Supplemental Methods and Results
Categorization of Crohn’s disease (CD) versus ulcerative colitis (UC)
To be categorized as CD, the patient must have had more CD diagnoses than UC, and the most recent diagnosis on or prior to the start of therapy had to be CD. An analogous algorithm was used to define UC. We also excluded patients who would otherwise meet the definition of UC but for whom there was a coded diagnosis of fistula or ostomy any time prior to the start of follow-up.
Exposure definitions
Patients were included in the study if they newly initiated a prolonged course of steroids or newly initiated anti-TNF therapy. New initiation of anti-TNF therapy required receipt of at least 1 dispensing for an anti-TNF drug with at least 1 filled prescription for a CS and no dispensing for any anti-TNF medication in the prior 12 months. Prolonged CS use was defined as either:
a)receipt of two or more non-budesonide CS prescriptions within 365 days where the second or later prescription results in a cumulative prescribed dose greater than 3000mg of prednisone (or equivalent), the latter prescription occurred more than 12 months after enrollment in Medicare, and there were no prescriptions for anti-TNF therapy within the 365 days prior to the prescription that made the cumulative dose >3000mg.
b)receipt of two or more budesonide prescriptions within 365 days where the second or later prescription results in a cumulative prescribed dose greater than 600mg of budesonide, the latter prescription occurred more than 12 months after enrollment in Medicare, and there were no prescriptions for anti-TNF therapy within the 365 days prior to the prescription that made the cumulative dose >600mg.
These definitions were based on the amount of corticosteroids prescribed in a high dose steroid taper. For example, treatment with prednisone 60mg daily for 1 week and then tapering by 5mg every week includes 2730 mg of prednisone. Similarly, a budesonide regimen including 9mg daily for 6 weeks, 6mg daily for 2 weeks, and 3mg daily for 2 weeks includes 504mg in total. In a sensitivity analysis, the threshold of traditional CS used to define prolonged CS use was lowered to 2000mg. This is approximately equal to a CS taper starting at 40mg of prednisone daily and tapering by 5mg per week to 10mg daily but with need for a second taper from 30mg daily. More than 99.5% of all steroid prescriptions were oral formulations and the remainder were intravenous.
Secondary outcome definitions
All outcomes other than cancer were assessed at any time after the start of therapy; cancer outcomes were measured 6 months after the start of therapy given the biological implausibility that a medical therapy would cause cancer within 6 months. The following criteria were used to identify key secondary outcomes:
Any cancer other than non-melanoma skin cancer was coded using the previously established modification of the algorithm developed by Setoguchi1,2.
Acute myocardial infarction required at least 1 inpatient claim with a discharge ICD-9 diagnoses for acute myocardial infarction (410 excluding 410.x2) and at least one night of inpatient stay except if the patient died. The validity of the algorithm has been evaluated in prior studies with positive predictive values (PPV) that exceed 90%3.
Stroke required ICD-9 codes 430, 431, 433.x1, 434.x1, or 436 in the primary diagnosis position of a hospitalization. This has been demonstrated to have a PPV that exceeds 90%4.
Cardiac arrest was identified by ICD-9 code 427.5 in the primary discharge diagnosis or ED diagnosis, which has been shown to have a PPV of 81.5%5.
Sudden death included those with cardiac arrest or those with a primary discharge or emergency department code of 798, 798.1, or 798.2 and those who died without an emergency department visit or hospitalization in the prior 90 days.
Major adverse cardiovascular event (MACE) including acute myocardial infarction, stroke, sudden death, or the need for revascuarization. The latter was based on CPT codes for percutaneous coronary intervention or coronary artery bypass graft surgery. Percutaneous coronary intervention was identified by ≥ 1 CPT codes 92980-92996 or ICD-9 procedure codes 00.66 or 36.01-36.09 from inpatient, outpatient, revenue center, or carrier line file claims. Coronary artery bypass graft surgery was identified by 1 CPT codes 33510-33536 or ICD-9 procedure codes 36.10-36.19 from inpatient, outpatient, revenue center, or carrier line file claims.
Pulmonary embolus was based on a hospitalized diagnosis code of 415.1x in any position as per prior validation studies. The PPV of this algorithm has varied from 31% to 97%6.
Hip fracture was identified based on the algorithm developed by Ray et al with 98% PPV7.
Serious bacterial or opportunistic infection defined as hospitalization with the infection as the primary discharge diagnosis. This definition has been previously demonstrated to have PPVs in excess of 90%8. Opportunistic infections were defined as in prior studies in similar cohorts in Medicare and included infection with the following organisms9-11: Aspergillus, Blastomyces, Coccidiodes, Cryptococcus, Histoplasma, Pneumocystis, Actinomyces, Legionella, Listeria, Norcardia, Salmonella, tuberculous and non-tuberculous mycobacteria, Toxoplasma, herpes zoster, and JC virus. For aspergillosis, a prescription for posaconazole, itraconazole, or voriconazole was required within 90 days of the diagnosis. For blastomycosis, coccidioidomycosis, cryptococcosis, histoplasmosis, and endemic mycosis, we also required a prescription for fluconazole, itraconazole, or voriconazole within 90 days of the diagnosis code. For tuberculosis, we required a prescription of pyrazinamide to assure that treatment was not for latent tuberculosis infection. For herpes zoster, we required a prescription of acyclovir, valacyclovir, or famcyclovir within 90 days of the diagnosis. For all other opportunistic infections, we did not require concomitant antimicrobial prescriptions.
Emergency bowel resection surgery was identified using our previously established codes and algorithms11,12. Emergency surgeries were those that occurred during the same hospitalization as an emergency department visit or after more than 24 hours of hospitalization13.
In addition, IBD related hospitalizations using our previously established algorithm11,12 were recorded.
Covariates
Potential confounding variables were measured including factors likely to be associated with the choice between chronic intermittent steroid use or anti-TNF therapy and the outcomes of interest. The covariates are listed in supplemental tables 1 and 2 along with a description of whether they were included in the baseline model, the time updating models, or both. For chronic conditions at baseline, the look back period used all available data. The following variables were measured at the index date: age, sex, calendar year of cohort entry, urban versus rural residency based on zip codes, cumulative dose of corticosteroids in the 6 months prior, and receipt of the following tests in the 56 days prior – colonoscopy or sigmoidoscopy, CT scan or MRI of the abdomen or pelvis, and small bowel follow-through study. The following potential confounders were measured at index date and as time updating variables every 28 days using a 6 month look back period unless otherwise specified: medical therapies received in the prior month, the combined Elixhauser-Charlson comorbidity index described by Gagne (Supplemental Table 3)14, procedures for fistula drainage or seton placement in the prior 365 days12, dehydration or hypovolemia, C. difficile infection, testing for C. difficile, completion of a stool culture, serious or opportunistic infections as defined above, weight loss or malnutrition, anemia from iron deficiency or other nutritional disorders, electrolyte disorders, receipt of a blood transfusion or intravenous iron, receipt of total parenteral nutrition, receipt of prescriptions for 5-ASA medications, thiopurine analogues, methotrexate, narcotics, oral or intravenous antibiotics which were categorized as quinolones, metronidazole, other antibiotics, antiviral medications for herpes, anti-HIV medications, anti-fungal medications, colonoscopy or sigmoidoscopy, CT scan or MRI of the abdomen or pelvis, small bowel follow-through study, presence of an ostomy, pyoderma gangrenosum, diabetes, hypertension, coronary artery disease, congestive heart failure, osteoporosis, bisphosphonate use, vitamin D in a dose greater than or equal to 2000 international units, history of stroke, Parkinson’s disease, problems with balance, hypercholesterolemia, statin use, fibrate use, IBD-related hospitalizations [categorized as none, short (<8 days) and long (8 or more days)], number of non-IBD related hospitalizations, and the number of non-IBD and non-narcotic medications.
Statistical Analysis
Data setup
In order to carry out a marginal structural model (MSM) analysis, the first step is to set up a longitudinal data set, with equally spaced time intervals appropriate for discrete-time Cox models. We used 28 day time intervals. That is, each row of data corresponded to one 28 day period for a given patient. For the kth time interval for a given patient, exposure (CS or anti-TNF) was defined as the treatment the patient was taking at the beginning of that time interval. Time-updated covariates in the kth time interval were the values that those covariates were at the start of the interval. This setup ensures that the time-varying propensity score model always conditions on covariates that preceded the current treatment value. Not all time-varying variables had documented updates in every interval. We carried forward values from the previous interval when no new information was available. In the kth time interval, the outcome was assigned a value of ‘1’ if it occurred anytime during the interval. Similarly, the censoring variable took a value of ‘1’ if censoring occurred anytime during the interval, and prior to the outcome occurring.
Treatment and censoring models
In order to estimate weights, we need to estimate the probability of the observed treatment, given covariate and treatment history, at each time. We also need to estimate the probability of censoring, given covariate and treatment history, at each time. To estimate the probability of treatment (CS or anti-TNF) we first fitted a baseline logistic regression model. This is a model for the initial treatment decision, given baseline (pre-treatment) covariates. Next, we fitted a repeated measures logistic regression model for all time periods after the initial time period. For the primary analysis, where medication exposure was unidirectional, prior exposure was not (directly) included in the treatment model. This is because CS users cannot have prior treatment of anti-TNF, and therefore the coefficient of past treatment would not be identifiable. However, we implicitly condition on past treatment in that the probability of being in the anti-TNF group at time t is set to 1 if they were in the anti-TNF group at baseline. For the ‘as-treated’ model, the treatment model conditioned on past treatment.For example, if a patient was on CS at baseline, at time period 2 we would model the probability of CS at time period 2, given that the patient was in the CS group at time 1, and given covariate history. We expect that a baseline model that implicitly does not condition on past treatment would be different from models that condition on past treatment, which is the reason we fitted two separate logistic regression models. The output from these logistic regression models was the probability of each patient’s observed treatment. By multiplying these probabilities, we obtain an estimate of each patient’s whole treatment history.
The procedure for estimating the probability of non-censoring was similar. We fitted a baseline logistic regression model, where the outcome was non-censoring and baseline covariates were predictors. We then fitted a repeated measures logistic regression model, where censoring history and covariate history was included in the models. The output from these models was the probability of not being censored up to the current time.
In order to obtain the weights, we first multiplied these censoring probabilities by the treatment probabilities. This produces an estimate of the probability of that particular treatment and censoring history. For example, if a patient was taking CS for 3 time intervals and was censored at the end of the 3rd interval, we would multiple the probability of CS at times 1, 2, and 3, by the probability of non-censoring at times 1 and 2 and by the probability of censoring at time 3.
Variable selection
There were 57 covariates that were a priori selected as potential confounders. We excluded variables from the treatment and censoring models if their prevalence was too low. Specifically, we decided to exclude variables if any cell count for any model was less than 10. The other covariates, 36 baseline and 46 time-varying for CD and 25 baseline and 39 time-varying for UC, were included in the treatment and censoring models (SupplementalTables 1 and 2).
Weights
The probability of observed treatment and censoring history, calculated as described above, is the denominator of the weights. If a value of 1 is used in the numerator, these are unstabilized weights. We used stabilized weights, which involve estimating the probability of treatment and censoring using only baseline, rather than time-updated, predictors. Stabilized weights are less variable and lead to more efficient estimates.
We trimmed the tails of the weights to make the groups more comparable and to make the positivity assumption more plausible. The positivity assumption states that the probability of treatment and non-censoring should be greater than 0 at every observed level of covariate history. In other words, at each time each patient should have some chance of being in either treatment group. The exact level of weights where the tails were trimmed was determined by qualitatively reviewing the distribution of the baseline propensity scores (Supplemental Figure 1) and the resulting standardized mean differences comparing the prevalence of baseline covariates between the treatment groups. The choice was made independent of knowledge of the impact that would result on the association of the primary and secondary outcome measures and the two treatments. For CD, patients were excluded if the propensity for corticosteroid therapy was greater than 0.98 or less than 0.30. For UC, patients were excluded if the propensity for corticosteroid therapy was greater than 0.98 or less than 0.40.
We truncated weights at the 2nd and 98th percentiles. What this means is that for any weight that was, e.g., greater than the 98th percentile, its value was set to that of the weight at the 98th percentile. Weights were truncated at the 2nd and 98th percentile to avoid excessive influence of patients with extremely low or high probability of receiving one of the treatments.
Marginal Structural Model (MSM)
For the primary analyses, the MSM included current treatment and time as predictors. These were estimated using a weighted generalized estimating equations approach (SAS PROC GENMOD). Asymptotic standard errors were used to obtain confidence intervals. Standard errors reflect the possibility that individual patients could contribute to more than one exposure group and in the as treated model could contribute multiple exposure periods to a single treatment. The ORs from these models are approximate hazard ratios.
Supplemental Table 1. Characteristics of the Crohn’s disease population at cohort entry
New initiators of prolonged steroids or anti-TNF / Final cohort excluding outliers from baseline PS / Final weighted cohortSteroids
(N=8818) / Anti-TNF
(N=2031) / Steroids
(N=7694) / Anti-TNF
(N=1879) / Steroids / Anti-TNF
Characteristic / Group / % / % / SMD / % / % / SMD / % / % / SMD
Age at index* / 1= 18-34.9 / 19.7 / 25.9 / 0.15 / 21.1 / 24.6 / 0.08 / 22.0 / 23.5 / 0.04
2= 35-49.9 / 24.2 / 30.5 / 0.14 / 25.9 / 30.1 / 0.09 / 26.9 / 27.8 / 0.02
3= 50-64.9 / 16.0 / 15.5 / 0.01 / 15.9 / 15.6 / 0.01 / 15.8 / 16.1 / 0.01
4= 65-69.9 / 13.7 / 11.8 / 0.06 / 13.0 / 12.2 / 0.02 / 12.8 / 12.0 / 0.02
5= 70-74.9 / 10.2 / 8.1 / 0.07 / 9.8 / 8.6 / 0.04 / 9.5 / 8.8 / 0.02
6= 75-79.9 / 7.6 / 5.0 / 0.11 / 7.3 / 5.3 / 0.08 / 6.8 / 6.5 / 0.01
7= 80-84.9 / 5.6 / 2.7 / 0.15 / 5.1 / 2.9 / 0.11 / 4.6 / 4.0 / 0.03
8= 85+ / 3.0 / 0.6 / 0.18 / 1.9 / 0.6 / 0.12 / 1.7 / 1.3 / 0.03
Gender* / Female / 62.4 / 65.1 / 0.06 / 62.8 / 64.7 / 0.04 / 63.1 / 62.5 / 0.01
Race* / White / 83.6 / 83.9 / 0.01 / 82.9 / 84.3 / 0.04 / 83.1 / 82.4 / 0.02
Calendar year at index* / 2001 / 11.1 / 6.5 / 0.16 / 11.6 / 6.9 / 0.16 / 10.7 / 11.5 / 0.03
2002 / 6.8 / 6.5 / 0.01 / 7.3 / 6.9 / 0.02 / 7.2 / 7.5 / 0.01
2003 / 6.0 / 5.1 / 0.04 / 6.2 / 5.3 / 0.04 / 6.0 / 6.8 / 0.03
2004 / 7.7 / 5.7 / 0.08 / 7.6 / 6.1 / 0.06 / 7.4 / 7.6 / 0.01
2005 / 6.4 / 4.9 / 0.07 / 6.4 / 5.1 / 0.06 / 6.5 / 7.4 / 0.04
2006 / 2.3 / 2.1 / 0.01 / 2.2 / 2.2 / 0.00 / 2.2 / 2.2 / 0.00
2007 / 13.7 / 10.5 / 0.10 / 12.4 / 10.6 / 0.06 / 11.9 / 11.0 / 0.03
2008 / 8.9 / 9.1 / 0.01 / 8.8 / 9.4 / 0.02 / 8.7 / 7.6 / 0.04
2009 / 8.2 / 8.8 / 0.02 / 8.0 / 8.9 / 0.03 / 8.1 / 8.5 / 0.01
2010 / 5.6 / 9.0 / 0.13 / 5.8 / 8.5 / 0.10 / 6.3 / 6.0 / 0.01
2011 / 6.5 / 9.4 / 0.11 / 6.7 / 8.4 / 0.06 / 6.9 / 7.0 / 0.00
2012 / 7.6 / 10.0 / 0.08 / 7.5 / 9.8 / 0.08 / 7.8 / 7.2 / 0.02
2013 / 9.3 / 12.4 / 0.10 / 9.3 / 11.8 / 0.08 / 10.2 / 9.8 / 0.01
Urban residence* / Yes / 72.1 / 69.4 / 0.06 / 71.6 / 69.8 / 0.04 / 71.5 / 72.1 / 0.01
Diabetes*^ / Yes / 13.1 / 11.1 / 0.06 / 12.6 / 11.3 / 0.04 / 12.3 / 11.8 / 0.02
Hypertension*^ / Yes / 34.4 / 29.6 / 0.10 / 33.3 / 30.3 / 0.06 / 32.6 / 30.9 / 0.04
Coronary artery disease^ / Yes / 11.4 / 9.5 / 0.06 / 10.7 / 9.6 / 0.04 / 10.2 / 10.4 / 0.01
Congestive heart failure^ / Yes / 4.9 / 3.1 / 0.09 / 4.5 / 3.3 / 0.06 / 4.3 / 4.0 / 0.02
Carotid artery disease / Yes / 0.7 / 0.4 / 0.04 / 0.6 / 0.5 / 0.01 / 0.6 / 0.5 / 0.01
Hypercholesterolemia*^ / Yes / 23.4 / 21.4 / 0.05 / 22.7 / 22.0 / 0.02 / 22.5 / 21.1 / 0.03
Osteoporosis*^ / Yes / 12.9 / 11.7 / 0.04 / 12.6 / 12.1 / 0.02 / 12.4 / 12.3 / 0.00
Stroke/Parkinsons disease^ / Yes / 1.8 / 1.4 / 0.03 / 1.7 / 1.5 / 0.02 / 1.6 / 1.4 / 0.02
Problems with balance*^ / Yes / 17.0 / 17.4 / 0.01 / 16.8 / 16.8 / 0.00 / 16.8 / 15.5 / 0.04
Falls in prior 365 days / Yes / 0.3 / 0.1 / 0.04 / 0.3 / 0.2 / 0.02 / 0.3 / 0.1 / 0.04
DXA scan in prior 365 days*^ / Yes / 13.6 / 13.3 / 0.01 / 13.3 / 13.4 / 0.00 / 13.3 / 13.7 / 0.01
Combined comorbidity score*^ / 0= <=0 / 31.8 / 25.0 / 0.15 / 31.6 / 25.9 / 0.13 / 30.4 / 30.0 / 0.01
1= 1 / 22.0 / 22.6 / 0.01 / 22.2 / 22.8 / 0.01 / 22.3 / 23.1 / 0.02
2= 2-3 / 24.8 / 28.2 / 0.08 / 25.1 / 28.1 / 0.07 / 25.8 / 26.2 / 0.01
3= 4+ / 21.4 / 24.2 / 0.07 / 21.1 / 23.2 / 0.05 / 21.4 / 20.7 / 0.02
Fistula w/in 365 & 56 days prior to index*^ / 0= none in past 365 dys / 78.0 / 65.6 / 0.28 / 76.7 / 67.3 / 0.21 / 74.7 / 73.9 / 0.02
1= w/in 57-365 dys prior / 14.9 / 17.7 / 0.08 / 15.7 / 17.8 / 0.06 / 16.1 / 16.4 / 0.01
2= w/in 56 dys prior / 7.1 / 16.7 / 0.30 / 7.5 / 14.9 / 0.24 / 9.3 / 9.7 / 0.01
Small bowel follow-through exam w/in 365 & 56 days prior to index*^ / 0= none in past 365 dys / 76.9 / 72.8 / 0.09 / 76.3 / 73.7 / 0.06 / 75.4 / 72.5 / 0.07
1= w/in 57-365 dys prior / 18.2 / 16.7 / 0.04 / 18.5 / 16.8 / 0.04 / 18.3 / 20.0 / 0.04
2= w/in 56 dys prior / 4.9 / 10.5 / 0.21 / 5.2 / 9.5 / 0.17 / 6.3 / 7.5 / 0.05
Colonoscopy or sigmoidoscopy w/in 365 & 56 days prior to index*^ / 0= none in past 365 dys / 51.2 / 39.0 / 0.25 / 49.8 / 41.1 / 0.18 / 47.8 / 45.7 / 0.04
1= w/in 57-365 dys prior / 38.5 / 36.7 / 0.04 / 39.1 / 37.8 / 0.03 / 38.8 / 39.3 / 0.01
2= w/in 56 dys prior / 10.3 / 24.3 / 0.38 / 11.2 / 21.0 / 0.27 / 13.4 / 15.0 / 0.05
CT or MRI scan of abdomen/pelvis w/in 365 & 56 days prior to index*^ / 0= none in past 365 dys / 55.2 / 48.3 / 0.14 / 54.1 / 48.7 / 0.11 / 52.9 / 49.9 / 0.06
1= w/in 57-365 dys prior / 30.7 / 28.5 / 0.05 / 30.9 / 28.8 / 0.05 / 30.4 / 31.5 / 0.02
2= w/in 56 dys prior / 14.1 / 23.2 / 0.24 / 15.0 / 22.5 / 0.19 / 16.7 / 18.6 / 0.05
C difficile w/in 183 & 56 days prior to index^ / 0= none in past 183 dys / 97.5 / 97.9 / 0.03 / 97.7 / 97.9 / 0.01 / 97.5 / 98.4 / 0.06
1= w/in 57-183 dys prior / 1.9 / 1.4 / 0.04 / 1.7 / 1.3 / 0.03 / 1.8 / 1.1 / 0.06
2= w/in 56 dys prior / 0.6 / 0.7 / 0.01 / 0.6 / 0.7 / 0.01 / 0.7 / 0.5 / 0.03
Testing for C difficile w/in 183 & 56 days prior to index*^ / 0= none in past 183 dys / 84.8 / 82.9 / 0.05 / 84.8 / 83.1 / 0.05 / 84.5 / 84.7 / 0.01
1= w/in 57-183 dys prior / 11.9 / 11.0 / 0.03 / 11.6 / 11.5 / 0.00 / 11.8 / 10.2 / 0.05
2= w/in 56 dys prior / 3.3 / 6.1 / 0.13 / 3.5 / 5.4 / 0.09 / 3.7 / 5.1 / 0.07
Ostomy w/in 183 days prior to index^ / Yes / 5.7 / 10.1 / 0.16 / 6.0 / 9.3 / 0.12 / 6.2 / 7.9 / 0.07
Bowel Resection w/in 183 days prior to index^ / Yes / 2.6 / 3.9 / 0.07 / 2.7 / 3.7 / 0.06 / 2.9 / 2.8 / 0.01
Serious infection w/in 183 days prior to index^ / Yes / 4.8 / 5.7 / 0.04 / 4.8 / 5.4 / 0.03 / 5.0 / 5.3 / 0.01
Opportunistic infection w/in 183 days prior to index / Yes / 0.7 / 0.4 / 0.04 / 0.7 / 0.5 / 0.03 / 0.7 / 0.5 / 0.03
Blood transfusion w/in 183 days prior to index^ / Yes / 4.3 / 5.6 / 0.06 / 4.5 / 5.4 / 0.04 / 4.6 / 4.9 / 0.01
Intravenous iron w/in 183 days prior to index^ / Yes / 1.2 / 1.3 / 0.01 / 1.2 / 1.3 / 0.01 / 1.2 / 1.4 / 0.02
Parenteral nutrition w/in 183 days prior to index^ / Yes / 2.1 / 4.0 / 0.11 / 2.3 / 3.7 / 0.08 / 2.5 / 3.2 / 0.04
Anemia from iron deficiency or NOS w/in 183 days prior to index*^ / Yes / 27.0 / 31.6 / 0.10 / 26.8 / 30.6 / 0.08 / 27.6 / 27.8 / 0.00
Other nutritional anemia (e.g. B12) w/in 183 days prior to index^ / Yes / 6.3 / 5.5 / 0.03 / 6.1 / 5.6 / 0.02 / 5.9 / 5.0 / 0.04
Weight loss or malnutrition w/in 183 days prior to index^ / Yes / 5.8 / 8.5 / 0.10 / 6.0 / 7.9 / 0.07 / 6.2 / 6.9 / 0.03
Dehydration/hypovolemia w/in 183 days prior to index*^ / Yes / 12.1 / 15.8 / 0.11 / 12.6 / 15.6 / 0.09 / 13.3 / 14.1 / 0.02
Electrolyte disorders w/in 183 days prior to index*^ / Yes / 14.2 / 17.9 / 0.10 / 14.4 / 16.9 / 0.07 / 14.9 / 15.0 / 0.00
Pyoderma Gangrenosum w/in 183 days prior to index^ / Yes / 0.6 / 0.7 / 0.01 / 0.7 / 0.8 / 0.01 / 0.7 / 0.8 / 0.01
Stool culture w/in 183 days prior to index*^ / Yes / 7.2 / 9.1 / 0.07 / 7.5 / 8.7 / 0.04 / 7.8 / 8.3 / 0.02
Quinolone Rx w/in 183 days prior to index*^ / Yes / 27.8 / 34.6 / 0.15 / 28.7 / 33.0 / 0.09 / 29.7 / 30.9 / 0.03
Metronidazole Rx w/in 183 days prior to index*^ / Yes / 21.2 / 31.7 / 0.24 / 22.5 / 29.7 / 0.16 / 24.2 / 25.9 / 0.04
Other Antibiotic Rx w/in 183 days prior to index*^ / Yes / 42.6 / 44.4 / 0.04 / 42.8 / 43.8 / 0.02 / 42.8 / 41.1 / 0.03
Herpes Antiviral Rx w/in 183 days prior to index^ / Yes / 3.1 / 3.4 / 0.02 / 2.9 / 3.4 / 0.03 / 3.0 / 2.8 / 0.01
Antifungal Rx w/in 183 days prior to index^ / Yes / 7.3 / 9.8 / 0.09 / 7.5 / 9.4 / 0.07 / 7.7 / 8.2 / 0.02
Mesalamine Rx w/in 183 days prior to index*^ / Yes / 50.7 / 51.4 / 0.01 / 51.3 / 51.3 / 0.00 / 51.5 / 53.3 / 0.04
AZA/6MP Rx w/in 183 days prior to index*^ / Yes / 25.6 / 34.8 / 0.20 / 27.1 / 33.7 / 0.14 / 28.5 / 30.7 / 0.05
Methotrexate Rx w/in 183 days prior to index^ / Yes / 2.1 / 3.5 / 0.08 / 2.3 / 3.4 / 0.07 / 2.2 / 3.7 / 0.09
Narcotics Rx (excluding Belladonna/Opium) w/in 183 days prior to index*^ / Yes / 52.4 / 62.4 / 0.20 / 53.8 / 61.0 / 0.15 / 55.3 / 56.9 / 0.03
PPI Rx w/in 183 days prior to index*^ / Yes / 41.6 / 41.0 / 0.01 / 41.3 / 40.6 / 0.01 / 41.1 / 40.8 / 0.01
Benzodiazepine Rx w/in 183 days prior to index*^ / Yes / 14.9 / 13.4 / 0.04 / 15.3 / 13.8 / 0.04 / 15.2 / 15.1 / 0.00
Bisphosphonate Rx w/in 183 days prior to index*^ / Yes / 13.2 / 8.8 / 0.14 / 12.5 / 9.1 / 0.11 / 11.7 / 11.4 / 0.01
High dose Vitamin D Rx w/in 183 days prior to index / Yes / 0.3 / 0.3 / 0.00 / 0.3 / 0.3 / 0.00 / 0.3 / 0.3 / 0.00
Stains Rx w/in 183 days prior to index*^ / Yes / 15.9 / 11.7 / 0.12 / 14.9 / 12.2 / 0.08 / 14.2 / 12.7 / 0.04
Fibrates Rx w/in 183 days prior to index^ / Yes / 3.6 / 3.2 / 0.02 / 3.3 / 3.2 / 0.01 / 3.3 / 2.9 / 0.02
Prednisone use in 183 days prior to index* / Yes / 63.3 / 73.3 / 0.22 / 67.2 / 72.8 / 0.12 / 68.6 / 71.5 / 0.06
Budesonide use in 183 days prior to index* / Yes / 40.6 / 17.1 / 0.54 / 34.5 / 18.0 / 0.38 / 31.1 / 30.0 / 0.02
# of non-IBD medication classes taking during 183 days prior to index*^ / 0-3 / 17.2 / 16.7 / 0.01 / 17.3 / 17.4 / 0.00 / 17.5 / 17.3 / 0.01
4-6 / 26.7 / 27.0 / 0.01 / 27.1 / 27.3 / 0.00 / 27.1 / 26.7 / 0.01
7-10 / 30.7 / 32.2 / 0.03 / 30.7 / 31.7 / 0.02 / 30.8 / 31.1 / 0.01
> 10 / 25.4 / 24.1 / 0.03 / 24.9 / 23.6 / 0.03 / 24.7 / 24.8 / 0.00
Total IBD hospitalization days in 183 days prior to index*^ / 0=none / 71.0 / 57.0 / 0.29 / 69.0 / 58.6 / 0.22 / 66.6 / 63.7 / 0.06
1= >= 1 day / 29.0 / 43.0 / 0.29 / 31.0 / 41.4 / 0.22 / 33.4 / 36.3 / 0.06
# of hospitalizations other than IBD in 183 days prior to index*^ / 0 / 83.3 / 86.6 / 0.09 / 83.7 / 86.4 / 0.08 / 84.2 / 84.8 / 0.02
1+ / 16.7 / 13.4 / 0.09 / 16.3 / 13.6 / 0.08 / 15.8 / 15.2 / 0.02
* Included in baseline model