Psychotic disorders hospitalizations associated with cannabis abuse or dependence: A nationwide big data analysis

Abstract Objectives We aimed to describe and correlate the hospital panorama of psychotic disorders (PD) with cannabis use (CU) trends in all Portuguese public hospitals. Methods We conducted a retrospective observational study that analysed all hospitalizations that occurred in Portuguese public hospitals from 2000 to 2015. Hospitalizations with a primary diagnosis of PD or schizophrenia were selected based on Clinical Classification Software diagnostic single‐level 659. Episodes associated with CU were identified by the International Classification of Diseases Version 9, Clinical Modification code 304.3/305.2 that correspond to cannabis dependence/cannabis abuse. Results The number of hospitalizations with a primary diagnosis of PD and schizophrenia associated with CU rose 29.4 times during the study period, from 20 to 588 hospitalizations yearly (2000 and 2015, respectively) with a total of 3,233 hospitalizations and an average episode cost of €3,500. Male patients represented 89.8% of all episodes, and the mean/median age at discharge were 30.66/29.00 years, respectively. From all hospitalizations with a primary diagnosis of PD or schizophrenia, the ones with a secondary diagnosis of CU rose from 0.87% in 2000 to 10.60% in 2015. Conclusions The increase on secondary diagnosis coding and the change on cannabis patterns of consumption in Portuguese population with an increasing frequency of moderate/high dosage cannabis consumers may explain the rise on PD hospitalizations.

development of a PD may also play a role in the presence of addictive behaviours (Ksir & Hart, 2016).
PD are one of the most serious group of medical conditions that may appear after CU and may generate comorbidities in an acute or chronic timeline (Gage, Hickman, & Zammit, 2016). Although it is difficult to establish a direct causal link between CU and PD, a large number of observational studies have found a clear association between CU and schizophrenia and the development of psychotic symptoms (Colizzi et al., 2018;Ferdinand et al., 2005;Hall & Degenhardt, 2000;Miettunen et al., 2008;Ortiz-Medina et al., 2018;Regier et al., 1990;Semple, McIntosh, & Lawrie, 2005). Patients diagnosed with a first episode of psychosis or with schizophrenia are more likely to report current or prior use of cannabis, compared with the general population. Cannabis properties also play a role with the incidence of a first psychotic event being linked to the dosage of the cannabis used by the patient (Andreasson, Allebeck, Engstrom, & Rydberg, 1987;Di Forti et al., 2019;Roncero et al., 2018;Zammit, Allebeck, Andreasson, Lundberg, & Lewis, 2002). Continued CU not only might play a role on the first psychotic event but also affects negatively the prognosis of a patient after the first episode of psychosis, increasing the relapse rate (Schoeler et al., 2016). According to Tennant et al., PD related with CU may occur in three distinct situations: (a) auto-limited psychosis caused by acute cannabis consumption that ends after stopping the consumption, (b) psychosis that develops during CU that requires medical treatment or hospitalization even after stopping the consumption, and (c) psychosis that manifests years after CU but is likely directly related to it (Rylander, Winston, Medlin, Hull, & Nussbaum, 2018;Tennant, 2005).
CU is linked to an increase in hospital visits and in the utilization rates of emergency and hospital services for psychosis (Rylander et al., 2018). In France, emergency department visits associated with CU have increased, and in the United States, there was a significant increase on CU-related hospitalizations in the recent years (Charilaou et al., 2017;Noel, Maghoo, Franke, Viudes, & Minodier, 2019 The main goal of this study was to analyse clinical, demographic, and administrative trends regarding PD hospitalizations associated with cannabis abuse and/or dependence in the recent years in Portugal. The secondary goal of this study was to use secondary data such as administrative databases in Psychiatry and Epidemiology observational research.

| MATERIALS AND METHODS
We conducted a retrospective observational study using a database     (Miettunen et al., 2008;Roncero et al., 2018). Also, risk behaviours are clearly more associated with younger age groups, which can be a possible explanation for the differ-  (SICAD, 2017).

| RESULTS
In the year 2012, an epidemiologic study stated that higher risk consumers represented 7.0‰ inhabitants aged from 15 to 64 years old and that the average potency of cannabis used in Portugal increased in the last years of our study (SICAD, 2014). The increasing rate of PD hospitalizations associated with CU per 100,000 inhabitants accompanied the rate previously described in those reports.
Higher consumer patterns and higher cannabis potency may be one of the reasons that could explain the increasing number of hospitalizations we found in our study (SICAD, 2017).
The mean LoS was 19.42 days, a long LoS that may refer to the need for inpatient dishabituation treatment, the treatment of organic comorbidities affected in the psychotic event, or even the possible difficulties in social reinsertion of the patient after the hospitalization.
This long LoS may also explain the high mean charges associated with PD and CU hospitalizations.

| Limitations
One of the limitations of our study is that we used an administrative database that was not previously conceived for the purpose of this

| CONCLUSIONS
The use of big data is currently making its way in the field of psychiatry and mental health, and studies like ours reinforce the importance of increasing the detailing of clinical coding in administrative clinical databases. Large databases allow researchers to have a more representative sample of the studied population even though the data quality may vary depending on the health records from where the database is created (Lejoyeux et al., 2014;Simon, 2019). Although previous studies demonstrated the possible effect that cannabis may exert on brain functionality, the purpose of this study was not to prove causality but to demonstrate and describe trends regarding PD and CU. Even if one cannot establish a causal effect by analysing raw administrative data, we can conclude that the number of hospitalizations due to PD and associated with CU increased tremendously during the study period. This shift may be due to the increasing consumption of cannabis and/or the increasing codification of secondary diagnosis in administrative hospital databases. It would be interesting to know similar trends in other countries with the same or different cannabis consumption patterns.