsened, while the other decreased when it worsened.
Table 3.
Spearman rank order correlation coefficients between PROMIS Global-10 physical and mental health scores and SDoH indices
| SDoH index | PROMIS-10 physical health | PROMIS-10 mental health |
| NSS | -0.14 | -0.13 |
| ADI | -0.24 | -0.22 |
| SVI | -0.17 | -0.17 |
p < 0.001 for all correlation coefficients. PROMIS = Patient-Reported Outcomes Measurement Information System; SDoH = social determinants of health; NSS = Neighborhood Stress Score; ADI = Area Deprivation Index; SVI = Social Vulnerability Index.
Which Geographically Based SDoH Metrics Are Most Strongly Associated With Physical and Mental Health?
ADI demonstrated the largest association with presenting physical and mental health and was confirmed after the partial r-squared was calculated for each SDoH index (physical health: ADI 0.04 versus SVI 0.02 versus NSS 0.01; mental health: ADI 0.04 versus SVI 0.02 versus NSS 0.01). When accounting for key sociodemographic factors, the ADI was associated with presenting physical health (regression coefficient: -0.12 [95% CI -0.14 to -0.12]; p < 0.001) (Table 4) and mental health (regression coefficient: -0.13 [95% CI -0.14 to -0.12]; p < 0.001) (Table 5). The negative association demonstrated that with more social deprivation (that is, increasing ADI score), physical and mental health worsened (that is, decreasing PROMIS Global-10 physical and mental health scores). Given the minimum clinically important difference estimates for PROMIS Global-10 physical and mental health (4.7 and 5.1, respectively), an increase in ADI score by approximately 36 and 39 represents a clinically meaningful worsening of physical and mental health, respectively. SVI (physical health regression coefficient: -5.12 [95% CI -5.5 to -4.6]; p < 0.001; mental health regression coefficient: -5.3 [95% CI -5.8 to -4.8]; p < 0.001) and NSS (physical health regression coefficient: -1.1 [95% CI -1.2 to -1.0]; p < 0.001; mental health regression coefficient: -1.2 [95% CI -1.3 to -1.0]; p < 0.001) were also associated with presenting patient physical and mental health symptoms. Although the regression coefficients for the SVI and NSS seem to be more strongly associated with presenting physical and mental health symptoms than the ADI, this was caused by the unit (or scale) of each of the variables.
Table 4.
Ordinary least squares regression analyses for PROMIS Global-10 physical health and NSS, ADI, and SVI scores, accounting for sociodemographic factors
| NSS model | ADI model | SVI model | ||||
| Sociodemographic characteristic | Regression coefficient (95% CI) | p value | Regression coefficient (95% CI) | p value | Regression coefficient (95% CI) | p value |
| Regression constant | 50.2 (49.8 to 50.6) | < 0.001 | 52.6 (52.2 to 53.0) | < 0.001 | 51.8 (51.4 to 52.3) | < 0.001 |
| Age | -0.09 (-0.10 to -0.09) | < 0.001 | -0.09 (-0.10 to -0.08) | < 0.001 | -0.09 (-0.10 to -0.09) | < 0.001 |
| Women | -1.0 (-1.2 to -0.8) | < 0.001 | -1.1 (-1.3 to -0.9) | < 0.001 | -1.0 (-1.2 to -0.8) | < 0.001 |
| Non-White racea | -0.2 (-0.5 to 0.1) | 0.18 | -0.6 (-0.9 to -0.3) | < 0.001 | -0.1 (-0.4 to 0.2) | 0.54 |
| Hispanic ethnicitya | -1.8 (-2.3 to -1.2) | < 0.001 | -1.7 (-2.2 to -1.2) | < 0.001 | -1.5 (-2.1 to -1.0) | < 0.001 |
| Non-English languageb | -3.2 (-3.8 to -2.6) | < 0.001 | -3.3 (-3.9 to -2.8) | < 0.001 | -3.0 (-3.6 to -2.4) | < 0.001 |
| Not married | -2.1 (-2.4 to -1.9) | < 0.001 | -2.2 (-2.4 to -1.9) | < 0.001 | -2.0 (-2.2 to -1.8) | < 0.001 |
| NSS | -1.1 (-1.2 to -1.0) | < 0.001 | ||||
| ADI | -0.13 (-0.14 to -0.12) | < 0.001 | ||||
| SVI | -5.1 (-5.5 to -4.6) | < 0.001 | ||||
An r-squared value suggests how much variation in a dependent variable (here, PROMIS Global-10 physical health) is being captured by the independent variables in the analysis; above, the r-squared values for the NSS, ADI, and SVI models are 0.07, 0.09, and 0.07, respectively. For continuous variables (for example, age or any of the SDoH indices), the regression coefficient is the amount (and numeric direction) the PROMIS Global-10 physical health score changes with a one-unit increase in the variable. For example, for each year older in age in the ADI model, the PROMIS Global-10 physical health score decreases by 0.09. For dichotomous variables (for example, non-White race), the regression coefficient is the amount (and numeric direction) the PROMIS Global-10 physical health score changes if the patient is of the identified variable; if the patient is not of the identified variable, then there is no change in PROMIS Global-10 physical health score. For example, in the ADI model, if a patient is of non-White race, the PROMIS Global-10 physical health score decreases by 0.6 points but by no points if the patient is of White race.
aRace and ethnicity were self-reported or selected by patients.
bPrimary language. PROMIS = Patient-Reported Outcomes Measurement Information System; SDoH = social determinants of health; NSS = Neighborhood Stress Score; ADI = Area Deprivation Index; SVI = Social Vulnerability Index.
Table 5.
Ordinary least-squares regression analyses for PROMIS Global-10 mental health and NSS, ADI, and SVI scores, accounting for sociodemographic factors
| NSS model | ADI model | SVI model | ||||
| Sociodemographic characteristic | Regression coefficient (95% CI) | p value | Regression coefficient (95% CI) | p value | Regression coefficient (95% CI) | p value |
| Regression constant | 54.7 (54.3 to 55.2) | < 0.001 | 57.2 (56.7 to 57.7) | < 0.001 | 56.4 (55.9 to 56.9) | < 0.001 |
| Age | -0.06 (-0.07 to -0.05) | < 0.001 | -0.06 (-0.07 to -0.05) | < 0.001 | -0.06 (-0.07 to -0.05) | < 0.001 |
| Women | -0.4 (-0.7 to -0.2) | 0.001 | -0.5 (-0.7 to -0.2) | < 0.001 | -0.4 (-0.6 to -0.2) | 0.001 |
| Non-White race | -0.3 (-0.6 to 0.1) | 0.12 | -0.7 (-1.0 to -0.3) | < 0.001 | -0.2 (-0.5 to 0.2) | 0.39 |
| Hispanic ethnicity | -1.0 (-1.6 to -0.5) | < 0.001 | -1.0 (-1.5 to -0.4) | 0.001 | -0.8 (-1.3 to -0.2) | 0.008 |
| Non-English language | -3.3 (-3.9 to -2.6) | < 0.001 | -3.4 (-4.0 to -2.8) | < 0.001 | -3.1 (-3.7 to -2.4) | < 0.001 |
| Nonmarried | -2.9 (-3.2 to -2.7) | < 0.001 | -2.9 (-3.2 to -2.7) | < 0.001 | -2.8 (-3.0 to -2.5) | < 0.001 |
| NSS | -1.2 (-1.3 to -1.0) | < 0.001 | ||||
| ADI | -0.13 (-0.14 to -0.12) | < 0.001 | ||||
| SVI | -5.3 (-5.8 to -4.8) | < 0.001 | ||||
An r-squared value suggests how much variation in a dependent variable (here, PROMIS Global-10 mental health) is being captured by the independent variables in the analysis; above, the r-squared values for the NSS, ADI, and SVI models are 0.05, 0.08, and 0.06, respectively. For continuous variables (for example, age or any of the SDoH indices), the regression coefficient is the amount (and numeric direction) the PROMIS Global-10 mental health score changes with a unit increase in the variable. For example, for each year older in age in the ADI model, the PROMIS Global-10 mental health score decreases by 0.06. For dichotomous variables (for example, non-White race), the regression coefficient is the amount (and numeric direction) the PROMIS Global-10 mental health score changes if the patient is of the identified variable; if the patient is not of the identified variable, then there is no change in PROMIS Global-10 mental health score. For example, in the ADI model, if a patient is of non-White race, the PROMIS Global-10 mental health score decreases by 0.7 points but by no points if the patient is of White race.
aRace and ethnicity were self-reported or selected by patients.
bPrimary language. PROMIS = Patient-Reported Outcomes Measurement Information System; SDoH = social determinants of health; NSS = Neighborhood Stress Score; ADI = Area Deprivation Index; SVI = Social Vulnerability Index.
Discussion
Health disparities are widely recognized in orthopaedic surgery [16], with prior research demonstrating an association between SDoH factors and presenting symptom severity [5, 6] and clinical outcomes [12]. Instead of assessing each SDoH factor (such as self-reported race or education level) individually, researchers and policy makers have turned to geographically based SDoH indices that incorporate several SDoH characteristics into a single score to conduct widescale analyses to assess initiatives to address health disparities more efficiently across a patient population and better risk-adjust payments. Commonly used geographically based SDoH indices include the ADI, NSS, and SVI. However, the reason for using one SDoH index over another is neither well outlined nor much discussed. Although they are all geographically derived using United States Census tract data, there are slight variations in how they are calculated, and it remains unclear how well they are correlated. This is important to consider because comparing initiatives and their outcomes may not be appropriate if they use different measures as “inputs” that are not sufficiently similar enough. Further, in orthopaedic surgery, it would be beneficial to know which geographically based SDoH index is more strongly associated with physical function and pain, because this would suggest that such a measure may be more beneficial. We found that physical and mental health were poorly correlated with SDoH indices, but ADI, SVI, and NSS were associated with PROMIS scores after accounting for individual patient sociodemographic characteristics. The ADI was more strongly associated with PROMIS scores than the SVI or NSS. Based on our discoveries, the preferred geographically based SDoH index for orthopaedic research is the ADI.
Limitations
First, only 35% (26,684 of 75,335) of possible new patient encounters were ultimately included in the final study cohort. There were some differences in patient characteristics between the excluded and included patients, with more patients of non-White race, Hispanic ethnicity, and those who speak a non-English language primarily missing complete data. It is also possible that patients with housing instability were excluded because the SDoH indices could not be calculated without recorded home addresses. Nonetheless, we believe this would not alter our core findings that the three different SDoH indices are not concurrently valid because the formulas are inherently different. We also do not believe it would alter our findings that health disparities exist regarding presenting symptom severity, although the degree to which this is true by SDoH index may slightly change. Future research would be warranted to assess this further, although we suspect our findings are generalizable to similar patient populations. Importantly, our included data captured a wide range of sociodemographic statuses, as measured by the three indices used, which further supports the generalizability of our findings across broad populations. In addition, because this study primarily focused on concurrent validity and the relationship between different geographically based SDoH indices and patient symptoms at presentation and not clinical or treatment outcomes, we do not believe that missing data (or data not analyzed, such as comorbidities) have a major effect on our key findings. Further, we controlled for these patient characteristics as part of our regression analyses. However, if our findings were impacted, the present study’s core results would likely be a lower bound estimate (or even underestimate) of the health disparities that currently exist in the relationship between geographically based SDoH indices and presenting physical and mental health. It is also possible that other patient factors such as health literacy were not captured and could be associated with completion of PROMs and inclusion in the present study; however, health literacy is likely correlated with education level [24], which is captured in geographically based SDoH measures.
Second, we dichotomized self-reported race and self-reported ethnicity. Unfortunately, health disparities are more often present in minority groups overall. The causes of health disparities among different racial and ethnic patient groups may differ; however, because this study did not evaluate targeted interventions in addressing health disparities, we believe such “buckets” are acceptable.
Third, these patient data were from a single Level I trauma academic medical center. As such, our findings may not be generalizable to other geographic regions or healthcare settings. However, our patient sample is large and diverse, and we believe our findings are most likely generalizable to similar clinic sites affiliated with tertiary referral academic medical centers with patients who have a wide range of sociodemographic characteristics. In addition, the geographically based SDoH indices we analyzed are designed for use across the United States and not only in our state or region.
Fourth, there is concern that the localized sample assessing the correlation among the three geographically based SDoH indices is not sufficiently robust to form a definitive conclusion, given that the scores are not individual-level characteristics but area-level characteristics. However, the comprehensive range of scores across all three indices suggests our sample captures patients of all backgrounds and provides a “pilot” type of analysis on which future research in this area can build. It also introduces this issue to practicing orthopaedic surgeons who otherwise may not be aware of it.
Lastly, we used the PROMIS Global-10 to assess functional and mental health among our new-patient sample. Although the PROMIS Global-10 is a commonly used and validated general PROM [10], the relationship between SDoH indices and more condition-specific or anatomy-specific PROMs in different patient subgroups may differ. Although some of these instruments are strongly correlated with the PROMIS Global-10, others are not as strongly associated [18, 20, 22]. We suspect that our findings may be similar to other general PROMs; however, similar research may be warranted with condition-specific or anatomy-specific PROMs.
Discussion of Key Findings
The lack of strong correlation among the geographically based SDoH indices likely arises from the different ways each index is calculated. For example, the ADI incorporates up to 17 factors that include aspects of income (median family income and income disparity), education, employment, and housing (not only home value and rent variables but also detailed characteristics, such as units without a motor vehicle, telephone, or complete plumbing) at the census block group, or “neighborhood,” level [13, 14, 19, 23]. Although the NSS also uses variables at the census block group level and incorporates some of the general factors incorporated into the ADI, it does so to a lesser extent, with only seven variables included in the calculation without as many variables [3, 8]. Lastly, the SVI, a measure developed by the Centers for Disease Control at the census tract level, assesses 16 social factors focused on poverty, vehicle access, and housing, grouping them into four themes: socioeconomic status; household composition; race, ethnicity, and language; and housing and transportation [1]. Locations are assigned rankings in each theme, as well as a composite overall ranking. Although each index was robustly created, our findings demonstrate they do not measure SDoH status in exactly the same way. This may be expected if one assesses how each is calculated; however, we suspect this is not well known in the orthopaedic surgery community.
It was not surprising that SDoH status was not strongly correlated with presenting physical function or mental health. This is because we assumed different constructs were being measured. However, we feel it is important to officially determine this to be the case because if the correlations were strong, one could argue that only PROMIS Global-10 physical and mental health questionnaires could be collected and tracked. This would reduce unnecessary data collection. Although there may be a relationship between SDoH status—regardless of the index being used—and symptom state, many other factors are also likely associated with function and mental health. Future research can determine whether other PROMs better capture SDoH status and physical function and mental health simultaneously; however, if so, it is possible that it captures neither sufficiently. A thorough analysis to ensure adequate performance would be needed.
Given the above finding that supports the idea that the three geographically based SDoH indices should not be interchanged freely, nor can PROMIS Global-10 physical and mental health scores capture SDoH factors adequately, it is most important to try to discern which index may be best to use in patients undergoing orthopaedic surgery.
Although all three geographically based SDoH indices were associated with presenting physical and mental health across a new orthopaedic patient sample, the ADI demonstrated the strongest such association. In addition, a reasonable change in ADI is associated with a clinical difference in presenting physical and mental health symptoms. In contrast, given the unit (or scale) of the SVI and NSS, it is highly unlikely that a patient could have a change in SDoH status large enough to be associated with a clinically appreciable change in presenting physical and mental health symptoms. For example, possible SVI values range from 0 to 1, and it would take nearly a change of that magnitude to be associated with PROMIS scores that are clinically relevant, as measured by the minimum clinically important difference. Prior research has suggested that the national ADI and self-reported insurance type captures the most variability because of social deprivation in orthopaedic patients’ self-reported functional and behavioral health well-being [7]. However, that study did not assess the SVI or NSS. Nonetheless, given that study’s findings and the results of our study, which demonstrated a stronger association between ADI and physical and mental health among new orthopaedic patients, we recommend using the ADI as the geographically based SDoH index in orthopaedic surgery. Additionally, the ADI is more easily calculated than the SVI or NSS via an interactive map on the internet [23], which we believe makes it more accessible to researchers and policymakers. Further inquiry is warranted into how individual component census variables contribute to the associations between these indices and patient health and outcomes.
Conclusion
Overall, this study demonstrates that geographically based SDoH indices do not appear to measure a consistent SDoH construct and are associated with presenting patient symptoms to varying degrees when accounting for common patient factors. Although robustly developed, the ADI incorporates the most SDoH factors (that is, 17) across the three indices evaluated in its calculation formula, is easily accessible online, and it demonstrates the strongest association with physical function and mental health symptoms in patients undergoing orthopaedic surgery. Considering these findings, we recommend the ADI as the geographically based SDoH metric in orthopaedic surgery research unless data limitations require the use of another. Ultimately, standardization of the choice and use of geographically based SDoH indices—globally or by country—will enable more efficient, patient-centered policy to address health disparities in orthopaedic care and throughout medicine. It is now important to use the ADI in initiatives to reduce health disparities and study the outcomes of such work to ensure that progress in promoting health equity is truly being made.
Footnotes
One of the authors (DNB) certifies receipt of personal payments or benefits, during the study period; in an amount of less than USD 10,000 from the Institute For Strategy And Competitiveness at Harvard Business School; in an amount of less than USD 10,000 from the National Academy of Medicine; in an amount of less than USD 10,000 from The Heritage Foundation; and in an amount of less than USD 10,000 from the PROMIS Health Organization.
All ICMJE Conflict of Interest Forms for authors and Clinical Orthopaedics and Related Research® editors and board members are on file with the publication and can be viewed on request.
Ethical approval for this study was obtained from Massachusetts General Hospital (IRB Protocol number: 2019P003521).
This work was performed at Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA.
Contributor Information
David N. Bernstein, Email: dtobert@mgh.harvard.edu.
David Shin, Email: david.shin@outlook.com.
Rudolf W. Poolman, Email: namloop@gmail.com.
Joseph H. Schwab, Email: jhschwab@mgh.harvard.edu.
Daniel G. Tobert, Email: dtobert@partners.org.
References
- 1.Agency for Toxic Substances and Disease Registry. CDC/ATSR/SVI fact sheet. United States Centers for Disease Control and Prevention. Available at: https://www.atsdr.cdc.gov/placeandhealth/svi/fact_sheet/fact_sheet.html. Accessed August 14, 2022.
- 2.Amen TB, Varady NH, Shannon EM, Chopra A, Rajaee S, Chen AF. Racial and ethnic disparities in hip fracture surgery care in the United States from 2006 to 2015: a nationwide trends study. J Am Acad Orthop Surg. 2022;30:e182-e190. [DOI] [PubMed] [Google Scholar]
- 3.Ash AS, Mick E. UMass risk adjustment project for MassHealth payment and care delivery reform: describing the 2017 payment model. Available at: https://www.mass.gov/files/documents/2017/11/03/1610-umass-modeling-sdh-summary-report.pdf. Accessed July 26, 2022.
- 4.Bernstein DN, Houck JR, Mahmood B, Hammert WC. Responsiveness of the PROMIS and its concurrent validity with other region- and condition-specific PROMs in patients undergoing carpal tunnel release. Clin Orthop Relat Res. 2019;477:2544-2551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Bernstein DN, Kurucan E, Fear K, Hammert WC. Evaluating the impact of patient social deprivation on the level of symptom severity at carpal tunnel syndrome presentation. Hand (N Y). 2022;17:339-345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bernstein DN, Merchan N, Fear K, Rubery PT, Mesfin A. Greater socioeconomic disadvantage is associated with worse symptom severity at initial presentation in patients seeking care for lumbar disc herniation. Spine (Phila Pa 1976). 2021;46:464-471. [DOI] [PubMed] [Google Scholar]
- 7.Cheng AL, McDuffie JV, Schuelke MJ, Calfee RP, Prather H. How should we measure social deprivation in orthopaedic patients? Clin Orthop Relat Res. 2022;480:325-339. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Commonwealth of Massachusetts. FAQs for MassHealth’s 2017 payment model. Available at: https://www.mass.gov/doc/social-determinants-of-health-sdh-faq-1/download. Accessed March 25, 2023.
- 9.Delanois RE, Tarazi JM, Wilkie WA, et al. Social determinants of health in total knee arthroplasty: are social factors associated with increased 30-day post-discharge cost of care and length of stay? Bone Joint J. 2021;103:113-118. [DOI] [PubMed] [Google Scholar]
- 10.Hays RD, Bjorner JB, Revicki DA, Spritzer KL, Cella D. Development of physical and mental health summary scores from the Patient-Reported Outcomes Measurement Information System (PROMIS) global items. Qual Life Res. 2009;18:873-880. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Huffstetler AN, Phillips RL, Jr. Payment structures that support social care integration with clinical care: Social deprivation indices and novel payment models. Am J Prev Med. 2019;57:S82-S88. [DOI] [PubMed] [Google Scholar]
- 12.Kamalapathy PN, Dunne PJ, Yarboro S. National evaluation of social determinants of health in orthopedic fracture care: decreased social determinants of health is associated with increased adverse complications after surgery. J Orthop Trauma. 2022;36:e278-e282. [DOI] [PubMed] [Google Scholar]
- 13.Kind AJ, Jencks S, Brock J, et al. Neighborhood socioeconomic disadvantage and 30-day rehospitalization: a retrospective cohort study. Ann Intern Med. 2014;161:765-774. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Knighton AJ, Savitz L, Belnap T, Stephenson B, VanDerslice J. Introduction of an area deprivation index measuring patient socioeconomic status in an integrated health system: Implications for population health. EGEMS (Wash DC) . 2016;4:1238. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.National Academies of Sciences, Engineering, and Medicine, Health and Medicine Division, Board on Population Health and Public Health Practice, Committee of Community-Based Solutions to Promote Health Equity in the United States. In: Baciu A, Negussie Y, Geller A, Weinstein JN, eds. Communities in Action: Pathways to Health Equity. National Academies Press; 2017. [PubMed] [Google Scholar]
- 16.Nelson CL. Disparities in orthopaedic surgical intervention. J Am Acad Orthop Surg. 2007;15 Suppl 1:S13-17. [DOI] [PubMed] [Google Scholar]
- 17.Richards SE. Who will get the vaccine first? Here’s where you might land in line. Available at: https://www.nationalgeographic.com/science/article/who-is-first-in-line-coronavirus-vaccine. Accessed August 15, 2022.
- 18.Saad MA, Kassam HF, Suriani RJ, Jr, et al. Performance of PROMIS global-10 compared with legacy instruments in patients with shoulder arthritis. J Shoulder Elbow Surg. 2018;27:2249-2256. [DOI] [PubMed] [Google Scholar]
- 19.Singh GK. Area deprivation and widening inequalities in us mortality, 1969-1998. Am J Public Health. 2003;93:1137-1143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Stoop N, Menendez ME, Mellema JJ, Ring D. The PROMIS global health questionnaire correlates with the QuickDASH in patients with upper extremity illness. Hand (N Y). 2018;13:118-121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Suleiman LI, Tucker K, Ihekweazu U, Huddleston JI, Cohen-Rosenblum AR. Caring for diverse and high-risk patients: surgeon, health system, and patient integration. J Arthroplasty . 2022;37:1421-1425. [DOI] [PubMed] [Google Scholar]
- 22.Suriani RJ, Kassam HF, Passarelli NR, Esparza R, Kovacevic D. Validation of PROMIS global-10 compared with legacy instruments in patients with shoulder instability. Shoulder Elbow. 2020;12:243-252. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.University of Wisconsin School of Medicine and Public Health Center for Health Disparities Research. Neighborhood atlas. Available at: https://www.neighborhoodatlas.medicine.wisc.edu/. Accessed March 25, 2023.
- 24.van der Heide I, Wang J, Droomers M, Spreeuwenberg P, Rademakers J, Uiters E. The relationship between health, education, and health literacy: results from the Dutch Adult Literacy and Life Skills Survey. J Health Commun. 2013;18:172-184. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Varady NH, d'Amonville S, Chen AF. Electronic patient portal use in orthopaedic surgery is associated with disparities, improved satisfaction, and lower no-show rates. J Bone Joint Surg Am. 2020;102:1336-1343. [DOI] [PubMed] [Google Scholar]
- 26.World Health Organization. Social determinants of health. Available at: https://www.who.int/teams/social-determinants-of-health. Accessed July 26, 2022.
- 27.Ziedas A, Abed V, Swantek A, et al. Social determinants of health influence access to care and outcomes in patients undergoing anterior cruciate ligament reconstruction: a systematic review. Arthroscopy. 2022;38:583-594.e584. [DOI] [PubMed] [Google Scholar]