Covariates associated with pulmonary complications including unplanned intubation after total knee arthroplasty

Abstract

Background

Total knee arthroplasty (TKA) is widely regarded as a safe procedure. Yet pulmonary complications do occur sometimes, requiring unplanned intubation, creating difficulties with ventilator weaning, or resulting in pneumonia. Our goal was to identify which risk factors are most strongly associated with these respiratory complications after TKA using the American College of Surgeons National Surgical Quality Improvement Program database (NSQIP) database.

Methods

The NSQIP database is a comprehensive de-identified registry of surgical outcomes. Our analytical approach proceeded in two phases. Initially, we conducted univariate analyses to screen individual factors for associations with our three primary outcomes: unplanned intubation, failure to wean from mechanical ventilation, and the development of pneumonia.

Factors showing statistical significance in this preliminary screening then advanced to multivariate logistic regression modelling. Variables included patient demographics, such as age and sex, relevant comorbidities (such as diabetes, chronic obstructive pulmonary disease (COPD), and congestive heart failure (CHF), functional status, key laboratory parameters (such as white blood cell (WBC) and platelet counts), American Society of Anesthesiologists (ASA) classification, and anesthesia approach. Statistical significance was defined as p < 0.05.

Results

The strongest predictors across the three outcomes of unplanned intubation, failure to wean, and pneumonia, in descending order, were age, ASA classification, history of COPD, and history of CHF.

Conclusion

Our study found that the risk of pulmonary complications within 30 days following total knee arthroplasty was independently associated with older age, higher ASA class, and history COPD or CHF, increased operative time, and elevated preoperative WBC count.

1. Introduction

Total knee arthroplasty (TKA), also commonly referred to as total knee replacement, is one of the most frequently performed orthopedic procedures worldwide and is widely recognized as a highly effective, affordable, and reliably successful intervention for end-stage degenerative joint disease of the knee.¹ The surgery is primarily indicated for patients with advanced osteoarthritis, but it may also be performed for rheumatoid arthritis, post-traumatic arthritis, or other conditions causing severe pain and functional impairment. By replacing the damaged joint surfaces with prosthetic components designed to mimic the biomechanics of the native knee, TKA reliably provides significant pain relief, improves joint stability and mobility, and enhances overall quality of life for millions of patients each year.² The success of TKA, coupled with an aging population and rising prevalence of knee osteoarthritis, has led to an exponential increase in its utilization, with projections suggesting continued growth in surgical volume over the coming decades.

Despite its success, TKA is not without risk. While the procedure is generally considered safe, a small but clinically meaningful proportion of patients experience postoperative complications. Among these, pulmonary complications are particularly concerning. These include events such as unplanned intubation, difficulty with ventilator weaning, and postoperative pneumonia, all of which can contribute to longer hospital stays, higher healthcare costs, and increased morbidity and mortality 3, 4, 5. Although the absolute incidence of these complications following TKA is relatively low compared to other major surgical procedures, their clinical consequences are significant and warrant careful attention.

Pulmonary complications after orthopedic surgery are influenced by a complex interplay of patient-, surgery-, and anesthesia-related factors. Patient-related risk factors include advanced age, obesity, smoking history, obstructive sleep apnea, chronic obstructive pulmonary disease, and other cardiopulmonary comorbidities. Surgical risk factors include the duration of anesthesia, intraoperative blood loss, and perioperative transfusion requirements. Moreover, in the context of TKA, the increasing number of elderly patients with multiple comorbidities undergoing the procedure places this population at heightened vulnerability.⁶ Even seemingly minor pulmonary complications may delay rehabilitation and recovery, undermining the functional benefits of the surgery. Recognizing which patients have a higher probability of experiencing these problems is essential for improving care.

The burden of postoperative pulmonary complications also extends beyond individual patient outcomes. These complications are associated with substantial economic costs to healthcare systems. Prolonged hospitalizations, need for intensive care, and readmissions contribute to increased resource utilization. Furthermore, pulmonary events can delay mobilization and rehabilitation, limiting early participation in physical therapy—an essential component of recovery after TKA. Consequently, preventing, identifying, and managing pulmonary complications represents a key target for improving the overall safety and cost-effectiveness of TKA.

Recognition of patients at highest risk for pulmonary complications is therefore an essential aspect of perioperative planning. Preoperative optimization strategies, including smoking cessation, pulmonary rehabilitation, and management of underlying cardiopulmonary disease, may reduce risk. Intraoperative anesthetic approaches, multimodal pain regimens designed to minimize opioid use, and postoperative enhanced recovery after surgery (ERAS) protocols may further decrease pulmonary morbidity. However, accurate risk stratification requires robust data from large, representative patient populations.

To address this need, large-scale clinical databases provide an invaluable resource. The American College of Surgeons National Surgical Quality Improvement Program (NSQIP) is one such tool. NSQIP collects standardized, prospectively gathered data on patient demographics, comorbidities, intraoperative variables, and 30-day postoperative outcomes across a wide range of surgical specialties.⁷ Its inclusion of granular information on complications such as unplanned intubation, pneumonia, and ventilator dependence makes it ideally suited for studying pulmonary outcomes following TKA. Analyses using NSQIP data allow for the identification of risk factors at a population level and the development of predictive models that may guide clinical decision-making and quality improvement initiatives.

2. Methods

We conducted a retrospective analysis of NSQIP data from January 1, 2018, to December 31, 2022. Inclusion criteria were patients who underwent primary TKA as identified by CPT code 27447⁸ .⁹ The data included both unilateral and bilateral TKA.

Compared with unilateral total knee arthroplasty, bilateral procedures impose a greater physiologic burden through prolonged anesthesia duration, increased embolic load, higher transfusion requirements, and amplified inflammatory response. These factors may independently contribute to postoperative pulmonary complications, including pneumonia and respiratory failure, and warrant consideration when assessing pulmonary risk profiles following TKA.

NSQIP focuses on high-quality data collection through a standardized process involving trained and certified Surgical Clinical Reviewers (SCRs) at participating hospitals. This rigorous process, which includes inter-rater reliability (IRR) audits, is designed to minimize the amount of missing data from the outset.

Any missing data were handled using a complete-case approach. Variables with missing values were not imputed, and patients were included in each multivariable model only if data for all covariates in that specific model were available. No multiple imputation or other statistical imputation techniques were performed as part of this analysis. The primary variables of interest (demographics, comorbidities, operative time, anesthesia type, and pulmonary outcomes) are routinely captured by NSQIP with high completeness, minimizing the overall impact of missingness on the analytic cohort.

Three primary pulmonary outcomes were assessed: unplanned intubation, failure to wean, and pneumonia, as coded by NSQIP.⁹ Failure to wean was defined as the inability to extubate the patient within 48 h postoperatively. Our variables were age, sex, race/ethnicity; comorbid conditions such as diabetes, hypertension, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), smoking status, and body mass index (BMI); preoperative white blood cell count (WBC) and platelet count; American Society of anesthesiology (ASA) class; anesthesia type; operative time; and preoperative functional independence (FNSTAUS simple), categorized as independent or dependent in activities of daily living. All data was de-identified⁸ ,¹⁰ and the study received a non-human subject determination from the Division of Research at Brown University.

Univariate analysis determined variables significantly associated with each outcome (p < 0.05), which were then included in multivariate logistic regression models. We evaluated logworth, false discovery rate (FDR)-corrected p-values, parameter estimates, odds ratios (ORs) with 95% CIs, and model performance using the area under the receiver operating characteristic curve (AUC)¹¹ .¹² In performing our regression analyses, we made the following assumptions: 1) the observations in the dataset are independent; this could be challenging in NSQIP cohorts, given that patients are clustered in participating hospitals. While the exact number of hospitals reporting TKA outcomes to NSQIP are not specified for TKA, there were over 308,000 patients in the 5-year cohort, which makes the assumption of independence reasonable; 2) A linear relationship exists between continuous predictors (such as age, BMI) and the log-odds of the outcome (such as unplanned intubation, failure to wean, and pneumonia); 3) Predictor variables should not be highly correlated with each other to ensure stable and interpretable coefficients (in other words, there should be absence of multicollinearity).

3. Results

3.1. Unplanned intubation

Unplanned intubation occurred in 295 patients, and was significantly associated with increasing age (p < 0.0001), ASA class (3 vs 2: OR 2.23; 4 vs 3: OR 2.33; both p < 0.0001), history of COPD (OR 3.11, 95% CI 2.18–4.43, p < 0.0001), and CHF (OR 2.76, 95% CI 1.72–4.49, p < 0.0001). Anesthesia type also played a role, with general anesthesia associated with increased risk compared to spinal/epidural (OR 1.81, 95% CI 1.37–2.40, <0.0001); regional/MAC did not show a significant effect. Elevated preoperative WBC (p = 0.0248) were additional independent predictors. The AUC for the model was 0.73376 and the R² was 0.0507 (Fig. 1).

Fig. 1.

Adjusted odds ratio for variables in the model. All were statistically significant except BMI, MAC anesthesia, preoperative steroid use, smoking and functional status, ASA = American Society of Anesthesiologists Classification; COPD = history of chronic obstructive pulmonary disease; CHF = history of congestive heart failure; GA = General anesthesia; WBC = preoperative white blood cell count; BMI = body mass index; MAC = monitored anesthesia care.

3.2. Failure to wean

Failure to wean was defined as inability to wean from the ventilator within 48 h after surgery. This complication occurred in 121 patients. Failure to wean was significantly associated with older age (p < 0.0001), ASA Class 3 vs 2 (p = 0.0002), COPD (p = 0.0013), CHF (p = 0.0059), and elevated preoperative WBC (estimate 0.0899, p < 0.0001). General anesthesia vs. spinal or epidural anesthesia (p = 0.1944, OR 1.32), general vs. monitored anesthesia care (MAC) (p = 0.4784), increased operative time, (p = 0.0863), and elevated preoperative WBC. The odds ratios are summarized in Fig. 2. The AUC was 0.73450, and the model R² was 0.0443.

Fig. 2.

Odds ratios for variables in the multivariate model for ventilator dependence/failure to wean.

3.3. Postoperative pneumonia

Postoperative pneumonia occurred in 714 patients. Factors significantly associated with having pneumonia included: ASA Class 3 vs 2, or 4 vs 3, history of COPD or CHF, elevated preoperative WBC, smoking, functional status, (all p < 0.0001), General anesthesia vs. spinal/epidural (p = 0.0004), increased operative time (p = 0.0002), and preoperative steroid use (p = 0.0214). The odds ratios are summarized in Fig. 3. The AUC was 0.74117, and the model R² was 0.0638.

Fig. 3.

Odds ratios for variables in the multivariate model for post-operative pneumonia.

3.4. Summary of outcomes

Table 1 summarizes all p-values, ORs, and 95% CI across the outcomes of unplanned intubation, failure to wean, and postoperative pneumonia.

Table 1.

Summary of all variables and their associated statistics with regard to each of the three outcomes.

Table 1: Summary of Multivariate models	Unplanned intubation (AUC = 0.73376, R2 = 0.0507)		Failure to wean (AUC = 0.73450, R2 = 0.0443)		Pneumonia (AUC = 0.74117, R2 = 0.0630)
	p-value	Odds ratio (95% CI)	p-value	Odds ratio (95% CI)	p-value	Odds ratio (95% CI)
ASA Class 3 vs. 2	p < 0.0001	2.23 (1.65–3.01)	p = 0.0002	2.23 (1.65–3.01)	p < 0.0001	1.81 (1.50–2.20)
ASA Class 4 vs. 3	P = 0.0002	2.33 (1.39–3.64)	NS (0.0827)	1.93 (0.92–4.07)	p < 0.0001	2.70 (2.05–3.56)
History of COPD	p < 0.0001	3.11 (2.18–4.43)	p = 0.0013	2.55 (1.434–4.50)	p < 0.0001	2.96 (2.36–3.71)
History of CHF	p < 0.0001	2.76 (1.72–4.49)	p = 0.0059	2.84 (1.35–5.99)	p < 0.0001	2.45 (1.77–3.39)
General anesthesia vs. spinal/epidural	p < 0.0001	1.81 (1.37–2.40)	P = 0.1944	1.32 (0.86–2.01)	p = 0.0004	1.39 (1.16–1.67)
General anesthesia vs. regional or MAC	NS (0.0719)	1.31 (0.98–1.76)	P = 0.4784	1.19 (0.74–1.90)	NS (0.6092)	1.05 (0.87–1.27)
Increased operative time	p = 0.0233	1.00 (1.00–1.00)	P = 0.0863	1.00 (0.99–1.00)	p = 0.0002	1.00 (1.00–1.00)
Elevated preoperative WBC	p = 0.0248	1.05 (1.00–1.09)	p = 0.0090	1.08 (1.02–1.13)	p < 0.0001	1.09 (1.07–1.11)
Preoperative steroid use	NS (0.3748)	1.26 (0.77–2.07)	NS (0.5665)	1.26 (0.58–2.72)	p = 0.0214	1.44 (1.06–1.96)
Age	p < 0.0001	1.04 (1.02–1.06)	p < 0.0001	1.04 (1.02–1.06)	p < 0.0001	1.05 (1.04–1.06)
Smoking	NS (0.2694)	0.75 (0.45–1.25)	NS (0.4328)	1.30 (0.68–2.50)	p < 0.0001	1.81 (1.41–2.32)
Functional status	NS (0.1726)	1.53 (0.83–2.82)	NS (0.5685)	1.34 (0.49–3.67)	p < 0.0001	2.34 (1.67–3.25)
Body Mass Index (BMI)	NS (0.5614)	1.00 (0.99–1.02)	NS (0.09221)	1.00 (0.97–1.03)	NS (0.3803)	1.00 (0.99–1.01)

Fig. 4 is a heat map depicting the contribution of each of the variables to the three outcomes of unplanned intubation, failure to wean from the ventilator within 48 h, and postoperative pneumonia.

Fig. 4.

Heatmap depicting the log odds ratio for each of the predictor variables against each of the three pulmonary outcomes.

4. Discussion

The current study highlights three potential pulmonary complications after total knee arthroplasty (TKA): unplanned intubation, failure to wean from the ventilator within 48 h of surgery, and postoperative pneumonia. The predictors that were common across all three pulmonary outcomes were older age, ASA Class 3 vs. 2, history of COPD or CHF, general vs. neuraxial anesthesia, increased operative time and elevated preoperative WBC.

Older age as a risk factor is consistent with previous research and reflects the natural decline in pulmonary reserve and immune function with aging.¹² Older adults tend to face a trifecta of challenges: undiagnosed lung problems, diminished muscle strength, and sluggish recovery from anesthesia, all of which can complicate the post-surgical period^{13 14}.¹⁵

Chronic obstructive pulmonary disease (COPD) and congestive heart failure (CHF) both demonstrated a clear link to increased rates of unexpected intubation and difficulty discontinuing mechanical ventilation. These findings align with prior research linking cardiopulmonary conditions to an increased risk of postoperative respiratory failure.¹⁶ In particular, COPD increased the odds of failure to wean by over 2.5 times, and CHF by nearly 3 times.

Anesthesia type also emerged as a significant factor. Interestingly, patients who underwent general anesthesia faced higher odds of unplanned intubation compared to those who received spinal or epidural anesthesia. This pattern reinforces prior work suggesting that neuraxial techniques may offer protective respiratory effects, especially in orthopedic surgical settings.¹⁷ We did not explicitly differentiate between failure to wean from a planned vs. an unplanned intubation. This point is worth mentioning, as an unplanned intubation may reflect something about those patients which imposed a greater risk to begin with, if they needed the unplanned intubation.

Another key finding was the role of operative time, which consistently appeared as a modest but significant predictor across all models. Longer surgeries reflect more complex cases or operative challenges and can increase exposure to anesthesia and blood loss, both of which contribute to impaired respiratory function postoperatively.¹⁸

We also observed that elevated preoperative white blood cell (WBC) count, an indicator of systemic inflammation or infection, was significantly associated with both intubation and failure to wean. Elevated WBC reflects an ongoing inflammatory process or increased physiologic stress, both of which could impair postoperative recovery.¹⁹ This suggests that patients with abnormal laboratory values before surgery may benefit from further evaluation or postponement of surgery until their values are optimized.20, 21

The American Society of Anesthesiologists (ASA) physical status classification proved remarkably robust across our entire analysis. Higher ASA classes consistently predicted worse outcomes. This isn't surprising as these classifications capture the cumulative burden of a patient's comorbidities. While ASA scoring involves some subjectivity, its predictive power remained consistent across every model we tested. This suggests that it continues to serve as a valuable shorthand for assessing surgical risk.

The logistic regression analyses showed moderate predictive capability. AUCs ranged from 0.73 to 0.74, and R² values for each of the models ranged from 0.04 to 0.06. This is hardly perfect, but these results align well with other NSQIP-based clinical tools already in widespread use. More importantly, they provide some guidance for meaningful preoperative risk stratification and surgical decision-making.

4.1. Limitations

This study has several limitations. First, it is based on retrospective data from NSQIP, which does not include information such as anesthesia management or exact operative time.¹⁰ Additionally, variables such as ASA classification and functional status are subjective and vary among clinicians. The dataset only includes U.S. hospitals that participate in the NSQIP, which may limit the applicability of the findings to international patient populations. Finally, the analysis is limited to cases from 2018 to 2022 and does not reflect trends before or after this period⁹ .¹¹

Second, as with all large registry-based studies, the accuracy of our findings depends on the fidelity of NSQIP data abstraction and reporting. Although NSQIP employs trained Surgical Clinical Reviewers and standardized definitions, rare outcomes—particularly those occurring in fewer than 0.1% of cases—remain susceptible to documentation error or undercapture. This limitation is especially relevant for infrequent pulmonary complications such as prolonged ventilator dependence or unplanned intubation, where even minor misclassification may influence effect estimates. Accordingly, associations involving very low-frequency events should be interpreted with appropriate caution.

Despite these limitations, the consistency of effect estimates across three distinct pulmonary outcomes supports the robustness of the identified risk factors.

5. Conclusion

This large-cohort NSQIP study identifies key predictors: age, COPD, CHF, BMI, platelet count, operative time, ASA class, anesthesia type, and WBC, for major pulmonary complications after TKA. Implementing comprehensive preoperative risk assessment and targeted interventions may reduce complication rates and improve outcomes in this population.

Ethical statement

The authors’ institutional IRB has determined that this NSQIP analysis is non-human subject research and is thus exempt.

The United States Department of Health and Human Services (HHS) determines that Quality Improvement activities, such as the ACS NSQIP do not satisfy the definition of “research” under (45 CFR part 46) if their purposes are limited to: (a) delivering healthcare, and (b) measuring and reporting provider performance data for clinical, practical or administrative uses.

Guardian or patient consent

Guardian/patient consent is not applicable to this manuscript as it is a bibliometric review and not a case report. There is no identifiable data in this review.

Authors contributions

The study was designed by TS and LG. TS and LG managed data capture. TS and LG supervised the conduct of the research. PG performed the literature search and drafted the initial manuscript. LG reviewed and edited the initial manuscript, and all authors approved the final version of the manuscript.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Declaration of competing interest

The authors declare that there are no competing or conflicts of interest.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.