The max‐intracerebral hemorrhage score predicts long‐term outcome of intracerebral hemorrhage

Summary

Aims

Little is known about the performance of the maximally treated intracerebral hemorrhage (max‐ICH) score in predicting unfavorable long‐term functional outcome and death in patients with intracerebral hemorrhage (ICH) in China. We aimed to validate the performance of the max‐ICH score and compared it with other recognized scores.

Methods

We derived data from the China National Stroke Registry (CNSR). Receiver‐operating characteristic (ROC) analysis and Hosmer‐Lemeshow test were used to measure the score performance. We compared the performance of max‐ICH score with six recognized models, including the ICH score, ICH functional outcome score (ICH‐FOS), Essen‐ICH score, modified intracerebral hemorrhage (MICH) score, intracerebral hemorrhage grading scale (ICH‐GS), and functional outcome (FUNC) score.

Results

A total of 2581 patients with spontaneous ICH were enrolled in the study. The max‐ICH score was similar or superior to the six existing scores in predicting long‐term unfavorable functional outcome after ICH with good discrimination (AUC 0.83, 95% confidence interval [CI] 0.81‐0.84) and calibration (Hosmer‐Lemeshow P = 0.19). For predicting death, the AUC of max‐ICH was 0.81 (95% CI 0.79‐0.83).

Conclusions

The easy‐to‐use max‐ICH score is a reliable tool to predict unfavorable long‐term (12‐month) functional outcome and death after intracerebral hemorrhage in the Chinese population.

1. INTRODUCTION

As one of the least treatable subtypes of stroke, the incidence and all‐age mortality of spontaneous intracerebral hemorrhage (ICH) increased over the last decades in China.1, 2 The 1‐month fatality of ICH was approximately 40% and severe high disability rate in survivors.3 In a recent study focused on palliative care encounter (PCE) in hospitalized stroke patients, ICH was associated with a higher rate of PCE use.4 Furthermore, several studies demonstrated that a substantial proportion of patients with ICH experienced improvement throughout 12 months after the discharge.5, 6 Scales predicting long‐term outcome of ICH patients were different from observed long‐term outcome according to several validation studies.7, 8 However, none of the previous prognosis‐prediction scales9, 10, 11, 12, 13, 14 account for the impact of medical decisions including early care limitations, PCE, or do‐not‐resuscitate order. Incorrect assessment of long‐term outcome might lead to biased clinical judgment of ICH patients. Even adverse consequences on clinical outcome might be caused if the decisions based on prognostic model that withhold beneficial treatments.15 Thus, it is essential to predict the prognosis for patients with ICH, especially for neurologists to make high‐quality decision or to inform patients’ family members choosing appropriate therapy.16

To evaluate the long‐term prognosis of ICH and improve risk stratification, the maximally treated intracerebral hemorrhage (max‐ICH) score was developed recently from a cohort without early care limitations.7 The max‐ICH score had a superior predictive validity comparing to previous long‐term prognosis scores in ICH patients in total ICH cohort of the derivation dataset. It has not been validated in Asian patients with ICH.

In this study, we aimed to assess the performance of the max‐ICH score in Chinese patients from the China National Stroke Registry (CNSR)17 cohort and compare its performance with other 6 recognized scores, including ICH functional outcome score (ICH‐FOS), Essen‐ICH score, modified intracerebral hemorrhage (MICH) score, intracerebral hemorrhage grading scale (ICH‐GS), and functional outcome (FUNC) score.9, 10, 11, 12, 13, 14

2. METHODS

2.1. Data source and subjects

Data of this study were derived from the CNSR. Detailed description of rationale and design as well as baseline characteristics was published previously.17 The CNSR recruited 21 902 patients who were hospitalized in 132 hospitals of 27 provinces and regions between September 2007 and August 2008 with diagnoses of acute stroke. Baseline data collections and etiologic evaluations were completed by trained research coordinators through direct clinical examination. Patient's follow‐up at 3, 6, and 12 months after disease onset was conducted by trained personnel through central telephone interview using standardized scripts. National Institutes of Health Stroke Scale (NIHSS) scores and modified Rankin Scale (mRS) were calculated by qualified neurologists at patients’ arrival at the emergency room. This registry was approved by the institutional ethics committees of all participating hospitals before enrollment. Written consent forms were obtained from all patients or their legal representatives before being recruited into the registry.

Patients aged 18 years or older, admitted to hospital within 14 days of symptoms onset of a spontaneous ICH, were included in the study. ICH was diagnosed according to World Health Organization criteria.17 Major exclusion criteria were secondary ICH (caused by trauma, underlying tumor, or brain structural abnormalities with concrete evidence), any of the components of max‐ICH score unavailable or incomplete follow‐up data (Figure 1).

Figure 1.

Flowchart of patient selection for the study. IS, indicates ischemic stroke; TIA, transient ischemic attack; ICH, intracerebral hemorrhage

2.2. Max‐ intracerebral hemorrhage score and other six predictive scores

The max‐ICH score was obtained by adding up items for NIHSS score (0‐6 = 0; 7‐13 = 1; 14‐20 = 2; ≥21 = 3), age group (≤69 = 0; 70‐74 = 1; 75‐79 = 2; ≥80 = 3); intraventricular hemorrhage (No = 0; Yes = 1); oral anticoagulant agents (No = 0, Yes = 1); hematoma volume for lobar origin (<30 = 0; ≥30 = 1); hematoma volume for nonlobar origin (<10 = 0; ≥10 = 1).7 ICH involving both lobar and nonlobar regions should be scored according to the location that ICH most likely originated from. If there were 2 distinct ICH (1 large lobar and 1 large nonlobar) in one patient, then the patient would be assigned more than 1 point referring to ICH volume (Table S1). The highest max‐ICH total score was 9 points as none of the ICH patients were with 2 distinct location of hemorrhage.

The ICH‐FOS score was obtained by adding up items for age group (≤59 = 0; 60‐69 = 1; 70‐79 = 2; ≥80 = 4), NIHSS assessed at arrival in emergency room (0‐5 = 0; 6‐10 = 2; 11‐15 = 3; 16‐20 = 4; ≥21 = 5), Glasgow Coma Scale (GCS) assessed in emergency room (15‐13 = 0; 12‐9 = 1; 8‐3 = 2), admission glucose (≤11.0 mmol/L = 0; ≥11.1 mmol/L = 1), hematoma volume for supratentorial origin (<40 cm³ = 0; 40‐70 cm³ = 2; >70 cm³ = 2), hematoma volume for infra‐tentorial origin (<10 cm³ = 0; 10‐20 cm³ = 2; >20 cm³ = 2), extension into the ventricles (no = 0; yes = 1).10 The volume of hematoma was assessed using the ABC/2 method.18

The ICH score is one of the most widely accepted scores.19, 20, 21 It was obtained by adding up items for: GCS (15‐13 = 0; 12‐5 = 1; 4‐3 = 2), hematoma volume (<30 cm³ = 0; ≥30 cm³ = 1), presence of intraventricular hemorrhage (no = 0; yes = 1), infra‐tentorial origin (no = 0; yes = 1), age (<80 = 0; ≥80 = 1).9

The Essen‐ICH score was calculated by adding up points assigned for: age (<60 = 0; 60‐69 = 1; 70‐79 = 2; ≥80 = 3), NIHSS (0‐5 = 0; 6‐10 = 1; 11‐15 = 2; 16‐20 = 3; ≥21 or coma = 3), and NIHSS level of consciousness (alert=0; drowsy=1; stupor=2; comatose=3).12

The modified intracerebral hemorrhage (MICH) score was obtained by adding up points for GCS (15‐13 = 0; 12‐5 = 1; 4‐3 = 2), hematoma volume (≤20 cm³ = 0; 21‐50 cm³ = 1; ≥51 cm³ = 2), presence of intraventricular hemorrhage, or hydrocephalus (no = 0; yes = 1).11

The ICH‐GS score was obtained by adding up points assigned for: age (<45 = 1; 45‐64 = 2; ≥65 = 3), GCS (15‐13 = 1, 12‐9 = 2, 8‐3 = 3), ICH location (supratentorial = 1; infra‐tentorial = 2), hematoma volume (for supratentorial origin: <40 cm³ = 1; 40‐70 cm³ = 2; >70 cm³ = 3; for infra‐tentorial origin: <10 cm³ = 1; 10‐20 cm³ = 2; >20 cm³ = 3).14

The FUNC score was obtained by adding up points for age (<70 = 2; 70‐79 = 1; ≥80 = 0), hematoma volume (<70 cm³ = 2; 70‐79 cm³ = 1; ≥80 cm³ = 2), ICH location (lobar = 2; deep = 1), GCS (≥9 = 2; ≤8 = 0), pre‐ICH cognitive impairment (no = 1; yes = 0).13

The lobar ICH was defined as ICH originating from brain cortex or cortical‐subcortical junction. Nonlobar ICH was defined as ICH originating from deep, cerebellar, and brainstem origin.22, 23 Deep ICH comprised ICH that exclusively involving basal ganglia, thalamus, internal capsule, and deep periventricular white matter.22

2.3. Clinical outcomes

The primary outcome of this study was unfavorable long‐term (12‐month) functional outcome after symptoms onset, which was defined as the mRS of 4‐6 based on the original article of derivation study.7

2.4. Statistical analysis

Categorical variables were presented as percentages; continuous variables were presented as mean and SD or median (IQR) depending on whether the variable is normally distributed or skewed. We used Shapiro‐Wilk test to examine the normality of continuous variables including the area under the receiver‐operator curve (AUC) of each included predicting models. Odds ratios with 95% confidence intervals (CIs) were calculated with multivariable logistic regression.

The discriminatory power of the max‐ICH score was assessed by AUC and 95% CI. An AUC statistic of 1.0 indicates perfect prediction, and of 0.5 indicates no better than random prediction. Calibration was assessed using Pearson correlation coefficient and Hosmer‐Lemeshow goodness‐of‐fit test. Mann‐Whitney and Delong method were used for comparisons of pairwise AUCs among max‐ICH score and 6 other scores.24 Level of significance was determined as P < 0.05, 2‐sided. All statistical analyses were performed using SAS version 9.4 software (SAS Institute Inc., Cary, NC, USA).

3. RESULTS

3.1. Patient characteristics

Our study included 2851 ICH patients who met the inclusion criteria (Figure 1). The mean age of included patients was 61.9 years, and 38.2% of them were female. The baseline characteristics, medical history and clinical characteristics of this study population, and the population which were used to develop the max‐ICH score were listed in Table 1.

Table 1.

Comparison of baseline characteristics and unfavorable long‐term outcome (12‐month) between the china national stroke registry cohort and the original germany institutional ICH registry

Intracerebral hemorrhage	CNSR (n = 2,851)	Germany ICH registry (n = 471)
Age, y (mean ± SD)	62.0 ± 13.1	70 ± 12
Female sex, n (%)	38.2%	45.4%
Medical history
Hypertension, n (%)	1959 (68.7%)	385 (81.7%)
Diabetes mellitus, n (%)	266 (9.3%)	127 (27.0%)
Hyperlipidemia, n (%)	216 (7.6%)	152 (32.3%)
Stroke, n (%)	771 (27.0%)	140 (29.8%)
Previous use of medication
Antiplatelet agents, n (%)	261 (9.2%)	137 (29.1%)
Anticoagulant agents, n (%)	28 (1.0%)	90 (19.1%)
GCS at admission, median (IQRa)	15 (10‐15)	13 (10‐15)
NIHSS at admission median (IQR)	8 (3‐15)	11 (5‐19)
ICH score, median (IQR)	1 (0‐2)	1 (0‐2)
Admitted to NCU or ICU, n (%)	449 (15.8%)	471 (100%)
Admitted to ward, n (%)	2402 (84.3%)	0 (0)
Length of hospital stay, d, median (IQR)	19 (12‐27)	12 (7‐19)
Unfavorable long‐term outcome, n (%)	1864 (65.4%)	NA

CNSR, China National Stroke Registry; GCS, Glasgow Coma Scale; ICH, intracerebral hemorrhage; IQR, interquartile range; ICU, intensive care units; NA, unavailable data; NIHSS, National Institutes of Health Stroke Scale; NCU, neurointensive care units.

^a25th‐75th percentile.

3.2. The max‐intracerebral hemorrhage score and risk of unfavorable functional outcome

Along with the increasing max‐ICH score, proportions of unfavorable long‐term (12‐month) functional outcome became higher. Each 1‐point increase in the max‐ICH score was associated with an OR of 1.24 for unfavorable outcome (95% CI 0.71‐0.91; P < 0.001). The proportions of each level of functional outcome evaluated by modified Rankin Scale (0‐6) were displayed in the gradient of max‐ICH score as shown in Figure 3.

Figure 3.

Distribution of modified Rankin Scale (mRS) score in relation to total max‐ICH scores. The size of each bar represents percentage of patients with a particular mRS score in each level of max‐ICH score per point

3.3. Comparing the max‐intracerebral hemorrhage score with 6 other predicting scores

For prediction of 12‐month unfavorable functional outcome, the ICH‐FOS score had the largest (0.85, 95% CI 0.83‐0.86) and FUNC score the smallest ROC (0.73, 95% CI 0.71‐0.75) (Table 2). Among the three scores with AUCs exceeding 0.8, ESSEN‐ICH score had the second largest AUC (0.84, 95% CI 0.83‐0.86). Likewise, max‐ICH score showed good discriminatory power with a ROC of 0.83 (95% CI 0.81‐0.84), which was greater than ICH score, MICH score, ICH‐GS score, and FUNC score (Figure 2A). Pairwise comparison of using nonparametric tests between max‐ICH and the other six scores yielded differing ROC for four scores (max‐ICH, ICH, Essen‐ICH, and ICH‐GS score). Differences in ROCs were observed between max‐ICH and three other included scores (ICH‐FOS, Essen‐ICH, and FUNC) (Table 2).

Table 2.

Comparisons of AUCs among max‐ICH score and other predictive scores for unfavorable long‐term (12‐month) functional outcome and death

Predicting scores	Unfavorable long‐term functional outcome					Death within 12‐month after ICH
	AUC	95%CI	ORa	95%CI	P valueb	AUC	95%CI	ORc	95%CI	P valueb
Max‐ICH	0.83	0.81‐0.84	1.24	1.11‐1.40	/	0.81	0.79‐0.83	1.093	0.95‐1.26	/
ICH score	0.77	0.76‐0.79	1.40	1.12‐1.76	0.003	0.81	0.79‐0.83	1.735	1.37‐2.21	0.85
ICH‐FOS	0.85	0.83‐0.86	1.05	0.94‐1.18	0.36	0.83	0.81‐0.85	0.981	0.87‐1.11	<0.001
Essen‐ICH	0.84	0.83‐0.86	1.38	1.25‐1.53	<0.001	0.83	0.81‐0.86	1.329	1.21‐1.46	0.005
MICH	0.77	0.75‐0.78	0.96	0.80‐1.14	0.60	0.79	0.77‐0.82	0.969	0.80‐1.18	0.16
ICH‐GS	0.80	0.78‐0.81	1.18	1.03‐1.34	0.02	0.82	0.80‐0.84	1.116	0.96‐1.30	0.45
FUNC	0.73	0.71‐0.75	1.00	0.92‐1.08	0.95	0.78	0.76‐0.80	0.947	0.87‐1.03	0.01

AUC, area under curve; CI, confidence interval; FUNC, functional outcome score; ICH indicates intracerebral hemorrhage; ICH‐FOS—ICH functional outcome score; ICH‐GS, intracerebral hemorrhage grading scale; MICH, modified intracerebral hemorrhage score; OR, odds ratio.

^aOR of each 1‐point increase in each score associated with unfavorable outcome.

^bPairwise P value between each predicting score and max‐ICH score.

^cOR of each 1‐point increase in each score associated with death.

Figure 2.

Receiver‐operating characteristic curves comparing the max‐ICH score with other predictive scores for prediction of long‐term unfavorable outcome (Figure 2 A) and death (Figure 2 B), respectively. ICH indicates the ICH score; ICH‐FOS, functional outcome score; MICH, modified intracerebral hemorrhage score; ICH‐GS, intracerebral hemorrhage grading scale, and FUNC, functional outcome score

For prediction of 12‐month death, Essen‐ICH score had the largest and FUNC score the smallest ROC (Figure 2B). Meanwhile, max‐ICH, ICH score, ICH‐FOS, and ICH‐GS performed well with ROCs of 0.81, 0.81, 0.83, and 0.82, respectively (Table 2).

The max‐ICH score also performed well in the calibration analysis (Figure S1). The Hosmer‐Lemeshow test correlation coefficient between observed and predicted probability of unfavorable or favorable outcome was 11.3. The significance level of the Hosmer‐Lemeshow test was 0.19. Only 1 (2.6%) patient reached favorable functional outcome in 38 patients with a score above 7 points. None of the patients with a total score of 9 points reached favorable outcome, and 91.7% of this interval deceased. More than 90% of patients with 0 point reached favorable outcome (Figure 3). Rates of unfavorable outcome (mRS of 4‐6) at 12‐month post‐ICH for max‐ICH score 0‐2, 3‐6, and 7‐9 points were 13.8%, 53.2%, and 91.4%, respectively (P _trend < 0.001).

The high correlation between the observed and max‐ICH score predicted probability of favorable outcome (Wald chi‐square 13.0) indicates the usability of the max‐ICH grading scale for identifying patients with potential probability to reach good prognosis.

4. DISCUSSION

Compared with 6 other recognized scales, we found that max‐ICH score had a reliable or even superior performance in predicting unfavorable long‐term functional outcome of ICH and stratifying patients with the probabilities of favorable outcome.

The max‐ICH score was developed in a maximally treated ICH patient cohort in a single center in Germany to minimize the confounding effects of early care limitations. Our study performed the external validation of the max‐ICH score in a nation‐wide multicenter cohort which avoid the specificity of single‐center characteristics. However, the max‐ICH score was not superior to the ICH‐FOS score, which was developed from the CNSR cohort with good result in internal validation. Unlike the requirement of admission glucose of the ICH‐FOS score,10 or the knowledge of cognitive status before ICH when calculating the FUNC score,13 as soon as getting the age of ICH patient, the result of physical and neurological examination and cranial CT scan that revealing the location and volume of the hematoma, the max‐ICH score could be obtained timely.

The items of the max‐ICH score were all established risk factors of unfavorable long‐term functional outcome.25 Correlations between the items were also reported to be associated with outcome of ICH. Incidence of ICH with different etiologies and mechanisms was varied by age.26 Cerebral amyloid angiopathy (CAA) attributed a large proportion of intracerebral hemorrhage in the elder population. Age‐dependent incidence of CAA was evidently increased from 13.8% of 60 and 69 years group to 44.8% of 80 years and older group.27 Baseline hematoma volume and hematoma expansion were associated with neurological deterioration after ICH and predicted unfavorable short‐term outcome.28 The correlation between age and hematoma volume was also discovered in ICH patients as hematoma volume tended to increase in lobar ICH patients older than 70 years.29 Extension of hematoma into ventricles and higher severity (evaluated by NIHSS score) were significantly associated with unfavorable outcome after spontaneous ICH.5, 30, 31, 32 The max‐ICH score is an integral tool including the quantifiable risk factors of unfavorable outcome.

We recognized superior prognostic utility of ICH‐FOS score and Essen‐ICH score compared with max‐ICH score. Differences between the max‐ICH score and ICH‐FOS score were categorical method of ICH location and inclusion of admission serum glucose. According to the analysis based on INTERACT‐2, worst short‐term (90‐day) outcome was seen in patients with ICH in posterior limb of internal capsule or thalamus.31 Both locations were included in the same category in the 2 scores. However, impaired glucose regulation independently associated with poor functional outcome of ICH patients.33, 34, 35 The inclusion of admission serum glucose might be the critical component which leads to a higher discrimination power of predicting the unfavorable outcome of ICH patients using ICH‐FOS score than the max‐ICH score. The level of consciousness was reported to be one of the most significant factors of poor outcome of ICH patients.36 The superior performance of Essen‐ICH score in recognizing unfavorable long‐term outcome of ICH patients might due to highlighting the importance of consciousness level.

There were certain strengths of this study. First, the CNSR was a prospective cohort of acute‐phase stroke, and multicenter registry, which allowed the validation as unbiased by single‐center characteristics as possible. Second, this study was the first external validation of max‐ICH in Chinese stroke patients up to date. Third, other than the commonly selected short‐term prognosis, the max‐ICH score was mainly used in predicting long‐term outcome of ICH patients, the longer time course would be beneficial for decision‐making in daily clinical practice. Additionally, we included ICH patients without exclusion of those received early care limitations, indicating a reliable application of max‐ICH score in all ICH patients.

Our study had several limitations. First, detailed information about decisions of care limitations such as life‐support withdraws or palliative care encounter was not available according to the CNSR database which resulted in carrying potential bias of medical decisions.37, 38 Second, the follow‐up by phone interview might introduce misjudgment into the functional outcome evaluation. Evidence from previous studies indicated that structured telephone assessment of the modified Rankin Scale is reliable and comparable with face‐to‐face interview with good agreement.39, 40 The phone interviewers were all personnel at Beijing Tiantan Hospital who received centralized training using identical questionnaire which would help reducing bias causing by potential inconsistency of the results to a certain extent. Interobserver and intraobserver consistency pretests among phone interviewers might be solutions to the concern of inconsistency. Third, the max‐ICH score did not include medical history or prestroke dependence status, which might lead to confounding effects of comorbidities. It is important to obtain detailed medical history for supplement to avoid underestimation of unfavorable outcome because of compromised organ function due to prestroke chronic diseases. Forth, with constant improvement in medical care in China, the CNSR can hardly represent the current situation of treatment strategies and secondary prevention of ICH. Fifth, the proportions of different etiologies causing ICH might have evolved as tendency of increasing senior citizens and better prevention in risk factors such as hypertension.41, 42 The CNSR enrolled patients from September 2007 and August 2008, which might not represent the etiology structure of ICH nowadays.

5. CONCLUSIONS

The max‐ICH score is a useful tool to evaluate the long‐term functional outcome and to identify patients in whom full medical support is potentially benefit in ICH patients. This could be useful for improving risk stratification of ICH patients in research and clinical practice.

DISCLOSURES

None.

Supporting information

 

Suo Y, Chen W‐Q, Pan Y‐S, et al. The max‐intracerebral hemorrhage score predicts long‐term outcome of intracerebral hemorrhage. CNS Neurosci Ther. 2018;24:1149–1155. 10.1111/cns.12846