Association of Insurance Type With Colorectal Surgery Outcomes and Costs at a Safety-Net Hospital

Objective: Association of insurance type with colorectal surgical complications, textbook outcomes (TO), and cost in a safety-net hospital (SNH). Background: SNHs have higher surgical complications and costs compared to low-burden hospitals. How does presentation acuity and insurance type influence colorectal surgical outcomes? Methods: Retrospective cohort study using single-site National Surgical Quality Improvement Program (2013–2019) with cost data and risk-adjusted by frailty, preoperative serious acute conditions (PASC), case status and open versus laparoscopic to evaluate 30-day reoperations, any complication, Clavien-Dindo IV (CDIV) complications, TO, and hospitalization variable costs. Results: Cases (Private 252; Medicare 207; Medicaid/Uninsured 619) with patient mean age 55.2 years (SD = 13.4) and 53.1% male. Adjusting for frailty, open abdomen, and urgent/emergent cases, Medicaid/Uninsured patients had higher odds of presenting with PASC (adjusted odds ratio [aOR] = 2.02, 95% confidence interval [CI] = 1.22–3.52, P = 0.009) versus Private. Medicaid/Uninsured (aOR = 1.80, 95% CI = 1.28–2.55, P < 0.001) patients were more likely to undergo urgent/emergent surgeries compared to Private. Medicare patients had increased odds of any and CDIV complications while Medicaid/Uninsured had increased odds of any complication, emergency department or observations stays, and readmissions versus Private. Medicare (aOR = 0.51, 95% CI = 0.33–0.88, P = 0.003) and Medicaid/Uninsured (aOR = 0.43, 95% CI = 0.30–0.60, P < 0.001) patients had lower odds of achieving TO versus Private. Variable cost %change increased in Medicaid/Uninsured patients to 13.94% (P = 0.005) versus Private but was similar after adjusting for case status. Urgent/emergent cases (43.23%, P < 0.001) and any complication (78.34%, P < 0.001) increased %change hospitalization costs. Conclusions: Decreasing the incidence of urgent/emergent colorectal surgeries, possibly by improving access to care, could have a greater impact on improving clinical outcomes and decreasing costs, especially in Medicaid/Uninsured insurance type patients.


INTRODUCTION
Frailty 1 and social risk factors 2 significantly impact colorectal surgical outcomes; however, value-based medicine risk adjustment models fail to account for these factors. Additionally, the impact of frailty on patient outcomes is not captured using standard Hierarchical Condition Category (HCC) risk adjustment. 3,4 High Social Vulnerability Index scores in colorectal surgery patients were associated with higher risk of postoperative complications and index hospitalization expenditures, 2 but are not included in risk adjustment. Safety-net hospitals (SNH) serve higher proportions of low socioeconomic status (SES) patients with higher severity of illness scores, higher rates of emergency surgeries, and longer hospital length of stay (LOS). 5 Hospital Readmission Reduction Program and other pay for performance (P4P) programs unintentionally contribute to widening disparities in healthcare and outcomes, penalizing SNH, and further limiting resources to treat vulnerable populations. [6][7][8] High-burden, SNH have higher postoperative complications and costs compared to low-burden hospitals. 5,9,10 Insurance status/type is a common proxy used for patient SES. [11][12][13] Moreover, dual-eligibility for Medicare and Medicaid is indicative of poverty and plays a fundamental role in predicting surgical outcomes, 12,14 along with urgent/emergent surgeries, 15 frailty, 1,11,13,[16][17][18] and open compared to laparoscopic surgeries. 12 Uninsured or Medicaid patients experience increased rates of emergency procedures, complications, and mortality. 14,19,20 Additionally, the current NSQIP risk calculator underestimates the risk of complications for emergency surgeries 21 and does not provide risk prediction for urgent cases. 15 Failure to account for factors beyond clinicians' control (eg, increased presentation acuity and social risk factors) in value-based medicine when assessing outcomes of high-burden facilities may continue to widen disparities in access, care, and outcomes in vulnerable populations. 6,18 Textbook outcomes (TO) is a composite metric that has been increasingly used to assess surgical outcomes. 22,23 Social risk factors may not consistently affect outcomes; therefore, including multiple outcomes of interest may more comprehensively assess the effects and costs of social risk factors, especially in underinsured/uninsured patients.
We assessed the association of Private, Medicare, and Medicaid/Uninsured (ie, Medicaid, dual-eligible Medicare/ Medicaid, self-pay, indigent programs) health insurance type with complications and cost after colorectal surgeries in an SNH with a large range of SES patients. We hypothesize that after adjusting for frailty and open versus laparoscopic procedure type, increased presentation acuity, measured by presenting with acute serious conditions and urgent or emergent cases in patients with Medicaid/Uninsured insurance type will be associated with higher complications and index hospitalization costs compared to patients with Private insurance.

Study Population and Data
This retrospective cohort study followed STROBE Reporting Guidelines 24 and used local, identified data on all patients undergoing colorectal procedures present in the 2013-2019 American College of Surgeons National Surgical Quality Improvement Program (NSQIP) at a single facility, serving as an academic medical center and SNH. NSQIP registry was used for cohort identification. NSQIP provides standardized definitions of preoperative risk factors and complications. 25 The Institutional Review Board of the University of Texas Health San Antonio approved this study.
Patients presenting with preoperative acute serious conditions (PASC) were defined using 6 NSQIP present at the time of surgery (PATOS) variables and NSQIP variables defining acute renal failure (with or without dialysis required) within 2 weeks before surgery, as previously described (Supplemental Table  1, http://links.lww.com/AOSO/A177 lists the NSQIP variable names). 16 Case status was determined from NSQIP variables with urgent cases being defined as neither elective nor emergency, as determined by "no" responses to the ELECTSURG and EMERGNCY variables. 15 Procedures were categorized as open or laparoscopic surgeries using NSQIP principal Current Procedural Terminology Codes (Supplemental Table 2, http:// links.lww.com/AOSO/A177).

Any and Clavien-Dindo IV 30-Day Complications
Clavien-Dindo classifies complications based on their treatments. 29 We approximated Clavien-Dindo IV (CDIV) complications using the NSQIP variables of postoperative septic shock, postoperative dialysis, pulmonary embolus, myocardial infarction, cardiac arrest, prolonged ventilation, reintubation, or stroke as previously reported. 16 Unplanned reoperations were defined using the NSQIP variable REOPERATION1 as present or absent. Any complication was defined using the CDIV and reoperation NSQIP variables plus an additional 11 NSQIP variables defining postoperative complications.

TO Composite Variable
TOs were defined as surgeries with the absence of 30-day CDIV complications, unplanned reoperations, 30-day mortality after the date of surgery, and 30-day after the date of discharge from the index hospitalization readmissions and emergency department or observations stays (EDOS).

Mortality
Mortality was defined as death within 30 days of the index colorectal surgical procedure. Dates of death were obtained from ACS NSQIP and cross referenced from our data warehouse using the electronic health records of the local health system as well as the Social Security Death Master File.

Insurance Type and Cost Data
The identified, local NSQIP data were merged with electronic health records and managerial accounting data to determine insurance type and the variable cost of the index hospitalization. Insurance type was categorized based upon billing data for the encounter supplemented by EHR data and defined as (1) Private including TRICARE and workers compensation; (2) Medicare; and (3) Medicaid/Uninsured including Medicaid, dual enrollment in Medicare/Medicaid, charity care, self-pay with ≤1% of charges paid, or county indigent care programs (Supplemental Table 3, http://links.lww.com/AOSO/A177). "Other" included encounters billed to the Veterans Administration, Department of Corrections, or self-pay with >1% of charges collected and were excluded (n = 13).
We defined variable costs as related directly to patient care occurring during the encounter, such as supplies/salaries and include direct variable costs that vary directly with the quantity of resources provided for patient care. 30,31 Direct variable costs are accounted primarily using direct measurements from a bottom-up approach rather than calculated estimates derived from charges. Hospital fixed costs, outpatient and professional fees were not included. We used variable costs, as fixed costs are not directly related to patient care and vary between hospitals. 31 The natural logarithm of variable costs was used, as previously described 30 after adjusting costs to 2019 dollars using the Personal Health Care Index. 32

Management of Missing Variables
Cases were excluded due to (1) perineal and transsacral only procedures; (2) missing or inaccurate cost variables; and (3) "Other" insurance type.

Study Outcomes
Clinical outcomes of interest were 30-day unplanned reoperations, any complication, severe/life-threatening CDIV complications, readmissions, EDOS, TO composite variable, and variable costs for the index surgery hospitalization adjusted for RAI, case status, open versus laparoscopic procedure, and insurance type.

Statistical Analysis
Categorical data were summarized using count and percentage and continuous data using mean and standard deviation (SD). Chi-square tests and F-tests were used to test for difference between groups for categorical and continuous variable. Logistic regression analyses were performed for (1) (4) TO subgroup analyses for elective and urgent/emergent cases. Natural logarithms were used to normalize the skewed LOS and variable costs for the index hospitalization, which reduces the impact of extreme values, as previously described. 30,33 Percent change/relative difference was calculated using the exponential function; %change = (e estimated coefficients − 1) × 100. Analyses were performed using R version 4.1.0 (2021-05-18).

Population Characteristics
Our cohort consisted of 1,078 cases of inpatient procedures at a major urban SNH (Fig. 1). Cases (Table 1) were more commonly performed on males (53.1%) and White patients (90.4%) with the majority identifying as Hispanic ethnicity (64.7%). Most cases were performed on patients with Medicaid/Uninsured insurance type (57.4%), followed by Private (23.4%) and Medicare (19.2%). Most cases were performed on robust (60.3%) and normal (27.5%) patients based on RAI scores. Only 9.7% and 2.5% of patients were frail and very frail, respectively, with Medicare patients exhibiting higher rates of frailty. Complication rates were higher in Medicare and Medicaid/Uninsured patients compared to Private.

Increased PASC in Medicaid/Uninsured Patients and Increased Urgent/Emergent Cases in Medicaid/Uninsured and Medicare Patients
Rates of patients presenting with PASC (13.8%) were highest in Medicaid/Uninsured (15.8%) and Medicare (15.5%) patients versus Private (7.5%, P = 0.004, Table 1 PASC variable distribution (Supplemental Table 1, http:// links.lww.com/AOSO/A177) was similar between insurance types, except for sepsis which was lowest in Medicare patients. Rates of urgent/emergent cases were highest in the Medicare (100.0%) and Medicaid/Uninsured (94.9%) patients presenting with PASC.

DISCUSSION
Medicare and Medicaid/Uninsured patients had higher odds of 30-day complications with decreased odds of achieving TO compared to the Private group. Contributing to the worse outcomes in Medicaid/Uninsured patients were the increased odds of presenting with PASC (aOR = 2.02) and undergoing urgent/ emergent surgeries (aOR = 1.80) versus Private. Presenting with PASC was associated with 94.6% rate and aOR = 26.65 of undergoing an urgent/emergent surgery. Consistent with our data, uninsured patients were 3.54 times more likely to undergo emergent colorectal surgeries. 19 Urgent/emergent surgeries had higher odds of complications in this study and in prior publications for urgent 15 and emergent 19,34,35 procedures, suggesting that Medicaid/Uninsured patients present under worse condition than privately insured patients. Urgent cases usually occur after a failed trial of medical management in unplanned hospitalizations. Numerous studies stratify cases into elective and emergent without categorizing urgent case status 19,34,35 or stating how urgent cases were classified. 19,34,35 Urgent cases were more common than emergent cases in all three insurance groups and highest (37.0%) in the Medicaid/Uninsured group (Table 1). Combining urgent and elective cases may disproportionately increase complication rates of vulnerable patients that have higher rates of urgent surgeries.  Open procedures in our study and others had higher odds of complications, 36 reoperations, 36 costs, 37 and lower odds of achieving TO compared to laparoscopic surgeries. Medicaid/ Uninsured patients displayed similar odds of undergoing open procedures as Private after adjusting for PASC and urgent/ emergent cases (Table 2). This suggests that the higher rates of open procedures in Medicaid/Uninsured patients were due to increased presentation acuity and need for urgent/emergent surgeries, as opposed to not providing the alternative of laparoscopic surgery.
Medicaid/Uninsured patients had higher odds of 30-day EDOS (aOR = 4.64) and readmissions (aOR = 1.69). Emergency departments often serve as the primary healthcare source for low-SES patients with limited access to care. 38 The strongest predictor for preventable readmissions was patients undergoing urgent/emergent colorectal procedures. 39 Higher odds of readmission were observed in our Medicaid/Uninsured group with longer LOS given their higher rates of emergent cases, consistent with previous studies. 12,19 Adjusting for patient SES showed similar readmission odds after major surgery between SNH and non-SNH, leading the authors to conclude that differences in patient case mix of low-SES patients, not quality of care, were responsible for higher readmission rates at SNH. 7 Medicare and Medicaid/Uninsured patients had decreased odds of achieving the composite TO measure, regardless of case status. We chose TO as our primary outcome because composite measures often provide more comprehensive assessments of surgical outcomes than single variables. [40][41][42] Our data demonstrate the utility of this approach. While all component variables had increased aOR for Medicare and Medicaid/Uninsured patients compared to the Private insurance group (Table 3), only CDIV complications were significant for Medicare and 30-day EDOS and readmissions for Medicaid/Uninsured patients. Thus, identification of healthcare disparities may be improved by composite variables secondary to the additive effects of each component. Prior studies using TO in colorectal surgery have been limited to oncological procedures. 43,44 Insurance type has consistently been recognized as an independent risk factor for worse surgical outcomes 12,14,20,45 and higher costs. 30,45,46 Consistent with our results, a study on diverticular disease demonstrated the association of Medicaid and Uninsured patients with complicated preoperative presentations and increased mortality. 20 The patients with Private insurance treated in our SNH had a similar complication rate (33.3%) to the previously reported complication rate (32.8%) of privately insured cohorts treated in low-burden hospitals. 47 This suggests that poor outcomes are not a result of lower quality of care in SNH but due to patientlevel differences. 48 While some groups have established SNH provide equitable care, 49,50 high-burden SNH have been associated with increased risk of complications 9,10 and costs. 5 Many colorectal surgery studies assess surgical outcomes across multiple healthcare systems based on safety-net burden 9,12 rather than within a healthcare system, which ultimately compares institutions with vastly different patient populations. The National Academy of Medicine recommends studying the effect of social risk factors within a hospital system, especially one that cares for a range of SES patients, to target factors that distinguish between high and low quality of care. 51 This study is one of the first to assess colorectal surgery outcomes following these recommendations and includes factors influencing the cost of care for patients in these insurance groups.
The Medicaid/Uninsured group was associated with increased adjusted odds of longer index hospitalization LOS and higher variable costs, but both were similar to Private after adjusting for urgent/emergent cases. Major factors impacting the %change in variable costs (Table 6) were urgent/emergent cases (28.19%), open abdominal surgeries (30.10%), and CDIV complications (153.07%).
Our findings suggest insurance type plays a significant role in outcomes and costs after colorectal surgery. Worse outcomes in Medicaid/Uninsured, low-SES patients were driven by increased presentation acuity, measured by increased odds of presenting with PASC and undergoing urgent/emergent surgeries, driving the increased complications and costs, consistent with a previous publication. 19 Healthcare providers are increasingly being held accountable for quality, outcomes, and cost of care. 52,53 Decreasing rates of nonelective surgeries is a potential target for policy change. Medicaid expansion was associated with a 1.8-percentage point increase in the probability of an early, uncomplicated presentation for several surgical conditions compared to states that did not expand Medicaid. 54 Improving access to health care to decrease the incidence of PASC and urgent/emergent operations may be a better approach to improving surgical outcomes and reducing costs than P4P programs. A Cochrane review 55 concluded that hospital P4P programs have an uncertain impact and effects on patient outcomes were at most small on quality of care, equity, or resource use. Further studies should assess the impact of improved healthcare access for vulnerable patients on reducing urgent/emergent surgeries, complications, and costs. Numerous studies indicate that SNH and academic medical centers are disproportionately penalized in current value-based medicine programs. 6,8,18 Integration of socioeconomic risk factors into evaluation of high-and low-burden hospitals could improve the distribution/allocation of resources, improving resources to SNH and mitigating disparities potentially propagated by P4P programs.

Limitations
This study is a retrospective review and cannot establish causal relationships. NSQIP provides a representative sample of surgeries but does not include all procedures performed at our institution. High complication and mortality rates may occur in cases performed for palliation, rather than for the purpose of extending life, but these data do not clearly define procedures that were performed specifically for palliation. While we included frailty, laparoscopic procedures and case status variables, multiple other variables could have been included.

CONCLUSIONS
Medicaid/Uninsured insurance type was associated with decreased odds of achieving TO and increased odds of presenting with PASC and urgent/emergent cases, driving higher odds of complications, and index hospitalization costs. This suggests that factors beyond the surgeons' control, such as increased presentation acuity and insurance type, impact surgical outcomes. Socioeconomic factors profoundly affect patient outcomes, which can account for the higher complication rates 47 and costs of SNH serving a higher proportion of low-SES patients. Improving access to health care could provide a more significant impact on patient outcomes and decrease index hospitalization costs, by decreasing the incidence of urgent/emergent colorectal surgeries, particularly in low-SES, vulnerable patients.