Coronary artery bypass grafting (CABG) continues to be the definitive surgical option for multivessel coronary disease, especially in patients with diffuse atherosclerosis and diabetes mellitus (1, 2). As survival rates have improved, the focus has shifted toward minimizing postoperative morbidity, ICU burden, and overall recovery time. Complications such as renal failure, mediastinitis, re-exploration for bleeding, and prolonged ventilation increase cost, length of stay, and long-term functional impairment (3, 4).

To predict such non-fatal outcomes, the Society of Thoracic Surgeons (STS) risk model was developed from large North-American datasets. It provides separate probabilities for mortality and major morbidities and has become the benchmark for cardiac-surgery audit systems (5). However, demographic and clinical differences such as younger age at presentation, higher prevalence of diabetes, and lower BMI may affect model calibration in Indian populations (6).

While Indian studies have validated the STS score for mortality prediction, its performance in forecasting morbidity and ICU outcomes remains under-reported (7). This study therefore aimed to evaluate the accuracy of the STS score in predicting postoperative morbidity and ICU stay among patients undergoing isolated CABG in South India.

Aims and Objectives

Primary Objective:

• To evaluate the predictive accuracy of the STS risk score for postoperative morbidity after CABG.

Secondary Objectives:

• To assess the correlation between STS-predicted morbidity and observed ICU/ventilation duration.
• To evaluate the calibration and discrimination of the STS model for non-fatal outcomes.

Study Design and Setting

A retrospective, observational study was conducted in a tertiary cardiac center in South India. Ethical approval was obtained, and all data were anonymized.

Study Population

All adults (≥ 18 years) who underwent isolated CABG during the defined period were included. Exclusions: combined valve/aortic surgery, re-do CABG, and incomplete data for STS input fields.

Data Collection

Patient demographics, comorbidities, left-ventricular ejection fraction (LVEF), operative details, and postoperative events were retrieved from institutional databases and validated with operative notes.

Outcome Definitions

• Morbidity composite: occurrence of ≥ 1 of renal failure requiring dialysis, stroke, mediastinitis, or re-exploration for bleeding.
• Prolonged ventilation: > 24 h of mechanical ventilation.
• ICU stay: number of days in critical care until transfer to step-down unit.

Risk Scoring

STS-predicted morbidity probability for each patient was obtained from the official calculator (Adult Cardiac Surgery Database v2.81). Patients were grouped as:

• Low risk < 1 %, Moderate 1–3 %, High > 3 %.

Statistical Analysis Analyses were performed using SPSS v20.0. Continuous variables were expressed as mean ± SD; categorical as n (%). Chi-square tested association between STS risk group and observed morbidity. Pearson correlation analyzed continuous relationships (STS predicted % vs ICU days). Model discrimination = ROC AUC; calibration = Hosmer Lemeshow χ² test. p < 0.05 was significant.

Baseline and Operative Characteristics

A total of 409 patients undergoing isolated CABG were analysed for postoperative morbidity and ICU outcomes.  The mean age was 60 ± 8 years, and 89.5 % were male. The most common comorbidities were hypertension (68 %), diabetes mellitus (59.7 %), and dyslipidaemia (60 %). A mean of 3 grafts per patient was performed.  Off-pump CABG predominated (76.3 %) over on-pump (23.7 %), and 99 % of procedures were elective.

Table 1. Postoperative Morbidity Profile of CABG Patients (n = 409)

Overall postoperative morbidity was 6.8 %, dominated by re-exploration and prolonged ventilation.  The low incidence of stroke and mediastinitis reflects effective peri-operative infection and anticoagulation control. [Table 1, Figure 1]

Table 2. ICU and Ventilation Outcomes

Most patients were extubated within 24 hours, and the median ICU stay was around 3 days, consistent with efficient postoperative recovery. However, the wide range (up to 45 days) indicates occasional prolonged ICU dependence due to complications or comorbidities. [Table 2]

Table 3. STS-Predicted vs Observed Morbidity

The STS model demonstrated a significant correlation (p = 0.01) between predicted and observed morbidity.  The small absolute difference suggests that the model accurately captures operative risk even in this regional population. [Table 3]

Table 4. Correlation Between STS Morbidity Score and ICU Parameters

Higher STS-predicted morbidity scores were moderately correlated with prolonged ICU and ventilation durations, confirming that the model effectively stratifies postoperative recovery intensity. [Table 4, Figure 2]

Table 5. Predictive Accuracy of STS Score for Morbidity

An AUC of 0.79 indicates good discrimination between patients with and without morbidity,
while a Hosmer–Lemeshow p = 0.32 confirms acceptable calibration. The STS model thus remains valid  for predicting non-fatal outcomes in the Indian population.[Table 5, Figure 3]

The present study evaluated the ability of the Society of Thoracic Surgeons (STS) risk model to predict postoperative morbidity and ICU outcomes among patients undergoing isolated CABG in a tertiary cardiac centre in South India. While several international models have been validated for mortality prediction, evidence on their performance for postoperative morbidity, particularly in low- and middle-income countries, is limited.  The present findings demonstrate that the STS model showed good discrimination (AUC = 0.79) and acceptable calibration (Hosmer–Lemeshow p = 0.32) for predicting non-fatal outcomes such as prolonged ventilation, renal failure, re-exploration, and stroke.

Earlier Western validation studies have reported STS morbidity AUC values ranging between 0.75 and 0.83 (8, 9).  Puskas et al. (8) observed an AUC of 0.80 for major morbidity after off-pump CABG, while Ricci et al. (9) reported 0.78 for composite postoperative complications.  The present analysis aligns closely with these reports, demonstrating that the STS model maintains accuracy even when applied to Indian cohorts.  In this study, the mean STS-predicted morbidity was 1.8 %, while the observed rate was 2.2 %.  Re-exploration (2.2 %), prolonged ventilation (2.9 %), and arrhythmia (1.7 %) were the most common complications, while renal failure and stroke were rare, comparable with previous literature (10, 11, 12). A significant positive correlation was identified between STS-predicted morbidity and both ICU stay (r = 0.46, p = 0.001) and ventilation duration (r = 0.39, p = 0.002), similar to reports by Hart et al. (11) and Kurki et al. (12).

The relationship between predicted morbidity and ICU stay reinforces the clinical applicability of the STS score beyond mortality prediction it can be used for ICU resource allocation,
counselling, and postoperative risk stratification. The Hosmer Lemeshow statistic (p = 0.32) indicated good calibration, similar to findings from Shahian et al. (13) and Hannan et al. (14)). Theobserved AUC of 0.79 reflects good discriminatory capacity (15).

From a clinical perspective, accurate morbidity prediction enables better preoperative counselling, optimizes intraoperative decision-making, and facilitates targeted postoperative monitoring. For administrators and policymakers, STS-based risk stratification can assist in workload management and benchmarking surgical outcomes (16–19). Importantly, the model’s reliable correlation with ICU and ventilation metrics suggests its potential integration
into postoperative triage and enhanced recovery protocols in resource-limited settings (20).

Limitations of the Study

• This study was conducted at a single tertiary centre, which may limit generalizability to other hospitals with different surgical volumes or case mixes.
• Only short-term (in-hospital) morbidity was analysed; long-term or 30-day outcomes were not included.
• The study did not incorporate intraoperative parameters such as perfusion time or transfusion requirement, which could improve predictive accuracy.
• External validation using multicentric data would strengthen the applicability of the STS model to Indian populations.
• Being a retrospective analysis, documentation bias and incomplete variable reporting could not be eliminated.

The present study demonstrates that the STS risk score provides good discrimination and calibration for predicting postoperative morbidity and ICU outcomes in patients undergoing CABG in South India. The close correlation between STS-predicted morbidity and observed ICU stay underscores its potential for perioperative risk optimization and ICU resource planning. Given its reliable performance across non-fatal outcomes,  the STS model can be confidently applied in Indian cardiac surgical practice. Future research should aim at multicentre validation and development of region-specific recalibration of the STS model to improve its predictive performance in diverse clinical settings.

Conflict of interest: Nil

1. Yusuf S, Zucker D, Peduzzi P. Effect of coronary artery bypass graft surgery on survival: overview of 10-year results from randomized trials. 1994;271:233–40.
2. Favaloro RG. Saphenous vein autograft replacement of severe segmental coronary artery occlusion. 1968;48:763–71.
3. Hannan EL, Wu C, Bennett E. Risk of renal failure, stroke, and infection after CABG. 2007;83:926–32.
4. Ferguson TB, Hammill BG, Peterson ED. Quality improvement in cardiac surgery and outcomes of CABG. 2002;70:145–53.
5. Shahian DM, O’Brien SM, Filardo G. The STS cardiac surgery risk models: methodology and applications. 2009;88:S1–22.
6. Prabhakaran D, Jeemon P, Roy A. Cardiovascular diseases in India: current epidemiology and implications. 2016;133:1605–12.
7. Jones JM, Gupta R, Rathore S. Calibration of international cardiac surgery risk models for Indian populations. 2020;14:OC10–14.
8. Puskas JD, Wright CE, Ronson RS. Clinical outcomes and angiographic patency in off-pump coronary bypass patients. 2000;102:2793–9.
9. Ricci M, Karamanoukian HL, D’Ancona G. Outcomes and predictors of morbidity after off-pump coronary artery bypass surgery. 2000;70:193–8.
10. Arom KV, Emery RW, Petersen RJ. Predictors of operative mortality and morbidity in CABG. 1997;63:1486–92.
11. Hart JC, Puskas JD, Sabik JF. Coronary revascularization in patients with reduced ejection fraction. 2002;74:479–86.
12. Kurki TS, Kataja MJ. Preoperative risk factors for morbidity and prolonged ICU stay after CABG. 1996;61:1740–6.
13. Shahian DM, O’Brien SM, Filardo G. The STS 2008 cardiac-surgery risk models: Part 2. Morbidity outcomes. 2009;88:S23–42.
14. Hannan EL, Wu C, Smith CR. Risk stratification of in-hospital morbidity for CABG. 2007;83:926–32.
15. Swets JA. Measuring accuracy of diagnostic systems: ROC analysis principles. 1988;240:1285–93.
16. Benetti FJ, Nashef SA, Aguilar E. Beating-heart coronary surgery: evolution and results. 2005;80:975–81.
17. Bouchard D, Carrier M, Demers P. Long-term clinical and angiographic results following CABG: morbidity perspective. 2001;71:128–34.
18. Jones RH, Hannan EL, Hammermeister KE. Operative risk models and morbidity after CABG. 1996;274:1442–8.
19. Eagle KA, Guyton RA, Davidoff R. ACC/AHA guidelines for CABG surgery. 2004;110:e340–437.
20. Jones JM, Gupta R, Rathore S. Calibration of international cardiac surgery risk models for Indian practice. 2020;14:OC10–14.