Assessment of Sex Differences for Treatment, Procedures, Complications and Associated Conditions among Adolescents Hospitalized with Crohn’s Disease

Abstract

Background

Sex differences among adults in healthcare treatment and outcomes have been reported, however, there is a paucity of literature regarding pediatric populations, particularly adolescents with Crohn’s disease (CD). The objective was to identify whether sex differences exist with respect to complications, procedures, and medication usage (corticosteroids, biologic agents, and total parenteral nutrition (TPN)) among hospitalized adolescents with CD.

Methods

Adolescents with CD (n=5782) hospitalized between April 1, 2004–June 30, 2012 were selected from the Pediatric Health Information System database with a 1:1 ratio of males to females by hospital. Frequency of disease complications, associated conditions, procedures performed, and medication usage were analyzed with non-parametric statistical tests for the existence of sex differences.

Results

5782 patients were included with a median age of 15 years. Females were slightly more likely to have anemia (29% vs. 25%, p=0.012), infection (12% vs. 8%, p=0.001), and mood disorder (9% vs. 6%, p<0.001), whereas males had more maturational delays (3% vs. 1%, p=0.004), and malnutrition (18% vs. 14%, p=0.027). Among procedures, only one category demonstrated a sex difference: females had more blood product transfusions (9% vs. 6%, p<0.001). Female rates for corticosteroids (62%), biologic agents (16%) and TPN (18%) were not statistically different from those for males (62%, 15%, and 20%, respectively). There were no differences in length of stay by sex.

Conclusions

This cross-sectional study of an administrative database identified few sex differences among adolescents with CD. The effect sizes were universally small and generally consistent with known sex differences unrelated to IBD.

Introduction

Understanding sex differences in disease symptoms and response to treatment may be an important pathway for understanding disease etiology, treatment and improving overall care^{1, 2}. Investigations into sex differences have alerted providers in diverse fields to potential biases in diagnosis and treatment^2–4 and suggested changes to traditional practice.^5–7. While sex differences in prevalence, treatment, and patient outcomes have been identified and studied among adults(e.g., depression, cardiovascular, and autoimmune diseases^{1, 8–13}), much less is known concerning such differences in pediatric populations, particularly for Crohn’s disease (CD), a chronic, relapsing, autoimmune disease.

Consistent evidence for sex differences in the treatment, course and outcomes of both adult and pediatric patients with CD is lacking. A large retrospective cohort analysis of adults in the Kaiser Permanente Inflammatory Bowel Disease registry found no differences in the risk for rehospitalization or colectomy.¹⁴ In other adult cohort studies, females were more likely to have extraintestinal manifestations (EIMs, such as arthritis, erythema nodosum, and ocular manifestations),^{15, 16} while males had higher disease remission rates¹⁷ but a greater decline in bone mineral density compared with non-CD controls.^{18, 19}

Prior studies in children are conflicting: demonstrating either greater severity in males, greater severity in girls, or finding no sex differences. For example, males appear to be at a higher risk than females for growth failure.^20–22 In contrast, females may be more likely to have hypoalbuminemia and mouth sores at diagnosis, and generally have a more severe disease course.²⁰ In a study assessing risk factors for intestinal resection surgery, females were at greater risk compared to males.²³ Several studies have reported no sex differences with respect to patient symptoms or EIMs,²⁴ risk of intestinal resection surgery,^{23, 25, 26} or in disease severity, BMI, height velocity, or medication use.²⁷

Given the conflicting evidence among mostly small-scale studies in the pediatric CD literature, the present study sought to address this gap by using a large administrative database with diagnostic and treatment data across multiple, geographically diverse children’s hospitals to evaluate a cross-section of hospitalized, male and female adolescent patients with respect to medical management, complications, and procedures. If sex plays a significant role in Crohn’s etiology and outcomes, this effect may be pronounced during puberty. Further, as adolescence is a period of peak onset of CD, it may be an optimal time to assess potential sex differences. The main objective of the present study was to identify whether sex differences exist with respect to complications, procedures, and medication usage (corticosteroids, biologic agents and total parenteral nutrition (TPN)) among hospitalized adolescent patients with CD.

Materials and Methods

Data Source

The Pediatric Health Information System (PHIS) is an administrative database established and maintained by the Children’s Hospital Association (CHA) on behalf of its 44 affiliated children’s hospitals which comprise approximately 25% of pediatric centers in the United States. Data (e.g. demographics, diagnoses, procedures, and medications) are collected in standardized formats for each hospitalized patient. Reliability and validity are assured by each hospital and by CHA before database inclusion. For this study period, data from 44 hospitals were available and all 44 hospitals were included.

Study Design

This was a cross-sectional study of male and female adolescent patients who were hospitalized with a diagnosis of CD. This study was approved by the Nationwide Children’s Hospital Institutional Review Board.

Study Cohort

The study population was derived from adolescent patients (age 13 to 18 years) identified in the PHIS database by a primary or secondary diagnosis of Crohn’s disease (ICD-9: 555.x) during their first (index) hospital admission within the study time period of January 1, 2004 and June 30, 2012. This age range was selected as a simple way to encompass the majority of pubertal/growth changes for both sexes. To be included, patients had to have sex recorded, and no admissions in the preceding three months. Patients with admission dates within the first three months (January 1, 2004 to March 31, 2004) were excluded in order to increase the likelihood that their initial hospitalization occurring at the beginning of the study period was not a readmission related to a prior hospitalization. Although this study did not consider encounters outside of the initial hospitalization, patients were also excluded if there were inconsistencies in the reporting of sex or birthdate across encounters. This more conservative approach allowed us to reduce reporting errors. Specific hospitalizations were excluded if the encounter was for an ostomy take down (CPT 44625, ICD-9 procedure code 46.5x) or if the encounter had the same admission and discharge date.

Subsequent to applying inclusion and exclusion criteria, a subset of male and female subjects was randomly selected to achieve a 1:1 ratio at the hospital level (Figure 1).

FIGURE 1.

Flowchart for the final cohort selection process and exclusion criteria. After applying exclusion criteria to the initial database sample of patients age 13 to 18 years, and subsequently applying a 1:1 ratio of males to females within each hospital, the final study cohort was determined. MRN = medical record number.

Measures

Demographic descriptors collected for each subject (distinguished by encrypted medical record number (MRN)) included sex, age, race, ethnicity, payor status, ZIP code (encrypted), median neighborhood income, and hospital location. Payor status was categorized into one of three groups: commercial insurance, Medicaid, or other (APPENDIX A). Individuals were initially assigned to one of 4 race categories: a specific race category (e.g., White, Black, Asian, if they consistently endorsed the same race across all encounters), “mixed race” (if they endorsed more than one race), “missing” (if race was consistently missing across encounters), or “inconsistent” (if there were any reporting discrepancies). For demographic analysis, race and ethnicity were defined as: non-Hispanic White, non-Hispanic Black, non-Hispanic other (including Asian, Pacific Islander, American Indian, mixed-race, and other), and Hispanic.

Additional variables collected included admission and discharge dates, discharge ID, intensive care unit (ICU) admission flag, length of stay (LOS), and use of specific medications (corticosteroids, biologic agents and TPN). Median neighborhood income was derived by the PHIS data analytics team using the United States Census-based values corresponding to their actual ZIP codes. Regional divisions (Northeast, South, Midwest, and West) were derived from hospital location based on the 2010 Census Regions and Divisions of the United States (www.census.gov).

To assess the use of the most frequent procedures and surgical complications among all patients in the final study cohort, lists of the top 200 procedural codes and complication diagnoses codes were generated. From these two lists, procedural and complication/associated diagnoses groups were constructed based on clinical relevance to CD. Compared to a simple search of specific ICD-9 codes, this approach was less likely to inadvertently overlook common procedural and diagnostic codes. (APPENDIX B)

Outcomes

The primary outcome measures included the presence of complications, procedure utilization, and medication utilization. The secondary outcome was LOS.

Analysis plan

The Mann-Whitney U test was used to identify sex differences on age at admission and median income. The Chi-square test for contingency tables was used to identify sex differences in the proportion of male versus female subjects on race/ethnicity and payor status, medications received, complications, and procedures performed. All tests were two-sided. For each set of analyses, we assumed a critical alpha of 0.05 and made Bonferroni corrections for multiple testing where appropriate.

Effect sizes were calculated (e.g. phi (ϕ) for Chi-square contingency tables) to describe the magnitude of statistically significant sex differences. We considered an effect size greater than 0.15 to be clinically meaningful; although a small effect may not be apparent in individual practice, it is important from a population health perspective where small effects are magnified in the larger population of affected individuals. With 2891 males and 2891 females, this study has at least 80% power to detect small effects using a X² test with a critical p-value of 0.05. Analyses were performed using R 3.0.1 (R foundation of Statistical Computing, Vienna, 2012).

Results

Demographics

Table 1 summarizes the demographics according to sex. By design (1:1 ratio of males to females within each hospital), there were no sex frequency differences by region.

TABLE 1.

Demographic Characteristics by Sex for a Hospitalized Adolescent Crohn’s Disease Cohort

Variable	Female n (%)	Male n (%)	P value
Total	2891 (50)	2891 (50)	n/a
Race/Ethnicity
Non-Hispanic White	2024 (70)	1936 (67)	0.194
Non-Hispanic Black	475 (16)	512 (18)
NH Other	173 (6)	204 (7)
Hispanic	133 (5)	150 (5)
Missing	57 (2)	63 (2)
Inconsistent	29 (1)	26 (1)
Age, Median [Q1–Q3]1	15 [14–17]	15 [14–17]	0.081
Region2
Midwest	857 (30)	857 (30)	n/a
Northeast	701 (24)	701 (24)
South	946 (33)	946 (33)
West	387 (13)	387 (13)
Payor3
Commercial	1276 (44)	1255 (43)	0.592
Medicaid	605 (21)	647 (22)
Other	969 (34)	946 (33)
Missing	41 (1)	43 (1)
Neighborhood income, Median4 [Q1–Q3]	$46,982 [$36,111–$61,059]	$46,565 [$35,975–$61,275]	0.769

¹Median age represented with quartile ranges: Q1 and Q3 refer to the first (25%) and third (75%) quartiles

²Region assignment: Midwest (OH, IL, MI, IN, MO, WI, MN, NE, ND, SD, KS, IA); Northeast (ME, NH, VT, NY, MA, CT, RI, PA, NJ); South (DC, MD, VA, DE, WV, KY, TN, NC, SC, GA, AL, MS, FL, LA, TX, AR, OK); West (AZ, CO, CA, WA, OR, MT, WY, ID, NV, UT, NM, HI, AK)

³Payor status was collated into the following categories: Commercial=Blue Cross, HMO, TRICARE, Commercial HMO, Commercial PPO, Commercial Other; Medicaid=Medicaid, In-state Medicaid (managed care), In-state Medicaid (other), Out-of-state Medicaid (all); Other=Medicare, Title V, Other government, Workers Compensation, other insurance company, self-pay, no charge, other payor, charity, hospital chose not to bill; Missing=not recorded, invalid code, unknown

⁴Median of median income based on 2010 US Census Data compared to ZIP code reported at index hospitalization

Medication usage

Female rates for corticosteroids (n = 1781, 62%), biologic agents (n = 460, 16%) and TPN usage (n = 509, 18%) were not statistically different from those for males (n = 1798, 62%, p = 1; n = 444, 15%, p = 1; and n = 570, 20%, p = 0.129 (adjusted based on Bonferroni correction), respectively).

Complications

Among the top 200 diagnoses reported, 75 were found to have clinical relevance to CD. The frequency of reported ICD-9 codes ranged from 0.4% for “Intestinal obstruction NOS” (ICD-9 code = 560.9; not otherwise specified (NOS)) to 11.1% for “Anemia NOS” (ICD-9 code=285.9). There were sex differences in 4 of 19 diagnostic groups identified. Females were slightly more likely to have anemia (29% versus 25%, p = 0.012, ϕ = 0.04), infection (12% versus 8%, p = 0.001, ϕ = 0.05), and mood disorder (9% versus 6%, p < 0.001, ϕ = 0.07), whereas males had more pubertal complications (such as delays in maturation, including short stature and sexual development; 3% versus 1%, p = 0.004, ϕ = 0.05), and more nutritional complications (including malnutrition and under nutrition, 18% versus 14%, p = 0.027) (Table 2). With respect to greater infections among females, it should be noted that these were largely explained by urinary tract infections (UTIs); when UTIs were removed from the infection category in a post hoc analysis, the sex difference for the remaining infections as a group was diminished and no longer significant (10% female versus 8% males, X² = 7.76, p = 0.102, ϕ = 0.04).

TABLE 2.

Crohn’s Disease Related Diagnoses at Index Hospitalization

Variable	Total	Female n (%)	Male n (%)	P value1
Anemia	1552	834 (29)	718 (25)	0.012
Nutrition complications	928	419 (14)	509 (18)	0.027
Dehydration/hypovolemia	599	313 (11)	286 (10)	1
Infection (others)	585	340 (12)	245 (8)	0.001
Obstruction	553	243 (8)	310 (11)	0.060
Perianal disease	473	207 (7)	266 (9)	0.102
Mood Disorders	439	270 (9)	169 (6)	<0.001
EIM	383	185 (6)	198 (7)	1
Intra-abdominal abscess	326	149 (5)	177 (6)	1
Electrolyte anomalies	220	111 (4)	109 (4)	1
Other	213	104 (4)	109 (4)	1
Fistula	167	79 (3)	88 (3)	1
Irritable Bowel Syndrome	154	90 (3)	64 (2)	0.760
C. difficile infection	145	72 (2)	73 (3)	1
Perforation/peritonitis	115	57 (2)	58 (2)	1
Pubertal complications	112	36 (1)	76 (3)	0.004
Autoimmune	63	42 (1)	21 (1)	0.190
Bacteremia/sepsis	48	29 (1)	19 (1)	1
Vitamin D deficiency	43	19 (1)	24 (1)	1

¹Bonferroni corrected p-values

Total cohort n=5782 (Female=2891 and Male=2891)

Bolded p values represent significant findings

Procedures

Among 84 CD-related procedures identified, the frequency of ICD-9 codes ranged from 0.1% for “Other laparoscopic partial excision of large intestine” (ICD-9 code = 17.39) to 34.8% for “Closed [endoscopic] biopsy of large intestine” (ICD-9 code = 45.25). Of 13 procedural categories (APPENDIX B), only one demonstrated a sex difference: females had more blood product transfusions compared to males (9% versus 6%, p=0.002, ϕ =0.05) (Table 3). There were no differences in LOS between males (mean = 6.5 ± 10.0, median = 4 days) and females (mean = 6.5 ± 7.4, median = 4 days) (p = 0.280, d = 0.002).

TABLE 3.

Crohn’s Disease Related Procedures at Index Hospitalization

Variable	Total	Female n (%)	Male n (%)	P value1
Endoscopies	2241	1116 (39)	1125 (39)	1
Colectomy	436	211 (7)	225 (8)	1
Blood product transfusions	426	251 (9)	175 (6)	0.002
Resection/anastomosis	227	97 (3)	130 (4)	0.390
Abdominal surgery	219	105 (4)	114 (4)	1
Wound/lesion related	184	87 (3)	97 (3)	1
Perianal/perineal procedures	181	80 (3)	101 (3)	1
Appendectomy	145	67 (2)	78 (3)	1
Ostomy	110	54 (2)	56 (2)	1
Central access	95	40 (1)	55 (2)	1
Other	55	32 (1)	23 (1)	1
Intra-abdominal fistula repair	25	11 (0)	14 (0)	1
Cholecystectomy	12	7 (0)	5 (0)	1

¹Bonferroni corrected p-values

Total cohort n=5782 (Female=2891 and Male=2891)

Bolded p values represent significant findings

Discussion

This study identified several small sex differences among CD related diagnoses and procedures, much of which is explained by known sex differences in the general population. The observed sex difference in anemia was expected given the increased prevalence of anemia in adolescent girls as compared to boys in the general population. The third National Health and Nutrition Examination Survey (NHANES III) found a prevalence for iron-deficiency anemia of 2–5% among non-pregnant female adults (2–3% among adolescents) and 1–2% in male adults (< 1% among adolescents).²⁸

Given that females have a higher baseline rate of anemia; it may not be surprising to have found a small difference in the rate of transfusions in our study. However, there are other sex related factors that may contribute to blood transfusions being more prevalent among adolescent females (9%) compared with males (6%). Androgens have a stimulating effect on erythropoiesis,²⁹ with lower androgen levels and therefore lower hemoglobin concentrations in adolescent females.³⁰ Also, given the wide inter-hospital variation in thresholds for transfusion,^31–33 such decisions are likely based on a combination of objective and subjective factors. Since females have been found to have a slightly greater tendency for self-disclosure,³⁴ they may be more willing to communicate (and providers more likely to respond to³⁵) subjective symptoms such as dizziness, headache, and fatigue, which may increase their likelihood for anemia-related transfusions in less severe cases. Finally, if females have a more severe disease course during initial hospitalization as previously suggested, a greater need for transfusion might be seen.²⁰

As with anemia, we found sex differences for disorders involving depression as well as anxiety that may be explained by sex differences in the general population, with a higher prevalence of mood disorders among females.³⁶ Sex-based differences in brain functioning as well as circulating gonadal steroids have different effects on cortisol response and hypothalamic-pituitary-adrenal axis reactivity. In addition to such biological differences, it has been speculated that the higher female incidence for depression and anxiety may reflect psychological differences, such as a differential response to stress from interpersonal life events.^37–40 Similarly, although we found differences in rates of infections, this was no longer significant after removing UTIs, a condition known to be more common among females during adolescence in the general population.⁴¹

In contrast, the difference in abnormal growth and development in our study are not readily explained by differences in the general population, and are consistent with findings from previous research. Growth failure, delayed puberty, subnormal height velocity, and short stature are all commonly observed among adolescents with CD generally^42–44. A previous study showed the risk for growth failure (height or height velocity at the fifth percentile) among girls was 0.28 times the risk in boys (p < 0.002) over a ten year follow-up, with a final cumulative incidence of 4.0% versus 12.6%.²⁰ The rate of complications of growth and development in our study was lower than previously reported. We expected to observe a lower prevalence (under-reporting) of these conditions in our study cohort given that our study population was strictly an inpatient cohort. Growth, height velocity, and puberty are more likely to be clinically assessed, managed, and followed longitudinally, and thus more likely to be entered as diagnoses codes in an outpatient setting.

There were no sex differences among corticosteroid, biologic agent, or TPN usage. While no prior studies of CD medications in a hospitalized pediatric population were identified for comparison, there were three relevant studies of adult populations. Females were found to receive immunosuppressive medications less frequently,¹⁷ have lower cumulative corticosteroid use,¹⁸ and were slightly less likely to initiate infliximab infusions (non-significant trend)⁴⁵ compared with males. Our results were consistent, however, with a large outpatient IBD study which also found no sex differences in medication use.²⁷ Perhaps some of the differences that were seen in adults could be explained by less disease severity among adult women, impact of disease duration, or effects of treatment decisions which occur during pregnancy or child-bearing years.

There were several important limitations in the present study. We evaluated only the first index hospitalization for each patient, which does not address potential differences in outpatient management (not captured in PHIS) or in subsequent hospitalizations. We recognize that CD is mainly treated on an outpatient basis without hospitalization. The outpatient management and clinic visits are integral components in the management of adolescents with CD. Further, the PHIS database is not weighted for extrapolation to national estimates, thus results may not be generalizable to all pediatric centers. Also, this was a retrospective study based on administrative data, so reporting bias or classification errors may be present, although there are data quality measures in place to minimize errors. Since the findings were not based on detailed medical record review, we were unable to assess or control for disease phenotype, location or symptoms. Finally, while we believe that small observed differences are likely due to biological causes, we cannot rule out the possibility that they may also be due to other factors such as differential access to care, adherence, or treatment based on sex^{46, 47}. The strengths of the study include its large sample size, regionally diverse population, and the focus on a hospitalized cohort with likely moderate to severe disease, thereby minimizing any confounding of severity with sex differences. We also focused on the adolescent population, which allowed us to better observe hormonal-based sex differences in a pediatric cohort.

In conclusion, this cross-sectional study of an administrative database identified few sex differences among adolescents with CD. The effect sizes were universally small and consistent with the expectations in the general population. Sex did not have a large effect on the outcomes assessed in this study.

Abbreviations

CD

Crohn’s disease

IBD

inflammatory bowel disease

LOS

length of stay

PHIS

Pediatric Health Information System

TPN

total parenteral nutrition