DOI: https://doi.org/10.21203/rs.3.rs-1886026/v1
Compared to normal platelet count (PC), absolute and relative platelet reductions are associated with worse clinical outcomes in septic patients. However, as to whether different variation trends of platelet count are associated with prognosis has not been investigated. The purpose of the research was to examine the effect of platelet count trajectories on prognosis and risk factors of septic patients with different PC trajectories.
Group-based trajectory models (GBTMs), which identify groups of individuals on the basis of similar developmental paths, were used to identify three subgroups of PC trajectories (increasing, low, and medium PC). Hospital mortality and 90-day mortality were the primary outcomes. Multivariable Cox, logistic, linear regression analysis were used to assess relationship between PC trajectories and clinical outcomes and multinomial logistic regression were used to identify risk factors of different trajectories
The proportion of patients with increasing, low, and medium PC was identified in 367 (6.2%), 3173 (53.4%), and 2406 (40.5%) patients. Compared with low PC, Medium PC was associated with lower hospital mortality [odds ratio (OR) 0.84, 95% confidence interval (CI) 0.72–0.99], the following relative risks (RRs) < 1: Male, Vasopressor use (1st 24h), RRT use (1st 24h), without hypertension, without diabetes, and CCU/CVICU
The Low PC trajectory was related to a higher risk of hospital mortality. Male, Vasopressor use (1st 24h), RRT use (1st 24h), without hypertension, without diabetes, and CCU/CVICU patients are risk factors of septic patients with low PC trajectories.
Sepsis is a potentially life-threatening multi-organ dysfunction caused by dysregulation of the body's immune system (1). Due to its high incidence and mortality, sepsis has become a major public health threat globally(2, 3). Since platelets play an important role in the immune response during sepsis/septic shock, thrombocytopenia is a famous risk factor of mortality(4).
Platelet count less than 150\(\times\) 109/L is generally referred to as thrombocytopenia and is the most common complication of intensive care (ICU), present in 50% of cases(5, 6). Nearly 50% of all thrombocytopenia in severely ill patients is due to sepsis (7). Previous investigations have found that both absolute and relative reduction in platelet count is associated with worse clinical outcomes for patients with sepsis(8–10).
However, as to whether different variation trends of platelet count are associated with prognosis has not been investigated. Previous studies have focused on trends in platelet variation between surviving and dead patients. Among ICU patients, a prospective and multicenter study reported biphasic temporal patterns in platelet counts.(11). Li et al. also reported that variation trend of PC between the dead and the non-dead was different among serious patients within first week of ICU admission (12).
Our research was conducted using data from Medical Information Mart for Intensive Care IV1.0 (MIMIC IV 1.0), which was a retrospective cohort study. The purpose of the research was to examine the effect of platelet count trajectories on clinical outcomes and the risk factors of septic patients with different PC trajectories. We checked the assumption that potential PC trajectories existed in septic patients during first week and were related to in-hospital mortality and some Prognostic evaluation indicators
Data from MIMIC IV 1.0 was extracted to conduct this study(13). Presently, MIMIC-IV consists of extensive and high-quality data on patients accepted into ICU at Beth Israel Deaconess Medical Center from 2008 to 2019. The original data were further processed using STATA software after extracted by Navicat software. Significantly, due to lack of out-of-hospital date of death in MIMIC-IV 1.0, MIMIC-IV 2.0 which was released on June 12, 2022 was used to complete the data. The database is free-to-access and anyone who has agreed a protocol for the use of the data and has finished course in the 'protection of human subjects' can apply for it to be accessed. (14). Therefore, we need no patient agreement or moral permission. Data were extracted by author DW (certification number: 44274909).
The inclusive criteria were: (1) adults admitted to hospital first time; (2) patients with sepsis-3 within 48 hours after ICU admission. exclusive criteria were:(1) patients were diagnosed with sepsis-3 by first sofa evaluation after ICU admission; (2) patients stayed in ICU less than 3 days; (3) Patients with less than 3 platelet count time-point data within a week. Ultimately, 5946 patients were selected into the research. (Fig. 1).
The outcomes of the research were hospital mortality, 90-days mortality, proportion of patients with WBC drop below 10×109/L, RBC transfusion, the time WBC drop below 10×109/L, maximum SOFA score, length of hospital stay and ICU stay. Patients were grouped into Low PC, Medium PC and Increasing PC according to their PC trajectories.
Many covariates were picked up from the MIMIC IV: Population features (sex, age, and weight); Laboratory tests (Hb, WBC and INR); SOFA score; Carlson comorbidity index; source of admission; treatments after admission to ICU (MV, RRT, platelet transfusion and drug therapy [antibiotics, anti-coagulants, anti-platelet drugs and vasopressor]); basic diseases diagnosed before the ICU admission (ICD-9 or ICD-10), covering diabetes mellitus, hypertension, coronary heart disease; bacterial culture results.
The definitions are as follow: vasopressor (dopamine, dobutamine, norepinephrine, adrenaline); Cephalosporins (cefazolin, ceftazidime, ceftriaxone, cefepime); G+ (Streptococcus pneumoniae; Viridans streptococci; Staph aureus coag+; Staphylococcus epidermidis; staphylococcus, coagulase negative; Enterococcus SP; Enterococcus faecium; Enterococcus faecalis); G- (Klebsiella pneumoniae; Pseudomonas aeruginosa; Escherichia coli; Klebsiella Oxytoca; Stenotrophomonas maltophilia; Serratia marcescens; Proteus mirabilis; Citrobacter koseri; Morganella morganii); fungus (Candida albicans; Yeast).
The GBTM was adopted to recognize potential sub-groups of septic patients who followed comparable PC change in the minimum daily PC measurement within first week of ICU admission. GBTM is a specific use of Finite Mixture Model, aimed at identifying groups that follow an analogous pattern of development.(15). The approach supposes that the human is heterogeneous, consisting of a limited number of different groups. Stata’s traj plug-in was used to conduct GBTM, estimating PC trajectories (16). The criteria for optimal model are the presence of some of the following criteria: BIC Closest to 0, Group average posterior probability > 0.70, Odds ratio for correct classification > 5, subgroup proportion > 5% and the estimated subgroup proportion is consistent with the actual subgroup proportion(17).
The supplement file Table S1 showed that the model with group number 3 fitted best and all classification accuracy parameters were at the desired level, so it was chosen as the best model.
In Fig. 2. Trajectory 1, defined as “increasing PC”, demonstrated a pattern that the patients began with a high PC (> 400\(\times\) 109/L) and then increased quickly over the next week. Trajectory 2, defined as “low PC”, demonstrated a pattern that the patients began with a low PC (< 150\(\times\) 109/L), dropped to a minimum over the next 2 days and then stayed up until day 7. Trajectory 3, defined as “medium PC”, demonstrated a pattern that the patients began with a medium PC (> 200 × 109/L), dropped to a minimum over the next day and then stayed up until day 7.
Continuous variables were presented as mean and standard deviation (SD) or median and interquartile range (IQR) as appropriate. Differences among PC groups were compared by one-way ANOVA or Kruskal-Wallis test for continuous variables, and categorical variables were presented as percentages and compared using chi-square test. Bonferroni post hoc test was chosen for multi-comparison.
Univariate and multi-variate Cox, logistic and linear regression analysis were used to evaluate the correlation between latent trajectory groups and outcome indicators (primary and secondary). Covariates were removed if p < 0.05 in univariate regression. Variance inflation factors (VIFs) were calculated to check for multi-collinearity in covariates; VIF values above 5 were usually considered to be multi-collinear and were excluded from model. Hazard ratios (HRs), odds ratios (ORs) or regression coefficient along with the 95% confidence intervals (CIs) are reported in addition. Multinomial logistic regression was applied to determine risk factors between latent trajectory groups by relative risk (RR). For sensitivity analysis, we eliminated patients who underwent platelet transfusion within the first week of ICU admission.
All statistical analyses were conducted with Stata 15.1 software. p-values < 0.05 were regarded as indicating statistical significance.
The baseline features of the research cohort based on the three PC trajectories are presented on Table 1. Septic patients with low PC were more probably elderly, triple infection (G+, G- and fungus), to use cephalosporins, to be CCU/CVICU patients and have coronary but less probably to have diabetes; median of Carlson comorbidity index was 6(4–8); male accounted for 61%; vasopressor use (1st 24 h) accounted for 44.9%; renal replacement likely therapy accounted for 3%; platelet transfusion (7 days) accounted for 16.1%; aspirin use accounted for 44.4%; heparin sodium use accounted for 62.9%.
| Characteristics | All(n = 5946) | platelet count trajectory patterns | P | ||
|---|---|---|---|---|---|
| Increasing PC (n = 367) | Low PC (n = 3173) | Medium PC (n = 2406) | |||
| Age(years) (Mean, SD%) | 65.3(17.1) | 63.4(17.0) | 65.7(16.9) a | 64.9(17.4) | 0.021 |
| Male, n (%) | 3338(56.1) | 165(45.0) | 1934(61.0) | 1239(51.5) a | < 0.001 |
| Weight (kg) (median, IQR) | 80.0(66.7–95.0) | 76.0(61.5–91.4) | 79.8(67.0–94.0) b | 80.1(67.0-96.7) | < 0.001 |
| Laboratory tests (median, IQR) | |||||
| Hemoglobin (× 1012/L) | 10.7(9.2–12.3) | 9.7(8.5–10.9) | 10.6(9.0-12.4) | 10.9(9.6–12.5) | < 0.001 |
| WBC (× 109/L) | 11.2(7.9–15.6) | 14.1(10.6–19.7) | 10.2(7.1–14.8) | 12.0(8.8–16.0) | < 0.001 |
| INR | 1.3(1.1–1.5) | 1.3(1.2–1.5) | 1.3(1.1–1.6) a | 1.2(1.1–1.5) | < 0.001 |
| SOFA (median, IQR) | 1(0–1) | 0(0–1) | 1(0–2) | 0(0–1) a | < 0.001 |
| Carlson comorbidity index (median, IQR) | 6(4–8) | 5(3–7) | 6(4–8) | 5(3–8) a | < 0.001 |
| Source of admission, n (%) | < 0.001 | ||||
| SICU/Trauma SICU/Neuro SICU | 1981(33.3) | 121(33.0) | 937(29.5) | 923(38.4) | |
| MICU | 1340(22.5) | 95(25.9) | 731(23.0) | 514(21.4) | |
| CCU/CVICU | 1467(24.7) | 47(12.8) | 927(29.2) | 493(20.5) | |
| MICU/SICU | 1029(17.3) | 101(27.5) | 524(16.5) | 404(16.8) | |
| Neuro Intermediate/ Neuro Stepdown | 129(2.2) | 3(0.8) | 54(1.7) | 72(3.0) | |
| Treatments, n (%) | |||||
| MV use (1st 24 h) | 4494(75.6) | 278(75.7) | 2392(75.4) | 1824(75.8) | 0.933 |
| RRT use (1st 24 h) | 117(2.0) | 1(0.3) | 95(3.0) | 21(0.9) a | < 0.001 |
| Vasopressor use (1st 24 h) | 2289(38.5) | 109(29.7) | 1426(44.9) | 754(31.3) a | < 0.001 |
| Platelet transfusion (7 days) | 572(9.6) | 2(0.5) | 512(16.1) | 58(2.4) a | < 0.001 |
| Aspirin | 2492(41.9) | 119(32.4) | 1410(44.4) | 963(40.0) | < 0.001 |
| Clopidogrel | 534(9.0) | 22(6.0) | 302(9.5) | 210(8.7) | 0.07 |
| Warfarin | 260(4.4) | 17(4.6) | 150(4.7) | 93(3.9) | 0.287 |
| Heparin sodium | 4007(67.4) | 292(79.6) | 1997(62.9) | 1718(71.4) | < 0.001 |
| Comorbidities, n (%) | |||||
| Hypertension | 1762(29.6) | 134(36.5) | 887(28.0) b | 741(30.8) a | 0.001 |
| coronary | 1610(27.1) | 73(19.9) | 934(29.4) | 603(25.1) a | < 0.001 |
| Diabetes | 1083(18.2) | 77(21.0) | 501(15.8) | 505(21.0) a | < 0.001 |
| Bacterial culture positive, n (%) | 2853(48.0) | 195(53.1) | 1500(47.2) | 1158(48.1) | 0.102 |
| Type of infection, n (%) | |||||
| G+ | 1443(24.3) | 99(27.0) | 756(23.8) | 588(24.4) | 0.399 |
| G- | 1140(19.2) | 69(18.8) | 648(20.4) a | 423(17.6) a | 0.028 |
| Fungus | 1445(24.3) | 114(31.1) | 760(24.0) b | 571(23.7) | 0.008 |
| Multi-infection, n (%) | |||||
| G- and G+ | 244(4.1) | 20(5.4) | 131(4.1) | 93(3.9) | 0.360 |
| G- and fungus | 226(3.8) | 15(4.1) | 134(4.2) | 77(3.2) | 0.135 |
| G + and fungus | 325(5.5) | 26(7.1) | 163(5.1) | 136(5.7) | 0.261 |
| G + and G- and fungus | 190(3.2) | 13(3.5) | 118(3.7) a | 59(2.5) a | 0.027 |
| Antibiotic, n (%) | |||||
| Cephalosporins | 4033(67.8) | 199(54.2) | 2284(72.0) | 1550(64.4) | < 0.001 |
| Piperacillin-Tazobactam | 1654(27.8) | 130(35.4) | 872(27.5) b | 652(27.1) | 0.003 |
| Meropenem | 543(9.1) | 48(13.1) | 307(9.7) a | 188(7.8) | 0.001 |
| Levofloxacin | 746(12.5) | 62(16.9) | 337(10.6) | 347(14.4) a | < 0.001 |
| Ciprofloxacin | 692(11.6) | 46(12.5) | 373(11.8) | 273(11.3) | 0.768 |
| Metronidazole | 1419(23.9) | 123(33.5) | 774(24.4) b | 522(21.7) | < 0.001 |
| Clindamycin | 199(3.3) | 13(3.5) | 97(3.1) | 89(3.7) | 0.409 |
| SD standard deviation, IQR interquartile range, WBC white blood cell, INR international normalized ratio, SOFA Sequential Organ Failure Assessment, MICU Medical Intensive Care Unit, SICU Surgical Intensive Care Unit, CCU Coronary Care Unit, CVICU Cardiac Vascular Intensive Care Unit, MV mechanical ventilation, RRT renal replacement therapy, G- gram negative, G + gram positive. a Compared with Increasing PC, P > 0.017, b Compared with medium PC, P > 0.017. | |||||
Septic patients with increasing PC were more probably younger, fungal infection, to use piperacillin-tazobactam, meropenem, levofloxacin, metronidazole, to be mixed ICU (MICU/SICU) patients and have hypertension but less probably to have coronary; median of Carlson comorbidity index was 5(3–7); male accounted for 45%; vasopressor use (1st 24 h) accounted for 29.7%; renal replacement therapy accounted for 0.3%; platelet transfusion (7 days) accounted for 0.5%; aspirin use accounted for 32.4%; heparin sodium use accounted for 79.6%.
The Kaplan-Meier survival estimates of three groups in 90-days mortality was shown in Fig. 3. The clinical outcomes were presented in Table 2. In comparison to low PC, increasing PC and medium PC had lower hospital mortality (14.7% vs. 14.3% vs. 20.1%), RBC transfusion (14.7% vs. 14.3% vs. 20.1%) and maximum SOFA score (5 vs. 6 vs. 8), medium PC had lower 90-days mortality (24.6% vs. 29.3%). However, the low PC had a higher infection control rate and less time to infection control than increasing PC and medium PC (87.7% vs. 63.2% vs. 78.4%) and median (0.5 vs. 2.8 vs. 1.2). The increasing PC had lower AKI incidence than the low PC and medium PC (81.2% vs. 86.5% vs. 83.4%).
| Characteristics | All(n = 5946) | platelet count trajectory patterns | P | ||
|---|---|---|---|---|---|
| Increasing PC (n = 367) | Low PC (n = 3173) | Medium PC (n = 2406) | |||
| Clinical outcomes, n (%) | |||||
| Hospital mortality | 1035(17.4) | 54(14.7) | 638(20.1) c, d | 343(14.3) | < 0.001 |
| 90-days mortality | 1617(27.2) | 95(25.9) | 931(29.3) d | 591(24.6) | < 0.001 |
| WBC drop below 10×109/L | 4902(82.4) | 232(63.2) | 2784(87.7) | 1886(78.4) | < 0.001 |
| RBC transfusion | 2037(63.2) | 96(26.2) | 1402(44.1) c, d | 539(22.4) | < 0.001 |
| AKI | 5048(84.9) | 298(81.2) | 2744(86.5) c | 2006(83.4) c | < 0.001 |
| WBC drop below 10×109/L(days) (median, IQR) | 0.7(0.1–2.9) | 2.8(0.6-7.0) | 0.5(0.1–2.2) | 1.2(0.2–3.5) | < 0.001 |
| Maximum SOFA score (median, IQR) | 7(5–9) | 5(4–7) | 8(5–11) | 6(4–8) | < 0.001 |
| Length of stay (days) (median, IQR) | |||||
| Hospital* | 11.4(7.6–18.1) | 12.8(8.7–19.1) | 11.4(7.4–18.7) | 11.3(7.7–17.4) | 0.864 |
| ICU* | 5.3(3.9–8.6) | 5.5(4.1–8.3) | 5.3(3.9–8.8) | 5.3(4.0-8.3) | 0.814 |
| AKI acute kidney injury, c Compared with Increasing PC, P < 0.017, d Compared with medium PC, P < 0.017, * For survivors. | |||||
Table 3 and Table 4 demonstrated correlation between PC trajectories and clinical outcomes of patients with sepsis. After controlling confounders, we observed that medium PC was highly related to a lower hospital mortality in comparison to low PC. (OR 0.84, 95% CI 0.72–0.99), RBC transfusion (OR 0.45, 95% CI 0.39–0.52) and maximum SOFA score (Regression coefficient − 1.3, 95% CI -1.45~-1.18), and increasing PC was similar with medium PC for the above. However, medium PC was significantly associated with a lower infection control rate and more time to infection control (OR 0.52, 95% CI 0.44–0.61) and (Regression coefficient 0.64, 95% CI 0.4 ~ 0.9). No significant association was observed between increasing PC and outcomes of hospital mortality, 90-days mortality and AKI.
| Clinical outcome | Events [n (%)] | Adjusted HR/OR (95% CI) | P |
|---|---|---|---|
| Hospital mortality | |||
| Low PC | 638(20.1) | 1.00 | |
| Medium PC | 343(14.3) | 0.84(0.72–0.99) | 0.039 |
| Increasing PC | 54(14.7) | 0.86(0.62–1.20) | 0.386 |
| 90-days mortality | |||
| Low PC | 931(29.3) | 1.00 | |
| Medium PC | 591(24.6) | 0.96(0.86–1.07) | 0.425 |
| Increasing PC | 95(25.9) | 0.93(0.75–1.16) | 0.532 |
| WBC drop below 10×109/L | |||
| Low PC | 2784(87.7) | 1.00 | |
| Medium PC | 1886(78.4) | 0.52(0.44–0.61) | < 0.001 |
| Increasing PC | 232(63.2) | 0.29(0.22–0.38) | < 0.001 |
| RBC transfusion | |||
| Low PC | 1402(44.1) | 1.00 | |
| Medium PC | 539(22.4) | 0.45(0.39–0.52) | < 0.001 |
| Increasing PC | 96(26.2) | 0.35(0.26–0.46) | < 0.001 |
| AKI | |||
| Low PC | 2744(86.5) | 1.00 | |
| Medium PC | 2006(83.4) | 0.98(0.83–1.15) | 0.789 |
| Increasing PC | 298(81.2) | 0.65(0.68–1.21) | 0.462 |
| PC Platelet count, AKI acute kidney injury, CI confidence interval, OR odds ratio, HR hazard ratio | |||
| Clinical outcome | Median (IQR) | Regression coefficient (95% CI) | P |
|---|---|---|---|
| WBC drop below 10×109/L (days) | |||
| Low PC | 0.5(0.1–2.2) | ||
| Medium PC | 1.2(0.2–3.5) | 0.64(0.40–0.90) | < 0.001 |
| Increasing PC | 2.8(0.6-7.0) | 2.24(1.68–2.80) | < 0.001 |
| Maximum SOFA score | |||
| Low PC | 8(5–11) | ||
| Medium PC | 6(4–8) | -1.3(-1.45~-1.18) | < 0.001 |
| Increasing PC | 5(4–7) | -1.9(-2.16~-1.62) | < 0.001 |
| PC Platelet count, IQR interquartile range, CI confidence interval, SOFA Sequential Organ Failure Assessment. | |||
In supplement file Table S2, low PC was used to be control group, relative risks of medium PC and increasing PC are as follow, medium PC: gender [RR 0.68, 95% CI 0.60–0.76]; Cephalosporins [RR 0.75, 95% CI 0.66–0.85]; Vasopressor use (1st 24h) [RR 0.71, 95% CI 0.63–0.81]; RRT use (1st 24h) [RR 0.49, 95% CI 0.29–0.83]; Platelet transfusion (7 days) [RR 0.16, 95% CI 0.12–0.22]; CCU/CVICU [RR 0.58, 95% CI 0.49–0.70]; Heparin sodium use [RR 1.34, 95% CI 1.17–1.52]; Diabetes [RR 1.32, 95% CI 1.13–1.55] and Levofloxacin [RR 1.32, 95% CI 1.11–1.57].. Increasing PCą¼gender [RR 0.71, 95% CI 0.55–0.90]; Cephalosporins [RR 0.51, 95% CI 0.40–0.65]; Vasopressor use (1st 24h) [RR 0.52, 95% CI 0.39–0.68]; RRT use (1st 24h) [RR 0.11, 95% CI 0.02–0.85]; Platelet transfusion (7 days) [RR 0.02, 95% CI 0.01–0.09]; CCU/CVICU [RR 0.52, 95% CI 0.35–0.77]; Heparin sodium use [RR 2.06, 95% CI 1.54–2.75]; Hypertension [RR 1.36, 95% CI 1.04–1.77] and Metronidazole [RR 1.68, 95% CI 1.29–2.18].
For sensitivity analysis, we eliminated 572 patients who received platelet transfusions to remove the possible effect of platelet transfusions on PC trajectories. In spite of this, we determined similar PC trajectories (Supplement file Fig S1). Increasing PC, low PC and medium PC were recognized in 320 (6.0%), 3137 (59.0%) and 1917 (35.7%) patients, separately.
In this retrospective cohort research, GBTM was applied to demonstrate PC trajectories in patients with sepsis and PC trajectories were related to hospital mortality. In addition, we observed that PC trajectories were also related to infection control rate, time to infection control and RBC transfusion during first week of ICU admission.
It is well known that the progression of sepsis is rapid, so the starting point of sepsis is difficult to determine accurately. In our study design, the starting point of sepsis was relatively unified (within 2 days of ICU admission) in the expectation of obtaining an accurate trend in platelet count changes before and after the start of sepsis, although the study was retrospective. It can be seen that there is a clear differentiation of platelet count (< 150, > 200, >400) at the beginning of sepsis or before sepsis in the 3 groups of patients, which may represent different pathological states and consequently different trajectories of development. Akca et al. observed an early sharp drop followed by an improvement in platelet count, which was consistent with our study. But we did not find a continuous low level PC trajectory (< 50\(\times\) 109/L), which may be because the proportion of this kind of patients is too low to form a trajectory, or because there will be a new change point in PC after 7 days. It is as expected that low PC trajectory is associated with hospital mortality rather than 90-days mortality. Surprisingly, low PC trajectory is associated with a higher infection control rate and less time to infection control, which may because platelets can induce an acute period response to infection(18, 19), and platelet–leukocyte interactions play the crucial role in defending against infections during inflammation and sepsis(20). The association of red blood cell transfusion with low PC trajectory can be explained by the fact that clinical decisions tend to red blood cell transfusion in septic patients with lower platelet counts under high pressure such as invasive procedures, surgery, etc. The mean difference in maximum SOFA score between three groups was consistent with trends in platelet variation, implying that there were no significant differences between the three groups in terms of systemic function other than blood system, although we did not perform further analysis of the individual system scores. We did not found the association of AKI incidence with PC trajectories,w, ich means platelet function played a more important role in AKI rather than quantity. Increased platelet activation marker levels in clinical and experimental models of AKI include P-selectin, TxB2, CCL5, PF4(21). Unfortunately, the limitations of the retrospective study prevented us from obtaining these data.
In the analysis of risk factors, the septic patients with Vasopressor use (1st 24h) and RRT use (1st 24h) are associated with low PC, which reflects deterioration in respiratory and renal function and so does the platelet function. Although we identified the respective effects of some antibiotics on platelet trajectories, this may be of little help in clinical work, as antibiotic use is often combined. Interestingly, we found that septic patients without diabetes and without hypertension had a higher risk of forming low PC trajectories. Jin-Young Hwang et al, reported that the incidence of type 2 DM increased as the serum platelet count at baseline increased within the normal range(22). In addition, platelet volumes and counts are inversely correlated, and the total platelet mass (platelet count\(\times\) MPV) remains stable(23, 24). In a meta-analysis, higher MPV and PDW in T2DM, but not a parallel reduction in platelet count means that the physiological negative feedback between counts and volume is dysregulated in T2DM, resulting in a higher platelet mass(25). Men have a high risk at low PC trajectory, which may because women generally have higher PC(26, 27).CCU/CVICU patients tend to use antiplatelet drugs which may result in low PC. The correlation between platelet transfusion and the use of heparin and trajectory may be inverted, in the sense that differences in trajectory drive changes in clinical decision making.
Our research is the first to examine the effects of PC trajectories on clinical outcomes in patients with sepsis. But our research still had some limitations. Firstly, the observation-based research cannot provide an explanation for the causality between PC trajectories and outcomes. Secondly, we probably did not correct for further possible confounding factors. Thirdly, we didn't think about platelet function in sepsis, Consequently, we are unable to presume a pathophysiologic mechanism for higher mortality in septic patients with low PC. Fourth, the correct understanding of GBTM is that the model is data driven and it tends to present data characteristics. Specifically for this study, we found 3 PC trajectories, but reality is often complex and there may still be these other PC change trends that were not found due to the limitations of this study.
The Low PC trajectory was related to a higher risk of hospital mortality. Male, Vasopressor use (1st 24h), RRT use (1st 24h), without hypertension, without diabetes, and CCU/CVICU patients are risk factors of septic patients with low PC trajectories.
Data available statement
The original data which support the conclusions of this paper will be provided by authors without improper reservation.
Acknowledgment
We thank the researchers and collaborative research groups at the MIT Computational Physiology Laboratory for maintaining the availability of the MIMIC IV database.
Authors’ contributions
DW collected, interpreted, analyzed data, drew graphs and wrote the manuscript. DW, ZPZ, XL and WPM, modified the manuscript and clarified the findings, commenting on and helping to amend drafts of the manuscript. SQC studied manuscript design, manuscript preparation, and review. All authors read and approved the final manuscript.
Ethical Approval
Not applicable
Funding
No funding.
Disclosure statement
The authors did not report potential conflicts of interest.