Discussion

In this single center study of nearly 41,000 patients who underwent select surgeries at an academic quaternary referral center in the Deep South, a high SVI is associated with a 50% increased risk of SSI. This association was present but not significant for more severe SSIs (i.e., deep incisional or organ space) and for patients with higher socioeconomic status theme as well as vulnerability of housing and transportation theme. To our knowledge, this is the largest study to examine the relationship between NHSN-defined SSIs and SVI across multiple procedure types. The increased risk of SSI remained after statistical adjustment with an increasing association gradient from low to high SVI. There is similar risk when categorized by depth of infection, although not statistically significant, likely due to low overall numbers.

The findings of this study affirm the significant association between SVI and SSI, and our results extend prior published work. DeWitt et al reported similar odds of SSI in the highest SVI quartile, with the strongest associations in colon surgery and spinal fusion. ^{Reference Dewitt, Reinke and Inman23} In exploratory analyses incorporating an interaction between procedure type and SVI tertile in the current models, we similarly observed stronger, although not statistically significant, associations for colon procedures (HR 1.68, 95% CI 0.78–3.61) and spinal fusion (HR 1.62, 95% CI 0.79–3.34). The largest effect estimate occurred among abdominal hysterectomy procedures (HR 3.93, 95% CI 0.54–28.55), though the number of SSIs among hysterectomy procedures was relatively few, resulting in wide confidence limits. Other studies have associated higher SVI with adverse postoperative outcomes, including morbidity, mortality, and infectious complications across smaller or demographically different populations, although findings specific to infection have varied and methodological differences limit direct comparison with our results. ^{Reference Diaz, O’Reggio, Norman, Thumma, Dimick and Ibrahim14,Reference Dyas, Carmichael and Bronsert16,Reference Carmichael, Dyas and Bronsert17,Reference Rahman, Shaw and Mehaidli24} Our findings add more direct evidence that social vulnerability is associated with SSI across diverse surgical populations.

Patients in the high-SVI tertile were younger, had more medical comorbidities, and more likely to identify as Black, which is consistent to other surgical populations. ^{Reference Diaz, Hyer, Barmash, Azap, Paredes and Pawlik6,Reference Dyas, Carmichael and Bronsert16} High-SVI patients were more likely to live farther from the hospital and have Medicaid or self-pay status than low-SVI patients. Medicaid beneficiaries are disproportionately concentrated in high SVI areas and are less likely to have a primary care visit. ^{Reference Herring, Park, Luo, Vichare, Erikson and Pittman25} If access to healthcare for primary care is minimal in patients with high-SVI, follow up after complex surgical procedure is likely also underutilized. Interestingly, we observed more trauma and emergent surgical procedures in high SVI tertile. These findings are consistent with other studies that have highlighted socially vulnerable communities experience higher trauma incidence and more emergent procedures. ^{Reference Hey, Alty and Uribe Leitz26–Reference Carmichael, Moore, Steward and Velopulos29} This data suggests that area-level vulnerability is tightly linked to access to care, timely care, and perioperative risk profiles. The persistence of an SVI-SSI association even after adjustment suggests additional unmeasured social and structural factors, such as barriers to transportation, home environment, and health literacy, among others.

Although ADI has been associated with adverse postoperative outcomes, it differs conceptually and methodologically from SVI; findings based on the two measures should not be considered interchangeable. ^{Reference Sun, Gross and Hamo30–Reference Morenz, Liao, Au and Hayes34} We selected SVI because its theme-level structure can identify specific dimensions of community vulnerability that may inform interventions. ^{Reference Rahman, Shaw and Mehaidli24,Reference Morenz, Liao, Au and Hayes34} In our study, the socioeconomic status and vulnerability of housing type and transportation themes are most strongly associated with SSI (Figure 1), suggesting potential contributions from factors such as poverty, lack of insurance coverage, crowded housing, and inadequate transportation. The limited prior literature examining SSI by SVI theme has produced variable findings. Rahman et al identified the strongest association in the themes of socioeconomic status and race/ethnic minority status, whereas Dewitt et al identified increased SSI risk with socioeconomic status and household characteristics. ^{Reference Dewitt, Reinke and Inman23,Reference Rahman, Shaw and Mehaidli24} These subthemes differences highlight the unique community needs that may be associated with SSI and targets for intervention, yet interestingly, socioeconomic status remains as one of the strongest associations with poor surgical outcomes. ^{Reference Paro, Hyer, Diaz, Tsilimigras and Pawlik13}

Our cohort identified the theme of housing type and transportation as significant in the risk of deep incisional and organ space SSI. Within this theme, we hypothesize that lack of vehicle and transportation access plays a role in the development of SSI. Patients in the high SVI tertile lived 25 miles farther away compared to medium and low SVI populations. In a large cohort of 15,000 patients from 66 countries, the use of telemedicine to assess surgical wounds postdischarge was a feasible method to accurately identify SSI and optimize timing to present for follow up. ^{35,Reference Zhu, Williamson and Lin36} Furthermore, mobile health technology is emerging as a tool to improve healthcare delivery, bridging patient care from the health system to medically underserved communities. ^{Reference McLean, Sgrò and Brown37,Reference Lesher, Gavrilova, Ruggiero and Evans38} Almost all Americans (96%) own a cellphone, but only 81% own smartphones; further, broadband internet penetrance varies across rural and urban settings, which may limit use of a smartphone intervention in this patient population. ^{Reference Lesher, Gavrilova, Ruggiero and Evans38} Use of patient engagement with smartphone technology differs significantly by sex, race, and insurance status and should be carefully assessed if implemented as an intervention to improve SSI. ^{Reference Theiss, Wood and Shao39} The race/ethnicity theme was not significantly associated with SSI, which is similar to findings by DeWitt et al. ^{Reference Dewitt, Reinke and Inman23}

The results of the current research have multiple direct implications that can be addressed to effect appropriate prevention of SSIs during pre- and postsurgical care. Prevention efforts such as comprehensive preoperative assessments, tailored postoperative care, and patient education are pivotal in preventing SSI. ^{Reference Tarabay, Nix and Doline5} Routine healthcare questionnaires may identify unmet needs for physicians to refer patients for social assistance. ^{Reference Berkowitz, Hulberg and Hong41} Further, tapping into community based resources such as weekly meals have been shown to reduce ED visits and inpatient admissions and may play a role in optimizing nutrition prior to surgery. ^{Reference Berkowitz, Terranova and Hill42} The theme of socioeconomic status includes living below the poverty level, unemployment, housing cost burden, education, and health insurance. Low health literacy, a surrogate of education, is associated with increased postoperative complications, including SSIs. ^{Reference Theiss, Wood and McLeod43–Reference Qin, Jin, Min, Wang and Shen45} Utilizing SVI to identify high risk patients can optimize individualized education for SSI prevention.

Our study has several key strengths, including the large study population across seven procedure types at a large academic medical center in the Deep South. To our knowledge, this is the largest study evaluating NHSN SSI outcomes stratified by both census-tract SVI tertiles as well as SVI theme-specific associations. Further, the use of the NHSN to determine SSI status minimizes potential biases due to outcome misclassification that would occur if center-level data were the only source of outcomes. Additionally, the Deep South in the United States represents a critical, understudied region regarding highly vulnerable populations, particularly due to concentrated racial health disparities, poverty, and structural barriers. Studying this population is crucial in bridging a critical gap between known SSI risks and addressing specific, intertwined social and structural barriers to care and SSI in the Deep South.

Our study has a few main limitations. First, this is a retrospective study from a single medical center; thus, our findings may not be generalizable outside of the region. We opted to use SVI in our study, which has inherent limitations and is subject to residual confounding and difficulty in teasing out differences in themes (i.e. housing vs transportation as an association). Related to the use of SVI, the use of ZCTA-level measures may result in misclassification of true SVI status if a more granular measurement level was used; however, there is no reason to suspect the misclassification is related to SSI or SVI status such that observed associations are likely underestimates of the true association. Second, though we used a longer period of five years, the range included the COVID year of 2020 in which surgical volume was decreased; however, exclusion of procedures performed in 2020 did not appreciably alter the point estimates of the associations, and statistical inferences were similar. Third, although we included a heterogenous cohort of patients undergoing different surgical procedures, our results may not correlate to other types of operations outside of these procedures. The lack of significant associations observed for more severe SSIs may be a result of the overall low incidence of severe SSIs, which potentially decreased statistical power. Finally, though models included multiple potential confounders as covariates, the observed associations may still be biased by residual confounding.