Prediction of postoperative pulmonary complications in geriatric patients with hip fracture using a nomogram incorporating bedside ultrasonography and clinical parameters

Abstract

Background

Postoperative pulmonary complications (PPCs) are common in geriatric patients with hip fractures and are associated with increased morbidity and mortality. Bedside lung ultrasonography may improve perioperative pulmonary risk stratification.

Objective

This study aimed to develop and validate a nomogram incorporating lung ultrasonography findings, diaphragmatic mobility, and clinical parameters to predict PPCs in geriatric patients with hip fractures.

Methods

This observational study included 786 geriatric patients with hip fractures undergoing surgery; patients were divided into a training cohort (n = 594) and a validation cohort (n = 192). Lung ultrasonography score and diaphragmatic mobility were assessed at the bedside one day before surgery. All patients received general anesthesia combined with a regional block and were followed for 2 weeks postoperatively for the occurrence of PPCs. Independent predictors were identified using logistic regression, and a nomogram was constructed. Model performance was evaluated using receiver operating characteristic analysis, calibration curves, and decision curve analysis.

Results

PPCs occurred in 32.84% (246/749) of patients. Nine independent variables were incorporated in the nomogram: American Society of Anesthesiologists physical status class, functional dependence, chronic obstructive pulmonary disease, lung ultrasonography score, diaphragmatic mobility, recent respiratory infection, hypoalbuminemia, anemia, and elevated N-terminal fragment of the pro–brain natriuretic peptide level. The model demonstrated strong discrimination, with an area under the curve of 0.897 (95% confidence interval, 0.870–0.924), good calibration, and favorable clinical utility. In addition, the nomogram outperformed the ARISCAT in the overall cohort.

Conclusions

A nomogram incorporating bedside ultrasonography and clinical parameters demonstrated favorable predictive performance for PPCs in geriatric patients with hip fractures. This tool may provide supplementary information for perioperative risk stratification and aid in the development of targeted preventive strategies in clinical practice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-026-07742-x.

Background

Morbidity in geriatric hip fractures is increasing with an increase in the aging population worldwide [1]. Postoperative pulmonary complications (PPCs) are common in geriatric patients undergoing hip fracture surgery, with a reported incidence of 33.3% (168/504) [2]. PPCs are associated with prolonged postoperative tracheal intubation, increased healthcare costs, and higher mortality [3]. Identifying risk factors for PPCs in geriatric patients with hip fractures is therefore essential for improving perioperative prevention and management strategies.

Although various predictive models for PPCs have been developed, most do not incorporate imaging assessments of pulmonary status [4–6]. Bedside lung ultrasonography has gained widespread acceptance for evaluating lung status and can effectively predict postoperative pulmonary outcomes and major adverse cardiac events [7, 8] It is rapid, repeatable, radiation-free, and non-invasive. Integrating bedside ultrasonographic findings into a predictive model may therefore enhance risk stratification for PPCs in geriatric patients with hip fractures and provide meaningful clinical value.

Given these findings, an observational study involving 786 geriatric patients with hip fractures who underwent surgical treatment at a Chinese tertiary orthopedic hospital was conducted. This study aimed to develop a nomogram incorporating the lung ultrasonography score, diaphragmatic mobility, and traditional clinical parameters to predict PPCs in geriatric patients undergoing hip surgery.

Methods

Ethics

This study was approved by the Ethics Committee of Honghui Hospital, Xi’an Jiaotong University, Xi’an, Shaanxi Province, China (Chairperson: Professor Kun Zhang) on September 14, 2024, with the ethics approval number 20,240,915. All study methods were performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from each patient or their family member. As a non‑interventional observational study, clinical trial registration was exempted for this research in accordance with ICMJE guidelines and ChiCTR requirements.

Sample size calculation

Sample size estimation for the training cohort was based on the events-per-variable (EPV) metric, with a minimum of 10 positive events required per independent variable to ensure model stability [9]. Previous evidence indicates an incidence of PPCs of 33.3% in geriatric patients with hip fractures [2]. Given the planned inclusion of 18 predictor variables, the sample size was calculated as follows:

Required number of PPCs eventsEPV × Number of predictive variables = 10 × 18 = 180

.

Considering a 10% loss rate, the sample size was adjusted to a minimum of 594 cases.

Sample size estimation for the validation cohort was based on the hypothesis that the expected area under the receiver operating characteristic curve (AUC) is greater than 0.65. A two-sided test with a significance level of α = 0.05 and a power of 90% (β = 0.10) was performed using PASS software, which indicated a required sample size of 174 cases. Considering a potential 10% loss to follow-up, the sample size was increased from 174 to 192 cases.

Participants

Inclusion criteria were patients aged > 65 years, hospital admission for hip fracture, and scheduled surgical management. Exclusion criteria comprised multiple fractures; acetabular, periprosthetic, or pathological hip fracture; chest wall skin contusions; perioperative cardiovascular or cerebrovascular accidents; unplanned surgery or discharge during follow-up; and refusal to participate.

Bedside ultrasonography

Bedside ultrasonography, including lung and diaphragmatic assessment, was performed by two experienced ultrasonologists who were blinded to the study objectives. Examinations were conducted prior to surgery using a convex transducer.

Twelve anatomical regions were defined on the chest wall (Fig. 1). Each region was scored according to predefined criteria (Fig. 2), and the cumulative score of the 12 regions was defined as the lung ultrasonography score [10, 11].

Fig. 1.

Chest wall regions assessed by ultrasonography. Schematic representation of chest wall regions evaluated during ultrasonographic examination. Based on the anatomical landmarks defined by the parasternal, anterior axillary, posterior axillary, and mid-spinal lines, each hemithorax is divided into anterior, lateral, and posterior regions. Each region is further subdivided into superior and inferior subregions by the bilateral nipple line

Fig. 2.

Lung ultrasound scoring criteria. The lung ultrasound scoring system is used for regional assessment. A score of 0 indicates A-lines or ≤ 2 B-lines (A or B); 1 indicates ≥ 3 well-spaced B-lines (C); 2 indicates coalescent B-lines (D); and 3 indicates a tissue-like pattern (E). The final score assigned to each region corresponds to the highest score observed within that region

Diaphragmatic mobility was assessed using a curved array probe positioned in the subcostal region between the midclavicular and anterior axillary line (Fig. 3). Participants were instructed to perform maximal exhale followed by maximal inhalation. The distance between the highest and lowest points of the diaphragm was measured (Fig. 4), and the mean of bilateral measurements was calculated as diaphragmatic mobility [12].

Fig. 3.

Probe placement for assessment of diaphragmatic mobility. Schematic diagram illustrating the positioning of the curved array ultrasound probe for diaphragmatic mobility assessment. The probe was placed in the subcostal region between the midclavicular and anterior axillary lines to visualize diaphragmatic excursion during respiration

Fig. 4.

Measurement of diaphragmatic mobility. Representative ultrasound image showing diaphragmatic excursion. Diaphragmatic mobility is defined as the distance (cm) between the highest and lowest positions of the diaphragm on the ultrasound image

Anesthesia protocol

General anesthesia combined with a fascia iliaca block was administered to all patients. Anesthesia was induced with midazolam 0.1 mg·kg⁻¹, sufentanil 0.25 µg·kg⁻¹, propofol 1.5–2.0 mg·kg⁻¹, and rocuronium 0.6 mg·kg⁻¹. Anesthesia was maintained with remifentanil 0.1–0.2 µg·kg⁻¹·min⁻¹ and sevoflurane 1.0–1.5%, targeting a bispectral index of 40–60. Fascia iliaca block was performed using 30–40 mL of 0.25% ropivacaine. Neuromuscular blockade was continuously monitored, and residual blockade was reversed with neostigmine.

After surgery, patients who met the standard extubation criteria (adequate spontaneous respiration, satisfactory oxygenation, stable hemodynamics, recovered muscle strength, and full consciousness) underwent tracheal extubation in the operating room. In contrast, patients with severe respiratory insufficiency, unstable hemodynamics, or impaired consciousness were transferred to the intensive care unit (ICU) with tracheal intubation retained for further supportive care and delayed extubation.

Intraoperative ventilation

Volume-controlled ventilation was used throughout surgery. Tidal volume was set at 6–8 mL·kg^{− 1} predicted body weight, with positive end-expiratory pressure of 6–8 cm H₂O. The inspiratory-to-expiratory ratio was 1:2, and the respiratory rate was adjusted to maintain end-tidal carbon dioxide at 35–40 mmHg. The fraction of inspired oxygen was maintained below 0.5. In case of SpO₂ < 95%, rescue therapy, including increased inspired oxygen and recruitment maneuvers, was initiated immediately [13].

Perioperative fluid and hemodynamic management

Before induction of anesthesia, lactated Ringer’s solution was administered intravenously at 5 mL·kg⁻¹, followed by continuous infusion at 5 mL·kg⁻¹·h⁻¹. Hemodynamic management was guided by a continuous non-invasive arterial pressure system. Boluses of 3 mL·kg⁻¹ lactated Ringer’s solution were administered every 15 min to maintain pulse pressure variation (PPV) ≤ 14%. Additional boluses were given if PPV remained > 14% for ≥ 5 min or stroke volume increased by > 10%, until PPV ≤ 14% was achieved. If the mean arterial pressure (MAP) remained < 60 mmHg despite fluid optimization or if MAP < 60 mmHg was accompanied by PPV ≤ 14% and stroke volume (SV) increase < 10%, vasopressor therapy was initiated with ephedrine 6 mg intravenously (maximum cumulative dose 30 mg). If required, norepinephrine was continuously infused at 0.05 µg·kg⁻¹·min⁻¹ [14].

Data collection

Collected variables included age, sex, height, weight, American Society of Anesthesiologists (ASA) status, history of chronic obstructive pulmonary disease (COPD) and diabetes, smoking within 2 months before surgery, and functional dependence (whether assistance was required for activities of daily living, including using the toilet, feeding, transferring, dressing, climbing stairs, bathing, as well as urinary and fecal incontinence). Other variables comprised a recent history of respiratory infection (within 1 month before surgery), time from injury to surgery, cough, expectoration symptoms, and hypoxemia (SpO₂ ≤ 90% or arterial oxygen tension < 8.0 kPa on room air). Lung ultrasonography score, diaphragmatic mobility, surgical type and duration, intravenous fluid volume, and laboratory parameters were recorded, including serum creatinine, N-terminal pro-B-type natriuretic peptide (NT-proBNP), hemoglobin, and serum albumin [4, 10, 15, 16].

Elevated serum creatinine was defined as > 1.3 mg·dL^{− 1} [17], hypoalbuminemia as < 35 g·L^{− 1} [18], and anemia as a hemoglobin < 12 g·dL^{− 1} in females and < 13 g·dL^{− 1} in males [19]. Elevated NT-proBNP was defined as ≥ 250 pg·mL^{− 1} in patients aged 65–74 years, and ≥ 500 pg·mL^{− 1} in individuals ≥ 75 years old [20].

All participants were followed up for 2 weeks postoperatively by an independent assessment team consisting of two senior attending physicians with extensive experience in respiratory medicine, to monitor the occurrence of PPCs. These assessors were blinded to the preoperative ultrasonographic parameters (lung ultrasonography score and diaphragmatic mobility), and had access only to basic clinical information as part of routine clinical workflow. PPCs were diagnosed based on a combination of clinical signs and symptoms (fever, cough, dyspnea, purulent sputum, and diminished breath sounds), laboratory examinations (elevated white blood cell count, C‑reactive protein, or procalcitonin), and chest imaging (chest radiograph or computed tomography demonstrating new infiltrates, atelectasis, or pleural effusion). The diagnosis of PPCs was in accordance with the European Perioperative Clinical Outcome Guidelines (Supplementary Digital Content) [21].

Statistical analysis

Statistical analyses were conducted using SPSS version 26.0 (IBM Corporation, Armonk, NY, USA) and R version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria). Categorical variables were analyzed using the chi-square test. Independent samples t-tests were used for normally distributed continuous variables, while the Mann–Whitney U test was used for non-normally distributed variables. Logistic regression was performed to identify independent predictors of PPCs, with assessment for multicollinearity. A nomogram was constructed based on multivariate logistic regression results. Model discrimination was evaluated using the area under the receiver-operating characteristic (ROC) curve (AUC), and calibration was assessed using the Hosmer–Lemeshow test and calibration curves. Decision curve analysis (DCA) was used to evaluate the clinical utility of the model. Model discrimination was compared via differences in AUC tested with DeLong’s method. Statistical significance was set at P < 0.05.

Results

The final training and validation cohorts comprised 562 and 187 patients, respectively. The study flowchart is shown in Fig. 5.

Fig. 5.

Study flowchart. PE, pulmonary embolism; ACS, acute coronary syndrome; PPCs, postoperative pulmonary complications

Patients’ baseline demographic and clinical characteristics

PPCs occurred in 32.84% (246/749) of patients. Specifically, 20.32% (50/246) patients presented with respiratory infection, 37.80% (93/246) with respiratory failure, 6.10% (15/246) with pleural effusion, 14.63% (36/246) with atelectasis, 1.62% (4/246) with pneumothorax, 3.66% (9/246) with bronchospasm, and 15.87% (39/246) with multiple PPCs (two or more PPCs mentioned above). Postoperative length of hospital stay, unexpected ICU admission, and mortality increased significantly in patients with respiratory infection and multiple PPCs (Table 1).

Table 1.

Postoperative length of hospital stay, ICU admission, and mortality according to the types of PPCs

	non-PPCs	Respiratory infection	Respiratory failure	Pleural effusion	Atelectasis	Pneumothorax	Bronchospasm	Multiple PPCs
Number	503	50	93	15	36	4	9	39
Length of hospital stay, days, median (IQR)	3.0 (2.0, 7.0)	14.0 (7.0, 20.0)	7.0 (3.0, 11.0)	7.0 (4.0, 12.0)	3.0 (2.0, 7.0)	7.0 (4.0, 10.0)	5.0 (3.0, 7.0)	17.0 (12.0, 21.0)
ICU admission, n (%)	13 (2.58%)	12 (24.00%)	10 (10.75%)	1 (6.66%)	0 (0.00%)	2 (50.00%)	1 (11.11%)	8 (20.52%)
30-day mortality, n (%)	2 (0.39%)	2 (4.00%)	1 (1.07%)	0 (0.00%)	0 (0.00%)	0 (0.00%)	0 (0.00%)	1 (2.56%)

Abbreviation: PPCs Postoperative pulmonary complications, ICU Intensive care unit

No significant differences were observed in baseline demographic and clinical characteristics between the training cohort and validation cohort (Table 2).

Table 2.

Baseline and clinical characteristics of patients in the training and validation cohorts

Variables	Training cohort	Validation cohort	Overall	P
Number	562	187	749
Age, years, median (IQR)	72.0 (68.0, 79.0)	71.0 (68.0, 77.0)	72.0 (68.0, 79.0)	0.181
Sex				0.460
Women, n (%)	280 (49.82%)	99 (52.94%)	379 (50.60%)
Men, n (%)	282 (50.18%)	88 (47.06%)	370 (49.40%)
BMI, kg·m− 2, mean ± SD	22.99 ± 3.87	23.47 ± 4.27	23.13 ± 4.08	0.192
ASA physical status class				0.280
≤ II, n (%)	351 (62.46%)	125 (66.84%)	476 (63.55%)
≥ III, n (%)	211 (37.54%)	62 (33.16%)	273 (36.45%)
Functional dependence				0.962
No, n (%)	468 (83.27%)	156 (83.42%)	624 (83.31%)
Yes, n (%)	94 (16.73%)	31 (16.58%)	125 (16.69%)
COPD				0.104
No, n (%)	421 (74.91%)	151 (80.75%)	572 (76.37%)
Yes, n (%)	141 (25.09%)	36 (19.25%)	177 (23.63%)
Diabetes				0.139
No, n (%)	488 (86.83%)	170 (90.91%)	658 (87.85%)
Yes, n (%)	74 (13.17%)	17 (9.09%)	91 (12.15%)
Smoking status				0.105
No, n (%)	451 (80.25%)	160 (85.56%)	611 (81.58%)
Yes, n (%)	111 (19.75%)	27 (14.44%)	138 (18.42%)
Time from injury to surgery				0.303
≤ 48 h, n (%)	161 (28.65%)	61 (32.62%)	222 (29.64%)
> 48 h, n (%)	401 (71.35%)	126 (67.38%)	527 (70.36%)
Lung ultrasonography score, median (IQR)	3.0 (3.0, 6.0)	3.0 (2.0, 5.0)	3.0 (2.0, 5.0)	0.072
Diaphragmatic mobility, cm, median (IQR)	3.2 (3.2, 5.3)	4.2 (3.2, 5.3)	3.2 (3.2, 5.3)	0.061
Recent respiratory infection				0.114
No, n (%)	422 (75.09%)	151 (80.75%)	573 (76.50%)
Yes, n (%)	140 (24.91%)	36 (19.25%)	176 (23.50%)
Cough and expectoration				0.915
No, n (%)	473 (84.16%)	158 (84.49%)	631 (84.25%)
Yes, n (%)	89 (15.84%)	29 (15.51%)	118 (15.75%)
Hypoxemia				0.429
No, n (%)	467 (83.10%)	160 (85.56%)	627 (83.71%)
Yes, n (%)	95 (16.90%)	27 (14.44%)	122 (16.29%)
Hypoalbuminemia				0.760
No, n (%)	480 (85.41%)	158 (84.49%)	638 (85.18%)
Yes, n (%)	82 (14.59%)	29 (15.51%)	111 (14.82%)
Anemia				0.299
No, n (%)	443 (78.83%)	154 (82.35%)	597 (79.71%)
Yes, n (%)	119 (21.17%)	33 (17.65%)	152 (20.29%)
Elevated serum creatinine				0.461
No, n (%)	537 (95.55%)	181 (96.79%)	718 (95.86%)
Yes, n (%)	25 (4.45%)	6 (3.21%)	31 (4.14%)
Elevated NT-proBNP				0.708
No, n (%)	459 (81.67%)	155 (82.89%)	614 (81.98%)
Yes, n (%)	103 (18.33%)	32 (17.11%)	135 (18.02%)
Surgical type				0.770
FNA, n (%)	154 (27.40%)	51 (27.27%)	205 (27.37%)
Hip arthroplasty, n (%)	408 (72.60%)	136 (72.73%)	544 (72.63%)
Duration of surgery, min, mean ± SD	122.24 ± 31.10	123.71 ± 29.37	123.00 ± 30.08	0.801
Intravenous fluid volumes, mL·kg⁻¹·h⁻¹, mean ± SD	7.58 ± 0.88	7.61 ± 0.32	7.59 ± 0.77	0.131

Abbreviation: BMI Body mass index, ASA American Society of Anesthesiologists, COPD Chronic obstructive pulmonary disease, PFNA Proximal femoral nail antirotation

Identification of risk factors for PPCs

In the training cohort, patients with PPCs differed from individuals without PPCs with respect to age, sex, ASA physical status, time from injury to surgery, lung ultrasonography score, and diaphragmatic mobility. Significant between-group differences were also observed for functional dependence, COPD, smoking status, recent respiratory infection, cough and expectoration, hypoxaemia, hypoalbuminemia, anemia, elevated serum creatinine, and elevated NT-proBNP (P < 0.05) (Table 3).

Table 3.

Demographic and clinical characteristics of patients in the training cohort

Variables	PPCs	non-PPCs	Overall	P
Number	186	376	562
Age, years, median (IQR)	77.0 (69.0, 82.0)	71.0 (68.0, 78.0)	72.0 (68.0, 79.0)	0.000*
Sex				0.002*
Women, n (%)	75 (40.32%)	205 (54.52%)	280 (49.82%)
Men, n (%)	111 (59.68%)	171 (45.48%)	282 (50.18%)
BMI, kg·m− 2, mean ± SD	22.59 ± 3.89	23.22 ± 4.05	22.99 ± 3.87	0.080
ASA physical status class				0.000*
≤ II, n (%)	90 (48.39%)	261 (69.41%)	351 (62.46%)
≥ III, n (%)	96 (51.61%)	115 (30.59%)	211 (37.54%)
Functional dependence				0.033*
No, n (%)	146 (78.49%)	322 (85.64%)	468 (83.27%)
Yes, n (%)	40 (21.51%)	54 (14.36%)	94 (16.73%)
COPD				0.000*
No, n (%)	109 (58.60%)	312 (82.98%)	421 (74.91%)
Yes, n (%)	77 (41.40%)	64 (17.02%)	141 (25.09%)
Diabetes				0.145
No, n (%)	167 (89.78%)	321 (85.37%)	488 (86.83%)
Yes, n (%)	19 (10.22%)	55 (14.63%)	74 (13.17%)
Smoking status				0.011*
No, n (%)	138 (74.19%)	313 (83.24%)	451 (80.25%)
Yes, n (%)	48 (25.81%)	63 (16.76%)	111 (19.75%)
Time from injury to surgery				0.015*
≤ 48 h, n (%)	41 (22.04%)	120 (31.91%)	161 (28.65%)
> 48 h, n (%)	145 (77.96%)	256 (68.09%)	401 (71.35%)
Lung ultrasonography score, median (IQR)	7.0 (3.0, 9.0)	3.0 (2.0, 4.0)	3.0 (3.0, 6.0)	0.000*
Diaphragmatic mobility, cm, median (IQR)	3.2 (2.3, 3.2)	4.3 (3.2, 5.4)	3.2 (3.2, 5.3)	0.000*
Recent respiratory infection				0.000*
No, n (%)	106 (56.99%)	316 (84.04%)	422 (75.09%)
Yes, n (%)	80 (43.01%)	60 (15.96%)	140 (24.91%)
Cough and expectoration				0.005*
No, n (%)	145 (77.96%)	328 (87.23%)	473 (84.16%)
Yes, n (%)	41 (22.04%)	48 (12.77%)	89 (15.84%)
Hypoxemia				0.006*
No, n (%)	143 (76.88%)	324 (86.17%)	467 (83.10%)
Yes, n (%)	43 (23.12%)	52 (13.83%)	95 (16.90%)
Hypoalbuminemia				0.001*
No, n (%)	143 (76.88%)	337 (89.63%)	480 (85.41%)
Yes, n (%)	43 (23.12%)	39 (10.37%)	82 (14.59%)
Anemia				0.000*
No, n (%)	124 (66.67%)	319 (84.84%)	443 (78.83%)
Yes, n (%)	62 (33.33%)	57 (15.16%)	119 (21.17%)
Elevated serum creatinine				0.001*
No, n (%)	170 (91.40%)	367 (97.61%)	537 (95.55%)
Yes, n (%)	16 (8.60%)	9 (2.39%)	25 (4.45%)
Elevated NT-proBNP				0.000*
No, n (%)	132 (70.97%)	327 (87.23%)	459 (81.67%)
Yes, n (%)	54 (29.03%)	49 (12.77%)	103 (18.33%)
Surgical type				0.846
PFNA, n (%)	50 (26.88%)	104 (27.66%)	154 (27.40%)
Hip arthroplasty, n (%)	136 (73.12%)	272 (72.34%)	408 (72.60%)
Duration of surgery, min, mean ± SD	123.47 ± 30.63	121.63 ± 31.35	122.24 ± 31.10	0.510
Intravenous fluid volumes, mL·kg⁻¹·h⁻¹, mean ± SD	7.60 ± 0.90	7.56 ± 0.88	7.58 ± 0.88	0.890

Abbreviation: PPCs Postoperative pulmonary complications, BMI Body mass index, ASA American Society of Anesthesiologists, COPD Chronic obstructive pulmonary disease, PFNA Proximal femoral nail antirotation

*P < 0.05 considered statistically significant

Univariate logistic regression identified age (odds ratio [OR]: 1.074, 95% confidence interval [CI]: 1.047–1.102, P < 0.001), ASA physical status (ASA physical status class ≥ III, OR: 2.421, 95% CI: 1.686–3.476, P < 0.001), functional dependence (OR: 1.634, 95% CI: 1.038–2.570, P = 0.034), COPD (OR: 3.444, 95% CI: 2.316–5.121, P < 0.001), smoking status (OR: 1.728, 95% CI: 1.129–2.645, P = 0.012), time from injury to surgery (> 48 h, OR: 1.658, 95% CI: 1.101–2.495, P = 0.016), lung ultrasonography score (OR: 1.508, 95% CI: 1.394–1.632, P < 0.001), diaphragmatic mobility (OR: 0.583, 95% CI: 0.497–0.684, P < 0.001), respiratory infection (OR: 3.975, 95% CI: 2.663–5.933, P = 0.001), cough and expectoration (OR: 1.932, 95% CI: 1.219–3.062, P = 0.005), hypoxemia (OR: 1.874, 95% CI: 1.195–2.937, P = 0.006), hypoalbuminemia (OR: 2.598, 95% CI: 1.615–4.180, P < 0.001), anemia (OR: 2.798, 95% CI: 1.847–4.238, P < 0.001), elevated serum creatinine (OR: 3.838, 95% CI: 1.626–8.860, P = 0.002), and elevated NT-proBNP (OR: 2.730, 95% CI: 1.765–4.223, P < 0.001) as factors associated with PPCs (Table 4).

Table 4.

Univariate and multivariate logistic regression analyses of risk factors associated with PPCs

Variables	Univariate		Multivariate
	OR (95%CI)	P	AOR (95%CI)	P
Age	1.074 (1.047–1.102)	0.000*	0.973 (0.936–1.011)	0.166
Sex
Women	1.0 (reference)		1.0 (reference)
Men	1.774 (1.242–2.534)	0.002*	1.179 (0.733–1.895)	0.498
BMI	0.961 (0.918–1.005)	0.080
ASA physical status class
≤ II	1.0 (reference)		1.0 (reference)
≥ III	2.421 (1.686–3.476)	0.000*	1.900 (1.136–3.179)	0.014*
Functional dependence
No	1.0 (reference)		1.0 (reference)
Yes	1.634 (1.038–2.570)	0.034*	2.061 (1.117–3.802)	0.021*
COPD
No	1.0 (reference)		1.0 (reference)
Yes	3.444 (2.316–5.121)	0.000*	3.176 (1.848–5.458)	0.000*
Diabetes
No	1.0 (reference)
Yes	0.664 (0.382–1.156)	0.148
Smoking status
No	1.0 (reference)		1.0 (reference)
Yes	1.728 (1.129–2.645)	0.012*	0.767 (0.404–1.458)	0.418
Time from injury to surgery
≤ 48 h	1.0 (reference)		1.0 (reference)
> 48 h	1.658 (1.101–2.495)	0.015*	0.749 (0.425–1.318)	0.316
Lung ultrasonography score	1.508 (1.394–1.632)	0.000*	1.449 (1.327–1.581)	0.000*
Diaphragmatic mobility	0.583 (0.497–0.684)	0.000*	0.601 (0.494–0.732)	0.000*
Recent respiratory infection
No	1.0 (reference)		1.0 (reference)
Yes	3.975 (2.663–5.933)	0.000*	1.914 (1.049–3.490)	0.034*
Cough and expectoration
No	1.0 (reference)		1.0 (reference)
Yes	1.932 (1.219–3.062)	0.005*	1.778 (0.933–3.385)	0.080
Hypoxemia
No	1.0 (reference)		1.0 (reference)
Yes	1.874 (1.195–2.937)	0.006*	0.916 (0.481–1.742)	0.788
Hypoalbuminemia
No	1.0 (reference)		1.0 (reference)
Yes	2.598 (1.615–4.180)	0.000*	2.307 (1.274–4.180)	0.006*
Anemia
No	1.0 (reference)		1.0 (reference)
Yes	2.798 (1.847–4.238)	0.000*	2.035 (1.274–4.180)	0.039*
Elevated serum creatinine
No	1.0 (reference)		1.0 (reference)
Yes	3.838 (1.626–8.860)	0.002*	2.067 (0.683–6.251)	0.199
Elevated NT-proBNP
No	1.0 (reference)		1.0 (reference)
Yes	2.730 (1.765–4.223)	0.000*	1.909 (1.045–3.486)	0.035*
Surgical type
PFNA	1.0 (reference)
Hip arthroplasty	1.040 (0.701–1.544)	0.846
Duration of surgery	1.002 (0.996–1.008)	0.509
Intravenous fluid volumes	0.957 (0.521–1.759)	0.887

Abbreviation: OR Odds ratio, AOR Adjusted odds ratio, CI Confidence interval, PPCs Postoperative pulmonary complications, BMI Body mass index, ASA American Society of Anesthesiologists, COPD chronic obstructive pulmonary disease, PFNA Proximal femoral nail antirotation

*P < 0.05 was considered statistically significant

Multivariate analysis revealed that ASA physical status ≥ III (OR: 1.900, 95% CI: 1.136–3.179, P = 0.014), functional dependence (OR: 2.061, 95% CI: 1.117–3.802, P = 0.021), COPD (OR: 3.176, 95% CI: 1.848–5.458, P < 0.001), respiratory infection (OR: 1.914, 95% CI: 1.049–3.490, P = 0.034), hypoalbuminemia (OR: 2.307, 95% CI: 1.274–4.180, P = 0.006), anemia (OR: 2.035, 95% CI: 1.274–4.180, P = 0.039), and elevated NT-proBNP (OR: 1.909, 95% CI: 1.045–3.486, P = 0.035) were independent risk factors for PPCs. Higher lung ultrasonography score (OR: 1.449, 95% CI: 1.327–1.581, P < 0.001) and reduced diaphragmatic mobility (OR: 0.601, 95% CI: 0.494–0.732, P < 0.001) were significantly associated with an increased risk of PPCs (Table 4).

Nomogram construction and validation

Based on the independent predictors identified, a predictive nomogram was developed to estimate the risk of PPCs (Fig. 6). The Hosmer–Lemeshow test yielded a good model fit (training cohort: χ² = 9.09, P = 0.334; validation cohort: χ² = 8, P = 0.553). Calibration analysis revealed close agreement between the predicted and observed outcomes, with a mean absolute error of 0.018 in the training cohort and 0.016 in the validation cohort (Fig. 7).

Fig. 6.

Nomogram for predicting postoperative pulmonary complications. Nomogram developed to estimate the probability of postoperative pulmonary complications in geriatric patients with hip fracture. Points are assigned for each predictor by drawing a vertical line to the top point row. The total points are summed, and a vertical line is drawn downward from the total point row to determine the probability of postoperative pulmonary complications

Fig. 7.

Calibration plot of the prediction nomogram (A, training cohort; B, validation cohort). Calibration curve comparing predicted and observed probabilities of postoperative pulmonary complications. The solid line represents nomogram predictions, and the diagonal line indicates ideal agreement between predicted and observed outcomes

ROC analysis demonstrated strong discriminative performance of the nomogram, with a higher AUC of 0.897 (95% CI, 0.870–0.924) in the training cohort and 0.869 (95% CI, 0.813–0.926) in the validation cohort (Fig. 8). DCA confirmed good clinical predictive value and applicability (Fig. 9). To enhance the clinical application, we developed a web-based calculator (https://chiwen1109.shinyapps.io/dynnomapp) to predict individual PPC risk.

Fig. 8.

Discriminative performance of the prediction nomogram (A, training cohort; B, validation cohort). The receiver operating characteristic curve illustrates the discriminative ability of the nomogram for predicting postoperative pulmonary complications. The area under the curve reflects overall model performance

Fig. 9.

Decision curve analysis of the prediction nomogram (A, training cohort; B, validation cohort). Decision curve analysis demonstrates the clinical net benefit of the nomogram across a range of threshold probabilities for postoperative pulmonary complications, compared with treat-all and treat-none strategies

Nomogram performance compared with the ARISCAT

In the overall cohort, the nomogram outperformed the ARISCAT. The AUC of the nomogram was 0.860, compared with 0.675 for the ARISCAT, with a statistically significant difference (P < 0.001) (Fig. 10). Additionally, DCA showed that the nomogram provided a greater net benefit across the range of threshold probabilities. (Fig. 11).

Fig. 10.

Comparison of the discriminative ability of the nomogram (red) and the ARISCAT (blue). In the ARISCAT model, four predictors were included while three were excluded: surgical incision site (all patients underwent hip surgery) and emergency status (all elective procedures) demonstrated no variability; duration of surgery > 3 h (n = 3) had an insufficient sample size for reliable estimation

Fig. 11.

Comparison of the decision curve analysis of the nomogram (red) and the ARISCAT (blue)

Discussion

Risk factors of PPCs

In this study, nine variables were incorporated into a nomogram to predict PPCs in geriatric patients undergoing hip fracture surgery: ASA physical status, functional dependence, COPD, lung ultrasonography score, diaphragmatic mobility, recent respiratory infection, hypoalbuminemia, anemia, and elevated NT-proBNP. The model demonstrated favorable discrimination, calibration, and clinical utility, supporting its potential role in perioperative risk stratification. In addition, the study results demonstrated that this nomogram outperformed the ARISCAT.

ASA physical status

ASA physical status is widely used in anesthetic practice and was originally developed to predict perioperative mortality [22]. Our study demonstrated ASA physical status ≥ III as an independent predictor of PPCs in geriatric patients with hip fractures. Geriatric patients with hip fractures classified as ASA physical status class III–IV had approximately 2–3 times the risk of pneumonia than patients in class I–II [23–25]. Duan and co-workers [26] found that in older patients with femoral neck fractures, ASA physical status class was strongly and independently associated with postoperative hypoxemia. Further, Crego-Vita and colleagues [27] suggested that in geriatric patients with hip fracture, ASA physical status class was a statistically significant predictor of postoperative mortality. Therefore, close monitoring should be implemented for geriatric patients with hip fractures classified as ASA physical status class ≥ III.

Functional dependence

Preoperative baseline frailty assessment is recommended for geriatric surgical patients [28]. A variety of validated frailty evaluation tools are currently available; however, comprehensive frailty tools generally require numerous variables and are not always rapidly, easily, and efficiently applicable to the entire patient population [29–31]. Functional dependence, based on activities of daily living, provides a reliable surrogate marker of frailty and can be readily identified during routine clinical evaluation. Although it cannot replace frailty assessment, preoperative functional status evaluation may be more practical due to its relatively straightforward definition. In fact, functional dependence is often regarded as an advanced manifestation of frailty. [32–33] In this study, functional dependence was associated with a twofold higher risk of PPCs, consistent with prior evidence linking dependency to adverse postoperative respiratory and survival outcomes [32–35]. Given its simplicity and clinical relevance, functional dependence may be particularly valuable when validated frailty assessment is not feasible.

COPD

COPD is a well-established risk factor for PPCs [36]. In this cohort, COPD emerged as a strong independent risk factor of PPCs. Previous studies demonstrated higher proportions of readmission, pneumonia, acute respiratory failure, prolonged hospitalization, and mortality in patients with hip fractures and co-existing COPD [37, 38]. These findings underscore the importance of meticulous preoperative respiratory assessment and optimization, including treatment of acute exacerbations, bronchodilator and systemic corticosteroid therapies where appropriate [39].

Lung ultrasonography score

Normal lung tissue maintains a precise air-fluid balance, which is critical for sustaining normal gas exchange function [40]. Pathological conditions, including insufficient ventilation or increased exudation in lung tissue, can lead to an imbalance in the air–fluid ratio of the lung parenchyma [41]. Changes in the lung tissue air–fluid ratio manifest as characteristic signs on lung ultrasonography images, including A-lines, B-lines, the shred sign, and the tissue-like pattern [42]. Quantitative lung ultrasonography scoring has become an established method for evaluating pulmonary disease severity and treatment response [10, 11]. In this study, higher lung ultrasonography scores were significantly correlated with an increased risk of PPCs. A prospective observational study conducted across 11 hospitals in Italy revealed that among older persons with hip fractures, patients who developed postoperative pneumonia had a significantly higher preoperative lung ultrasonography scores than patients in the non-pneumonia group [7]. Additionally, the preoperative lung ultrasonography scores had high predictive efficacy for postoperative major adverse cardiovascular events [8]. Therefore, for patients with high lung ultrasonography scores, preoperative interventions and selection of optimal surgical timing based on ultrasound scoring may yield clinical benefits.

Diaphragmatic mobility

The diaphragm is the primary respiratory muscle in the human body, contributing to more than 60% of the changes in thoracic volume during respiration [43]. Diaphragmatic mobility is an important indicator of diaphragmatic function [44]. Patients with decreased diaphragmatic mobility present with impaired cough capacity, leading to sputum retention. This pathological state may be associated with the progression of postoperative respiratory tract infections and pneumonia [45–47]. Additionally, patients with decreased diaphragmatic mobility exhibit altered respiratory mechanics due to reduced pressure and diaphragmatic pressure [46, 48, 49]. Diaphragmatic mobility is correlated with both forced expiratory volume in 1 s (FEV₁) and maximal voluntary ventilation (MVV) [50]. Both reduced FEV₁ and MVV are associated with the occurrence of PPCs [51, 52]. In this study, decreased diaphragmatic mobility measured by ultrasonography was significantly associated with an increased risk of PPCs. This aligns with prior evidence linking diaphragmatic mobility with PPCs following thoracic surgery [53]. Identifying impaired diaphragmatic mobility preoperatively may allow implementation of targeted respiratory interventions, including incentive spirometry, diaphragmatic breathing exercises, and manual diaphragmatic release techniques both preoperatively and postoperatively, to reduce the risk of PPCs [54, 55].

Recent respiratory infection

Recent respiratory infection is a strong risk factor for PPCs [56, 57], due to persistent airway hyper-reactivity, decreased pulmonary function, and impaired immune function [6]. This risk may be amplified in patients with trauma, with disrupted immune homeostasis, predisposing patients to infections [58, 59]. Given the persistence of airway inflammation following the resolution of infectious pathogens, Lumb and co-workers [60] suggested that for patients with respiratory infections or inadequately treated respiratory diseases, surgery should be performed only after complete treatment of the condition, although this may be inappropriate for patients with trauma. Nebulized lidocaine may reduce airway hyperreactivity induced by upper respiratory tract infection, with recent respiratory tract infections leading to excessive tracheal secretions, potentially associated with postoperative pneumonia [61]. Prophylactic mucolytics may be beneficial in patients with a recent history of respiratory infection. Ambroxol can reduce the viscosity of bronchial sputum, thereby aiding sputum expectoration.

Hypoalbuminemia

Preoperative malnutrition is a potential risk and prognostic factor for poor outcomes following hip fracture surgery [62]. Serum albumin is a routine indicator of malnutrition, and patients with serum albumin levels < 35 g·L^{− 1} are considered to have poor nutritional status [63]. Hypoalbuminemia is relatively common in older patients. In this study, 14.81% of the patients had preoperative hypoalbuminemia, comparable to the 17.36% previously reported [64]. Our findings demonstrated that hypoalbuminemia was an independent risk factor of PPCs in older patients with hip fractures. Older patients with femoral neck fractures who had preoperative hypoalbuminemia are at a markedly higher risk of postoperative pneumonia [64]. Hypoalbuminemia is an important contributor to severe complications following total knee arthroplasty, including PPCs [65]. Therefore, in older patients with hip fractures, patients with preoperative hypoalbuminemia should be prioritized. Preoperative nutritional support, including increased protein and energy intake, is beneficial for reducing postoperative complications [66–68].

Anemia

Anemia on admission is relatively frequent among patients with hip fractures, ranging from 12.3% to 40.4% [69]. In our study, 20.29% of patients with hip fractures had preoperative anemia, and preoperative anemia doubled the risk of PPCs. Preoperative anemia is closely associated with postoperative pulmonary infection, respiratory failure, and postoperative mortality risk in patients with hip fractures [70, 71]. However, findings from studies regarding the effect of preoperative anemia intervention on clinical outcomes are contradictory. Kouyoumdjian and co-workers [72] suggested that perioperative blood transfusion may confer potential benefits for gastric cancer patients undergoing gastrectomy. Nevertheless, among patients with hip fractures complicated with anemia, the incidence of postoperative complications does not show a significant reduction following iron supplementation therapy [73]. Additionally, blood transfusions could increase the risk of mortality in patients [74].

N-terminal pro-B-type natriuretic peptide

Preoperative heart failure is closely associated with the PPCs in patients undergoing hip surgery [75, 76]. NT-proBNP is an established indicator of cardiac dysfunction and heart failure severity [77]. Elevated NT-proBNP levels were defined as: a level of ≥ 250 pg·mL^{− 1} for patients aged 65–74 years, and ≥ 500 pg·mL^{− 1} for patients ≥ 75 years old [20]. In our study, individuals with elevated NT-proBNP levels exhibited a twofold risk of PPCs. Incorporating NT-proBNP into preoperative assessments may therefore improve identification of patients at high risk of PPCs and inform perioperative management strategies.

Limitations

This study has some limitations. First, proficiency in bedside lung ultrasound is a prerequisite for the effective clinical application of this predictive model, and thus training in ultrasound operation for anesthesiologists is essential. Second, only a 2-week postoperative follow-up was conducted in this study, and the long-term survival status of patients was not analyzed. Third, not all potential factors influencing PPCs were evaluated, including the timing of postoperative tracheal extubation and anesthetic technique. A previous study confirmed that delayed extubation (> 4 h) was associated with an increased risk of postoperative pulmonary infection and prolonged length of hospital stay [78]. Existing evidence regarding the impact of anesthesia type on PPCs in hip fracture surgery remains inconsistent [79–81], and further research is needed to clarify this relationship. Fourth, although the developed nomogram may facilitate perioperative risk stratification and provide supportive information for clinical evaluation, the current investigation does not demonstrate that implementation of this model alters clinical decision‑making or improves patient outcomes. Therefore, the clinical utility of the nomogram remains preliminary and should be interpreted with considerable caution. Fifth, given that both model development and validation were performed within a single‑center cohort, the generalizability of the findings is limited. Future multicenter external validation and prospective interventional studies are warranted to verify whether this predictive tool can be effectively translated into clinical practice and improve perioperative care.

Conclusions

This study evaluated bedside ultrasonography alongside conventional clinical parameters in geriatric patients with hip fractures. Nine variables were incorporated into a nomogram to predict PPCs: ASA physical status class, functional dependence, COPD, lung ultrasonography score, diaphragmatic mobility, recent respiratory infection, hypoalbuminemia, anemia, and elevated NT-proBNP level. Validation indicated that the nomogram had favorable predictive ability for PPCs. This preliminary model may serve as a supplementary reference for perioperative risk stratification, although multicenter prospective intervention trials are still required to verify the reliability and stability of the model.

Abbreviations

PPCs

Postoperative pulmonary complications

ARISCAT

Assess Respiratory Risk In Surgical Patients In Catalonia

ASA

American Society of Anesthesiologists

AUC

Area under the receiver operating characteristic curve

PPV

Pulse pressure variation

MAP

Mean arterial pressure

COPD

Chronic obstructive pulmonary disease

NT-proBNP

N-terminal pro-B-type natriuretic peptide

ROC

Receiver-operating characteristic curve

DCA

Decision curve analysis

OR

Odds ratio

CI

Confidence interval

Authors’ contributions

Conceptualization: Jianhong Hao, Mei Yang. Protocol development: Jianhong Hao, Peng Pang, Zhirong Wang. Data collection: Xiaobing Liu, Wen Chi, Wenbo Cai, Li Zhang, Wangyang Li. Data analysis: Xiaobing Liu, Zhenguo Luo. Writing and editing manuscript: Jianhong Hao, Zhenguo Luo. All authors read and approved the final manuscript.

Funding

This work was funded by the Key Research and Development Project of Shaanxi Province (NO. 2023-YBSF-069).

Data availability

The data analyzed in this study is subject to the following licenses/restrictions: the datasets used and analysed during the current study are available from the corresponding author on reasonable request. Requests to access these datasets should be directed to Jianhong Hao, haojianhong@xashhyy9.wecom.work, or haojianhong722@163.com .

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Honghui Hospital, Xi’an Jiaotong University, Xi’an, Shaanxi Province, China (Chairperson: Professor Kun Zhang) on September 14, 2024, with the ethics approval number 20240915. All study methods were performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from each patient or their family member. As a non‑interventional observational study, clinical trial registration was exempted for this research in accordance with ICMJE guidelines and ChiCTR requirements.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.