This retrospective cohort study was based on the MIMIC-IV database (version 2.2) [16], which contains data from patients admitted to the ICU of the Beth Israel Deaconess Medical Centre in Boston, Massachusetts, between 2008 and 2019. One author (JLX) obtained access to the database and was responsible for data extraction. The establishment of the MIMIC-IV database was approved by the institutional review boards of Beth Israel Deaconess Medical Centre and the Massachusetts Institute of Technology Affiliates. The informed consent requirement was waived because the data from all patients in the database were anonymized.

We included all patients who were first admitted to the ICU diagnosed with COPD [17] in the MIMIC-IV database. We excluded (i) patients with dementia; (ii) patients younger than 18 years; (iii) patients without a CAM-ICU estimate; (iiii) patients with a duration of stay in the ICU stay < 2 days; (v) patients with mild cognitive impairment (MCI). Navicat Premium (version 16.0) was used to extract raw data relative to patients diagnosed with COPD from the MIMIC-IV database (version 2.2). The data extracted included demographics, laboratory tests results, vital signs, comorbidities, and administered drugs. The following demographic information was extracted: age, sex, length of hospitalization, and duration of stay in the ICU. Vital signs such as systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate (HR), respiratory rate (RR), and oxygenated hemoglobin saturation (SpO2) were collected. Comorbidities including diabetes, asthma, coronary artery disease (CAD), peripheral vascular disease (PVD), congestive heart failure (CHF), malignant cancer, cerebrovascular disease (CVD), liver disease, and renal disease were extracted. Laboratory data included white blood cell (WBC) count, hemoglobin (HGB) level, platelets (PLT), blood glucose, anion gap, potassium, sodium, and calcium levels. We also extracted whether the patient received mechanical ventilation. The use of vasoactive drugs (norepinephrine, vasopressin, epinephrine) and antibiotics was also included. Whether the patient received oral or intravenous glucocorticoids was also determined. Whether the patient received ACEI/ARB, β-blockers, and statins 3 days after entering the ICU were extracted. A Simplified Acute Physiology Score (SAPS II), Charlson Comorbidity Index (CCI), and the Oxford Acute Severity of Illness Score (OASIS), which represents the severity of the disease, were also included.

We defined patients with records of 3 days of statin use before or after admission to the ICU as statin-exposed and others as non-statin-exposed. We searched drug ILIKE “statin” and NOT ILIKE “nystatin”, “mycostatin”, “imipenem-cilastatin”, “pentostatin”, and “sandostatin” in Navicat Premium (version 16.0). The medication prescriptions were recorded in the MIMIC-IV table (version 2.2) under “mimic-hospital, prescription”.

The primary outcome was the occurrence of delirium during the stay in the ICU. The secondary outcome was the 30-day mortality rate. The Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) method was used to assess delirium in patients [18].

The characteristics of the patients are described in general and by group (statin-exposed and non-statin-exposed). The measured data are expressed as mean (standard deviation) or median (interquartile interval) according to whether they were normally distributed. A one-way analysis of variance (ANOVA) or Kruskal–Wallis H test was performed depending on whether the data were normally distributed. Categorical variables were expressed as percentages and treated with Chi-square tests.

Propensity score matching (PSM) was used to adjust for confounders between the non-statin and statin groups. The following prognostic variables related to the outcome at p < 0.2 in the univariate analysis (Additional file 1: Table S1) were included in the propensity score: age, HGB, WBC, sodium and calcium, CAD, asthma, liver disease, malignant cancer, SBP, DBP, HR, and SpO2 on the first day of admission, ACEI/ARB, antibiotic and glucocorticoids after admission to the ICU, and norepinephrine, vasopressin, epinephrine, SAPII, OASIS score and length of ICU stay for delirium were forced in the PSM. The variables that were included in the 30-day mortality analysis are shown in Additional file 1: Table S2. Using the estimated propensity scores as weights, an inverse probability weighting (IPW) model was used to generate a weighted cohort [19]. The probability of being exposed to statin was estimated using a logistic regression for each patient. We matched the two groups in a 1:1 ratio with a caliper width of 0.2. The standardized mean difference (SMD) was used to examine the degree of PSM. The R software packages (http://www.R-project.org, The R Foundation) and Free Statistics software versions 1.7 were used to perform all statistical analyses. Statistical differences were considered significant at p < 0.05.

Patients were re-grouped according to whether they were on statins or not before ICU admission. There were three groups, no statin users, statin use after ICU admission and statin use both before and after ICU admission. There were no patients who were statin users before ICU and not statin users after ICU admission. Multivariate logistic regression analysis was used to ascertain the relationship between statin use and the incidence of delirium in patients with COPD adjusting covariates as for PSM. We did a subgroup analysis in logistic regression to investigate the association between statin use and delirium and 30-day mortality, as it differed between various subgroups classified by age, sex, CAD, CHF, antibiotic and glucocorticoids use, CCI, SAPII, OASIS score and length of ICU stay.