Ulcerative colitis-associated hospitalization costs : a population-based study

1Department of Medicine, 2Community Health Sciences, and 4Surgery, University of Calgary, Calgary, Alberta; 3Alberta Health Services, Calgary, Alberta Correspondence: Dr Gilaad G Kaplan, Departments of Medicine and Community Health Sciences, University of Calgary, 3280 Hospital Drive Northwest, 6D56, Calgary, Alberta T2N 4N1. Telephone 403-592-5015, fax 403-592-5090, e-mail ggkaplan@ucalgary.ca Received for publication February 1, 2015. Accepted March 7, 2015 Ulcerative colitis (UC) is primarily diagnosed in late adolescence to early adulthood. The burden of UC is lifelong, and it afflicts individuals physically, mentally and financially (1). In North America, 0.25% of the population has UC and the incidence of UC appears to be increasing in many parts of the world (2). Despite advances in medical management, 15% of UC patients will undergo a total abdominal colectomy within 10 years of diagnosis (3). Colectomy is associated with considerable postoperative morbidity and impairs long-term quality of life (4-8). Overall, colectomy rates have decreased over time (3); however, while elective colectomy rates have steadily decreased, the rates of emergent colectomies have remained stable (9). These findings suggest that the health and economic burden of UC continues to be high. The direct health care costs of inflammatory bowel disease (IBD) in the United States exceed USD$6 billion annually (10). The average direct cost of UC has been estimated to be >$3,500 per patient, with a large portion of these costs attributed to hospitalizations (11,12). IBD hospitalizations account for nearly $395 million in health care spending in Canada, and are predicted to increase (11,12). Drivers of inpatient UC costs are multifactorial (13), but include infliximab, which was shown in 2005 to reduce the risk of colectomy among UC patients who failed to respond to intravenous S Coward, SJ Heitman, F Clement, et al. Ulcerative colitisassociated hospitalization costs: A population-based study. Can J Gastroenterol Hepatol 2015;29(7):357-362.


RÉSULTATS :
Accordingly, we conducted a population-based study to identify the primary drivers of in-hospital cost for UC and to assess whether these factors changed following the introduction of infliximab.

METHODS
The present study was a population-based costing analysis of medical and surgical hospitalization admissions for adults (18 years of age) with UC in the Calgary Health Zone (CHZ) from January 1, 2001 to December 31, 2009.

Data sources
The Discharge Abstract Database used by Alberta Health Services (AHS) captures all diagnostic and procedural codes that occur during a hospital admission.AHS' Data Integration, Measurement and Reporting department (DIMR) identified patients coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and the Tenth Revision (ICD-10-CM).All patients who were admitted with UC (ICD-9-CA code 556.9;ICD-10-CM code K51) from January 1, 2001 to December 31, 2009 in the CHZ were captured; this method of identifying admissions has been previously validated (15).The CHZ contains the City of Calgary and >20 surrounding villages, towns, smaller cities and hamlets, and had a population of 1,408,606 in 2011.AHS is a single-payer, publically funded health care system that is responsible for all medical and surgical care within the CHZ.
The total cost for each individual admission was attained from Activity and Costing -AHS Finance Department and through the Physician Claims Database.This database includes claims submitted for payment by Alberta physicians for services provided to registrants of the Alberta Health Care Insurance Plan, a universal plan that covers >99% of Alberta residents (16).The total costs associated with each hospital admission account for both direct and indirect costs.Direct costs are all costs associated with direct patient care.These include -but are not limited to -nonphysician salaries, drugs, equipment depreciation cost, and the allocated costs for the nursing units and supporting care areas (ie, diagnostic imaging, cardiac labs, ambulatory services).Indirect costs are expenses that account for hospital overhead (eg, administration, support services, site utilities, human resources, information technology services, etc).Virtually all (99.7%) of the admissions had matched finance data and were included in the analysis (the remaining 0.3% were excluded).

Study population
The DIMR database identified 1062 UC patients who had either a colectomy or flare admission.If patients had >1 UC-related admission, a single admission was randomly selected for analysis; this method of analysis has previously been validated (15).After chart review of the 1062 admissions, a total of 318 patients were excluded from the study because the charts were unavailable (n=85), they had Crohn disease (n=54), they did not have IBD (n=42), they underwent a previous colectomy (n=72) or UC was not the primary reason for admission (n=65).The remaining 744 admissions were then stratified into one of three admission categories: responsive to inpatient medical therapy (n=309), emergent colectomy during admission (n=227) and elective colectomy during admission (n=208).These admissions were then submitted to the AHS Finance Department for retrieval of costing data.Two additional admissions were excluded due to a lack of matching cost data, resulting in a study population of 742 patients.

Data extraction
Each patient underwent a comprehensive chart review to capture the following data: age at admission; residence; admit and discharge dates; duration of flare; date of UC diagnosis; disease extent; sex; smoking; UC-validated comorbidities (17,18); UC medications including corticosteroids, 5-aminosalicylic acid, azathioprine, infliximab, adalimumab, budesonide and methotrexate; disease activity (stool frequency, blood in stool and hemoglobin level at admission); in-hospital complications both medical and surgical (assigned according to Clavien class [19]); and length of stay.
The data attained through the chart review were used to stratify patients into the three categories: responsive to medical management (flare); elective colectomy; or emergent colectomy.Patients were deemed responsive to inpatient medical therapy if they came to hospital with a UC-related flare and were discharged without surgery; during analysis, this was the referent group.Emergent colectomy was defined as a UC flare that required an unplanned colectomy.Elective colectomy was defined as admission for a scheduled colectomy.

Statistical analysis
Due to the skewed nature of the cost data, the annual median costs, with interquartile ranges for each of the admission categories (ie, responsive to inpatient medical management, elective colectomy, and emergent colectomy) were calculated.All costs were adjusted to 2013 Canadian dollars using the Consumer Price Index (20).
Linear regression was used to determine the effect that the different admission types had on the annual trend in hospital costs incurred for the admissions.Admission type was modelled as a categorical variable with 'responsive to inpatient medical management' as the referent level.The hospital costs were logarithmically transformed because of their skewed distribution and adjusted for the following covariates: year of admission as continuous variable; age stratified according to tercile (18 to 31, 32 to 47 and ≥48); sex; residence as defined as residing within or outside the CHZ; smoking as current, ex-smoker, never or unknown; comorbidities as 0 or ≥1; disease extent as pancolitis versus left-sided/undetermined; flare duration (<2 weeks, two to eight weeks, >8 weeks or unknown); disease severity as presence of blood in stool or stool frequency (>5 or ≤5 days), and hemoglobin level at admission (>100 g/L versus ≤100 g/L); in-hospital complications; UC medications at admission (5-aminosalicylic acid, prednisone, azathioprine,) and/or during hospital (infliximab); and length of stay.All of these clinically relevant covariates were a priori included in the adjusted analysis.
Using linear regression, the annual cost increase was calculated for each UC admission type with the regression model including the adjustment for confounders.The beta coefficients were exponentiated to give the final annual percentage increase and their corresponding 95% CIs.A Joinpoint analysis assessed for significant inflection points for both the mean and median cost of the aggregate data over the study period.The a priori analysis assessed for a significant cost inflection point after 2005 when infliximab was introduced (14).The statistical analysis was performed using SAS version 9.3 (SAS Institute, USA).The Joinpoint analysis was performed with Joinpoint Regression Program version 4.1.0(21).

RESULTS
Patient characteristics stratified according to type of admission are presented in Table 1.The median costs for each of the three admission types were: UC flare $5,499 (interquartile range [IQR] $3,374 to $8,904), elective colectomy $14,316 (IQR $11,932 to $18,331) and emergent colectomy $23,698 (IQR $17,981 to $32,385).The inflationadjusted cost of all admissions increased by 6.0% (95% CI 4.5% to 7.5%) per year (Table 2).The median costs stratified according to admission type per year are presented in Figure 1.Significant Joinpoints were not identified in the median or mean costs.
UC patients who underwent an elective colectomy had a 179.8% (95% CI 151.6% to 211.1%) increase in cost versus those who were medically responsive to in-patient medical management after adjusting for covariates.Those who underwent emergent colectomies were significantly higher with a 211.1% (95% CI 183.2% to 241.6%) increase in cost (Table 2).

DISCUSSION
It has previously been shown that during our study period, the rates of colectomy fell in the CHZ, particularly the rate of elective colectomies (9).Despite the reduction in colectomies, in-hospital costs for all UC patients increased by 6.0% per year.After adjusting for disease severity, surgical admissions were associated with the highest magnitude of increased hospital costs.Emergent operations are associated with a greater length of stay and interventions performed in hospital due to higher occurrence of in-hospital complications and longer recovery periods (5,7).Other drivers of cost were age, disease severity, in-hospital complications, length of stay, smoking status and in-hospital prescription of infliximab.By identifying some of the factors that contributed to the rising cost of UC care, we may consider alternative ways to mitigate these costs while providing appropriate patient care.The present study evaluated the temporal trends of UC hospitalization costs in Canada since the introduction of infliximab in 2005 for the treatment of hospitalized UC patients refractory to intravenous corticosteroid therapy (14).Since 2005, infliximab has become a mainstay inhospital rescue agent for UC patients.A previous study demonstrated a sharp uptake of in-hospital prescriptions of infliximab in the CHZ (9).In the current study, infliximab was an independent risk factor for increasing hospital costs after adjusting for patient factors, disease severity and surgery.However, an expected inflection point of rising hospital costs after 2005 was not observed, which suggests that multiple factors contributed to increased hospital costs and/or money spent on infliximab may have been offset by alternative cost saving.
Moreover, infliximab is a modifiable cost factor that could prove to be a significant cost savings to hospital budgets.In Alberta, in-hospital use of infliximab is paid for by the hospital budget, whereas outpatient infliximab use is covered by provincial drug plans or by private drug coverage.Thus, timelier introduction of infliximab in the outpatient setting would avoid this expense in hospital and, additionally, may reduce the need for hospitalizations or lead to an elective colectomy.Among UC patients with fulminant colitis that requires infliximab in hospital, an alternative method of funding anti-tumour necrosis factor therapy that does not impact patient care should be implemented.Policy changes could lead to shifting in-hospital infliximab costs to drug insurance plans, or existing biologic infusion centres could be expanded to include in-hospital care where additional efficiencies may be generated.
In Canada, from 2000 to 2009, annual health care inflation rate rose by an average of approximately 1.6% per year (20).Costs for inhospital admissions for UC increased significantly faster than the  inflation rate.This is due, in part, to the overall rise in health care costs that are outpacing the national inflation rate.Over time, hospitals have been admitting older and sicker patients with a greater number of chronic comorbidities, which has led to an increase in-hospital expenditures (22).In addition, medical technology and the labour force have been found to be significant cost drivers in the United States (23).In Alberta, the salary of unionized health care work force increased by 2% to 3% per year during our study period (24).Also, patients are staying longer in emergency departments due to bed shortages in hospital, and delays exceeding 12 h have been shown to increase cost by 11% (25).These factors may have played a part in the increasing expenditures.
A few limitations of our study should be considered (26).While we were able to explore both direct and indirect in-hospital costs, we were unable to differentiate whether the direct or indirect costs were the major cost drivers.In addition, we did not account for outpatient costs that patients incurred to the health care system.However, the purpose of the present study was to evaluate the evolution of in-hospital costs independent of the effect of outpatient management.By focusing on in-hospital costs, our study informs health care resource allocation planning for inpatient care, but is not generalizable to outpatient management.Because we undertook a retrospective chart review, some clinical factors, such as outpatient medication use, were missing in some patients.While the linkage to the AHS Finance Department that reported both direct and indirect hospital costs was 99.7% complete, indirect outpatient costs (eg, loss of work productivity) were not assessed in our study.Finally, administrative data were used to initially identify UC patients.While administrative data are subject to misclassification errors, we have previously validated this approach (15).

Figure 1 )
Figure 1) Temporal trends from 2000 to 2009 of median costs stratified according to each admission type: medically responsive flare, elective colectomy and emergent colectomy

TAble 1 Characteristics of the study population
*Data presented as % (n) unless otherwise indicated.*Fisher's exact test, unless otherwise indicated; † Row mean scores test; ‡ Defined as missing data regarding stool frequency, the presence of blood in stool or both; § Defined as medication being taken at the time of admission to hospital or in the past; no refers to no record of drug use in the medical chart; ¶ Defined as medication prescribed during the hospital stay; no refers to no record of drug use in the medical chart; **Kruskal-Wallis test to 43.8%]); disease extent (pancolitis versus left-sided: 11.4% [95% CI 3.8% to 19.6%]); smoking (current versus never: −11.6% [95% CI −21.4% to −0.5%]); infliximab prescribed in hospital (69.5% [95% CI 49.2% to 92.5%]); and length of stay (1.7% per day [95% CI 1.5% to 1.9%]).