The impact of social determinants of health on outcomes and complications following total knee arthroplasty: an analysis of multiple health deprivation indices

In the age of value-based medicine, total knee arthroplasty (TKA) is the most effective treatment to improve quality of life for patients with knee osteoarthritis (S.T. Skou ). al, 2018, para 3.). As demands for the procedure steadily increase annually, surgeons are challenged to become more efficient while simultaneously optimizing outcomes (Hamilton et al., 2015, para. 1). For these reasons, there is an emerging interest in patient optimization and risk stratification, with particular focus on social determinants of health (SDoH). Social determinants of health are defined as the immediate and structural conditions in which people live that have an influence on the well-being and healthcare accessibility of individuals (Wilkinson & Marmot, 2003, p. 5). The World Health Organization further defines specific variables that qualify as a SDoH, including income, education, job security and unemployment, food insecurity, housing environment, social inclusion, structural conflict, and access to affordable health services of decent quality (Marmot & Wilkinson, 2003, p. 7).

Recent evidence associates outcomes following total joint replacement to SDoH (Holbert et al., 2022, para. 1). Some studies have demonstrated a correlation between lower socioeconomic status and minority status to a greater likelihood of emergency department visit (ED visit), readmission, and discharge to institutional care following total joint arthroplasty (Holbert et al., 2022, para. 3). Based on this evidence, SDoH can have a direct impact on costs of care. Yet it remains uncommon practice to assess and counsel patients with regards to SDoH.

A reason for this is that it is difficult to holistically assess a patient for SDoH and objectively interpret health disparity data. Multiple indices have been created to help better quantify a patient’s risk of healthcare inaccessibility based on the location of their home. Examples include the Center of Disease Control and Prevention’s Social Vulnerability Index (SVI) and the Health Resources and Services Administration’s Area of Deprivation Index (ADI). The SVI is a summative, national percentile score that quantifies a patient’s risk of healthcare vulnerability, with additionally available sub-domain/theme scores (socioeconomic status, household characteristics, racial and ethnic minority status, and housing type/transportation) (Centers for Disease Control and Prevention (CDC), 2020). The ADI is a national percentile score that similarly quantifies an individual’s risk of healthcare disparity based on different census variables (Kind & Buckingham, 2018; University of Wisconsin School of Medicine and Public Health, 2018). A higher score in either index demonstrates greater healthcare disadvantage. Both indices have been validated in identifying at-risk neighborhoods and have been consistently shown to correlate with health outcomes in multiple medical disciplines (CDC, 2020).

However, orthopaedic literature seems to be a laggard in this area of interest. A handful of studies have retrospectively looked at the ADI in relation to total joint arthroplasty, but results vary (Karimi et al., 2022, p 1.). One study demonstrated no correlation between ADI and ED visits (Shaw et al., 2021, p 2.). Other studies demonstrated a significant correlation of higher ADI to poorer outcomes and greater postoperative readmissions (Karimi et al., 2022, p 1.). A few studies have correlated higher SVI to increased costs and worse outcomes (Karimi et al., 2022, p 1.).  
Better understanding of SDoH may allow surgeons to identify at-risk patients and provide resources for optimization. Given the utility of the SVI and ADI in estimating an individual’s risk of health disparity, the primary purpose of our study was to utilize database derived outcomes data to determine the correlation between SVI and ADI with discharge disposition and complications following total knee arthroplasty. A secondary aim was to correlate SVI and ADI to patient reported outcome measures (PROMs) up to a year from surgery. We hypothesize that higher SVI subdomain scores, higher SVI scores, and higher ADI scores will be associated with greater rates of discharge to intuitional care facilities, ED visits, and readmission in addition to poorer PROMs scores.

Methods or Case Description:

After institutional review board approval, a retrospective chart review was conducted on all TKAs performed between August 2013 and March 2021 at a single multi-center hospital system. Data was collected from 6 hospitals from within this hospital system. Cases included in the investigation were limited to primary, elective TKAs. Revision TKAs, unicompartmental knee arthroplasties including medial, lateral, or patellofemoral knee arthroplasty, and TKAs performed as urgent or emergent procedures in the setting of traumatic or pathologic indications were excluded. Patients were also excluded if their address was unavailable at the time of review.

An inquiry was submitted to the Michigan Arthroplasty Research Collaborative Quality Initiative (MARCQI) database to abstract data on all TKA cases that qualified under the inclusion and exclusion criteria. Cases were limited to the study period based on the initiation of MARCQI database collection at the index hospital system. Upon consideration of inclusion and exclusion criteria, a total of 19,321 total knee arthroplasty cases were included in the investigation. Patient demographics and outcomes were descriptively summarized and presented in Table 1. Collected baseline demographics and medical history included age, sex, smoking history, alcohol use history, history of deep venous thrombosis (DVT), preoperative use of assistive devices, body mass index, diabetes, American Society of Anesthesiologists (ASA) score, and use of robotic assistance.

Our primary testing variables were the SVI sub-theme, SVI total, and ADI national percentile ranks. Patient addresses were first retrieved via MARCQI database inquiry. Patient addresses were subsequently geocoded to 9- or 12-digit census tract codes. Using appropriate census-tract codes, the patients were matched to respective SVI total and SVI theme percentile ranks for socioeconomic status, household characteristics, race and minority status, and housing type/transportation. A similar methodology was used to match patients to their respective ADI percentile rank. A breakdown of included metrics in each SVI theme and ADI is presented in Table 2.
All statistical analyses were performed in the SPSS software version 26 (IBM, Armonk, NY). Patient demographics and outcomes were descriptively summarized. Patients were stratified into quartiles based on their national percentiles in SVI sub-themes, SVI total score, and ADI score (0-24.99%, 25-49.99%, 50-74.99%, 75-100%). Patient characteristics and outcomes were descriptively summarized by groups stratified based upon the SVI and ADI quartiles, and analysis of variance (ANOVA) and Pearson’s chi-square analyses were used to compare continuous and categorical variables, respectively.  Subsequent backwards, stepwise, multivariate binary logistic regressions were conducted for the need for blood transfusion, discharge disposition (home versus non-home), and incidence of complication versus SVI/ADI variables and patient covariates (age, sex, BMI, smoking status, alcohol abuse, history of DVT, DM, ASA score, use of assistive devices, use of robotic assistance). Backwards, stepwise, multivariate linear regressions were conducted for LOS and change in KOOS, JR versus SVI/ADI variables and the same patient covariates. SVI themes, SVI and ADI percentiles were treated as a continuous variable for regressions. Separate regressions were conducted for each SVI theme, SVI total, and ADI to avoid these metrics confounding one another. The threshold for exclusion in subsequent regressions was a p-value > 0.1. An alpha-level of 0.05 was used to determine statistical significance.

Outcomes variables included length of stay (LOS), discharge disposition (home versus skilled nursing facility versus inpatient rehabilitation), need for blood transfusion, and incidence of postoperative complications. Postoperative complications were documented by the database for the 90 days following surgery. Complications evaluated included death, deep venous thrombosis/pulmonary embolism, periprosthetic fracture, hardware failure, prosthetic joint infection, any purpose ED visit, hospital readmission, and all-purpose reoperation. To account for reoperations past 90 days, the database was cross-referenced for revision surgeries for the same patient and limb; if the patient had undergone revision at any point past 90 days, the procedure date and reason was documented. Cross-referencing was similarly conducted to account for chronic septic revisions past 90 days. Lastly, Knee Osteoarthritis and Injury Severity Score, Joints Replacement (KOOS, JR) scores were collected at the preoperative and 6 months to 1-year timepoints. The difference between preoperative and postoperative KOOS, JR was calculated. Only 1,895 patients from the total cohort had complete KOOS, JR data for calculation of the change in KOOS, JR and were therefore included in this sub-analysis.

Outcomes:

 The results of the bivariate comparisons of patients grouped by SVI and ADI quartiles are presented in Tables 3-8. Interpretation of the results demonstrate statistically significant differences in TKA outcomes between SVI/ADI quartiles. Greater SVI and ADI scores resulted in greater rates of discharge to a skilled nursing facility. Incidence of ED visits seemed to follow a stepwise, positive trend with higher SVI sub-themes, total SVI, and ADI quartiles. Incidence of readmissions after surgery followed a similar trend with an association to ADI and all SVI quartiles except the household characteristics SVI theme. Postoperative change in KOOS, JR was not found to have a statistically significant difference when compared amongst all SVI and ADI quartiles (p > 0.05).

The results of subsequent multivariate regressions are summarized in Table 9. All outcome variables, except death and change in KOOS, JR, were found to correlate to at least one SVI theme, SVI, or ADI.

Conclusion:

In an effort to provide healthcare equity, it is critical for adult reconstructive surgeons to account for patient differences in healthcare access and well-being. This study demonstrates that social determinants of health, as measured by the social vulnerability index and area of deprivation index, have important relationships with outcomes following total knee arthroplasty. Only by holistically assessing patients can arthroplasty surgeons look towards optimizing at-risk patients. Investigation of healthcare education and social work interventions are needed for our socially disadvantaged patient population.