Admission CRP as an Independent Predictor of Prolonged Length of Stay in Calculous Acute Pyelonephritis: A Comparative Evaluation With the HALP Score

Abstract

Background

Stone-associated acute pyelonephritis (APN) carries an increased risk of complicated hospitalization, but prognostic markers for prolonged length of stay (LOS) remain limited. We sought independent predictors of prolonged LOS and compared discriminative performances of C-reactive protein (CRP) and the immuno-nutritional HALP [hemoglobin, albumin, lymphocyte, and platelet] score.

Material/Methods

This retrospective, single-center study included 120 patients (n = 118 for HALP-based analyses) with stone-associated APN diagnosed based on clinical, laboratory, and computed tomography findings. Firth-penalized multivariable logistic regression was used to identify independent predictors of prolonged hospital stay (> 5 days). The discriminative performances of CRP and the HALP score were assessed using receiver operating characteristic analysis and compared via the DeLong test.

Results

Mean LOS was 7.0 ± 4.2 days, and 53.3% of patients had a prolonged stay. CRP was the only independent predictor of prolonged stay in multivariable analysis. CRP showed better discrimination (area under the curve [AUC], 0.815; cutoff, ≥ 103.0 mg/L; sensitivity, 68.8%; specificity, 87.5%) compared with the HALP score (AUC, 0.691; cutoff, < 38; sensitivity, 78.1%; specificity, 57.4%); this difference was statistically significant (DeLong P = 0.013). Both markers were moderately correlated with LOS (CRP: ρ = 0.571; HALP: ρ = −0.376; P < 0.001 for both).

Conclusions

In stone-associated APN, high admission CRP was an independent predictor of prolonged LOS. A low HALP score was associated with prolonged stay in univariate analysis but not in the multivariable model. Admission CRP may help identify patients at risk of prolonged hospitalization; additional prognostic value of the HALP score requires prospective validation.

Introduction

Acute pyelonephritis (APN) is a bacterial infection of the upper urinary tract responsible for a substantial proportion of emergency urology visits and healthcare costs worldwide [1,2]. The clinical course of APN is inherently unpredictable and becomes more complex in the presence of stone-induced urinary tract obstruction. In this setting, persistent bacteremia, systemic inflammatory responses, and renal parenchymal damage occur more frequently, potentially prolonging hospitalization, increasing resource utilization, and raising the risk of long-term renal impairment [3,4].

In stone-associated APN, the duration of hospitalization is influenced by various host- and infection-specific factors, including comorbidity burden, pathogen virulence, stone location, and the degree of ureteral obstruction [5,6]. The development of sepsis further worsens the clinical course and prolongs recovery [7,8]. Because urine culture results are typically available at 48 to 72 hours after admission, early prognostic assessment must rely on data obtained during the initial evaluation.

C-reactive protein (CRP), a well-established acute-phase reactant, has long been used as a real-time indicator of inflammatory activity in urological infections [9,10]. CRP levels at presentation reflect the severity of the host inflammatory response and have been associated with infection severity and clinical outcomes in APN [11,12]. However, inflammation represents only a single aspect of the recovery process. Even when pathogen burden and antibiotic therapy are similar among patients, nutritional reserves and immune capacity may substantially influence the rate at which the inflammatory burden resolves.

The HALP score—a composite index derived from hemoglobin (H), albumin (A), lymphocyte count (L), and platelet count (P)—was originally developed as a prognostic tool in oncology [13,14] and has recently been investigated in acute inflammatory conditions, including sepsis and coronavirus disease 2019 (COVID-19) [15,16]. Whereas albumin and hemoglobin reflect nutritional status and oxygen-carrying capacity, lymphocyte and platelet counts provide information regarding systemic immune activity. This multidimensional structure may provide additional information about patients’ nutritional and immune statuses, factors that might influence recovery beyond the acute inflammatory response reflected by CRP.

In this context, the present study was designed to address 3 related but distinct questions in patients hospitalized for calculous APN: (1) whether high admission CRP is associated with prolonged hospital stay (LOS > 5 days); (2) whether the HALP score is independently associated with prolonged hospital stay after adjustment for established predictors; and (3) how the 2 markers compare in terms of discriminative performance for this outcome. We hypothesized that CRP, as a direct inflammatory marker, would demonstrate superior discriminative performance, whereas the HALP score would potentially reflect immuno-nutritional characteristics associated with recovery. Accordingly, the present study evaluates the independent associations and comparative discriminative performances of these markers; the potential benefit of their combined use for improving risk classification remains a hypothesis for future validation.

Material and Methods

Study Design and Patient Selection

This retrospective, single-center study examined the electronic medical records of all consecutive patients who presented to our institution with a diagnosis of APN between January 2023 and December 2025. The study was designed and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines; the completed STROBE checklist is provided as an additional file.

The diagnosis of APN was established based on clinical presentation (fever, flank pain, and lower urinary tract symptoms), elevated inflammatory markers, and characteristic findings on contrast-enhanced computed tomography. To ensure etiologic homogeneity, the study included only patients with calculous pyelonephritis, defined as APN occurring in the presence of urinary tract stones confirmed by computed tomography. Patients aged 18 years and older who met all diagnostic criteria were eligible for the study. The exclusion criteria were age younger than 18 years, pregnancy, and incomplete clinical or laboratory data.

Clinical Management and Surgical Interventions

Upon arrival at the emergency department, all patients were evaluated for hemodynamic stability, degree of urinary tract obstruction, and severity of the systemic inflammatory response. Patients with obstructive pyelonephritis or clinical deterioration despite initial antibiotic therapy underwent emergency urinary decompression; the remaining patients received conservative treatment, including intravenous broad-spectrum antibiotics and fluid resuscitation.

Percutaneous nephrostomy (PCN) was performed under local anesthesia with the patient in the prone position, guided by fluoroscopy or ultrasound. Retrograde double-J (DJ) stent placement was performed under sedoanalgesia or general anesthesia, depending on the patient’s American Society of Anesthesiologists (ASA) classification. Emergency decompression was performed in 100 patients (83.3%): DJ stenting in 69 patients (median time to intervention, 24.0 hours; interquartile range [IQR], 12.0–48.0) and PCN in 31 patients (median, 18.0 hours; IQR: 12.0–30.0).

Data Collection and HALP Score Calculation

Demographic characteristics, comorbidities, ASA physical status scores, vital signs at admission, and baseline laboratory parameters (complete blood count, CRP, procalcitonin, albumin, and creatinine) were obtained from electronic medical records. Although vital signs at admission and serum procalcitonin and creatinine levels were recorded during data collection, they are not reported in the study results because they were not part of the current analytical objective.

The HALP score was calculated using the following formula, as previously reported [17]: HALP = hemoglobin (g/L) × albumin (g/L) × lymphocyte count (10⁹/L) ÷ platelet count (10⁹/L). Hemoglobin values recorded in g/dL were converted to g/L (multiplied by 10) before calculation, and albumin values were expressed in g/L. Thus, the reported HALP values and threshold were directly comparable with those in the published literature.

Sepsis was defined according to the Sepsis-3 criteria as suspected infection plus an acute increase of at least 2 points in the Sequential Organ Failure Assessment (SOFA) score [18].

Outcome Definition

The primary outcome was prolonged length of stay (LOS), defined as a stay exceeding 5 days. This threshold was selected on clinical grounds: published guidelines indicate that patients with APN who adequately respond to intravenous antibiotics typically achieve clinical stability and become eligible for transition to oral therapy within 48 to 72 hours of admission. Therefore, a hospital stay exceeding 5 days represents a meaningful deviation from the expected course of recovery [19]. Because LOS is influenced by local discharge practices and care pathways in addition to infection severity, it was considered not merely a direct measure of infection severity but a pragmatic, recovery-related healthcare outcome [20].

Time to defervescence was defined as the interval between hospital admission and the first 24-hour period without fever (body temperature ≤ 38.0 °C); time to clinical stabilization was defined as the interval between hospital admission and the time at which the attending urologist documented hemodynamic stability and sustained clinical recovery. Both variables were derived from clinical records and analyzed as continuous variables. Their associations with LOS were assessed via Spearman correlation; comparisons between the short- and prolonged-stay groups were performed using the Mann-Whitney U test. These analyses were conducted to determine whether LOS was associated with clinically relevant measures of recovery, and corresponding findings are reported in the Results section.

Discharge was approved when all of the following criteria were met: absence of fever for at least 48 to 72 hours (body temperature ≤ 38.0 °C), clinical stabilization confirmed by the attending urologist, tolerance of oral antibiotic therapy, and hemodynamic stability.

Statistical Analysis

Descriptive, comparative, and receiver operating characteristic (ROC) analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA). Normality was assessed using the Kolmogorov-Smirnov test. Continuous variables are presented as mean ± standard deviation to permit comparison with previously published cohorts; categorical variables are shown as frequencies and percentages. Because several continuous variables were nonnormally distributed, all between-group comparisons of continuous variables were performed using the Mann–Whitney U test.

Comparisons between the short-stay (≤ 5 days) and prolonged-stay (> 5 days) groups were performed using the Mann-Whitney U test for continuous variables and the Pearson chi-square test for categorical variables. Fisher’s exact test was used when any expected cell count was less than 5. The primary multivariable binary logistic regression model was specified to evaluate CRP and the HALP score as predictors of prolonged hospital stay; sepsis at admission was retained as a clinically relevant confounder. Because the HALP score is a composite index incorporating hemoglobin and albumin, its individual components were not entered separately to avoid structural multicollinearity. Although leukocyte count and age were significantly associated with prolonged hospital stay in univariate analyses, they were not included in the primary model because the primary analysis was designed to evaluate the independent prognostic contributions of CRP and the HALP score, with adjustment for sepsis as a clinically relevant confounder, rather than to develop a comprehensive clinical prediction model. Accordingly, leukocyte count and age were evaluated separately in sensitivity analyses. These analyses were not prespecified but were performed after inspection of the univariate results and are thus considered exploratory. They were conducted to determine whether the primary findings were sensitive to exclusion of these variables and were not used for variable selection.

Multicollinearity among predictors was assessed using variance inflation factors, and model adequacy was evaluated using Nagelkerke R². Discriminative performance was further assessed using repeated stratified 5-fold cross-validation to obtain an optimism-corrected estimate of the area under the curve (AUC). The same cross-validation procedure was applied to the CRP-only and CRP + HALP models. The incremental contribution of the HALP score was additionally evaluated using a likelihood-ratio test comparing these 2 nested models. The Hosmer-Lemeshow test was not used because of its limited statistical power in the present sample.

The type of intervention (DJ stenting vs PCN) was not included in the regression model because it was considered a consequence of disease severity rather than an independent predictor; its inclusion could have introduced multicollinearity with sepsis status and baseline inflammatory markers. Results are presented as odds ratios (ORs) with 95% confidence intervals (CIs). Given complete separation for the sepsis variable (all patients with sepsis experienced prolonged hospitalization), the multivariable model was estimated using Firth’s penalized maximum likelihood method (R version 4.3.1, logistf package), which provides finite, bias-reduced estimates in the presence of separation.

The discriminative performances of CRP and the HALP score for predicting prolonged hospital stay were evaluated using ROC curve analysis, and AUCs were compared using the DeLong method. ROC analysis was performed using CRP values on their original scale (mg/L). As a secondary analysis, Spearman rank correlation was used to assess the relationships of CRP and the HALP score with LOS. Two-sided P-value < 0.05 were considered statistically significant. Two patients had missing admission albumin values; therefore, analyses involving albumin or the HALP score were based on 118 patients, without imputation of missing data.

Ethics Approval

This retrospective study was approved by the Clinical Research Ethics Committee of Van Training and Research Hospital (IRB approval number: GOKAEK/2026-04/07; date: April 10, 2026). All procedures involving human participants were conducted in accordance with the ethical standards of the institutional research committee and the 1964 Declaration of Helsinki and its subsequent amendments. Given the retrospective study design and use of anonymized data, the requirement for individual informed consent was waived by the ethics committee.

Results

Patient Characteristics

In total, 120 patients met the inclusion criteria. The mean age was 50.6 ± 16.4 years; 76.7% (n = 92) of patients were men and 23.3% (n = 28) were women. The mean LOS was 7.0 ± 4.2 days (median, 6 days). Prolonged hospitalization (LOS > 5 days) occurred in 64 patients (53.3%). Sepsis was present in 5 patients (4.2%) at admission. Baseline demographic and clinical characteristics are presented in Table 1. The following clinical and anatomical comparisons are reported in the text only. Among the 100 patients who underwent emergency decompression, LOS was significantly longer among those treated with PCN than among those who underwent DJ stenting (median, 9.0 vs 5.0 days; P < 0.001), consistent with the more severe clinical presentation in the PCN subgroup. Type of intervention was not included in the multivariable model because it was considered a consequence of disease severity, rather than an independent risk factor. No significant difference in LOS was observed between patients who underwent decompression within the first 24 hours and those who underwent later decompression (7.0 vs 6.0 days; P = 0.659). Several anatomical variables related to stones and obstruction were also associated with prolonged hospital stay, including proximal or pelvic stone location (P = 0.005), bilateral stones (17.2% vs 3.6%; P = 0.036), complicated pyelonephritis (15.6% vs 3.6%; P = 0.034), and higher-grade hydronephrosis (P = 0.033).

Table 1.

Baseline demographic and clinical characteristics.

Variable	Total (n = 120)	Short stay ≤ 5 days (n = 56)	Prolonged stay > 5 days (n = 64)	P
Age (years), mean ± SD	50.6 ± 16.4	47.0 ± 15.0	53.7 ± 17.1	0.027
Male sex, n (%)	92 (76.7%)	44 (78.6%)	48 (75.0%)	0.644
Diabetes mellitus, n (%)	20 (16.7%)	6 (10.7%)	14 (21.9%)	0.102
Sepsis at admission, n (%)	5 (4.2%)	0 (0%)	5 (7.8%)	0.060
LOS (days), mean ± SD	7.0 ± 4.2	3.9 ± 0.9	9.7 ± 4.0	< 0.001

Abbreviations: LOS, length of stay; SD, standard deviation. Notes: Continuous variables were compared using the Mann-Whitney U test and categorical variables using the chi-square test, except for sepsis at admission—the expected cell count was < 5 and Fisher’s exact test was performed.

Univariate and Multivariable Analyses

In univariate analyses, age, admission CRP, hemoglobin, albumin, white blood cell count, and the HALP score were significantly associated with prolonged LOS (Tables 1, 2). Sepsis at admission was more frequent among patients with a prolonged stay (7.8% vs 0%), although the difference was not statistically significant (Fisher’s exact test, P = 0.060). Fisher’s exact test was used because the expected cell count was less than 5. Nevertheless, sepsis was included in the multivariable model as a clinically relevant confounder. White blood cell count was higher in the prolonged-stay group (P = 0.037) and was evaluated separately in a sensitivity analysis. When added to the primary model, white blood cell count was not independently associated with prolonged LOS (adjusted OR, 1.06; 95% CI, 0.96–1.16; P = 0.25).

Table 2.

Comparison of admission laboratory parameters between groups.

Parameter	Short stay ≤ 5 days (n = 56)	Prolonged stay > 5 days (n = 64)	P
WBC (× 103/μL)	12.9 ± 4.3	14.9 ± 5.4	0.037
CRP (mg/L)	50.0 ± 53.9	157.5 ± 107.0	< 0.001
Hemoglobin (g/dL)	14.8 ± 1.8	13.2 ± 2.2	< 0.001
Albumin (g/dL)	3.79 ± 0.47	3.49 ± 0.46	0.001
HALP score	45.0 ± 25.0	30.4 ± 22.6	< 0.001

Abbreviations: CRP, C-reactive protein; HALP, hemoglobin, albumin, lymphocytes, platelets; WBC, white blood cell count. Notes: The HALP score was calculated using hemoglobin and albumin values expressed in g/L; hemoglobin values recorded in g/dL were converted to g/L by multiplying by 10. Albumin and HALP score data were available for 118 patients (short stay, n = 54; prolonged stay, n = 64); all other variables were analyzed in 120 patients (short stay, n = 56; prolonged stay, n = 64). Group comparisons were performed using the Mann-Whitney U test.

In the multivariable model, CRP was the only independent predictor of prolonged hospital stay (Table 3). Each 1 mg/L increase in baseline CRP was associated with a 1.4% increase in the odds of a prolonged stay (OR, 1.014; 95% CI, 1.008–1.021; P < 0.001). Because prolonged hospital stay occurred in all 5 patients with sepsis, the sepsis variable was affected by complete separation. When Firth’s penalized estimation method was used, the OR for sepsis remained high but imprecise and lacked statistical significance (OR, 7.76; 95% CI, 0.25–239.9; P = 0.242).

Table 3.

Independent predictors of prolonged length of stay: multivariable binary logistic regression.

Variable	OR	95% CI	P
Sepsis at admission	7.76	0.25–239.9	0.242
Baseline CRP (per 1 mg/L)	1.014	1.008–1.021	< 0.001
HALP score (per 1 unit)	0.993	0.975–1.012	0.46

Abbreviations: CI, confidence interval; CRP, C-reactive protein; HALP, hemoglobin, albumin, lymphocytes, platelets; OR, odds ratio. Notes: Estimates were obtained using Firth’s penalized logistic regression. The sepsis term was affected by complete separation because all patients with sepsis had a prolonged hospital stay, resulting in a wide CI. Nagelkerke R² = 0.40 for the CRP-HALP model (model P < 0.001). The model was fitted using data from 118 patients with complete admission albumin data.

The HALP score was significantly associated with prolonged stay in univariate analysis (OR, 0.975; 95% CI, 0.958–0.991; P = 0.003) but did not remain independently associated in the multivariable model (adjusted OR, 0.993; 95% CI, 0.975–1.012; P = 0.46). CRP and the HALP score were moderately and inversely correlated (Spearman ρ = −0.48). Excluding the separated sepsis term, the CRP-HALP model had a Nagelkerke R² of 0.40 (P < 0.001). Sensitivity analyses incorporating leukocyte count and age are reported separately below and in Table 4.

Table 4.

Sensitivity analyses: Firth-penalized multivariable models for prolonged length of stay.

Model/Variable	OR	95% CI	P	n
Sensitivity model 1 (primary model + leukocyte count)				118
CRP (per 1 mg/L)	1.014	1.007–1.021	< 0.001
HALP score (per 1 unit)	0.994	0.975–1.013	0.52
Sepsis at admission6	6.04	0.21–177.8	0.30
Leukocyte count (per 1 × 103/μL)	1.056	0.962–1.159	0.25
Sensitivity model 2 (primary model + age)				118
CRP (per 1 mg/L)	1.014	1.008–1.021	< 0.001
HALP score (per 1 unit)	0.996	0.976–1.015	0.67
Sepsis at admission	7.03	0.20–245.5	0.28
Age (per 1 year)	1.022	0.994–1.050	0.13

Abbreviations: CI, confidence interval; CRP, C-reactive protein; HALP, hemoglobin, albumin, lymphocytes, platelets; OR, odds ratio. Notes: Both models were estimated using Firth’s penalized logistic regression in the 118 patients with complete admission albumin data. The primary model included CRP, the HALP score, and sepsis at admission; each sensitivity model included 1 additional covariate. The sepsis term was affected by complete separation in both models, resulting in wide CIs. These analyses were not prespecified; they were performed after inspection of the univariate results and are reported as exploratory. They were not used for variable selection.

In repeated stratified 5-fold cross-validation, both the CRP-only and CRP + HALP models yielded an optimism-corrected AUC of 0.80 (ΔAUC = 0.00), indicating no improvement in discriminative performance upon addition of the HALP score. This finding was consistent with the nonsignificant HALP coefficient in the multivariable model (P = 0.46). The optimism-corrected AUC for the CRP-only model (0.80) was also similar to the apparent AUC from the primary ROC analysis (0.815), suggesting limited optimism in the apparent estimate.

ROC Curve Analysis

CRP demonstrated significantly greater discriminative ability than the HALP score (AUC, 0.815 vs 0.691; ΔAUC, 0.124); the DeLong test indicated a statistically significant difference between curves (Z = 2.48; P = 0.013). The optimal CRP cutoff of at least 103.0 mg/L yielded high specificity (87.5%) and moderate sensitivity (68.8%). Conversely, a HALP score cutoff of less than 38 yielded higher sensitivity (78.1%) but lower specificity (57.4%). These findings are summarized in Table 5 and Figure 1.

Table 5.

Receiver operating characteristic curve analysis: CRP vs HALP score for prolonged length of stay.

Variable	AUC	95% CI	Cutoff	Sensitivity (%)	Specificity (%)	P (vs CRP)
CRP (mg/L)	0.815	0.74–0.89	≥ 103.0	68.8	87.5	—
HALP score	0.691	0.59–0.79	< 38	78.1	57.4	0.013

Abbreviations: AUC, area under the curve; CI, confidence interval; CRP, C-reactive protein; HALP, hemoglobin, albumin, lymphocytes, platelets.

Figure 1.

Receiver operating characteristic (ROC) curve analysis comparing the predictive performance of the HALP (hemoglobin, albumin, lymphocytes, platelets) score and C-reactive protein (CRP) for prolonged hospital stay (length of stay [LOS] > 5 days). The orange line represents CRP (area under the curve [AUC] = 0.815). ROC analysis was performed using CRP on its original scale (mg/L); the reported AUC and clinical cutoff (≥ 103.0 mg/L) both refer to this scale. Both curves were generated using data from the 118 patients with complete admission albumin data, corresponding to the sample used for the DeLong comparison. The blue line represents the HALP score, which showed moderate discriminative ability (AUC = 0.691). The diagonal dashed line represents a nondiscriminatory model (AUC = 0.50).

Spearman correlation analyses confirmed moderate associations of both markers with LOS: ρ = 0.571 (P < 0.001) for CRP and ρ = −0.376 (P < 0.001) for the HALP score.

Clinical Relevance of LOS as an Outcome Measure

LOS was positively correlated with both clinically recorded measures of recovery: time to defervescence (Spearman ρ = 0.611; P < 0.001) and time to clinical stabilization (ρ = 0.616; P < 0.001). Both intervals were significantly longer in patients with a hospital stay of more than 5 days. The median time to defervescence was 48 hours (IQR, 48–72) in the prolonged-stay group versus 24 hours (IQR, 24–48) in the short-stay group (P < 0.001); the median time to clinical stabilization was 72 hours (IQR, 48–90) versus 48 hours (IQR, 24–48), respectively (P < 0.001). These findings support the clinical relevance of LOS as an outcome measure in the present cohort.

Sensitivity Analyses

Two exploratory sensitivity models were fitted after inspection of the univariate results, each adding 1 covariate to the primary Firth-penalized model (CRP, HALP score, and sepsis at admission). Both analyses were based on the 118 patients with complete admission albumin data. When leukocyte count was added, CRP remained the only independent predictor of prolonged stay (adjusted OR, 1.014; 95% CI, 1.007–1.021; P < 0.001), whereas the HALP score (OR, 0.994; 95% CI, 0.975–1.013; P = 0.52), sepsis (OR, 6.04; 95% CI, 0.21–177.8; P = 0.30), and leukocyte count (OR, 1.056; 95% CI, 0.962–1.159; P = 0.25) were not statistically significant. A similar pattern was observed when age was added: CRP remained significant (OR, 1.014; 95% CI, 1.008–1.021; P < 0.001), whereas the HALP score (OR, 0.996; 95% CI, 0.976–1.015; P = 0.67), sepsis (OR, 7.03; 95% CI, 0.20–245.5; P = 0.28), and age (OR, 1.022; 95% CI, 0.994–1.050; P = 0.13) were not. Full model estimates are presented in Table 4. A likelihood-ratio test comparing the CRP-only and CRP + HALP models showed that adding the HALP score did not significantly improve model fit (χ² = 0.81; df = 1; P = 0.37), consistent with the identical optimism-corrected AUCs obtained for these models.

Discussion

In this study, high CRP was the only independent predictor of prolonged hospital stay in a well-defined cohort of patients with stone-associated APN. Although a low HALP score was significantly associated with prolonged hospital stay in univariate analysis, it did not remain independently associated after adjustment for CRP. Furthermore, the contribution of sepsis could not be reliably estimated due to complete separation within this small subgroup. Although CRP and the HALP score reflect partially distinct biological processes (acute inflammatory burden and immuno-nutritional reserve, respectively), the moderate inverse correlation between them (ρ = −0.48) indicates substantial shared variability. Consequently, much of the prognostic information conveyed by the HALP score may be captured by CRP in the adjusted model.

The large but imprecise point estimate for sepsis is consistent with previous studies showing that systemic inflammatory responses in obstructive pyelonephritis significantly prolong hospitalization [4,21,22]. However, this finding should be interpreted with caution given the very small number of patients with sepsis (n = 5), which resulted in a wide CI and limited estimate precision. Prospective studies involving larger sepsis subgroups are needed to confirm this association.

Consistent with its known sensitivity to acute inflammatory activity in urological infections [9,11,23], CRP outperformed the HALP score in ROC analysis (AUC, 0.815 vs 0.691). The optimal cutoff of at least 103.0 mg/L provides a practical reference point at hospital admission and, considering its high specificity (87.5%), may help identify patients likely to require prolonged hospitalization. The difference in AUCs was statistically significant according to the DeLong test (P = 0.013), indicating that CRP had greater discriminative ability as a single marker; however, the HALP score showed higher sensitivity at its optimal cutoff.

In our study cohort, the HALP score was associated with prolonged hospital stay in univariate analysis; to our knowledge, this represents a rare evaluation of this score specifically in the context of stone-related APN. Rather than representing a major innovation, this finding should be viewed as an extension of the established prognostic role of the HALP score [24–26] to the specific clinical context of obstruction-related upper urinary tract infection. At an optimal cutoff of less than 38, the HALP score identified patients at risk of prolonged hospitalization with higher sensitivity (78.1%) but lower specificity than CRP. Low albumin and lymphocyte counts have been independently associated with impaired host defense and delayed recovery from systemic infections [27]; furthermore, albumin and hemoglobin have both been associated with clinical outcomes in febrile urological infections [28,29]. In univariate analysis, the HALP score was associated with prolonged hospital stay, reflecting an immuno-nutritional dimension of patient vulnerability. However, in the adjusted model, this information largely overlapped with data provided by CRP and did not contribute independently.

In addition to biochemical markers, several anatomical variables related to stones and obstruction were associated with prolonged hospital stay in the present cohort. Given that urinary tract obstruction, rather than stone size, is the primary pathophysiological factor in stone-related APN [30,31], these findings are consistent with expectations; they also reinforce that biochemical markers alone are insufficient to fully characterize clinical risk. Along with anatomical and infectious factors, genetic predisposition contributes to stone formation and recurrence risk [32,33], underscoring the multifactorial nature of urolithiasis. Definitive stone treatment, including modern laser lithotripsy and minimally invasive endourological approaches, remains essential to prevent recurrent obstructive episodes [34,35]. Current treatment strategies for APN continue to evolve, including investigation of adjunctive measures intended to limit renal scarring [36]; early risk stratification may help tailor the intensity of inpatient management to individual patients. Composite urological scoring systems, such as the STONE score and Guy’s Stone Score, were not routinely recorded in our dataset; this represents a limitation and an area for investigation in future prospective studies [37].

Given that CRP remained independently associated with prolonged hospital stay, whereas the HALP score was associated with prolonged stay only in univariate analysis, these 2 markers may capture partly distinct biological dimensions while providing overlapping prognostic information. CRP primarily reflects the acute inflammatory response; the components of the HALP score reflect broader hematologic, nutritional, and immune characteristics that may also be relevant to recovery. In a direct comparison of the 2 nested models, addition of the HALP score to CRP did not improve optimism-corrected discrimination (AUC, 0.80 vs 0.80). Reclassification measures (net reclassification improvement/integrated discrimination improvement) and calibration of the combined model were not evaluated. Thus, further prospective evaluation is warranted concerning whether the HALP score provides clinically meaningful incremental value beyond CRP.

The predominance of male patients (76.7%) in our study cohort warrants consideration. Clinicians should incorporate this demographic characteristic when assessing the generalizability of these findings to populations with uncomplicated APN. Furthermore, because hemoglobin levels differ by sex and constitute a component of the HALP score, the strong predominance of male patients may have influenced HALP values and represents a potential source of confounding. Accordingly, although the current findings support admission CRP as a primary laboratory marker for the early identification of patients at risk of prolonged hospitalization, the role of the HALP score should be considered exploratory pending external validation.

This study has several limitations. The retrospective, single-center design introduces the possibility of selection bias and limits generalizability. The sepsis subgroup was small (n = 5), and no patients with sepsis were present in the short-stay group. The complex immune response accompanying severe urinary tract infections and sepsis is becoming better understood through recent studies of NETosis and pyroptosis [38]. Established biomarkers for bacteremic pyelonephritis, including procalcitonin and presepsin [39,40], were not systematically included in the present analysis. Prospective comparison of these markers with CRP and the HALP score would help determine the optimal biomarker combination for predicting LOS. Finally, composite urological scoring indices were not systematically recorded, limiting the comprehensiveness of the predictive model.

Conclusions

In patients with stone-associated APN, high CRP at admission was independently associated with prolonged hospital stay; an optimal cutoff of at least 103.0 mg/L showed high specificity in the present cohort. A low HALP score was associated with prolonged hospital stay in univariate analysis and showed higher sensitivity at its optimal cutoff; however, it did not remain independently associated with prolonged stay in the multivariable model, and CRP showed significantly better overall discriminative performance in ROC analysis. Admission CRP may thus contribute to early risk assessment in conjunction with standard clinical and imaging findings. Prospective multicenter studies are needed to validate the identified CRP cutoff and determine whether the HALP score provides additional prognostic value beyond CRP before these markers are incorporated into routine clinical decision-making.

Data Availability

The datasets analyzed during this study are available from the corresponding author upon reasonable request.