Early Postoperative Prognostic Nutritional Index, Dementia, and Walking Independence at Discharge in Patients with Hip Fracture: A Multicenter Observational Study

ABSTRACT

Objectives:

This study investigated whether early postoperative Prognostic Nutritional Index (PNI), dementia, and physical function are associated with walking independence at discharge in patients with hip fracture.

Methods:

This multicenter retrospective observational study included 173 complete-case patients aged 65 years or older who had been independent ambulators before hip fracture surgery. The primary outcome was walking independence at discharge. Early postoperative PNI was calculated from serum albumin concentration and total lymphocyte count at 1 week postoperatively. Physical function was assessed using single-leg standing time and weight-bearing ratio on the non-affected side. Multivariable logistic regression, receiver operating characteristic (ROC) analysis, and sensitivity analyses were performed.

Results:

Among 173 patients, 85 were independent walkers and 88 were non-independent walkers at discharge. Multivariable logistic regression analysis showed that lower PNI at 1 week postoperatively [odds ratio (OR) per 1-point increase, 0.86; 95% confidence interval (CI), 0.80–0.92; P <0.001] and dementia (OR, 6.32; 95% CI, 2.75–15.87; P <0.001) were independently associated with non-independent walking at discharge. These associations remained consistent in sensitivity analyses including adjustment for log-transformed C-reactive protein and a parsimonious model including age, dementia, and PNI. The area under the ROC curve for PNI was 0.66 (95% CI, 0.58–0.74).

Conclusions:

Early postoperative PNI and dementia were independently associated with non-independent walking at discharge. PNI may serve as an adjunctive marker for early risk stratification in patients with hip fracture.

INTRODUCTION

Hip fracture is a major injury in older adults and is associated with increased mortality and substantial declines in mobility and activities of daily living.¹⁾ Regaining walking independence by discharge is a key determinant of subsequent functional recovery and long-term outcomes.

Early postoperative nutritional status has recently gained attention as a potentially modifiable factor influencing wound healing and muscle recovery.²⁾ Identifying simple and feasible nutritional indicators for early risk stratification may facilitate timely rehabilitation planning.

The Prognostic Nutritional Index (PNI), calculated from serum albumin concentration and peripheral lymphocyte count, reflects both nutritional and immunological status. It has been widely used as a prognostic marker in oncology and perioperative settings.³⁾ However, few multicenter studies have examined the association between early postoperative PNI and walking independence at discharge while simultaneously adjusting for dementia status and physical function.

Low PNI in the early postoperative period may reflect a combination of nutritional, inflammatory, and immunological conditions after surgery. Previous experimental studies have suggested that inflammatory pathways, including interleukin-6-related mechanisms, may be involved in impaired muscle protein metabolism.⁴⁾ In this study, we examined the association between early postoperative PNI and walking independence at discharge. Inflammatory cytokines and muscle metabolism were not directly measured in this study, and these mechanisms should be regarded only as a possible biological background.

MATERIALS AND METHODS

Study Design and Participants

This multicenter retrospective observational study was conducted at three institutions between April 2023 and March 2025. Patient inclusion was based on the following criteria: (1) age 65 years or older; (2) surgical treatment for hip fracture; (3) independent ambulation prior to injury; (4) availability of blood test data at 1 week postoperatively; and (5) assessable walking ability at discharge. Dementia was not an inclusion criterion but was adjusted for in the analysis. Patients were excluded according to the following criteria: (1) severe perioperative complications significantly affecting rehabilitation; and (2) history of previous ipsilateral or contralateral hip fracture. Severe perioperative complications were defined as complications that substantially interrupted or prevented postoperative rehabilitation, such as reoperation, severe infection requiring additional treatment, severe cardiopulmonary complications, or postoperative intensive care management.

The sample size was determined by the number of eligible patients treated during the study period at the participating institutions. Of the 180 eligible patients, 7 patients were excluded because of missing data in variables required for the main analyses. Therefore, the final complete-case analysis population was 173 patients. The patient selection process is shown in Fig. 1. This study was approved by the Ethics Committee of Chiba University Graduate School of Medicine (Approval No. 2888). Written informed consent was obtained from all patients or their legal representatives.

Fig. 1.

 Flow diagram of patient selection.

OUTCOME MEASURES

Walking independence at discharge was defined as the ability to walk continuously for at least 50 m without supervision or physical assistance, with the use of a cane permitted.⁵⁾ This threshold reflects a practical walking distance required for basic mobility within home and community settings.

Primary Explanatory Variable

PNI at 1 week postoperatively was calculated as: PNI = 10 × Serum albumin (g/dL) + 0.005 × Total lymphocyte count (/µL). Blood samples were obtained 7 ± 1 days after surgery.

Covariates

Covariates included age, sex, surgical procedure [internal fixation (IF) or bipolar hemiarthroplasty (BHA)], presence of dementia, hemoglobin level at 1 week postoperatively, young adult mean percentage of lumbar spine bone mineral density (lumbar spine YAM), and physical function (single-leg standing time and weight-bearing ratio on the non-affected side). Fracture type is presented in Table 1 as a baseline characteristic, whereas surgical procedure was used as a covariate in the multivariable model because it may reflect both fracture characteristics and postoperative rehabilitation course.

Table 1.  Baseline characteristics of the complete-case analysis population.

Variable	All patients (n = 173)	Independent walking group (n = 85)	Non-independent walking group (n = 88)	P value
Age, years	79.9 ± 9.6	77.2 ± 10.7	82.5 ± 7.7	<0.001
Female sex	157 (90.8)	81 (95.3)	76 (86.4)	0.064
BMI, kg/m2	20.0 ± 3.2	20.3 ± 2.6	19.6 ± 3.6	0.497
Lumbar spine YAM, %	72.1 ± 21.3	78.3 ± 22.5	66.1 ± 18.4	<0.001
Fracture type				0.102
Femoral neck fracture	86 (49.7)	38 (44.7)	48 (54.5)
Trochanteric fracture	87 (50.3)	47 (55.3)	40 (45.5)
Surgical procedure				0.109
BHA	84 (48.6)	36 (42.4)	48 (54.5)
IF	89 (51.4)	49 (57.6)	40 (45.5)
Dementia	43 (24.9)	11 (12.9)	32 (36.4)	<0.001
PNI a	35.3 ± 8.1	37.4 ± 9.8	33.2 ± 5.4	<0.001
Hemoglobin, g/dL a	9.8 ± 1.1	10.1 ± 1.1	9.6 ± 1.0	<0.001
CRP, mg/dL a	2.29 [0.99–4.34]	1.47 [0.58–3.06]	3.25 [2.19–4.80]	<0.001
Single-leg standing time, s b	0.0 [0.0–7.1]	3.4 [0.0–13.0]	0.0 [0.0–0.0]	<0.001
Weight-bearing ratio, % c	77.3 ± 35.1	85.7 ± 29.1	69.2 ± 38.6	<0.001
Discharge destination				<0.001
Home	52 (30.1)	42 (49.4)	10 (11.4)
Rehabilitation hospital	104 (60.1)	41 (48.2)	63 (71.6)
Long-term care facility	17 (9.8)	2 (2.4)	15 (17.0)

Data are based on the complete-case analysis population (n = 173). Values are presented as mean ± standard deviation for variables with approximately normal distributions and as median [interquartile range] for variables with skewed distributions. Categorical variables are expressed as number (percentage). Between-group comparisons were performed using Student's t-test or the Mann–Whitney U test for continuous variables, and the chi-square test or Fisher's exact test for categorical variables, as appropriate. P <0.05 was considered statistically significant.

^a Recorded at 1 week postoperatively; ^b Non-affected side; ^c Non-affected limb.

Dementia was defined based on a documented diagnosis in the medical records before or at the time of admission. Postoperative delirium alone was not included in the definition of dementia.

Physical Function Assessment

Weight-bearing ratio on the non-affected side was assessed at 1 week postoperatively using two digital scales. Patients stood with one foot on each scale, and the load on the non-affected side was recorded under therapist supervision. The weight-bearing ratio was calculated as the load on the non-affected side divided by body weight and expressed as a percentage. Measurements were performed twice, and the average value of the two trials was used for analysis. The same basic measurement procedure was used across the participating institutions.

Single-leg standing time on the non-affected side was also assessed at 1 week postoperatively under therapist supervision. Patients were instructed to stand on the non-affected limb for as long as possible while maintaining safety. The test was discontinued if the patient required physical assistance, touched the floor with the affected limb, or could not maintain balance. Patients who were unable to perform the test were given a reading of 0 s rather than treated as missing values. The same basic safety procedure was used across the participating institutions.

Statistical Analysis

Multivariable logistic regression analysis was performed with non-independent walking at discharge as the dependent variable. Prespecified clinically relevant covariates were simultaneously entered into the multivariable model to reduce confounding. Body mass index (BMI) was not included in the multivariable model because it did not differ significantly between groups and the model was restricted to prespecified clinically relevant variables to avoid overfitting. Facility was included as a fixed effect, and cluster-robust standard errors were applied at the facility level. Receiver operating characteristic (ROC) analysis was conducted to evaluate the discriminative ability of PNI, and the optimal cutoff value was determined using the Youden index. Multicollinearity was assessed using variance inflation factors. Complete-case analysis was performed because the proportion of missing data was low. All primary analyses, including group comparisons, multivariable logistic regression analysis, ROC analysis, and sensitivity analysis, were conducted using the same complete-case analysis population (n = 173). As an additional sensitivity analysis to assess model stability, a parsimonious logistic regression model including age, dementia, and PNI was also examined.

RESULTS

Among the 173 patients included in the final complete-case analysis, 85 were classified into the independent walking group and 88 into the non-independent walking group according to walking status at discharge. Baseline characteristics of the complete-case analysis population are shown in Table 1.

In the multivariable model, lower PNI and dementia remained independently associated with non-independent walking at discharge (Table 2). Age was also significantly associated with non-independent walking (odds ratio per 1-year increase, 1.05). Sex, surgical procedure, hemoglobin, lumbar spine YAM, and early postoperative physical function measures were not independently associated after adjustment. The events-per-variable value was 9.8, calculated as 88 non-independent walkers divided by 9 covariates included in the multivariable model. The variance inflation factors for all covariates were below 2.0, indicating no serious multicollinearity. The VIF values for PNI and hemoglobin were 1.25 and 1.18, respectively.

Table 2.  Multivariable logistic regression analysis for non-independent walking at discharge.

Variable	Odds ratio	95% confidence interval	P value
Age (per 1-year increase)	1.05	1.01–1.09	0.013
Female sex (yes)	0.58	0.20–1.65	0.304
Surgical procedure (BHA)	1.49	0.74–3.02	0.266
Dementia (yes)	6.32	2.75–15.87	<0.001
Hemoglobin (per 1 g/dL)	0.88	0.63–1.22	0.434
Lumbar spine YAM (per 10%)	0.94	0.81–1.09	0.403
Single-leg standing time (per 1 s)	0.99	0.97–1.01	0.212
Weight-bearing ratio (per 10%)	0.92	0.77–1.10	0.358
PNI (per 1-point increase)	0.86	0.80–0.92	<0.001

Analysis was performed using complete-case data (n = 173). All variables were simultaneously entered into the multivariable logistic regression model. Continuous variables are expressed per unit increase as shown. Surgical procedure was coded as BHA (reference: IF). Facility was included as a fixed effect, and cluster-robust standard errors were applied at facility level. The events-per-variable value was 9.8. All variance inflation factors were below 2.0.

The ROC analysis of PNI is summarized in Table 3. The optimal PNI cutoff value was 33.2, with a sensitivity of 68.2% and specificity of 68.2%. Results of the sensitivity analyses, including the model with log-transformed C-reactive protein (CRP) and the parsimonious model, are shown in Table 4. In the parsimonious model, dementia and PNI remained significantly associated with non-independent walking at discharge, whereas age was not significantly associated.

Table 3.  ROC analysis of PNI for non-independent walking at discharge.

Variable	Value
Area under ROC curve	0.66 (95% CI 0.58–0.74)
Optimal cutoff value (PNI)	33.2
Sensitivity, %	68.2
Specificity, %	68.2
Positive predictive value, %	69.0
Negative predictive value, %	67.4

ROC analysis was performed using the complete-case analysis population (n = 173). Optimal cutoff value of PNI was determined using the Youden index. Outcome of interest was non-independent walking at discharge.

CI, confidence interval.

Table 4.  Sensitivity analyses for non-independent walking at discharge.

Variable	Primary model			Model including log CRP			Parsimonious model
	OR (95% CI)	P value		OR (95% CI)	P value		OR (95% CI)	P value
Age (per 1-year increase)	1.05 (1.01–1.09)	0.013		-	-		1.02 (0.98–1.06)	0.293
Dementia (yes)	6.32 (2.75–15.87)	<0.001		6.05 (2.58–14.62)	<0.001		5.69 (2.21–14.68)	<0.001
PNI (per 1-point increase)	0.86 (0.80–0.92)	<0.001		0.87 (0.81–0.93)	<0.001		0.87 (0.81–0.95)	0.001
log CRP (per 1-unit increase)	-	-		1.21 (0.98–1.51)	0.074		-	-

Analyses were performed using the complete-case analysis population (n = 173). The primary model included age, sex, surgical procedure, dementia, hemoglobin at 1 week postoperatively, lumbar spine YAM, single-leg standing time on non-affected side, weight-bearing ratio of non-affected limb, and PNI. In sensitivity analysis including inflammatory status, log CRP was also included in the primary model. Parsimonious model included age, dementia, and PNI. Odds ratio (OR) and 95% confidence interval (CI) were estimated using multivariable logistic regression analysis. The outcome variable was non-independent walking at discharge.

DISCUSSION

This study was not designed to develop a predictive model but rather to examine the association between the PNI in the early postoperative period and walking independence at discharge. The results demonstrated that PNI at postoperative week 1 was independently associated with non-independent walking at discharge after adjustment for age, sex, surgical procedure, bone mineral density, anemia, and early postoperative physical function measures. The presence of dementia was also identified as an independent associated factor. These findings were supported by sensitivity analyses including additional adjustment for log-transformed CRP and a parsimonious model including age, dementia, and PNI. In contrast, measures of physical function that were assessed during the early postoperative period did not show independent associations in the multivariable analysis.

Physical function assessed as early as 1 week postoperatively may be strongly influenced by acute-phase factors such as surgical stress, pain, perioperative inflammation, and anemia. Therefore, physical performance at this time point may reflect acute stress responses rather than true recovery potential. It is thus reasonable that early postoperative physical function did not emerge as an independent factor in the multivariable model.

PNI, calculated from serum albumin levels and peripheral lymphocyte counts, may reflect postoperative nutritional, inflammatory, and immunological conditions. Low postoperative PNI may therefore indicate an impaired systemic condition during the early recovery phase. PNI has also been associated with sarcopenia and mortality in a population-based study.⁶⁾ Previous experimental studies have suggested that inflammatory pathways may be involved in impaired muscle protein metabolism.⁴⁾ However, inflammatory cytokines, muscle mass, and muscle protein metabolism were not directly measured in the present study. Therefore, the present findings cannot establish causal mechanisms. Our results should instead be interpreted as showing an association between postoperative PNI and walking independence at discharge, and as hypothesis-generating evidence that postoperative systemic condition may be related to short-term walking recovery.

A notable feature of this study is that it examined the association between PNI assessed at postoperative week 1 and walking independence at discharge. Previous studies examining the relationship between nutritional status and functional prognosis in patients with hip fracture have primarily evaluated nutritional indices at admission or during the perioperative period.^2,7) Our findings suggest that early assessment of nutritional and inflammatory status may contribute to risk stratification of walking outcomes at discharge and to the identification of patients who may require closer rehabilitation and nutritional monitoring.

The presence of dementia was also independently associated with walking independence at discharge. Dementia may affect comprehension of rehabilitation instructions, participation in therapy, physical activity, and recovery of daily mobility. Previous research has also reported that cognitive function affects physical activity, physical function, and quality of life after hip fracture.⁸⁾ However, this study did not include quantitative cognitive assessments such as the Mini-mental State Examination or the revised Hasegawa Dementia Scale. Therefore, our findings should be interpreted as showing an association between documented dementia and walking independence, rather than comprehensive cognitive function.

The discriminatory ability of PNI alone was moderate (area under the curve = 0.66) and insufficient to definitively predict walking independence at discharge. This finding suggests that PNI should be positioned not as a standalone determinant but rather as an adjunctive marker that may contribute to early risk stratification when combined with dementia status and physical function.

Lumbar spine YAM was not identified as an independent factor in the multivariable analysis. For the short-term outcome of walking independence at discharge, factors such as dementia status, postoperative PNI, and physical function may be more directly associated with walking independence than bone mineral density. In addition, lumbar spine dual-energy X-ray absorptiometry measurements are susceptible to degenerative changes, and limitations in measurement accuracy may have influenced the results.

No significant independent association was observed between sex and walking independence at discharge. Although previous studies have suggested potential sex differences in functional prognosis, Kristensen⁹⁾ reported that after adjustment for age, comorbidities physical function, and cognitive-related factors, the effect of sex becomes limited in multivariable analysis. Our findings are consistent with their report.⁹⁾ Although the proportion of women was higher in the independent walking group in univariate analysis, sex did not remain independently associated with walking independence in the multivariable model. This discrepancy may reflect the influence of confounding factors or characteristics of the selected study population. The present study included only patients who were ambulatory prior to injury and who could be evaluated at postoperative week 1, and a certain degree of selection bias cannot be excluded.

In sensitivity analyses, PNI remained associated with walking independence after additional adjustment for log-transformed CRP. However, because CRP is only one inflammatory marker and inflammatory cytokines were not measured, this result should be interpreted with caution. This finding suggests that postoperative PNI may provide information related to walking recovery beyond CRP alone, but it does not establish that PNI reflects systemic recovery reserve independently of inflammation.

LIMITATIONS

This study was a retrospective observational study and does not establish causality. PNI was measured at a single time point (postoperative week 1), and temporal changes during the perioperative period were not evaluated. In addition, postoperative PNI may be influenced not only by nutritional status but also by surgical stress, inflammation, infection, fluid balance, albumin synthesis, and changes in lymphocyte count. Rehabilitation content and nutritional interventions were not standardized, and residual confounding caused by unmeasured factors—such as frailty, severity of comorbidities (e.g., chronic heart failure, chronic kidney disease), and social background—may not have been fully controlled. Furthermore, because the data were derived from only three institutions within Japan, caution is warranted in generalizing the findings. Given the low rate of missing data, complete-case analysis was performed; however, whether the missing data were completely at random was not formally verified, and the possibility of selection bias cannot be excluded. Therefore, the findings may be most applicable to older Japanese patients with hip fracture who were ambulatory before injury and underwent postoperative rehabilitation in acute-care settings. In addition, the events-per-variable value was 9.8, which approached the lower limit of recommended thresholds; therefore, instability of estimates or overfitting cannot be entirely ruled out. Furthermore, although facility fixed effects and cluster-robust standard errors were used to account for institutional differences, the number of participating facilities was small. Therefore, the stability of standard error estimates related to facility-level clustering may be limited.

CONCLUSION

Early postoperative PNI was independently associated with non-independent walking at discharge. Because the discriminative ability of PNI alone was moderate, PNI should be interpreted as an adjunctive clinical marker rather than as a standalone measure for determining walking outcomes. These findings suggest that postoperative PNI, together with dementia status and other clinical factors, may help inform early rehabilitation planning in patients with hip fracture.