Abstract
Background
Few studies have evaluated frailty in the setting of aneurysmal subarachnoid hemorrhage (aSAH) using large-scale data. The risk analysis index (RAI) may be implemented at the bedside or assessed retrospectively, differentiating it from other indices used in administrative registry-based research.
Methods
Adult aSAH hospitalizations were identified in the National Inpatient Sample (NIS) from 2015 to 2019. Complex samples statistical methods were performed to evaluate the comparative effect size and discriminative ability of the RAI, the modified frailty index (mFI), and the Hospital Frailty Risk Score (HFRS). Poor functional outcome was determined by the NIS-SAH Outcome Measure (NIS-SOM), shown to have high concordance with modified Rankin Scale scores > 2.
Results
42,300 aSAH hospitalizations were identified in the NIS during the study period. By both ordinal [adjusted odds ratio (aOR) 3.20, 95% confidence interval (CI) 3.05, 3.36, p < 0.001] and categorical stratification [frail aOR 3.59, 95% CI 3.39, 3.80, p < 0.001; severely frail aOR 6.67, 95% CI 5.78, 7.69, p < 0.001], the RAI achieved the largest effect sizes for NIS-SOM in comparison with the mFI and HFRS. Discrimination of the RAI for NIS-SOM in high-grade aSAH was significantly greater than that of the HFRS (c-statistic 0.651 vs. 0.615). The mFI demonstrated the lowest discrimination in both high-grade and normal-grade patients. A combined Hunt and Hess-RAI model (c-statistic 0.837, 95% CI 0.828, 0.845) for NIS-SOM achieved significantly greater discrimination than both the combined models for mFI and HFRS (p < 0.001).
Conclusion
The RAI was robustly associated with poor functional outcomes in aSAH independent of established risk factors.
Keywords: Aneurysm, Database, Frailty, Hemorrhage, Subarachnoid
Introduction
Aneurysmal subarachnoid hemorrhage (aSAH) affects an estimated 5–10 per 100,000 persons annually and confers high rates of morbidity and mortality [1, 2]. Age, in addition to measures of acute neurological condition on presentation, including the Hunt and Hess and World Federation of Neurosurgical Societies grading systems, are well-established metrics for prognostication of outcome [3, 4]. Few studies, however, have investigated the impact of frailty on outcomes following aSAH, and evidence regarding the prognostic significance of baseline frailty status in this disease process, and in cerebrovascular pathologies more broadly, remains limited. Previous investigations have applied the modified frailty index (mFI) and the hospital frailty risk score (HFRS) to large national cohorts of aSAH hospitalizations and have demonstrated significant associations of these indices with poor clinical outcomes, increased lengths of stay, and development of medical and neurological complications [5, 6].
The risk analysis index (RAI) represents yet another quantitative metric by which to assess frailty, but has yet to be evaluated in the aSAH population. In contrast to the mFI and the HFRS, which are predominantly indices of comorbidity burden, the RAI includes only parameters directly related to domains of frailty, defined as a state of diminished physiological reserve, with particular attention paid to functional status [7]. Advantages include wide validation in a variety of surgical contexts, as well as point-of-care testing at the bedside which can guide real-time clinical decision making in addition to retrospective applications with administrative registries [8–10]. This duality represents both a unique attribute of the RAI and a limitation of other existing indices. We hypothesize that the RAI will robustly predict clinical outcomes following aSAH independent of other established prognostic factors, and that the RAI will outperform other existing measures of frailty status and comorbidity burden in a large nationally-representative sample.
Methods
Data source
The National Inpatient Sample (NIS), developed and maintained by the Healthcare Cost and Utilization Project (HCUP), is among the largest publicly accessible inpatient care databases in the United States. Yearly unweighted data approximate 7,000,000 patients, reflecting a 20% stratified sample of all HCUP-participating community hospitals nationally. More information regarding the NIS and data access can be found at: www.hcup-us.ahrq.gov. Given the public accessibility and de-identified nature of the information in this database, this study did not meet the requirements for institutional review board approval. For the same reason, patient consent was neither sought nor required. The data utilized in this analysis are available upon reasonable request of the corresponding author following completion of onboarding and verification procedures as specified by the HCUP, and a comprehensive list of billing codes used to define the clinical variables in this study are included in the manuscript supplement. This manuscript was composed in accordance with Reporting of Studies Conducted Using Observational Routinely-Collected Data (RECORD) guidelines.
Patient selection and cohort development
Adult hospitalizations (> 17 years of age) with primary admission diagnoses of aSAH were identified in the NIS during the period of 2015 (fourth quarter, October through December) to 2019 using International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes. Patients with diagnoses related to trauma or congenital cerebrovascular malformations, elective admissions, and patients not treated by microsurgical clipping or endovascular embolization were excluded from the analysis to eliminate non-aneurysmal causes of SAH. Patient age and sex as well as aneurysm location were identified. Baseline neurological status was evaluated by the Hunt and Hess grade, derived from the previously validated NIS Subarachnoid Hemorrhage Severity Score (NIS-SSS) optimized for utilization in this dataset [11].
Measures of baseline frailty status and comorbidity burden: the risk analysis index, the modified frailty index, and the hospital frailty risk score
The RAI [12] was initially derived from the minimum data set mortality risk index instrument, implemented to predict mortality at 6 months in a Veterans Affairs population of patients older than 65 years, and is based on an accumulating deficits model of frailty quantification. The index is composed of fourteen parameters: age, sex, weight loss, poor appetite, congestive heart failure, dyspnea, renal failure, presence of disseminated cancer, functional status, cognitive decline, and institutional living status at time of admission. Determination of functional status was made using billing codes encompassing impairments to activities of daily living, fall risk assessment, bed ridden status, use of cane or wheelchair, and chronic mobility impairments. The RAI has been previously validated in two forms, the original “clinical RAI” (RAI-C), a weighted calculation based on responses to a patient survey, and subsequently the “administrative RAI” (RAI-A) for application to national registries with less available baseline clinical data, and calculated from variables contained in the VA Surgical Quality Improvement Project and American College of Surgeons-National Surgical Quality Improvement Program databases. Components of the RAI were identified with ICD-10-CM diagnosis codes in accordance with adaptations made for the RAI-A from the original RAI-C and in collaboration with the creators of the index (complete coding presented in Supplemental Table 1). The 5-factor mFI [13] is composed of five factors present on admission: congestive heart failure, diabetes mellitus, chronic obstructive pulmonary disease or recent pneumonia, hypertension, and dependent functional status. Comorbid conditions are weighted equally, each assigned a value of “1”, and are summed to create a composite score. The HFRS [14] was composed through initial clustering analysis identifying a priori designated frailty and disability-related ICD-10-CM codes most highly represented in a cohort of administrative registry patients aged 75 years or older. The effect sizes of each of the 109 parameters composing the HFRS for various adverse events including death, in-patient complications, and hospital length of stay are summed to create a weighted composite score. In preparation for statistical analysis, the NIS aSAH cohort was stratified into three categories of increasing frailty based on varying thresholds within each index [RAI (robust and pre-frail, 0–20; 21–30, frail 21–30, severely frail 31+), mFI (robust and pre-frail 0–1, frail 2, severely frail 3+), HFRS (low risk 0–5, intermediate risk 6–15, high risk 16+)].
Clinical endpoints and statistical analysis
Poor functional outcome, as determined by the NIS Subarachnoid Hemorrhage Outcome Measure (NIS-SOM) [11], was evaluated as the primary clinical endpoint of the analysis. The dichotomous NIS-SOM was previously validated in a national cohort of aSAH hospitalizations and defines poor outcome as discharge disposition to intermediate, long-term care, or skilled nursing facility, in-hospital mortality, or placement of tracheostomy or gastrostomy tube, and was shown to have strong concordance with modified Rankin Scale scores greater than 2 (suggestive of high levels of disability at discharge). In-hospital mortality was evaluated separately as both a secondary endpoint and form of sensitivity testing (as this parameter is contained within the NIS-SOM).
All analyses were performed within a complex samples function with appropriate stratum and cluster variables and discharge weights per HCUP guidelines to account for NIS sampling design and to ultimately yield accurate national estimates. Summary statistics were calculated for baseline clinical characteristics of the aSAH cohort and rates of clinical endpoints were identified within the three frailty categories for each of the three indices evaluated in this analysis. Multivariable logistic regression analysis was performed to evaluate the independent association between each index with clinical endpoints following adjustment for Hunt and Hess grade as well as aneurysm location. Each index was evaluated in its pseudocontinuous, ordinal, and categorical forms. The mFI and HFRS were secondarily evaluated in age-adjusted models, as neither parameter incorporates age into its calculation. For sensitivity testing, all multivariable models were performed following a revision in the stratification of the RAI and mFI to better match the distribution of the HFRS, in which the size of the intermediate risk category is largest, followed by the low risk and high risk categories, respectively. A Bonferroni correction was applied for determination of statistical significance to account for multiple comparisons.
Receiver operating characteristic (ROC) curve analysis was performed to evaluate the discriminative capacity of the three indices for the NIS-SOM. Because frailty must always be considered in conjunction with acute neurological condition, index discrimination was evaluated in sub-groups of high-grade (Hunt and Hess 4 and 5) and normal-grade (Hunt and Hess 1–3) aSAH. Additionally, combined models of Hunt and Hess grade with each index (based on predicted probabilities generated from logistic regression) were evaluated. Generated c-statistics were compared by the DeLong test, computed using MedCalc for Windows, Version 19.4 (MedCalc Software, Ostend, Belgium). All other statistical analyses were performed with IBM SPSS Version 26 software (Armonk, NY).
Preliminary validation of the risk analysis index in the national inpatient sample
Since the RAI has not previously been applied to the NIS, a brief preliminary validation was performed using a broader cohort of 2019 NIS data, from which all hospitalizations in which a medical or surgical procedure was performed in the category of “Central Nervous System or Cranial Nerves” (ICD-10 Procedural Coding System 00XXXXX) were identified. Records without discharge disposition or age were excluded from the analysis. Discrimination of the RAI for routine discharge (to home without services) and in-hospital mortality was assessed. This query of 2019 NIS data yielded 757,730 hospitalizations. The RAI demonstrated an AUC of 0.772 (95% CI 0.770, 0.774; p < 0.001) for routine discharge and an AUC of 0.674 (95% CI 0.672, 0.677; p < 0.001) for mortality.
Results
42,300 aSAH hospitalizations were identified in the NIS during the study period. Baseline clinical characteristics are shown in Table 1 and distribution of clinical endpoints as a function of increasing frailty or risk category is reported in Table 2. By both ordinal [adjusted odds ratio (aOR) 3.20, 95% confidence interval (CI) 3.05, 3.36, p < 0.001] and categorical stratification [frail aOR 3.59, 95% CI 3.39, 3.80, p < 0.001; severely frail 6.67, 95% CI 5.78, 7.69, p < 0.001], the RAI achieved the largest effect sizes for NIS-SOM in comparison with the mFI and HFRS (Table 3). A larger effect size of the RAI was also seen for mortality. Sensitivity testing by re-stratifying the RAI and mFI categorical groupings to match the proportion of patients in the HFRS risk categories confirmed and extended the findings of the primary analytical models (Supplemental Tables 2, 3).
Table 1.
Summary of cohort characteristics for aneurysmal subarachnoid hemorrhage patients
| Characteristic | Total cohort (n = 42,300) |
|---|---|
|
| |
| Age (median years) (IQR) | 57 (47–66) |
| Hunt and Hess Grade (median grade) (IQR) | 3 (1–5) |
| High-Grade aSAH (Hunt and Hess 4 and 5) | 18,585 (43.9) |
| Aneurysm location | 1565 (3.7) |
| Carotid siphon and bifurcation | 1565 (3.7) |
| Middle cerebral artery | 5815 (13.7) |
| Anterior communicating artery | 12,675 (30.0) |
| Posterior communicating artery | 9065 (21.4) |
| Basilar artery | 2240 (5.3) |
| Vertebral artery | 665 (1.6) |
| Unknown or unspecified location | 11,020 (26.1) |
| Frailty and risk indices Risk analysis index | 17 (14–22) |
| 5-factor modified frailty index | 1 (1–1) |
| Hospital frailty risk score | 9 (5–13) |
| Treatment modalities | |
| Endovascular coiling | 31,115 (73.6) |
| Microsurgical clipping | 11,185 (26.4) |
| Clinical endpoints | |
| Poor functional outcome (NIS-SOM) | 23,280 (55.0) |
| In-hospital mortality | 5245 (12.4) |
Values presented as n (%) for categorical parameters and as median [interquartile range (IQR)] for continuous parameters
Table 2.
Clinical outcomes stratified by frailty/risk category of risk analysis index (RAI), modified frailty index (mFI)), and hospital frailty risk score (HFRS)
| Frailty or risk index | Frailty or risk stratification | ||
|---|---|---|---|
|
| |||
| Robust and pre-frail (0–20) | Frail (21–30) | Severely frail (> 30) | |
| RAI | 29,695 (70.2) | 11,110 (26.3) | 1495 (3.5) |
| Poor functional outcome | 13,845 (46.6) | 8255 (74.3) | 1180 (78.9) |
| In-hospital mortality | 2,940 (9.9) | 1970 (17.7) | 335 (22.4) |
| Robust and pre-frail (0–1) | Frail (2) | Severely frail (3–5) | |
| mFI | 32,230 (76.2) | 8330 (19.7) | 1740 (4.1) |
| Poor functional outcome | 16,635 (51.6) | 5385 (64.6) | 1260 (72.4) |
| In-hospital mortality | 3635 (11.3) | 1290 (15.5) | 320 (18.4) |
| Low risk (0–5) | Intermediate risk (6–15) | High risk (> 15) | |
| HFRS | 9855 (23.3) | 26,650 (63.0) | 5760 (13.6) |
| Poor functional outcome | 2445 (24.8) | 16,085 (60.4) | 4725 (80.2) |
| In-hospital mortality | 645 (6.5) | 3985 (15.0) | 605 (10.5) |
Values presented as n (%). Frailty or risk stratification was performed as follows: robust and pre-frail/low risk (RAI 0–20, mFI 0–1, HFRS < 5), frail/intermediate risk (RAI 21–30, mFI 2, HFRS 5–15), and severely frail/high risk (RAI > 30, mFI 3–5, HFRS > 15)
Table 3.
Multivariable analysis—association of risk analysis index, modified frailty index, and hospital frailty risk score with clinical endpoints
| Primary analytical models (aOR, 95% CI) |
Age-adjusted models (aOR, 95% CI) |
||||
|---|---|---|---|---|---|
| RAI | mFI | HFRS | mFI | HFRS | |
|
| |||||
| Poor functional outcome | |||||
| Complete index | 1.12 (1.11, 1.13) | 1.39 (1.35, 1.43) | 1.11 (1.10 1.12) | 1.12 (1.08, 1.15) | 1.11 (1.10, 1.12) |
| Ordinal stratification | 3.20 (3.05, 3.36) | 1.56 (1.52, 1.66) | 2.38 (2.28, 2.48) | 1.21 (1.16, 1.27) | 2.31 (2.21, 2.42) |
| Categorical stratification | |||||
| Non-frail, pre-frail, and low risk (reference) | – | – | – | – | – |
| Frail and intermediate risk | 3.59 (3.39, 3.80) | 1.62 (1.53, 1.72) | 2.50 (2.35, 2.65) | 1.22 (1.14, 1.29) | 2.39 (2.25, 2.55) |
| Severely frail and high risk | 6.67 (5.78, 7.69) | 2.38 (2.10, 2.69) | 5.49 (5.02, 6.00) | 1.45 (1.28, 1.65) | 5.21 (4.74, 5.72) |
| In-hospital mortality | |||||
| Full index | 1.05 (1.04, 1.06) | 1.13 (1.09, 1.17) | 0.94 (0.93, 0.95) | 1.03 (0.99, 1.07)N.S | 0.94 (0.93, 0.95) |
| Ordinal stratification | 1.69 (1.60, 1.78) | 1.23 (1.17, 1.30) | 0.65 (0.61, 0.69) | 1.10 (1.05, 1.17) | 0.63 (0.59, 0.67) |
| Categorical stratification | |||||
| Non-frail, pre-frail, and low risk (reference) | – | – | – | – | – |
| Frail and intermediate risk | 1.68 (1.57, 1.79) | 1.26 (1.17, 1.36) | 0.98 (0.89, 1.08)N.S | 1.12 (1.04, 1.21) | 0.94 (0.86, 1.04)N.S |
| Severely frail and high risk | 2.88 (2.50, 3.31) | 1.45 (1.27, 1.66) | 0.43 (0.38, 0.49) | 1.19 (1.04, 1.36) | 0.40 (0.35, 0.46) |
Effect sizes for frailty or risk presented as adjusted odds ratios (aOR) with 95% confidence intervals (CI) following multivariable logistic regression analysis adjusting for Hunt and Hess grade and aneurysm location in primary analytical models. Age-adjusted models (including age in addition to Hunt and Hess grade and aneurysm location as a model covariate) were performed for the mFI and HFRS as neither of these indices incorporates age
Denotes non-significant association following Bonferroni correction for multiple statistical comparisons (values without this notation are statistically significant)
Discrimination of the RAI for NIS-SOM in high-grade aSAH was significantly greater than that of the HFRS (c-statistic 0.651 vs. 0.615, p = 0.026), however there was no difference in normal-grade patients (Fig. 1, Supplemental Table 4). The mFI demonstrated the lowest discrimination in comparison with the two other indices for both high-grade and normal-grade patients. The combined Hunt and Hess-RAI model (c-statistic 0.837, 95% CI 0.828, 0.845) for NIS-SOM achieved significantly greater discrimination than both the combined models for mFI and HFRS (p < 0.001) (Fig. 1, Supplemental Table 4).
Fig. 1.
Receiver operating characteristic curve analysis—discrimination of risk analysis index, modified frailty index, and hospital frailty risk score for poor functional outcome. C-statistics presented with 95% confidence intervals and compared by De Long test. Normal-grade aSAH (Hunt and Hess 1–3) [RAI > mFI (p < 0.001); RAI, HFRS (p = 0.1923); HFRS > mFI (p < 0.001)]. High-grade aSAH (Hunt and Hess 4–5) [RAI > mFI (p < 0.001); RAI > HFRS (p = 0.026); HFRS > mFI (p < 0.001)]. Combined Hunt and Hess Model [Hunt and Hess + RAI > Hunt and Hess + HFRS (p < 0.001); Hunt and Hess + RAI > Hunt and Hess + mFI (p < 0.001); Hunt and Hess + HFRS > Hunt and Hess + mFI (p < 0.001)]
Discussion
The present analysis of over 42,000 aSAH hospitalizations using population-level data from a national registry is the first to evaluate the prognostic utility of the RAI in a cerebrovascular disease population, and is the first to apply this index to the NIS. The RAI was robustly associated with poor functional outcome (the primary clinical endpoint of this analysis), independent of established predictors including Hunt and Hess grade and aneurysm location. Severely frail patients, as categorized by the RAI, were nearly 7 times more likely to experience poor outcomes in comparison with frail and non-frail counterparts. Evaluation of three indices chosen for comparative analysis determined that the RAI was superior to the mFI and HFRS, two other indices that have been utilized in the literature to quantify frailty, in terms of effect sizes and discrimination for the primary clinical endpoint. Concomitant consideration of the RAI and Hunt and Hess grade was shown to have favorable prognostic utility, as a combined model achieved a c-statistic greater than 0.80, indicative of excellent discrimination.
The mFI and HFRS were chosen for comparative evaluation against the RAI as they have been recently applied to aSAH populations [5, 6]. These indices are primarily comorbidity metrics which lack fidelity to a broader concept of frailty beyond this singular domain. The ideal index represents a reliable clinical tool that is simple, concise, and can be rapidly calculated in real time at the patient bedside or applied retrospectively (in analyses such as that of the present study). The mFI fulfills the criterion for ease of use, but is limited in its lack of weighting of individual components and binary approach to functional status assessment. The HFRS is impractical for real-time clinical use due to an excessive number of variables required for calculation (greater than 100 parameters) and inclusion of data unrelated to frailty assessment (such as hyponatremia). Additionally, ideal frailty tools account for increasing patient age as physiological reserve and chronological age cannot be delinked from one another [15], however both indices lack this component.
Previous critics have elucidated methodological flaws pertaining to the initial validation of the HFRS [15–18], the discussion of which is beyond the scope of this study, however, sensitivity evaluation of our secondary clinical endpoint demonstrated poor discrimination and an inverted effect size of the HFRS for in-hospital mortality (increasing HFRS was significantly associated with decreased mortality), raising significant concerns regarding the utilization of this index in cerebrovascular patients. A separate aSAH NIS analysis was recently published evaluating the relationship between the Johns Hopkins Frailty Score and clinical outcomes [19]. Like the HFRS, the index is composed of numerous “frailty-defining diagnoses” (most of which are also inappropriate metrics of frailty), and, like the HFRS, the same inverse relationship was demonstrated between increasing scale score and decreased mortality. Taken together, these studies illustrate the importance of careful ICD code selection when constructing an administrative registry-compatible frailty index and that quantifying frailty requires a theoretical model prior to a statistical model [18]. Beyond phenotypic assessment, parameters such as temporalis muscle thickness and sarcopenia are currently being evaluated as surrogates for frailty in aSAH [20].
The primary limitation of this analysis is its observational retrospective design and utilization of ICD billing codes to compose the NIS version of the RAI. Although billing codes are unable to completely capture the nuance and complexity of a frailty phenotype, adaptation of existing indices has previously been undertaken (such as the conversion of the RAI-C to the RAI-A for use with administrative databases, analogous to our methodology). Additionally, important prognostic data such as aneurysm recurrence, morphology, specific location, and comprehensive radiographic parameters, are absent from this analysis and therefore could not serve as adjusting factors in multivariable analysis. However, acute neurological condition, arguably the most important covariate, was accounted for using a validated index. Moreover, only treated patients were included in this analysis to isolate aneurysmal causes of SAH, however this would exclude many high-severity aSAH patients for whom treatment was not offered, as well as those who died prior to arrival at the hospital, and must be considered a source of bias. Finally, prior to this study, the RAI had yet to be applied to NIS data and therefore has not undergone rigorous validation and calibration. However, we provide a brief preliminary validation in the manuscript supplement. A more general validation and recalibration of the RAI for use in the NIS is currently being undertaken by this group of authors.
Conclusion
The RAI was robustly associated with poor functional outcome independent of established predictors, including Hunt and Hess grade and aneurysm location, in this large national cohort of 42,300 treated aSAH patients. The RAI had superior discrimination for the primary clinical endpoint when compared to the mFI and HFRS, and the effect sizes were far greater for RAI than for mFI or HFRS, when comparing equivalent frailty groupings. Combining RAI score with Hunt and Hess grade was shown to have excellent prognostic utility, with a c-statistic greater than 0.80. We propose that the RAI, which may be rapidly implemented at the bedside, be routinely considered for prognostication and risk stratification in addition to Hunt and Hess grade. Prospective clinical data, currently being collected at our institution for the past two years, will provide further evidence and validation of the RAI validation.
Supplementary Material
Funding
No funding was secured for this study.
Footnotes
Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s00415-023-11805-z.
Declarations
Conflict of interest Authors disclose no conflicts of interest or funding.
Ethics approval Informed consent was not sought for the present study due to the de-identified nature of the publicly available data provided by the Healthcare Cost and Utilization Project.
Data availability
The data utilized in this analysis are available upon reasonable request of the corresponding author following completion of onboarding and verification procedures as specified by the Healthcare Cost and Utilization Project, and a comprehensive list of billing codes used to define the clinical variables in this study are included in the manuscript supplement.
References
- 1.Linn FH, Rinkel GJ, Algra A, van Gijn J (1996) Incidence of subarachnoid hemorrhage: role of region, year, and rate of computed tomography: a meta-analysis. Stroke 27(4):625–629 [DOI] [PubMed] [Google Scholar]
- 2.Long B, Koyfman A, Runyon MS (2017) Subarachnoid hemorrhage: updates in diagnosis and management. Emerg Med Clin N Am 35(4):803–824 [DOI] [PubMed] [Google Scholar]
- 3.Shimamura N, Munakata A, Ohkuma H (2011) Current management of subarachnoid hemorrhage in advanced age. Acta Neurochir Suppl 110(Pt 2):151–155 [DOI] [PubMed] [Google Scholar]
- 4.Mahta A, Murray K, Reznik ME, Thompson BB, Wendell LC, Furie KL (2021) Early neurological changes and interpretation of clinical grades in aneurysmal subarachnoid hemorrhage. J Stroke Cerebrovasc Dis 30(9):105939 [DOI] [PubMed] [Google Scholar]
- 5.Dicpinigaitis AJ, McIntyre MK, Al-Mufti F et al. (2022) Association of baseline frailty status with clinical outcome following aneurysmal subarachnoid hemorrhage. J Stroke Cerebrovasc Dis 31(5):106394 [DOI] [PubMed] [Google Scholar]
- 6.Koo AB, Elsamadicy AA, Renedo D, Sarkozy M, Sherman J, Reeves BC, Havlik J, Antonios J, Sujijantarat N, Hebert R, Malhotra A, Matouk C (2023) Higher Hospital Frailty Risk Score is associated with increased complications and healthcare resource utilization after endovascular treatment of ruptured intracranial aneurysms. J Neurointerv Surg 15(3):255–261 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hall DE, Arya S, Schmid KK et al. (2017) Development and initial validation of the risk analysis index for measuring frailty in surgical populations. JAMA Surg 152(2):175–182 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Shah R, Borrebach JD, Hodges JC et al. (2020) Validation of the risk analysis index for evaluating frailty in ambulatory patients. J Am Geriatr Soc 68(8):1818–1824 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Varley PR, Borrebach JD, Arya S et al. (2021) Clinical utility of the risk analysis index as a prospective frailty screening tool within a multi-practice, multi-hospital integrated healthcare system. Ann Surg 274(6):e1230–e1237 [DOI] [PubMed] [Google Scholar]
- 10.Hall DE, Arya S, Schmid KK et al. (2017) Association of a frailty screening initiative with postoperative survival at 30, 180, and 365 days. JAMA Surg 152(3):233–240 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Washington CW, Derdeyn CP, Dacey RG Jr, Dhar R, Zipfel GJ (2014) Analysis of subarachnoid hemorrhage using the Nationwide Inpatient Sample: the NIS-SAH Severity Score and Outcome Measure. J Neurosurg 121(2):482–489 [DOI] [PubMed] [Google Scholar]
- 12.Arya S, Varley P, Youk A et al. (2020) Recalibration and external validation of the risk analysis index: a surgical frailty assessment tool. Ann Surg 272(6):996–1005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Subramaniam S, Aalberg JJ, Soriano RP, Divino CM (2018) New 5-factor modified frailty index using American College of Surgeons NSQIP data. J Am Coll Surg 226(2):173–181.e178 [DOI] [PubMed] [Google Scholar]
- 14.Gilbert T, Neuburger J, Kraindler J et al. (2018) Development and validation of a Hospital Frailty Risk Score focusing on older people in acute care settings using electronic hospital records: an observational study. Lancet 391(10132):1775–1782 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Dent E, Martin FC, Bergman H, Woo J, Romero-Ortuno R, Walston JD (2019) Management of frailty: opportunities, challenges, and future directions. Lancet 394(10206):1376–1386 [DOI] [PubMed] [Google Scholar]
- 16.Estes E, Rumalla K, Schmidt MH, Bowers C (2023) Correspondence on “Higher hospital frailty risk score is associated with increased complications and healthcare resource utilization after endovascular treatment of ruptured intracranial aneurysms” by Koo et al. J Neurointerv Surg 15(3):305–306 [DOI] [PubMed] [Google Scholar]
- 17.Bowers CA, Varela S, Kazim SF, Gurgel R (2022) Letter: association of patient frailty with vestibular schwannoma resection outcomes and machine learning development of a vestibular schwannoma risk stratification score. Neurosurgery 91(5):e139–e140 [DOI] [PubMed] [Google Scholar]
- 18.Reitz KM, Arya S, Hall DE (2022) Quantifying frailty requires a conceptual model before a statistical model. JAMA Surg 157(11):1065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Guo Y, Wu H, Sun W, Hu X, Dai J (2022) Effects of frailty on postoperative clinical outcomes of aneurysmal subarachnoid hemorrhage: results from the National Inpatient Sample database. BMC Geriatr 22(1):460. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Lim JX, Lim YG, Kumar A et al. (2022) Relevance of presenting risks of frailty, sarcopaenia and osteopaenia to outcomes from aneurysmal subarachnoid haemorrhage. BMC Geriatr 22(1):333. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data utilized in this analysis are available upon reasonable request of the corresponding author following completion of onboarding and verification procedures as specified by the Healthcare Cost and Utilization Project, and a comprehensive list of billing codes used to define the clinical variables in this study are included in the manuscript supplement.