Comorbidity Before and After a Diagnosis of Inflammatory Bowel Disease

Charles N. Bernstein; Zoann Nugent; Seth Shaffer; Harminder Singh; Ruth Ann Marrie


Aliment Pharmacol Ther. 2021;54(5):637-651. 

In This Article


Data Source & Population

The University of Manitoba IBD Epidemiology Database includes all persons with IBD in the province of Manitoba (population 1.35 million, 2018) from April 1984 through 31 March 2018.[14] All residents of Manitoba have a unique personal health identification number through which all health system contacts can be tracked including all outpatient physician contacts and hospitalisations. These data are available from 1 April 1984 onwards. The Discharge Abstract Database captures administrative, clinical and demographic information on hospital discharges (including deaths, discharges and transfers). Up to 25 diagnoses are recorded using the International Classification of Disease (ICD)-9-clinical modification (CM) until 2004, following which they have been recorded using ICD-10-CA. Physician claims capture one diagnosis recorded using ICD-9-CM codes, date of service and type of service delivered.

Persons were identified as having IBD, either Crohn's disease (CD) or ulcerative colitis (UC) using a validated administrative definition.[14] This definition includes all persons with at least five health system contacts (either unique outpatient visits or hospitalisations) using ICD-9 CM code 555 or ICD-10 code K50 for CD and ICD-9-CM code 556 or ICD-10 code K51 for UC. Cases of IBD with diagnosis dates between 1 April 1987 and 31 March 2018 were extracted from the University of Manitoba IBD Epidemiology Database. This allowed a lookback period of 33 months of hospital and physician records for the earliest IBD cases. The index date for all cases of IBD was the date of the first diagnostic contact for IBD. For each IBD case, we identified up to 10 controls without IBD matched on year of birth, sex and geographic residence by postal forward sortation area. Socioeconomic status was measured with the Socioeconomic Factor Index.[15] The Socioeconomic Factor Index is a factor score derived from Canadian Census data that reflects non-medical social determinants of health and is used as a proxy measure of socioeconomic status. When interpreting the Socioeconomic Factor Index values, scores less than zero indicate more favourable socioeconomic conditions, while scores greater than zero indicate less ideal socioeconomic conditions.

Data Source & Population

The analyses include 9247 IBD cases (4253 CD, 4994 UC) and 85 691 controls matched up to 1:10 by age, sex and geographic residence by postal forward sortation area. Socioeconomic status was measured with the Socioeconomic Factor Index.[15]

Charlson Comorbidity Index

We searched the Discharge Abstract Database and physician claims for diagnostic codes for the CCI, including all 17 diagnoses except AIDS. AIDS was excluded from this analysis since it was too rarely reported in persons with IBD to meet minimum cell size reporting requirements in Manitoba (cell sizes must be greater than 5 to be reported). Myocardial infarctions and congestive heart failure were combined to increase the numbers of positives available (Cardiac). The two subdivisions of each of liver disease, diabetes and cancer were also combined into one category each. Twelve conditions remained for consideration.

We identified CCI conditions in hospital and physician claims using ICD-9-CM codes or ICD-10-CA codes (see Appendix 1). To assess the CCI, data from the year before diagnosis in question are used, however, we used all available data to assess for these comorbidities. Cases and controls had the same period of follow-up prior to IBD diagnosis and post-diagnosis. First, we report the prevalence of these diagnoses before IBD diagnosis (or the diagnosis date of the associated case for controls), and compare the prevalence between groups using relative rate (RR). CIs were adjusted for Bonferroni correction. Second, we report the incidence of these diagnoses after IBD. For these analyses, we excluded any IBD case or control who had already developed the condition before diagnosis and excluded all matched controls of excluded cases.

We compared time to diagnosis of each condition after IBD diagnosis using Cox proportional hazards regression, and report hazard ratios (HR). CIs were adjusted for Bonferroni correction. For example, for 12 comparisons, the 95% CI is replaced by 99.57% CI. Numbers of IBD cases needed to produce an extra case of the comorbidity in question (number needed to diagnose comorbidity, NNDC) are presented 10 and 20 years post-IBD for significantly increased HR values. All calculations were carried out with combined Discharge Abstract Database and physician claims data and with Discharge Abstract Database alone. We report our HR results for all IBD overall and for CD and UC separately, for males and females separately, and by age distribution (<25 years, 25–50 years, 51–65 years and over 65 years). We calculated the median age and interquartile range for diagnosis of each comorbidity in IBD cases and controls and compared them using Kruskal–Wallis test. A Bonferroni correction was applied to the HR. Relative rates pre-IBD diagnosis were reported for IBD, CD and UC.

Statistical analyses used SAS V9.4 (SAS Institute Inc).