Between 2010 and 2018, altogether 36,135 patients were included in SIR as cardio-thoracic ICU patients. After excluding 1,897 individuals dying in the first quarter period after admission to the ICU and 38 individuals receiving methadone or buprenorphine, 34,200 patients made up the final study cohort (Figure 1). The final study cohort is presented in Table 1. Opioid use (with 95% confidence intervals (CI)) for the final cohort (n = 34,200) and for a subset of opioid naïve individuals (not using any opioids during a 12 month-period preceding ICU admission, n = 29,390) are presented in Figure 2a and b.
Mean opioid use increased in the quarter period before admission to critical care. After admission, there was a peak in the first three-month period followed by a decline during the first year. After 12 months, the mean consumption returned to baseline levels (equaling 9-12 months before admission), as shown in the Supplementary Information, Table S2. Characteristics for the patients with and without prior opioid consumption are presented in the Supplementary Information, Table S3. Among opioid-naïve patients, the mean opioid consumption remained unchanged between 3 months after admission up to 24 months of follow up.
Cardiac patients admitted to critical care who subsequently developed persistent opioid usage (n = 4,050) were younger, less likely to be male, had lower level of education and income, and had more psychiatric and somatic disorders as well as more substance dependence at the time of admission to critical care. In addition, more patients had emergency surgery, length of stay in the ICU was longer, 40% of persistent opioid users also used opioids before admission to critical care (Table 2). Also among patients not using opioids prior to critical care, similar differences were seen between patients with and without persistent opioid use (Supplementary Information, Table S4).
In the multivariable logistic regression analysis, determinants associated with higher odds of chronic opioid use included female sex, psychiatric and somatic comorbid conditions, substance dependence, preadmission opioid usage, and critical care stay for 3-7 days. In contrast, high age, high income and education, ICU stay for more than 7 days, and admission year 2016-2018 were all associated with lower odds of persistent opioid usage (Table 3). Patients not using opioids prior to critical care, determinants associated with increased odds of chronic opioid use included female sex, medium level of education, psychiatric and somatic comorbidities, substance abuse, and length of stay in the ICU for 3-7 days. Higher age and admission year 2012-2013 or 2016-2018 were associated with lower odds of chronic opioid use (Supplementary Information, Table S5).
During follow-up of 3 to 8 quarters after admission to critical care, 680 patients passed away, of which 164 were persistent opioid users. In the Cox proportional hazards regression analysis (unadjusted), persistent opioid usage was associated with higher mortality, hazard ratio (HR) of 2.3 (95% CI: 2.0-2.8; P < 0.001). After adjusting for covariates (age, sex, psychiatric and somatic comorbid conditions, substance dependence and critical care length of stay, the association remained significant, HR of 2.2 (95% CI: 1.8-2.6; P < 0.001). Also among opioid naïve patients, increased mortality was associated with persistent opioid usage, adjusted HR of 2.3 (95% CI: 1.8-2.9; P < 0.001).
Sensitivity analysis
Results did not change when accounting for non-random dropout due to death (data not shown).
Missing data
Smaller numbers of missing data were found on income (n = 121, 0.4 %) and education (n = 514, 1.5 %).