RI-1

Health Services Use during Transition from Pediatric to Adult Care for
Inflammatory Bowel Disease: A Population-Based Study Using
Health Administrative Data

Xinbei Zhao, MSc1,2,3, Lise M. Bjerre, MDCM, PhD2,3,4,5, Geoffrey C. Nguyen, MD, PhD3,6, David R. Mack, MD1,7, Douglas G. Manuel, MD, MSc2,3,4,8, Steven Hawken, PhD2,3,8, Nassim Mojaverian, MSc3, Natasha Bollegala, MD, MSc9, and
Eric I. Benchimol, MD, PhD1,2,3,7

Objectives To evaluate the impact of the transfer from pediatric to adult care on health services use for adoles- cents with inflammatory bowel disease (IBD).
Study design A population-based retrospective cohort study identified all children diagnosed with IBD from 1994 to 2008 and treated by pediatric gastroenterologists in Ontario, Canada, using health administrative data. Self- controlled case series analyses compared health service use in the 2 years before and 2 years after transfer with adult gastroenterologists, with a 6-month washout period at transfer. Outcomes evaluated included IBD-specific and IBD-related hospitalizations, emergency department use, outpatient visits, and laboratory use. The relative in- cidence (RI) in the post-transfer period was compared with pretransfer periods using Poisson regression analysis controlling for transfer starting age. Analyses were stratified by IBD type: Crohn’s disease (CD) and ulcerative colitis (UC).
Results There were 536 patients included in the study (388 CD, 148 UC). Emergency department use rate was higher after transfer for both CD (RI, 2.12; 95% CI, 1.53-2.93) and UC (RI, 2.34; 95% CI, 1.09-5.03), as were out- patient visits (CD: RI, 1.56; 95% CI, 1.42-1.72; UC: RI, 1.48; 95% CI, 1.24-1.76), and laboratory investigations (CD: RI, 1.43; 95% CI, 1.26-1.63; UC: 1.38; 95% CI, 1.13-1.68). There was no change in the hospitalization rate (CD: RI, 0.70; 95% CI, 0.42-1.18; UC: RI, 2.41; 95% CI, 0.62-9.40).
Conclusions Health services use in Canada increases after transfer from pediatric to adult care for outpatient visits, emergency department use, and laboratory tests, but not hospitalizations. This study has implications for
the planning and budgeting of care for adolescents transitioning to adult care. (J Pediatr 2018;■■:■■-■■).

he inflammatory bowel diseases (IBD), and their 2 main subtypes Crohn’s disease (CD) and ulcerative colitis (UC), are complex, multifactorial diseases characterized by chronic inflammation of the gastrointestinal tract.1 Canada has one of the highest reported incidence and prevalence of IBD in the world.2 In particular, in children <16 years old the in- cidence and prevalence was reported to be 9.68 and 38.25 per 100 000 children, respectively,3 second only to Norway’s inci- dence of 10.6 per 100 000.4 In addition, the incidence of IBD is increasing most rapidly in the youngest age groups of children.3,5 Children with IBD are being cared for more frequently by pediatric gastroenterologists compared with adult gastroenterolo- gists or generalists.6 As these children age, more children undergo transition from pediatric care to an adult model of healthcare. Transition is the entire process of transfer from a pediatric model of health- care to an adult provider, including medical, psychosocial, and educational prepa- ration, and physical transition.7 The transfer of care between providers is only part of the full transition process. Children with chronic diseases may have difficulty with the transition from pediatric to adult care, which may result in adverse out- comes.8,9 Although the effectiveness of transition models has been studied, the impact CD Crohn’s disease ED Emergency department IBD Inflammatory bowel disease ICD International Classification of Diseases OCCC Ontario Crohn’s and Colitis Cohort OHIP Ontario Health Insurance Plan RI Relative incidence SCCS Self-controlled case series UC Ulcerative colitis From the 1Children’s Hospital of Eastern Ontario IBD Centre, Division of Gastroenterology, Hepatology and Nutrition, CHEO Research Institute, Children’s Hospital of Eastern Ontario; 2School of Epidemiology and Public Health, University of Ottawa, Ottawa; 3Institute for Clinical Evaluative Sciences, Toronto; 4Department of Family Medicine, University of Ottawa; 5Bruyère Research Institute, Ottawa; 6Mount Sinai Hospital Centre for Inflammatory Bowel Disease, Department of Medicine, University of Toronto, Toronto; 7Department of Pediatrics, University of Ottawa; 8Ottawa Hospital Research Institute, Ottawa; and 9Women’s College Hospital, Division of Gastroenterology, Department of Medicine, University of Toronto, Toronto, Canada Funding and Disclosures available at www.jpeds.com. Portions of this study were presented as a poster at the Canadian Association of Health Services and Policy Research Annual Meeting, May 28, 2015, Montreal, Quebec, Canada, and as a poster at the World Congress of Pediatric Gastroenterology, Hepatology and Nutrition, October 5-8, 2016, Montreal, Quebec, Canada. 0022-3476/$ - see front matter. © 2018 Elsevier Inc. All rights reserved. https://doi.org10.1016/j.jpeds.2018.08.021 1 FLA 5.5.0 DTD ■ YMPD10213_proof ■ September 25, 2018 of the transition process on the health system has not been well- evaluated. In a small cohort of 96 patients, Bollegala et al found fewer clinic visits in the year after transfer from pediatric to adult IBD care.10 Ontario is Canada’s most populous province (with a popu- lation of 14 million) and has a universal single payer health- care system in which all basic outpatient, inpatient, and emergency care is provided by government funding and records of health system use are contained within population-based health administrative data. We used these data to determine how the transfer from pediatric to adult care in youth with IBD impacts a universal healthcare system. We measured health ser- vices use before and after transfer, including outpatient phy- sician visit rates, hospitalizations, emergency department (ED) use, and laboratory tests to determine the rates of both desir- able and undesirable health services outcomes. Methods This study was approved by the Research Ethics Boards of the Children’s Hospital of Eastern Ontario and The Ottawa Hos- pital. We conducted a retrospective cohort study using health administrative data of all patients in Ontario, Canada, with pediatric-onset IBD who were treated initially by a pediatric gastroenterologist and transferred to an adult gastroenterolo- gist. Ontario administrative data contain all residents of the province diagnosed with IBD during the study period of fiscal years 1994 to 2008 (April 1, 1994, to March 31, 2009) who quali- fied for universal government healthcare insurance (>99% of the population).
Adolescents treated by a pediatric gastroenterologist (and underwent transfer to adult care) would be systemically dif- ferent from adolescents never treated by a pediatric gastroenterologist (and never transferred to adult care). For example, children with IBD only treated by an adult gastro- enterologist may have less severe disease, be less likely to require a biologic or immunomodulator, live farther from a pediatric care center, or have less access to specialized health- care services. Therefore, unmeasured or measured confounding may result in bias when comparing outcomes in those treated by pediatric gastroenterologists compared with those who were not. Therefore, we used a self-controlled case series (SCCS) design, in which a patient acts as his or her own control, to adjust for unmeasured time-invariant confound- ing factors such as disease severity at the time of transfer. This strategy has been used successfully in determining rare vaccine-related outcomes, in which vaccinated people cannot reliably be compared with unvaccinated people.11,12 Each sub- ject’s control period started with the date of diagnosis of IBD and ended with the first visit to an adult gastroenterolo- gist. We defined the transfer period as the time between the first adult gastroenterologist visit and the last pediatric gas- troenterologist visit. If the subject had shared care from both a pediatric and adult gastroenterologist during this period, the transfer period duration was assigned as the number of days from the first adult gastroenterologist visit to the final pediatric gastroenterologist visit. If the subject did not go

back to a pediatric gastroenterologist after their first adult gastroenterologist visit, their transfer period length was as- signed as 1 day. We defined a washout period of 180 days (6 months) after the first adult gastroenterologist visit to ensure that events after transfer were due to the transfer and not due to clinical status during the control period. The risk period started from either the last day of the transfer period if the transfer period was longer than the washout period, or otherwise from the last day of the washout period (Figure 1). We conducted sensitivity analyses to vary the washout periods, assessing no washout period, as well as washout periods of 3 months, 6 months (our original analysis), 9 months, and 12 months. In addition, a sensitivity analysis was performed to assess the risk of events in a 1-year risk period (instead of 2 years), compared with one-year control period.
The study cohort consisted of all patients <18 years old di- agnosed with IBD between April 1, 1994, and March 31, 2009. Patients were included if they had ≥2 visits to each of an adult and a pediatric gastroenterologist, ≥730 days in both the control and risk periods, and a transfer period that was ≤365 days. Pa- tients were excluded if they never saw a pediatric gastroen- terologist and, therefore, did not experience transfer or if they had <2 visits to an adult gastroenterologist beyond their 18th birthday (considered failed transfer). Patients were also ex- cluded if they could not be classified as having either CD or UC using validated algorithms.13 Data Sources Ontario’s health administrative data are maintained by the In- stitute for Clinical Evaluative Sciences through an agreement with the Ontario Ministry of Health and Long-Term Care, with all data available in uncleaned and unedited format to re- searchers.14 The Ontario Crohn’s and Colitis Cohort (OCCC) is a patient-derived cohort created using validated algo- rithms to identify patients with IBD using the health admin- istrative data.13,15 Accuracy of the identification algorithms was determined to be a sensitivity of 89.6%-91.1%, specificity of 99.5%-100%, positive predictive value of 59.2%-76.0%, and negative predictive value of >99.9%.13 Patients were classified as having CD or UC using validated algorithms that deter- mined subtype using the most recent physician billing codes, which distinguished patients with CD from patients with UC with a sensitivity of 95.1%, specificity of 86.0%, positive pre- dictive value of 92.0%, and negative predictive value of 91.2%.13 Full health administrative data were available from April 1, 1991. Validated look-back periods to distinguish incident from preva- lent cases were 3 years for childhood-onset IBD (diagnosed at
<18 years of age).13 Therefore, incidence cases were available from 1994. The date of diagnosis was defined as the date of first contact with a physician or first hospitalization with an associated IBD-specific diagnostic code. The OCCC also con- tains a unique patient identification number that allows in- dividual deterministic linkage to all other provincial health administrative and population data. To determine hospitalization rates, we used discharge ab- stract data collected from all hospitals and reported to the Ca- nadian Institute for Health Information, with diagnostic codes 2 Zhao et al Figure 1. Examples of timelines of the control, transfer, washout, and risk periods used in the study. A, Standard timeline (in which the last pediatric gastroenterologist visit occurred after the first adult gastroenterologist visit). B, Timeline in which the transfer period is one day (in which the last pediatric gastroenterologist visit occurred before the first adult gastroenterologist visit). known to be 92%-99% accurate.16 Outpatient healthcare use included IBD-specific and IBD-related physician visits, ED visits, and laboratory visits for investigations, and were derived from the Ontario Health Insurance Plan (OHIP) database. The Reg- istered Persons Database provided date of birth, date of death (where applicable), and mean neighborhood income quintile on the date of diagnosis. Information of the specialty of health- care provider for each visit came from Institute for Clinical Evaluative Sciences Physician Database and OHIP database. Certification status of physicians in Ontario recorded in the Institute for Clinical Evaluative Sciences Physician Database were used to classify visits to pediatric and adult gastroenter- ologists, in addition to visits to internists and general pedia- tricians. Not all physicians functioning as gastroenterologists are certified as such owing to more recent inception of spe- cialty certification examinations. Therefore, internists who per- formed >100 sigmoidoscopies or colonoscopies per year and pediatricians who performed >5 sigmoidoscopies or colonos- copies per year were considered gastroenterologists. Scopes were identified by using OHIP fee code Z555, Z580, Z496A, Z497A, Z498A, and Z499A.

Variables
The outcomes of interest included IBD-specific and IBD- related rates of hospitalization, ED visits, outpatient physi- cian visits, and laboratory investigations. Only hospitalizations
of >24 hours in duration were included. Only a single outpa- tient visit per day was counted for ED, physician, and labo- ratory visits. IBD-specific visits were any visit associated with the International Classification of Disease, Ninth Edition (ICD- 9) diagnostic codes for CD (555.x) or UC (556.x) or ICD, Tenth Edition (ICD-10) diagnostic codes for CD (K50.x) or UC (K51.x). IBD-related visits included IBD-specific visits, and visits associated with diagnoses for signs, symptoms, and extraintestinal manifestations of IBD (Table I; available at www.jpeds.com).6,17 We also determined the risk of failure of transfer, defined as the proportion of patients without a single IBD-related outpatient visit to an adult gastroenterologist during the first year of the risk period.

Statistical Analyses
For descriptive statistics, we calculated medians with IQR or means with SD where appropriate. SCCS analyses were re- ported as relative incidence (RI) with 95% CI in the post- transfer risk period, with the pretransfer control period as the reference, adjusting for within-individual clustering. Pub- lished SAS macros (SAS Institute, Cary, North Carolina) were used to fit a Poisson regression model.18 All analyses were strati- fied by IBD type (CD, UC). Sensitivity analyses were per- formed to assess the risk of events in a 1-year risk period (instead of 2 years) compared with a 1-year control period. Additional sensitivity analysis varied the washout period. The

Figure 2. Study flow diagram. *IBD type could not be classified as CD or UC using validated administrative data algorithm based on the last seven outpatient visit diagnoses.13 **Lost to follow-up defined as the age of last IBD-related visit was beyond the 18th birthday, and had <2 adult gastroenterologist visits at the end of the observation period. ***Transfer period >365 days, and/or control period <2 years, and/or risk period <2 years. crude model only compared event rates without adjustment for transfer starting age. In a separate adjusted model, we in- cluded 3 transfer age categories as an independent variable: <18, 18-19, and ≥19. All analyses were conducted using SAS v9.1.3 (SAS Institute). Results Descriptive Characteristics The OCCC contained 4314 children <18 years of age diag- nosed with IBD in Ontario from fiscal years 1994 to 2008, of which 536 qualified for the study with sufficient pretransfer and post-transfer follow-up time by both pediatric and adult gastroenterologists (Figure 2). Of these, 1969 patients had ≥1 visit to both pediatric and adult gastroenterologists and 1357 patients had ≥2 visits to both adult and pediatric gastroenterologists. Of these, 602 patients were diagnosed with either CD or UC and were followed for ≥2 years before trans- fer. Sixty-six patients were lost to follow-up and could not be followed for 2 years in the risk period, resulting in 536 pa- tients included in this study. Descriptive characteristics of the cohort are presented in Table II. Of 536 children included in this study, 388 (72.4%) had CD and 148 (27.6%) had UC. In this cohort, 59% of the patients were boys. More patients came from families with higher socioeconomic status. Sixty-one patients (11.4%) ex- perienced transfer failure (no visits to an adult gastroenter- ologist in the first year of the risk period). Most patients (473 [88.2%]) had a transfer period of only 1 day in duration and only 8 patients (1.5%) had a transfer period longer than the washout period. All patients who died (<6) did so after the risk period. 4 Zhao et al Table II. Demographic characteristics of the study population adolescents beyond a specific age (in Ontario and most prov- inces, this age is 18 years). Optimally, adolescents with IBD are prepared for this transfer from pediatric to adult gastroenter- Characteristics Median age at diagnosis, y, IQR Latest diagnosis CD UC Sex Male Female Neighborhood income quintile at diagnosis 1 (lowest) 2 3 4 5 (highest) Median age at first adult gastroenterologist visit, y (IQR) Median age at start of risk period, y (IQR) n = 536 13.6 (2.9) 388 (72.4%) 148 (27.6%) 316 (59.0%) 220 (41.0%) 59 (11.0%) 86 (16.0%) 100 (18.7%) 126 (23.5%) 165 (30.8%) 18.1 (0.8) 18.6 (0.8) ology care well before their 18th birthday (during a transi- tion period). They are then transferred to an adult provider or care is shared between pediatric and adult providers for a brief period. Increased ED visit rates after transfer suggests either the disease severity worsened after transfer to adult care, neces- sitating emergency care, or patients had perceived or true dif- ficulty accessing urgent non-ED outpatient care. By contrast, the hospitalization rate did not change after transfer. Hospi- talization was more likely a marker of true worsening of disease severity and was less subject to patients’ distress or reduced access to care after transfer. Although we were not able to de- termine disease severity or the underlying reason for the visits (beyond diagnostic code), our findings indicate that most ED visits did not result in hospitalization and, therefore, could be preventable with proper planning for outpatient care. Health Services Use After Transfer The raw rates of health services use in risk compared with control periods are presented in Figure 3, with adjusted RI of rates in Table III. In the adjusted models, ED use rate was higher in the risk period after transfer for both CD (RI, 2.12; 95% CI, 1.53-2.93) and UC (RI, 2.34; 95% CI, 1.09-5.03). However, there was no significant increase in hospitalization rate (CD: RI, 0.70; 95% CI, 0.42-1.18; UC: RI, 2.41; 95% CI, 0.62- 9.40). Outpatient visits occurred more frequently in the risk period (CD: RI, 1.56; 95% CI, 1.42-1.72; UC: RI, 1.48; 95% CI, 1.24-1.76), as did laboratory investigations (CD: RI, 1.72; 95% CI, 1.64-1.81; UC: 1.69; 95% CI, 1.56-1.82). RIs for the risk period were larger in the adjusted model than the crude model, indicating the importance of the age at which trans- fer began. Younger age at transfer was associated with an in- creased outpatient visit rate and laboratory investigation rate in the risk period after transfer (in CD and UC) and in- creased hospitalization rates (in UC only; Table IV; available at www.jpeds.com). The sensitivity analyses (using a 1-year risk and control period, and varying the duration of the washout period) dem- onstrated similar RI of health services use to our main analy- sis with 2-year risk and control periods and a 6-month washout period (Table V; available at www.jpeds.com). The increased number of laboratory investigations re- quested after transfer may be due to adult gastroenterolo- gists ordering baseline measures for their new patients to adequately assess disease activity. Alternatively, it is plausible that patients were more likely clinically stable at the time of transfer and therefore had no recent bloodwork before trans- fer. In addition, the increased laboratory investigation rate may be a function of the current laboratory systems in place in Ontario, because the electronic health records rarely share labo- ratory data between pediatric and adult providers. The increase in outpatient visits after transfer was unex- pected; we expected that adult gastroenterologists would see patients with IBD less frequently than pediatric providers, as indicated in a previous study.10 The present study included pa- tients treated by adult gastroenterologists in private prac- tices, hospital-based practices, and academic health centers. There are differences in the organization and delivery of care between pediatric and adult gastroenterologists in Ontario. Almost all pediatric gastroenterologists in Ontario practice in academic centers. In addition, pediatric providers are usually on an alternate payment model involving salary, whereas adult gastroenterologists in Ontario are remunerated on a fee-for- service basis. Finally, pediatric centers were more likely to include specialized IBD multidisciplinary teams, unlike most community adult gastroenterology practices. Telephone or email access to this multidisciplinary team may have resulted in urgent Discussion Using a population-based cohort of patients with pediatric- onset IBD, we demonstrated increased healthcare use rates after transfer from pediatric to adult gastroenterology care. This in- crease was found for outpatient visits, laboratory investiga- tions, and ED visits, but not hospitalizations. These findings have important implications for health system planning, because they suggest that undesirable health services use (such as ED use for nonurgent problems) could be prevented with ad- equate transition planning. In Canada, pediatric gastroenter- ologists are typically prohibited or discouraged from following issues being addressed without the need for an outpatient or ED visit. Further work should assess and optimize various models of care of patients with IBD to reduce unnecessary health services use. Patients who were younger at the beginning of the trans- fer period had a greater relative increase in health services use. Younger patients may have required more support after the transfer and, therefore, were seen more frequently by their adult gastroenterologists. However, we cannot exclude that younger patients had greater severe disease activity, there- fore, requiring greater use of outpatient facilities and labora- tory investigations. Figure 3. Number of IBD-related health services events during the control and risk periods. A, Rates of outpatient physician visits and laboratory visits. B, Rates of ED visits and hospitalization. In recent years, research on the transition from pediatric to adult IBD care has expanded.19-24 This has resulted in mul- tiple guidelines from various professional societies.25-27 However, few studies have examined the impact of transition or struc- tured transition programs on health services use of children with chronic diseases. One population-based Ontario study using health administrative data examined 11 chronic dis- eases, including IBD.28 Investigators found ED visits rates in- creased for those with noncomplex chronic conditions including IBD. Although there was no increase in outpatient visits and overall costs decreased for patients with IBD after transfer, this study only assessed general health services and not IBD- related care. In addition, the authors included youth who were never treated by pediatric specialists and, therefore, were never transferred from pediatric to adult care. A single-center study of 95 patients with IBD from Hamilton, Ontario, revealed de- creased outpatient visits after transfer and no difference in ED visits, hospitalizations, surgical interventions, or endosco- pies10; however, adjustment for socioeconomic status, age at transfer, and pretransfer use was not performed. Our study ad- dressed these concerns with the multivariable SCCS analysis. Studies in other pediatric-onset chronic diseases have dem- onstrated significant concerns about increased ED visits after transfer from pediatric to adult care.29-33 A recent study of Ontario youth with IBD demonstrated that adult gastroen- terology practice type (academic or community) was not as- sociated with a change in the rate of ED visits after transfer.34 Interventions to ensure appropriateness of ED use could de- crease the cost and burden on patients and the health system and should be a focus of structured transition programs. The use of health administrative data to determine health- care patterns has limitations. Our databases are limited in that they contain no clinical information; therefore, disease sever- ity and extent or phenotype of IBD was not available, nor were medication records. Other potential confounders could not be measured, such as parental education, ethnic or cultural 6 Zhao et al Table III. RI of health services use in risk compared with control periods in patients with IBD diagnosed between 1994 and 2008 CD UC Models Control: mean number per year ± SD Risk: mean number per year ± SD RI 95% CI P Control: mean number per year ± SD Risk: mean number per year ± SD RI 95% CI P Hospitalizations Crude 0.32 ± 0.8 0.19 ± 0.62 0.59 0.44 0.79 .0004 0.11 ± 0.36 0.06 ± 0.29 0.56 0.26 1.20 .1361 Adjusted 0.70 0.42 1.18 .178 2.41 0.62 9.40 .2043 ED visits Crude 0.35 ± 0.81 0.70 ± 1.53 2.06 1.68 2.53 <.0001 0.24 ± 0.70 0.43 ± 1.04 1.75 1.16 2.64 .0074 Adjusted 2.12 1.53 2.93 <.0001 2.34 1.09 5.03 .0291 Outpatients visits Crude 8.19 ± 7.20 7.68 ± 6.11 0.94 0.90 0.99 .022 5.27 ± 4.7 5.62 ± 4.9 1.07 0.97 1.18 .1712 Adjusted 1.56 1.42 1.72 <.0001 1.48 1.24 1.76 <.0001 Laboratory investigations Crude 4.34 ± 6.18 5.01 ± 5.25 1.53 1.49 1.57 <.0001 3.94 ± 6.26 4.59 ± 5.66 1.43 1.37 1.49 <.0001 Adjusted 1.72 1.64 1.81 <.0001 1.69 1.56 1.82 <.0001 Crude models did not include age at time of transfer. Adjusted model included age at transfer by age category: <18, 18-19, and ≥19 years. background, and smoking status. However, because patients acted as their own controls in the SCCS analysis, these con- founders should be static before and after transfer. In addi- tion, although we can assess the diagnoses associated with each visit, we could not determine the reason that a patient visited an ED or whether the visit was appropriate. We could not measure rates of surgery or endoscopies for IBD because there were too few events to assess statistical differences. Misclassification bias is always a potential issue in research using health administrative data.35 However, we used algorithms vali- dated in pediatric patients in Ontario to identify children with IBD.13 Another limitation is that the minority of our population-based cohort of pediatric-onset IBD (n = 4314) had ≥2 visits to a pediatric gastroenterologist and 2 visits to an adult gastroenterologist in our data (n = 1357). Of these, a further minority had ≥2 years of data before and after the transfer period and were diagnosed with either CD or UC (n = 536). Therefore, the included cohort of patients may represent a select group of adolescents with IBD. For example, many adoles- cents diagnosed at >15 years of age may be diagnosed and treated by an adult gastroenterologist in Canada and may never require transition of care. However, this study focused on pa- tients who required transition and transfer after diagnosis, and were treated by both pediatric and adult gastroenterologists.
An assumption of SCCS analyses is that each health service event was independent. If the first outpatient visit led to the following one, this assumption would be violated, and may have resulted in overestimation of the increased rates after trans- fer. However, the SCCS method allowed us to compare rates before and after transfer without comparing patients who did or did not undergo the transfer process. Children treated by pediatric gastroenterologists are more likely to have severe disease or complex care requirements when compared with children treated by adult providers through their childhood. In addition, there is increased acknowledgement that chil- dren with pediatric-onset IBD have unique needs and risks compared with adult-onset IBD, which has resulted in an in- creased proportion of IBD care being provided by pediatric
gastroenterologists.6 Finally, the SCCS method controls for po- tential confounding factors that are fixed over time. However, any unmeasured confounders or patient characteristics that changed over time in association with transfer would not be controlled for in this analysis. For example, different medica- tion adherence rates in children compared with older adoles- cents may have altered our results.
In this population-based cohort study, we demonstrated sig- nificant changes in health services use after transfer from pe- diatric to adult care for patients with childhood-onset IBD. This finding was particularly prominent for ED use, outpatient visits, and laboratory investigations, and was consistent for both CD and UC. The biggest change after transfer was observed in ED use, where rates more than doubled after transfer. However, there was no significant change in hospitalization rates after transfer. Our findings will help healthcare providers and policymakers to engage in health system planning as the in- cidence of childhood-onset IBD has increased significantly in recent years. A concerted effort to evaluate models of transi- tion of care should include assessment of the impact of these models of health services use and health system resources. ■

We thank the following people for their contributions: Meltem Tuna (lead analyst at the Institute for Clinical Evaluative Sciences [ICES] uOttawa), Wenbin Li (former ICES analyst), and Pauline Quach (research coordinator).

Submitted for publication Apr 22, 2018; last revision received Jul 2, 2018; accepted Aug 9, 2018
Reprint requests: Eric I. Benchimol, MD, PhD, FRCPC, Division of Gastroenterology, Hepatology and Nutrition, Children’s Hospital of Eastern Ontario, 401 Smyth Rd, Ottawa, ON K1H 8L1, Canada. E-mail: ebenchimol@ cheo.on.ca

Data Statement

Parts of this material is based on data and information com- piled and provided by the Canadian Institute for Health In- formation (CIHI). However, the analyses, conclusions, opinions,

and statements expressed herein are those of the authors, and not necessarily those of CIHI. The dataset from this study is held securely in coded form at The Institute for Clinical Evalu- ative Sciences. Although data sharing agreements prohibit The Institute for Clinical Evaluative Sciences from making the dataset publicly available, access may be granted to those who meet prespecified criteria for confidential access, available at www.ices.on.ca/DAS.

References
1.Glick SR, Carvalho RS. Inflammatory bowel disease. Pediatr Rev 2011;32:14-24, quiz 5.
2.Ng SC, Shi HY, Hamidi N, Underwood FE, Tang W, Benchimol EI, et al. Worldwide incidence and prevalence of inflammatory bowel disease in the 21st century: a systematic review of population-based studies. Lancet 2017;390:2769-78.
3.Benchimol EI, Bernstein CN, Bitton A, Carroll MW, Singh H, Otley AR, et al. Trends in epidemiology of pediatric inflammatory bowel disease in Canada: distributed network analysis of multiple population-based pro- vincial health administrative databases. Am J Gastroenterol 2017;112:1120- 34.
4.Perminow G, Brackmann S, Lyckander LG, Franke A, Borthne A, Rydning A, et al. A characterization in childhood inflammatory bowel disease, a new population-based inception cohort from South-Eastern Norway, 2005- 07, showing increased incidence in Crohn’s disease. Scand J Gastroenterol 2009;44:446-56.
5.Benchimol EI, Manuel DG, Guttmann A, Nguyen GC, Mojaverian N, Quach P, et al. Changing age demographics of inflammatory bowel disease in Ontario, Canada: a population-based cohort study of epidemiology trends. Inflamm Bowel Dis 2014;20:1761-9.
6.Benchimol EI, Guttmann A, To T, Rabeneck L, Griffiths AM. Changes to surgical and hospitalization rates of pediatric inflammatory bowel disease in Ontario, Canada (1994-2007). Inflamm Bowel Dis 2011;17:2153-61.
7.Rosen DS, Blum RW, Britto M, Sawyer SM, Siegel DM. Transition to adult health care for adolescents and young adults with chronic conditions: po- sition paper of the Society for Adolescent Medicine. J Adolesc Health 2003;33:309-11.
8.Nakhla M, Daneman D, To T, Paradis G, Guttmann A. Transition to adult care for youths with diabetes mellitus: findings from a Universal Health Care System. Pediatrics 2009;124:e1134-41.
9.Samuel SM, Nettel-Aguirre A, Hemmelgarn BR, Tonelli MA, Soo A, Clark C, et al. Graft failure and adaptation period to adult healthcare centers in pediatric renal transplant patients. Transplantation 2011;91:1380-5.
10.Bollegala N, Brill H, Marshall JK. Resource utilization during pediatric to adult transfer of care in IBD. J Crohns Colitis 2013;7:e55-60.
11.Farrington CP, Nash J, Miller E. Case series analysis of adverse reactions to vaccines: a comparative evaluation. Am J Epidemiol 1996;143:1165- 73.
12.Benchimol EI, Hawken S, Kwong JC, Wilson K. Safety and utilization of influenza immunization in children with inflammatory bowel disease. Pe- diatrics 2013;131:e1811-20.
13.Benchimol EI, Guttmann A, Griffiths AM, Rabeneck L, Mack DR, Brill H, et al. Increasing incidence of paediatric inflammatory bowel disease in Ontario, Canada: evidence from health administrative data. Gut 2009;58:1490-7.
14.Institute for Clinical Evaluative Sciences (ICES). ICES data dictionary. Toronto: Institute for Clinical Evaluative Sciences; 2018.
15.Benchimol EI, Guttmann A, Mack DR, Nguyen GC, Marshall JK, Gregor JC, et al. Validation of international algorithms to identify adults with inflammatory bowel disease in health administrative data from Ontario, Canada. J Clin Epidemiol 2014;67:887-96.
16.Canadian Institute for Health Information. CIHI data quality study of the 2009-2010 discharge abstract database. Ottawa: Canadian Institute for Health Information, Standards and Data Submission; 2012.

17.Benchimol EI, Mack DR, Nguyen GC, Snapper SB, Li W, Mojaverian N, et al. Incidence, outcomes, and health services burden of very early onset inflammatory bowel disease. Gastroenterology 2014;147:803-13.e7, quiz e14-e5.
18.Whitaker H, Spiessens B. The self controlled case series method. The Open University; 2015.
19.Carlsen K, Haddad N, Gordon J, Phan BL, Pittman N, Benkov K, et al. Self-efficacy and resilience are useful predictors of transition readiness scores in adolescents with inflammatory bowel diseases. Inflamm Bowel Dis 2017;23:341-6.
20.Fishman LN, Mitchell PD, Lakin PR, Masciarelli L, Flier SN. Are expec- tations too high for transitioning adolescents with inflammatory bowel disease? Examining adult medication knowledge and self-management skills. J Pediatr Gastroenterol Nutr 2016;63:494-9.
21.Gumidyala AP, Plevinsky JM, Poulopoulos N, Kahn SA, Walkiewicz D, Greenley RN. What teens do not know can hurt them: an assessment of disease knowledge in adolescents and young adults with IBD. Inflamm Bowel Dis 2017;23:89-96.
22.Jeganathan J, Lee CH, Rahme A, Tiao DK, Weston C, Dutt S, et al. Pediatric- to-adult transition and medication adherence in patients with inflam- matory bowel disease. Inflamm Bowel Dis 2017;23:1065-70.
23.Yerushalmy-Feler A, Ron Y, Barnea E, Nachum A, Matalon S, Dali-Levy M, et al. Adolescent transition clinic in inflammatory bowel disease: quan- titative assessment of self-efficacy skills. Eur J Gastroenterol Hepatol 2017;29:831-7.
24.Benchimol EI, Walters TD, Kaufman M, Frost K, Fiedler K, Chinea Z, et al. Assessment of knowledge in adolescents with inflammatory bowel disease using a novel transition tool. Inflamm Bowel Dis 2011;17:1131-7.
25.Baldassano R, Ferry G, Griffiths A, Mack D, Markowitz J, Winter H. Tran- sition of the patient with inflammatory bowel disease from pediatric to adult care: recommendations of the North American Society for Pediat- ric Gastroenterology, Hepatology and Nutrition. J Pediatr Gastroenterol Nutr 2002;34:245-8.
26.Brooks AJ, Smith PJ, Cohen R, Collins P, Douds A, Forbes V, et al. UK guideline on transition of adolescent and young persons with chronic di- gestive diseases from paediatric to adult care. Gut 2017;66:988-1000.
27.Elli L, Maieron R, Martelossi S, Guariso G, Buscarini E, Conte D, et al. Transition of gastroenterological patients from paediatric to adult care: a position statement by the Italian Societies of Gastroenterology. Dig Liver Dis 2015;47:734-40.
28.Cohen E, Gandhi S, Toulany A, Moore C, Fu L, Orkin J, et al. Health care use during transfer to adult care among youth with chronic conditions. Pediatrics 2016;137:e20152734.
29.Gurvitz MZ, Inkelas M, Lee M, Stout K, Escarce J, Chang RK. Changes in hospitalization patterns among patients with congenital heart disease during the transition from adolescence to adulthood. J Am Coll Cardiol 2007;49:875-82.
30.Blinder MA, Duh MS, Sasane M, Trahey A, Paley C, Vekeman F. Age- related emergency department reliance in patients with sickle cell disease. J Emerg Med 2015;49:513-22.e1.
31.Lee JM, Sundaram V, Sanders L, Chamberlain L, Wise P. Health care uti- lization and costs of publicly-insured children with diabetes in Califor- nia. J Pediatr 2015;167:449-54.e6.
32.Mann JR, Royer JA, Turk MA, McDermott S, Holland MM, Ozturk OD, et al. Inpatient and emergency room visits for adolescents and young adults with spina bifida living in South Carolina. PM R 2015;7:499-511.
33.Son MB, Sergeyenko Y, Guan H, Costenbader KH. Disease activity and transition outcomes in a childhood-onset systemic lupus erythemato- sus cohort. Lupus 2016;25:1431-9.
34.Bollegala N, Benchimol EI, Griffiths AM, Kovacs A, Steinhart AH, Zhao X, et al. Characterizing the posttransfer period among patients with pe- diatric onset IBD: the impact of academic versus community adult care on emergent health resource utilization. Inflamm Bowel Dis 2017;23:1483- 91.
35.Benchimol EI, Smeeth L, Guttmann A, Harron K, Moher D, Petersen I, et al. The REporting of studies Conducted using Observational Routinely- collected health Data (RECORD) statement. PLoS Med 2015;12:e1001885.

8 Zhao et al

Funding and Disclosures

E.B. and G.N. received support from New Investigator Awards from the Canadian Institutes of Health Research, Crohn’s and Colitis Canada, and the Canadian Association of Gastroen- terology. E.B. also received support from the Career Develop- ment Award and the Career Enhancement Program from the Canadian Child Health Clinician Scientist Program. The authors declare no conflicts of interest
This work was made possible with the support of the In- stitute for Clinical Evaluative Sciences which receives funding from the Ontario Ministry of Health and Long-Term Care (MOHLTC). The results and conclusions are those of the authors; no official endorsement by the Ontario MOHLTC should be inferred.

Table I. Classification of IBD-related health services use Table I. Continued

Conditions IBD specific
CD
UC
IBD signs/symptoms Anorexia
Abnormal weight gain Abnormal weight loss Underweight
Failure to thrive, child Failure to thrive, adult
Symptoms involving digestive
system, including (787.0) Nausea and vomiting (787.01) Nausea w/ vomiting (787.02) Nausea, alone (787.03) Vomiting, alone (787.1) Heartburn
(787.2) Dysphagia (787.3) Gas/bloating (787.6) Encopresis, fecal
incontinence
(787.9) Other symptoms involving
digestive system (787.91) Diarrhea, NOS
Abdominal pain Dyspepsia Cachexia Esophagitis

GERD
Gastric ulcer Duodenal ulcer Peptic ulcer
GJ ulcer Gastritis/duodenitis Intestinal obstruction

ICD-9

555.x
556.x 783.0
783.1
783.2 783.22 783.41

783.7 787.x

789.0
536.8
799.4
530.1 530.81
531
532
533
534 535.x 560.8
OHIP diagnostic code

555
556

787

787

787
536

530

531
532

534
535
560

ICD-10

K50.x
K51.x

R63.0
R63.5
R63.4
R62.8
R62.8
R62.9
R62.8
R62.9
R11.x
R12.x
R13.x
R14.x
R15.x
R19.x

R10.x
K30.x
R64.x
K20.x
K22.1
K21.x
K25.x
K26.x
K27.x
K28.x
K29.x
K31.5

Conditions Malabsorption

Extraintestinal manifestations Anal fistula
Anal abscess Ureteral fistula Urethral fistula
Fistula of stomach and
duodenum Vesical fistula
Fistula involving female GU Hemorrhoids, including
(455.9) Anal skin tags Rheumatoid arthritis

Arthropathy associated with
erythema nodosum Inflammatory spondylopathies,
including
(720.0) Ankylosing spondylitis (720.1) Spinal enthesopathy (720.2) Sacroiliitis
(720.8) Other inflammatory (720.9) Other unspecified
inflammatory Scleritis and episcleritis Unspecified iridocyclitis
(uveitis NOS) Chorioretinitis, unspecified
(uveitis, posterior NOS)

ICD-9 262.x
263.0
263.1
263.2
263.8
263.9
579.8
579.9

565.1 593.82
599.1
537.4

596.2 619.x 455.x

714.x

713.3 720.x

379.0
364.3

363.2
OHIP diagnostic code
579

565
566

455

714

720

379
364

363

ICD-10 E43.x
E44.x
E45.x
E46.x
K90.8
K90.9

K60.3

N28.8
N36.0
K31.6

N32.2
N82.x
I84.x M06.99
M05.3
M08.x
M09.x
M14.8

M45.x
M46.x
M47.x
M48.x

H15.x
H20.9

H30.9

560.9 Rectal/anal hemorrhage 569.3
Other disorder of 569.4
rectum/anus, including: (569.41) Ulcer
(569.42) Pain
(569.43) Sphincter tear (healed) (569.44) Dysplasia
(569.45) Other specified, including
proctitis, inflammation
Abscess of the intestine 569.5
Other disorders of intestine, 569.8
including:
(569.81) Fistula (excl rectum) (569.82) Ulcer of intestine (569.83) Perforation
(569.84) Angiodysplasia, no
hemorrhage
(569.85) Angiodysplasia, with
hemorrhage (569.86) Dieulafoy (569.89) Other, including

569
569

569
569
K56.6
K62.5
K62.6
K62.8

K63.0
K63.2
K63.3
K63.1
K55.2
K63.8
Acute and subacute iridocyclitis 364.0
Erythema nodosum 695.2
Acrodermatitis enteropathica 686.8
Pyoderma 686.0
Pyogenic granuloma of the skin 686.1
and soft tissue
Oral aphthae 528.2
Short stature 783.43
Osteoporosis 733.0

Osteomyelitis 730.0 730.1 730.2
Acute glomerulonephritis 580.x

Nephrolithiasis 592.x
Primary sclerosing cholangitis 576.1
Venous embolism/thrombosis 453.x
364
695

686
686

528

733

730

580

592
576
H20.0
L52.x
E83.2
L08.0
L98.0

K12.0
E34.3
M80.x
M81.x
M82.x
M83.x
M86.x

N00.x
N01.x
N05.x
N20.x
K83.0
I82.x

– Enteroptosis
– Granuloma of intestine
– Prolapse of intestine
– Pericolitis
– Perisigmoiditis
– Visceroptosis
GERD, gastroesophageal reflux disease; GU, genitourinary; NOS, not otherwise specified. Note that description of IBD-related diagnoses in the text refer to all diagnostic codes classi- fied as IBD-specific, IBD signs/symptoms and extraintestinal manifestations.

(continued)

8.e2 Zhao et al

Table IV. RI of health services use in risk compared with control periods in patients with IBD diagnosed between 1994 and 2008, stratified by age at transfer to adult care

Transfer age categories, y

RI
CD 95% CI

P value

RI
UC 95% CI

P value

Hospitalizations
<18 1.18 0.65 2.14 .5901 6.30 1.35 29.41 .0192 18-19 0.56 0.33 0.94 .0297 0.98 0.24 3.97 .9804 ≥19 ED visits Reference . Reference . <18 1.04 0.71 1.54 .8323 1.40 0.59 3.32 .4441 18-19 0.98 0.74 1.30 .9107 0.64 0.33 1.24 .1821 ≥19 Outpatients visits Reference . Reference . <18 1.91 1.71 2.13 <.0001 1.56 1.27 1.92 <.0001 18-19 1.15 1.06 1.25 .0005 1.18 1.01 1.37 .0322 ≥19 Laboratory investigations Reference . Reference . <18 1.39 1.20 1.61 <.0001 1.27 1.01 1.59 .0449 18-19 1.22 1.10 1.35 .0001 0.77 0.64 0.92 .0044 ≥19 Reference . Reference . Table V. Sensitivity analyses: RI of health service use in risk vs control period with varying wash-out periods Health service Model RI CD 95% CI P value RI UC 95% CI P value 2-Year risk/control periods Hospitalization (no washout) Adjusted 1.308 0.861 1.987 .208 4.307 1.654 11.217 .0028 Hospitalization (3 months washout) Adjusted 0.904 0.566 1.445 .6732 1.320 0.363 4.801 .6738 Hospitalization (6 months washout, original) Adjusted 0.701 0.418 1.175 .178 2.413 0.619 9.403 .2043 Hospitalization (9 months washout) Adjusted 0.829 0.488 1.408 .4875 1.646 0.383 7.077 .5032 Hospitalization (12 months washout) Adjusted 0.796 0.462 1.371 .4103 2.889 0.651 12.821 .163 ED visits (no washout) Adjusted 2.585 1.934 3.456 <.0001 1.992 1.005 3.949 .05 ED visits (3 months washout) Adjusted 2.162 1.590 2.940 <.0001 1.720 0.824 3.590 .15 ED visits (6 months washout, original) Adjusted 2.120 1.534 2.930 <.0001 2.341 1.090 5.027 .0291 ED visits (9 months washout) Adjusted 2.087 1.493 2.919 <.0001 2.000 0.918 4.357 .08 ED visits (12 months washout) Adjusted 1.920 1.359 2.714 <.0001 2.538 1.158 5.562 .02 Hospitalization/ED visits (no washout) Adjusted 1.830 1.540 2.175 <.0001 2.003 1.317 3.047 <.0001 Hospitalization/ED visits (3 months washout) Adjusted 1.653 1.286 2.124 <.0001 1.733 0.923 3.255 .09 Hospitalization/ED visits (6 months washout, original) Adjusted 1.537 1.178 2.006 .0016 2.556 1.316 4.963 .0056 Hospitalization/ED visits (9 months washout) Adjusted 1.503 1.200 2.088 .0012 2.108 1.066 4.171 .0321 Hospitalization/ED visits (12 months washout) Adjusted 1.468 1.103 1.953 .0086 2.893 1.450 5.770 .0026 Outpatients visits (no washout) Adjusted 2.062 1.901 2.236 <.0001 1.991 1.710 2.317 <.0001 Outpatients visits (3 months washout) Adjusted 1.529 1.395 1.676 <.0001 1.420 1.202 1.679 <.0001 Outpatients visits (6 months washout, original) Adjusted 1.562 1.417 1.723 <.0001 1.478 1.239 1.763 <.0001 Outpatients visits (9 months washout) Adjusted 1.592 1.436 1.765 <.0001 1.505 1.250 1.812 <.0001 Outpatients visits (12 months washout) Adjusted 1.577 1.417 1.756 <.0001 1.690 1.398 2.043 <.0001 Laboratory investigations (no washout) Adjusted 1.908 1.827 1.993 <.0001 1.872 1.746 2.008 <.0001 Laboratory investigations (3 months washout) Adjusted 1.661 1.585 1.741 <.0001 1.469 1.642 1.905 <.0001 Laboratory investigations (6 months washout, original) Adjusted 1.718 1.635 1.806 <.0001 1.686 1.560 1.822 <.0001 Laboratory investigations (9 months washout) Adjusted 1.796 1.704 1.892 <.0001 1.761 1.624 1.910 <.0001 RI-1
Laboratory investigations (12 months washout)
1-Year risk/control periods
Adjusted 1.587 1.504 1.675 <.0001 1.649 1.518 1.790 <.0001 Hospitalization (no washout) Adjusted 1.338 0.842 2.126 .22 2.706 1.011 7.315 .0497 Hospitalization (3 months washout) Adjusted 0.836 0.494 1.415 .50 0.827 0.194 3.520 .797 Hospitalization (6 months washout, original) Adjusted 0.611 0.336 1.113 .1072 1.035 0.189 5.678 .9686 Hospitalization (9 months washout) Adjusted 0.714 0.385 1.325 .2854 0.247 0.023 2.697 .2517 Hospitalization (12 months washout) Adjusted 0.675 0.356 1.279 .2279 0.660 0.079 5.490 .7004 ED visits (no washout) Adjusted 2.547 1.827 3.552 <.0001 1.577 0.750 3.315 .2295 ED visits (3 months washout) Adjusted 2.179 1.522 3.119 <.0001 1.465 0.633 3.388 .3721 ED visits (6 months washout, original) Adjusted 2.021 1.378 2.964 .0003 1.704 0.673 4.314 .2611 ED visits (9 months washout) Adjusted 1.999 1.343 2.975 .0006 0.907 0.332 2.483 .8498 ED visits (12 months washout) Adjusted 1.630 1.079 2.463 .0202 1.377 0.520 3.647 .52 Hospitalization/ED visits (no washout) Adjusted 2.038 1.561 2.660 <.0001 1.949 1.074 3.534 .0281 Hospitalization/ED visits (3 months washout) Adjusted 1.598 1.196 2.135 .0015 1.334 0.654 2.720 .4288 Hospitalization/ED visits (6 months washout, original) Adjusted 1.418 1.037 1.940 .0288 1.642 0.735 3.669 .2263 Hospitalization/ED visits (9 months washout) Adjusted 1.475 1.063 2.046 .0201 0.826 0.337 2.024 .6763 Hospitalization/ED visits (12 months washout) Adjusted 1.260 0.897 1.770 .1829 1.376 0.586 3.232 .4636 Outpatients visits (no washout) Adjusted 2.124 1.934 2.334 <.0001 2.335 1.953 2.791 <.0001 Outpatients visits (3 months washout) Adjusted 1.453 1.304 1.619 <.0001 1.487 1.217 1.819 .0001 Outpatients visits (6 months washout, original) Adjusted 1.533 1.367 1.720 <.0001 1.499 1.209 1.857 .0002 Outpatients visits (9 months washout) Adjusted 1.606 1.431 1.825 <.0001 1.447 1.153 1.815 .001 Outpatients visits (12 months washout) Adjusted 1.557 1.380 1.779 <.0001 1.636 1.299 2.060 <.0001 Laboratory investigations (no washout) Adjusted 1.878 1.783 1.978 <.0001 1.809 1.662 1.968 <.0001 Laboratory investigations (3 months washout) Adjusted 1.497 1.416 1.583 <.0001 1.717 1.570 1.877 <.0001 Laboratory investigations (6 months washout, original) Adjusted 1.682 1.535 1.727 <.0001 1.705 1.551 1.874 <.0001 Laboratory investigations (9 months washout) Adjusted 1.901 1.787 2.021 <.0001 1.963 1.779 2.167 <.0001 Laboratory investigations (12 months washout) Adjusted 1.713 1.606 1.826 <.0001 1.805 1.634 1.994 <.0001 Adjusted model included age at transfer by age category: <18, 18-19, and ≥19 years. Bolded values represent those that are statistically significant (P <.05). 8.e4 Zhao et al