Associations between static and dynamic changes of platelet counts and in-hospital mortality in critical patients with acute heart failure

Source of data

All data of this study were obtained from the Medical Information Mart for Intensive Care III (MIMIC-III) database (version 1.4). The MIMIC-III database records medical data on patients in the intensive care unit (ICU) at Beth Israel Deaconess Medical Center between 2001 and 2012, with a collection of 53,423 ICU admissions, including a large amount of physiological data and medical records12. All data are authentic and publicly accessible for free. No informed consent was required from the patients as their basic personal information was anonymized ( Data for the present study was extracted by one author, Tao Liu, passed the training test and obtained permission to download and use the database (ID: 9008147).

study population

We initially recruited 3614 critical patients with the diagnosis of AHF from the MIMIC-III database, of which 607 participants were excluded due to repeated hospitalization and 42 participants were excluded due to death within 3 days of admission. Of the remaining 2,965 participants, 35 participants were excluded due to missing data. Ultimately, 2930 critical patients with AHF were identified for analysis, including 2720 survivors and 210 non-survivors.

Data extraction

The involved data was extracted with PostgreSQL tool. Demographic information: gender, age, ethnicity, and marital status; Vital signs: heart rate (HR), respiratory rate (RR), systolic blood pressure (SBP), diastolic blood pressure (DBP), pulse oxygen saturation (SpO2), and temperature; 24 h urine output; Indicators of laboratory tests: platelet, partial pressure of oxygen (PaO2), partial pressure of carbon dioxide (PaCO2), potential of hydrogen (pH), N-Terminal pro-brain natriuretic peptide (NT-proBNP), white blood cell (WBC), neutrophile (N), lymphocyte (L), C reactive protein (CRP), and estimated glomerular filtration rate (eGFR); Comorbidities: sepsis, stroke, pneumonia, acute myocardial infarction (AMI), atrial fibrillation (AF), liver disease, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), respiratory failure, and cancer; Therapeutic measures: renal replacement therapy (RRT), ventilation, and vasoactive drugs; Scoring scales for ICU disease: the Sequential Organ Failure Assessment (SOFA) Score and the Simplified Acute Physiology Score II (SAPS II). Length of hospital: length of ICU stay and length of hospital stay. All of the above baseline values were the first measured within the first 24-h after admission to the ICU. In addition, the registration information of admission and discharge was recorded, including the length of hospitalization, time of death, and length of stay in the ICU.

Outcomes and main exposure variables

The outcome of this study was in-hospital mortality, which was death from all causes during the patient’s hospitalization. The independent variables were baseline platelet count and dynamically changed platelet count. Dynamically changed platelet counts were platelet counts measured daily during the one week stay in the ICU. The interval time between repeated measurements of platelet counts was irregular.

Statistical analysis

All data were analyzed with R Studio software (Version 4.2.1). Continuous variables that conformed to normal distribution were expressed as mean (standard deviation) and the t-tests were applied; Continuous variables not satisfying normal distribution were shown as median (interquartile spacing) [M (Q1, Q3)] and the Wilcoxon rank-sum tests were carried out; Categorical variables were presented as percentages (%) and the X2 tests were conducted. Sequentially, logistic regression models were constructed to assess the association of baseline platelet count and baseline tertile platelet count with in-hospital mortality in critical patients with AHF, and the results were displayed in the form of odds ratios (ORs) and 95% confidence intervals (CIs). Model 1: unadjusted variables; Model 2: Adjusted for age, gender, ethnicity, and marital status; Model 3: Adjusted for age, ethnicity, HR, RR, SBP, DBP, temperature, PO2, PCO2, pH, 24-h urine output, sepsis, AMI, AF, respiratory failure, RRT, ventilation, vasoactive drugs, SOFA, SASP II, NT-proBNP, WBC, N, L, CRP, eGFR, length of ICU stay, and length of hospital stay. Meanwhile, the relationship between baseline platelet counts and in-hospital mortality was re-analyzed utilizing stepwise regression to avoid multicollinearity among the covariates (Supplementary Table S1). The nonlinear relationship between baseline platelet count levels and in-hospital mortality were explored by adjusting the multivariate restricted cubic spline (RCS) regression model. The inflection point was calculated based on the recursive algorithm, and then the threshold effect analyses were conducted by the segmented logistic regression model (Supplementary Table S3). In addition, we conducted ROC for PLT, eGFR, and NT-proBNP on predicting in-hospital mortality, and we also examine correlation analysis of in-hospital mortality with WBC, CRP, neutrophil, lymphocyte, and PLT.

Differences in dynamical platelet counts between survivors and non-survivors during the first week after ICU admission were compared (Supplementary Table S4). Finally, generalized additive mixed models (GAMM) were constructed to investigate the relationship between dynamic platelet counts in the first week and in-hospital mortality in critical patients with AHF admitted to ICU. GAMM is usually performed to the analyse data obtained repeatedly, especially when the data have missing values or are duplicated irregularly13,14. P-value < 0.05 was considered to be statistically significant.

Ethics statement

Data of the present study was from the MIMIC-III database. The MIMIC III database was approved to build by the Institutional Review Boards of Beth Israel Deaconess Medical Center (Boston, MA) and the Massachusetts Institute of Technology (Cambridge, MA). Data for the present study was extracted by one author, Tao Liu, who has completed the online training course and passed the exam (ID: 9008147).


Leave a Reply

Your email address will not be published. Required fields are marked *