Background
Delirium is an acute confusional state that is characterized by sudden alteration and fluctuation of mental status, inattention, and disorganized thinking. Although delirium mostly disappears within a few days, it can result in a delayed medical recovery, prolonged hospitalization, and increased costs [
1]. For example, early mobilization, which is critical in patients with hip fracture, can be hampered until the recovery of delirium. Delirium is known to be reversible, but recent reports have suggested that the cognitive decline experienced during delirium lasts after recovery [
2,
3]. Various abnormal physical conditions have been reported to be associated with delirium, though the pathophysiology of this condition remains a topic of debate [
4]. Delirium is one of the common complications in surgical patients. Postoperative delirium has occurred with a wide range of the incidence rates in surgical units. Hip fracture has a particularly high association with postoperative delirium. The higher incidence after hip fracture surgery may be the result of not only advanced age and more preexisting medical comorbidities but also with the traumatic event [
5].
In most studies, risk factors of delirium have mainly been considered in terms of medical and physical aspects [
6]. While delirium has been regarded as a result of systemic organic changes, psychological factors have been rarely considered in the occurrence of delirium [
7]. This forms a striking contrast to cognitive factors in that cognitive impairment has been reported as a risk factor for delirium in most studies [
8]. Some studies reported that preoperative anxiety was a predictive factor for emergence of delirium and major depression was significantly associated with the incidence of delirium [
9,
10]. However, these studies did not consider the results with respect to multiple candidate factors and the results were also heterogeneous depending on the variety of collected data. In terms of the relationship with personality, only one report showed that Type D personality, characterized by social inhibition and negative affect, had a nearly significant association with delirium [
10].
Although the effect of personality on delirium has not been addressed for a long time, personality traits have been shown to influence health [
11]. In particular, a study on the effect of personality on cognitive and psychological aspects has reported that individuals with high neuroticism, who are more likely to experience anxiety, anger, guilt, and depressed mood, showed increased probability of dementia [
12]. Hostility, a sub factor of neuroticism, has been suggested to be a strong risk factor for the development of physical diseases such as coronary heart disease [
13].
Given the evidence of the contribution of psychological factors on physical health, it is possible that the occurrence of delirium may also be associated with psychological factors. However, there is no prospective systemic study on this issue. In order to define the effects of psychological factors on the occurrence of delirium, we collected psychological data such as anxiety, depression, and personality type before surgery. In this study, we focused on patients with hip fracture because fragility in the setting of mental and physical stress and the higher rate of post-operative delirium in elderly patients may highlight unidentified risk factors for delirium compared with other surgical units. We hypothesized that patients with vulnerable psychological factors would be more likely to develop postoperative delirium. In addition, based on pathophysiological and etiological differences among patients with postoperative delirium, we hypothesized that a data-driven analysis of clinical data using Topological Data Analysis (TDA), which is a recently developed partial clustering method and has been applied in various clinical data [
14,
15], would identify meaningful sub-groups of delirium.
Discussion
In this study on psychological predictive factors of postoperative delirium, neuroticism and conscientiousness remained a risk factor in the final model. Other factors such as anxiety, depression, and other personality domains were not associated with postoperative delirium according to the logistic regression approach. Overall, we observed that lower MMSE score, higher neuroticism, lower conscientiousness, and regional anesthesia seemed to be related to more frequent occurrence of delirium in patients with advanced age undergoing hip fracture surgery. In addition, topology-based data analysis detected three distinct subgroups of delirium in the dimensions of MMSE, neuroticism, conscientiousness, and anesthesia method.
Among the psychological factors, personality traits which remained significant in the forward selection of multivariate analysis were neuroticism and conscientiousness. Considering that females showed higher conscientiousness and neuroticism than males [
27], these two factors might be affected by the gender effect. However, because no significant sex difference in personality traits was observed in our dataset, these two personality factors may be associated with postoperative delirium regardless of the gender effect. In TDA patients with postoperative delirium showed higher neuroticism and lower conscientiousness scores in the patient-patient network. Given that TDA is one of clustering techniques to find hidden patterns or grouping in data but allowing overlaps among clusters [
25,
14], a node (or a cluster) in the patient-patient network represents a group of patients with similar patterns of clinical and psychological characteristics and an edge represents the existence of an overlap between two nodes (or clusters). During the time of hospitalization patients with hip fracture may struggle to manage a stressful situation. Unexpected fracture, pain, and surgery as well as being in an unfamiliar hospital setting can be stressors. Therefore, an individual with high neuroticism or low conscientiousness may not be able control himself/herself effectively and may be less likely to cope with a crisis and interpret the situation as threatening and hopelessly difficult. This may be possible because an individual with high neuroticism is more likely to experience anxiety, anger, guilt, and depressed mood [
28].
In fact, preoperative anxiety and depression scores were not regarded as a predictive factor of postoperative delirium and were not included in TDA. The contribution of these two psychological factors to delirium has not been conclusive because of conflicting results among previous studies [
29]. In the collinearity test, neuroticism was not significantly correlated with anxiety and depression scores though the two scores were correlated each other. Nonetheless, mapping clinical information on topology showed significant positive correlations of filter metrics with HAS and HRSD scores. Filter matric is the first principal component and thus contains all important information of input features. Given that filter metrics were positively correlated neuroticism scores, personality traits related to anxiety and depression rather than these two factors themselves are supposed to cause psychological and physiological changes in patients to develop delirium after surgery.
A mechanism of neuroticism- or conscientiousness-related physical changes has been investigated in terms of the genetic, cellular, and metabolic level. For example, individuals with the top quartile of neuroticism scores or the lowest quartile of conscientiousness scores had a threefold increased risk of Alzheimer’s disease [
30]. In metabolic syndrome, neuroticism was associated with its prevalence, whereas individuals who scored in the top 10 % on conscientiousness were 40 % less likely to have it [
31]. Neuroticism was identified as a risk factor for heart problems [
13,
15]. In addition, neurotransmitter systems that modulate affective states and stress responses were associated with neuroticism scores [
32]. Serum cortisol responses to an opioid receptor antagonist were higher in subjects with high neuroticism than with low neuroticism [
33]. On the other hand, the conscientiousness element was correlated with cellular immune response [
34]. High neuroticism and low conscientiousness were also associated with higher levels of interleukin-6, a pro-inflammatory cytokine [
35]. Therefore, neuroticism and conscientiousness may be related to the development of an acute illness such as delirium. Regarding the relationship between stress hormones and systemic inflammation, these personality traits may also be involved in delirium from the genetic to system level via inflammation mechanisms.
Alternatively, neuroticism and conscientiousness may be related to nutritional factors in the development of delirium. According to TDA, the probability of postoperative delirium increased as filter metrics increased. This was not a surprising result because filter metrics were significantly associated with risk factors of postoperative delirium, e.g., the negative correlation with MMSE and conscientiousness scores and positive correlation with neuroticism scores. Furthermore, although levels of serum albumin and total protein were not identified as risk factors of postoperative delirium in logistic regression analysis, they were negatively correlated with filter metric. Considering that reduced protein intake decreases the serum albumin concentration [
36,
37] and then contributes to malnutrition and inflammation [
38], our results suggest that higher neuroticism and lower conscientiousness scores, which were positively correlated with filter metrics, might contribute to malnutrition in patients with postoperative delirium.
Meanwhile, we found no significant result in other personality traits, including extraversion, agreeableness, and openness. These traits can have a possible link to delirium by influence on the physical state. For example, low agreeableness or high extraversion has been suggested to be associated with drug users [
39,
40], and openness is associated with reduced risk of cardiovascular disease [
41]. However, because these associations may be limited some specific people, personality factors other than neuroticism and conscientiousness seem to be not important enough to have an effect on the occurrence of postoperative delirium.
In addition to a personality trait, regional anesthesia was associated with postoperative delirium. When we checked the effect of regional anesthesia using TDA, delirious patients with regional and general anesthesia were clearly separated in the patient-patient network by mapping the ratio of regional anesthesia. Studies on the contribution of the anesthetic method to delirium have shown conflicting results. Some studies have postulated that general anesthesia is more frequently associated with delirium because anesthetic agents including propofol influence neuronal processes, leading to changes in neurotransmitter signaling [
42]. However, there is also a study reporting no distinct effects on delirium between general and regional anesthesia [
43]. Although the difference in the incidence rate of delirium was not statistically significant between the two anesthetic methods, it was higher in regional anesthesia (50 %) than in general anesthesia (38 %) for patients with hip fracture in a prospective study [
44]. With regards to detailed information such as delirium severity or onset, there has never been a study according to the anesthetic method. In our study, delirium occurred earlier when regional anesthesia was used compared to general anesthesia. The duration and severity of delirium and cognitive decline during delirium were less in regional anesthesia compared to general anesthesia, though these were not significant. It might be possible that symptoms of delirium, which were more likely to pass unnoticed, may be detected by thorough prospective approaches and result in a higher incidence rate in regional anesthesia. Among possible explanations of regional anesthesia as a risk factor for postoperative delirium, deep sedation levels consistent with general anesthesia were frequently observed during regional anesthesia in elderly patients for hip fracture repair [
45]. There was other explanation that a decrease in systolic blood pressure or oxygen saturation during regional anesthesia might be associated with postoperative delirium [
46]. However, patients included in our study were not kept under sedation during regional anesthesia and hemodynamic features were not significantly different between patients with and without delirium after regional anesthesia. In the light of these facts, higher incidence of postoperative delirium after regional anesthesia needs to be explained by the effect of psychological factors rather than the complications of this method.
Our results showed that the interactive effects among regional anesthesia, neuroticism, and conscientiousness were associated with subgroups of postoperative delirium as observed in the patient-patient network (see insets in the second column of Fig.
3b). TDA provided more intuitive information by mapping personality traits and a ratio of regional anesthesia as a color in the node. Herein, it may be important to pay attention to the characteristics of ‘hip fracture repairs.’ During regional anesthesia, patients are usually in a voice-reacting conscious state [
47], and thus can respond to some stimuli. Specifically, the operative setting in an orthopedic unit such as being undressed around unfamiliar peoples, surgical positioning, unfamiliar environment, and sounds of a drill or hammer during surgery may cause stress in many patients. These uncomfortable experiences may provoke psychological disturbances, particularly in patients with high neuroticism. Taken together, regional anesthesia in an orthopedic unit may cause postoperative delirium by placing the patients in an environment that affects psychological aspects. In this sense, the implications of our results for patients with hip fracture are that personality traits should be considered when the anesthetic method is chosen and the environment should be optimized for a less stimulating setting when regional anesthesia is used.
Meanwhile, MMSE score remained a risk factor in the final model even after confounding factors were added for correction. Although the statistical power was somewhat low, there is no doubt that preoperative cognitive ability was associated with the occurrence of postoperative delirium. This finding is consistent with previous reports that cognitive impairment is one of the risk factors for delirium [
1,
8]. Although MMSE score and neuroticism proved to be an independent risk factor, it should be considered that cognitive ability including attention, perception, memory, problem solving, and comprehension may affect psychological factors such as anxiety and depression [
48]. In addition, most other studies have reported that advanced age is a risk factor for delirium [
1]. However, age itself was not a significant factor in our study and had marginal significance with a
p-value of 0.07 in univariate analysis. This may be due to the fact that age in our inclusion criteria (aged ≥ 70) was relatively high and the variation of age was relatively small. Previous studies have also documented that pain or inadequate pain control is likely to increase the risk of delirium [
49], whereas our study showed that the degree of pain and the following opioid medication were not associated with the occurrence of postoperative delirium.
There are some limitations in this study. First, our data were collected within a restricted period in only one center which resulted in a relatively small size for a prospective study. Although TDA produced informative outputs regarding personality traits and anesthesia types, identification of clinically meaningful subgroups of postoperative delirium was not achieved due to a small sample size. Second, although TDA has an advantage in deeper understanding of a disease, its application has been limited in several medical diseases such as diabetes, breast cancer [
50], and attention deficit hyperactivity disorder [
51]. More studies are necessary to determine its usefulness in the clinical domain. Third, more comprehensive psychological assessments were not performed because all patients were acutely ill due to hip fracture and other traumatic injuries. Because of the same reason, personality traits were measured using the short version instead of the full version of the BFI. However, this short version may be rather useful to acutely ill patients because it was proved to be reliable and valid [
21]. Forth, because patients were acutely ill and preoperative assessment was done in a supine position in order to prevent the aggravation of pain, their baseline MMSE scores seemed to be underscored.
Acknowledgements
The authors thank In Jung Son, M.D., M.S., Ho Jun Choi, M.D., Min Suk Ko, M.D., Jun Yeol Lim, M.D., Sang Yok Seok, M.D., and Jun Young Kim, M.D. for helping to collect data, the nursing staffs of emergency center and orthopedic unit for supporting to enroll patients, and Han Na Yoo, Ph.D. for helping statistical analysis.