Abstract
PURPOSE:
Post-operative complication rates may vary among racial and/or ethnic groups and have not been previously described in individuals with spina bifida (SB) undergoing urologic surgery. The aim of this study was to compare in-hospital complication frequencies of individuals with SB following urologic surgery by race/ethnicity.
METHODS:
The Nationwide Inpatient Sample was used to identify pediatric patients with SB who underwent inpatient urologic procedures. A pediatric cohort (<18 years old) with SB that underwent urologic surgery were assessed. All analyses report weighted descriptive statistics, outcomes, and race/ethnicity was the primary predictor variable. The primary outcome of interest was post-operative complications which were defined using NSQIP ICD-9 code definitions. Secondary analysis included length of stay (LOS), and encounter cost was estimated using the cost-to-charge ratio files provided by the Healthcare Cost and Utilization Project.
RESULTS:
The unadjusted model showed no differences in complications, LOS, and cost. In the adjusted model there were no differences in complications, LOS, and cost between Black and White encounters. However, Hispanic ethnicity was associated with a 20%(95%CI: 4–40%) increase in LOS and 18%(95%CI: 2–35%, p = 0.02) increase in cost compared to White encounters.
CONCLUSION:
There was no evidence of variation for in-hospital complication rates among racial/ethnic groups undergoing urologic surgery. Hispanic ethnicity was associated with higher costs and longer LOS in pediatric SB encounters.
Introduction
Spina bifida (SB) is the most common permanently disabling birth defect in the US [1, 2]. This population requires routine urologic evaluation including surveillance of bladder function and upper tract evaluation [3, 4]. In order to treat or avert urologic complications, many individuals with SB undergo urologic reconstructive surgery such as bladder augmentation, catheterizable channel creation, or a bladder outlet procedure. However, these complex procedures carry a significant risk of in-hospital, post-operative complications. In a recent multi-institutional case series, the complication and readmission rates for SB patients undergoing laparotomy for urologic indications were 92%and 42%, respectively [5].
Previous reports have demonstrated that surgical complication rates may vary among racial and/or ethnic groups; specifically, that racial/ethnic minorities may be at higher risk of complications [6, 7]. In the setting of high overall complication rates among individuals with SB, a better understanding of potential effects of race and/or ethnicity would be helpful. Understanding the relative importance of these non-modifiable risk factors for complications would allow clinicians to identify at-risk populations and facilitate improvement of healthcare delivery at the system and patient level.
Thus, the primary objective of this study was to compare the odds of an in-hospital, post-operative complication among White, Black and Hispanic individuals with SB who underwent urologic surgery after accounting for available non-modifiable risk factors. The secondary objective was to compare length of stay (LOS) and cost among the same groups. The hypothesis was that the odds of in-hospital, post-operative complications, mean LOS, and mean inflation-adjusted costs would be lower in White SB encounters compared to their Black or Hispanic counterparts.
Methods
Data source
The Nationwide Inpatient Sample (NIS) is an all-payer database managed by the Healthcare Cost and Utilization Project (HCUP) and sponsored by the Agency for Healthcare Research and Quality. NIS data are drawn from a 20%stratified probability sample of US hospitals based on five hospital characteristics including ownership status, number of beds, teaching status, urban/rural location, and geographic region. NIS includes post-stratification discharge weights that can be used to make national estimates.
Cohort
The cohort assessed for this study was comprised of pediatric encounters (0–18 years) in the NIS database between 1998 and 2014 (inclusive) with an ICD-9 diagnosis code for SB (741.X and 756.17), a non-missing race/ethnicity variable and an ICD-9 procedure code for inpatient urologic surgery (ureteral reimplant [56.74, 59.3], ureteroureterostomy [56.75], pyeloplasty [55.86, 55.87], nephrectomy [55.5], partial/heminephrectomy [55.4], appendico-vesicostomy [57.88], enterocystoplasty [57.87], cystectomy [57.6], vesicostomy [57.21], or bladder outlet procedure [59.4, 59.5, 59.6]). Because patients frequently undergo multiple procedures during the same inpatient encounter (e.g., enterocystoplasty and bladder outlet procedure), not all of which will necessarily be correctly coded, urologic surgeries were thus analyzed in aggregate rather than by any given specific procedure.
Outcomes and variable definitions
The primary outcome, in-hospital post-operative complication, was defined according to the NSQIP ICD-9 code (Appendix A) and was dichotomized to reflect the presence of at least one in-hospital, post-operative complication. Secondary outcomes were inpatient LOS (days) and cost of encounter.
Race and ethnicity are combined into a single variable in NIS, with the following possible values: White, Black, Hispanic, Asian/Pacific Islander, Native American, missing, or other. Due to data sparseness, missing, Asian/Pacific islander, and Native American were grouped together as “other”, which was not included in the analysis due to low incidence of SB in this population (∼5%of the cohort total). Thus, the primary predictor of interest was non-Hispanic White, non-Hispanic Black, or Hispanic.
The cohort was examined by age, gender, primary payer (public, private, or other), hospital bed size (small, medium or large), hospital type (metropolitan teaching, metropolitan non-teaching and non-metropolitan), Van Walraven comorbidity Index (VWI) and renal failure. The VWI is a calculated score based on medical comorbidities that predicts in-hospital mortality. The score ranges from –19 to 89 with higher scores associated with an increased risk for hospital mortality. VWI can also be used as a surrogate for overall health, which was utilized in this manner for this study in order to adjust for baseline patient comorbidity [8]. Because renal failure can simultaneously act as an outcome, a confounder, and a surgical indication, it was defined separately from other components of VWI as the presence of at least one of the following ICD-9 diagnostic codes: 58.4x, 58.5x, 58.6x, V42.0, V45.1, V56.0-V563.2, V56.8, V45.11-V45.12 or at least one of the following ICD-9 procedural codes: hemodialysis (39.95), peritoneal dialysis (54.98) or kidney transplant (55.61, 55.69).
Cost of encounter was estimated using the cost-to-charge ratio (CCR) files provided by HCUP. This provides a hospital level adjustment factor that is multiplied by the total charge of an encounter. CCR was only released for 2001 to 2014; thus, 1998–2000 were excluded from the cost analysis. Inflation was adjusted using the Consumer Price Index (CPI)-standard index. The CPI standard was used because the medical CPI may not adequately account for technological improvement, quality change, and improved health outcomes [9]. CPI values were obtained from the Bureau of Labor Statistics. Inflation was calculated based on 2014 USD. Estimated cost outcome was defined as inflation-adjusted cost. Cost was not adjusted for LOS.
Statistical methods
As recommended by HCUP, weighted descriptive statistics were used to describe the demographic characteristics for both cohorts. Per HCUP data use agreements, cell counts less than 15 were censored. For simplicity, weighted encounters are referred to as ‘encounters,’ but all analyses presented are weighted.
For the primary outcome, an initial weighted bivariate analysis was performed using the Wald Chi Square test to compare differences in proportions of in-hospital post-operative complications by race/ethnicity. A weighted logistic regression model was then used to account for both survey weighting and correlation within hospitals by allowing the variance of hospitals within strata to change by year. The primary predictor of this multivariable model was race/ethnicity and the primary outcome was at least one in-hospital, post-operative complication. The model was adjusted for age, gender and primary payer. Since the primary hypotheses were focused only on differences between White, Black and Hispanic encounters, only these encounters were considered for the unadjusted analysis. For adjusted models all encounters were included.
Similarly, secondary outcomes were initially analyzed using an unadjusted weighted ANOVA to test for differences by race/ethnicity in means for inpatient LOS and inflation-adjusted costs. These outcomes were then separately modeled using weighted generalized estimating equations (GEEs), which also accounted for both survey weighting and correlation within hospitals. For LOS, GEE was employed with a negative binomial distribution and log link function. For inflation-adjusted cost, GEE was employed with a Gamma distribution and log link function. The primary predictor for all models was race/ethnicity, adjusted for age, gender and primary payer.
Using contrasts, differences were tested between White and Black encounters and also between White and Hispanic encounters. This resulted in two tests per model. To account for multiple testing, Bonferroni correction was utilized which adjusted the alpha to 0.025 as criteria for statistical significance. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC).
Results
Demographics & cohort characteristics
There was a total of 6,445 pediatric encounters between 1998 and 2014 with a diagnosis of SB, a urological surgical intervention, and documented race (Table 1). The mean age of the cohort was 9.4 (SE: 0.15) years. Additional demographics of the cohort included the following: 57.8%were female, 58.2%White, 28%Hispanic, 8.6%Black, and 42.8%had private insurance. The mean, median, and range VWI were 1.5 (SE: 0.11), –0.35 (IQR: –0.7 to 1.7), and –7 to 18. Renal failure was present in 3.9%of encounters. Also, 56.6%of encounters were seen in a large bed-count hospital, and 91.7%were seen in urban teaching facilities.
Cohort demographic features
Cohort demographic features
When comparing the unadjusted primary outcome of interest by White, Black, and Hispanic, there was no difference in proportion of in-hospital post-operative complications (Table 2; 19.7%vs 18.7%vs 22.3%respectively). After adjusting for covariates (Table 3), there was no difference in the odds of in-hospital post-operative complications for Black compared to White (OR: 0.86 [95%CI: 0.51, 1.45]) nor Hispanic compared to White (OR 1.10 [95%CI: 0.70, 1.73]).
Unadjusted Outcomes for Pediatric SB Encounters
Unadjusted Outcomes for Pediatric SB Encounters
1Wald-Chi-Square. 2Weighted ANOVA.
Model results for in-hospital post-operative complications
Models adjusted for age (continuous), gender and insurance type (private vs. other/missing).
When comparing the unadjusted secondary outcomes by race/ethnicity (Table 2), there were no differences in mean inpatient LOS (8.0 [SE: 0.32] vs 8.78 [SE: 0.55] vs 10.27 [SE:1.00] respectively) or mean inflation-adjusted cost ($28,831 [SE: $1,258] vs $29,281 [SE: $2,350] vs $32,966 [SE: $2,270] respectively).
After adjusting for covariates (Table 4), there was no evidence of a difference in inpatient LOS between Black versus White encounters (RR: 1.10 [95%CI: 0.96–1.28]). There was evidence of an increase in inpatient LOS for Hispanic compared to White encounters (RR 1.20, 95%CI: 1.04–1.40, p = 0.02).
Model results for inpatient length of stay
Model results for inpatient length of stay
Models adjusted for age (continuous), gender and insurance type (private vs. other/missing).
After adjusting for covariates (Table 5), there was no evidence of a difference of inflation-adjusted cost between White compared to Black (RR: 1.07 [95%CI: 0.91–1.27]). There was evidence of an increase in inflation-adjusted cost for Hispanic compared to White encounters (RR 1.18, 95%CI: 1.02–1.35, p = 0.02).
Model for estimated inpatient cost
Models adjusted for age (continuous), gender and insurance type (private vs. other/missing).
In this administrative database analysis, Hispanic individuals with SB had an increase in inflation-adjusted cost and a longer LOS when compared to their White counterparts. The higher cost associated with the Hispanic pediatric encounters may be associated with the longer LOS. There was insufficient evidence to identify disparities of in-hospital post-operative complications, inflation-adjusted cost, or LOS between White and Black SB encounters.
Understanding the combination of disease-specific risk stratification and non-modifiable risk factors, including race and/or ethnicity, is important to provide more effective healthcare delivery in individuals with SB and to address inequalities on a system level. These findings demonstrate variation among racial/ethnic groups, specifically between Hispanic and White encounters regarding cost and LOS, although not in-hospital complications. This finding is particularly relevant given the increased incidence of SB in Hispanic populations, implying that Hispanic individuals may be disproportionately impacted both by SB and by complications from the surgical management of SB-associated neurogenic bladder. These findings echo a similar trend within the literature [7, 10–12]. Bloo et al. performed a meta-analysis comparing operative mortality and complications between minority groups and White patients in the general population. They found that minority groups were associated with longer post-operative hospital stays [6]. They also reported a higher risk of mortality and overall complication rate, which were not specifically addressed by the current study. The mechanisms driving these differences are probably complex and multifactorial. Explanations for racial/ethnic differences in surgical patients in other studies include socioeconomic status, presurgical risk, cultural factors, and biological/genetic factors [6, 10]. Likewise, it seems likely that other nonclinical factors, including geographic or population density cost differences, may be playing an as-yet unexplained role here.
This trend of racial/ethnic differences in post-operative outcomes has been previously reported in other pediatric urologic studies. Chu et al. noted racial differences in 30-day complication rates following urologic surgery in pediatric patients using the National Surgical Quality Improvement Program-Pediatrics (NSQIP-P). They found that non-Hispanic Black patients were at higher risk of 30-day complications and hospital-acquired infections compared to their non-Hispanic White counterparts [10]. That study included all urologic surgery patients under age 17 while the current study focused only on SB encounters. Additionally, it was a prospective study that bundled the specific post-operative complications with the procedure. It is possible that these findings in the general pediatric urology population are not as prevalent within the SB population. One explanation could be the organization and standardization of SB care. In the current study, 91%of patients had their surgery performed at urban, teaching hospitals. Many of these hospitals, including the authors’, provide a SB-specific multispecialty practice or participate in national registries to facilitate appointment scheduling and follow-up, provide multispecialty provider collaboration, and share outcome data across institutions [13, 14]. While these groups and registries do not specifically standardize surgical care of patients with SB, the data from these studies may be used to alter practice patterns and indirectly standardize care. Standardization of surgical care has been shown to decrease post-operative complications and may explain the inability to find sufficient evidence of a difference between racial/ethnic groups [15–18].
Perhaps most importantly, these findings must be interpreted in the context of this study’s design limitations. Length of stay is a multifactorial outcome that may be influenced by unmeasured variables such as patient, provider, or hospital system factors. The Nationwide Inpatient Sample represents a 20%stratified sample of US hospital admissions; therefore, the reported results may not be applicable to encounters beyond the database. Despite this, NIS does provide meticulous tracking of hospital weights and discharges to minimize the risk of sampling bias [19]. As with all retrospective, administrative databases, NIS may be affected by miscoding bias. The analysis relied on the accuracy of the diagnostic and procedure codes included in NIS; in particular, accurate diagnosis and racial/ethnic information was relied on in order to define this cohort and to identify post-operative complications. While the accuracy of NIS is high for an administrative database, it is possible that at least some portion of the cohort may be incorrectly coded. While it is not likely that this potential error is a source of bias impacting the results, a higher-than-expected degree of coding error among a particular racial/ethnic group could potentially have biased the results. Likewise, it is important to note that this NIS analysis was explicitly performed using ICD-9 data in order to limit miscoding bias during the era of the ICD-9 to ICD-10 changeover; the unintended consequence of this choice is that the data may be “out of date”. Future analyses using more recent NIS datasets would be of interest in order to verify these findings.
The analysis is subject to variability based on NIS reporting, which tracks encounters, not patients. Due to the rarity of SB and the number of years examined, it is possible that the same patient appeared multiple times and each encounter was treated as unique. This may result in overestimation of the prevalence of SB encounters or complications thereof. Similarly, because it is impossible to track a given patient across time, post-hospital outcomes and whether individual patients were readmitted or underwent repeat procedures were not assessed. Additionally, the picture of post-operative complications is limited by the inpatient-only nature of NIS, precluding analysis of outpatient encounters not captured in the database. Finally, it is assumed that any NSQIP-coded complication was a surgically-related complication, although there is a chance that comorbidity or unrelated medical illness could be at fault instead; if true, this would potentially result in overestimation of the complication rates. However, it can be assumed that such random events would have been dispersed equally among racial/ethnic groups. Alternatively, defining complications with NSQIP may underestimate the true overall complication rate as some complications may not be counted. However, use of NSQIP in post-operative patients is a well validated method to assess complications. Although ethnicity data is tracked separately for some HCUP databases, NIS provides only the Race variable, which considers Hispanic ethnicity as a “racial” group. Though patients may individually identify within multiple racial and ethnic groups, this categorization rendered them mutually exclusive to a specific racial group. Despite these limitations, NIS is rigorously monitored and audited for coding accuracy and represents a reasonably reliable panorama of the characteristics of an inpatient surgical cohort.
Though a difference in cost and LOS between Hispanic and White SB cohorts was detected, the basis of these differences was not able to be addressed due to the retrospective nature of the study and data limitations inherent to a large national administrative database. It is noteworthy that there were differences between encounters for Hispanic and White patients but not between encounters for Black and White patients. This is most likely due to unmeasured confounding or possibly to cultural responses to treatments/providers, but based on data limitations, this can only be speculated. In order to confirm and ideally move beyond these findings to address causality, more (ideally prospective) studies in this population are needed. Hopefully, future analyses will also be able to avoid NIS’s reliance on White race as a de facto comparator for other groups, and instead, adopt a more nuanced approach to a very complex research topic.
Conclusion
Hispanic ethnicity was associated with higher costs and longer LOS in pediatric SB surgical in-hospital encounters. A better understanding of the effect of these non-modifiable risk factors can help identify the need to address racial disparities and establish change at the individual practice and hospital system level.
Footnotes
Acknowledgments
Dr. Routh was also supported in part by grant K08-DK100534 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). The funding sources had no role in the collection, analysis and interpretation of data; in the writing of the manuscript; or in the decision to submit the manuscript for publication.
The Duke Biostatistics Core’s support of this project was made possible (in part) by Grant Number UL1TR001117 from the National Center for Advancing Translational Sciences (NCATS), a component of the National Institutes of Health (NIH), and NIH Roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of NCATS or NIH. The funding sources had no role in the collection, analysis and interpretation of data; in the writing of the manuscript; or in the decision to submit the manuscript for publication.
Conflict of interest
The authors have no conflict of interest to report.
Appendix
Appendix A
Post-operative NSQIP Complication
ICD-9 Code
SSI
998.32, 998.31
Peritoneal abscess
567.22
UTI
032.84, 590.00, 590.01, 590.10, 590.11, 590.2, 590.3, 590.80, 590.81, 590.9, 595.0, 595.1, 595.2, 595.3, 595.4, 595.81, 595.82, 595.89, 595.9, 597.0, 597.80, 597.81, 597.89, 598.00, 598.01, 599.0
Urinary Complication
997.5
Respiratory complication
997.3, 518.81, 518.82, 514, 518.4
Pneumonia
481, 482, 483, 484, 485, 486, 487, 507
Post-operative respiratory complications
518.5
Acute Respiratory Insufficiency
518.82
Acute Respiratory Failure
518.81
Systemic Sepsis
790.7, 038
Pulmonary emboli infarct
415.1, 415.11, 415.19
Post-operative CVA
997.02
Cardiac complications
997.1
MI
410
Cardiac Arrest
427.5
Post-operative Bleeding
285.1, 998.11
DVT
4539, 453.4x, 453.8x, 451.11, 451.19, 451.2, 451.9, 415.11, 415.13, 415.19