The study was conducted according to the principles of the declaration of Helsinki, approved by the local ethical committee (EKNZ BASEC Req-2020-02023) and registered at ClinicalTrials.gov (ID NCT05225246). The ethical committee has waived the need to obtain informed consent.

We constructed the single centre retrospective observational DREAMS (Database for REAdmission after Major cardiac Surgery) registry including all patients after major cardiac surgery from January 2012 to September 2020. Patients with minor operations such as pacemaker implantation, pericardiocentesis, or in-hospital death were excluded from this analysis. Additionally, we excluded patients with aortic dissection or left ventricular assist device since these patients represent a highly special cardiac surgical patient cohort with a high-risk for readmission. Furthermore, re-operations/revisions within the same admission were not registered twice.

Using a prospectively maintained institutional registry (Intellect 1.7, Dendrite Clinical Systems, Henley-on-Thames, United Kingdom), we identified all patients who underwent major cardiac surgery in our institution. The clinical data were exported from this registry where data have been regularly controlled for completeness and accuracy. Intraoperative data were collected prospectively in a standardized fashion. For each admission with major cardiac surgery, further referred to as index admission, we extracted data on stationary and ambulatory readmissions within 30 days after discharge from the electronic health record using the hospitals datawarehousing solution.

All readmissions to our tertiary centre within 30 days after discharge from index admission were analyzed. To maintain consistency, reasons for readmission were adjudicated by the same research fellow (J.M.) according to all available medical records obtained during clinical care including history, physical examination, results of laboratory testing, radiologic testing, electrocardiogram and echocardiography. In case of uncertainty, patient records were reviewed by two other members of the research group (L.K., O.R.).

Based on previous literature and medical considerations, causes for readmission were categorized in cardiac related, infections, bleeding, neurological, gastrointestinal disease excluding bleeding (such as gastritis, cholecystitis or diverticulitis), pain excluding angina, noninfectious wound complications, musculoskeletal disorders, urological and other (7, 10). Detailed information about the categories and the definitions is provided in Online Supplementary Table S1.

To investigate whether patient sex is associated with the risk of re-admission within 30 days, we calculated crude and adjusted incidence rate ratios (IRR) with 95% confidence intervals using negative binomial regression. For the adjustment, we included age, COPD, EuroSCORE 2, emergency, type of surgery (coronary artery bypass grafting, aortic valve surgery, mitral valve surgery, aortic surgery) and time of extracorporeal circulation (min) as covariates. We first included one of these covariates into the model together with sex, then we included an interaction term of the covariate and sex to check for different associations in women and men. Lastly, we included all variables. To give an overview of the association of sex and risk of re-admission, we showed all IRRs in a forest plot. We visualized time to first re-admission using Kaplan-Meier plots. Moreover, we conducted four subgroup analysis. First, we focussed on type of surgery, e.g. isolated coronary artery bypass grafting, isolated aortic valve surgery, and isolated mitral valve surgery, then we derived risk of readmission by type of cardiac rehabilitation (in-patient vs. out-patient) as a sensitivity analysis.

Continuous variables were shown as median with interquartile range and compared using Wilcoxon-Mann-Whitney test. Categorical variables were shown as number with percentage and tested using Fisher's exact test. All analyses were carried out by a biostatistician (B.G.) using Stata 16 (Stata Corp., College Station, Texas).

A total of 5,512 standard cardiac operations were carried out in our centre between January 2012 and September 2020. After exclusion of surgery for aortic dissection (n = 295), left ventricular assist device implantation (n = 16), minor procedures (n = 116), consecutive operations within the same admission (n = 42) and in hospital death (n = 175), the final study cohort comprised a total of 4,868 patients, 1,149 patients (24%) were female. Median [Interquartile range (IQR)] age was 68 (60 to 74) years, while female patients were significantly older compared to male patients (70 [63 to 76] vs. 67 [59 to 74] years; p < 0.001). Isolated valve surgery, e.g., single, double or triple valve procedure without further concomitant procedures, was more frequent in female [41% (n = 474) vs 22% (n = 815)] while coronary artery bypass grafting was more often performed in male (27% [n = 305] vs. 46% [n = 1,700]). Baseline characteristics and procedure related information are depicted in Tables 1, 2 and postoperative data in the Online Supplementary Table S2.

In the entire study cohort, frequency of readmissions within 30 days after discharge was 8.3% (n = 403), 7.0% (n = 81) in women compared to 8.7% (n = 322) for men. Six women (0.52%) and 31 men (0.83%) were re-admitted twice, one woman (0.09%) and three men (0.08%) were re-admitted three times, hence 444 readmissions in total, 356 in men and 88 in women. This corresponds to a readmission rate of 9.2 (8.4 to 10.1) per 100 patients within 30 days and an IRR of 0.8 (95% CI 0.63 to 1.03, p = 0.078) of female gender. Figure 1 displays the cumulative incidence curves for 30-day readmission.

In male patients, readmitted patients significantly more often were current smokers and had renal disease. In female patients, only preoperative NYHA class III or IV was significantly different between the groups with higher values in patients which were readmitted. A more detailed depiction of readmitted patients for both cohorts is provided in Online Supplementary Table S3.

The main reasons for readmission within 30 days were cardiac related and infectious (female: 36% [n = 32] and 31% [n = 28], male: 28% [n = 99] and 23% [n = 84]; respectively, p = 0.12 and p = 0.13). While bleeding complications were seen twice as often in men than in female (18% [n = 64] vs. 7.9% [n = 7], p = 0.023), non-infectious wound complications were 15% (n = 13) in female and 5.6% (n = 20) in men (p = 0.006). Detailed information about the frequency of specific reasons for readmission are provided in Figure 2 and Table 3.

We checked aforementioned variables for association with the risk of readmission. We found a slightly increased readmission rate with EuroSCORE II increase (IRR 1.01, CI 1.00 to 1.03, p = 0.012), mitral valve surgery (IRR 1.3, CI 1.03 to 1.63, p = 0.025) and ECC time increase (IRR 1.04 per 10 min, CI 1.02 to 1.06, p < 0.001) and a slightly decreased rate with age increase (IRR 0.92, CI 0.84 to 1.00, p = 0.047). We did not see an association with COPD, CABG, aortic valve surgery, and aortic surgery. Moreover, patients visiting an inpatient cardiac rehabilitation after the operation had a reduced risk for readmission (IRR 0.53, CI 0.39 to 0.71, p < 0.001). An overview is provided in Table 4 and Figure 3, as well as in Online Supplementary Table S4. Female sex did not show a significant interaction with any predictor, so we did not find evidence that associations differ between women and men.

Concerning the association of complications and readmissions, all complications were associated with increased incidence of readmission. Sternal infection (IRR 5.35), sepsis (IRR 2.83) and reoperation for bleeding (IRR 1.88) showed the strongest association. However, we did not observe substantial changes of the IRR of sex. (Online Supplementary Table S5).

In 2005 patients with isolated CABG surgery (85% male [n = 1,700], 15% female [n = 305]) readmission rate was 8.0 (CI 6.9 to 9.4) per 100 patient years, and therefore slightly lower than in the entire cohort. Again, readmission frequency of male and female patients were similar (7.2% men, 122 patients with 139 readmissions vs. 6.6% women, 20 patients with 22 readmissions, p = 0.81). Most common reason for readmission in male patients was cardiac related [n = 44 (32%)] while in female patients, infection was the most frequent reason [n = 12 (55%); Online Supplementary Table S6, S7].

In 695 patients with isolated aortic valve surgery (65% male [n = 452], 35% female [n = 243]) readmission rate was 8.6 (6.7 to 11.1) per 100 patient years, concordant with the entire cohort. As in the entire cohort, readmission frequency of readmission for male and female patient was similar (9.3% [n = 42] male and 6.6% [n = 16] female, rate ratio 0.74 (95% CI 0.4 to 1.3) and p = 0.28). In this subgroup the main reason for readmission were cardiac related (26% [n = 11] male and 25% [n = 4] female, p = 1.00) and infections (26% [n = 11] male and 25% [n = 4] female, p = 1.00). Online Supplementary Tables S8, S9.

In 464 patients with isolated mitral valve surgery (61% male [n = 283], 39% female [n = 181]) overall readmission rate was 0.74 (0.4 to 1.3) per 100 patient years. Male patients were readmitted more often than female patients (11.3% [n = 32] vs. 4.46% [n = 8], with a rate ratio 0.74 (95% CI 0.4 to 1.3), p = 0.014). The main cause for readmission in the male cohort were infection [28% (n = 9)], followed by cardiac related [22% (n = 7)]. In the female cohort, the main cause for readmission were cardiac related [38% (n = 3)] followed by equal rate of readmission for infection, bleeding, gastrointestinal disease, muskuloskelettal disorder and others [each 13% (n = 1)]. Online Supplementary Tables S10, S11.

In patients with available data [n = 1,646, (34%)] concerning the type of postoperative cardiac rehabilitation (inpatient vs. outpatient), a subgroup analysis was performed (Online Supplementary Table S12). Patients undergoing inpatient rehabilitation were older (69 (61 to 75) years vs. 63 (56 to 70) years; p < 0.001), had higher preoperative EuroSCORE II (2.0 (1.1 to 4.2) % vs. 1.3 (0.79 to 2.4) %; p < 0.001), as well as a higher number of comorbidities. The subgroup analysis showed that inpatient rehabilitation was associated with a lower risk of readmission [IRR 0.53 (0.39 to 0.71), p < 0.001]. This reduction remained unchanged when patient sex was included into the model as covariate.