Heart failure (HF) is a life-threatening condition and is associated with significant morbidity, poor functional capacity, and decreased quality of life [1]. In 2017, more than 64 million people worldwide were affected with HF [2], and the number is likely to rise. Savarese G et al. [3] reported in their survey that annual health care costs per HF patient amount up to €25,000 in the Western world, resulting in a substantial economic burden. The prevalence of valvular heart disease (VHD) among the elderly, as well as VHD-related HF, is rising as the population is ageing [4]. VHD is one of the most common types of cardiac surgery. According to a multicenter study conducted in China [5], the overall mortality rate for VHD surgery was around 2%. However, it is much higher in patients with HF, and can be greater than 3%. Nearly all risk prediction models for cardiac surgery include HF as an independent predictor. There is a growing demand for risk assessment for these surgical patients, however current risk scores do not provide a reliable estimate of the exact operative mortality in an individual HF patient [6, 7].

In order to better assess the risk of surgery for these patients, the aim of the present study is to establish a simplified scoring risk model based on the Chinese Cardiac Surgery Registry (CCSR) database to accurately predict the postoperative mortality of suspected HF patients undergoing VHD surgery.

This risk score is an effective and simple tool for mortality prediction after valvular surgery in patients with HF. Unlike the EuroSCORE II, which has 18 predictors, this model includes only 9 predictors which are easily accessible in clinical practice. The model is convenient for clinical use and could be a reliable bedside tool.

HF is a major health-care issue that is related to high resource usage and health-care costs [3]. HF is also the leading cause of hospitalization in people over the age of 65 [12]. VHD is amongst the most common primary causes of HF, and many VHD patients require surgery. HF has long been a focus of clinical perioperative evaluation as an independent risk factor for cardiac surgery. The definitive diagnosis of HF, on the other hand, is challenging, especially for HF with preserved LVEF, which necessitates a combination of clinical symptoms and signs, as well as a variety of objective laboratory and ultrasound indicators. Many patients cannot receive a precise diagnosis of HF prior to surgery due to the wide discrepancies in preoperative examination of VHD patients among different cardiac institutes in China. The target population of this study was therefore identified as suspected HF, which could be quickly diagnosed based on symptoms, signs, and valvular abnormalities, thereby enhancing the clinical application of this model. Previous prediction models may no longer be able to reliably estimate current surgical risk due to improvements in surgical techniques and perioperative treatment. Prediction models are time-sensitive: an excellent prediction model must be continuously updated. For instance, consider EuroSCORE II was released in 2012, and has nearly fully replaced EuroSCORE I, which was first published in 1999. Therefore, in our study, we developed a prediction model based on the most recent clinical data that could objectively reflect current VHD features and surgical outcomes.

Furthermore, the prediction model is region-specific, because people in different regions of the world have distinct disease features [13], and there is regional variation in therapeutic concepts and techniques [6, 14]. Currently, the most of widely used clinical prediction models (such as EuroSCORE II and the society of thoracic surgeons (STS) score) were based on western populations. These western models may not be ideal for Asia or the Chinese population. The EuroSCORE II has underperformed in the Chinese suspected HF population, according to our findings. Our subgroup analysis (Fig. 5) indicates that in terms of discrimination, our model was significantly better than EuroSCORE II among NYHA III or IV patients. As a result, developing a prediction model for Chinese suspected HF patients who require VHD surgery is important in clinical practice. Wessler et al. [6] published a study showing that many VHD prediction models performed poorly in validation. They suggested that one probable explanation is a lack of sample size or a poor representative of the sample population for model derivation. Fortunately, one of the advantages of our study is that the sample population is well-representative using the CCSR data. The CCSR is the largest Chinese multicenter cardiac surgery database, analogous to the STS in North America, and includes almost all high-quality cardiac hospitals in China. As a result, this risk model’s validation performance was satisfactory, and considerably better than EuroSCORE II. However, pending testing and practice in real world, this score’s clinical significance will have to be determined for other populations.

Preoperative renal function indicators (eGFR and prior dialysis) and cardiac function indicators (NYHA IV and LVEF 35%) account for the large proportion of predictors in this model. The renal function indicators have the highest weights, implying that renal function has a significant impact on the prognosis of HF patients. Metra M et al. [15] suggested that worse renal function might result in poorer clinical outcomes in HF patients. And a meta-analysis [16] showed inadequate renal function was found in 23% of the HF patients, and was associated with a two-fold increased risk of all-cause death, with greater magnitude of the association whether LVEF was higher. In those who suffer from HF, preoperative CKD is a predictor of poor outcomes in all cardiac surgical patients and nearly all cardiac surgery prediction models include preoperative renal function as a predictor. In contrast to coronary artery bypass grafting (CABG) surgery, there are many different valvular surgical techniques. The surgical method, however, is not an independent risk factor for postoperative mortality, as shown in this study. It indicates that, as valvular surgical techniques have advanced in recent years, the effects of disparities between surgical methods on prognosis are decreasing, highlighting the importance of the patient’s underlying medical conditions in determining prognosis.

The predictive probability of many existing VHD surgical risk models is not good [7, 17]. One of the reasons might be the diversity of VHD surgical methods and the relatively small sample sizes for model derivation [18]. Some investigations had proposed that, in addition to traditional factors, risk models should include more predictors to increase their effectiveness [7]. Given the vast number of intraoperative uncertainties in VHD surgery, we added certain essential intraoperative predictors to the model, in addition to some fundamental preoperative variables, to improve the model’s prediction capability. CPB time, for example, was chosen as a predictor in this study because CPB time provides a thorough reflection of surgical complexity and surgeon proficiency. The longer the CPB time, the more complex the surgery and/or the less skilled the surgeon.

Our risk score model’s primary goal is to offer patients and health care practitioners more accurate information about the risk of VHD surgery and to aid in decision-making. This simplified score model is simple to use, as it is based on nine predictors that are routinely accessed in VHD patients. When considering VHD surgery, it aids in stratifying the risk of mortality.

There is still a gap in nationwide representativeness between CCSR and STS. CCSR includes only data from high-quality cardiac centers in China, hospitals with lower operation volumes are not included. A definitive diagnosis of HF requires objective laboratory and ultrasound indicators. Unfortunately, there are many missing data of these indicators in the current database. Although this model can be used for preoperative evaluation, it is not a complete preoperative evaluation model due to the inclusion of intraoperative predictors. This model also needs external validation in real world practice to evaluate its clinical applicability. In addition, although all patients had HF at admission, after preoperative medical treatment, some of them had improved cardiac function by the time of surgery, and this updated information might not be collected in time. As a result, the data in CCSR might not actually reflect the latest status of every patient before surgery, and is one of the major limitations of this study.

The new risk score is an effective and concise tool that could accurately predict postoperative mortality rates in suspected HF patients after valve surgery.

Not applicable.

The CCSR registry is not publicly available, but the datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. The permission to access and use the raw data was granted by Prof. Shengshou Hu and from the Fuwai hospital.

HYL and JFH designed the research study. HYL, JMG and KA performed the research. HYL analyzed the data and wrote the manuscript. YJW and ZZ supervised the research. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.

This study had been approved by institutional review board of Fuwai hospital, Peking union medical college and Chinese academy of medical sciences (2021-1477). All methods were performed in accordance with the relevant guidelines and regulations. The written informed consents were obtained from all participants.

This study was conducted on behalf of the CCSR. We thank the colleagues from CCSR coordinating board for their excellent work on data management and all the members from the 94 participated hospitals. We thank the colleagues at Fuwai Hospital and the NCCD for their assistance with data audit and entry.

The current study and Drs Hou, Zheng, and Lin were supported by National Key R&D Program of China (2020YFC2008100).

The authors declare no conflict of interest.