Development and validation of a predictive model for surgical site infection following joint surgery

Abstract

Background

Surgical site infections (SSIs) are common complications after joint arthroplasty, leading to increased morbidity and healthcare costs. Traditional models, like the National Nosocomial Infections Surveillance (NNIS) system, have limitations in predicting SSI risk due to a lack of patient-specific factors. This study aimed to create and validate a predictive model focusing on hypoproteinemia to enhance SSI risk assessment in joint surgery patients.

Methods

A retrospective cohort study of 726 patients undergoing joint arthroplasty between 2020 and 2022 was conducted. Data included demographics, laboratory values, and surgical details. Univariate and multivariate analyses identified key predictors, including hypoproteinemia, to develop a predictive nomogram. Model validation was performed using receiver operating characteristic curves, calibration, and decision curve analysis (DCA), comparing it to the NNIS model.

Results

Hypoproteinemia was a significant independent predictor of SSI, with the new model outperforming the NNIS system (area under the curve: 0.829 vs. 0.534). Calibration analysis showed excellent agreement between predicted and observed probabilities, with a mean absolute error of 0.009. DCA further confirmed the model’s clinical utility, showing a higher net benefit across various thresholds compared to traditional approaches.

Conclusions

Hypoproteinemia is a critical risk factor for SSI in joint arthroplasty. The new predictive model offers improved risk stratification, supporting a more personalized approach to perioperative management in orthopedic surgery.

Highlight box.

Key findings

• Hypoproteinemia, defined as serum albumin below 35 g/L, emerged as an independent risk factor for surgical site infection (SSI) following joint arthroplasty. A novel predictive model that integrates hypoproteinemia, body mass index (BMI), and other key clinical variables achieved an impressive accuracy, with an area under the curve (AUC) of 0.829—substantially outperforming the traditional National Nosocomial Infections Surveillance (NNIS) system, which recorded an AUC of only 0.534. Furthermore, the model demonstrated excellent calibration (mean absolute error of 0.009) and provided enhanced clinical utility across a range of risk thresholds, as evidenced by decision curve analysis.

What is known and what is new?

• SSI remains a formidable complication in joint surgery, and established predictive tools like the NNIS system have typically overlooked patient-specific factors such as nutritional status.

• Our work introduces hypoproteinemia as a critical predictive factor and presents a validated nomogram that tailors risk stratification by incorporating individual variables. This approach marks a significant step forward in personalizing risk assessments for orthopedic surgical patients.

What is the implication, and what should change now?

• Routine preoperative screening should incorporate assessment for hypoproteinemia, and targeted nutritional interventions, including albumin supplementation when appropriate, should be considered to mitigate SSI risk. Clinicians are encouraged to adopt this more individualized predictive model in favor of generic systems like NNIS. Additionally, multi-center, prospective studies are warranted to further validate the model’s generalizability and to refine its integration into standardized perioperative protocols.

Introduction

Globally, over 1.5 million joint arthroplasties are performed annually, driven by aging populations and rising osteoarthritis prevalence (1). Surgical site infections (SSIs) are a major concern in orthopedic surgery, particularly following joint arthroplasty, where they can significantly impact patient recovery and healthcare costs. Based on the data from the European Centre for Disease Prevention and Control (ECDC’s) coordinated SSI surveillance in 13 European Union/European Economic Area (EU/EEA) countries during 2018–2020, the percentage of SSIs varied from 0.6% in knee prosthesis surgery to 9.5% in open colon surgery, depending on the type of surgical procedure (2). Despite advancements in surgical techniques and perioperative care, SSI rates in primary and revision joint procedures remain substantial, ranging from 0.5% to 10% depending on the complexity of the case (3-5). Accurate identification of high-risk patients is crucial for implementing targeted preventive measures.

Traditional predictive tools like the National Nosocomial Infections Surveillance (NNIS) system are widely used to estimate SSI risk across different surgical fields (6). However, these models often fall short in orthopedic contexts due to their reliance on general factors such as wound classification and operative duration, failing to account for specific patient characteristics like nutritional status and comorbidities (7). This limitation underscores the need for a more tailored predictive approach in joint surgery.

Recent studies have highlighted the role of patient-specific factors—such as body mass index (BMI), diabetes, and nutritional indicators—in predicting postoperative complications, including SSIs (8,9). Hypoproteinemia, as measured by low serum albumin levels, has been associated with poor wound healing and increased infection risk, making it a potential key predictor in orthopedic cases (7). This study aims to develop a predictive model focusing on hypoproteinemia, integrating patient and surgical variables to enhance the accuracy of SSI risk assessment compared to the traditional NNIS model. We present this article in accordance with the TRIPOD reporting checklist (available at https://aoj.amegroups.com/article/view/10.21037/aoj-25-14/rc).

Methods

Study design

This study was a retrospective cohort analysis conducted to develop a predictive model for SSI following joint arthroplasty. We reviewed medical records from a consecutive series of patients who underwent joint surgery at North China Healthcare Group Xingtai General Hospital between January 2020 and December 2022. Patients were identified using the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) codes (M00-M25) and manually cross-verified. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the ethics committee of the North China Healthcare Group Xingtai General Hospital (No. ZCKT-2021-0025) and informed consent was taken from all the patients.

Inclusion criteria:

1. Patients aged 18 years or older;

2. Undergoing joint surgery;

3. Availability of complete preoperative, intraoperative, and postoperative data;

4. No evidence of infection at the time of surgery.

Exclusion criteria:

1. Patients with active systemic infections or immunocompromised status;

2. History of joint surgery within the past 6 months;

3. Patients with severe liver or renal dysfunction prior to surgery;

4. Patients with significant data missing.

Definition of SSI

SSI was defined according to the guidelines provided by the United States Centers for Disease Control and Prevention (CDC) (3). In this study, an SSI was considered present if any of the following criteria were met within 90 days of surgery:

1. Presence of purulent drainage from the surgical site;

2. Positive microbial culture from an aseptically obtained specimen of fluid or tissue from the surgical site;

3. Diagnosis of infection by the attending surgeon based on clinical symptoms, including localized pain, redness, swelling, or fever, confirmed with laboratory or radiological findings.

Data collection

Data were collected retrospectively from the hospital’s electronic medical record system. Preoperative data included demographic information (age, gender, BMI), comorbidities (diabetes mellitus, cardiovascular disease, hypertension), laboratory values (serum albumin, hemoglobin), and smoking status. Intraoperative variables included surgical duration, estimated blood loss, type of anesthesia, and the use of cemented or uncemented prosthesis. Postoperative data encompassed wound characteristics, antibiotic usage, length of hospital stay, and the occurrence of SSI. Hypoproteinemia was defined as serum albumin <35 g/L (10).

Perioperative management

All patients underwent a standardized perioperative management protocol. Preoperative assessments included routine blood tests, imaging studies, and optimization of existing medical conditions such as diabetes and hypertension. Patients were advised to cease smoking at least two weeks prior to surgery. Prophylactic antibiotics (cefazolin, or vancomycin in case of penicillin allergy) were administered within 60 minutes before the surgical incision and continued for 24 hours postoperatively (2). Surgeries were performed by an experienced team of orthopedic surgeons, adhering to strict aseptic techniques. Postoperatively, wound dressings were changed every 48 hours until healing was confirmed. Deep vein thrombosis prophylaxis included low-molecular-weight heparin and early mobilization.

Statistical analysis

Normality tests were performed for continuous variables. Normally distributed data were expressed as mean ± standard deviation (SD), while non-normally distributed data were presented as median (interquartile range). Categorical variables were described using frequencies and percentages. An independent t-test was used to compare continuous variables between patients with and without SSI, while the Chi-squared test or Fisher’s exact test was utilized for categorical data. Candidate variables with a P value <0.05 in the univariate analysis were selected for the multivariate logistic regression model to identify independent risk factors for SSI. BMI was analyzed as a continuous variable in both univariate and multivariate models.

A predictive nomogram was developed using significant variables from the multivariate analysis. Calibration of the model was performed using the Hosmer-Lemeshow goodness-of-fit test and by plotting a calibration curve, comparing predicted probabilities with observed outcomes. Decision curve analysis (DCA) was employed to evaluate the clinical utility of the predictive model. Statistical analyses were conducted using SPSS version 27.0 (IBM Corp, Armonk, NY) and R software version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria). A P value <0.05 was considered statistically significant for all analyses.

Results

Patient characteristics

A total of 726 patients who underwent joint surgery were included in the study, of whom 11 cases developed SSI (Figure 1). The median age was 61 years, with an interquartile range (IQR) of 51–68 years. The median BMI was 25.71 kg/m² (IQR, 23.46–28.39 kg/m²). Of the cohort, 62% were female, and 38% were male. No significant differences in age and BMI were observed between patients who developed SSI and those who did not, although the median BMI in patients with SSI tended to be slightly higher (Table 1).

Figure 1.

Study flow diagram. A flowchart illustrating the inclusion and exclusion of patients in the study cohort, detailing the process of data extraction and analysis. SSI, surgical site infection.

Table 1. Patient demographics and clinical characteristics.

Variables	Total (n=726)	Non-SSI group (n=715)	SSI group (n=11)
Age (years)	61 [51, 68]	61 [51, 68]	56 [52, 68]
BMI (kg/m2)	25.71 [23.46, 28.39]	25.71 [23.51, 28.34]	27.34 [22.7, 31.28]
Gender
Female	448 [62]	440 [62]	8 [73]
Male	278 [38]	275 [38]	3 [27]
Operating time (min)	60.04 [48.41, 74.36]	60 [48.06, 74.31]	65.7 [53.15, 74.52]
ASA
I	116 [16]	115 [16]	1 [9]
II	533 [73]	525 [73]	8 [73]
III	77 [11]	75 [10]	2 [18]
NNIS
Class 0	642 [88]	633 [89]	9 [82]
Class 1+2	84 [12]	82 [11]	2 [18]
Drainage
No	567 [78]	557 [78]	10 [91]
Yes	159 [22]	158 [22]	1 [9]
Standardized grafting
Yes	607 [84]	597 [83]	10 [91]
No	38 [5]	38 [5]	0 [0]
Unprepared	81 [11]	80 [11]	1 [9]
Rational antibiotic utilization
Yes	308 [42]	302 [42]	6 [55]
No	145 [20]	142 [20]	3 [27]
Non-use	273 [38]	271 [38]	2 [18]
Hypertension
No	469 [65]	462 [65]	7 [64]
Yes	257 [35]	253 [35]	4 [36]
Diabetes
No	638 [88]	628 [88]	10 [91]
Yes	88 [12]	87 [12]	1 [9]
Preoperative hospitalization duration (days)	3 [2, 4]	3 [2, 4]	2 [2, 3]
WBC (109/L)	6.16 [5.13, 7.43]	6.16 [5.12, 7.43]	7.34 [5.65, 7.49]
Neutrophils (109/L)	3.7 [2.9, 4.87]	3.68 [2.9, 4.87]	4.22 [3.22, 4.93]
Lymphocyte (109/L)	1.69 [1.38, 2.09]	1.69 [1.38, 2.09]	1.81 [1.48, 1.98]
RBC (1012/L)	4.4±0.5	4.4±0.5	4.4±0.4
Hemoglobin (g/L)	133 [123, 144]	133 [123, 144]	134 [123, 139.5]
Platelet (109/L)	238.5 [200, 283.75]	238 [198, 283]	238.5 [200, 283.75]
AST (U/L)	18 [15, 22]	18 [15, 22]	18 [15, 22]
ALT (U/L	17 [13, 25]	17 [13, 25]	17 [13, 25]
Albumin (g/L)	41.8 [39.5, 43.9]	41.8 [39.5, 43.9]	41.8 [39.5, 43.9]
Total bilirubin (μmol/L)	15.45 [12.22, 19.78]	15.5 [12.3, 19.9]	15.45 [12.22, 19.78]
Direct bilirubin (μmol/L)	3.3 [2.5, 4.5]	3.3 [2.5, 4.5]	3.3 [2.5, 4.5]
ALP (U/L)	82 [67, 100]	81 [67, 100]	82 [67, 100]
GGT (U/L)	21 [15, 31]	21 [15, 31]	21 [15, 31]
Creatinine (μmol/L)	59.9 [51.8, 69.77]	59.9 [51.8, 69.85]	59.9 [51.8, 69.77]
Cystatin C (mg/L)	0.89 [0.8, 1.02]	0.89 [0.8, 1.02]	0.89 [0.8, 1.02]
PT (s)	11.6 [11.1, 12.3]	11.6 [11.1, 12.3]	11.6 [11.1, 12.3]
INR	1.01 [0.96, 1.07]	1.01 [0.96, 1.07]	1.01 [0.96, 1.07]
APTT (s)	27.6 [26.2, 29.28]	27.6 [26.2, 29.25]	27.6 [26.2, 29.28]
D-dimer (μg/mL)	0.46 [0.23, 1.08]	0.46 [0.23, 1.06]	0.46 [0.23, 1.08]

Data are presented as median [interquartile range], mean ± standard deviation or No. [%]. ALP, alkaline phosphatase; ALT, alanine aminotransferase; APTT, activated partial thromboplastin time; ASA, American Society of Anesthesiologists; AST, aspartate aminotransferase; BMI, body mass index; GGT, gamma-glutamyl transferase; INR, international normalized ratio; IQR, interquartile range; NNIS, National Nosocomial Infections Surveillance; PT, prothrombin time; RBC, red blood cell; SSI, surgical site infection; WBC, white blood cell.

Independent risk factors for SSI

Univariate and multivariate logistic regression analysis identified BMI, hemoglobin levels, hypoproteinemia status and platelet count as significant predictors of SSI. The adjusted odds ratio (OR) with 95% confidence interval (CI) of 1.040 (1.033–1.216) for BMI, 1.018 (1.013–1.075) for hemoglobin levels, 2.576 (2.166–106.321) for Hypoproteinemia status and 1.008 (1.004–1.023) for platelet count, suggesting that higher BMI and hypoproteinemia status, lower hemoglobin levels and platelet count are associated with an increased risk of SSI (Table 2).

Table 2. Univariate and multivariate analysis of risk factors for SSI.

Characteristics	Univariate			Multivariate
	OR (95% CI)	P value		OR (95% CI)	P value
(Intercept)	–	–		21.934 (0.001–0.002)	0.001
Age (years)	1.004 (0.963–1.054)	0.87		–	–
BMI (kg/m2)	1.088 (1.002–1.159)	0.02		1.040 (1.033–1.216)	0.002
Gender	0.600 (0.131–2.094)	0.45		–	–
Operating time (min)	1.001 (0.979–1.016)	0.95		–	–
ASA	1.751 (0.549–5.507)	0.34		–	–
NNIS	1.715 (0.259–6.799)	0.49		–	–
Drainage	0.353 (0.019–1.861)	0.32		–	–
Standardized grafting	0.764 (0.163–1.864)	0.63		–	–
Rational antibiotic utilization	0.654 (0.295–1.299)	0.25		–	–
Hypertension	1.043 (0.271–3.489)	0.94		–	–
Diabetes	0.722 (0.039–3.837)	0.76		–	–
Preoperative hospitalization duration (days)	1.056 (0.809–1.243)	0.60		–	–
WBC (109/L)	1.066 (0.759–1.420)	0.69		–	–
Hemoglobin (g/L)	1.023 (1.004–1.046)	0.01		1.018 (1.013–1.075)	0.003
Neutrophils (109/L)	1.073 (0.723–1.490)	0.69		–	–
Lymphocyte (109/L)	0.918 (0.300–2.195)	0.87		–	–
RBC (1012/L)	0.829 (0.258–2.706)	0.75		–	–
Hypoproteinemia	4.903 (0.727–20.052)	0.048		2.576 (2.166–106.321)	0.002
Platelet (109/L)	1.01 (1.004–1.016)	0.001		1.008 (1.004–1.023)	0.001
AST (U/L)	0.937 (0.817–1.013)	0.28		–	–
ALT (U/L)	1.005 (0.973–1.018)	0.58		–	–
Total bilirubin (μmol/L)	0.944 (0.837–1.035)	0.29		–	–
Direct bilirubin (μmol/L)	0.892 (0.589–1.168)	0.51		–	–
ALP (U/L)	0.997 (0.974–1.008)	0.75		–	–
GGT (U/L)	0.998 (0.960–1.008)	0.83		–	–
Creatinine (μmol/L)	1.003 (0.958–1.044)	0.89		–	–
Cystatin C (mg/L)	0.261 (0.009–0.684)	0.41		–	–
PT (s)	0.822 (0.405–1.600)	0.57		–	–
INR	0.138 (0.001–297.119)	0.62		–	–
APTT (s)	0.945 (0.742–1.183)	0.64		–	–
D-dimer (μg/mL)	1.123 (0.920–1.268)	0.11		–	–

ALP, alkaline phosphatase; ALT, alanine aminotransferase; APTT, activated partial thrombin time; ASA, American Society of Anesthesiologists; AST, aspartate transaminase; BMI, body mass index; CI, confidence interval; GGT, γ-glutamine acylase; INR, international normalized ratio; NNIS, National Nosocomial Infections Surveillance; OR, odds ratio; PT, prothrombin time; RBC, red blood cell; SSI, surgical site infection; WBC, white blood cell.

Comparative analysis between the hypoproteinemia and normal groups

Patients were categorized based on serum protein levels into two groups: the normal protein group (n=693) and the hypoproteinemia group (n=33). The hypoproteinemia group was significantly older, with a median age of 70 years (IQR, 66–75 years), compared to 61 years in the normal group (P=0.001). Additionally, the median BMI was lower in the hypoproteinemia group at 23.88 kg/m² compared to 25.81 kg/m² in the normal group (P=0.001) (Table 3).

Table 3. Comparative analysis between hypoproteinemia and normal protein groups.

Variables	Total (n=726)	Normal protein (n=693)	Hypoproteinemia (n=33)	P value
Age (years)	61 [51, 68]	61 [51, 67]	70 [66, 75]	0.001
BMI (kg/m2)	25.71 [23.51, 28.4]	25.81 [23.66, 28.58]	23.88 [20.96, 25.71]	0.001
Gender				0.43
Female	448 [62]	425 [61]	23 [70]
Male	278 [38]	268 [39]	10 [30]
Operating time (min)	60.04 [48.41, 74.36]	60.12 [48.38, 74.58]	56.47 [49.85, 70.25]	0.95
ASA				0.001
I	116 [16]	116 [17]	0 [0]
II	533 [73]	511 [74]	22 [67]
III	77 [11]	66 [10]	11 [33]
NNIS				0.001
Class 0	642 [88]	622 [90]	20 [61]
Class 1+2	84 [12]	71 [10]	13 [39]
Drainage				0.02
No	567 [78]	547 [79]	20 [61]
Yes	159 [22]	146 [21]	13 [39]
Standardized grafting				0.33
Yes	607 [84]	576 [83]	31 [94]
No	38 [5]	38 [5]	0 [0]
Unprepared	81 [11]	79 [11]	2 [6]
Rational antibiotic utilization				0.001
Yes	308 [42]	284 [41]	24 [73]
No	145 [20]	139 [20]	6 [18]
Non-use	273 [38]	270 [39]	3 [9]
Hypertension				0.76
No	469 [65]	449 [65]	20 [61]
Yes	257 [35]	244 [35]	13 [39]
Diabetes				>0.99
No	638 [88]	609 [88]	29 [88]
Yes	88 [12]	84 [12]	4 [12]
Preoperative hospitalization duration (days)	3 [2, 4]	3 [2, 4]	4 [2, 6]	0.003
SSI				0.08
No	715 [98]	684 [99]	31 [94]
Yes	11 [2]	9 [1]	2 [6]
WBC (109/L)	6.16 [5.13, 7.43]	6.16 [5.13, 7.43]	6.12 [5.49, 7.61]	0.55
Neutrophils (109/L)	3.7 [2.9, 4.87]	3.67 [2.89, 4.82]	4.17 [3.5, 5.66]	0.039
Lymphocyte (109/L)	1.69 [1.38, 2.09]	1.71 [1.4, 2.11]	1.17 [1.03, 1.62]	0.001
RBC (1012/L)	4.43±0.51	4.46±0.49	3.89±0.57	0.001
Hemoglobin (g/L)	133 [123, 144]	134 [124, 145]	115 [108, 122]	0.001
Platelet (109/L)	238.5 [199.25, 283.75]	241 [201, 284]	218 [193, 267]	0.19
AST (U/L)	18 [15, 22]	18 [15, 22]	17 [15, 23]	0.98
ALT (U/L)	17 [13, 25]	18 [13, 25]	15 [11, 25]	0.13
Albumin (g/L)	41.8 [39.5, 43.9]	41.9 [39.8, 44.1]	33 [30.6, 33.8]	0.001
Total bilirubin (μmol/L)	15.4 [12.22, 19.78]	15.4 [12.3, 19.7]	14.3 [11.7, 19.9]	0.76
Direct bilirubin (μmol/L)	3.3 [2.5, 4.5]	3.2 [2.5, 4.5]	4.8 [3, 6.3]	0.003
ALP (U/L)	81 [67, 101]	82 [68, 101]	75 [61, 108]	0.55
GGT (U/L)	21 [15, 31]	21 [15, 31]	17 [12, 32]	0.11
Creatinine (μmol/L)	59.95 [51.9, 69.7]	60.3 [51.9, 69.9]	56.7 [51.8, 62.9]	0.07
Cystatin C (mg/L)	0.89 [0.8, 1.03]	0.89 [0.8, 1.03]	0.92 [0.83, 1.02]	0.39
PT (s)	11.6 [11.1, 12.3]	11.6 [11.1, 12.3]	12.4 [11.5, 13.1]	0.001
INR	1.01 [0.96, 1.07]	1.01 [0.96, 1.07]	1.08 [1, 1.13]	0.001
APTT (s)	27.6 [26.13, 29.3]	27.6 [26.1, 29.1]	29.6 [28.2, 31.8]	0.001
D-dimer (μg/mL)	0.46 [0.23, 1.07]	0.44 [0.22, 0.94]	2.41 [1.14, 5.42]	0.001

Data are presented as median [interquartile range], mean ± standard deviation or No. [%]. ALP, alkaline phosphatase; ALT, alanine aminotransferase; APTT, activated partial thromboplastin time; ASA, American Society of Anesthesiologists; AST, aspartate aminotransferase; BMI, body mass index; GGT, gamma-glutamyl transferase; INR, international normalized ratio; IQR, interquartile range; NNIS, National Nosocomial Infections Surveillance; PT, prothrombin time; RBC, red blood cell; SSI, surgical site infection; WBC, white blood cell.

Development of the predictive model

The predictive model for SSI following joint arthroplasty was constructed based on the multivariate analysis. Significant factors, including BMI, were integrated into the final model. A nomogram was developed to visually represent the model, allowing clinicians to estimate the probability of SSI by assigning a score to each predictor (Figure 2).

Figure 2.

Nomogram for predictive model of SSI. A visual representation of the predictive model, allowing clinicians to estimate the probability of SSI based on patient-specific variables such as BMI (kg/m²), hemoglobin, platelet count, and serum albumin. BMI, body mass index; SSI, surgical site infection.

Calibration of the predictive model

The model’s calibration was evaluated using a calibration curve, which demonstrated a high correlation between predicted and observed probabilities of SSI. The mean absolute error for the model was 0.009, suggesting reliable prediction performance (Figure 3).

Figure 3.

Calibration curve for predictive model. A graph showing the agreement between predicted and observed probabilities of SSI, with calibration accuracy indicated by the closeness to the ideal line. SSI, surgical site infection.

Comparison with the NNIS prediction model

The predictive efficiency of the developed model was compared with the NNIS system. Using receiver operating characteristic (ROC) curve analysis, the area under the curve (AUC) for the developed model was 0.829 (95% CI: 0.701–0.958), outperforming the NNIS model, which had an AUC of 0.534 (95% CI: 0.413–0.654) (Figure 4). Furthermore, the proposed model had a higher negative predictive value (NPV) of 0.996 compared to 0.986 for the NNIS model. Although the NNIS system exhibited a higher specificity of 0.885, the new model showed superior sensitivity at 0.818 versus 0.182 for the NNIS, indicating a better ability to identify true positive cases of SSI (Table 4).

Figure 4.

ROC curve comparing predictive model with NNIS system. A ROC curve comparing the predictive performance of the developed model and the NNIS, showing the AUC for both. AUC, area under the curve; CI, confidence interval; NNIS, National Nosocomial Infections Surveillance; ROC, receiver operating characteristic.

Table 4. Predictive accuracy metrics for developed model vs. NNIS system.

Models	NPV	PPV	Threshold	Specificity	Sensitivity	Accuracy	Precision	Recall
Nomogram model	0.996	0.042	0.01	0.716	0.818	0.718	0.042	0.818
NNIS system	0.986	0.024	0.5	0.885	0.182	0.875	0.024	0.182

NNIS, National Nosocomial Infections Surveillance; NPV, negative predictive value; PPV, positive predictive value.

DCA

DCA was used to assess the clinical utility of the new predictive model. The net benefit curve for the proposed model consistently outperformed both the NNIS model and the “treat all” or “treat none” strategies across a wide range of threshold probabilities, suggesting that the new model provides a greater net benefit in clinical decision-making (Figure 5).

Figure 5.

DCA for clinical utility. A DCA demonstrating the net benefit of the predictive model versus the NNIS system across various risk thresholds. DCA, decision curve analysis; NNIS, National Nosocomial Infections Surveillance.

Discussion

This study presents a novel predictive model for SSI in patients undergoing joint surgery with a particular focus on hypoproteinemia as a key independent risk factor. The inclusion of this factor significantly enhances the predictive accuracy compared to traditional models like the NNIS system (6). Our findings suggest that addressing nutritional deficits, particularly hypoproteinemia, could play a critical role in reducing SSI rates and improving patient outcomes in orthopedic surgery. For patients with hypoproteinemia, preoperative nutritional interventions (e.g., albumin infusion or dietary supplementation) may mitigate SSI risk.

The development of a tailored predictive model for SSIs in joint surgery, represents a significant advance in the field of orthopedic surgery. The use of hypoproteinemia as a central variable is a novel approach, providing a more nuanced assessment of risk. Hypoproteinemia, indicated by low serum albumin levels, has been shown to correlate with poor wound healing and increased infection risk, making it a critical, yet often overlooked, predictor of postoperative outcomes (7,11). Our model, by directly incorporating this factor, offers a more accurate and clinically relevant risk assessment than the NNIS model, which relies on more generic criteria such as wound classification and operative duration (12-14).

The ability of our model to achieve a higher sensitivity (0.818) compared to the NNIS (0.182) has important clinical implications. It suggests that the model is particularly effective in identifying true positive cases, which is crucial for implementing preventive strategies and optimizing antibiotic use. This aligns with recent calls in the surgical literature for more personalized and patient-specific approaches to SSI risk stratification, especially in high-risk populations (15,16). While our model outperformed the NNIS system in our cohort, its generalizability requires validation in external settings.

In recent years, predictive modeling has gained traction in the domain of surgical outcomes, driven by advances in machine learning and big data analytics. Similar studies have emphasized the importance of integrating patient-specific variables, such as comorbidities and laboratory markers, into predictive models. For instance, a study demonstrated that including preoperative inflammatory markers such as C-reactive protein (CRP) and white blood cell counts could improve the accuracy of SSI prediction in colorectal surgery (17). Similarly, Other relevant studies incorporated frailty indices and nutritional assessments into a predictive model for complications following spinal surgery, underscoring the relevance of nutritional status in postoperative risk assessments (11,18).

These studies echo our findings that serum albumin, a marker of nutritional status, is a critical predictor of postoperative infections (19,20). However, our study is unique in directly comparing the new model with the NNIS system, demonstrating clear advantages in terms of sensitivity and clinical applicability. Moreover, our emphasis on hypoproteinemia is supported by growing evidence linking hypoalbuminemia to adverse outcomes across various surgical specialties, from cardiac to gastrointestinal surgery (21,22).

While many studies have acknowledged the role of nutritional status in surgical outcomes, few have placed it at the core of a predictive model specifically designed for orthopedic joint surgery. Our model’s superior performance highlights the importance of this approach. Unlike the NNIS, which does not account for nutritional factors, our model integrates BMI, serum albumin, and other relevant variables, allowing for a more precise and context-specific risk assessment (17).

Age and BMI are recognized as key predictors of surgical outcomes, and our data (Table 3) suggested a link between these factors, serum protein status, and SSI risk. However, age and BMI are not independent; advanced age often comes with comorbidities that cause cachexia, altering BMI and serum protein levels. For example, older patients with cancer or heart failure are prone to involuntary weight loss. Our study has limitations in untangling these complex relationships. Future research should use multivariate analysis to better understand the independent and combined impacts of age, comorbidities, and BMI on SSI risk, facilitating the creation of more accurate risk prediction models and personalized prevention strategies.

Another distinction of this study is its comprehensive validation process, including calibration and DCA. The model demonstrated excellent calibration, with a mean absolute error of 0.009, indicating high reliability. DCA further highlighted the model’s clinical utility, revealing a greater net benefit across a wide range of threshold probabilities compared to traditional strategies such as “treat all” or “treat none” (23). This reinforces the model’s potential to inform clinical decision-making and optimize resource use in high-volume orthopedic settings.

Despite the strengths and innovations presented in this study, several limitations must be acknowledged. The retrospective design may introduce inherent biases, including selection bias, as data were collected from existing medical records, potentially limiting the generalizability of our findings. Although we used a relatively large dataset, the number of SSI cases remained limited, which could impact the statistical power of the model. More researches with larger sample sizes are required to validate its superiority over the NIHS model. The absence of hemoglobin A1c (HbA1c) data for diabetic patients may limit the assessment of glycemic control’s impact on SSI risk. Prospective multi-center studies with larger and more diverse populations are needed to validate these findings. Further investigation is required to assess the inclusion of variables such as surgical approach, anesthesia type, and intraoperative temperature control, thereby enhancing the model’s generalizability across different patient populations.

Additionally, while hypoproteinemia emerged as a significant independent predictor, its measurement requires timely and accurate preoperative laboratory tests, which may not always be feasible in emergency settings. Moreover, factors like preoperative nutritional interventions and postoperative rehabilitation, which could influence serum albumin levels, were not considered in our model. Due to the retrospective nature of this study, in the future, we will conduct a prospective interventional study to clarify the impact of correcting hypoproteinemia before surgery on the prevention of SSI.

Conclusions

In conclusion, this study underscores the pivotal role of hypoproteinemia as an independent risk factor for SSI following joint surgery and presents a predictive model that might be outperforms the NNIS system. By integrating patient-specific variables, including BMI and serum protein levels, our model offers a more accurate and reliable tool for SSI risk stratification.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the ethics committee of the North China Healthcare Group Xingtai General Hospital (No. ZCKT-2021-0025) and informed consent was taken from all the patients.

Data Sharing Statement

Available at https://aoj.amegroups.com/article/view/10.21037/aoj-25-14/dss

aoj-10-21-dss.pdf ^{(77.4KB, pdf)}

DOI: 10.21037/aoj-25-14