Area Deprivation, Fragmented Care, and Colectomy Case Acuity in the Veterans Health Administration

Abstract

BACKGROUND:

Colectomy for benign or malignant disease may be elective, urgent, or emergent. Data suggest successively worse outcomes for nonelective colectomy. Limited data exist regarding the contribution of high area deprivation index and care fragmentation to nonelective colectomy.

OBJECTIVE:

Determine the association between area deprivation and nonelective colectomy in the Veterans Health Administration and assess whether accounting for differences in care fragmentation alters the association across indications and for benign and malignant conditions separately.

DESIGN:

Retrospective cohort with multivariable multinomial logit models to evaluate associations between high-deprivation care fragmentation and the adjusted odds of nonelective colectomy. We calculated total, direct, and indirect effects to assess whether the association varied by levels of care fragmentation.

SETTING:

Veterans receiving care in the private sector and Veterans Health Administration.

PATIENTS:

Veterans aged 65 years or older undergoing colectomy between 2013 and 2019 were included.

MAIN OUTCOME MEASURES:

Colectomy case acuity.

RESULTS:

We identified 6538 colectomy patients, of whom 3006 (46.0%) were identified for malignancy. The odds of emergent colectomy were higher for patients in high-deprivation areas when the indication was for benign pathology (adjusted OR 1.51; 95% CI, 1.15–2.00). For malignant indications, there was no association between high deprivation and nonelective colectomy. More fragmented care was associated with higher odds of urgent and emergent colectomy for both benign and malignant indications, but the association between deprivation and nonelective colectomy did not vary by care fragmentation.

LIMITATIONS:

Inherent to large administrative retrospective databases.

CONCLUSIONS:

Veterans living in high-deprivation areas are at higher risk for emergent colectomy for benign conditions. Care fragmentation is also associated with a higher risk of emergent colectomy across indications. Efforts to reduce care fragmentation and promote early detection of IBD and diverticular disease in high-deprivation neighborhoods may lower the risk for nonelective colectomy in veterans. See Video Abstract.

Colectomy is one of the most common surgical procedures in the United States, with more than 300,000 annual hospitalizations.¹ Various colon and rectal pathologies, including malignancy, premalignant polyps, IBD, benign neoplasms, diverticular disease, obstruction, and volvulus, can lead to surgical resection. Some patients present with severe symptoms and require nonelective surgery, whereas others are identified early and receive nonsurgical treatment or undergo elective surgical procedures. Urgent and emergent surgery both fall under the umbrella of “nonelective” but are distinct entities and reflect important differences in acuity.² Nonelective and elective presentations have different pathways for benign and malignant colon pathology. For colorectal cancer, risk stratification and screening modalities identify malignant or premalignant entities so that endoscopic therapy or elective surgery can be used. For benign conditions, such as IBD or diverticular disease, early identification also leads to medical or endoscopic therapy, either eliminating the need for a colectomy or reducing the urgency. These pathways are summarized broadly in the top row of Figure 1. If screening is not adequate and symptoms are not evaluated, the resulting colectomy will tend to be more urgent or emergent (bottom row, Fig. 1).

Figure 1. Conceptual Model for non-elective colectomy.

CRC, colorectal cancer

Social risk factors produce health disparities, creating worse patient outcomes, including for patients undergoing colectomy.^3-6 Composite measures of area-level socioeconomic stress, such as the area deprivation index (ADI), encompasses multiple domains, such as income, education, and environment, and data suggest that ADI is an important risk factor for poor health.^7-10 However, data are limited on whether living in high-deprivation areas limits access to health care and opportunities for early identification of colorectal pathology to plan for elective surgery or avoid surgery if possible.

Care fragmentation is another risk factor for poor health outcomes.¹¹ Care fragmentation and high ADI might be comorbid, but their association remains an open question. High ADI might lead to emergent colectomy via fragmented care with unobserved mechanisms, including poor communication, delays in care, and redundancy, eventually leading to advanced presentation necessitating nonelective colectomy.

We aimed to determine the association between high area deprivation and urgent or emergent colectomy in the Veterans Health Administration (VHA) and to assess whether this association is accounted for by differences in care fragmentation (care within and outside of VHA). We also aimed to determine whether care fragmentation is associated with urgent or emergent colectomy. We hypothesized that high area deprivation is associated with nonelective colectomy and care fragmentation and that the association between high deprivation and nonelective colectomy is explained in part by adjusting for care fragmentation. We also hypothesized that care fragmentation on its own is associated with nonelective colectomy.

MATERIALS AND METHODS

Design and Population

We conducted a retrospective cohort study of VHA patients older than 65 years undergoing colectomy using the Veterans Affairs Surgical Quality Improvement Program (VASQIP) national database from the years 2013 to 2019. VASQIP is the surgical quality database in VHA and includes nurse-abstracted clinical data from the electronic medical record, including preoperative risk factors, diagnoses, operative procedures, and 30-day morbidity and mortality.^12,13 VASQIP has been used in a wide breadth of health services research efforts to study surgical quality for GI and colorectal surgery in VHA.^14-19 Patients were identified using procedure codes for colectomy found in Appendix 1 at http://links.lww.com/DCR/C461. Veterans were excluded if they 1) lacked Medicare enrollment 1 year before surgical admission or enrollment in a Centers for Medicare and Medicaid Systems (CMS)–managed care plan, 2) did not use VHA services during the 2 years before surgery, 3) had a missing or unreliable address, and 4) were missing clinical variables for measuring risk. A complete cohort waterfall diagram may be found in Figure 2.

Figure 2. Flow diagram of the study cohort.

CPT, Current Procedural Terminology; HMO, Health Maintenance Organizations; SDoH, social determinants of health; VHA, Veterans Health Administration

Data Sources

We created an expansive data set, including linkages to several different sources in and out of the VHA. Diagnosis (malignant and benign), procedure, and enrollee demographics were included from the VHA Corporate Data Warehouse using the International Classification of Disease, 9th or 10th Revision, Clinical Modification. Diagnosis codes were used to identify malignant and nonmalignant diseases leading to colectomy (see Appendix 2 at http://links.lww.com/DCR/C462). Additional data sources included the following: 1) the VHA Program Integrity Tool and Veterans Affairs (VA) fee-based domain provided claims for care provided by community providers under the VHA Care in the Community program; 2) CMS enrollment and claims data provided Medicare enrollment, mortality, and fee-for-service Medicare claims among VHA enrollees; and 3) the VA Planning Systems Support Group provided geocoded patient addresses for assigning rurality and neighborhood level area deprivation. Scrambled Social Security numbers were crosswalked to patient medical record numbers (within the VHA, Integrated Control Numbers) and real Social Security numbers. Patients were excluded when this scrambled Social Security number was associated with more than 1 patient.

Variables

Age at the time of surgery, sex, race, and ethnicity were determined from VHA administrative data supplemented with CMS enrollment information. The Risk Analysis Index (RAI) and the Gagne score measured patient frailty and comorbidity, respectively.^20,21 The ADI combines Census-derived measures of income, education, employment, and housing quality to summarize neighborhood socioeconomic conditions. It allows neighborhoods, defined as census block groups, to be percentile ranked by socioeconomic disadvantage. ADI was dichotomized according to prior literature using the 85th percentile as a threshold.²² Two measures of care fragmentation were defined. The first was the non-VHA Usual Site of Care (USC), which was defined as the proportion of days receiving non-VHA care during the 12 months before surgical admission relative to total days of care analyzed in quintiles. The second measure was the Sequential Continuity of Care (SCOC). The SCOC considers the order of visits, not just their concentration among providers. It represents the number of handoffs that exist between providers. It equals the fraction of sequential visit pairs at which the same provider is seen, that is, the same provider being seen at both the previous and current visits.²³ For consistency with the USC, we reversed the direction of the SCOC so that it was calculated as the fraction of sequential visits that occur in different health systems. For both USC and SCOC, higher numbers represent greater fragmentation.

We captured preoperative acute serious conditions; medical issues such as ventilator use, pneumonia, coma, sepsis, blood transfusions, or acute renal failure that typically arise in the setting of an acute surgical problem can increase the urgency for operative intervention and the risk of complication or death after surgery.²⁴ In private health sector analyses, this variable is often used as a measure of medical acuity, and work from our group is the first to incorporate this variable into VHA data analyses.^2,24-26 Our outcome was surgical case status classified in 3 levels as 1) elective, 2) urgent, or 3) emergent. Emergent cases were defined as those flagged as “emergent” in the VASQIP database. Cases were considered urgent if VHA surgical scheduling data indicated they were scheduled as 1) “urgent,” 2) “add-on,” 3) “stand-by,” or 4) “emergency” but with a "non-emergent” VASQIP flag. Prior data support the need for such a distinction between purely elective, urgent, and emergent.^2,27 All other cases were considered elective unless the scheduling type was missing.

Statistical Analysis

Descriptive summary statistics were presented in numbers and percentages for categorical variables (sex, race, malignancy, and case status) and in means and SDs for continuous variables (age, RAI, Gagne comorbidity score, and care fragmentation). The Pearson χ² test and ANOVA were used to compare differences in categorical and continuous variables between patients residing in census tracts with high ADI (85 or more) or low ADI (less than 85).

Multivariable multinomial logit models were created to assess the association of ADI with the odds of urgent or emergent colectomy relative to elective colectomy. Models controlled for year of surgery, age, sex, race, RAI, Gagne comorbidity score, malignancy status, and presence of a preoperative acute serious condition. Models also controlled for care fragmentation, measured alternatively as the proportion of days of care during the 12 months before surgery that occurred in a VHA facility, or using the Sequential Care Index.²³ Models were also generated separately for patients with and without a diagnosis of colorectal cancer, given that the pathways for colorectal cancer and benign disease requiring colectomy are likely different. All models used robust standard errors to account for the clustering of patients within hospitals. Finally, analyses were conducted to decompose the relationship between ADI and case status into the natural direct effect and natural indirect effect of ADI on the probability of emergent or urgent colectomy.²⁸ We hypothesized that ADI impacts case status both directly and indirectly through impacts on care fragmentation. We additionally performed a sensitivity analysis using ADI cutoffs of 80 and 90 for high deprivation to assess variability using different cutoffs for ADI. We also conducted post hoc exploratory analyses to test for associations between high-deprivation areas, care fragmentation, and fecal diversion alone without colectomy and included these findings in the supplemental materials. All p values were reported with the level of significance set at a p value of <0.05. Data analysis was performed using SAS software, SAS Enterprise Guide version 8.3 (SAS Institute Inc, Cary, NC).

RESULTS

We identified 6538 patients undergoing colectomy, of whom 3006 (46.0%) were identified for malignant indications. Patients were nearly all men and the average age was 72 years (Table 1). Patients living in high-deprivation areas were less likely to be White (72% vs 82%, p < 0.001) but had similar levels of frailty, medical comorbidities, malignant indications for surgery, and care fragmentation (Table 1). More than 60% of the colectomy procedure codes were billed as either open or laparoscopic partial colectomy with anastomosis or colectomy with ileocolostomy (see Appendix 1). A total of 689 Hartmann procedures (10.5%) were performed either open or laparoscopically.

Table 1.

Patient Characteristics

	Low Deprivation* (n=5427)	High Deprivation** (n=1111)
Age, mean(SD)	72.92(6.50)	72.49(6.25)
Male, N(%)	5316(97.95)	1079(97.12)
Race
Black	642(11.83)	257(23.13)
Hispanic	236(4.35)	45(4.05)
Other	76(1.40)	14(1.26)
White	4473(82.42)	795(71.56)
RAI, mean(SD)	28.99(6.65)	28.94(6.70)
Gagne Score, mean(SD)	4.05(3.49)	4.13(3.56)
PASC	300 (5.53)	62 (5.58)
Percent of Total Days in Care outside VA (mean, SD)	0.74(0.36)	0.76(0.35)
Indications for colectomy was malignancy, N(%)	2685(49.47)	549(49.41)
Case status, N(%)
Elective	4215(77.67)	833(74.98)
Urgent	610(11.24)	122(10.98)
Emergent	602(11.09)	156(14.04)

ADI, Area Deprivation Index; PASC, Preoperative Acute Serious Condition; RAI, Risk Analysis Index; VA, Veterans Affairs

^*ADI < 85

^**ADI ≥ 85

For all colectomy indications (benign and malignant), there was an association between high area deprivation and emergent surgery (adjusted OR [aOR] 1.45; 95% CI, 1.11–1.89) but not urgent surgery. When stratified by indication, there was no association between high area deprivation and urgent or emergent colectomy for malignancy, but high area deprivation was associated with a higher risk of emergent colectomy for benign conditions (aOR 1.51; 95% CI, 1.15–2.00; Table 2). There was no association between ADI and care fragmentation (coefficient –0.17; 95% CI, –0.40 to 0.06; p = 0.147). Results from the sensitivity analysis are included in the supplemental materials. Overall, the odds of emergent colectomy increases as the ADI cutoff increases, with odds of 1.29, 1.45, and 1.59 using cutoffs of 80, 85, and 90, respectively. In post hoc exploratory analyses of fecal diversion alone, there was an association between high-deprivation areas and higher risk for emergent diversion for malignant disease. There was a lower risk of urgent diversion in benign disease and no association with emergent diversion (see all Supplemental material).

Table 2.

Association of ADI with urgent or emergent colectomy based on multinomial logit model

	OR (95% CI; p-value)	OR (95% CI; p-value)
	Urgent vs Elective	Emergent vs Elective
All indications for colectomy*	1.06 (0.84-1.34; p=0.61)	1.45 (1.11-1.89; p=0.006)
Colorectal Cancer**	1.23 (0.85-1.78; p=0.27)	1.19 (0.79-1.78; p=0.41)
Other non-malignant indications**	0.94 (0.71-1.23; p=0.64)	1.51 (1.15-2.00; p=0.004)

ADI, Area Deprivation Index

^*Models adjusted for year, age, sex, race, RAI, Gagne score, PASC, cancer indication, and percent of total care in the VA

^**Models adjusted for year, age, sex, race, RAI, Gagne score, PASC, and percent of total care in the VA

There was an association between more fragmented care and urgent and emergent colectomy. For every 10% increase in non-VHA care, there were 9% higher odds of emergent colectomy (aOR 1.09; 95% CI, 1.07–1.11) and a modest 3% higher odds of urgent colectomy (aOR 1.03; 95% CI, 1.00–1.06). These findings were generally similar for benign and malignant indications (Table 3). Results were similar using the Sequential Care Index (see Appendix 3 at http://links.lww.com/DCR/C463).

Table 3.

Association of fragmentation (per 10% increase in non-VA care) with urgent or emergent colectomy based on multinomial logit model

	OR (95% CI; p-value)	OR (95% CI; p-value)
	Urgent & Emergent versus Elective	Emergent versus Urgent & Elective
All indications for colectomy*	1.03 (1.00-1.06; p=0.067)	1.09 (1.07-1.11; p<.001)
Colorectal Cancer**	1.02 (0.98-1.07; p=0.27)	1.09 (1.02-1.16; p=0.007)
Other non-malignant indications**	1.03 (1.00-1.07; p=0.043)	1.09 (1.06-1.12; p<.001)

^*Models adjusted for year, age, sex, race, RAI, Gagne score, PASC, and cancer indication

^**Models adjusted for year, age, sex, race, RAI, Gagne score, and PASC

Total, direct, and indirect effects of care fragmentation are presented in Figure 3. For all indications, the majority of the effects of high area deprivation occurred directly, with little (1%–2%) impact through care fragmentation across indications stratified by benign and malignant disease. The magnitude and direction of indirect effects using the Sequential Care Index were similar to care fragmentation measured by the proportion of days of care during the 12 months before surgery.

Figure 3. Total effects, direct effects and indirect effect of care fragmentation on urgent and emergent colectomy.

DISCUSSION

To our knowledge, these data are the first to demonstrate an association between high area deprivation and urgent or emergent colectomy in a large national cohort of the VHA. Our data demonstrate this association in nonmalignant indications for colectomy (eg, diverticulitis, IBD, benign polyps) but not for colorectal cancer. This association was not sensitive to levels of care fragmentation. However, there was an association between fragmented care and urgent and emergent colectomy across benign and malignant conditions.

Few data provide insight into the social risk factors that may lead to nonelective colectomy. Using the National Inpatient Sample (NIS), Schlottmann et al²⁹ analyzed more than 140,000 hospitalizations between 2008 and 2015 for colorectal cancer and found that insurance type (Medicaid and Medicare compared to private), race (Black patients compared to White patients), and household income (low compared to high) were associated with emergent surgery and that emergent surgery was associated with significant postoperative complications and death.²⁹ Similar findings have also been reported in the National Surgery Quality Improvement Program.^30,31 The NIS only contains data from the index hospitalization, whereas our data tracks 1 year of preoperative care. In addition, the NIS lists elective or emergent admission, whereas our data use elective, urgent, or emergent surgery. Emergent admissions in the NIS may have an urgent (not emergent) colectomy, and our 3-level outcome may be difficult to interpret alongside the 2-level NIS outcome (admission, not surgery). Taken together, for colorectal cancer specifically, individual social risk factors that are associated with emergency surgery at the individual level may be confounded by area, according to our findings.

Data from the Medicare Provider Analysis and Review file from 2014 to 2018 demonstrated that patients undergoing common surgery from high-deprivation areas going to low-quality hospitals had higher adjusted odds of 30-day mortality compared to patients from low-deprivation areas going to high-quality hospitals.³² This study included 5 common surgical procedures, including colon resection (highest proportion of patients) along with appendectomy, coronary artery bypass grafting, cholecystectomy, and incisional hernia. Our data align with this analysis and add that separating benign and malignant conditions may add nuance to understanding these pathways. In another study at a safety-net hospital in San Antonio, Texas, patients living in high-deprivation areas had higher hospital readmission after colorectal surgery than patients living in low-deprivation areas.³³ These data also support the finding that living in high-deprivation areas places patients at risk for adverse outcomes in colorectal surgery, including nonelective surgery.

There are limited data on care fragmentation for dual-use veterans with GI conditions requiring surgery. Analyzing patients with IBD in the Corporate Data Warehouse between 2002 and 2014, Cohen-Mekelburg et al³⁴ used the Bice-Boxerman Continuity of Care Index to show that veterans with low levels of care coordination were at higher risk of outpatient flares requiring steroids (treatment failures), hospitalizations, and surgery. Other disease entities such as diabetes, congestive heart failure, and poor mental health show similar but sometimes conflicting results.^35-37 Our data suggest that care fragmentation is associated with urgent or emergent colectomy for both benign and malignant disease conditions. However, our findings do not suggest that the association between area deprivation and urgent or emergent colectomy is sensitive to care fragmentation. This makes sense as social risk factors that make it difficult to access 1 health system may similarly make it difficult to access another. Therefore, coordination between health systems may not be a challenge in this scenario if access to any health care is limited.

Findings from our study may have important implications for the management of benign conditions where elective surgery may avoid an emergent operation, such as with diverticular disease or IBD. In a single study of 947 patients at a single center hospitalized with acute diverticulitis, Hamdan et al³⁸ did not demonstrate an association between ADI and inpatient outcomes such as length of stay and need for procedural intervention. However, only 6% of this cohort underwent surgery compared to our cohort, where all patients underwent a colectomy. This distinction highlights the fact that when it comes to diverticular disease, few patients in the acute presentation are likely to need urgent or emergent colectomy at the time of initial presentation. By using this group as the denominator, it may obscure the association of the long-term management of diverticular disease that involves important decisions about who may benefit from elective colectomy. Knowing that living in a high-deprivation area is associated with a higher risk of emergency colectomy (all benign conditions, not only diverticular disease) may be an important consideration when counseling patients about pursuing elective resection. In our sensitivity analysis of the various ADI cutoffs, the odds of emergent colectomy increase as the ADI cutoff increases, suggesting that even among areas of relatively high deprivation, there nevertheless exists a discriminatory effect such that patients in the most deprived areas are most likely to incur emergent surgery. This pattern persisted for nonmalignant indications but not for malignant ones, as in the main analysis.

The findings of this study align with the American Society of Colon and Rectal Surgery clinical practice guidelines on left-sided colonic diverticular disease, which state that the decision for elective surgery should be individualized, taking into account the entirety of patient circumstances.³⁹

These data did not demonstrate an association between high ADI and urgent or emergent colectomy in colorectal cancer. The interpretation of this finding should consider that another important outcome to measure in colorectal cancer is the colectomy that never happens due to screening. Obviously, this outcome is impossible to identify, not to mention the denominator would have to include patients who were eligible for screening in addition to those who have surgery. It is difficult to hypothesize how and if the estimates would change with an expanded numerator and denominator, but it is nonetheless important to remain vigilant about improving access to colorectal cancer screening and treatment for veterans in high-deprivation areas. ADI was not related to urgent or emergent surgery for malignant cases, regardless of the cutoff used in the sensitivity analysis.

This study has several important limitations. First, all large, retrospective database investigations are limited by the quality of the data reporting, but our approach leveraged the longitudinal character of data not shared in prior studies of strictly cross-sectional data. Our regression analyses included variables specific to the pathways we hypothesized, but unmeasured confounding can influence results. Specifically, race and ADI are correlated, and statistical adjustment for race was used to address this fact. Even so, structural racism inherent in where we live and the differences between those locations are difficult to quantify and completely adjust for in statistical analyses. We did not include colonic stenting as an outcome because it would heavily complicate coding and ascertainment of the outcomes as it relates to nonelective colectomy, particularly as it relates to colorectal cancer. There is a possibility that more patients with colorectal cancer underwent placement of stents as a bridge to elective surgery, where this was not common for benign conditions. This type of event would potentially obscure the relationship between area deprivation and emergency presentation requiring surgery. However, most commonly, the placement of colonic stents in patients who are candidates for surgery will undergo surgery at the initial hospitalization, which our data collection captured. Given we did not collect data on colectomies performed at non-VHA hospitals, it is also possible that patients with symptomatic colorectal cancer requiring surgery may preferentially be cared for at non-VHA facilities. Given there is some variability between clinical services of VHA hospitals, these findings are not generalizable to VHA hospitals doing little or no colon surgery. Also, these findings may not be generalizable to patients older than 65 years receiving care outside of VHA facilities.

CONCLUSIONS

Despite these limitations, our data, including more than 6000 colectomies nationwide, demonstrate that high area deprivation is associated with nonelective colectomy in VHA hospitals and that this association is most clearly present among noncancer colorectal pathology. This association is not sensitive to fragmented care; however, fragmented care is associated with urgent or emergent colectomy across benign and malignant conditions. These data support additional resources for care coordination among dual-use veterans for colorectal pathology that may require surgery. These findings may also influence the patient-centered discussions around elective colon resection for benign pathology, such as diverticular disease, when patients live in highly deprived areas.

Supplementary Material

Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML and PDF versions of this article on the journal’s website (www.dcrjournal.com).