Introduction
Intracerebral hemorrhage (ICH) is a global life-threatening disease, the prognosis of ICH patients is commonly poor [1]. In the past 40 years, the incidence of hemorrhagic stroke has increased worldwide [2]. The overall prevalence, morbidity and mortality of stroke in China in 2020 are 2.6%, 50.2% and 34.4% respectively [3]. ICH is the most destructive form of stroke, with a high risk of early deterioration and short-term death [4]. Hematoma expansion and early deterioration are common within the first few hours after ICH onset [5]. The 2019 Chinese guidelines for Clinical Management of Intracerebral Hemorrhage have pointed out that once a patient is diagnosed with ICH, he should be triaged to a stroke unit or neurointensive care unit (Class I recommendation, level A evidence) [6]. In most medical institutions, patients with ICH after admission are usually sent to the intensive care unit (ICU) to detect and fully address their potential complications, such as elevated intracranial pressure and blood pressure, airway damage and other complex medical problems [7, 8].
Long-term stay in ICU will lead to high utilization of resources and increase of patients’ personal medical costs [9]. The previous study has found that 23% of ICU beds are occupied by long-term hospitalized patients [10]. Although most patients with ICH need to receive intervention at the nursing level of ICU or similar environment, the shortage of ICU beds and intensive care nurses is a major problem in China [11]. Understanding the critical care needs of ICH patients is very important to identify the patients who may not need intensive care to stay in ICU [12]. Although ICU care is essential for most patients with ICH, transfer or escalation of care for some patients may be unnecessary. Currently, the standard about whether and when to transfer patients with ICH to a higher level of care is not clear. With the shortage of intensive care resources, tools to guide inter-hospital transfer and help stroke patients allocate intensive care resources properly are becoming more and more important and urgent [13]. At present, the need to predict the intensive care of patients with cerebral hemorrhage is not clear. Therefore, the purpose of this study is to evaluate the status and influencing factors of critical care needs in patients with ICH, to construct a risk prediction model for the critical care needs in patients with ICH, to provide useful evidence for the clinical ICH treatment and nursing care.
Anzeige
Methods
This study was a retrospective cohort design, and this study had obtained the ethical approval of hospital with approval number: 2023 − 312.
Study population
This study included patients with ICH who were admitted to the department of neurosurgery of our hospital from January 2018 to September 2023 as the study population. The inclusion criteria of ICH patients were as following: The ICH was confirmed and diagnosed by cranial computed tomography (CT) and MRI examination in accordance with the diagnostic criteria of Chinese guidelines for diagnosis and treatment of intracerebral hemorrhage regardless of type of ICH; the age ≥ 18 years old. The exclusion criteria were as following: patients with subarachnoid hemorrhage, epidural and subdural hematoma; patients with structural vascular lesions such as arteriovenous malformations (AVMs); those who were discharged or die within 48 h after admission to the neurosurgery department.
Data collection
Two investigators collected following data from the medical records, including: gender, age, hypertension, diabetes, coronary heart disease, past cerebral infarction, past cerebral hemorrhage, pre-hospital use of anticoagulants, the vital signs recorded for the first time (temperature, respiration, systolic blood pressure (SBP), diastolic blood pressure (DBP), pulse oxygen, heart rate), Glasgow Coma Scale (GCS) score, intravenous injection of antihypertensive drugs, as well as laboratory data and brain computed tomography (CT) results (location of bleeding, volume of bleeding, Intraventricular hemorrhage). The CT scanned the location, appearance, intraventricular hemorrhage and leukoaraiosis of the head. The bleeding volume was calculated using the ABC/2 method, and all imaging data were reviewed by imaging physicians and deputy chief neurosurgeons.
Outcomes of interest
The main outcome was whether critical care intervention was needed at any time during hospitalization. After reviewing the literatures [14‐16] and consulting neurosurgical experts to develop an operational definition of critical care needs, critical care interventions included uncontrolled hypertension, the need for titration of intravenous antihypertensive drugs (continuous infusion or intravenous administration exceeds the frequency permitted by non-ICU environment), and the use of vasoactive drugs to treat symptomatic hypotension (epinephrine, dopamine, norepinephrine, etc.), Symptomatic bradycardia or tachycardia required maintenance of the heart rate, invasive hemodynamic monitoring, uncontrolled hyperglycemia requiring continuous intravenous insulin, respiratory damage requiring mechanical ventilation, management of elevated intracranial pressure and brain edema, and the need for cardiac compressions, as well as neurosurgical interventions such as extraventricular drainage or bone flap decompression.
Anzeige
Our hospital has a special stroke ward for patients with ICH. The patients in the stroke ward was evaluated every 2 h, and the patient’s condition is evaluated hourly in ICU. Once the patient is diagnosed with ICH, it should be triaged to the stroke unit or ICU immediately. For patients with ICH whose systolic blood pressure exceeds 150mmHg and there is no contraindication of acute antihypertensive therapy, we would reduce and control the systolic blood pressure to 140mmHg. Blood pressure should be monitored during antihypertensive treatment to avoid excessive blood pressure variability. In addition, we closely monitored blood glucose levels and deal with them accordingly to avoid hyperglycemia and hypoglycemia. When the GCS was less than 8, we would transfer the patient to ICU. We chose the thresholds for ICH volumes and SBP based on the clinical guidelines [17, 18].
Statistical analysis
SPSS 25.0 was used to process data in this study. The quantitative data was expressed as mean ± standard deviation, independent sample t test was used for comparison between the two groups, median (quartile) [M (P 25, P 75)] was used for quantitative data of non-normal distribution, and Mann Whitney U test was used for inter-group comparison. The qualitative data were expressed by frequency and constituent ratio (%), χ 2 test or Fisher accurate test was used for comparison between groups. Mann independent Whitney U test was used for grade data comparison. The risk factors of critical care needs were determined by univariate analysis and logistic regression analysis. In this study, stepwise regression was used to screen variables, and all potential covariates were included in the initial multivariable model. In the refinement of the model, covariates exhibiting the weakest association with the outcome variable were sequentially excluded based on their P values. This process continued until all covariates retained within the model achieved a statistically significant threshold, defined as P < 0.05. Subsequently, the Receiver Operating Characteristic (ROC) curves were utilized to evaluate the predictive performance of the model. The area under the curve (AUC) and Hosmer Lemeshow goodness-of-fit test were used to evaluate the discrimination and calibration of the model, and the sensitivity and specificity of the model were calculated and validated. All the statistical tests were two-sided, and the difference was statistically significant if p < 0.05.
Results
A total of 611 patients with ICH were included for model development. 61.21% (374/611) ICH patients had received critical care interventions, and 151 patients had been admitted to ICU. As presented in Table 1, there were statistical differences in the critical care needs of ICH patients with different age, gender, GCS, use of antihypertensive drugs, systolic blood pressure, diastolic blood pressure, limb hemiplegia, intraventricular hemorrhage, volume of bleeding, past history of hypertension, past use of anticoagulants, past course of hypertension, past blood pressure control (all p < 0.05). The interventions required and their frequency for patients with critical care needs are presented in Table 2.
Table 1
Univariate factor analysis of critical care needs of patients with ICH
Variable | ALL ICH Patients (n = 611) | Do not need critical care (n = 237) | Need critical care (n = 374) | t/Z | p |
|---|---|---|---|---|---|
Age, mean and standard deviation, years | 57.01 ± 14.22 | 55.58 ± 15.28 | 57.92 ± 13.44 | -1.983 | 0.048 |
Gender | 14.569 | < 0.001 | |||
Male | 477(78.1%) | 166(70.0%) | 311(83.2%) | ||
Female | 134(21.9%) | 71(30.0) % | 63(16.8%) | ||
GCS score, median (IQR) | 12(8,15) | 15(14,15) | 9(7,13) | -14.540 | < 0.001 |
Use antihypertensive drugs | 39.281 | < 0.001 | |||
No | 200(32.7%) | 113(47.7%) | 87(23.3%) | ||
Yes | 411(67.3%) | 124(52.3%) | 287(76.7%) | ||
SBP | 34.332 | < 0.001 | |||
< 140mmHg | 160(26.2%) | 90(38.0%) | 70(18.7%) | ||
140-179mmHg | 371(60.7%) | 131(55.3%) | 240(64.2%) | ||
> 180mmHg | 80(13.1%) | 16(6.7%) | 64(17.1%) | ||
DBP | 8.720 | 0.013 | |||
< 140mmHg | 385(63.0%) | 141(59.5%) | 244(65.2%) | ||
140-179mmHg | 194(31.8%) | 89(37.5%) | 105(28.1%) | ||
> 180mmHg | 32(5.2%) | 7(3.0%) | 25(6.7%) | ||
Limb hemiplegia | 9.374 | 0.002 | |||
No | 316(51.7%) | 141(59.5%) | 175(46.8%) | ||
Yes | 295(48.3%) | 96(40.5%) | 199(53.2%) | ||
Location of bleeding | 3.237 | 0.519 | |||
Lobar | 163(26.7%) | 66(27.8%) | 97(25.9%) | ||
Basal ganglia | 266(43.5%) | 95(40.2%) | 171(45.7%) | ||
Cerebellum | 37(6.1%) | 13(5.5%) | 24(6.4%) | ||
Cerebral ventricle | 62(10.1%) | 25(10.5%) | 37(9.9%) | ||
Brain stem | 83(13.6%) | 38(16.0%) | 45(12.1%) | ||
Intraventricular hemorrhage | 34.312 | < 0.001 | |||
No | 360(58.9%) | 178(75.1%) | 192(51.3%) | ||
Yes | 251(41.1%) | 59(24.9%) | 182(48.7%) | ||
Volume of bleeding(mL) | 357.607 | < 0.001 | |||
< 10 | 187(30.6%) | 168(70.9%) | 19(5.1%) | ||
10–30 | 154(25.2%) | 63(26.6%) | 91(24.3%) | ||
> 30 | 270(44.2%) | 6(2.5%) | 264(70.6%) | ||
Hypertension | 377(61.7%) | 134(56.5%) | 243(65.0%) | 4.366 | 0.037 |
Diabetes | 72(11.8%) | 21(8.9%) | 51(13.6%) | 3.183 | 0.074 |
Coronary artery disease | 33(5.4%) | 11(4.6%) | 22(5.9%) | 0.437 | 0.508 |
Previous ischemic stroke | 41(6.7%) | 10(4.2%) | 31(8.3%) | 3.838 | 0.050 |
Previous history of ICH | 33(5.4%) | 12(5.1%) | 21(5.6%) | 0.086 | 0.769 |
Use of anticoagulant | 45(7.4%) | 10(4.2%) | 35(9.4%) | 5.615 | 0.018 |
Past course of hypertension | 3(0,10) | 2(0,7) | 4(0,10) | -2.368 | 0.018 |
Past blood pressure control | 13.838 | < 0.001 | |||
Good | 200(32.7%) | 113(47.7%) | 87(23.3%) | ||
Poor | 411(67.3%) | 123(52.3%) | 287(76.7%) |
Table 2
All interventions required and their frequency for patients with critical care needs
Interventions | Frequency |
|---|---|
Continuous infusion of antihypertensive drugs | 256 |
Extraventricular drainage | 176 |
Invasive hemodynamic monitoring | 135 |
Mechanical ventilation | 87 |
Bone flap decompression | 79 |
The use of vasoactive drugs | 60 |
Continuous intravenous insulin | 51 |
Cardiac compressions | 29 |
We took ICH critical care needs as dependent variables, and univariate statistically significant variables as independent variables, and carried out multivariate logistic regression analysis. As presented in Table 3, the results showed that GCS, SBP, intraventricular hemorrhage, hemorrhage volume and past blood pressure control were the influencing factors of critical care needs of ICH patients (all p < 0.05).
Table 3
Logistic regression analysis of critical care needs of patients with ICH
Independent variable | β | S.E. | p | OR(95%CI) |
|---|---|---|---|---|
Constant | -1.658 | 0.828 | 0.045 | 0.190 |
GCS | -0.124 | 0.049 | 0.012 | 0.852(0.786–0.925) |
SBP | ||||
140-179mmHg | 0.521 | 0.352 | 0.139 | 1.683(0.845–3.381) |
> 180mmHg | 1.377 | 0.546 | 0.012 | 3.963(1.378–11.793) |
Volume of bleeding | ||||
10–30 | 2.172 | 0.341 | < 0.001 | 8.773(4.584–17.532) |
> 30 | 5.399 | 0.538 | < 0.001 | 25.214(10.039–63.327) |
Intraventricular hemorrhage | 0.706 | 0.290 | 0.015 | 2.026(1.150–3.603) |
Past blood pressure control | 0.799 | 0.340 | 0.019 | 2.223(1.148–4.382) |
In this study, a diagram of critical care needs of patients with ICH was established based on the predictive model (see Fig. 1).
Anzeige
Fig. 1
The line diagram of critical care needs of patients with ICH (notes: intraventricular hemorrhage (IVH), past blood pressure control, systolic blood pressure (SBP), Glasgow coma scale (GCS), hemorrhage volume are the risk factors, we draw a vertical line according to the corresponding direction of each factor of the patient, and we can get its corresponding score, and so on, we find out the corresponding score in the state of each variable. finally, the total score of the patient is obtained by adding up the scores of all variables, and on the basis of the total score, a vertical line is drawn down to know the risk of the patient)
×
![]()
Figure 1 The line diagram of critical care needs of patients with ICH (notes: intraventricular hemorrhage (IVH), past blood pressure control, systolic blood pressure (SBP), Glasgow coma scale (GCS), hemorrhage volume are the risk factors, we draw a vertical line according to the corresponding direction of each factor of the patient, and we can get its corresponding score, and so on, we find out the corresponding score in the state of each variable. finally, the total score of the patient is obtained by adding up the scores of all variables, and on the basis of the total score, a vertical line is drawn down to know the risk of the patient).
We randomly split the data (in 7:3 ratio), The area under ROC curve (AUC) was 0.946 (95%CI: 0.929 ~ 0.963) (see Fig. 2). According to the maximum Yoden index (0.750), the optimal critical value was 0.649, linear prediction cutoff value was 0.613. The sensitivity and specificity of the model were 90.56% and 78.22%, The positive predictive value was 0.935, and the negative predictive value was 0.785. Bootstrap self-sampling internal verification was used for 1000 times, and the AUC was 0.943. The result of Hosmer-Lemeshow test showed that χ2 = 6.800 and p = 0.558. The prediction probability of the model was in good agreement with the actual occurrence probability. The calibration curve shows that the model fits well and is in good agreement with the ideal model, as presented in Fig. 3.
Fig. 2
The AUC of the risk prediction model
×
![]()
Anzeige
Fig. 3
The Calibration curve of modeling group
×
![]()
The data for validating the model was 260, and the model was tested on new data. The area under ROC curve (AUC) was 0.927(95%CI:0.896–0.957) (see Fig. 2). Accuracy was 0.827, Bootstrap self-sampling internal verification was used for 1000 times, and the AUC was 0.921. The result of Hosmer-Lemeshow goodness-of-fit test showed that χ2 = 8.617 and p = 0.376. The calibration curve shows that the model fits well and is in good agreementwith the ideal model, as presented in Fig. 4.
Fig. 4
The Calibration curve of validation group
×
![]()
Discussion
The incidence of critical care needs of patients with ICH in this study is 61.19% (533/871). The risk factors for intensive care needs of patients with ICH are gender, GCS score, intraventricular hemorrhage, past blood pressure control, systolic blood pressure on admission and blood loss, which are similar to the previous findings [19, 20]. It is well known that GCS score, intraventricular hemorrhage and bleeding volume are the risk factors of death and functional prognosis in patients with cerebral hemorrhage [21‐23]. Some scholars [24, 25] have reported that GCS is the useful baseline clinical predictor of short-term and long-term mortality of ICH. ICH exists in about 45% of 50% of patients with spontaneous cerebral hemorrhage [26, 27]. The existence and continuous expansion of intraventricular hemorrhage are useful and independent predictors of functional outcome after intracerebral hemorrhage [28, 29]. Compared with patients without intraventricular hemorrhage, patients with intraventricular hemorrhage generally had lower GCS at admission, and patients with intraventricular hemorrhage have a higher mortality rate of 30 days and 1 year [30]. Therefore, treatment should be taken as soon as possible after ICH, because the risk of hematoma expansion is highest within the first three hours after hemorrhage [31, 32].
Anzeige
Hypertension is the main risk factor for ICH and recurrence of ICH [33, 34]. Studies [2, 35] have shown that hypertension is related to hematoma enlargement and poor prognosis after intracerebral hemorrhage. Initial systolic blood pressure > 200mmHg is called acute hypertensive response of ICH and is an independent predictor of hematoma expansion and increased mortality [36]. It mainly occurs within the first hour, and most current studies advocate that early intensive hypotension can improve the clinical prognosis by inhibiting the expansion of hematoma. It has been reported that the target range of systolic blood pressure within 6 h in patients with cerebral hemorrhage is 130-140mmHg, which is safe and may improve functional outcomes [2]. The incidence of poor blood pressure control in survivors of cerebral hemorrhage was 50–80% [37]. It has been found that poor blood pressure control still exists within 6 months after intracerebral hemorrhage, and the incidence of poor blood pressure control is higher in patients with deep cerebral hemorrhage [38]. Guidelines for the prevention of secondary cerebral hemorrhage mainly focus on short-term and long-term blood pressure control in patients. Clinicians need to emphasize the importance of blood pressure control in practice [39]. Effective blood pressure control is the key to primary and secondary prevention of ICH [40]. Blood pressure managements after intracerebral hemorrhage include health education for patients and their families, the use of mobile or electronic medical tools, family blood pressure monitoring, team-assisted care systems, etc [41‐43].
In this study, the intensive care needs model of patients with cerebral hemorrhage is constructed, and a number of risk factors of intensive care needs are integrated and quantified for comprehensive risk assessment, the model has good predictive efficiency. It is important to provide scoring criteria for clinical doctors and nurses to accurately evaluate and identify patients with different monitoring levels of cerebral hemorrhage through early warning and precise control. Thus, it is conducive to the formulation of individual management measures combined with clinical practice, and the formulation of adaptive intervention plan is the focus of the following studies. The establishment of an accurate outcome prediction model in neurosurgery can provide important information for medical staff, patients and their families to make the best management decisions, such as whether to be admitted to ICU or not and the corresponding level of nursing care.
A promising avenue for future research lies in the application of large-scale or contemporary machine learning methodologies to the prediction of clinical problems. The current approach presented in this paper is traditional, relying on a limited set of features to predict the critical care needs of ICH patients. Recent advancements in healthcare analytics have witnessed the development of models that leverage extensive feature sets, potentially numbering in the hundreds or thousands. For instance, Mišić et al. [44] have demonstrated the utility of standard machine learning in predicting emergency department readmissions using a comprehensive feature set derived from laboratory data. Similarly, Lee et al. [45] have employed a deep neural network to predict postoperative mortality in surgical patients, they have shown that while deep learning models can forecast in-hospital mortality based on automatically extracted intraoperative data, they have not yet surpassed the efficacy of established methods. While the models referenced may not offer a direct solution to the predictive challenges posed by the current study, they do illuminate the possibility of enriching predictive models with a broader spectrum of medical record data. This could encompass a variety of clinical, demographic, and biochemical parameters that, when integrated into machine learning frameworks, may enhance the accuracy and reliability of model predictions.
This study, while contributing valuable insights, is not without its limitations. Firstly, despite the relatively large sample size of cerebral hemorrhage cases examined, the findings of this study have not been subjected to prospective validation. It is imperative for future research to undertake multicenter, prospective studies to corroborate and refine the predictive model presented here. Secondly, the criteria for intensive care requirements were established through a literature review and expert consultation, which introduces a degree of subjectivity into the study design. This subjectivity could potentially account for the observed large area under the AUC curve of the model. To mitigate this, future iterations of the model could benefit from the objectification of all indicators, thereby enhancing the model’s generalizability and potentially reducing variability in outcomes. In light of these considerations, it is recommended that subsequent studies employ a more rigorous, objective approach to defining critical care needs, possibly through the development of standardized protocols or the incorporation of additional quantitative measures. This would not only bolster the validity of the model but also facilitate more consistent and reliable predictions across diverse clinical settings.
Conclusion
In conclusion, the results of this study show that the influencing factors of critical care needs of patients with ICH include GCS, systolic blood pressure, intraventricular hemorrhage, bleeding volume and past blood pressure control. This study has developed a clinical predictive model specifically designed for the assessment of critical care needs for patients presenting with ICH. This risk prediction model introduces a standardized scoring system that serves as a valuable tool for clinicians, including physicians and nurses, to precisely evaluate and stratify ICH patients based on varying levels of monitoring needs. The implementation of this model is instrumental in the accurate identification of the appropriate monitoring intensity for ICH patients. It aids in the customization of care plans to contribute to the optimization of critical care resource allocation, ensuring that patients receive the necessary level of attention and intervention.
Acknowledgements
None.
Declarations
Ethics approval and consent to participate
In this study, all methods were performed in accordance with the relevant guidelines and regulations. The study has been reviewed and approved by the ethics committee of First Affiliated Hospital of Soochow University (approval number: 2023 − 312). And written informed consents had been obtained from all the included patients.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
