Despite advances in care, IBD hospitalization and readmission continues to have a major impact both on the cost of care and QOL of IBD patients. Our observation of 18% readmission within 30 days and 26.5% within 90 days is in keeping with findings from other recent analyses from large academic centers, cohort studies and the National Readmissions Database (NRD). Our findings of more frequent readmission among patients with longer index hospitalizations and higher comorbidities have also been observed previously [4, 5, 10]. We observed significant associations between cannabis use and younger age, male gender, African American race, tobacco use, anxiety, depression and former alcohol use. We noted that former rather than current alcohol use was associated with cannabis. The study design would not allow for further analysis of this observation, though we speculate that cannabis may also serve as a “substitute” for those with prior alcohol use. For our primary outcomes, cannabis use was noted to be associated with 30-day, but not 90-day readmission in the UC cohort, and was not associated with either 30- or 90-day readmission for those with CD.

Notably, we did not find an association between opioid use during hospitalization and readmission. Prior analysis of this issue has been mixed. Charilaou et al’s recent analysis of opioid use disorder (OUD) from the NRD from 2010 to 2014 observed a significant association with OUD and 30-day readmission, hazard ratio 1.47, 95% CI 1.28 - 1.69, P < 0.001 [20]. However, OUD, which was identified in only 1.4% of patients, appears to be a distinct entity from the far more common inpatient opioid use. We observed opioid use by almost a third of patients, also consistent with findings of other large cohort studies [21, 22] and recent meta-analysis [14]. A recent Cleveland Clinic cohort report from Hazratjee et al also found no association between inpatient opioid use and readmission, while noting that those discharged without opioids had a 2.2-fold increased risk for readmission [6].

Our study is among the first to examine cannabis use among patients hospitalized for IBD exacerbation. We found that slightly over 7% of patients reported pre-admission cannabis use, with a majority self-categorizing as recreational users. It is important to emphasize that various aspects of cannabis use vary across countries. Ransing et al demonstrated a higher prevalence of cannabis use among teenagers in countries such as South Africa, Italy and Spain [23]. Thailand in comparison demonstrated an overall decline in cannabis use across ages [23]. While the prevalence demonstrated in our study is lower than that of recent outpatient studies showing rates of active cannabis use of 10-12% [24, 25], it is similar to the findings of Dalal et al’s recent cohort reporting 8% active cannabis use among IBD inpatients [21]. As was the case with descriptions of opioid use, the method of analysis of cannabis use appears to impact this figure. Micic et al’s study of the NRD, utilizing the ICD code for cannabis dependence, reported a rate of only 1.6%. Cannabis dependence was associated with an increased risk of 30-day hospital readmission on univariate analysis, but not multivariate analysis [9]. Our analysis of cannabis use, as distinct from cannabis dependence, revealed an association between cannabis and 30-day readmission for patients with UC. In such a large registry as the NRD, it is impossible to say how precise the application of the diagnosis of cannabis dependence truly was, but the lower prevalence suggests a distinct and perhaps more medically significant entity from the more common cannabis use. Also, while our cohort showed similar rates of cannabis use to Dalal et al, we did not repeat their observation of higher inpatient opioid use by cannabis users, despite high rates of opioid use in each cohort. This leaves open the question of whether cannabis use is a “gateway” drug/predictor of future opioid use. Considering the increasing availability of cannabis, and the dismal history of opioids for IBD [26], this is an important question for future study.

Limitations to the study are those common to retrospective analysis of real-world data pulled from an electronic health record. While the methodology of the natural language search offered the ability to capture documentation of cannabis use beyond specific data point entries, it seems likely that many cannabis users were missed. Though documentation of cannabis use is encouraged by the electronic health record (EHR), it is not required to complete the admission process, at which time this history is typically recorded. Also, though cannabis use is becoming more accepted by society in general, it seems likely that the history of stigma and illegality associated with its use might cause some patients to withhold information of their own cannabis use. Also, while it would have been helpful for study purposes, urine toxicology for cannabis for objective evidence of use was (and is) not required at the study institutions, and was rare among those hospitalized for an IBD exacerbation. Though many of the typical cofactors associated with readmission were accessible by inpatient EHR data extraction, such as age, length of stay, and CCI, the data pull could not account for other factors that may have been associated with readmissions. Disease duration and severity quantified by accepted disease activity indices were not calculated during hospitalization for the study population and were therefore unknown. Also, post-discharge factors such as corticosteroid tapering and the timeliness of outpatient follow-up, which has also been observed to impact IBD readmissions [8], were not available for inclusion in our analysis. Also, we chose to include patients with a secondary admitting diagnosis of either CD or UC, but maintained the same standard of inpatient corticosteroid and/or biologic consistent with disease flare. This was intended to avoid excluding admissions for IBD for whom the admitting general medicine doctor may have preferred the use of an alternate primary admitting diagnostic code. Numbers of cannabidiol (CBD) users were also too few to perform any meaningful sub-analysis. Finally, the main limitation of the study was our inability to track cannabis use post discharge. Given the palliative properties of cannabis, it would be reasonable to assume that its use would continue or even increase following IBD hospitalization. This remains an assumption, leaving an unfilled gap in our knowledge of how cannabis use may impact rehospitalization.

In conclusion, our analysis of 30- and 90-day hospital readmissions in a large cohort of patients hospitalized with an IBD exacerbation confirmed risk factors for readmission including longer index hospitalization/LOS, and increased medical comorbidities. Inpatient opioid use was not associated with hospital readmission. A history of cannabis use was not associated with more frequent readmissions following an CD exacerbation, but was for UC patients within 30 days. As cannabis use becomes more prevalent, and as its use as a palliative therapy becomes more accepted, further analyses will be needed to better understand its impact on long-term outcomes. The current study suggests a possible “red flag” in regards to cannabis use and rehospitalization, but further studies incorporating the outpatient IBD/cannabis experience post discharge are needed to further endorse its use in this important population.