Incidence and Risk Factors for Postoperative Delirium in Patients Undergoing Spine Surgery: A Systematic Review and Meta-Analysis

Background The present study aims to investigate the incidence and risk factors associated with postoperative delirium in patients undergoing spine surgery. Methods PubMed, EMBASE, Cochrane Library, and Science Citation Index were searched up to August 2019 for studies examining postoperative delirium following spine surgery. Incidence and risk factors associated with delirium were extracted. Odds ratios (OR) and 95% confidence intervals (CI) were calculated for outcomes. The Newcastle–Ottawa Scale (NOS) was used for the study quality evaluation. Results The final analysis includes a total of 40 studies. The pooled analysis reveals that incidence of delirium is 8%, and there are significant differences for developing delirium in age (OR 1.07; 95% CI 1.04–1.09), age more than 65 (OR 4.77; 95% CI 4.37–5.16), age more than 70 (OR 15.87; 95% CI 6.03–41.73), and age more than 80 (OR 1.91; 95% CI 1.78–2.03) years, male (OR 0.81; 95% CI 0.76–0.86), a history of alcohol abuse (OR 2.11; 95% CI 1.67–2.56), anxiety (OR 1.74; 95% CI 1.04–2.44), congestive heart failure (OR 1.4; 95% CI 1.21–1.6), depression (OR 2.5; 95% CI 1.52–3.49), hypertension (OR 1.12; 95% CI 1.04–1.2), kidney disease (OR 1.41; 95% CI 1.16–1.66), neurological disorder (OR 4.66; 95% CI 4.22–5.11), opioid use (OR 1.86; 95% CI 1.18–2.54), psychoses (OR 2.77; 95% CI 2.29–3.25), pulmonary disease (OR 1.81; 95% CI 1.27–2.35), higher mini-mental state examination (OR 0.7; 95% CI 0.5–0.89), preoperative pain (OR 1.88; 95% CI 1.11–2.64), and postoperative urinary tract infection (OR 5.68; 95% CI 2.41–13.39). Conclusions A comprehensive understanding of incidence and risk factors of delirium can improve prevention, diagnosis, and management. Risk of postoperative delirium can be reduced based upon identifiable risk factors.


Introduction
Postoperative delirium is a common complication after surgery in the elderly and causes difficulty in postoperative care [1,2]. It is defined as an acute change in the cognitive status characterized by fluctuating consciousness, attention, memory, perceptions, and behavior postoperatively [3].
Postoperative delirium often brings out many adverse outcomes, such as functional disability, increased health care costs, and higher morbidity and mortality rates [4]. us, a further understanding and prevention of delirium may help reduce these problems and the associated costs. Some previous studies have reported the incidence and risk factors for delirium. However, incidences of postoperative delirium differ greatly, and risk factors of these studies are inconsistent. erefore, we perform a systematic review and meta-analysis to explore incidence and risk factors for developing postoperative delirium following spine surgery.

Quality Assessment.
Two authors independently evaluated the quality of the studies, and the level of agreement between them was recorded. Any disagreements between the 2 authors were resolved by discussion with a third author. Newcastle-Ottawa Scale (NOS) was utilized to assess the quality of each study [7] since no studies were randomized controlled trials. Studies with 7-9 points could be identified as high quality, 5-6 points as moderate quality, and 0-4 as poor quality.

Statistical Analysis.
e meta-analysis of comparable data was performed using Stata/SE version 15.0 software. All adjusted odds ratio (OR) with 95% confidence interval (CI) Full-text articles excluded, with reasons: Small sample size (n = 1) No ORs and 95% CI or raw data to calculate ORs and 95% CI (n = 17) No risk factors or incidence (n = 44) Only abstract available (n = 2)         were collected and pooled to evaluate the relationships between various risk factors and postoperative delirium in patients undergoing spine surgery. In addition, crude ORs with 95% CIs were calculated based on the frequency reported in the original literature. Inconsistency was quantified with I 2 statistic, and an I 2 of >50% was considered to indicate substantial heterogeneity. e random-effects model or the fixed-effect model was used depending on the heterogeneity of studies included. A random-effects model was used for heterogeneous data. Otherwise, a fixed-effect model was used. Begg's and Egger's test were used to estimate publication bias, when 10 or more studies are presented. For any variable presenting with large heterogeneity, sensitive analysis or subgroup analysis was used to investigate the potential origin of heterogeneity.

Search Results.
ere were 1360 relevant studies included according to the search strategy. After the titles and abstracts were reviewed, 1256 of them were removed. A fulltext review was evaluated in the 104 records maintained, and 64 of them were excluded because they did not meet the inclusion criteria. Finally, 40 studies representing 712820 patients were included in the present meta-analysis ( Figure 1).

Study Characteristics and Quality Assessment.
e characteristics of the included studies are summarized in Table 1. 22 studies were conducted in Asian countries, 16 studies in North America, and 2 studies in Europe. 31 studies were retrospective, and 9 were prospective in design. e sample size ranged from 35 to 578457 patients. e reported incidence of delirium ranged from 0.49% to 31.43% for patients after spinal surgery. To evaluate the quality of each study, the NOS was utilized. In those studies, all of them were of moderate to high quality (range, 6-8) ( Table 1).

Incidence of Postoperative Delirium after Spine Surgery.
e final meta-analysis included 40 studies [1, from 7 different countries, and the pooled incidence was 8% ( Figure 2). ere was high heterogeneity (I-squared > 50%, P < 0.001). Interestingly, the heterogeneity remained high with each of the subgroups of study type, countries, or operated levels (Figure 2(a)-2(c)). After sensitive analysis, 3 studies [11,25,41] showed great influence on the pooled result (Figure 2(d)). e asymmetry Begg's funnel plot suggested the presence of publication bias for incidence of postoperative delirium after spine surgery (P < 0.001) (Figure 2(e)).

Risk Factors for Postoperative Delirium after Spine
Surgery.
e ORs and 95% CIs of the risk factors are displayed in Table 2. Among these, 33 factors were examined in 2 or more studies and 18 factors demonstrated statistical significance.

Discussion
Delirium is thought to be a less transient disorder than previously believed in several studies [8,11]. In addition, it has been reported that patients with postoperative delirium have a higher mortality rate than in those without it [4]. Due to the fact that delirium is varying and multifactorial, it will be helpful for prevention of delirium through identifying predictable risk factors. is systematic review and meta-analysis were performed to pool and identify the incidence and risk factors of postoperative delirium after spine surgery. e pooled incidence of delirium in this meta-analysis is 8%. However, the present study showed wide variation and heterogeneity in incidence of delirium. A previous meta-analysis of 6 studies reported incidence of delirium after spine surgery varies from 0.84% to 21.3% [47]. Interestingly, the heterogeneity remained high with each of the subgroups of study type, countries, or operated levels (Figures 2(a)-2(c)). We found that patients with spinal deformity have higher rate of delirium (10%) and lower rate in patients with lumbar spine (1%). Meanwhile, prospective studies have a higher incidence of postoperative delirium than retrospective studies. After sensitive analysis, 3 studies [11,25,41] showed great influence on the pooled result (Figure 2(d)). All these 3 studies have relatively a larger sample size (range, 13188 to 578457), low incidence of delirium (range, 0.49 to 5.1%), and retrospective nature of study design, which may contribute to the heterogeneity. e asymmetry Begg's funnel plot suggested the presence of publication bias for incidence of postoperative delirium after spine surgery, and lower incidence values could be missing (Figure 2(e)). One of the most important risk factors was older age, especially in patients over 65. is may be attributed to the fact that elderly patients are more likely influenced by agerelated physical and psychical changes. Aging is also associated with a higher incidence of comorbidity such as hypertension, diabetes mellitus, and pulmonary disease [12,30]. e highest rate of delirium in our meta-analysis is 31.43% in a multicenter prospective study with patient's age more than 90 [21]. Another significant demographic factor is male as a protective factor. rough subgroup analysis, we found that study design may contribute to the heterogeneity and prospective studies showing relatively a higher risk of developing delirium in females (Figure 3(a)). For publication bias, Begg' funnel plot demonstrated no significant bias (Figure 3(b)).
e present study showed that comorbidities significantly increase the risk of postoperative delirium after spine surgery. A history of alcohol abuse, congestive heart failure, hypertension, neurological disorder, opioid use, psychoses, and pulmonary disease are related to develop delirium. However, diabetes mellitus, history of surgery, and cerebral vascular diseases were not found to be related to developing   delirium, which was consistent with the previous metaanalysis [47]. For the cardiovascular comorbidity, the pooled result of 10 studies [8,11,15,23,26,30,31,34,43,45] showed no significance (OR 0.81; 95% CI 0.34-1.29) with low heterogeneity (I 2 0%) (Figure 4(a)). Only one study found cardiovascular comorbidity as a risk factor for delirium [11]. e symmetry Begg's funnel plot suggested no presence of publication bias for cardiovascular comorbidity (Figure 4(b)). Interestingly, however, pooled results showed congestive heart failure as a significant factor. is may be due to the severity of heart diseases.
For neurological or mental diseases, neurological disorder, psychoses, anxiety, and depression were found to be associated with developing delirium. e meta-analysis of 5 studies showed that mild cognitive impairment is not related to the occurrence of delirium (OR 2.43; 95% CI 0.99-3.86; I 2 0%). Meanwhile, parkinsonism was also not found to be related to postoperative delirium (OR 5.37; 95% CI 0.63-10.1). However, there is still controversy in the role of parkinsonism for postoperative delirium. Kim et al. [23] Note: weights are from random effects analysis       found that that parkinsonism is not a risk factor for postoperative delirium after multivariable analysis. Interestingly, Pan et al. [8] found an opposite result, which may be attributed to relatively a smaller sample of patients with parkinsonism in their study. Notably, the result should be explained with caution since the heterogeneity is high (I 2 88%). After subgroup analysis, there was a high heterogeneity between retrospective studies (Figure 6(a)). Moreover, the result of sensitive analysis showed two studies [24,25] contributing greatly to the high heterogeneity ( Figure 6(b)). Both studies were retrospective design and focus on patients with parkinsonism, which may result in high heterogeneity. Mental states, as assessed by MMSE, were associated with the development of delirium (OR 0.7; 95% CI 0.5-0.89).
rough subgroup analysis, we found that geographical factors may contribute to heterogeneity (Figure 7). is measure of the state of mental health appears to have a clearer association with postoperative delirium compared to Charlson Comorbidity Index (CCI) which assesses the number of specific medical comorbidities. ese findings are also seen in other studies where CCI appears less clearly associated with the incidence of delirium in older patients [12,49]. e finding that preoperative pain and opioid use is associated with increased probability of delirium has been previously reported in patients with or without hip fracture or patients with cancer [49,50]. In addition, elderly patients are more sensitive to opioid-related adverse events [51]. In patients with spine disease, pain may lead to stress reaction and changes of nerve conduction if not effectively controlled [34]. However, the accumulation of active metabolites in patients receiving opioid may contribute to the psychotic features such as delirium [52]. Hence, it is suggested that a less toxic drug, buprenorphine patch other than morphine, should be considered for patients with osteoarthrosis and other types of lumbago when pain continues despite adequate administrations of nonopioid analgesics [53].
In our study, intraoperative factors do not appear to influence the prevalence of delirium based on normal clinical practice such as blood loss, blood transfusion, cervical surgery, dural tear, operated levels, and operation time. Notably, for intraoperative blood loss, there was high heterogeneity among studies (Figure 8(a)). After sensitive analysis, we found that one study [23] focused on patients with parkinsonism lead to the high heterogeneity. In addition, high heterogeneity was also seen in the meta-analysis of blood transfusion (Figure 9(a)).
e sensitive analysis showed that the heterogeneity comes from one study [43], which had more fusion levels (2.27 ± 1.34) and blood loss (1263 ± 903) than other studies (Figure 9(b)). Postoperatively, patients experiencing complications such as UTI had a higher probability to develop delirium.
ere are some limitations in our study. First, no randomized controlled trials were included despite our exhausted search from literatures, which may influence the quality of the result. Second, although subgroup analyses were used, the pooled result of incidence was still reported with high heterogeneity, which should be explained with caution.

Conclusions
In summary, the study reveals that pooled incidence of delirium is 8% and age, gender, history of alcohol abuse, anxiety, congestive heart failure, depression, hypertension, kidney disease, neurological disorder, opioid use, psychoses, pulmonary disease, MMSE, preoperative pain, and postoperative UTI were significant factors for delirium after spine surgery. A comprehensive understanding of incidence and risk factors of delirium can improve prevention, diagnosis, and management.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Mingsheng Tan designed the study, and Xinjie Wu wrote this manuscript. Xinjie Wu and Wei Sun searched database, reviewed studies, and collected and analyzed data. All of the authors have read and approved the final manuscript.