MIMIC-III database (version 1.4) is open to the public and contains information on over 50,000 patients hospitalized to the Beth Israel Deaconess Medical Center's intensive care unit (ICU) between 2001 and 2012.42 This database can be accessed at https://mimic.physionet.org/. Enqian Liu obtained access to the database (No. 35919439). The use of the data was approved by the Beth Israel Deaconess Medical Center (Boston, MA) and Institutional Review Boards of the Massachusetts Institute of Technology (Cambridge, MA).43 The requirement for written informed consent was waived, because each patient information in the database was anonymized and de-identified.43 The ethics committee of Lishui Municipal Central Hospital approved this study (no. 2021184).

Data from each patient's initial ICU admission were analyzed. Patients with COPD were selected using International Classification of Diseases, Ninth Revision (ICD-9) codes (491.20, 491.21, 491.22, and 496). We excluded patients aged <18 years, those admitted to the ICU for <24 h, those with survival times <0, and those with missing RDW data.

Data were extracted using PostgreSQL (version 9.6) and a structured query language. The following variable data were extracted: age, sex, ethnicity, weight, heart rate, systolic blood pressure (SBP), diastolic blood pressure (DBP), temperature, respiratory rate, percutaneous oxygen saturation (SpO2), mean blood pressure (MBP), vasopressor use, renal replacement therapy (RRT), congestive heart failure, hypertension, diabetes, cardiac arrhythmias, sepsis, liver disease, acute kidney injury (AKI), renal failure, serum sodium, serum creatinine, serum potassium, blood urea nitrogen (BUN), serum hemoglobin, serum bicarbonate, serum glucose, serum hematocrit, anion gap, serum chloride, platelet, and white blood cell (WBC), Sequential Organ Failure Assessment (SOFA) score, and Simplified Acute Physiology Score II (SAPS II). The baseline data were obtained within the first 24 h after ICU admission. The initial value was considered for a variable that was measured multiple times within 24 h after ICU admission. Comorbidities, except AKI, were diagnosed according to ICD-9 codes. Severe sepsis was defined as the presence of either (a) a combination of ICD-9 codes for infection and one or more organ dysfunctions,44 or (b) the ICD-9 code for severe sepsis (995.92) or septic shock (785.52).45 According to the Kidney Disease: Improving Global Outcomes (KDIGO) Clinical Practice Guidelines, AKI was induced within the first 24 h of ICU admission and was identified based on serum creatinine and urine output.46 Anemia was diagnosed as a hemoglobin level <12 g/dL in women and <13 g/dL in men according to the World Health Organization guidelines.47 The clinical outcome was the 28-day all-cause mortality after ICU admission.

Participants were divided into tertiles based on RDW levels. Continuous data were expressed as mean (standard deviation) or median (interquartile range), while categorical variables were presented as percentages. The baseline characteristics of the different RDW tertile groups were analyzed using chi-square test for categorical variables, one-way analysis of variance test for normally distributed data, and Kruskal-Wallis H test for non-normally distributed data. The relationship between RDW and 28-day all-cause mortality was determined using multivariate logistic regression analysis. Values of the variation inflation factor (VIF) were used to assess multicollinearity. More than 10 VIFs showed multicollinearity. We constructed three models, namely, model 1 that was unadjusted, model 2 that was adjusted for age, sex, and ethnicity, and model 3 that was adjusted for age, sex, ethnicity, and other variables with P < 0.1 in univariate analysis or those with >10% change in effect estimates (weight, SBP, MBP, heart rate, respiratory rate, SpO2, hematocrit level, platelet level, anion gap, creatinine level, bicarbonate, chloride, glucose, BUN, WBC, and potassium levels, SAPS II, SOFA, cardiac arrhythmias, hypertension, sepsis, liver disease, vasopressor use, ventilation, RRT, AKI, and anemia). To investigate the non-linearity further, we turned RDW into a categorical variable based on the tertiles, and then into a continuous variable by entering the tertiles’ median values into the variable. For the sensitivity analysis, interaction and subgroup analyses were based on age (<75 and ≥75 years), sex, ethnicity, congestive heart failure, diabetes, renal failure, hypertension, cardiac arrhythmias, liver disease, sepsis, anemia, AKI, RRT, vasopressor use, ventilation, SAPS II (<37 and ≥37), and SOFA score (<4 and ≥4). Missing values were not found in the category variables. Missing values for continuous variables were imputed using the mean (normal data) or median (non-normal data). The statistical significance was considered at p < 0.05. Data analysis was performed using the R statistical software package (version 4.1.1).

A total of 3,244 patients in the MIMIC-III database satisfied the inclusion criteria (Fig. 1). The baseline characteristics of the RDW tertile groupings are shown in Table 1. The mean patient age was 72.3 ± 11.3 years, and approximately 53.6% patients were men. RDW values at baseline varied from 11.50% to 28.20% (median 14.70%; mean 15.18%). No significant differences were observed between the different groups concerning sex, heart rate, platelet level, serum chloride, glucose, and serum sodium levels, hypertension, and vasopressor use (all P > 0.05). Patients in the highest RDW tertile group were likely to develop congestive heart failure, cardiac arrhythmias, diabetes, real failure, liver disease, sepsis, anemia, AKI, and RRT, and were less likely to require ventilation than patients in the lowest group. As the RDW increased, respiratory rate, weight, anion gap, potassium, creatinine, and BUN levels, SOFA score, and SAPS II increased, whereas SBP, DBP, MBP, temperature, SpO2, and hematocrit, hemoglobin, WBC, and bicarbonate levels decreased.

Fig. 1.

Flow diagram of patient recruitment according to the cohort selection and exclusion criteria. MIMIC-III, multiparameter intelligent monitoring in intensive care III; COPD, chronic obstructive pulmonary disease; RDW, red blood cell distribution width

(0.16MB).

The independent effects of RDW on 28-day all-cause mortality in critically ill COPD patients were evaluated by constructing three different logistic regression models.

Logistic regression analysis for 28-day mortality (Table 2) show that RDW was positively related to the risk of 28-day all-cause mortality (unadjusted odds ratio [OR] 1.20; 95% CI 1.15−1.25). The crude ORs were 1.24 (95% CI 0.98−1.58) and 2.36 (95% CI 1.89−2.96) in the second and third tertile groups of RDW, respectively, with the first RDW tertile group as the reference. After adjusting for age, sex, and ethnicity, higher RDW values were correlated to higher risks of 28-day mortality (OR 1.21; 95% CI 1.16−1.27). In comparison with the first tertile group, the ORs were 1.18 (95% CI 0.93−1.51) and 2.37 (95% CI 1.89−2.98) in the second and third tertile groups, respectively. RDW was strongly correlated with 28-day all-cause mortality (OR 1.15; 95% CI 1.09−1.21) in Model 3. Furthermore, a higher RDW value was related to a greater risk of 28-day all-cause mortality in the second RDW tertile group (OR 1.03; 95% CI 0.78−1.34) and the third RDW tertile group (OR 1.70; 95% CI 1.29−2.22) after adjusting for age, sex, ethnicity, weight, SBP, MBP, heart rate, respiratory rate, SpO2, hematocrit level, platelet level, anion gap; creatinine, bicarbonate, chloride, glucose, BUN, WBC, and potassium levels, SAPS II, SOFA score, cardiac arrhythmias, hypertension, sepsis, liver disease, vasopressor use, ventilation, RRT, AKI, and anemia. The linear trend tests for 28-day mortality yielded remarkable results in the three different models.

To evaluate the underlying clinical heterogeneity, we used interaction and stratified analyses (Fig. 2). We assessed the relationship between RDW and 28-day mortality in different subgroups. Interaction and stratified analyses were not detected in terms of age (<75 and ≥75 years), sex, ethnicity, anemia, diabetes, hypertension, cardiac arrhythmias, renal failure, liver disease, congestive heart failure, sepsis, anemia, AKI, RRT, vasopressor use, ventilation, SAPS II (<37 and ≥37), and SOFA score (<4 and ≥4).

Fig. 2.

Effect size of red blood cell distribution width on 28-day mortality in prespecified and exploratory subgroups. The effect size was adjusted for age, sex, ethnicity, weight, systolic blood pressure, mean blood pressure, heart rate, respiratory rate, percutaneous oxygen saturation, hematocrit level, platelet level, anion gap, serum creatinine, bicarbonate, chloride, glucose, blood urea nitrogen, white blood cell, and potassium levels, Simplified Acute Physiology Score II, Sequential Organ Failure Assessment score, cardiac arrhythmias, hypertension, sepsis, liver disease, vasopressor, ventilation, renal replacement therapy, acute kidney injury, and anemia, except for the subgroup variable.

(0.47MB).