Concept: Inflammatory Bowel Disease (IBD) - Defining in Administrative Data

Introduction

This concept discusses the definition of inflammatory bowel disease (IBD), including Crohn's Disease (CD) and Ulcerative Colitis (UC). It includes a literature review describing how IBD is defined in various research conducted with administrative data. The concept begins with a description of the method used to develop the University of Manitoba IBD Epidemiology Database using Manitoba Health data, including diagnostic criteria used for identifying true cases of IBD, followed by definitions of IBD used by other Manitoba Centre for Health Policy (MCHP) deliverables.

Description of Inflammatory Bowel Disease (IBD)

IBD is a group of disorders characterized by inflammation of intestines (i.e., the intestines become red and swollen). The most common inflammatory bowel diseases are CD and UC. Crohn's disease is a chronic autoimmune disease that can affect any part of the gastrointestinal tract but most commonly occurs in the ileum (the area where the small and large intestine meet). Colitis is an inflammation of the large intestine that is caused by many different disease processes, including acute and chronic infections, primary inflammatory disorders (ulcerative colitis, Crohn's colitis, lymphocytic and collagenous colitis), lack of blood flow (ischemic colitis), and history of radiation to the large bowel.

Literature Review

Table 1 summarizes seven studies, published prior to 2010, that use administrative data to ascertain cases of IBD. The author, data source, diagnosis codes, algorithms, study cohort, validation methodology, and the purpose of study are described in Table 1.

IBD Algorithms Using Manitoba Health Data

The following studies provide descriptions of how IBD is defined by Bernstein, using administrative data from Manitoba Health.

Bernstein et al. (1999)

Bernstein et al. (1999) used administrative data from Manitoba Health to identify cases of IBD in Manitoba.

University of Manitoba Inflammatory Bowel Disease Epidemiology Database (UMIBDED)

These findings were used to develop a population-based database of IBD, known as the University of Manitoba Inflammatory Bowel Disease Epidemiology Database (UMIBDED). Since 2004, ICD-10-CA codes have been used in hospital discharge abstracts data, and thus included in the diagnosis of CD (ICD-10-CA K50) and UC (ICD-10-CA K51) in the UMIBDED.

This database has been used in numerous subsequent studies to analyze various aspects of IBD. The following is a selection of other publications by Bernstein et al. that have used UMIBDED.


• Bernstein, CN, Wajda, A, Blanchard JF. The clustering of other chronic inflammatory diseases in Inflammatory Bowel Disease: A population-based study. Gastroenterology . 2005;12:827-836.
  
  
• Bernstein, CN, Blanchard, JF, Metge, C, Yogendran, M. Does the use of 5-Aminosalicylates in Inflammatory Bowel Disease prevent the development of Colorectal Cancer?. The American Journal of Gastroenterology. 2003;98(12):2784-2788.
  
  
• Bernstein, CN, Blanchard, JF, Metge, C, Yogendran, M. The association between corticosteroid use and development of fractures among IBD patients in a population-based database. The American Journal of Gastroenterology . 2003;98(8):1797-1801.

All diagnostic codes from medical services/physician claims and hospital discharge abstracts were searched for diagnoses of CD or UC between April 1, 1984 and March 31, 1995. Individuals with at least one diagnosis of CD or UC (n=10,451) were found. Of these individuals, Bernstein et al. (1999) selected those with at least three IBD medical contacts occurring between 1984 and 1992 and those with at least one IBD medical contact after 1993 to study further using a self-administered questionnaire (n=5182). Chart reviews were randomly completed among those individuals who responded to the questionnaire and gave written consent (n=2725).

Based on findings from the medical contact search, questionnaire responses, and medical chart review, Bernstein et al. (1999) developed the following algorithms with a high degree of specificity and sensitivity to classify individuals as having CD, UC or no IBD using administrative health data.

Crohn's Disease (CD)


• For an individual registered with Manitoba Health for at least two years, he/she was considered to have CD if he/she had five or more hospitalizations or physician claims with any one of the recorded ICD-9-CM diagnostic codes as 555.
  
  
• For an individual registered with Manitoba Health for less than two years, he/she was considered to have CD if he/she had three or more hospitalizations or physician claims with any one of the recorded ICD-9-CM diagnostic codes as 555.

Ulcerative Colitis (UC)


• For an individual registered with Manitoba Health for at least two years, he/she was considered to have UC if he/she had five or more hospitalizations or physician claims with any one of the recorded ICD-9-CM diagnostic codes as 556.
  
  
• For an individual registered with Manitoba Health for less than two years, he/she was considered to have UC if he/she had three or more hospitalizations or physician claims with any one of the recorded ICD-9-CM diagnostic codes as 556.

If an individual had records that satisfied the criteria for both CD and UC, the nine most recent medical contacts were considered and the majority diagnosis was used. (Bernstein et al., 1999)

No IBD


• An individual was considered to have no IBD if he/she did not have any medical contact related to IBD (Bernstein et al., 1999).

Validation

This diagnostic criterion yielded sensitivity and specificity of approximately 90%. It was validated by Bernstein et al. (1999) by comparing the identified cases of IBD (from the medical services/physician claims and hospital discharge abstracts data) with self-reported information and medical chart information.

MCHP Studies - Algorithms

The following describes IBD algorithms used in studies conducted by the MCHP.

1. Fransoo et al. (2005)

Fransoo et al. (2005) analyzed the percentage of Manitobans with CD or UC (termed IBD treatment prevalence), with administrative health data using the same definition as Bernstein et al. (1999) because the algorithm provided a high degree of sensitivity and specificity. A case of IBD was defined as having at least 5 diagnoses of CD or UC, based on hospital discharge abstracts and medical services/physician claims data (ICD-9-CM codes 555 or 556, respectively), between 1994/95 and 2003/04, provided that the individual had been registered with Manitoba Health for two years or more. For situations in which an individual had been registered with Manitoba Health for less than two years, only three diagnoses of CD or UC, with the above ICD codes, were required for identification as an IBD case.

2. Raymond et al. (2010)

Over a ten year period (1997/98-2008/09), Raymond et al. (2010) assessed the use of tumor necrosis factor (TNF) alpha inhibitors in adults with IBD. An individual was considered to have IBD if he/she had at least one hospital discharge abstract or medical services/physician claim record with an ICD-9-CM code 555 or 556 AND/OR ICD-10-CA code K50 or K51.

3. Lix et al. (2021)

In Lix et al. (2021) an IBD sub-group was selected using the following case definition: individuals had to have at least one day of health insurance coverage prior to the index date in order to be included in the IBD sub-group. The IBD case definition is:

1. For an individual with at least two years of health insurance coverage on or before the index date, that individual is considered to be an IBD case if he/she had five or more hospital discharge abstracts or physician billing claims with a relevant ICD diagnosis code (in any diagnosis position in hospital discharge abstracts);
   
2. For an individual with less than two years of health insurance coverage on or before the index date, that individual is considered to be an IBD case if he/she had three or more hospital discharge abstracts or physician billing claims with a relevant ICD diagnosis code (in any diagnosis position in hospital discharge abstracts);
   
3. The relevant ICD diagnosis codes were ICD-9-CM 555 or 556, or ICD-10-CA K50 or K51.

Cautions

Bernstein's validated algorithm, as replicated by Fransoo et al. (2005), is the preferred method to use when calculating the prevalence of IBD. The algorithm used by Raymond et al. (2010) required only one health care contact to define cases of IBD, however it must be noted that in some cases using one hospitalization or physician visit to identify the prevalence of a disease is not recommended.

Limitations

• Because IBD is a disease with relapse and remission of symptoms, prevalent cases of CD or UC may be misclassified as incident cases (Bernstein et al., 1999). A long medical history must be obtained from administrative data to determine whether an individual is truly a new case of IBD or if the individual has been diagnosed in the past, experienced remission for a long duration and is currently experiencing a relapse of symptoms (may be evidenced by a long break between IBD-related hospital episodes or physician visits). Bernstein et al. (1999) cautions that prevalence and incidence of IBD may be under- and over-estimated, respectively.
  
  
• Individuals with IBD often experience symptoms long before diagnosis (Burgmann et al., 2006). The onset of IBD may be estimated inaccurately based on administrative data because (1) the condition may be misdiagnosed as Irritable Bowel Syndrome (IBS) or (2) an individual's symptoms may be mild and therefore, further investigation does not occur (Burgmann et al., 2006).