Delays in Treatment of Pediatric Appendicitis: A More Accurate Variable for Measuring Pediatric Healthcare Inequalities?

Racial and socioeconomic factors may cause barriers to healthcare access that result in delayed treatment. Because perforated appendicitis (PA) in children is thought to result from delays in treatment, it is often used as an index of barrier to access. Recent literature suggests that PA is not an inevitable consequence of delayed treatment, so it may not be the best marker for evaluating such barriers. Therefore we investigated whether racial and socioeconomic factors led directly to delays in treatment. We performed a retrospective study of 667 children undergoing appendectomy in a tertiary care center over 12.5 years. Univariate and multivariable regression analyses were used to determine if racial and socioeconomic variables were associated with increased risk of PA and increased risk of symptom duration greater than 48 hours. Hispanic children have higher rates of PA regardless of delays in treatment whereas Black children had higher PA rates likely due to delays in treatment. These differences were not from socioeconomic factors in our cohort. PA, a heterogeneous disease whose course is determined by multiple factors, is not a good metric for evaluation healthcare disparities in the pediatric population. Delays in treatment may be a more appropriate measure of healthcare inequalities in children.

HEALTHCARE INEQUALITIES are a well-documented issue in the United States. Racial and ethnic minority populations in the United States are at increased risk for morbidity and mortality in various disease states. These differences in outcomes are thought to result from barriers to access for timely and appropriate health care.1-3 One example where health disparities may be present is acute appendicitis (AA) in children, a common problem that accounts for approximately 85,000 hospitalizations each year at a cost of nearly $2.2 billion.4 The traditional understanding of the pathophysiology of AA is that it will progress from inflammation to necrosis and perforation if surgical intervention is delayed.5 Moreover, timely intervention prevents progression to perforation and its associated complications.5, 6 Several studies have confirmed that racial and ethnic minority children have a higher rate of perforated appendicitis (PA) compared with white children.7-14

Healthcare policy groups have suggested that rates of PA reflect the equity of access to medical care, especially in children. Their reasoning is that poor access to care leads to delays in treatment and a higher likelihood of PA.5, 6 Others have called into question whether PA is a valid metric of access to health care or at best an incomplete one. Livingston and Fairlie6 demonstrated demographic, socioeconomic, and hospital-level factors only account for a small fraction of observed racial and minority differences in rates of pediatric PA. Others suggest that there are phenotypic and anatomic differences that may make racial and ethnic minority children more prone to PA than their white counterparts.6, 15-17

Large inpatient databases cannot account for many of the issues that may lead to delayed treatment. Jablonsky and Guagliardo10 list examples: the degree of acculturation, language barriers, preferences for traditional healing, and fears among undocumented immigrants preventing free contact with healthcare systems sponsored by governmental agencies. Moreover, administrative inpatient databases do not capture the time from symptom onset to definitive treatment, a factor thought to affect rates of perforation and itself a more direct reflection of access to health care.

In this study our aim was to examine racial, ethnic, and socioeconomic factors in the rates of PA while controlling for delays in treatment. We wanted to evaluate barriers to health care by directly investigating whether racial and socioeconomic factors are associated with delays in treatment.

Methods

Included in our review were all patients aged 17 years and younger who underwent an appendectomy at our hospital for presumed AA over a 12.5-year time period. Only patients whose operations were performed by one of the board-certified pediatric surgeons on staffwere included. Patients treated on other nonpediatric surgical services were excluded. Also excluded were patients who had an incidental appendectomy as part of another operation, had an interval appendectomy after nonoperative treatment for complicated appendicitis, or had incomplete clinical data in the medical record. To evaluate education and income data linked to North Carolina state zip codes, only patients with instate zip codes were included in education and income data analyses.

Clinical data included age, sex, and race/ethnicity. Patients were assigned to four groups based on age: 0 to 4, 5 to 9, 10 to 14, and 15 to 17 years. Symptom duration was defined as the time between the onset of the patient's symptoms and the incision time at appendectomy estimated to the nearest half-day increment.11, 18 Patients then were assigned into three groups based on symptom duration: less than 24 hours, 24 to less than 48 hours, and 48 hours and greater. These broad categories thus accommodated less exact estimations of symptom duration (e.g., half-day or full-day increments).

Socioeconomic variables included payor type (Medicaid and self pay vs private insurance), an estimate of household income, an estimate of the education level of the parents, and whether a patient had an established primary care provider. Household income and educational status were determined by correlating U.S. Census data to patient zip codes. Per-capita income (in 1999 dollars) and educational census data from the 2000 census were used as a proxy for the individual income and educational status of the child's parents.19 From these data, patients were divided into low and high income groups. Low income was defined as a per-capita income of less than $16,480 per year, 200 per cent of the federal poverty level for a single individual, and the defined cutofffor ''low income'' recommended by the National Center for Children in Poverty.20, 21 For each zip code, education status was determined by the percentage of individuals from that zip code who had attained an associate's degree or higher,22 categorized as low (less than 25%), mid (25% to less than 50%), and high (50% or more). Although it would be preferable to have assessments of parental income and education specific to each child, the use of aggregate data is a common practice in health care-related research and has been validated as a substitute for individual-level data.7, 23

An established primary care provider (PCP) was defined as a patient having a specific healthcare provider or group practice and was identified by review of emergency department records and addressees of the postoperative clinic letters. Those directed to the child's county public health department were considered as having no established PCP.

The diagnosis of PA was determined from the operative note. Patients' charts provided all clinical data, including length of stay, postoperative complications (abscess formation, wound infection, re-exploration, and rehospitalization). Financial records provided cost data.

All descriptive results are presented as mean (standard deviation [SD]) and number (percentage) unless otherwise specified. The marginal (univariable) association of each variable with PA was tested using x2, Student's t tests, and Wilcoxon rank-sum tests as appropriate. We modeled the effect of race/ethnicity on the probability of PA adjusting for confounding/ mediating factors using multivariable logistic regression in a manner previously described.24We created models that progressively induced a greater degree of control for confounding/mediating factors beginning with an unadjusted model that only considered the effect of race/ethnicity. The second model then added adjustments for the biological factors of sex and age; third, socioeconomic factors (per-capita income based on zip code and insurance payor type); and fourth, symptom duration. Another multivariable logistic regression model compared symptom duration greater than 48 hours to 48 hours or less, a threshold recognized to be associated with a plateau of increasing PA rates with symptom duration.11 Statistical significance was at P # 0.05. All statistical analyses were performed with JMP (Version 8.0.1; SAS Institute, Cary, NC) or the R Statistical Computing Environment (Version 2.15; R Foundation for Statistical Computing, Vienna, Austria).

Results

From 1999 to mid-2011, there were 667 patients with AA that were included in the analysis, of which 290 (43.5%) had PA (Table 1), a rate comparable to that reported in the literature.7, 8, 11, 25 The average age of patients with PA (8.27 years, SD 4 4.34) was significantly lower than those without PA (10.14 years, SD 4 4.05, P < 0.0001). Using the 5-to 9-year age group as a reference group, very young patients (age 0 to 4 years) had a higher risk of PAwith an odds ratio (OR) of 2.81 and a 95 per cent confidence interval (CI) of 1.84 to 4.31 (P < 0.0001). The two older age groups (10 to 14 and 15 to 17 years) had significantly decreased risk of PA (OR, 0.62; CI, 0.45 to 0.85; P 4 0.0032 and OR, 0.55; CI, 0.34 to 0.90; P 4 0.0154, respectively; Table 1). There was no difference in perforation rates between sexes.

We found differences in perforation rates among racial and ethnic groups. The white non-Hispanic group had decreased odds of PA (OR, 0.4; 0.29 to 0.54; P < 0.0001), whereas Hispanic ethnicity had increased odds of PA (OR, 2.52; 1.77 to 3.58; P < 0.0001). Neither black race nor other ethnicities had an association with PA (Table 1).

The Medicaid and self-pay groups were associated with an increased PA rate (OR, 2.01; 1.47 to 2.75; P < 0.0001). Comparing those who had an established primary care provider and those who did not, there was no difference in perforation rates (P40.12) (Table 1).

To analyze the role of delays in seeking treatment and the incidence of PA, we treated symptom duration as both a continuous and categorical variable. As a continuous variable, symptom duration in patients with PA had longer mean and median symptom durations (mean, 3.30; SD 2.50; median, 3 days; range, 0.5 to 21 days) than those with non-PA (mean, 1.67; SD, 1.46; median, 1 day; range, 0 to 14 days; Table 1; P < 0.0001). When evaluated by category, both symptom duration of less than 24 hours (OR, 0.12; 0.05 to 0.29; P < 0.0001) and 24 to 48 hours (OR, 0.33; 0.24 to 0.45; P < 0.0001) had decreased odds of PA, whereas greater than 48 hours had increased odds of PA (OR, 6.34; 4.45 to 9.03; P < 0.0001) (Table 1).

Per-capita income as a continuous variable was significantly lower between those that had PA (mean, $19,876; SD, $4,823; median, $18,912; range, $13,153 to $43,127) compared with thosewith non-PA ($21,070; SD, $5,691;median, $19,201; range, $13,153 to $39,744) (Table 2; P 4 0.0082). As a categorical variable, the low-income group was associated with PA (OR, 1.54; 1.02 to 2.33; P 4 0.041). Education level was not associated with PA (Table 2).

The first multivariable logistic regression analysis generated four models to determine the association of race and PA using white children as the reference group (Table 3). The first model demonstrated that both black and Hispanic children had increased odds of perforation (ORs, 1.97 and 2.92, respectively). Controlling for sex had no effect, but with increasing age, the odds of PA decreased, and both black and Hispanic ethnicities still had increased odds of PA. Further control by payor type and income level had no effect on the model. A fourth model controlling for symptom duration had the largest effect on the odds of PA with duration greater than 48 hours with the greatest effect (OR, 20.93) with older age groups continuing to have lower risk of PA. In the final model, Hispanic ethnicity, but not black race, was a significant risk factor.

A second multivariable logistic regression analysis tested the correlation of symptom duration with the other racial, ethnic, clinical, and socioeconomic factors (Table 4). Both black and Hispanic children had increased odds of experiencing symptom duration of greater than 48 hours. Adjusting for sex and age, black but not Hispanic children had higher odds of symptom duration 48 hours or greater, whereas older age was negatively correlated with longer symptom duration. Neither payor type nor per-capita income had an association with symptom duration.

All outcome measures were significantly increased in PA compared with non-PA (Table 5), including length of hospital stay (mean, 6.74 vs 2.0 days; P < 0.0001), inpatient cost ($12,941 vs $5,787; P < 0.0001), deep space abscess rates (6.6 vs 0.5%; P < 0.0001), wound infections (6.6 vs 1.9%; P 4 0.0019), re-exploration rates (6.6 vs 1.3%; P 4 0.0003), and rehospitalization rates (7.6 vs 2.7%; P 4 0.0031).

Discussion

We found that two factors, symptom duration and age, had the strongest effects on the risk of PA on both univariate and multivariate analyses regardless of race and ethnicity. Delayed time to presentation, termed symptom duration in our study, has been previously studied and positively correlated to rates of PA in children.5, 11, 25, 26We observed a median difference in symptom duration between the PA and non-PA groups of approximately 2 days, and patients who presented after 2 days of symptoms had significantly higher odds of having PA on univariate analysis. Usingmultivariable logistic regression, symptom duration had a very large effect (Table 3). This agrees with Brender et al.11 who showed that rates of PA increased as delay in treatment increased up to a limit of 49 hours, after which the PA rate reached a plateau and further delays did not increase rates in PA.

The youngest age group (0 to 4 years) had increased odds of PA with a perforation rate of 64.6 per cent, a result observed by several others.11, 13, 14, 27 As the age of our patients increased, the perforation rate decreased such that the perforation rate of age groups 5 to 9, 10 to 14, and 15 to 17 years were 45.7, 36.3, and 31.4 per cent, respectively, with the older two age groups having decreased odds of PA on multivariate analysis. Our findings confirm those of several others and are attributed to the fact that these age groups are able to understand their perception of pain and communicate it more clearly to parents and providers than the youngest age group.5

Our initial statistical analysis confirmed observations by others8-10, 12, 28, 29 that ethnicity was associated with an increased rate of PA. Both Hispanic ethnicity and black race had significantly higher rates of perforation for the first three models when accounting for age, sex, income, and education. When also controlling for symptom duration, only Hispanic ethnicity still had a higher PA rate, whereas the correlation between black race and PA disappeared. The persistence of significantly high rates of PA among Hispanic children defies an easy social, anatomical, or physiological explanation. The strength of symptom duration in our analysis implies there are other important cultural or economic factors not adequately explored in our analysis. Our results from the second logistic regression model suggest that black children are more likely to present later (greater than 48 hours) even when controlling for biologic and socioeconomic factors. Thus, when considering the racial disparity in the PA rates of black children in our study population, we can surmise that their overall increased rates of PA are likely the result of a delay in the time to presentation to the hospital for surgery, yet because this analysis controlled for socioeconomic factors, we are still unable to conclude that the delay in treatment of black children is a result of socioeconomic factors that act as a barrier to healthcare access.

PA is considered an inevitable consequence of delay in treatment of AA, so rates of PA have been used as a surrogate metric of access to care6 and higher rates of PA the result of barriers to care.7-10, 12 Several of these studies use national administrative inpatient databases that fail to control for symptom duration. We confirmed the clinical relationship of delayed operation and the incidence of PA that overcomes the effect of black race on rates of PA. Other socioeconomic issues thatmay affect access to care such as education, income level, and having an established PCP had no effect. For Hispanic children, we had the surprising result that their ethnicity by itself was a risk factor. Therefore, assignment of risk of PA as an index of barrier to care is problematic. Refinement of analysis and use of patientlevel data in our study suggest that a delay in seeking care may be a better index of barrier to care in appendicitis than perforation rate.

Livingston and coworkers, in fact, suggest that PA and nonperforated AA are different diseases; i.e., PA is not an inevitable result of delayed treatment of AA.6, 15, 16 From their analyses they concluded that only a small amount of difference in PA among races and ethnicities is attributable to insurance and economic factors.6 The group noted that the incidence of PA has been slowly increasing since the 1970s, whereas nonperforated AA has not.15 The incidence of nonperforating AA and nonperforated diverticulitis cointegrated over time, whereas PA and nonperforated AA did not, a suggestion that nonperforated AA has more in common with nonperforated diverticulitis than with PA.16 Longer symptom duration may be simply a descriptor of PA rather than the cause of it. If this hypothesis is true, it may account for some of our puzzling observations of the rate of PA in Hispanic children. For instance, they may be more susceptible to PA than other children.

This study has limitations inherent in retrospective studies. Records may contain conflicting information. Although our sample size is relatively small compared with studies using large administrative databases, it was large enough to generate significant results and thus appeared to be adequately powered. Basing income status and education level on zip codes, a common practice in health systems research, still may lack a sufficient definition for adequate analysis.7, 23 The same may be true in our decision, limited by inexact data in the clinical records, to round symptom duration to the nearest half day.

In summary, our multivariable logistic regression analysis of PA suggests that symptom duration, and not rate of PA, is a more appropriate metric than PA rate for evaluating healthcare disparities. Healthcare disparities in AA and PA require further study using the prospective collection of more exact measures of socioeconomic indices and symptom duration.

REFERENCES

&lt;p>1. Brinjikji W, Rabinstein AA, Lanzino G, CloftHJ. Racial and ethnic disparities in the treatment of unruptured intracranial aneurysms: a study of the Nationwide Inpatient Sample 2001-2009. Stroke 2012;43:3200-6.

2. Joshi S. J Gaynor J, Ciancio G. Review of ethnic disparities in access to renal transplantation. Clin Transplant 2012;26:E337-43.

3. Simard EP, Fedewa S, Ma J, et al. Widening socioeconomic disparities in cervical cancer mortality among women in 26 states, 1993-2007. Cancer 2012;118:5110-6.

4. Agency for Healthcare Research and Quality web site. HCUPnet. Available at: http://hcupnet.ahrq.gov/HCUPnet.jsp. Accessed January 2013.

5. Bernard M, Jaffe DHB. The appendix. In: Brunicardi FC, Billiar TR, Hunter JG, Pollock RE, eds. Schwartz's Principles of Surgery. 9th ed. New York, NY: McGraw Hill. September 2009.

6. Livingston EH, Fairlie RW. Little effect of insurance status or socioeconomic condition on disparities in minority appendicitis perforation rates. Arch Surg 2012;147:11-7.

7. Lee SL, Yaghoubian A, Stark R, et al. Are there differences in access to care, treatment, and outcomes for children with appendicitis treated at county versus private hospitals? Perm J. 2012;16:4-6.

8. Lee SL, Shekherdimian S, Chiu VY, Sydorak RM. Perforated appendicitis in children: equal access to care eliminates racial and socioeconomic disparities. J Pediatr Surg 2010;45:1203-7.

9. Kelley-Quon LI, Tseng C-H, Jen HC, et al. Hospital type as a metric for racial disparities in pediatric appendicitis. J Am Coll Surg 2013;216:74-82.

10. Jablonski KA, Guagliardo MF. Pediatric appendicitis rupture rate: a national indicator of disparities in healthcare access. Popul Health Metr 2005;3:4.

11. Brender JD, Marcuse EK, Koepsell TD, Hatch EI. Childhood appendicitis: factors associated with perforation. Pediatrics 1985;76:301-6.

12. Guagliardo MF, Teach SJ, Huang ZJ, et al. Racial and ethnic disparities in pediatric appendicitis rupture rate. Acad Emerg Med 2003;10:1218-27.

13. Bratton SL, Haberkern CM, Waldhausen JH. Acute appendicitis risks of complications: age and Medicaid insurance. Pediatrics 2000;106:75-8.

14. Gadomski A, Jenkins P. Ruptured appendicitis among children as an indicator of access to care. Health Serv Res 2001;36: 129-42.

15. Livingston EH, Woodward WA, Sarosi GA, Haley RW. Disconnect between incidence of nonperforated and perforated appendicitis: implications for pathophysiology and management. Ann Surg 2007;245:886-92.

16. Livingston EH, Fomby TB, Woodward WA, Haley RW. Epidemiological similarities between appendicitis and diverticulitis suggesting a common underlying pathogenesis. Arch Surg 2011;146:308-14.

17. Guidry SP, Poole GV. The anatomy of appendicitis. Am Surg 1994;60:68-71.

18. Augustin T, Cagir B, Vandermeer TJ. Characteristics of perforated appendicitis: effect of delay is confounded by age and gender. J Gastrointest Surg 2011;15:1223-31.

19. U.S. Census Bureau. Per Capita Income from American Fact Finder. Available at: http://factfinder2.census.gov/faces/nav/ jsf/pages/index.xhtml. Accessed January 2013.

20. Mailman School of Public Health Columbia University. National Center for Children in Poverty (NCCP). NCCP.org. 2012. Available at: http://www.nccp.org/profiles/NC_profile_6.html. Accessed January 2013.

21. U.S. Department of Health and Human Services. The 1999 HHS Poverty Guidelines. Available at: http://aspe.hhs.gov/poverty/ 99poverty.htm. Accessed January 2013.

22. U.S. Census Bureau. Profile of Selected Social Characteristics: 2000 from American Fact Finder. Available at: http:// factfinder2.census.gov/faces/nav/jsf/pages/index.xhtml. Accessed January 2013.

23. Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health 1992;82:703-10.

24. R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2009.

25. O'Toole SJ, Karamanoukian HL, Allen JE, et al. Insurancerelated differences in the presentation of pediatric appendicitis. J Pediatr Surg 1996;31:1032-4.

26. Schweitzer J, Fairman N, Schreyer K, Waxman K. Appendicitis, 2002: relationship between payors and outcome. Am Surg 2003;69:902-8.

27. Nwomeh BC, Chisolm DJ, Caniano DA, Kelleher KJ. Racial and socioeconomic disparity in perforated appendicitis among children: where is the problem? Pediatrics 2006;117: 870-5.

28. Smink DS, Fishman SJ, Kleinman K, Finkelstein JA. Effects of race, insurance status, and hospital volume on perforated appendicitis in children. Pediatrics 2005;115:920-5.

29. Ponsky TA, Huang ZJ, Kittle K, et al. Hospital- and patient-level characteristics and the risk of appendiceal rupture and negative appendectomy in children. JAMA 2004;292: 1977-82.

MITCHELL R. LADD, PH.D.,* NICHOLAS M. PAJEWSKI, PH.D.,[dagger] ROBERT D. BECHER, M.D.,* JOHN M. SWANSON, B.S.,* JARED R. GALLAHER, M.D.,* THOMAS PRANIKOFF, M.D.,* LUCAS P. NEFF, M.D.*[double dagger]

From the *Section of Pediatric Surgery and the [dagger]Department of Biostatistical Sciences, Division of Public Health Sciences, Wake Forest School of Medicine, Winston-Salem, North Carolina; and the [double dagger]Department of General Surgery, University of California at Davis, Sacramento, California

Presented at the Southeastern Pediatric Surgical Congress at the Southeastern Surgical Congress Annual Meeting, Jacksonville, FL, February 9-12, 2013.

Address correspondence and reprint requests to Lucas P. Neff, M.D., MC, USAF, Major, USAF, MC, Assistant Clinical Professor of Surgery, University of California at Davis, Department of General Surgery/Division of GI Surgery, Department of Surgery, Cypress Building, 2221 Stockton Boulevard, Room 3107, Sacramento, CA 95817-1418. E-mail: [email protected].