[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.  Study Flowchart of Inclusion and Exclusion Criteria
Study Flowchart of Inclusion and Exclusion Criteria

This figure illustrates how the initial data set of 45 436 patients was narrowed to a study cohort of 35 735 patients, as well as their designation as cases of perforated appendicitis or acute appendicitis. ICD-9 indicates International Classification of Diseases, Ninth Revision (ICD-9) and LOS, length of stay.

Figure 2.  Illustration of Moran Index for Spatial Autocorrelation
Illustration of Moran Index for Spatial Autocorrelation

This figure illustrates the range of possible results for Moran index for Spatial Autocorrelation and how these values (from −1 to 1) are interpreted.

Figure 3.  Incidence of Acute Appendicitis (AA) and Perforated Appendicitis (PA) in the Tricounty Puget Sound Region
Incidence of Acute Appendicitis (AA) and Perforated Appendicitis (PA) in the Tricounty Puget Sound Region

This figure depicts the population-based incidence of AA and of PA in the study area. The Moran index of 0.30 (P < .001) for AA and of 0.16 (P < .001) for PA indicate a statistically significant nonrandom geographic distribution of incidence for both conditions. The strength of the association between geographic location and incidence is twice is strong for AA as for PA.

Figure 4.  Getis-Ord Gi* Clustering Analysis of Hot Spots and Cold Spots of Incidence of Acute Appendicitis (AA) and Perforated Appendicitis (PA)
Getis-Ord Gi* Clustering Analysis of Hot Spots and Cold Spots of Incidence of  Acute Appendicitis (AA) and Perforated Appendicitis (PA)

Getis-Ord Gi* is a regression-based geographic method that identifies clusters of spatially associated geographic subunits such as census tracts at 90, 95, and 99 percent confidence levels. A, Analysis for AA and identifies hot and cold spots for age- and sex-standardized incidence of AA. B, Analysis for PA and identifies hot and cold spots for age- and sex-standardized incidence of PA.

Table.  Comparing Incidence and Socioeconomic Factors Between High-Incidence and Low-Incidence Census Tract Clustersa
Comparing Incidence and Socioeconomic Factors Between High-Incidence and Low-Incidence Census Tract Clustersa
1.
Boomer  L, Freeman  J, Landrito  E, Feliz  A.  Perforation in adults with acute appendicitis linked to insurance status, not ethnicity.  J Surg Res. 2010;163(2):221-224. doi:10.1016/j.jss.2010.04.041PubMedGoogle ScholarCrossref
2.
Nwomeh  BC, Chisolm  DJ, Caniano  DA, Kelleher  KJ.  Racial and socioeconomic disparity in perforated appendicitis among children: where is the problem?  Pediatrics. 2006;117(3):870-875. doi:10.1542/peds.2005-1123PubMedGoogle ScholarCrossref
3.
Braveman  P, Schaaf  VM, Egerter  S, Bennett  T, Schecter  W.  Insurance-related differences in the risk of ruptured appendix.  N Engl J Med. 1994;331(7):444-449. doi:10.1056/NEJM199408183310706PubMedGoogle ScholarCrossref
4.
Luckmann  R.  Incidence and case fatality rates for acute appendicitis in California: a population-based study of the effects of age.  Am J Epidemiol. 1989;129(5):905-918. https://www.ncbi.nlm.nih.gov/pubmed/2784936. doi:10.1093/oxfordjournals.aje.a115224PubMedGoogle ScholarCrossref
5.
Luckmann  R, Davis  P.  The epidemiology of acute appendicitis in California: racial, gender, and seasonal variation.  Epidemiology. 1991;2(5):323-330. https://www.ncbi.nlm.nih.gov/pubmed/1742380. doi:10.1097/00001648-199109000-00003PubMedGoogle ScholarCrossref
6.
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(6):886-892. doi:10.1097/01.sla.0000256391.05233.aaPubMedGoogle ScholarCrossref
7.
Andersson  R, Hugander  A, Thulin  A, Nyström  PO, Olaison  G.  Indications for operation in suspected appendicitis and incidence of perforation.  BMJ. 1994;308(6921):107-110. doi:10.1136/bmj.308.6921.107PubMedGoogle ScholarCrossref
8.
Addiss  DG, Shaffer  N, Fowler  BS, Tauxe  RV.  The epidemiology of appendicitis and appendectomy in the United States.  Am J Epidemiol. 1990;132(5):910-925. https://www.ncbi.nlm.nih.gov/pubmed/2239906. doi:10.1093/oxfordjournals.aje.a115734PubMedGoogle ScholarCrossref
9.
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(16):1977-1982. doi:10.1001/jama.292.16.1977PubMedGoogle ScholarCrossref
10.
Smink  DS, Fishman  SJ, Kleinman  K, Finkelstein  JA.  Effects of race, insurance status, and hospital volume on perforated appendicitis in children.  Pediatrics. 2005;115(4):920-925. doi:10.1542/peds.2004-1363PubMedGoogle ScholarCrossref
11.
Pieracci  FM, Eachempati  SR, Barie  PS, Callahan  MA.  Insurance status, but not race, predicts perforation in adult patients with acute appendicitis.  J Am Coll Surg. 2007;205(3):445-452. doi:10.1016/j.jamcollsurg.2007.04.010PubMedGoogle ScholarCrossref
12.
Bickell  NA, Aufses  AH  Jr, Rojas  M, Bodian  C.  How time affects the risk of rupture in appendicitis.  J Am Coll Surg. 2006;202(3):401-406. doi:10.1016/j.jamcollsurg.2005.11.016PubMedGoogle ScholarCrossref
13.
Gadomski  A, Jenkins  P.  Ruptured appendicitis among children as an indicator of access to care.  Health Serv Res. 2001;36(1 pt 1):129-142.PubMedGoogle Scholar
14.
Sicard  N, Tousignant  P, Pineault  R,  et al.  Provider density and health system facility factors and their relationship to rates of pediatric perforated appendicitis in US counties.  J Pediatr Surg. 2010;117(2):290-293. doi:10.1001/archsurg.2010.328Google Scholar
15.
Camp  M, Chang  DC, Zhang  Y,  et al.  Provider density and health system facility factors and their relationship to rates of pediatric perforated appendicitis in US counties.  Arch Surg. 2010;145(12):1139-1144. doi:10.1001/archsurg.2010.271PubMedGoogle ScholarCrossref
16.
Paquette  IM, Zuckerman  R, Finlayson  SRG.  Perforated appendicitis among rural and urban patients: implications of access to care.  Ann Surg. 2011;253(3):534-538. doi:10.1097/SLA.0b013e3182096d68PubMedGoogle ScholarCrossref
17.
Akhtar-Danesh  G-G, Doumouras  AG, Flageole  H, Hong  D.  Geographic and socioeconomic predictors of perforated appendicitis: a national Canadian cohort study.  J Pediatr Surg. 2018;(November). doi:10.1016/j.jpedsurg.2018.10.065PubMedGoogle Scholar
18.
Sicard  N, Tousignant  P, Pineault  R, Dubé  S.  Non-patient factors related to rates of ruptured appendicitis.  Br J Surg. 2007;94(2):214-221. doi:10.1002/bjs.5428PubMedGoogle ScholarCrossref
19.
Lee  SL, Shekherdimian  S, Chiu  VY.  Effect of race and socioeconomic status in the treatment of appendicitis in patients with equal health care access.  Arch Surg. 2011;146(2):156-161. doi:10.1001/archsurg.2010.328PubMedGoogle ScholarCrossref
20.
Walker  A, Hatch  Q, Drake  T,  et al.  Predictors of appendiceal perforation in an equal access system.  J Surg Res. 2014;190(1):87-92. doi:10.1016/j.jss.2014.02.028PubMedGoogle ScholarCrossref
21.
Flum  DR, Morris  A, Koepsell  T, Dellinger  EP.  Has misdiagnosis of appendicitis decreased over time? a population-based analysis.  JAMA. 2001;286(14):1748-1753. https://www.ncbi.nlm.nih.gov/pubmed/11594900. doi:10.1001/jama.286.14.1748PubMedGoogle ScholarCrossref
22.
Grubesic  TH, Matisziw  TC.  On the use of ZIP codes and ZIP code tabulation areas (ZCTAs) for the spatial analysis of epidemiological data.  Int J Health Geogr. 2006;5:58. doi:10.1186/1476-072X-5-58PubMedGoogle ScholarCrossref
23.
Health NYSD. Rates based on small numbers. https://www.health.ny.gov/diseases/chronic/ratesmall.htm. Accessed November 15, 2018.
24.
Ilves  I, Paajanen  HEK, Herzig  KH, Fagerström  A, Miettinen  PJ.  Changing incidence of acute appendicitis and nonspecific abdominal pain between 1987 and 2007 in Finland.  World J Surg. 2011;35(4):731-738. doi:10.1007/s00268-011-0988-8PubMedGoogle ScholarCrossref
25.
von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.  Lancet. 2007;370(9596):1453-1457. doi:10.1016/S0140-6736(07)61602-XPubMedGoogle ScholarCrossref
26.
Winkleby  MA, Jatulis  DE, Frank  E, Fortmann  SP.  Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease.  Am J Public Health. 1992;82(6):816-820. doi:10.2105/AJPH.82.6.816PubMedGoogle ScholarCrossref
27.
Kelli  HM, Mehta  A, Tahhan  AS,  et al.  Low educational attainment is a predictor of adverse outcomes in patients with coronary artery disease.  J Am Heart Assoc. 2019;8(17):e013165. doi:10.1161/JAHA.119.013165PubMedGoogle Scholar
28.
Andersson  MA.  Higher education, bigger networks? differences by family socioeconomic background and network measures.  Socius Sociol Res a Dyn World. 2018;4:237802311879721. doi:10.1177/2378023118797217Google Scholar
29.
Pinsky  JL, Leaverton  PE, Stokes  J  III.  Predictors of good function: the Framingham Study.  J Chronic Dis. 1987;40(1)(suppl 1):159S-167S, 181S-2. doi:10.1016/S0021-9681(87)80045-0PubMedGoogle ScholarCrossref
30.
Huisman  M, Kunst  AE, Bopp  M,  et al.  Educational inequalities in cause-specific mortality in middle-aged and older men and women in eight western European populations.  Lancet. 2005;365(9458):493-500. doi:10.1016/S0140-6736(05)17867-2PubMedGoogle ScholarCrossref
31.
Kubota  Y, Heiss  G, MacLehose  RF, Roetker  NS, Folsom  AR.  Association of educational attainment with lifetime risk of cardiovascular disease: the atherosclerosis risk in communities study.  JAMA Intern Med. 2017;177(8):1165-1172. doi:10.1001/jamainternmed.2017.1877PubMedGoogle ScholarCrossref
32.
Drewnowski  A, Rehm  CD, Solet  D.  Disparities in obesity rates: analysis by ZIP code area.  Soc Sci Med. 2007;65(12):2458-2463. doi:10.1016/j.socscimed.2007.07.001PubMedGoogle ScholarCrossref
33.
Moudon  AV, Cook  AJ, Ulmer  J, Hurvitz  PM, Drewnowski  A.  A neighborhood wealth metric for use in health studies.  Am J Prev Med. 2011;41(1):88-97. doi:10.1016/j.amepre.2011.03.009PubMedGoogle ScholarCrossref
34.
Park  HC, Kim  MJ, Lee  BH.  Randomized clinical trial of antibiotic therapy for uncomplicated appendicitis.  Br J Surg. 2017;104(13):1785-1790. doi:10.1002/bjs.10660PubMedGoogle ScholarCrossref
35.
Fitz  RH. Perforating Inflammation of the Vermiform Appendix; with Special Reference to Its Early Diagnosis and Treatment. Philadelphia, PA: WM J. Dornan, Printer; 1886.
36.
Søreide  K.  The research conundrum of acute appendicitis.  Br J Surg. 2015;102(10):1151-1152. doi:10.1002/bjs.9890PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Views 955
    Citations 0
    Original Investigation
    March 4, 2020

    Geographic Association Between Incidence of Acute Appendicitis and Socioeconomic Status

    Author Affiliations
    • 1Department of Surgery, Boston University School of Medicine, Boston, Massachusetts
    • 2Department of Geography, San Francisco State University, San Francisco, California
    • 3Department of Surgery, University of Washington School of Medicine, Seattle
    JAMA Surg. Published online March 4, 2020. doi:10.1001/jamasurg.2019.6030
    Key Points

    Question  Is the population-based incidence of acute appendicitis (AA) or perforated appendicitis (PA), at the census tract level, uniformly distributed across a densely populated geographic area?

    Findings  In this study, AA and PA were clustered geographically into high-incidence and low-incidence regions, but geographic autocorrelation was twice as strong for AA compared with PA. Areas of low AA incidence have significantly increased secondary education and income vs high-incidence areas.

    Meaning  These findings challenge the conventional view that AA occurs randomly and has no predisposing characteristics beyond age/sex; higher socioeconomic status is associated with lower incidence of both AA and PA.

    Abstract

    Importance  Some studies based on proportions of patients with perforated appendicitis (PA) among all patients with acute appendicitis (AA) have found an association between socioeconomic status (SES) and risk of perforation. A potential limitation is their use of proportions, which assumes that incidence of AA is evenly distributed across populations at risk. This assumption may be invalid, and SES may have a more complex association with both AA and PA.

    Objective  To generate population-based incidences of AA and PA and to examine geographic patterns of incidence alongside geographic patterns of SES.

    Design, Setting, and Participants  Retrospective study of data from Washington’s Comprehensive Hospital Abstract Reporting System and the 2010 US census. Geographic methods were used to identify patterns of age- and sex-standardized incidence in Washington State between 2008 and 2012. The study included all patients discharged with International Classification of Diseases, Ninth Revision codes for AA or PA. Data were analyzed between November 2016 and December 2018.

    Exposures  Location of primary residence.

    Main Outcomes and Measures  Age- and sex-standardized incidence for AA and PA was generated for each census tract (CT). Global spatial autocorrelation was examined using Moran index (0.0 = completely random incidence; 1.0 = fully dependent on location). Clusters of low-incidence CTs (cold spots) and high-incidence CTs (hot spots) were identified for AA. Census-based SES data were aggregated for hot spots and cold spots and then compared.

    Results  Statewide, over the 5-year study period, there were 35 730 patients with AA (including 9780 cases of PA), of whom 16 574 were women (46.4%). Median age of the cohort was 29 years (IQR, 16-48 years). Statewide incidence of AA and PA was 106 and 29 per 100 000 person-years (PY), respectively. Crude incidence was higher within the male population and peaked at age 10 to 19 years. Age- and sex-standardized incidence of AA demonstrated significant positive spatial autocorrelation (Moran index, 0.30; P < .001), but autocorrelation for PA was only half as strong (0.16; P < .001). Median incidence of AA was 118.1 per 100 000 PY among hot spots vs 86.2 per 100 000 PY among cold spots (P < .001). Socioeconomic status was higher in cold spots vs hot spots: mean proportion of college-educated adults was 56% vs 26% (P < .001), and mean per capita income was $44 691 vs $30 027 (P < .001).

    Conclusions and Relevance  Age- and sex-standardized incidence of appendicitis is not randomly distributed across geographic subunits, and geographic clustering of AA is twice as strong as PA. Socioeconomic advantages, such as higher income and secondary education, are strongly associated with lower incidence of AA. These findings challenge conventional views that AA occurs randomly and has no predisposing characteristics beyond age/sex. Socioeconomic status, and likely other geographically circumscribed factors, are associated with incidence of AA.

    Introduction

    Many investigators have considered acute appendicitis (AA) a useful model to study disparities in access to high-quality surgical care.1-3 One study noted that “[AA] is considered an ideal subject of investigation regarding health care disparities because it is characterized by a consistent natural history … and an absence of known biological predisposition to perforation in any racial or ethnic group.”2 Other investigators have disputed the concept that AA is random in onset and consistent in progression.4,5 Luckmann4 raised the provocative “possibility that perforating and nonperforating appendicitis may represent two distinct entities.”4 He used California discharge data from 1984 to generate age-adjusted incidences of appendicitis and found significant age-related variation in the incidence of nonperforating appendicitis but much less variation in the incidence of perforated appendicitis (PA) and abscess. He hypothesized that a high proportion of PA among elderly individuals was associated more with a lower overall incidence of appendicitis than with a higher risk for perforation.4 Livingston et al6 examined secular trends from the 1970s to 2000s and showed that the population-based incidences of perforating and nonperforating appendicitis followed different patterns. He suggested that the relatively recent increase in nonperforating appendicitis was owing to early diagnosis of mild appendicitis on computed tomography and lower thresholds to take patients to surgery in the era of laparoscopic appendectomy. This is compared with previous management strategies in which patients with an indeterminate diagnosis of appendicitis would be observed, and, under this hypothesis, some patients with mild appendicitis would resolve with supportive care.6 Andersson et al7 evaluated 7 population-based studies and found that the incidence of non-PA varied widely among populations based on aggressiveness of surgical approach to suspected appendicitis, but that the incidence of PA was essentially unchanged. Similarly, Addiss et al8 noted in 1990 that “differences in surgical practices may also have played a role [in geographic variation], since the regions with the highest rates of appendicitis were generally also highest for negative and incidental appendectomy.”8

    These researchers all used population-based methods to study incidence of AA and rates of perforation. However, most studies of risk factors for perforation have based their findings on proportional methods: generally, the number of cases of PA divided by all cases of AA. Such proportion-based studies have identified disparities in the rate of perforation associated with race/ethnicity,9,10 insurance status,1,3,9,11 age,12 relationships with primary care clinicians,13-15 rural residence,16,17 and even hospital resource levels.14,18 Each of these studies demonstrated an association between markers of reduced access to health care and an increase in proportion of perforation. This association became so accepted that several articles were written demonstrating that improved health care access eliminated disparities in perforation.19,20

    Two assumptions underlie these proportion-based studies: (1) appendicitis has no predisposing characteristics (other than age and sex) and (2) all cases of appendicitis progress toward perforation without intervention. However, a close reading of the epidemiologic data presented by Luckmann,4 Andersson et al,7 Livingston et al,6 and others suggests that those assumptions should be revisited. We hypothesized that in a statewide data set with near-100% capture of hospital admissions, we could detect patterns in the epidemiology of AA that would clarify the validity of these previous assumptions. Furthermore, we hypothesized that areas of higher socioeconomic status (SES) would have a higher incidence of AA (because of greater access to early diagnosis) but that rates of PA would be more constant. A substantial body of high-quality research has documented the pervasive influence of economic, racial/ethnic, sex, and other social determinants on health care outcomes. The purpose of this study is not to dispute whether such disparities exist in AA, but to clarify more accurately how they may influence outcomes.

    Methods
    Generating the Washington State Comprehensive Hospital Abstract Reporting System Appendicitis Data Set

    We obtained data from the Washington State Comprehensive Hospital Abstract Reporting System (CHARS) for 5 years, 2008 to 2012, chosen because of temporal proximity to the 2010 US census. Some population-based studies have used procedure codes for appendectomy to identify patients with AA,21 but not all patients with appendicitis undergo operations at their index admission, in particular, those with advanced disease such as abscess or phlegmon. Therefore, we queried CHARS using International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes in addition to procedure codes (eFigure 1 in the Supplement). We began with all patients with diagnosis codes for AA or procedure codes for appendectomy (Figure 1). Non-Washington residents and patients coded for incidental appendectomy were excluded. The ICD-9 codes for appendiceal cancer are 153.5 and 209.11, and these patients were included only if they also had a code for AA. We excluded patients who had no ICD-9 code for AA but underwent a nonincidental appendectomy. These patients make up the “negative appendectomy” segment of the overall appendectomy population (eTable 1 in the Supplement). Perforated appendicitis was defined as an ICD-9 code of 540.0 or 540.1. In this article, AA includes all cases of appendicitis, including nonperforated and perforated; moreover, PA includes all instances of complicated appendicitis including perforation, abscess, and inoperable phlegmon.

    Including patients who had an appendicitis ICD-9 code but not an appendectomy risked including patients who were admitted with possible appendicitis but ultimately diagnosed as having something else. However, to keep patients in the data set who had complicated appendicitis treated nonoperatively, we could not rely on lack of appendectomy as an exclusion criteria. Therefore, we excluded patients who had an ICD-9 code for AA but did not have an appendectomy and were discharged within 24 hours. These patients were unlikely to have AA (and certainly not PA) because this data set is from well before antibiotics-only treatment of uncomplicated appendicitis was common.

    Generating Geographic Data

    The CHARS data set includes patient zip code but not home address. Zip codes are not ideal to spatially analyze epidemiologic data because they are based on linear postage routes and not neighborhoods.22 Thus, we chose to examine incidence at the census tract (CT) level. Census tracts contain greater statistical accuracy for measuring populations because they are relatively permanent and more representative of neighborhoods. Because zip codes are larger than CTs, data must be disaggregated from zip codes into CTs. We used a Census Bureau–approved method to systematically allocate patients within zip codes to CTs based on percentages of occupied residential housing units per CT.

    Substantial rural portions of Washington are too sparsely populated for incidence to be reliably calculated. The National Center for Health Statistics does not publish rates if the absolute number of cases is less than 20.23 Thus, although we generated statewide data, detailed analysis of geographic incidence patterns and SES was performed only within the 3 counties immediately adjacent to the Puget Sound region (King, Snohomish, and Pierce), hereafter referred to as the tricounty area. This area contains 51% of Washington’s population, with just more than 3.6 million residents, and represents a racially and socioeconomically diverse urban and dense suburban population. If a tricounty resident had appendicitis care at a hospital in another county, data for that patient would be correctly designated to the county of residence. Furthermore, none of these counties are on a state line, making it unlikely a tricounty resident would travel out of state for AA care.

    Age- and Sex-Standardized Incidence

    Several articles have demonstrated age and sex differences in the incidence of AA.5,8,24 There is substantial age variation across the tricounty Puget Sound area, with several university districts and certain neighborhoods with higher proportions of elderly residents. Thus, we performed direct age and sex standardization using the overall state population as the standard population. Standardization was performed at zip code level prior to disaggregation.

    Geographic Analysis

    Incidences of AA and PA were first examined using the Moran index to determine whether the spatial distribution of incidence was owing to chance. The output of Moran index is a value between −1.0 (dispersion) and 1.0 (clustering), and 0 indicates a random distribution (Figure 2). Because geospatial distribution was not random, we then proceeded to perform a Getis-Ord Gi* analysis. This is a regression-based technique to assess whether the incidence in one CT is influenced by nearby CTs at a particular confidence level. We identified groups of high-incidence CTs that are spatially related or clustered (designated as hot spots) and groups of clustered low-incidence CTs (cold spots).

    Socioeconomic Status and Incidence of AA

    Socioeconomic status data for each CT were taken from the 2010 census. Socioeconomic status data were aggregated for hot spots and cold spots. All data on race/ethnicity are reported as designated by the US Census Bureau. Importantly, percentages of race/ethnicity are aggregate variables per CT (as are all SES variables reported in this study); these variables are not based on patient data reported in the CHARS data set. Means were compared via t tests, with statistical significance set at a 2-sided P value greater than .05. The analytic plan and reporting of results for this study accord with the STROBE statement.25

    Human Participants Review

    Human participants review was waived by the San Francisco State University institutional review board because the data set provided by the Washington Department of Health contains no direct identifiers.

    Sensitivity Analysis

    To ensure that our results were not confounded by race/ethnicity, we used multivariable linear regression to study the association between incidence of AA/PA and SES variables across all CTs in the study area (ie, not just hot spot vs cold spot CTs). Based on prior work, census variables were included in the model if they had a univariate association with incidence of AA or PA at a significance level P of less than .05, which included race/ethnicity, median home value, median income, and percentage bachelor’s degree. All P values were 2-sided. Not all census categories of race/ethnicity had a univariate association with incidence, but because some did, all were included in the multivariable model.

    Results

    Incidence of AA across Washington State was 106.0 cases per 100 000 person-years (PY) and was 102.4 per 100 000 PY for the tricounty Puget Sound area. For PA, this was 29.0 and 28.2 per 100 000 PY, respectively. Both AA and PA were more common within the male population compared with the female population, with respective incidences of 114.0 and 31.4 per 100 000 PY (male population) and 94.3 and 24.5 per 100 000 PY (female population). For both sexes, the age decile with the highest incidence of AA was 10 to 19 years; children 9 years and younger had the lowest incidence (eTable 2 in the Supplement).

    Incidence of AA and PA both demonstrated significant nonrandom spatial variability (Figure 3); however, the maps depict a higher degree of geographic autocorrelation for AA compared with PA, a finding that was determined statistically using Moran index: AA = 0.30 (P < .001) and PA = 0.16 (P < .001). Geographic clustering was twice as robust for AA compared with PA, with a less than 1% chance that these spatial distributions were owing to chance. Getis-Ord Gi* analysis identified 235 CTs within cold spots and 234 CTs within hot spots at the 99% confidence level (Figure 4).

    Census tracts with high AA incidence were concentrated in the northern and southern ends of the study area (ie, Tacoma and Everett). The Seattle and Bellevue metropolitan areas had the lowest incidence. Perforated appendicitis incidence reflects a similar nonrandom pattern of distribution, although overall incidence is more homogenous. Median standardized incidence of AA was 86.2 per 100 000 PY among cold spots and 118.1 per 100 000 PY among hot spots. For PA, incidence between cold and hot spots was 24.4 per 100 000 PY and 29.5 per 100 000 PY, respectively. This incidence gap (the proportional difference from cold spots to hot spots) was almost double for AA compared with PA (21% for PA vs 37% for AA). (Table; eFigure 2 in the Supplement).

    Finally, we compared socioeconomic characteristics of AA hot spots vs cold spots. Socioeconomic status markers, such as mean proportion of adults with a college education (56% vs 26%, P < .001), mean median income ($79 841 vs $70 110; P < .001), and mean per capita income ($44 691.62 vs $30 027.60; P < .001), were substantially higher in cold spots compared with hot spots. Other SES markers, including rates of high school graduation, employment, and public assistance, were also significantly different, although to a lesser degree, and followed a similar pattern: areas of higher SES had lower incidence of both AA and PA. Racial and ethnic differences were statistically significant, likely because of the large data set, but the differences were small and provide somewhat of a mixed picture. More affluent, low-incidence CTs (cold spots) had more African American residents but fewer Latino residents; they had a higher nonwhite population and more residents not born in the United States (Table).

    After adjustments via our multivariable linear regression model, increasing percentage of bachelor’s degrees within CTs was independently associated with reduced incidence of both AA and PA. Increasing proportion of black race/ethnicity was also associated with reduced incidence of AA, but the other race/ethnicity variables had no association with incidence. The AA model had an adjusted R2 value of 0.20. For PA, in addition to bachelor’s degrees, increases in median income and proportion of black and Latino residents were all associated with decreases in incidence of PA. The PA model had an adjusted R2 value of 0.11 (eTable 3 in the Supplement). Therefore, approximately 20% of the variation in incidence of AA but only 11% of the variation in incidence of PA across the entire study area was explainable by the SES and race/ethnicity variables included in our 2 models.

    Discussion

    This analysis identified significant geographic clustering in the incidence of AA, a disease thought to be random in onset, and weaker, although still present, geographic clustering in PA, a disease thought to have substantial associations with SES and other social determinants of health. Consistent with the observation that PA incidence was more stable across regions, the incidence rate ratio of PA to AA was higher within cold spots for AA. Put another way: lower incidence of overall appendicitis, even among affluent CTs, did not result in proportionally lower incidence of PA. Moreover, contrary to our hypothesis, the incidence of AA was not higher in regions of higher SES. In fact, college education and income were substantially higher in areas of low incidence. This analysis focused particularly on education and income because these 2 parameters plus occupation are considered the chief components of overall SES.26-28 Indeed, some public health investigators consider education to be the best single marker of SES as it relates to health,26,29 and the difference in college education between AA hot spots and cold spots was especially large at 26% vs 56%. As with higher education, per capita income was substantially higher in cold spots vs hot spots (nearly 50% higher). In fact, the only income/wealth variable that did not follow the trend was home ownership, which was 10% higher in hot spots, and we suspect this is associated with the extremely expensive real estate market in Seattle and a young, tech-based work force that is more likely to rent than buy (this is supported by the small but significant difference in rent-to-income ratio, suggesting that cold spot renters pay less of their income in rent than those who rent in hot spots). In summary, essentially all of the education and income data support the conclusion that areas with low AA incidence are more educated and wealthier compared with those with high incidence. In a multivariable model of incidence over the entire study area, secondary education remained independently and robustly associated with incidence after adjustments for race/ethnicity and income.

    Previous studies have identified educational attainment as a potent predictor of health outcomes,26,27,29-31 and similar to our own study, other investigators have demonstrated associations between geography, SES, and health in the Puget Sound region.32 These have suggested that geography can function as its own marker of SES, with direct, measurable associations between where populations live and their health and health care outcomes.33 Although our study detected an unexpected association between the incidence of AA and SES, it is not the first to do so. A 2018 Canadian study17 (notably, in the setting of universal health care) used a similar geographic method to ours (although proportional, not population-based rates were calculated) and reported that high SES was associated with nearly twice the odds of perforation compared with the lowest quartile of SES. Our finding that AA has an age- and sex-standardized incidence with such significant, nonrandom geographic variation is a novel and important observation, and the association between incidence and SES requires validation in other populations along with investigations to ascertain mechanisms.

    Our overall estimations of incidence (106 and 29 per 100 000 PY for AA and PA, respectively) are in line with the existing literature. Addis et al8 evaluated 15 years of the National Hospital Discharge Survey (United States, 1970-1984) and calculated an overall incidence of AA of 110 per 100 000 PY8 in a study that also detected regional differences. Luckmann’s4 1984 data (California), with an age- and sex-adjusted incidence of 99.9 per 100 000 PY, were consistent with the temporal decrease observed in the Addis et al article.8 Ilves et al24 (Finland) also showed a decrease from 145 per 100 000 PY in 1987 to 98 per 100 000 PY in 2007,24 the latter number quite similar to our tricounty incidence. Lee et al19 (South Korea, 2005-2007) calculated an incidence of PA (29.1 per 100 000 PY) that is identical to our data.

    These similarities provide external validity to our methods for calculating incidence, which strengthens our finding of such variation within the highly-populated tricounty study area. With the prevailing assumption that AA is random in onset, one would expect a more homogenous distribution across the state. Not only is there variability, but hot spot and cold spot CTs were located in statistically significant clusters. There are several possibilities for why that might be: environmental factors that increase susceptibility, infectious processes that are geographically circumscribed, microbiome differences, socioeconomic factors, health care–seeking behaviors, or practice patterns at nearby hospitals. Likely, these differences in incidence are multifactorial.

    Regarding health care–seeking behaviors, Luckmann4 proposed that some observed trends among hospitalized patients with AA could be explained by differences in the number of patients who present with non-PA rather than differences among those with PA. If some patients with mild or self-limiting AA get better on their own and do not present to the hospital, evaluating proportions of PA among all patients with AA who do present to the hospital may generate misleading conclusions (assuming that all or most patients with PA become sick enough to require hospitalization).4 Inherent in this hypothesis is the notion that appendicitis can onset in more or less severe forms, and data are being accumulated to support this view, not only in terms of appendicitis that is successfully treated with antibiotics but also treated without antibiotics.34 Fitz,35 in the original description of appendicitis, indicated that up to one-third of the appendices in autopsy specimens showed evidence of prior inflammation. The disease appendicitis is almost certainly more complex than an obstructive process that progresses inexorably to perforation, and the ways in which socioeconomic factors influence both presentation and outcomes also appear to be more complex than a simple association between diminished health care access and increased risk of perforation.36 This population-based geographic analysis suggests that a more nuanced understanding of pathophysiology and the influence of SES is necessary to fully understand risk factors for both perforating and nonperforating appendicitis.

    Limitations

    This observational study has some limitations. Data sets based on discharge diagnoses can have misclassification. We designed our analysis to accommodate for this, but it is likely that some mischaracterization persists in the data set. Patients cared for at the Veterans Administration Medical Center or a military hospital in the state are not included, but prior research suggests this is only 0.5% of the cases of appendicitis in the state.21 Anyone traveling out of state when they had appendicitis would not be included, slightly reducing our calculated incidence. Disaggregation from zip codes to CTs has limitations, and an ideal data set would provide street addresses so that disaggregation was unnecessary. Finally, our study population, residents of Washington State and the tricounty Puget Sound, have high rates of insurance coverage and access to health care. Given Luckmann’s hypothesis4 that some patients with mild or self-limiting appendicitis may not present to the hospital, and given the traditional hypothesis that some patients delay presentation because of inadequate insurance coverage, variations in insurance status within a population may be associated with incidence of AA or PA. Thus, if this study was to be repeated in a population with less insurance coverage, the results might differ.

    Conclusions

    In summary, population-based incidence of AA is not randomly or uniformly spread across geographic space. Not only is there wide variability in incidence, areas of high incidence and low incidence are geographically clustered. Areas with low incidence of AA are more prosperous and more educated compared to areas with high incidence. Incidence of PA was also nonrandomly distributed, but the measurable association between geography and PA was only half as strong compared with AA. Although most studies have focused on associations between SES and rates of PA, our data suggest that associations between SES and incidence of overall AA may be equally or even more relevant. This population-based analysis also suggests that using proportions of PA of all cases of AA as a marker of access to high-quality surgical care is not appropriate. Our understanding of the pathophysiology of appendicitis and what pathways may lead to perforation or resolution is limited, but future investigators should prioritize elucidating the biologic, behavioral, and socioeconomic factors that drive the epidemiologic patterns we have detected.

    Back to top
    Article Information

    Accepted for Publication: December 1, 2019.

    Corresponding Author: F. Thurston Drake, MD, MPH, Boston Medical Center, Department of Surgery, Boston University School of Medicine, 820 Harrison Ave, FGH Building, 5th Floor, Boston, MA 02118 (frederick.drake@bmc.org).

    Published Online: March 4, 2020. doi:10.1001/jamasurg.2019.6030

    Author Contributions: Dr Drake and Mr Golz had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Golz, Flum, Liu, Donovan, Drake.

    Acquisition, analysis, or interpretation of data: Golz, Sanchez, Liu, Donovan, Drake.

    Drafting of the manuscript: Golz, Flum, Liu, Donovan, Drake.

    Critical revision of the manuscript for important intellectual content: Golz, Sanchez, Liu, Donovan, Drake.

    Statistical analysis: Golz, Liu, Drake.

    Administrative, technical, or material support: Golz, Flum, Sanchez, Drake.

    Supervision: Liu, Donovan, Drake.

    Conflict of Interest Disclosures: None reported.

    Meeting Presentations: This research was presented from the podium at the 2017 Annual Meeting of the American Association of Geographers; April 18, 2017; Boston, Massachusetts; and at the 2019 Academic Surgical Congress; February 8, 2019; Houston, Texas.

    References
    1.
    Boomer  L, Freeman  J, Landrito  E, Feliz  A.  Perforation in adults with acute appendicitis linked to insurance status, not ethnicity.  J Surg Res. 2010;163(2):221-224. doi:10.1016/j.jss.2010.04.041PubMedGoogle ScholarCrossref
    2.
    Nwomeh  BC, Chisolm  DJ, Caniano  DA, Kelleher  KJ.  Racial and socioeconomic disparity in perforated appendicitis among children: where is the problem?  Pediatrics. 2006;117(3):870-875. doi:10.1542/peds.2005-1123PubMedGoogle ScholarCrossref
    3.
    Braveman  P, Schaaf  VM, Egerter  S, Bennett  T, Schecter  W.  Insurance-related differences in the risk of ruptured appendix.  N Engl J Med. 1994;331(7):444-449. doi:10.1056/NEJM199408183310706PubMedGoogle ScholarCrossref
    4.
    Luckmann  R.  Incidence and case fatality rates for acute appendicitis in California: a population-based study of the effects of age.  Am J Epidemiol. 1989;129(5):905-918. https://www.ncbi.nlm.nih.gov/pubmed/2784936. doi:10.1093/oxfordjournals.aje.a115224PubMedGoogle ScholarCrossref
    5.
    Luckmann  R, Davis  P.  The epidemiology of acute appendicitis in California: racial, gender, and seasonal variation.  Epidemiology. 1991;2(5):323-330. https://www.ncbi.nlm.nih.gov/pubmed/1742380. doi:10.1097/00001648-199109000-00003PubMedGoogle ScholarCrossref
    6.
    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(6):886-892. doi:10.1097/01.sla.0000256391.05233.aaPubMedGoogle ScholarCrossref
    7.
    Andersson  R, Hugander  A, Thulin  A, Nyström  PO, Olaison  G.  Indications for operation in suspected appendicitis and incidence of perforation.  BMJ. 1994;308(6921):107-110. doi:10.1136/bmj.308.6921.107PubMedGoogle ScholarCrossref
    8.
    Addiss  DG, Shaffer  N, Fowler  BS, Tauxe  RV.  The epidemiology of appendicitis and appendectomy in the United States.  Am J Epidemiol. 1990;132(5):910-925. https://www.ncbi.nlm.nih.gov/pubmed/2239906. doi:10.1093/oxfordjournals.aje.a115734PubMedGoogle ScholarCrossref
    9.
    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(16):1977-1982. doi:10.1001/jama.292.16.1977PubMedGoogle ScholarCrossref
    10.
    Smink  DS, Fishman  SJ, Kleinman  K, Finkelstein  JA.  Effects of race, insurance status, and hospital volume on perforated appendicitis in children.  Pediatrics. 2005;115(4):920-925. doi:10.1542/peds.2004-1363PubMedGoogle ScholarCrossref
    11.
    Pieracci  FM, Eachempati  SR, Barie  PS, Callahan  MA.  Insurance status, but not race, predicts perforation in adult patients with acute appendicitis.  J Am Coll Surg. 2007;205(3):445-452. doi:10.1016/j.jamcollsurg.2007.04.010PubMedGoogle ScholarCrossref
    12.
    Bickell  NA, Aufses  AH  Jr, Rojas  M, Bodian  C.  How time affects the risk of rupture in appendicitis.  J Am Coll Surg. 2006;202(3):401-406. doi:10.1016/j.jamcollsurg.2005.11.016PubMedGoogle ScholarCrossref
    13.
    Gadomski  A, Jenkins  P.  Ruptured appendicitis among children as an indicator of access to care.  Health Serv Res. 2001;36(1 pt 1):129-142.PubMedGoogle Scholar
    14.
    Sicard  N, Tousignant  P, Pineault  R,  et al.  Provider density and health system facility factors and their relationship to rates of pediatric perforated appendicitis in US counties.  J Pediatr Surg. 2010;117(2):290-293. doi:10.1001/archsurg.2010.328Google Scholar
    15.
    Camp  M, Chang  DC, Zhang  Y,  et al.  Provider density and health system facility factors and their relationship to rates of pediatric perforated appendicitis in US counties.  Arch Surg. 2010;145(12):1139-1144. doi:10.1001/archsurg.2010.271PubMedGoogle ScholarCrossref
    16.
    Paquette  IM, Zuckerman  R, Finlayson  SRG.  Perforated appendicitis among rural and urban patients: implications of access to care.  Ann Surg. 2011;253(3):534-538. doi:10.1097/SLA.0b013e3182096d68PubMedGoogle ScholarCrossref
    17.
    Akhtar-Danesh  G-G, Doumouras  AG, Flageole  H, Hong  D.  Geographic and socioeconomic predictors of perforated appendicitis: a national Canadian cohort study.  J Pediatr Surg. 2018;(November). doi:10.1016/j.jpedsurg.2018.10.065PubMedGoogle Scholar
    18.
    Sicard  N, Tousignant  P, Pineault  R, Dubé  S.  Non-patient factors related to rates of ruptured appendicitis.  Br J Surg. 2007;94(2):214-221. doi:10.1002/bjs.5428PubMedGoogle ScholarCrossref
    19.
    Lee  SL, Shekherdimian  S, Chiu  VY.  Effect of race and socioeconomic status in the treatment of appendicitis in patients with equal health care access.  Arch Surg. 2011;146(2):156-161. doi:10.1001/archsurg.2010.328PubMedGoogle ScholarCrossref
    20.
    Walker  A, Hatch  Q, Drake  T,  et al.  Predictors of appendiceal perforation in an equal access system.  J Surg Res. 2014;190(1):87-92. doi:10.1016/j.jss.2014.02.028PubMedGoogle ScholarCrossref
    21.
    Flum  DR, Morris  A, Koepsell  T, Dellinger  EP.  Has misdiagnosis of appendicitis decreased over time? a population-based analysis.  JAMA. 2001;286(14):1748-1753. https://www.ncbi.nlm.nih.gov/pubmed/11594900. doi:10.1001/jama.286.14.1748PubMedGoogle ScholarCrossref
    22.
    Grubesic  TH, Matisziw  TC.  On the use of ZIP codes and ZIP code tabulation areas (ZCTAs) for the spatial analysis of epidemiological data.  Int J Health Geogr. 2006;5:58. doi:10.1186/1476-072X-5-58PubMedGoogle ScholarCrossref
    23.
    Health NYSD. Rates based on small numbers. https://www.health.ny.gov/diseases/chronic/ratesmall.htm. Accessed November 15, 2018.
    24.
    Ilves  I, Paajanen  HEK, Herzig  KH, Fagerström  A, Miettinen  PJ.  Changing incidence of acute appendicitis and nonspecific abdominal pain between 1987 and 2007 in Finland.  World J Surg. 2011;35(4):731-738. doi:10.1007/s00268-011-0988-8PubMedGoogle ScholarCrossref
    25.
    von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.  Lancet. 2007;370(9596):1453-1457. doi:10.1016/S0140-6736(07)61602-XPubMedGoogle ScholarCrossref
    26.
    Winkleby  MA, Jatulis  DE, Frank  E, Fortmann  SP.  Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease.  Am J Public Health. 1992;82(6):816-820. doi:10.2105/AJPH.82.6.816PubMedGoogle ScholarCrossref
    27.
    Kelli  HM, Mehta  A, Tahhan  AS,  et al.  Low educational attainment is a predictor of adverse outcomes in patients with coronary artery disease.  J Am Heart Assoc. 2019;8(17):e013165. doi:10.1161/JAHA.119.013165PubMedGoogle Scholar
    28.
    Andersson  MA.  Higher education, bigger networks? differences by family socioeconomic background and network measures.  Socius Sociol Res a Dyn World. 2018;4:237802311879721. doi:10.1177/2378023118797217Google Scholar
    29.
    Pinsky  JL, Leaverton  PE, Stokes  J  III.  Predictors of good function: the Framingham Study.  J Chronic Dis. 1987;40(1)(suppl 1):159S-167S, 181S-2. doi:10.1016/S0021-9681(87)80045-0PubMedGoogle ScholarCrossref
    30.
    Huisman  M, Kunst  AE, Bopp  M,  et al.  Educational inequalities in cause-specific mortality in middle-aged and older men and women in eight western European populations.  Lancet. 2005;365(9458):493-500. doi:10.1016/S0140-6736(05)17867-2PubMedGoogle ScholarCrossref
    31.
    Kubota  Y, Heiss  G, MacLehose  RF, Roetker  NS, Folsom  AR.  Association of educational attainment with lifetime risk of cardiovascular disease: the atherosclerosis risk in communities study.  JAMA Intern Med. 2017;177(8):1165-1172. doi:10.1001/jamainternmed.2017.1877PubMedGoogle ScholarCrossref
    32.
    Drewnowski  A, Rehm  CD, Solet  D.  Disparities in obesity rates: analysis by ZIP code area.  Soc Sci Med. 2007;65(12):2458-2463. doi:10.1016/j.socscimed.2007.07.001PubMedGoogle ScholarCrossref
    33.
    Moudon  AV, Cook  AJ, Ulmer  J, Hurvitz  PM, Drewnowski  A.  A neighborhood wealth metric for use in health studies.  Am J Prev Med. 2011;41(1):88-97. doi:10.1016/j.amepre.2011.03.009PubMedGoogle ScholarCrossref
    34.
    Park  HC, Kim  MJ, Lee  BH.  Randomized clinical trial of antibiotic therapy for uncomplicated appendicitis.  Br J Surg. 2017;104(13):1785-1790. doi:10.1002/bjs.10660PubMedGoogle ScholarCrossref
    35.
    Fitz  RH. Perforating Inflammation of the Vermiform Appendix; with Special Reference to Its Early Diagnosis and Treatment. Philadelphia, PA: WM J. Dornan, Printer; 1886.
    36.
    Søreide  K.  The research conundrum of acute appendicitis.  Br J Surg. 2015;102(10):1151-1152. doi:10.1002/bjs.9890PubMedGoogle ScholarCrossref
    ×