Social determinants of health (SDoH) may profoundly influence individual health outcomes.1 These factors are intrinsically linked to the circumstances in which people are born, grow, work, live, and age.2 These determinants encompass various critical elements such as income, social protection, education, employment, working conditions, food security, housing, stress levels, childhood development, social support, and access to health care.3 Interindividual perception of pain is an important characteristic that may contribute to the expectations and outcomes of patients4,5. It is essential to investigate the association between SDoH factors and daily pain experiences to gain a deeper understanding of how various components of SDoH may be related to reported pain. To address this gap, the objective of this study was to explore the association between various patient-reported SDoH measures and pain scores using the diverse database from the All of Us Research Program.6
METHODS
Consent requirements for this retrospective observational study were waived by the internal review board (Human Research Protections Program) due to the study’s retrospective nature using deidentified data. The All of Us dataset is a national registry consisting of approximately 1 million participants with a study population consisting of a significant proportion of participants from traditionally under-represented social groups. All participants with available survey data related to SDoH were analyzed from the seventh version of the National Institute of Health All of Us Research Program database.6 In addition, participants reported their average pain score experienced daily over the last 7 days (based on the Numeric Rating Scale [NRS]).7
| “How often are you treated with less respect than other people when you go to a doctor’s office or other health care provider?” | “Do you speak a language other than English at home?” | “The crime rate in my neighborhood makes it unsafe to go on walks during the day. Would you say that you…” | “Within the past 12 mo, were you worried whether the food you had bought just didn’t last and you didn’t have money to get more?” | “Within the past 12 mo, were you worried whether your food would run out before you got money to buy more?” | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Coefficient (95% CI) | P value | Coefficient (95% CI) | P value | Coefficient (95% CI) | P value | Coefficient (95% CI) | P value | Coefficient (95% CI) | P value | |
| (intercept) | 1.26 (1.16–1.37) | <.001 | 1.47 (1.35–1.58) | <.001 | 1.29 (1.18–1.4) | <.001 | 1.05 (0.94–1.15) | <.001 | 1.02 (0.92–1.13) | <.001 |
| Response to questiona | 0.14 (0.13–0.15) | <.001 | 0.08 (0.04–0.12) | <.001 | 0.15 (0.13–0.16) | <.001 | 0.33 (0.31–0.35) | <.001 | 0.32 (0.3–0.34) | <.001 |
| Body mass index (kg/m2) | 0 (0–0) | .75 | 0 (0–0) | .84 | 0 (0–0) | .76 | 0 (0–0) | .75 | 0 (0–0) | .75 |
| Age (y) | 0 (0–0) | <.001 | 0 (−0.01 to 0) | <.001 | 0 (0–0) | <.001 | 0 (0–0) | <.001 | 0 (0–0) | <.001 |
| Female | 0.12 (0.09–0.14) | <.001 | 0.12 (0.1–0.15) | <.001 | 0.13 (0.1–0.15) | <.001 | 0.11 (0.09–0.13) | <.001 | 0.11 (0.09–0.13) | <.001 |
| Sex (other) | 0.08 (0.01–0.15) | .03 | 0.08 (0.01–0.15) | .04 | 0.09 (0.01–0.16) | .03 | 0.05 (−0.02 to 0.12) | .16 | 0.06 (−0.02 to 0.13) | .13 |
| History of opioid dependence | 0.46 (0.41–0.51) | <.001 | 0.49 (0.44–0.54) | <.001 | 0.46 (0.41–0.52) | <.001 | 0.41 (0.36–0.46) | <.001 | 0.4 (0.35–0.45) | <.001 |
| Race—Asian | −0.42 (−0.54, −0.29) | <.001 | −0.43 (−0.55, −0.3) | <.001 | −0.32 (−0.45, −0.19) | <.001 | −0.32 (−0.45, −0.2) | <.001 | −0.33 (−0.46, −0.21) | <.001 |
| Race—Black or African American | 0.12 (0.03–0.21) | .01 | 0.19 (0.09–0.28) | <.001 | 0.17 (0.07–0.26) | <.001 | 0.13 (0.05–0.22) | .00 | 0.14 (0.05–0.23) | .00 |
| Race—Mixed | −0.12 (−0.23, −0.01) | .03 | −0.06 (−0.18 to 0.05) | .29 | −0.04 (−0.16 to 0.08) | .52 | −0.05 (−0.16 to 0.05) | .32 | −0.06 (−0.17 to 0.05) | .28 |
| Raced—MENA | −0.19 (−0.37, −0.01) | .04 | −0.2 (−0.38, −0.01) | .04 | −0.08 (−0.28 to 0.11) | .40 | −0.17 (−0.34 to 0.01) | .07 | −0.19 (−0.37, −0.01) | .04 |
| Race—NHPI | 0.26 (−0.15 to 0.68) | .22 | 0.3 (−0.12 to 0.73) | .17 | 0.23 (−0.21 to 0.68) | .32 | 0.21 (−0.2 to 0.62) | .32 | 0.27 (−0.13 to 0.68) | .20 |
| Race—White | −0.21 (−0.3, −0.13) | <.001 | −0.18 (−0.27, −0.09) | <.001 | −0.13 (−0.22, −0.04) | .01 | −0.16 (−0.25, −0.08) | <.001 | −0.17 (−0.25, −0.08) | <.001 |
| Ethnicity—Hispanic | 0.12 (0.04–0.19) | .00 | 0.09 (0.01–0.17) | .03 | 0.11 (0.03–0.19) | .01 | 0.08 (0–0.16) | .04 | 0.08 (0–0.15) | .04 |
| “In the last month, how often have you felt that things were going your way?” | “In the last month, how often have you felt that you were on top of things?” | “How much you agree or disagree that people in your neighborhood can be trusted?” | “How often do you feel that you can find companionship when you want it?” | “How much you agree or disagree that your neighborhood is safe?” | ||||||
| Coefficient (95% CI) | P value | Coefficient (95% CI) | P value | Coefficient (95% CI) | P value | Coefficient (95% CI) | P value | Coefficient (95% CI) | P value | |
| (intercept) | 1.9 (1.8–2.01) | <.001 | 1.91 (1.8–2.01) | <.001 | 1.93 (1.82–2.03) | <.001 | 2.01 (1.9–2.12) | <.001 | 1.27 (1.15–1.39) | <.001 |
| Response to questiona | −0.15 (−0.16, −0.14) | <.001 | −0.14 (−0.15, −0.13) | <.001 | −0.12 (−0.13, −0.11) | <.001 | −0.15 (−0.16, −0.13) | <.001 | −0.11 (−0.12, −0.1) | <.001 |
| Body mass index (kg/m2) | 0 (0–0) | 0.56 | 0 (0–0) | 0.57 | 0 (0–0) | 0.86 | 0 (0–0) | 0.80 | 0.02 (0.02–0.02) | <.001 |
| Age (y) | 0 (0–0) | <.001 | 0 (0–0) | <.001 | 0 (0–0) | <.001 | 0 (0–0) | <.001 | 0 (0–0) | <.001 |
| Female | 0.13 (0.1–0.15) | <.001 | 0.11 (0.08–0.13) | <.001 | 0.12 (0.1–0.15) | <.001 | 0.14 (0.11–0.16) | <.001 | 0.11 (0.09–0.13) | <.001 |
| Sex (other) | 0.06 (−0.02 to 0.13) | .13 | 0.08 (0–0.15) | .04 | 0.08 (0.01–0.15) | .03 | 0.07 (−0.01 to 0.14) | .07 | 0.05 (−0.02 to 0.12) | .15 |
| History of opioid dependence | 0.45 (0.4–0.5) | <.001 | 0.45 (0.4–0.5) | <.001 | 0.46 (0.41–0.51) | <.001 | 0.46 (0.41–0.51) | <.001 | 0.48 (0.43–0.52) | <.001 |
| Race—Asian | −0.38 (−0.51, −0.25) | <.001 | −0.38 (−0.51, −0.26) | <.001 | −0.38 (−0.5, −0.25) | <.001 | −0.4 (−0.53, −0.28) | <.001 | −0.27 (−0.4, −0.15) | <.001 |
| Race—Black or African American | 0.15 (0.06–0.25) | .00 | 0.17 (0.08–0.27) | <.001 | 0.13 (0.04–0.22) | 0.01 | 0.14 (0.05–0.23) | .00 | 0.11 (0.02–0.2) | .02 |
| Race—Mixed | −0.08 (−0.19 to 0.04) | .18 | −0.07 (−0.18 to 0.04) | .23 | −0.1 (−0.21 to 0.01) | .08 | −0.1 (−0.21 to 0.02) | .09 | −0.08 (−0.19 to 0.03) | .16 |
| Raced—MENA | −0.21 (−0.39, −0.02) | .03 | −0.15 (−0.33 to 0.03) | .10 | −0.15 (−0.33 to 0.03) | .09 | −0.22 (−0.4, −0.04) | .02 | −0.14 (−0.31 to 0.04) | .13 |
| Race—NHPI | 0.22 (−0.21 to 0.67) | .31 | 0.25 (−0.15 to 0.67) | .23 | 0.29 (−0.12 to 0.72) | .17 | 0.23 (−0.18 to 0.66) | .27 | 0.24 (−0.16 to 0.66) | .24 |
| Race—White | −0.19 (−0.28, −0.11) | <.001 | −0.19 (−0.28, −0.1) | <.001 | −0.19 (−0.27, −0.1) | <.001 | −0.2 (−0.29, −0.12) | <.001 | −0.16 (−0.24, −0.07) | <.001 |
| Ethnicity—Hispanic | 0.11 (0.03–0.19) | .01 | 0.12 (0.04–0.19) | .00 | 0.1 (0.02–0.18) | 0.01 | 0.1 (0.03–0.18) | .01 | 0.1 (0.03–0.18) | .01 |
The study population included all participants diagnosed with a history of chronic pain and answered the Overall Health Survey question related to pain score in the past 7 days (“In the past 7 days, how would you rate your pain on average?”). SDoH was assessed by 81 questions from the SDoH Survey. We excluded 3 questions because their respective responses were not ordinal in nature. Therefore, only questions with ordinal responses (eg, completely disagree to completely agree) were analyzed and responses were converted to Likert-scale responses (1–3 or 1–4 depending on number of responses). Any participant who did not fill out the SDoH survey and did not report pain scores were excluded.
Statistical Analysis
Analyses were conducted on the All of Us data release version 7 using the secure Researcher Workbench platform and Python. The primary outcome of interest was determining the association between responses to SDoH questions to reported pain scores. The primary dependent variable (pain score on the 0–10 NRS) followed a nonparametric distribution based on the Kolmogorov-Smirnov test, thus we used negative binomial regression as our statistical model. First, we performed a univariate analysis separately for each of the 78 SDoH questions and reported the coefficient, 95% confidence interval (CI), and P value. Then, we performed a multivariable regression modeling the questions that had the top 5 positive and 5 negative associations with pain. We included the following potential confounders in the model as they have been previously reported to be associated with pain scores: age, sex, body mass index, race, ethnicity, and history of opioid dependence. Given we performed 10 separate multivariable regression models, a value of P < .005 was considered statistically significant (based on Bonferroni correction) and a coefficient with absolute value ≥0.1 was considered clinically significant. For the multivariable models, we calculated the variance inflation factor for each independent variance to assess for collinearity.
RESULTS
There were 82,661 patients identified to have chronic pain. After exclusion, the study sample size was 23,309, which includes all patients who reported pain scores and completed the SDoH questions. The median [quartiles] NRS for pain was 3.0 [1–6] (Supplemental Digital Content 1, Supplementary Table 1, https://links.lww.com/AA/F8). First, we performed a univariate negative binomial regression for each of the 78 SDoH questions to model their association with pain scores (Supplemental Digital Content 2, Supplementary Table 2, https://links.lww.com/AA/F9). The Table lists the questions with the top 5 positive and top 5 negative associations with pain (based on univariate analysis) when modeled via a multivariable negative binomial regression model controlling for age, body mass index, opioid dependence, race, and ethnicity. The variance inflation factor for each independent variable for each model was <0.4 (Supplemental Digital Content 3, Supplementary Table 3, https://links.lww.com/AA/F10).
DISCUSSION
In this retrospective observational study, the All Of Us Research Program dataset was leveraged to measure the associations of various SDoH measures with a patient’s reported daily pain scores. Questions related to food insecurity, stress levels, social support, and neighborhood safety were found to be mostly associated with pain scores. Specifically, higher concerns for food insecurity were associated with higher pain scores in daily life. This is consistent with a previous study reporting that food insecurity may be a social determinant of chronic pain.8 Unique to our study was our analysis measuring SDoH associations with reported pain scores among chronic pain patients. Furthermore, lower concerns with daily stress levels were associated with lower pain scores in daily life. In patients with chronic pain, stress may be associated with worsening of pain symptoms.9 This highlights the importance of understanding SDoH and how it may be associated with patient outcomes, specifically with pain outcomes. Major limitations of the study include bias related to exclusion of patients with missing data and uncertainly related to severity of chronic pain. Thus, further studies should aim to validate these results while limiting such bias. Addressing these social determinants is crucial for improving overall population health outcomes and reducing health inequalities.