Utility of frailty as a predictor of acute kidney injury in patients with aneurysmal subarachnoid hemorrhage

Abstract

Introduction

Acute kidney injury (AKI) is associated with poor outcome in aneurysmal subarachnoid hemorrhage patients (aSAH). Frailty has recently been demonstrated to correlate with elevated mortality and morbidity; its impact on predicting AKI and mortality in aSAH patients has not been investigated.

Objective

Evaluating risk factors and predictors for AKI in aSAH patients.

Methods

aSAH patients from a single-center's prospectively maintained database were retrospectively evaluated for development of AKI within 14 days of admission. Baseline demographic and clinical characteristics were collected. The effect of frailty and other risk factors were evaluated.

Results

Of 213 aSAH patients, 53 (33.1%) were frail and 12 (5.6%) developed AKI. Admission serum creatinine (sCr) and peak sCr within 48 h were higher in frail patients. AKI patients showed a trend towards higher frailty. Mortality was significantly higher in AKI than non-AKI aSAH patients. Frailty was a poor predictor of AKI when controlling for Hunt and Hess (HH) grade or age. HH grade ≥ 4 strongly predicted AKI when controlling for frailty.

Conclusion

AKI in aSAH patients carries a poor prognosis. The HH grade appears to have superior utility as a predictor of AKI in aSAH patients than mFI.

Introduction

Aneurysmal subarachnoid hemorrhage (aSAH) makes up approximately 80% of total spontaneous non-traumatic SAH and accounts for 2–5% of new stroke patients per year. ¹ Treatment of aSAH has undergone significant advances in surgical, interventional, and medical management, resulting in improved outcomes. ² Despite this, patients often require admission to the intensive care unit (ICU) due to a high rate of complications, including early rebleeding, vasospasm, delayed cerebral ischemia, acute kidney injury (AKI) and hydrocephalus.^2,3 These have been associated with poor outcomes with a fatality rate as high as 51%, especially when considering factors such as level of consciousness on admission, age, and amount of blood on initial imaging^1–3. Although limited investigations have been conducted on the specific impact of AKI on non-traumatic SAH outcomes, AKI is associated with respiratory failure, pneumonia, and increased length of stay (LOS), disability, and mortality^2,4–9. The use of contrast in neuroendovascular imaging and intervention is a known cause of renal impairment in aSAH patients, however the specific etiology of AKI in aSAH patients has yet to be elucidated.^10,11

The Risk, Injury, Failure, Loss, and End-stage renal disease (RIFLE) criteria is utilized frequently in the literature to identify AKI, and was developed by the Acute Dialysis Quality Initiative (ADQI) Group in 2004 to establish consistent and sensitive guidelines for the detection and management of this condition. ¹² Subsequent criteria, such as the AKI Network (AKIN) and Kidney Disease Improving Global Outcomes (KDIGO), incorporate the RIFLE criteria in their prognostication algorithms. ¹³ In all three criteria, serum creatinine levels and urine output are used to define risk and stages of AKI.

Frailty is an important prognostication tool in determining outcomes in surgical and critically ill patients. It is an increasingly common variable of interest across disciplines and its usefulness to predict outcomes has been studied extensively.^14–22 Past literature has demonstrated that frailty is an independent predictor for increased AKI incidence in elderly patients, and non-frail patients with AKI are also more at risk to develop frailty.^23–25 A few studies show frailty to be associated with poor outcomes in aSAH patients, but its association may be relatively weak compared to age.^15,26

The modified frailty index 11 scale (mFI-11) integrates 11 factors related to relevant systemic comorbidities into a generalized predictor of frailty to help guide efficient clinical decision-making. It was originally derived from the Canada Study of Health and Aging 70-item scale (CHSA-FI), which provides a marker for relative fitness and risk of adverse outcomes. ²⁷ The mFI has been used widely in the literature to predict outcomes across disciplines and procedures^28–30. Its simplicity allows for stratification of frailty in the most appropriate way per project.^28,29 In the present study, mFI-11 appropriately encompasses history of cerebrovascular accidents and myocardial infarction, conditions directly related to AKI.

The Hunt and Hess (HH) Scale is one of the most widely used classification scales for aSAH.^31–35 Derived as a modification of Botterell's classification, the HH Scale classifies aSAH as Grades I-V based on symptomatology, alertness, and focal neurological deficits. The intended use was in aiding surgeons on deciding indication and timeline for surgical intervention. ³⁵ In the literature, the HH grade is used prolifically to study outcomes in aSAH and is a well-established prognostic indicator.^2,33,34,36 However, HH grade as a predictor of AKI in aSAH remains unstudied.

Although early detection of AKI is important to lower post-aSAH complications and mortality, robust prediction methods for AKI in aSAH patients do not yet exist. Our study investigates the utility of the 11-point modified frailty index (mFI), age, and HH grade in predicting AKI in aSAH patients.

Methods

Study design/setting

This is a single center retrospective cohort study of patients with aSAH presenting between June 2014 and July 2018. Data was collected from the electronic medical record system after approval from our Institutional Review Board.

Participants

217 aSAH patients, previously evaluated for age, frailty and their effects on outcomes in a recently published article, were assessed for inclusion. ¹⁵ Potentially eligible patients were identified through review of the departmental patient database and digital subtraction angiogram (DSA) institutional database. All individuals with aSAH, defined as presence of an aneurysm on DSA coupled with a non-traumatic SAH presentation on initial CT or MRI, were included. The exclusion criteria were as follows: traumatic SAH, non-aneurysmal SAH, incomplete/unavailable records, concurrent non-SAH neurosurgical or acute medical illness on presentation, non-acute SAH presentation, and no hemorrhage on imaging. From the original cohort, 213 patients were retained for the purposes of this study. Four patients were excluded due to lack of data, including baseline sCr.

Variables

We identified aSAH patients who developed AKI by trending serum creatinine (sCr) within 14 days of admission. AKI was defined as an increase in baseline serum creatinine by a threshold of 1.5 times or greater. This threshold was used because it is the most sensitive threshold for detecting AKI consistent across different AKI criteria (RIFLE, AKIN, and KDIGO). Baseline sCr was identified as the lowest sCr on admission. The primary outcome was development of AKI amongst the entire group of aSAH patients. Variables collected included age, gender, BMI, comorbidities of diabetes mellitus, hypertension, smoker status, Hunt and Hess (HH) grade, Fisher score, Glasgow Coma Scale (GCS), hospital and ICU LOS, and race. Data related to external ventricular drain (EVD), pneumonia, deep venous thrombosis, and radiographic vasospasm were also collected.

The variables of the 11-point mFI, which include diabetes mellitus, functional status, chronic obstructive disease or pneumonia, congestive heart failure, myocardial infarction, previous percutaneous coronary intervention or cardiac surgery or angina, hypertension requiring medication, peripheral vascular disease, impaired sensorium, transient ischemic attack or cerebrovascular accident, and neurologic deficit after previous cerebrovascular accident, were extracted from patient charts.^37,38 Each variable counted as 1 point. This score has been stratified to classify a patient's level of frailty using different cut off points.^15–17,38 Frailty was dichotomized based on the mean mFI and one standard deviation above. A frail state was defined when mFI was 2 or greater in our patient cohort. Potential confounders of mFI considered were age and severity of aSAH.

Bias

Sources of bias in this cohort study included selection bias, information bias, and confounding bias. Selection bias cannot be ruled out as we have limited the follow up period to length of hospitalization. Well defined inclusion criteria was used to attempt to reduce the risk of selection bias. Information bias is not completely avoidable given the nature of the electronic medical records; commonly collected variables and recognized scales may contribute towards avoidance of this. Confounding bias risk was high, and this was addressed by performing multivariable logistic regression as described in the statistics section.

Statistical methods

Statistical analysis was performed using SPSS (IBM SPSS Statistics for Windows, Version 28.0. Armonk NY:IBM Corp) and R statistical software (R, version 3.5.1, R Project, www.r-project.org). Baseline clinical and demographic characteristics were provided though descriptive statistics. Continuous variables collected were evaluated by the Shapiro-Wilk test for presence of a normal distribution. The Student t-test and Mann-Whitney U test were then used to compare normally and non-normally distributed continuous variables, respectively. Fisher's exact test (expected frequency was <5) or Pearson's chi-squared tests were used to compare categorical variables, as appropriate. Binary logistic regression was used to evaluate independent variables for risk factors of primary and secondary outcomes. Variables were also further evaluated using multivariate logistic regression depending on their univariate significance. Receiver operating characteristics (ROC) were used to calculate area under the curve (AUC). The DeLong-DeLong method was used to test AUC model performance, and the partial AUC (pAUC) was compared using bootstrapping. ³⁹ AUC values were categorized as follows: 0.9–0.99 = excellent, 0.8–0.89 = good, 0.7–0.79 = fair, 0.51–0.69 = poor, and 0.5 and below = no value. We set statistical significance to p < 0.05.

Results

Demographics and clinical characteristics

A total of 213 patients admitted to our neurocritical care unit with aSAH were identified and used in the final analysis. We explored the differences between non-frail (mFI<2, n = 160, 75.1%) and frail (mFI≥2, n = 53, 24.9%) patient groups. Average age was higher in patients with mFI≥2 in comparison to mFI<2 (66 ± 13.7 vs. 54.6 ± 13.5, p < 0.001). There was no significant difference in gender or race between the groups (p = 0.638 and p = 0.424, respectively). Body mass index (BMI) and current smoker status showed no statistically significant difference between groups (p = 0.296 and 0.722, respectively). There was a higher incidence of Fisher Grade of 4 (88.7% vs. 70%, p = 0.008) and increased HH grade in frail patients (p = 0.031). Admission sCr and peak sCr within 48 h were higher in the frail patient group (0.85 ± 0.24 vs. 0.78 ± 24, p = 0.012 and p = 0.002, respectively) (Table 1). Average LOS for frail and non-frail patients was 20.26 and 21.10 days, respectively (p = 0.049). Average LOS for patients with AKI and without AKI was 11.25 and 21.47, respectively (p = 0.813).

Table 1.

Comparison of patient demographics and baseline characteristics for mFI <2 versus mFI≥2 in patients with aSAH.

Variables	Total (n = 213)	mFI<2 (n = 160)	mFI≥2 (n = 53)	P value
Age(yr)	57.4 ± 14.4	54.6 ± 13.5	66 ± 13.7	<0.001
Old age(>65)	63 29.6%	35 21.9%	28 52.8%	<0.001
Sex				0.638
Male	74 34.7%	57 35.6%	17 32.1%
Female	139 65.3%	103 64.4%	36 67.9%
BMI	27.7 ± 6.8	27.4 ± 6.6	28.7 ± 7.3	0.296
Diabetes Mellitus	24 11.3%	3 1.9%	21 39.6%	<0.001
Hypertension	109 51.2%	60 37.5%	49 92.5%	<0.001
Current Smoking	84 39.4%	62 38.8%	22 41.5%	0.722
Fisher grade 4	160 75.1%	113 70%	47 88.7%	0.008
Admission sCr (mg/dL)	.80 ± .24	.78 ± 24	.85 ± .24	0.012
Peak sCr within 48 h (mg/dL)	.90 ± .45	.85 ± 26*	1.0 ± .77	0.002
Race				0.424
White	122 57.3%	92 57.5%	30 56.6%
Black	23 10.8%	15 9.4%	8 15.1%
Asian	10 4.7%	6 3.8%	4 7.5%
Hispanic	49 23%	39 24.4%	10 18.9%
Other	9 4.2%	8 5.0%	1 1.9%
Hunt and Hess grade				0.031
1.0	30 14.1%	28 17.5%	2 3.8%
2.0	55 25.8%	45 28.1%	10 18.9%
3.0	63 29.6%	44 27.5%	19 35.8%
4.0	33 15.5%	21 13.1%	12 22.6%
5.0	32 15%	22 13.8%	10 18.9%

*Peak sCr within 48 h contained n = 154 instead of n = 160 for our sample with mFI<2. BMI: Body mass index; sCr: serum creatinine; mFI: modified frailty index.

AKI versus non-AKI aSAH patients

Mortality was significantly increased in patients who developed AKI (OR = 16.45, 95% CI 4.22–64.20, p < 0.001). Patients with AKI had significantly higher rates of HH grade 4 or greater aSAH (OR = 7.77, 95% CI 2.03–29.74, p = 0.001). Patients with AKI were also more likely to be frail, although this did not reach statistical significance (OR = 3.28, 95% CI 1.01–10.64, p = 0.077). There was no difference in the rates of radiographic vasospasm or EVD requirement. Elderly patient age (>65 years) was also not statistically different between AKI and non-AKI aSAH patients (OR = 1.76, 95% CI 0.54–5.77, p = 0.344). (Table 2) There was no difference between patients undergoing endovascular treatment for aSAH and the development of AKI (OR = 2.73 p = 0.469).

Table 2.

Bivariate analysis showing differences between aSAH patients without AKI and those with AKI.

Relevant aSAH Clinical Features	Total (n = 213)	No AKI (n = 201)	AKI (n = 12)	OR(95% CI)	P value
Age > 65	63 29.6%	58 28.9%	5 41.7%	1.76(0.54–5.77)	0.344
EVD	161 75.6%	152 75.6%	9 75%	0.97(0.25–3.72)	1.00
Radiographic vasospasm	123 57.7%	118 58.7%	5 41.7%	0.50(0.15–1.64)	0.246
Hunt Hess ≥4	65 30.5%	56 27.9%	9 75%	7.77(2.03–29.74)	0.001
mFI≥2	53 24.9%	47 23.4%	6 50%	3.28(1.01–10.64)	0.077
Mortality	40 18.8%	31 15.4%	9 75%	16.45(4.22–64.20)	<0.001

EVD: external ventricular drain; mFI: modified frailty index; AKI: acute kidney injury; aSAH: aneurysmal subarachnoid hemorrhage.

Risk factors for AKI

Univariate risk factors for development of AKI among all aSAH patients included increasing age (OR = 1.05, 95% CI 1.00–1.10, p = 0.043), mFI ≥ 2 (OR = 3.28, 95% CI 1.01–10.64, p = 0.048), and HH grade ≥ 4 (OR = 7.77, 95% CI 2.03–29.74, p = 0.003). Gender, black ethnicity, smoker status, baseline creatinine level, and BMI were not significant predictors of AKI on univariate analysis). In multivariate analysis including mFI and HH, only HH grade ≥ 4 was a significant predictor for AKI (OR = 6.93, 95% CI 1.79–26.88, p = 0.005). In multivariate analysis including age and mFI, neither age nor mFI were significant predictors of AKI in aSAH (Table 3).

Table 3.

Univariate and multivariate logistic regression analysis for predictors of AKI in patients with aSAH.

	Univariate		Multivariate 1		Multivariate 2
	OR (95% CI)	p-value	OR (95% CI)	p-value	OR (95% CI)	p-value
Age(yr)	1.05 (1.00 – 1.10)	.043			1.04 (.99–1.09)	0.154
Female	2.80 (0.86–9.15)	.088
Black ethnicity	3.02 (0.75–12.06)	.118
mFI≥2	3.28 (1.01–10.64)	.048	2.58 (.76–8.73)	.128	2.25 (.62–8.11)	0.215
Smoking status	.756(0.22–2.60)	.657
Admission sCr (mg/dL)	4.22(0.93–19.27)	.063
Hunt Hess ≥4	7.77 (2.03–29.74)	.003	6.93 (1.79–26.88)	.005
BMI	.93 (.83–1.04)	.204

mFI: modified frailty index; HH: Hunt and Hess; sCr: serum creatinine.

ROC curve analysis revealed that HH grade (AUC: 0.76, 95% CI 0.59–0.92, p = 0.003) was a “fair” predictor for AKI in a SAH patients, while mFI (AUC: 0.61, 95% CI 0.43–0.80, p = 0.186) was a “poor” predictor (Figure 1(a)). The DeLong-DeLong test compared multivariate models for prediction of AKI, demonstrating no significant improvement with addition of mFI (Z = 0.08, p = 0.5) (Figure 1(b)).

Figure 1.

(a) ROC curve analysis comparing AUC of predictors mFI, Age, and HH in all aSAH patients; and (b) comparing AUC of multivariable predictor models with and without mFI.

Discussion

This study shows that HH grade outperforms mFI in predicting AKI in the setting of aSAH. Although mFI is a positive predictor of AKI on univariate analysis, the significance does not hold on multivariate analysis when controlling for HH grade. Furthermore, AUC analysis using ROC demonstrates that mFI has poor discriminatory value for predicting development of AKI compared to HH grade.

Characteristics of aSAH patients with AKI

HH grade greater than or equal to 4 was higher among patients with AKI. Frail patients were higher in the AKI group, but this result did not reach statistical significance. Interestingly there is no difference in the number of elderly (>65 years) patients between the groups. This is a finding that is not surprising given the younger age of our cohort. The low incidence of AKI in our aSAH patients is in marked contrast to its 75% mortality rate, representing almost 25% of all deaths. The decreased LOS in frail and AKI patients in our study may be explained by increased mortality in these groups. The association of AKI with these findings is widely recognized, with contributions from large studies available in the literature.^{2,4,5,7–9,40,41} Specific examples include AKI's association with worse outcomes at 3 months and death.^4–6,8,9

Predictors for AKI

Worsening clinical severity of aSAH presentation as quantified with HH grade, higher mFI, and age were significantly associated with development of AKI. This is consistent with recent literature, including one study with 260,885 aSAH patients where age, a large number of medical comorbidities including hypertension and diabetes, worse initial GCS and black ethnicity were described as risk factors for development of AKI.^8,9,42 However, our study demonstrates additionally that the strongest statistically significant predictor is HH grade. Gender has shown an inconsistent association with AKI.^8,9 Laboratory abnormalities associated with AKI include increased sCr, mean serum chloride and C-reactive Protein.^7–9,42,43 Our study also shows that admission sCr was not a significant predictor of AKI, strengthening the correlation between HH grade and the development of AKI.

Frailty and AKI in aSAH patients

Poor outcome with AKI is recognized in studies of aSAH patients. However, literature comparing frailty with aSAH is limited and varied, with one report showing frailty is inferior compared to age, and another using surrogates other than the mFI for frailty.^15,44 Our findings show that frailty is inferior to HH grade in predicting AKI. To our knowledge, no clear correlation between AKI and HH grade is described among aSAH patients. However, the literature does show increased complications and worse outcomes with increasing HH grade.^2,4 These findings suggest that although frailty may be applicable to a wide range of disciplines and procedures, more nuanced and specific scales may be necessary in certain disease processes.

Limitations

This is a single center retrospective study and as such has inherent limitations. A relatively small sample size and the reliance on a potentially incomplete electronic medical record are limitations. There are various criteria for AKI, and a broadly accepted version is yet to be defined for this particular patient population. Our clinical data lacked certain variables such as urine output. Although these limitations exist, our study contributes to the literature as the first to explore the utility of frailty as a predictor of development of AKI in the context of aSAH patients.

Conclusion

The development of AKI in aSAH patients is associated with increased mortality. Identification of accurate predictors for AKI in these patients may improve efficiency and outcomes. We demonstrate that frailty does not improve the accuracy of predicting AKI in aSAH patients and that HH grade is a more accurate predictor than mFI. Additional research to identify predictors and therefore guide resource allocation to mitigate poor outcomes is warranted.