Development and validation of a nomogram to predict postoperative delirium in older patients after major abdominal surgery: a retrospective case-control study

Global aging is an increasing concern, and the percentage of older patients is expected to double over a 35-year period from 8.5% in 2015 (Janssen et al. 2019a, b). More than an estimated 12 million older patients need surgical procedures every year, and they are more susceptible than younger patients to postoperative complications (Karalapillai et al. 2020). Delirium, the most common postoperative complication in adults over the age of 65, is expected to increase simultaneously (Marcantonio 2017, Shi et al. 2019).

Postoperative delirium (POD) is an acute organic brain syndrome characterized by a transient disturbed attention span or altered levels of consciousness that can prolong the length of hospital stays and increase medical expenses (Abelha et al. 2013). Furthermore, patients with POD have been shown to have poorer long-term outcomes, including a reduction in cognition and quality of life, increased readmissions, and shorter survival (Abelha, Luís 2013, Aitken et al. 2017, Brown et al. 2016, Gleason et al. 2015). Therefore, it is necessary to identify risk factors for delirium and establish an effective risk-predicting model to prevent POD. A retrospective cohort study conducted by Ying et al. identified carbamazepine, hemoglobin, and blood urea nitrogen (BUN) as major risk factors for delirium (Wang et al. 2020). In addition, Bowman et al. identified 55 risk factors of delirium, including cognitive impairment or mental illness, psychoactive drugs, frailty, and infection (Bowman et al. 2020). However, these studies were conducted either in a single type of surgery with small sample sizes or in Western populations, which does not represent the growing aging population undergoing surgery in China.

In 2017, the European Society of Anesthesiology summarized the risk factors for POD, but these risk factors were not differentiated according to the type of surgery (Aldecoa et al. 2017). However, the occurrence of POD is known to be related to the type of surgery (Zipser et al. 2020), and major abdominal surgery, which is often performed on older patients, is associated with a high incidence of POD (Olin et al. 2005). Generally, major abdominal surgery is relatively complicated and is associated with a relatively large trauma burden, which can trigger a series of physiological stress reactions (Romain et al. 2020). The incidence of delirium after major abdominal surgery is reported to be 3–39% in the older population, which places a significant burden on society (Janssen et al. 2019a, b, Probst et al. 2017). Therefore, we conducted a retrospective study in older patients undergoing major abdominal surgery at the Chinese People’s Liberation Army (PLA) General Hospital to investigate the risk factors and construct a prediction model for POD.

In this study, we developed and validated a nomogram for POD in older patients who underwent major abdominal surgery. The model included eight clinical variables that can be routinely obtained, namely age, COPD, WBC, glucose, total protein, creatinine, emergency surgery, and anesthesia time. The cut-off values were 175 and 199, and patients could be divided into low-risk, medium-risk, and high-risk groups based on the total nomogram score.

POD, the most common postoperative complication in older patients undergoing major abdominal surgery, has a complex pathophysiological mechanism and often indicates a poor prognosis (Hughes et al. 2020, Siddiqi et al. 2006). We identified age, COPD, emergency surgery, anesthesia time, WBC, glucose, total protein, and creatinine as independent risk factors for POD. According to previous studies, age is a definite risk factor for almost all types of surgeries (Hughes, Boncyk 2020). Yamamoto et al. revealed that emergency surgery was an independent predictor for sepsis-associated delirium (Yamamoto et al. 2020). However, previous studies have not definitively established an association between a history of COPD and POD. COPD is a systemic inflammatory disease; long-term hypoxemia and repeated infections have been shown to be significant factors in POD (Bowman, Jones 2020). In a prospective observational study, Cui et al. found that low cerebral tissue oxygen saturation (SctO₂) may be associated with an increased risk of delirium (Cui et al. 2021). COPD may affect the occurrence of POD through this pathway. Additionally, extended anesthesia time implies greater levels of fluid input, more complicated surgery, more challenging intraoperative conditions, and the use of multiple drugs; and these factors are all associated with the occurrence of POD (Brown, LaFlam 2016).

We also found that four preoperative biochemical indicators (total protein, WBC, glucose, and creatinine) were closely associated with POD. Ishihara et al. found that serum albumin concentration (≤ 3.7 g/dL) was an independent risk factor for POD (Ishihara et al. 2021). Preoperative malnutrition is significantly associated with POD (Dong et al. 2023). Furthermore, globulin and WBC are mainly related to the immune function of the body and infection. Duceppe et al. found that active infection or sepsis, usually accompanied by elevated globulin and white blood cells, significantly increased the risk of delirium (Duceppe et al. 2019). Our current findings suggest that preoperative hyperglycemia is an independent risk factor for POD while preoperative diabetes is not; these findings are consistent with those by Windmann et al. who found that patients without diabetes with preoperative hyperglycemia may be at high risk for POD (Windmann et al. 2019). This indicates that perioperative blood glucose control is also vital for people without diabetes. In summary, these eight independent risk factors are relatively consistent with our clinical expectations, and the four perioperative biochemical indicators can provide potentially viable targets for further clinical intervention.

We developed and validated a predictive nomogram based on these eight predictors. Previous research has also attempted to establish predictive models of POD. Racine et al. failed to establish a prediction model for POD in older patients after non-cardiac surgery using various statistical methods (Racine et al. 2021). Wong et al. confirmed that hospital-acquired delirium can be predicted using machine learning methods (Wong et al. 2018); however, machine learning relies on a large number of clinical variables to improve the accuracy of prediction. More importantly, a model built using machine learning cannot guide targeted interventions, which greatly affects its clinical application. It is worth mentioning that Li et al. developed a prediction model for POD in a population across all age groups undergoing major abdominal surgery. The model contained six predictors, including age, history of arrhythmia, history of mental illness or cognitive impairments, albumin levels, coagulation function, and surgical Apgar score. The surgical Apgar score was calculated by intraoperative estimated blood loss, lowest mean arterial pressure, and lowest heart rate (Li et al. 2021). Their model involved nine variables, as the surgical Apgar score could not be obtained easily. Additionally, older patients represented only a low proportion (35.3%) of their training cohort. Compared with the model developed by Li et al., our model was developed with a much larger sample (n = 7577) involving the older population. Our model does not perform as well as their model in terms of AUC; however, our model is more applicable in clinical practice. First, POD mainly occurs in the older population, resulting in serious outcomes; therefore, we developed our model specifically for the older population. Second, in terms of variable selection, our model was more suitable for clinical application. The indicators included in our model (age, COPD history, emergency surgery, and anesthesia time) are relatively easy to access, while the other four biochemical parameters (WBC, glucose, total protein, and creatinine) are objectively obtained through routine preoperative blood testing, with only a small risk of error. Our findings imply that the identification of more effective biomarkers for delirium will lead to greater progress in the prediction of delirium.

The occurrence of POD prolongs the length of hospitalization for older patients and increases the average inpatient costs. It has been suggested that a fair amount of POD can be prevented (up to 40% in some studies) and multicomponent interventions are useful in POD prevention measures and postoperative recovery (Chen et al. 2017; Clemmesen et al. 2018; Hughes, Boncyk 2020; Humeidan et al. 2021, Ito et al. 2019, Khan et al. 2020, Perbet et al. 2018). The model we developed may facilitate the clinical application of these interventions and provide potential intervention indicators for POD treatment in the future. At the same time, our model may contribute to the postoperative rehabilitation of older patients and obtain better health and economic benefits.

Similar to other retrospective studies, this study may not accurately reflect the incidence of POD and preoperative patient status. The overall incidence of POD (~ 5%) for major abdominal surgery in older adults is low. This is likely the consequence of a retrospective approach used to diagnose POD. Hyperactive or mixed cases of POD were detected, whereas hypoactive cases of POD, which accounted for a large proportion, were missed with the employed retrospective approach. However, a series of medical interventions are applied for hyperactive or mixed cases of POD; therefore, the model we developed meets the requirement for clinical application. On the other hand, we did not use variables of perioperative pain, which are important risk factors for POD, since it would be difficult to obtain this information retrospectively. For the same reason, we were unable to validate the model described by Li et al. using our dataset because we excluded patients with a preoperative history of mental illness or cognitive impairment. Considering this, a multicenter prospective study will be useful to further validate the predictive value of our study. Therefore, we have initiated a multicenter, large-scale, prospective study on POD in older patients.

In conclusion, our study identified eight variables (age, COPD, emergency surgery, anesthesia time, WBC, glucose, total protein, and creatinine) as independent risk factors for POD in older patients after major abdominal surgery and developed a nomogram that was capable of predicting POD with high degrees of accuracy and stability.