A Frailty-integrated Prognostic Prediction Tool for Long-term Outcomes after Shunt Surgery in Idiopathic Normal Pressure Hydrocephalus: Development and Prospective Validation

Abstract

Long-term functional independence after shunt surgery for idiopathic normal pressure hydrocephalus varies widely, particularly in older patients with multiple comorbidities. We aimed to develop and prospectively validate a frailty-integrated prognostic prediction tool to support patient selection and preoperative decision-making. A retrospective cohort of 82 patients with idiopathic normal pressure hydrocephalus who underwent shunt surgery between 2005 and 2014 was used to develop a weighted idiopathic normal pressure hydrocephalus–specific frailty index based on multivariable logistic regression. The index incorporated neurological and non-neurological comorbidities, including dementia, cancer, depression, and movement disorders. The weighted idiopathic normal pressure hydrocephalus–specific frailty index was combined with preoperative modified Rankin Scale scores to estimate the probability of achieving functional independence (modified Rankin Scale ≤2) at 1, 2, and mid- to long-term follow-up (up to 4 years). Prospective validation was performed in an independent cohort of 30 patients treated between 2018 and 2023. The weighted idiopathic normal pressure hydrocephalus–specific frailty index demonstrated improved predictive performance, particularly for mid- to long-term outcomes (2-4 years), compared with conventional frailty indices, with higher correlation coefficients and lower Akaike Information Criterion values. Predicted probabilities of long-term independence declined stepwise with increasing frailty burden. In the validation cohort, predicted and observed outcomes showed good agreement, with adequate calibration confirmed by the Hosmer–Lemeshow test. This frailty-integrated prognostic prediction tool provides individualized estimates of long-term postoperative independence, particularly in the context of frailty and comorbidity burden. By incorporating comorbidity burden into preoperative assessment, the tool supports patient selection and shared decision-making and may facilitate realistic counseling and long-term care planning in older patients.

Introduction

Idiopathic normal pressure hydrocephalus (iNPH) is a treatable disorder predominantly affecting older individuals, and is characterized by the classical triad of gait disturbance, cognitive impairment, and urinary incontinence.^1-3) Cerebrospinal fluid (CSF) shunt surgery is an established and effective treatment, particularly for improving gait and functional mobility. However, despite favorable short-term responses, long-term postoperative independence varies widely among patients, posing a persistent challenge for neurosurgeons in daily practice.

In older patients undergoing shunt surgery, long-term outcomes are influenced not only by disease-specific factors but also by systemic comorbidities and geriatric frailty. Though previous studies, including our own, have identified predictors of short-term postoperative improvement such as preoperative functional status, tap test response, and radiological features, including disproportionately enlarged subarachnoid space hydrocephalus (DESH), these factors alone are insufficient to estimate sustained independence over several years.⁴⁾ As survival after shunt surgery improves, non-neurosurgical conditions such as dementia, malignancy, and cerebrovascular disease increasingly determine long-term functional trajectories.

Frailty has emerged as a clinically relevant construct for risk stratification in older surgical populations. Conventional indices such as the Charlson Comorbidity Index (CCI) and the modified Frailty Index (mFI) provide general estimates of systemic vulnerability but may not adequately capture iNPH-specific conditions that critically affect postoperative independence.^5-7) In particular, comorbid dementia, movement disorders, depression, and cancer—frequently encountered in patients with iNPH—are insufficiently weighted in existing indices, limiting their utility in preoperative counseling.

To address this gap, we developed a frailty-integrated prognostic prediction tool (PPT) that combines a weighted iNPH-specific frailty index (W-iNPH-FI) with preoperative functional status to estimate the probability of achieving long-term postoperative independence. In contrast to our previous work, which focused on short-term postoperative outcomes and surgical results, this study emphasizes long-term functional trajectories and the role of comorbidities in determining sustained independence after shunt surgery.⁴⁾ By integrating frailty into preoperative risk stratification, this study aims to provide a practical tool to support surgical candidacy assessment, patient counseling, and shared decision-making in older patients with iNPH.

Materials and Methods

Study design and patients

We conducted a retrospective cohort study of 82 patients with iNPH who underwent ventriculoperitoneal or lumboperitoneal shunt surgery between 2005 and 2014. Inclusion criteria were based on Japanese iNPH guidelines,²⁾ including DESH-compatible imaging and ≥10% clinical improvement in the CSF tap test. Baseline patient characteristics are summarized in Table 1. All patients had at least 3 years of postoperative follow-up.

Table 1.

Baseline Characteristics

Demographic data	Mean values	Range
Age (years)	77.26	64-89
M:F	2.04	55:27
Factors
Severity (mRS)	3.5	2-5 (0-6)
Duration (months)	15.91	2-60
Follow-up period (months)	58.57	36-132
Improvement in tap test (0, 1, 2)	1.45	1-2 (0: no improvement, 1: slight improvement, 2: ≥10% improvement)
DESH (0, 1, 2)	1.46	1-2 (0: non-DESH, 1: incomplete DESH, 2: complete DESH)
Indexes
Charlson Comorbidity Index	1.57	0-5 (0-39)
modified Frailty Index	2.3	0-5 (0-11)
iNPH Frailty Index	3.06	0-6 (0-15)
Evaluation methods
Preoperative mRS	3.5	2-5 (0-6)
Preoperative iNPHGS-GD	2.59	2-4 (0-4)
Preoperative iNPHGS-cognitive dysfunction	0.71	0-2 (0-4)
Preoperative iNPHGS-urinary incontinence	1.33	0-3 (0-4)

DESH: disproportionately enlarged subarachnoid space hydrocephalus; iNPH: idiopathic normal pressure hydrocephalus; iNPHGS-CD: cognitive domain of iNPH grading scale; iNPHGS-GD: gait domain of iNPH grading scale; iNPHGS-UD: urinary domain of iNPH grading scale; mRS: modified Rankin Scale

To validate the prediction model, we prospectively evaluated a separate cohort of 30 patients diagnosed between 2018 and 2023 who underwent shunting and were followed for 3 years. Preoperative informed consent and institutional review board approval were obtained for all cases.

Development of the prediction tool

We constructed three frailty indices: the CCI⁶⁾ (Supplementary Table S1), the mFI⁷⁾ (Supplementary Table S2), and the newly designed iNPH Frailty Index (iNPH-FI) (Supplementary Table S3), which added four iNPH-relevant conditions (dementia, cancer, depression, and movement disorders) to the standard mFI.^8,9)

The W-iNPH-FI (Table 2) was developed using logistic regression to assign variable weights to each comorbidity based on its impact on 3-year postoperative modified Rankin Scale (mRS). Dementia and cancer were weighted highest due to their strong negative influence on outcomes.

Table 2.

Weighted iNPH-Specific Frailty Index (W-iNPH-FI)

	Condition	Coding	Score
1	DM	Insulin dependent/Non-insulin dependent DM	1
2	Functional status	Partially dependent or total dependent	1
3	Respiratory problems	COPD, current pneumonia	1
4	Congestive heart failure	Congestive heart failure	1
5	Myocardial infarction	Previous myocardial infarction	1
6	Other cardiac problems	Previous percutaneous coronary intervention, coronary surgery, angina	1
7	Hypertension	Hypertension requiring medication	1
8	Peripheral vascular disease	Peripheral vascular disease or resting pain	1
9	Impaired sensorium	Impaired sensorium	1
10	Cerebrovascular disease	TIA or cerebrovascular accident	1
11	CVD with neurological deficit	CVD with deficit	1
12	Dementia	AD, DLB, VaD, etc.	3
13	Movement disorder	PD, PSP, etc.	1
14	Cancer	Cancer with metastases or without metastasis	4
15	Depression	Depression requiring medication	1

AD: Alzheimer disease; COPD: chronic obstructive pulmonary disease; CVD: cerebrovascular disease; DLB: dementia with Lewy bodies; DM: diabetes mellitus; iNPH: idiopathic normal pressure hydrocephalus; PD: Parkinson disease; PSP: progressive supranuclear palsy; TIA: transient ischemic attack; VaD: vascular dementia

The W-iNPH-FI and preoperative mRS were incorporated into a logistic regression model to predict the probability of achieving functional independence (mRS ≤2) at 1, 2, and 3 years postoperatively. This model was formatted into a clinical prediction table (PPT) for practical application.

Model evaluation and prospective validation

Model performance was assessed using the Akaike Information Criterion (AIC), R-squared statistics, and Hosmer–Lemeshow goodness-of-fit tests.¹⁰⁾ AIC was used to compare model fit, as it is widely applied in predictive model evaluation. The number of non-independent cases (mRS ≥3) in the retrospective cohort was 33 at 1 year, 39 at 2 years, and 48 at 4 years. These event counts were considered sufficient for model development. The derived PPT was then applied to the prospective validation cohort to compare predicted versus actual outcomes across 3 years. Discrepancies due to unpredictable events (e.g., cancer recurrence, coronavirus disease 2019 pneumonia, and trauma) were noted.

Ethical approval

This study was conducted in accordance with good clinical practice guidelines and the World Medical Association's Declaration of Helsinki (2002), and was registered with the Ethics Committee of the Kitami Medical Association and the Institutional Review Board (No. 15-B-129).

Results

Patient characteristics

A total of 82 patients were included in the retrospective derivation cohort (mean age, 77.3 years; 67% male), with a median follow-up period of 58.6 months (range, 36-132 months). Baseline demographic and clinical characteristics are summarized in Table 1.

The mean preoperative mRS score was 3.5. Most patients demonstrated positive responses to the CSF tap test and exhibited typical imaging findings of DESH.

The mean CCI, mFI, and iNPH-FI scores were 1.57, 2.3, and 3.06, respectively.

Frailty and comorbidity analysis

The newly constructed iNPH-FI comprised 15 frailty-related items, including dementia, malignancy, depression, and movement disorders (Supplementary Table S3). Multivariable logistic regression analysis was performed to evaluate the association between each frailty component and poor functional outcome at 3 years (Supplementary Table S4).

Based on the regression coefficients, weighted scores were assigned to each component, resulting in the W-iNPH-FI. Dementia and malignancy showed the strongest associations with unfavorable outcomes and were assigned the highest weights (3 and 4 points, respectively; Table 1, Supplementary Table S5).

The W-iNPH-FI demonstrated a stronger correlation with CCI compared with mFI or the unweighted iNPH-FI (R² = 0.76 vs. 0.38 and 0.70, respectively; Figure 1, Supplementary Table S6).

Figure 1.

R squared values between indices and CCI.

Correlation between the Charlson Comorbidity Index (CCI) and three frailty indices in patients with idiopathic normal pressure hydrocephalus.

The weighted iNPH-specific frailty index (W-iNPH-FI) demonstrated the strongest correlation with CCI (R² = 0.76), compared with the unweighted iNPH Frailty Index (iNPH-FI; R² = 0.70) and the modified Frailty Index (mFI; R² = 0.38). These findings indicate that the W-iNPH-FI more accurately reflects comorbidity burden relevant to long-term postoperative outcomes.

CCI, Charlson Comorbidity Index; iNPH, idiopathic normal pressure hydrocephalus; mFI, modified Frailty Index; W-iNPH-FI, weighted iNPH Frailty Index

Model performance and development of the PPT

In multivariable logistic regression models predicting functional independence (mRS ≤2) at 3 years after shunt surgery, the model incorporating W-iNPH-FI showed superior goodness of fit, with the lowest AIC of 100.3, compared with models using CCI (AIC = 108.4), mFI (AIC = 117.5), or unweighted iNPH-FI (AIC = 113.3) (Figure 2, Table 3).

Figure 2.

AIC values for four logistic regression models.

Comparison of Akaike Information Criterion (AIC) values among logistic regression models predicting long-term postoperative independence (modified Rankin Scale ≤2) at 3 years after shunt surgery.

The model incorporating the weighted iNPH Frailty Index (W-iNPH-FI) and preoperative modified Rankin Scale (mRS) achieved the lowest AIC value (100.3), indicating superior model performance compared with models using the Charlson Comorbidity Index (CCI), modified Frailty Index (mFI), or unweighted iNPH Frailty Index.

AIC, Akaike Information Criterion; CCI, Charlson Comorbidity Index; iNPH, idiopathic normal pressure hydrocephalus; mFI, modified Frailty Index; mRS, modified Rankin Scale; W-iNPH-FI, weighted iNPH Frailty Index

Table 3.

AIC Values Comparison of Prognostic Models

	1 year	2 years	3 years
CCI	116.2	117.5	108.4
mFI	102.9	119.2	117.5
iNPH-FI	106.9	117.8	113.3
W-iNPH-FI	108.0	110.6	100.3

AIC: Akaike Information Criterion; CCI: Charlson Comorbidity Index; iNPH-FI: idiopathic normal pressure hydrocephalus–specific frailty index; mFI: modified Frailty Index; W-iNPH-FI: weighted iNPH-specific frailty index

Based on the regression model including W-iNPH-FI and preoperative mRS, a PPT was constructed to estimate the probability of achieving mRS ≤2 at 1, 2, and 3 years after surgery (Supplementary Tables S7-S9).

Predicted probabilities of long-term functional independence declined progressively with increasing frailty burden.

The number of non-independent cases (mRS ≥3) increased over time, from 33 at 1 year to 48 at 4 years, reflecting progressive functional decline. For example, among patients with a preoperative mRS score of 3, the predicted probability of achieving mRS ≤2 at 3 years was 0.979 for those with a W-iNPH-FI score of 0, whereas it decreased to 0.029 for those with a score of 8 (Table 4). The distribution of mRS scores at baseline and at each follow-up time point is summarized in Supplementary Table S10.

Table 4.

Predicted Probability of mRS ≤2 at 3 Years

		W-iNPH-FI score
		0	1	2	3	4	5	6	7	8
Preoperative mRS	2	0.996	0.99	0.976	0.941	0.865	0.718	0.504	0.288	0.139
	3	0.979	0.95	0.883	0.75	0.545	0.323	0.16	0.071	0.029
	4	0.899	0.78	0.586	0.361	0.184	0.082	0.035	0.014	0.006

mRS: modified Rankin Scale; W-iNPH-FI: weighted idiopathic normal pressure hydrocephalus–specific frailty index

Prospective validation

The PPT was prospectively applied to an independent validation cohort of 30 patients. The observed proportions of patients achieving functional independence (mRS ≤2) at 3 years closely aligned with the model-predicted probabilities across most subgroups (Supplementary Table S11-S13).

Supplementary Table S14 summarizes patients who remained functionally dependent (mRS ≥3) at 3 years after shunt surgery. These patients were predominantly distributed in higher W-iNPH-FI score categories, particularly among those with preoperative mRS scores of 3 or 4.

Overall concordance between predicted and observed outcomes confirmed the validity of the model. Notably, most discrepancies were attributable to severe, unpredictable new-onset comorbidities, such as newly diagnosed malignancies or cerebrovascular events unrelated to iNPH. The Hosmer–Lemeshow test demonstrated good calibration between predicted and observed outcomes (p = 0.48).

Discussion

This study developed and prospectively validated a frailty-integrated PPT for patients with iNPH undergoing shunt surgery. The key finding is that the W-iNPH-FI–based model demonstrated improved predictive performance, particularly for mid- to long-term outcomes (2-4 years), compared with conventional frailty indices.

Importantly, the superiority of the W-iNPH-FI–based model was not evident at 1 year, but became more pronounced over longer follow-up periods. This suggests that short-term outcomes after shunt surgery are primarily influenced by surgical responsiveness and CSF dynamics, whereas long-term outcomes are more strongly affected by frailty and comorbidity burden.

These findings have important clinical implications. Conventional decision-making in iNPH has largely focused on short-term responsiveness to CSF drainage tests. However, our results indicate that long-term functional independence is more closely associated with systemic vulnerability, as captured by frailty indices. Therefore, integrating frailty assessment into preoperative evaluation may improve patient selection and facilitate more realistic counseling regarding long-term outcomes.

Another strength of this study is the prospective validation of the proposed model. Though many previous prognostic studies have relied solely on retrospective data, our model demonstrated consistent performance in an independent validation cohort. This supports its robustness and potential applicability in real-world clinical settings.

Previous studies have identified individual predictors of outcome in iNPH, such as age, comorbidities, and imaging findings. However, these factors have rarely been integrated into a comprehensive prediction model. This study extends prior work by incorporating multiple comorbid conditions into a weighted frailty index, thereby providing a more holistic assessment of patient vulnerability.

Taken together, our findings suggest that the clinical course of iNPH is determined not only by disease-specific factors but also by the broader context of patient frailty. This may explain why some patients initially respond to shunt surgery but subsequently experience functional decline over time.

This study has several limitations. First, it was conducted at a single center with a moderate sample size. Second, although prospective validation was performed, external validation in larger and more diverse populations is warranted. Third, the model does not account for all possible factors influencing long-term outcomes, such as postoperative management and social support.

In conclusion, we developed and prospectively validated a frailty-integrated PPT for patients with iNPH. The model provides improved prediction of mid- to long-term functional outcomes and highlights the importance of frailty in determining postoperative prognosis. This approach may enhance clinical decision-making and support individualized patient care.

Future directions

Future studies should focus on external validation of the proposed model in larger, multicenter populations and on its integration into clinical workflows. Development of user-friendly tools, such as web-based or electronic medical record–integrated calculators, may further enhance its applicability in routine practice. In addition, incorporating longitudinal changes in frailty may improve prediction accuracy and enable more adaptive, patient-centered decision-making.

Conclusions

We developed and prospectively validated a frailty-integrated PPT for patients with iNPH undergoing shunt surgery. By combining a W-iNPH-FI with preoperative functional status, the tool provides individualized estimates of long-term postoperative independence.

Unlike conventional approaches that emphasize on short-term surgical outcomes, this prognostic framework highlights the critical role of comorbidities and frailty in determining long-term functional trajectories. The proposed tool is not intended to define rigid surgical indications, but to support patient selection and preoperative decision-making by aligning surgical intervention with realistic long-term expectations.

Incorporation of this frailty-based prediction tool into routine clinical practice may facilitate shared decision-making, improve preoperative counseling, and enhance long-term care planning for older patients with iNPH. Further validation in larger, multicenter cohorts is warranted.

Conflicts of Interest

All authors have no conflict of interest.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this manuscript, the authors used ChatGPT (OpenAI) to assist with language editing, clarification of wording, and improvement of manuscript structure. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.