Table of Contents
Study design
This study employed a historical cohort design to assess the risk of mortality, ICU admission, and need for intubation among hospitalized COVID-19 patients.
Study setting
The investigation was carried out in the Alborz province, Iran, during the early months of the COVID-19 pandemic. The Alborz University of Medical Sciences Local Ethics Committee approved the study protocol under the IR code. COVID-19 confirmation was made through a positive polymerase chain reaction (PCR) result. Opium use disorder was defined as a compulsive urge to use opioids or opioid-derivative drugs. In this study, the recent definition was considered, which involved inhaling the smoke from burning dark crude material obtained from Papaver somniferum.
Participants
Participants aged 18 years and above with confirmed COVID-19 were included in the study, as opium use disorder is rare among younger age groups. To form the control group, non-opioid abusers who contracted COVID-19 were included. These individuals were selected based on the absence of opioid use disorder (OUD), and they were matched with the opium-abusing group by age, gender, and other relevant demographic factors. In the current study “Opium” use is considered as a substance use disorder (SUD) which is defined as a mental disorder that has influence on patient’s brain and behavior, results in inability to control use of substances [15].
Patients who were transferred to other hospitals during the course of their treatment and individuals with missing data for any of the studied variables were excluded from the study. The exclusion of patients transferred to other hospitals was necessary because we could not follow their outcomes, potentially introducing variability. Excluding patients with missing data was essential to maintain dataset completeness and minimize potential bias resulting from incomplete information. It is important to note that the excluded patients were not substantially different from those who remained in the study in terms of the variables considered, thereby minimizing the risk of sampling bias.
Variable
The study focused on three primary outcomes: mortality, ICU admission, and mechanical ventilation. The researchers contacted the families of deceased patients to verify the recorded outcomes and complete any missing information on the data collection forms. Additionally, if individuals were discharged from the hospital and gave consent, telephone follow-ups were conducted one or two weeks after discharge to gather further information. Secondary outcomes included the severity of CT scan involvement and O2 saturation.
Trained nurses conducted interviews to gather data on demographic characteristics, medical history, medication use, and smoking habits [16]. Venous blood samples were collected from participants after an overnight fast of 12–14 h, between 07:00 and 09:00 a.m. Centrifugation was performed within 30–45 min of sample collection. The serum total cholesterol (TC) level was measured using the enzymatic colorimetric method with cholesterol esterase and cholesterol oxidase. Other metabolic measurements such as high-density lipoprotein cholesterol (HDL-C) and triglyceride (TG) levels were determined using previously reported methods [17].
Participants were categorized as current smokers (those who were currently smoking) or non-smokers (those who had never smoked or had quit smoking). Regular opium use was identified if a participant had a history of chronic OUD (through respiratory or gastrointestinal routes) for at least one year and experienced withdrawal symptoms upon cessation of opium use. Weight measurements were taken with minimal clothing and without shoes, using digital scales (Seca 707; range 0.1–150 kg), and recorded to the nearest 100 g. Height was measured while participants were standing without shoes, with their shoulders in a normal alignment. Body mass index (BMI) was calculated by dividing the weight in kilograms by the square of the height in meters.
Type 2 diabetes (T2D) was defined as a fasting plasma glucose (FPG) level of ≥ 7.0 mmol/L and/or a 2-h post-challenge plasma glucose (PCPG) level of ≥ 11.1 mmol/L, or current use of anti-diabetic medications [18]. Fever was characterized as an axillary temperature of at least 37.3°C. Secondary infection was diagnosed when patients exhibited clinical symptoms or signs of pneumonia or bacteremia, and a new pathogen was identified through positive culture from lower respiratory tract specimens (qualified sputum, endotracheal aspirate, or bronchoalveolar lavage fluid) or blood samples after admission. Acute kidney injury was diagnosed according to the Kidney Disease: Improving Global Outcomes (KDIGO) clinical practice guidelines [19], while acute respiratory distress syndrome (ARDS) was diagnosed based on the Berlin Definition [20].
In this study, coronary heart disease (CHD) events were considered as outcomes and included cases of definite myocardial infarction (MI) diagnosed by electrocardiogram (ECG) and biomarkers, probable MI (positive ECG findings plus cardiac symptoms or signs, but biomarkers showing negative or equivocal results), unstable angina pectoris (new cardiac symptoms or changing symptom patterns with positive ECG findings and normal biomarkers), angiography-proven CHD, and CHD-related deaths (any death due to CHD based on the aforementioned criteria that occurred in the hospital or sudden cardiac death from cardiac disease within 1 h of symptom onset based on verbal autopsy documents outside the hospital).
Data sources/measurement
To gather patients' demographic and clinical data, blood biochemical factors, and information related to COVID-19 signs, symptoms, treatment, and outcomes, a data collection form was developed and utilized. To ensure consistent distribution and completion of the forms, a series of three two-hour training sessions were conducted for 32 nurses from various hospitals in Alborz province, Iran, encompassing a total of 17 hospitals. Subsequently, the data collection teams within each hospital submitted the completed forms to the leading researchers of this study through an integrated electronic inter-sectoral system.
Data collection occurred during the admission phase, treatment period, and when the outcome (death or recovery) was observed. To maintain the quality and accuracy of the recorded information, continuous quality control of the completed questionnaires was implemented. As a first step, a subset of randomly selected questionnaires was assessed for item accuracy through telephone interviews with the responsible nurses. Throughout the data collection process, five individuals independently entered and coded the questionnaire responses using Excel software. Discrepancies were resolved by reviewing the questionnaire, contacting the study subjects, or excluding them from the study if necessary.
Bias
To mitigate potential sources of bias in this study, several measures were implemented. First, the use of propensity score matching (PSM) was employed to balance the treatment and control groups, minimizing the impact of confounding variables and selection bias. This technique allowed for a more rigorous comparison of outcomes. Additionally, data collection quality was controlled continuously through a series of training sessions for 32 nurses from various hospitals. These training sessions aimed to ensure uniform data collection across multiple healthcare institutions and minimize variability in data collection. Furthermore, to address potential information bias, a subset of randomly selected questionnaires was reviewed for item accuracy through telephone interviews with the responsible nurses. Any discrepancies were resolved through rigorous data validation procedures, enhancing the accuracy and reliability of the recorded information.
Study size
The study included a total of 442 participants. Given the retrospective cohort design, all eligible individuals meeting the inclusion criteria were incorporated into the study, and no specific sampling was conducted.
Statistical analysis
Continuous baseline demographic and clinical data are presented as mean ± standard deviation and grouped data as frequencies and percentages. The comparison of means between the discharged and deceased groups, for quantitative variables with approximately normal distribution, was performed using independent t-tests. The distribution of non-normal quantitative data between the two groups was assessed using the exact Mann–Whitney U test. Chi-square test or Fisher’s exact test were used to determine the independence of the two categorical variables. To assess the treatment effect on the outcome variables, the PSM method was employed. Initially, potential confounding variables were identified. Logistic regression was then employed to determine the variables associated with the outcome or treatment, which were subsequently included in the model. For the ICU outcome, covariates included Age, Sex, Smoking, Alcohol, fever, Cancer, Ghalyan history, ARDS, BP, PLT, NUT, BUN, LDH, TG, Ventilator need, Time ventilation, Posterior Lower Lobe predication, Inpatients duration, ceftazidime, clindamycin, and vancomycin. For the mortality outcome, covariates included Smoking, Temperature, Age, Sex, Alcohol, ceftazidime, Ghalyan history, Ventilator need, INR, Time ventilation, PLT, BP, Cancer, vancomycin, Impatient, LDL, CRP, AST, FBS, Air bronchogram, BUN, LYMP, NUT, Bad breath, and linezolid. For the need to ventilation outcome, covariates included Age, Sex, Smoking, Alcohol, first sign time, Nasua, Cancer, Ghalyan history, ICU, ARDS, BP, GCS, PLT, LYMP, BUN, LDH, TG, AST, Ventilator need, Time ventilation, Posterior Lower Lobe predication, Inpatients duration, ceftazidime, amikacin, clindamycin, vancomycin, linezolid, and oseltamivir. PSM was conducted using the nearest neighbor algorithm, matching each individual in the treatment group with three individuals from the control group, resulting in a 1:3 matching ratio [21]. Various methods, including the absolute standardized mean difference (SMD) [22], variance ratio [23], and Kolmogorov–Smirnov (KS) statistic [24], were used to assess balance and achieve an optimal matched set that would provide a reliable estimation of the treatment effect after PSM. A good balance between the two groups was considered when the absolute SMD approached zero, the variance ratio approached one, and the KS statistic approached zero.
To estimate the hazard ratio (HR) and its standard error, a weighted proportional hazards Cox model with robust standard errors and clustering by subgroup was utilized. The weighted data were subsequently used to construct the Kaplan–Meier plot. Furthermore, for the analysis of treatment effect, both the full adjustment method, which involved adjusting for all variables in a logistic regression model alongside the treatment effect, and the optimal model selection method, utilizing the stepwise backward algorithm, were employed.
All statistical analyses were performed using the R software. Statistical significance was determined based on a significance level of p < 0.05.
