Abstract
Background:
The effects of socioeconomic factors other than insurance status and race on outcomes following cardiac surgery are not well understood. We hypothesized that the Distressed Communities Index (DCI), a comprehensive socioeconomic ranking by zip code, would predict operative mortality following coronary artery bypass grafting (CABG).
Methods:
All patients who underwent isolated CABG (2010–2017) in the Virginia Cardiac Services Quality Initiative database were analyzed. The DCI accounts for unemployment, education level, poverty rate, median income, business growth, and housing vacancies, with scores ranging from 0 (no distress) to 100 (severe distress). Patients were stratified by DCI quartiles (I: 0–24.9, II: 25–49.9, III: 50–74.9, IV: 75–100) and compared. Hierarchical linear regression modeled the association between DCI and mortality.
Results:
A total of 19,756 CABG patients were analyzed, with mean PROM of 2.0±3.5%. Higher DCI scores were associated with increasing PROM. Overall operative mortality was 2.1% [n=424] and increased with increasing DCI quartile (I: 1.6% [95], II: 2.1% [77], III: 2.4% [114], IV: 2.6% [138], p=0.0009). The observed-to-expected ratio for mortality increased as level of socioeconomic distress increased. After risk adjustment for STS predicted risk of mortality, year of surgery, and hospital, the DCI remained predictive of operative mortality after CABG (OR 1.14 for each 25-point increase in DCI, 95% confidence interval 1.04–1.26, p=0.007).
Conclusions:
The Distressed Communities Index independently predicts risk-adjusted operative mortality following CABG. Socioeconomic status, although not part of traditional risk calculators, should be considered when building risk models, evaluating resource utilization, and comparing hospitals.
Keywords: coronary artery bypass grafts, CABG; database; health economics; outcomes; statistics, risk analysis/modeling
INTRODUCTION
Socioeconomic factors such as employment status, financial security, and education level are “fundamental determinants of human functioning” according to the American Psychological Association.[1] Many studies have evaluated the impact of socioeconomic status (SES) on surgical outcomes using proxies such as insurance status and race since these variables are routinely captured in patient databases.[2–6] These analyses are inherently limited in their ability to define a true relationship between health disparities and outcomes. Other studies have used components such as household income[7–9] and employment status[10, 11] to estimate overall SES and evaluate their relationship with outcomes. The most comprehensive studies to date have used multiple indicators from census data to determine a patient’s SES and its impact on failure to rescue after cancer surgery[12], preoperative quality of life[13], and mortality after cardiac surgery.[14–16] Consistently, an inverse correlation between SES and health outcomes is demonstrated.
The data points chosen to estimate a patient’s SES limit the findings of any particular study. In order to accurately determine the effect of health disparities on outcomes, use of a comprehensive estimate of social, economic, and financial status is necessary. Recently, the Economic Innovation Group, a “bipartisan public policy organization, founded in 2013, combining innovative research and data-driven advocacy to address America’s most pressing economic challenges,” released the Distressed Communities Index (DCI).[17] DCI is a composite ranking by zip code that accounts for seven component metrics that encompass unemployment, education level, poverty rate, median income, business establishments, job growth, and housing vacancies.[18] Based on DCI, one in six Americans lives in a zip code that ranks in the top 20% in terms of socioeconomic distress. The ability for these 52.3 million Americans to obtain acceptable healthcare compared to the remainder of the country is an important challenge that needs attention.
The objective of this study was to evaluate the impact of SES on outcomes after coronary artery bypass grafting (CABG), using the Economic Innovation Group’s comprehensive DCI. Considering that previous research to date supports a significant relationship between SES and health outcomes, we hypothesized that DCI would predict operative mortality after CABG for patients in a multi-state Society of Thoracic Surgeons (STS) database.
METHODS
Study Population
All patients undergoing isolated CABG between January 1, 2010 and January 31, 2017 (n=19,756) in the Virginia Cardiac Services Quality Initiative (VCSQI) database were included in the analysis. The VCSQI database (ARMUS Corporation, San Mateo, CA) captures STS demographic, preoperative, clinical, and 30-day outcomes data on patients who undergo cardiac surgery at 18 centers in Virginia and 1 center in North Carolina.[19] Cost data derived from Uniform Billing-04 files, which identifies charges classified by International Classification of Diseases ninth revision-based revenue codes, are matched to STS data. [20] Costs are estimated using cost-to-charge ratios submitted to Centers for Medicare and Medicaid Services. To account for medical-specific inflation, the market basket for the Medicare inpatient prospective payment system was used to adjust data to 2017 dollars. Patients were excluded for missing zip code (<5%), as this was required to link patients to DCI. Matching occurred at the time of data extraction allowing for analysis of a de-identified dataset. The Institutional Review Board at the University of Virginia granted exemption for this study.
Socioeconomic Status
DCI is available for all zip codes with more than 500 residents, which captures 99% of the American population. It is a composite score based on seven metrics: no high school degree, housing vacancy rate, adults not working, poverty rate, median income ratio, change in employment, and change in business establishments (definitions shown in Table 1).[18] The seven evenly weighted variables are used to calculate a zip code’s rank compared to its geographic peers and then normalized to obtain a raw distress score that ranges from 0 (no distress) to 100 (severe distress). The seven SES indicators were obtained from the American Communities Survey 2014 5-year Estimates and the Census Bureau County and Zip Code Business Patterns.
TABLE 1:
Components of the Distressed Communities Index[18]
| Component | Definition |
|---|---|
| No high school degreea | Percent of population age ≥ 25 years without a high school diploma |
| Housing vacancy ratea | Percent of habitable housing that is unoccupied and not for seasonal, recreational, or occasional use |
| Adults not workinga | Percent of the population age ≥ 16 years not currently in work |
| Poverty ratea | Percent of the population living under the poverty line |
| Median income ratioa | Ratio of a geography’s median income to that of its state |
| Change in employmentb | Percent change in the number of jobs (2010–2013) |
| Change in business establishmentsb | Percent change in the number of business establishments (2010–2013) |
Source: American Communities Survey 2014 5-year Estimates
Source: Census Bureau County and Zip Code Business Patterns
Risk Adjustment and Study End Point
The STS predicted risk of mortality (PROM) was used to adjust patient risk, along with year of operation (to account for improvements in quality over time) and hospital (to account for differences in outcomes at different centers). The primary endpoint for this study was operative mortality.[21]
Statistical Analysis
Continuous variables are presented as mean±standard deviation or median[interquartile range] and categorical variables are presented as number and percentage of total. Patients were stratified by DCI quartiles (I: 0–24.9, II: 25–49.9, III: 50–74.9, IV: 75–100) due to sample size (as opposed to quintiles which are presented by the Economic Innovation Group). Distribution of the seven components of the DCI across quartiles was analyzed with Kruskal-Wallis test given the non-normal distribution of the data. Commonly used proxies for SES (age, sex, insurance status), predicted risk scores, 30-day outcomes, and cost were likewise analyzed between DCI quartiles using Kruskal-Wallis test for continuous variables and Chi-square test for categorical variables. Hierarchical multivariate generalized linear regression modeled the association between DCI and operative mortality, with risk adjustment using PROM and year of surgery, while hospital was included as a random effect. The composite DCI score was used in the final model as it proved to be more predictive than any of the seven components individually. The utility of the addition of DCI was assessed by effect size (odds ratio) and significance, the change in area under the curve (AUC) for the nested models, and the net reclassification improvement (NRI). NRI is an index measure of how well a new model reclassifies subjects compared to an old model (correct versus incorrect changes in prediction for cases and controls separately).[22] Therefore, it is a combination of these proportions with a maximum value of 2. In this case a continuous or category free NRI was utilized. While the change in AUC represents a population level assessment of model performance, NRI assessed the change in risk prediction at the individual level. Interactions between DCI and predicted risk scores were evaluated and no correlations were identified (Supplementary Figure 1). The model was also run to determine if DCI was a significant predictor of in-hospital or 30-day mortality to help clarify when the effect of SES is most apparent. Alpha level for significance was 0.05. All analyses were performed with SAS, version 9.4 (SAS Institute, Cary, NC). Observed-to-expected ratios for operative mortality by DCI quartile were modeled in Prism 7 (GraphPad Software Inc, La Jolla, CA).
RESULTS
Characteristics and Outcomes after CABG
A total of 19,756 patients underwent isolated CABG during the study period. Mean age was 64.5±10.3 years and PROM was 2.0±3.5%, with 62.1% (n=12,271) of cases being urgent and 75.9% (n=14,958) requiring 3-vessel bypass. The most common insurance status was government payor (56.6%, n=11,145). Demographics, perioperative characteristics, and operative details for the entire cohort are shown in Table 2.
TABLE 2:
Demographics and operative details for CABG patients
| All patients (n=19,756) | |
|---|---|
| Demographics | |
| Female | 5113 (25.9)a |
| Age | 64.5±10.3b |
| BMI (kg/m2) | 30.2±11.4 |
| Comorbidities | |
| Hypertension | 17320 (87.7) |
| Diabetes | 9135 (46.3) |
| End-stage renal disease | 564 (2.9) |
| Prior PCI | 5567 (28.2) |
| Reoperation | 575 (2.9) |
| Insurance status | |
| Commercial | 5891 (29.9) |
| Government | 11145 (56.6) |
| HMO | 1012 (5.1) |
| Self pay | 1652 (8.4) |
| Ejection fraction (%) | 51.5±12.2 |
| Preoperative IABP | 1398 (7.1) |
| PROM (%) | 2.0±3.5 |
| PROMM (%) | 15.4±12.0 |
| Operative Characteristics | |
| Case Status | |
| Elective | 6700 (33.9) |
| Urgent | 12271 (62.1) |
| Emergent | 747 (3.8) |
| Emergent Salvage | 35 (0.2) |
| Number of bypassed vessels | |
| 0 | 19 (0.1) |
| 1 | 820 (4.2) |
| 2 | 3919 (19.9) |
| 3 | 14958 (75.9) |
| Bypass time (min) | 97.6±35.9 |
| Cross clamp time (min) | 69.8±26.1 |
| ICU length of stay (hrs) | 45.8 [24.7–74]c |
| Hospital length of stay (days) | 5 [4–7] |
Number (%), all such values
Mean±standard deviation, all such values
Median[interquartile range], all such values
BMI=Body mass index, CABG=coronary artery bypass grafting, HMO=health maintenance organization, IABP=intra-aortic balloon pump, ICU=intensive care unit, PCI=percutaneous coronary intervention, PROM=predicted risk of mortality, PROMM=predicted risk of morbidity or mortality
Operative mortality was 2.1% (n=424) while major morbidity occurred in 11.9% (n=2,356). The most common in-hospital postoperative event was atrial fibrillation (21.6%, n=4275). Outcomes for the entire cohort are shown in Table 3.
TABLE 3:
Outcomes after CABG
| Variable | All patients (n=19,756) |
|---|---|
| In-hospital postoperative event | 7092 (35.9)a |
| Stroke | 252 (1.3) |
| Atrial fibrillation | 4275 (21.6) |
| Cardiac arrest | 308 (1.6) |
| Pneumonia | 414 (2.1) |
| Renal failure | 468 (2.8) |
| Dialysis | 258 (1.3) |
| Deep sternal wound infection | 22 (0.1) |
| Reoperation | 700 (3.5) |
| Bleeding reoperation | 303 (1.5) |
| Major morbidity | 2356 (11.9) |
| In-hospital mortality | 279 (1.4) |
| 30-day mortality | 333 (1.7) |
| Operative mortality | 424 (2.1) |
| Major morbidity or mortality | 2496 (12.6) |
| Total cost (USD) | 45,703±31,934b |
Number (%), all such values
Mean±standard deviation, all such values
CABG=coronary artery bypass grafting, USD=United States dollar
Demographics and Outcomes by DCI Quartile
Comparison by DCI quartile for the distribution of patients, the seven components of DCI, and mean population per zip code are shown in Table 4. All DCI components and mean population were significantly different between quartiles.
TABLE 4:
Components of DCI for CABG patients stratified by DCI quartile
| Distress Score 0–24.9 | Distress Score 25–49.9 | Distress Score 50–74.9 | Distress Score 75–100 | p-value | |
|---|---|---|---|---|---|
| Number of patients | 6025 | 3726 | 4733 | 5272 | - |
| Components of DCIa | |||||
| No high school degree | 7.2±3.6b | 11.0±4.3 | 15.3±4.9 | 20.5±5.0 | < 0.0001 |
| Housing vacancy rate | 4.8±2.1 | 6.9±2.8 | 9.6±2.9 | 13.1±4.0 | < 0.0001 |
| Adults not working | 34.6±5.0 | 39.3±5.3 | 43.5±5.2 | 50.4±6.0 | < 0.0001 |
| Poverty rate | 5.9±2.3 | 9.6±3.0 | 15.9±5.8 | 21.7±7.1 | < 0.0001 |
| Median income ratio | 143.3±39.9 | 100.4±17.4 | 78.7±14.3 | 66.6±14.0 | < 0.0001 |
| Change in employment | 10.1±17.1 | 1.9±13.6 | 1.5±13.0 | −3.9±19.8 | < 0.0001 |
| Change in establishments | 4.1±7.7 | −1.7±6.1 | −1.0±7.6 | −5.2±7.3 | < 0.0001 |
| Population per zip code | 34,640±18,107 | 25,708±17,043 | 23,422±16,651 | 15,495±12,501 | < 0.0001 |
Components of Distressed Communities Index displayed as percentages and defined in Table 1
Mean±standard deviation, all such values
CABG=coronary artery bypass grafting, DCI=Distressed Communities Index
Demographics stratified by DCI quartile are shown in Table 5 (and Supplementary Table 1) and outcomes are shown in Table 6 (and Supplementary Table 2). Higher DCI scores were associated with incrementally increasing PROM (I: 1.9±3.3%, II: 2.0±3.8%, III: 2.1±3.5%, IV: 2.1±3.5%, p<0.0001). Operative mortality increased with each increasing DCI quartile (I: 1.6% [95], II: 2.1% [77], III: 2.4% [114], IV: 2.6% [138], p=0.0009). Patients in the highest DCI quartile (lowest SES) had 1.6-fold greater odds of operative mortality compared with patients in the lowest quartile. There was no significant association between DCI and total hospital cost or rate of major morbidity (both p>0.05). The observed-to-expected (O:E) ratio for operative mortality increased as level of socioeconomic distress increased (Figure 1), with worse than expected outcomes for any patient not in the first quartile (distress score 0–25).
TABLE 5:
Demographics for CABG patients stratified by DCI quartile
| Distress Score 0–24.9 | Distress Score 25–49.9 | Distress Score 50–74.9 | Distress Score 75–100 | p-value | |
|---|---|---|---|---|---|
| Number of patients | 6025 | 3726 | 4733 | 5272 | - |
| Female | 1286 (21.3)a | 964 (25.9) | 1306 (27.6) | 1557 (29.5) | < 0.0001 |
| Age | 64.4±10.1b | 64.8±10.5 | 64.8±10.4 | 64.1±10.3 | 0.0007 |
| Comorbidities | |||||
| Hypertension | 5120 (85) | 3250 (87.3) | 4190 (88.6) | 4760 (90.3) | < 0.0001 |
| Diabetes | 2601 (43.2) | 1657 (44.5) | 2219 (46.9) | 2658 (50.4) | < 0.0001 |
| ESRD | 139 (2.3) | 95 (2.6) | 139 (2.9) | 191 (3.6) | 0.0003 |
| Prior PCI | 1549 (25.7) | 1066 (28.6) | 1406 (29.7) | 1546 (29.3) | < 0.0001 |
| Reoperation | 141 (2.3) | 106 (2.8) | 152 (3.2) | 176 (3.3) | 0.008 |
| Insurance status | |||||
| Commercial | 2119 (35.3) | 1120 (30.1) | 1282 (27.2) | 1370 (26.0) | < 0.0001 |
| Government | 3059 (50.9) | 2083 (56.0) | 2802 (59.5) | 3201 (60.9) | |
| HMO | 393 (6.5) | 210 (5.7) | 219 (4.6) | 190 (3.6) | |
| Self Pay | 441 (7.3) | 305 (8.2) | 408 (8.7) | 498 (9.5) | |
| PROM (%) | 1.9±3.3 | 2.0±3.8 | 2.1±3.5 | 2.1±3.5 | < 0.0001 |
| PROMM (%) | 14.7±11.8 | 15.2±11.9 | 15.6±12.0 | 16.2±12.3 | < 0.0001 |
| Case Status | |||||
| Elective | 2104 (34.9) | 1236 (33.2) | 1622 (34.3) | 1738 (33) | 0.0002 |
| Urgent | 3628 (60.2) | 2346 (63) | 2947 (62.3) | 3350 (63.5) | |
| Emergent | 279 (4.6) | 137 (3.7) | 160 (3.4) | 171 (3.2) | |
| Emergent Salvage | 12 (0.2) | 7 (0.2) | 3 (0.1) | 13 (0.3) | |
Number (%), all such values
Mean±standard deviation, all such values
CABG=coronary artery bypass grafting, DCI=Distressed Communities Index, ESRD=end-stage renal disease, HMO=health maintenance organization, PCI=percutaneous coronary intervention, PROM=predicted risk of mortality, PROMM=predicted risk of morbidity or mortality
TABLE 6:
Outcomes after CABG stratified by DCI quartile
| Distress Score 0–24.9 | Distress Score 25–49.9 | Distress Score 50–74.9 | Distress Score 75–100 | p-value | |
|---|---|---|---|---|---|
| Number of patients | 6025 | 3726 | 4733 | 5272 | - |
| Major morbidity | 671 (11.1)a | 474 (12.7) | 567 (12.0) | 644 (12.2) | 0.1 |
| Major morbidity or mortality | 699 (11.6) | 501 (13.5) | 610 (12.9) | 686 (13.0) | 0.029 |
| In-hospital mortality | 65 (1.1) | 54 (1.5) | 66 (1.4) | 94 (1.8) | 0.018 |
| 30-day mortality | 69 (1.2) | 69 (1.9) | 82 (1.7) | 113 (2.1) | 0.0004 |
| Operative mortality | 95 (1.6) | 77 (2.1) | 114 (2.4) | 138 (2.6) | 0.0009 |
| Total cost (USD) | 44,676±28,656b | 47,718±37,330 | 45,132±31,345 | 45,980±31,726 | 0.15 |
Number (%), all such values
Mean±standard deviation, all such values
CABG=coronary artery bypass grafting, DCI=Distressed Communities Index, USD=United States Dollar
Figure 1.
Operative mortality observed-to-expected ratios for CABG patients stratified by DCI quartile. * denotes significance compared to 1. P=0.59 for Distress Score 0–24.9: p=0.59; Distress Score 25–49.9: p=0.82; Distress Score 50–74.9: p=0.13; Distress Score 75–100: p=0.009.
DCI Predicts Operative Mortality
After risk adjustment, DCI remained predictive of operative mortality after CABG with an odds ratio of 1.14 for each 25-point increase in DCI (95% confidence interval (CI) 1.04–1.26, p=0.007, c-statistic=0.798, Hosmer-Lemeshow (HL)=0.56, Table 7). Addition of DCI to the model significantly improved the predictive power with AUC difference of 0.0046 (95% CI 0.0032–0.0061, p=0.0479) and with NRI of −0.000437 (95% CI −0.0014, −0.0005, p=0.048). DCI was a stronger predictor than insurance status (OR 0.63, 95% CI 0.35–1.15, p=0.43, c-statistic=0.794, HL p=0.54) with better model fit. DCI was also a significant predictor of in-hospital mortality (OR 1.13, 95% CI 1.02–1.26, p=0.019, c-statistic=0.809, HL=0.64) and 30-day mortality (OR 1.17, CI 1.07–1.29, p=0.001, c-statistic=0.791, HL=0.64).
TABLE 7:
Association between DCI and operative mortality modeled with hierarchical multivariate generalized linear regression
| Variable | Odds Ratio | 95% Confidence Interval | p-value |
|---|---|---|---|
| DCI | 1.14 | 1.04 – 1.26 | 0.007 |
| Year | 0.93 | 0.88 – 0.97 | 0.002 |
| PROM | 2.97 | 2.72–3.25 | < 0.001 |
c-statistic=0.7982, Hosmer-Lemeshow p=0.560
DCI=Distressed Communities Index, PROM=predicted risk of mortality
COMMENT
Socioeconomic disparities are an inescapable truth that impacts patients throughout the US healthcare system. It remains difficult to quantify the varying impact of SES and to integrate potential solutions into care delivery. The present study evaluated if DCI, a comprehensive risk score based on seven individual metrics, could accurately estimate patients’ SES and its effect on cardiac surgical outcomes.[18] This was accomplished using patients who underwent CABG during a seven-year period and were captured in the VCSQI database. The study found that a patient’s distress score was highly predictive of operative mortality after CABG even after adjusting for STS PROM, year of surgery, and hospital. Every 25-point increase in a patient’s distress score, which ranges from 0 (no distress) to 100 (severe distress), was associated with a 14% risk increase in odds of mortality following CABG. This finding highlights the importance of a patient’s SES on their health outcomes and provides an easy-to-use metric (DCI) for quantifying their associated risk. This metric can easily be added to national databases as a stand-alone reportable field that can be used for retrospective analyses, but may also be valuable as an included variable in preoperative risk prediction models.
In the study cohort of 19,756 CABG patients, mean PROM was 2.0±3.5% and operative mortality was 2.1%, highlighting the fact that the STS database collects an appropriate amount of data to allow for accurate risk prediction.[23] However, the incorporation of a patient’s distress score, based on their home zip code, correlated closely with risk of mortality. Not only did increasingly distressed communities have increasingly high-risk patients, they were associated with increasing O:E for operative mortality. Performance after CABG switched from better than expected (O:E=0.84) for patients in the lowest DCI quartile to worse than expected (O:E=1.24) for patients in the highest DCI quartile.
Although DCI was closely associated with operative mortality, there was no significant difference in rate of morbidity following CABG between DCI quartiles. Thus, it appears as though the majority of risk related to SES that is not accounted for with current prediction models pertains to mortality. The addition of a socioeconomic indicator such as DCI may help improve that risk prediction. When in-hospital mortality and 30-day mortality were modeled, the DCI was a significant predictor of both outcomes. Future studies with long-term follow-up will help quantify the effect of SES on outcomes after patients return home.
In a study of patients undergoing cancer surgery, Reames and colleagues found that patients in the lowest quintile of SES had significantly higher rates of failure to rescue, defined as fatality in patients with 1 or more major complications.[12] Additionally, they reported 1.3-fold greater odds of perioperative mortality for patients in the lowest SES quintile compared with patients in the highest (10.2% vs. 7.7%). One can hypothesize that patients who are discharged home after CABG to environments with limited resources and support may be less likely to receive proper follow-up care if a morbidity occurs. Centers can consider performing short-term phone call follow-ups after discharge with patients from highly distressed communities, with the goal of identifying patients who may be developing a postoperative complication that requires intervention. Other potential solutions include more frequent use of home health visits, development of outreach clinics, and improved care coordination in distressed communities.
Our study found that patients with low SES based on DCI (quartile IV) had 1.6-fold greater odds of operative mortality following CABG compared with patients with high SES (quartile I). In two studies by Koch and colleagues, block socioeconomic data from the 2000 US Census was used to estimate cardiac surgery patients’ socioeconomic position.[13, 16] Although slightly different metrics were used to create a composite estimate of SES compared with DCI, the findings were similar. The authors concluded that lower SES was associated with lower quality of life and higher risk-adjusted mortality among cardiac surgery patients. Compared with those studies, DCI is a more efficient metric to estimate SES as it is generated by an independent entity and can be incorporated easily into clinical databases in a de-identified manner.
The findings of the present study, along with other studies evaluating the impact of SES on surgical outcomes, have all found an association between lower SES and worse outcomes. These findings highlight the impact of SES on cardiac surgical outcomes and should be considered when building risk models, evaluating resource utilization, and comparing hospitals. Since DCI is calculated independently and appears to be a comprehensive estimate of patients’ SES, consideration should be given into incorporating it into surgical databases such as the STS National Database. This can be done at the database level based on zip code to allow for de-identified datasets. While the findings of the present study are promising, future research is needed to validate these findings on a national level prior to large-scale implementation. Improving our understanding of how socioeconomic risk affects outcomes will allow us to tailor healthcare in a way that accounts for these differences. If future studies support an effect of SES on long-term outcomes, patients with low SES based on DCI could trigger additional post-discharge efforts not routinely provided. Modifying our risk prediction models and postoperative follow-up protocols are two other potential ways in which a better understanding of socioeconomic risk may allow us to improve delivery of care. The decision to include DCI in future models will require investigation for collinearity and use of interaction terms with each included variable, which is beyond the scope of the present study and should be performed with a large, national dataset.
This study is limited by its retrospective nature precluding demonstration of causality. DCI is a comprehensive socioeconomic metric based on zip code, which limits its ability to accurately estimate each patient’s specific risk profile. While this is undoubtedly a limitation, it also strengthens the argument for including it in databases as it can be incorporated efficiently and de-identified for analysis. An additional advantage of DCI is that it includes community-based factors such as business growth and employment changes which are not captured by recording patient-specific socioeconomic factors only. Future studies should compare DCI to other potential socioeconomic metrics to identify the optimal approach to account for a patient’s SES. DCI scores used in this analysis were calculated from a single census report and thus do not account for changes over time related to a community’s SES. Additionally, the STS category for insurance status does not differentiate between Medicare and Medicaid, which will be important to investigate with future studies. The findings of the current study in CABG patients will need to be expanded to other surgical populations and to national datasets in order to confirm the utility of the DCI to consistently estimate patients’ socioeconomic risk in an accurate and meaningful way.
In conclusion, DCI, a composite socioeconomic score based on seven independent factors, was independently associated with operative mortality following CABG in a regional cohort of approximately 20,000 patients. As level of socioeconomic distress increases, patients’ surgical risk increases, yet the mortality rate increases even more than predicted. These findings demonstrate the impact of SES on healthcare outcomes, particularly after CABG. As the US healthcare system continues to focus on improving quality and outcomes, incorporation of patients’ SES into national databases, risk prediction models, and treatment plans is prudent.
Supplementary Material
ACKNOWLEDGEMENTS
This work was supported by NHLBI grants UM1 HL088925 and T32 HL007849 (ILK).
ABBREVIATIONS
- AUC
Area under the curve
- CABG
Coronary artery bypass grafting
- CI
Confidence interval
- DCI
Distressed Communities Index
- HL
Hosmer-Lemeshow
- NRI
Net Reclassification Index
- O:E
Observed-to-expected ratio
- OR
Odds ratio
- PROM
Predicted risk of mortality
- SES
Socioeconomic status
- STS
Society of Thoracic Surgeons
- VCSQI
Virginia Cardiac Services Quality Initiative
Footnotes
Meeting: Oral presentation, STS Annual Meeting, 0½018, Fort Lauderdale, FL
DISCLOSURES
The authors have no conflicts of interest to report. The findings expressed in this manuscript are solely those of the listed authors and not necessarily those of The Economic Innovation Group. The Economic Innovation Group does not guarantee the accuracy or reliability of, or necessarily agree with, the information provided herein.
REFERENCES
- 1.Report of the American Psychological Association Task Force on Socioeconomic Status Available at External link http://www.apa.org/pi. Accessed January 20, 2018.
- 2.Chan SY, Suwanabol PA, Damle RN, et al. Characterizing Short-Term Outcomes Following Surgery for Rectal Cancer: the Role of Race and Insurance Status. J Gastrointest Surg 2016;20:1891–1898. [DOI] [PubMed] [Google Scholar]
- 3.Hayes S, Napolitano MA, Lent MR, et al. The effect of insurance status on pre- and post-operative bariatric surgery outcomes. Obes Surg 2015;25:191–194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Rohlfing ML, Mays AC, Isom S, Waltonen JD. Insurance status as a predictor of mortality in patients undergoing head and neck cancer surgery. Laryngoscope 2017;127:2784–2789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Weyh AM, Lunday L, McClure S. Insurance Status, an Important Predictor of Oral Cancer Surgery Outcomes. J Oral Maxillofac Surg 2015;73:2049–2056. [DOI] [PubMed] [Google Scholar]
- 6.Charles EJ, Johnston LE, Herbert MA, et al. Impact of Medicaid Expansion on Cardiac Surgery Volume and Outcomes. Ann Thorac Surg 2017;104:1251–1258. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ultee KH, Bastos Goncalves F, Hoeks SE, et al. Low Socioeconomic Status is an Independent Risk Factor for Survival After Abdominal Aortic Aneurysm Repair and Open Surgery for Peripheral Artery Disease. Eur J Vasc Endovasc Surg 2015;50:615–622. [DOI] [PubMed] [Google Scholar]
- 8.Poole L, Leigh E, Kidd T, Ronaldson A, Jahangiri M, Steptoe A. The combined association of depression and socioeconomic status with length of post-operative hospital stay following coronary artery bypass graft surgery: data from a prospective cohort study. J Psychosom Res 2014;76:34–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Armenia SJ, Pentakota SR, Merchant AM. Socioeconomic factors and mortality in emergency general surgery: trends over a 20-year period. J Surg Res 2017;212:178–186. [DOI] [PubMed] [Google Scholar]
- 10.Lazzarino AI, Hamer M, Stamatakis E, Steptoe A. Low socioeconomic status and psychological distress as synergistic predictors of mortality from stroke and coronary heart disease. Psychosom Med 2013;75:311–316. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Pell JP, Pell AC, Norrie J, Ford I, Cobbe SM. Effect of socioeconomic deprivation on waiting time for cardiac surgery: retrospective cohort study. BMJ 2000;320:15–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Reames BN, Birkmeyer NJ, Dimick JB, Ghaferi AA. Socioeconomic disparities in mortality after cancer surgery: failure to rescue. JAMA Surg 2014;149:475–481. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Koch CG, Li L, Shishehbor M, et al. Socioeconomic status and comorbidity as predictors of preoperative quality of life in cardiac surgery. J Thorac Cardiovasc Surg 2008;136:665–672. [DOI] [PubMed] [Google Scholar]
- 14.Boscarino JA, Chang J. Survival after coronary artery bypass graft surgery and community socioeconomic status: clinical and research implications. Med Care 1999;37:210–216. [DOI] [PubMed] [Google Scholar]
- 15.Kim C, Diez Roux AV, Hofer TP, Nallamothu BK, Bernstein SJ, Rogers MA. Area socioeconomic status and mortality after coronary artery bypass graft surgery: the role of hospital volume. Am Heart J 2007;154:385–390. [DOI] [PubMed] [Google Scholar]
- 16.Koch CG, Li L, Kaplan GA, et al. Socioeconomic position, not race, is linked to death after cardiac surgery. Circ Cardiovasc Qual Outcomes 2010;3:267–276. [DOI] [PubMed] [Google Scholar]
- 17.Economic Innovation Group Available at External link http://eig.org/about-us-mission2018. Accessed January 20, 2018.
- 18.Economic Innovation Group Distressed Communities Index Available at External link http://eig.org/dci. Accessed January 20, 2018.
- 19.Mehaffey JH, Hawkins RB, Byler M, et al. Cost of individual complications following coronary artery bypass grafting. J Thorac Cardiovasc Surg 2018;155:875–82. [DOI] [PubMed] [Google Scholar]
- 20.Ailawadi G, LaPar DJ, Speir AM, et al. Contemporary costs associated with transcatheter aortic valve replacement: a propensity-matched cost analysis. Ann Thorac Surg 2016;101:154–60. [DOI] [PubMed] [Google Scholar]
- 21.Overman DM, Jacobs JP, Prager RL, et al. Report from the Society of Thoracic Surgeons National Database Workforce: clarifying the definition of operative mortality. World J Pediatr Congenit Heart Surg 2013;4:10–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Pencina MJ, D’Agostino RB, Demler OV. Novel metrics for evaluating improvement in discrimination: Net reclassification and integrated discrimination improvement for normal variables and nested models. Stat Med 2012;31:101–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Jacobs JP, Shahian DM, Prager RL, et al. The Society of Thoracic Surgeons National Database 2016 Annual Report. Ann Thorac Surg 2016;102:1790–1797. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.