Pathophysiological factors of delirium among critically ill elders after non-cardiac surgery based on artificial neural networks: a pilot study

Mengting Ji, Shunpeng Xing, Yan Yang

Correspondence: Yan Yang, 160 Pujian Road, Shanghai, 200127, (China)
Mengting Ji, Mobile: 86 18817840389; E-mail: jimengting2012@163.com
ABSTRACT
Objectives: To utilize artificial neural networks (ANN) for early identification of pathophysiological risk factors of delirium among elders in surgical intensive care unit (SICU) after non-cardiac surgery.
Methodology: This prospective, single-center observational study was conducted at an SICU with 19 beds. Patients aged 65 and over were included. Delirium was screened by the Richmond Agitation-Sedation Scale and Confusion Assessment Method for the ICU. Factors analyzed were age, gender, disease, symptoms, sedatives and analgesics used, and biochemical parameters. Mean impact value (MIV) of each variable was calculated by MATLAB. Then, ANN was established based on the SPSS 20.0.
Results: Data from 134 patients were analyzed. The mean age was 77.045 ± 7.375 years (65-94), 50.7% were male and 11.94% had delirium. There were 13 important risk factors based on MIV, which were included to build ANN. The important pathophysiological risk factors were age, the use of sedatives, dose of propofol, dose of remifentanil, acidosis, fever, hyponatremia, hyperkalemia, albumin level, pre-albumin level and Child-Pugh score, and anemia. The area under ROC curve was 0.893.
Conclusion: Application and significance of artificial neural networks mainly lies on mean impact value, which can be used as a relative stable reference for early screening of pathophysiological risk factors of delirium among critically ill elders.
Keyword: Artificial Neural Networks; Critically ill; Elders; Delirium; Surgical intensive care unit; Risk factors
Citation: Ji M, Xing S, Yang Y. Pathophysiological factors of delirium among critically ill elders after non-cardiac surgery based on artificial neural networks: a pilot study. Anaesth Pain & Intensive Care 2018;22(4):­­424-430
Clinical Trial registration: Chinese Clinical Trial Register No. ChiCTR-OOC-16008154.
Received: 28 Sep 2018, Reviewed: 16 Oct, 30 Oct 2018, Corrected: 01 Nov 2018, Accepted: 02 Nov 2018

INTRODUCTION

Delirium is an acute organic brain dysfunction with fluctuating course of attention and cognition.1 It is an unpleasant experience for all involved, including patients, caregivers, and nursing staff.2 Patients describe their delirium experience as frightening, life-threatening, fearful, panic, or unsafe. Patients’ relatives are distressed by delirium and nurses report a higher burden when caring for delirious patients.3 The incidences of delirium in the surgical intensive care units (SICU) vary from 2.48 to 79 %,4,5 depending on the types of surgical procedures, patient settings, and screening tools for diagnosis of delirium. Although delirium can be experienced in all age groups, it has an increased frequency in elders.6 As reported that prevalence of delirium among older intensive care unit (ICU) patients ranges from 78% to 87%.5 Generally, delirium is a serious problem because of its negative clinical outcomes, such as prolonged intensive care unit (ICU) stay and hospitalization, decreased cognitive functioning, higher mortality rate, and increased odds of re-admission.7,8 Data from previous researches indicated that the median ICU, postoperative hospital, and hospital length of stay were approximately 2.5, 2, and 4 days longer for patients with delirium respectively.7 The duration of delirium in the ICU predicts long-term cognitive impairment, physical disability and death up to a year after discharge.9 Once delirium develops in patients, pharmacotherapy options are limited.10
For all that, diagnosis of delirium is not sufficiently made. Delirium is not noticed in 64–84% and not diagnosed in 33–66% 11,12 for it is “expected” to happen in patients with severe illness, and medical resources are dedicated to those who are more immediate “life-threatening” patients preferentially. Besides, clinicians generally give less attention to acute brain syndrome as a predictor of poorer overall outcome than acute dysfunction of other organ systems and regard delirium as transitory with no long-term adverse effect.13 A delay in delirium therapy is associated with higher mortality.14
Fortunately, delirium can be prevented. Evaluation of patients at risk may lead to better recognition, earlier intervention and decreasing delirium prevalence. Possible risk factors include medication, acute illness, environmental and patient characteristics, and underlying chronic pathology.15 However, some of these factors are not modifiable, and some are unavailable for assessment for ICU patients. Data of pathophysiological factors are convenient to acquired, especially at the first 24 hours in the ICU settings, so that clinicians and nurses are allowed to pay key attention and deliver adequate care to patients at risk at an early stage.

With regard to risk factors identification, bivariate logistic regression was most widely used before. Recently, artificial neural networks (ANN) has been viewed as a robust technique with higher accuracy than logistic regression analyses.16,17 The architecture for ANN employed a systematic method where the number of neurons is changed incrementally.16 It has no limitation on the form of relationship between response and predictor variables. Also, ANN finds the form of relationship which is not necessarily linear. Furthermore, the data are implicitly analyzed in ANN, so it has a high probability of finding the correct solution even if a part of the network layers is deleted or works incorrectly.18 This study was aimed to utilize ANN for early identification of pathophysiological risk factors of delirium in critically ill elders in SICU after non-cardiac surgery.

METHODOLOGY

This was a prospective single-center observational study in a 19-bed surgical ICU of our Hospital from March 2016 to April 2017. The study was approved by hospital ethical committee (No. 2014[012]). Informed consent was obtained as patient privacy rights must always be observed.

Inclusion criteria were patients aged 65 years or more who stayed in surgical ICU for at least 24 hours. Following patients were excluded: (1) with cognitive impairment before operation, (2) with history of delirium, (3) with visual and (or) hearing impairment , (4) taking antipsychotic drugs in recent six months, (5) received cerebral operations, (6) operation time less than one hour, (7) ICU stay less than 24 hours, (8) refused to participated in the study.

Delirium Assessment:

Delirium was screened by ten-point rating scale for assessing sedation status based on the Richmond Agitation-Sedation Scale (RASS) and Confusion Assessment Method for the ICU (CAM-ICU) criteria. Trained nurses applied the CAM for delirium evaluation during three shifts a day at 0600, 1400, 2200 hours during the SICU stay. The training program for nurses consisted of three parts, namely information about delirium, one-to-one instruction and evaluations at the patients’ bedside completed by a delirium expert compared with the evaluations completed by the nurses.

Factors

Factors analyzed were age, gender, disease (stroke, diabetes, hypertension, coronary heart disease, arrhythmia, hyperlipidemia, asthma and COPD), symptoms (fever, hypertension and hypotension), sedatives and analgesics (propofol, midazolam, dexmedetomidine, fentanyl and remifentanil) and biochemical parameters (white blood cell (WBC), red blood cell (RBC), hemoglobin (Hb), Hematocrit (Hct), platelet count (PLT), total protein (TP), albumin (Alb), pre-albumin (PA), alanine aminotransferase (ALT), aspartate transaminase (AST), total bilirubin (TBIL), direct bilirubin (DBIL), creatinine (Cr), blood urea nitrogen (BUN), blood sodium (Na+), blood potassium (K+), blood calcium (Ca++), alkalinity (pH), arterial blood gas- carbon dioxide partial pressure (PaCO2), oxygen partial pressure (PaO2), and blood lactate (Lac)) during first 24-hour stay in ICU . We did not include morphine and pethidine because they are not used at our center. Child-Pugh score was calculated. Data of these factors were observed once during the first 24 h in the ICU setting.

Statistical Analysis: Statistical analyses were performed by MATLAB software and SPSS 20.0. Mean impact value (MIV) of each variable was calculated by MATLAB. First, the MIV was used to calculate the influence of factors to the dependent variable, for positive or negative impact. The second step was ANN establishment based on the SPSS 20.0 software. Furthermore, a 70% sample for the learning and a 30% sample for the test and prediction of the topology of the final network were used. Finally, the area under ROC curve (AUC) was used to interpret predictive accuracy. .
RESULTS

From the 174 patients screened during the study period, 134 were eligible and were analyzed (Figure 1). The mean age of participants was 77.045 ± 7.375 years (65-94), 50.7% were male and 16 (11.940%) had delirium during their ICU stay. The sequence of the factors according to their absolute MIVs was sorted.

There were 13 important risk factors, whose absolute value of MIV was greater than 0.05, were included to build ANN (Table I). The SPSS Neural Network software automatically re-scales the negative pseudo-probabilities (less than 0) and pseudo-probability more than 1 irrespective of the value of the dependent variable in this case. According to the absolute value of MIV, the rank of the 13 factors was as follows: alkalinity (pH), body temperature, blood sodium (Na+), albumin (Alb), Child-Pugh score, use of sedatives, arterial carbon dioxide partial pressure (PaCO2), hemoglobin (Hb), remifentanil, propofol, age, pre-albumin (PA), and blood potassium (K+).

There were 93 cases originally assigned to the training sample, and 41 to the testing sample. According to data from the classification table (Table 2), the correct percentage was 93.5% in the learning set, and 75.6% in the testing set.

The network performance was in a good pattern when the network contained 13 risk factors. The ROC curve (Figure 2) described the sensitivity and specificity for all possible cutoffs in a single plot. The chart shown here displayed two curves, one for the category “Delirium” and one for the category “Non-Delirium”. The curves were sharply upward to the upper left corner of the chart and followed with a stability to the upper right corner. The area under ROC curve (AUC) was 0.893, which was used to interpret predictive accuracy.

DISCUSSION

The CAM-ICU adapted from the CAM, is an excellent bedside tool19 for non-psychiatrists to evaluate patients in the ICU setting for delirium even if mechanically ventilated (MV), who is frequently under-diagnosed if monitoring instruments are not used.20 The CAM-ICU evaluates: (1) acute onset and fluctuating course; (2) inattention; (3) disorganized thinking; and (4) altered level of consciousness. Delirium is considered in presence of 1 and 2 and additional either 3 or 4.21,22 Factors included in this study were based on literatures review, including cohort studies,6,23-34 systematic reviews,35,36 and meta-analysis.

In this study, the rate of delirium among elderly aged 65 years and above admitted in ICU after non-cardiac surgery was 11.94%, similar with Ranhoff et al23 report of delirium in 13.73% among elders (aged 60 and over) during their stay in the SICU.

Delirium is a temporary mental disorder that is characterized by abnormal arousal, language and cognition, perception, orientation, mood, sleeping patterns, and neurologic functioning in the context of a medical etiology.7 The incidences of delirium reported in previous studies various according to the research population, types of settings, and evaluation tools for diagnosis. The prevalence rate was 100% among 107 survivors treated with therapeutic hypothermia after cardiac arrest.37 Similarly, patients suffering from open heart surgery,38 liver transplantation,39 esophageal cancer,40 major trauma,26 or burn20 in ICU settings are also reported to have a high incidence of delirium, 41.7%, 47.4%, 50%, 59% and 77% respectively.

Numerous studies have found ICU delirium to be associated with many negative outcomes such as: increased time on the mechanical ventilation, longer ICU and hospital lengths of stay, increased costs, higher mortality –both in-hospital and after discharge, and greater long-term cognitive dysfunction.4,41,42 Recognition of risk factors, therefore, is a key component for the prevention and early identification of delirium.

The ANN is a nonlinear dynamic system, an information-deal-with system invoked by biology neural network for its structure, function and some basic characters, but being abstracted and simplified. It has distributed storage form and parallel disposing form of information.43 The ANN with self-study ability did not require the distribution form and independence of variables and also could handle the problem of collinearity better.18 The MIV was an important index of calculating influence of factors to the dependent variable, whose absolute data in rank can be used to select factors before ANN analysis.44
In this study, we preliminarily screened pathophysiological factors, using MIV and examined reliability of results through ANN analysis. According to the absolute values of MIV, we could determine the ranking of importance of each factor and provide the basis for further analysis. Results of this study indicated that pathophysiological factors of delirium among elders after non-cardiac surgery in the SICU contains: age, use of sedatives, dose of remifentanil, dose of propofol, body temperature, alkalinity (pH), arterial carbon dioxide partial pressure (PaCO2), albumin (Alb), pre-albumin (PA), Child-Pugh score, hemoglobin (Hb), blood sodium (Na+), and blood potassium (K+). Factors above could be collected at the first 24 hours in the ICU setting.

Increased age has been universally identified as a risk factor of delirium,28 especially for those who are admitted for surgery.45 Delirium in elderly ( >  65 years) patients was associated with an increased risk of death.46 Increasing age is a universally identified risk factor of delirium, suggesting that the naturally aged brain might be the basis of the occurrence of delirium.47
Sedatives and analgesics factors contributing to delirium are controversial in previous studies. In this study, the use of sedatives and higher dose of propofol both had impact on delirium.  But midazolam or dexmedetomidine did not. Since excessive sedation can prolong length of stay, it may lead to increased risk of delirium. However, some but not all clinical studies demonstrated that propofol led to delirium.48 Midazolam, as one of benzodiazepines, is an independent predictor of transitioning to delirium (p < 0.01),26 and even associated with longer duration of delirium.5 But, Zaal et al49 confirmed that benzodiazepines increases the risk for delirium in critically ill adults seems to be limited to continuous infusion only. Similarly, dexmedetomidine may indirectly decrease the incidence of delirium as reported in some researches. However, its direct effect on the prevention of delirium remains theoretical and unproven.50
With regard to association between analgesics and delirium, result of this study demonstrated that higher dose of remifentanil was considered as a protective factor of delirium. In a prospective randomized controlled trial from China reports that, patients receiving midazolam had lower incidence (p < 0.01) and shorter duration of delirium (p < 0.001). The reported no impact of fentanyl to delirium. But, Pandharipande et al reported that fentanyl was a risk factor for delirium in the SICU (p < 0.01) but not in the TICU (p > 0.05).26 Therefore, future prospective study is needed to investigate the use of sedatives and (or) analgesics and occurrence of delirium.

Acidosis was viewed as an important risk factor leading to delirium, which was matched with data from several studies before.6,30,51 There’s still no consensus on types of acidosis, metabolic, respiratory, or both leads to delirium. Fever was always deemed as a sign of infections which leads to mental changes because of cytokines and/or bacterial toxins and cerebral metabolic changes.6 However, WBC as a laboratory index of infection had little impact on delirium in this study. As a result, fever associated with delirium may result from other reasons except from infections. Among value of blood electrolytes, hyponatremia and hyperkalemia were found to be important risk factors contributing to delirium in present study. While, hypocalcemia was not. Low Alb, PA and high Child-Pugh score, index of renal function, were important predictors of delirium in present study. The ALT, AST, TBIL, and DBIL were not. Anemia was also a predictor of delirium in previous researches.4,6
One of the limitations in the study is that not all risk factors, including medication, acute illness, environmental and patient characteristics, and underlying chronic pathology, were considered in the analysis, and some of information may be lost. Besides, this is a pilot study based on single center. Therefore, it is promising for the application of ANN based on the MIV into a large scale, longitudinal study to determine the association between delirium and modifiable risk factors for timely and optimal intervention to reduce the burden resulting from delirium in critically ill elder patients.

CONCLUSION

Delirium was found to occur in nearly 12% of critically ill elders in SICU after non-cardiac surgery. The important pathophysiological risk factors associated with delirium were age, the use of sedatives, dose of propofol, dose of remifentanil, acidosis, fever, hyponatremia, hyperkalemia, albumin level, pre-albumin level and Child-Pugh score, and anemia. Furthermore, this study suggests that application and significance of artificial neural networks (ANN) mainly lies on mean impact value (MIV), which can be used as a relative stable reference for early screening of variables.

Financial declaration: This research was supported by grants from Shanghai Pudong New Area Municipal Commission of Health and Family Planning Commission of Science and Technology Development Special Fund [Grant number: PW2014E-1] and Shanghai Municipal Education Commission- Gaoyuan Nursing [Grant number: hlgy16016kyx], Shanghai, China.

Authors’ contribution:

MJ: Conduction of the work and manuscript writing
SX: Data collection
YY: Study design and manuscript editing

REFERENCE
  1. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 5th ed. Washington: American Psychiatric Association Publishing; 2013.
  2. Breitbart W, Gibson C, Tremblay A. The delirium experience: delirium recall and delirium-related distress in hospitalized patients with cancer, their spouses/caregivers, and their nurses. Psychosomatics. 2002;43(3):183-94. [PubMed] DOI: 10.1176/appi.psy.43.3.183
  3. Hasemann W, Tolson D, Godwin J, Spirig R, Frei IA, Kressig RW. A before and after study of a nurse led comprehensive delirium management programme (DemDel) for older acute care inpatients with cognitive impairment. Int J Nurs Stud. 2016;53:27-38. [PubMed] DOI: 10.1016/j.ijnurstu.2015.08.003
  4. Horacek R, Krnacova B, Prasko J, Latalova K. Delirium as a complication of the surgical intensive care. Neuropsychiatr Dis Treat. 2016;12:2425-34. [PubMed] DOI: 10.2147/NDT.S115800
  5. Pisani MA, Murphy TE, Araujo KL, Slattum P, Van Ness PH, Inouye SK. Benzodiazepine and opioid use and the duration of intensive care unit delirium in an older population. Crit Care Med. 2009;37(1):177-83. [PubMed] DOI: 10.1097/CCM.0b013e318192fcf9
  6. Aldemir M, Ozen S, Kara I, Sir A, Bac B. Predisposing factors for delirium in the surgical intensive care unit. Crit Care. 2001;5(5):265-70. [PubMed]
  7. Zhang WY, Wu WL, Gu JJ, Sun Y, Ye XF, Qiu WJ, et al. Risk factors for postoperative delirium in patients after coronary artery bypass grafting: A prospective cohort study. J Crit Care. 2015;30(3):606-12. [PubMed] DOI: 10.1016/j.jcrc.2015.02.003
  8. Zaal IJ, Devlin JW, Peelen LM, Slooter AJC. A systematic review of risk factors for delirium in the ICU. Crit Care Med. 2015;43(1):40-7. [PubMed] DOI: 10.1097/CCM.0000000000000625
  9. Brummel NE, Girard TD. Preventing delirium in the intensive care unit. Crit Care Clin. 2013;29(1):51-65.
  10. Farina N, Smithburger P, Kane-Gill S. Screening and management of delirium in critically ill patients. Hosp Pharm. 2015;50(8):667-71. [PubMed] DOI: 10.1310/hpj5008-667
  11. Bilge EU, Kaya M, Senel GO, Unver S. The Incidence of delirium at the postoperative intensive care unit in adult patients. Turk J Anaesthesiol Reanim. 2015;43(4):232-9. [PubMed] DOI: 10.5152/TJAR.2015.93798
  12. Ely E, Margolin R, Francis J, May L, Truman B, Dittus R, et al. Evaluation of delirium in critically ill patients: validation of the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU). Crit Care Med. 2001(29):1370-9. [PubMed]
  13. Lahariya S, Grover S, Bagga S, Sharma A. Delirium in patients admitted to a cardiac intensive care unit with cardiac emergencies in a developing country: incidence, prevalence, risk factor and outcome. Gen Hosp Psychiatry. 2014;36(2):156-64. [PubMed] DOI: 10.1016/j.genhosppsych.2013.10.010
  14. Heymann A, Radtke F, Schiemann A, Lutz A, MacGuill M, Wernecke KD, et al. Delayed treatment of delirium increases mortality rate in intensive care unit patients. J Int Med Res 2010;38(5):1584-95. [PubMed] DOI: 10.1177/147323001003800503
  15. Kalabalik J, Brunetti L, El-Srougy R. Intensive care unit delirium: a review of the literature. J Pharm Pract. 2014;27(2):195-207. [PubMed] DOI: 10.1177/0897190013513804
  16. Isma'eel HA, Cremer PC, Khalaf S, Almedawar MM, Elhajj IH, Sakr GE, et al. Artificial neural network modeling enhances risk stratification and can reduce downstream testing for patients with suspected acute coronary syndromes, negative cardiac biomarkers, and normal ECGs. Int J Cardiovasc Imaging. 2016;32(4):687-96. [PubMed] DOI: 10.1007/s10554-015-0821-9
  17. Freeman R, Eagle K, Bates E, Werns S, Kline-Rogers E, Karavite D, et al. Comparison of artificial neural networks with logistic regression in prediction of in-hospital death after percutaneous transluminal coronary angioplasty. Am Heart J. 2000;140(3):511-20. [PubMed] DOI: 10.1067/mhj.2000.109223
  18. Adavi M, Salehi M, Roudbari M. Artificial neural networks versus bivariate logistic regression in prediction diagnosis of patients with hypertension and diabetes. Med J Islam Repub Iran. 2016;30:312. [PubMed]
  19. Wong C, Holroyd-Leduc J, Simel D, Straus S. Does this patient have delirium? value of bedside instruments. JAMA. 2010;304(7):779-86. [PubMed] DOI: 10.1001/jama.2010.1182
  20. Agarwal V, O'Neill PJ, Cotton BA, Pun BT, Haney S, Thompson J, et al. Prevalence and risk factors for development of delirium in burn intensive care unit patients. J Burn Care Res. 2010;31(5):706-15. [PubMed] DOI:
10.1097/BCR.0b013e3181eebee9
  1. Inouye S, Dyck Cv, Alessi C, Balkin S, Siegal A, Horwitz R. Clarifying confusion: the confusion assessment method. a new method for detection of delirium. Ann Intern Med. 1990;113(12):941-8. [PubMed]
  2. Ely E, Truman B, Shintani A, Thomason J, Wheeler A, Gordon S, et al. Monitoring sedation status over time in ICU patients: reliability and validity of the Richmond Agitation-Sedation Scale (RASS). JAMA. 2003;289(22):2983-91. [PubMed] DOI: 10.1001/jama.289.22.2983
  3. Ranhoff A, Rozzini R, Sabatini T, Cassinadri A, Boffelli S, Trabucchi M. Delirium in a sub-intensive care unit for the elderly: occurrence and risk factors. Aging Clin Exp Res. 2006;18(5):440-5. [PubMed]
  4. Dubois M, Bergeron N, Dumont M, Dial S, Skrobik Y. Delirium in an intensive care unit: a study of risk factors. Intensive Care Med. 2001;27(8):1297-304. [PubMed]
  5. Ouimet S, Kavanagh B, Gottfried S, Skrobik Y. Incidence, risk factors and consequences of ICU delirium. Intensive Care Med. 2007;33(1):66-73. [PubMed] DOI: 10.1007/s00134-006-0399-8
  6. Pandharipande P, Cotton BA, Shintani A, Thompson J, Pun BT, Morris JA, Jr., et al. Prevalence and risk factors for development of delirium in surgical and trauma intensive care unit patients. J Trauma. 2008;65(1):34-41. [PubMed] DOI: 10.1097/TA.0b013e31814b2c4d
  7. Van Rompaey B, Elseviers MM, Schuurmans MJ, Shortridge-Baggett LM, Truijen S, Bossaert L. Risk factors for delirium in intensive care patients: a prospective cohort study. Crit Care. 2009;13(3):R77. [PubMed] DOI: 10.1186/cc7892
  8. Cheng-mei S, Dong-xin W, Kai-sheng C, Xiu-e G. Incidence and risk factors of delirium in critically ill patients after non-cardiac surgery. Chin Med J (Engl). 2010;123(8):993-9. [PubMed]
  9. Serafim R, MFDutra, Saddy F, Tura B, Castro Jd, Villarinho L, et al. Delirium in postoperative nonventilated intensive care patients: risk factors and outcomes. Ann Intensive Care. 2012;2(1):51. [PubMed] DOI: 10.1186/2110-5820-2-51
  10. Sharma A, Malhotra S, Grover S, Jindal SK. Incidence, prevalence, risk factor and outcome of delirium in intensive care unit: a study from India. Gen Hosp Psychiatry. 2012;34(6):639-46. [PubMed] DOI: 10.1016/j.genhosppsych.2012.06.009
  11. Li W, Cha Y, Wang J. Factors Associated with the development of delirium in elderly patients in intensive care units. J Nurs Res. 2015;23(4):322-9. [PubMed] DOI: 10.1097/JNR.0000000000000082
  12. Pipanmekaporn T, Chittawatanarat K, Chaiwat O, Thawitsri T, Wacharasint P, Kongsayreepong S, et al. Incidence and risk factors of delirium in multi-center Thai surgical intensive care units: a prospective cohort study. J Intensive Care. 2015;3:53. [PubMed] DOI: 10.1186/s40560-015-0118-z
  13. van den Boogaard M, Schoonhoven L, Hoeven JG, Achterberg T, Pickkers P. Incidence and short-term consequences of delirium in critically ill patients: A prospective observational cohort study. Int J Nurs Stud 2012;49(7):775-83. [PubMed] DOI: 10.1016/j.ijnurstu.2011.11.016
  14. van den Boogaard M, Pickkers P, Slooter AJ, Kuiper MA, Spronk PE, H. van der Voort PH, et al. Development and validation of PRE-DELIRIC (prediction of delirium in ICU patients) delirium prediction model for intensive care patients：observational multicentre study. BMJ. 2012;344:e420. [PubMed] DOI: 10.1136/bmj.e420
  15. Van Rompaey B, Schuurmans MJ, Shortridge-Baggett LM, Truijen S, Bossaert L. Risk factors for intensive care delirium: a systematic review. Intensive Crit Care Nurs. 2008;24(2):98-107. [PubMed] DOI: 10.1016/j.iccn.2007.08.005
  16. Mattar I, Chan MF, Childs C. Risk factors for acute delirium in critically ill adult patients: a systematic review. ISRN Crit Care. 2013;2013:1-10. DOI: 10.5402/2013/910125
  17. Pollock JS, Hollenbeck RD, Wang L, Holmes B, Young MN, Peters M, et al. Delirium in survivors of cardiac arrest treated with mild therapeutic hypothermia. Am J Crit Care. 2016;25(4):e81-9. [PubMed] DOI: 10.4037/ajcc2016581
  18. Chang YL, Tsai YF, Lin PJ, Chen MC, Liu CY. Prevalence and risk factors for postoperative delirium in a cardiovascular intensive care unit. Am J Crit Care. 2008;17(6):567-75. [PubMed]
  19. Wang SH, Wang JY, Lin PY, Lin KH, Ko CJ, Hsieh CE, et al. Predisposing risk factors for delirium in living donor liver transplantation patients in intensive care units. . PLoS One. 2014;9(5):e96676. [PubMed] DOI: 10.1371/journal.pone.0096676
  20. Takeuchi M, Takeuchi H, Fujisawa D, Miyajima K, Yoshimura K, Hashiguchi S, et al. Incidence and risk factors of postoperative delirium in patients with esophageal cancer. Ann Surg Oncol. 2012;19(12):3963-70. [PubMed] DOI: 10.1245/s10434-012-2432-1
  21. CIBS Center. Outcomes associated with delirium [Online]. [cited 2016 Oct 10]; Available from: URL: http://www.icudelirium.org/delirium/associated-outcomes.html
  22. Salluh JIF, Wang H, Schneider EB, Nagaraja N, Yenokyan G, Damluji A, et al. Outcome of delirium in critically ill patients: systematic review and meta-analysis. BMJ. 2015;350:h2538. [PubMed] DOI: 10.1136/bmj.h2538
  23. He XM. Study on artificial neural networks method in multianalysis of medical data (Chinese). Shanghai, China: The Second Military Medical University; 2001.
  24. Shi F, Wang X-C, Yu L, Li Y, Zhang Y-L. Matlab artificial neural networks—30 case analysis (Chinese). Beijing: Beihang University Press; 2010.
  25. Kim MY, Park UJ, Kim HT, Cho WH. Delirium prediction based on hospital information (delphi) in general surgery patients. Medicine (Baltimore). 2016;95(12):e3072. [PubMed] DOI: 10.1097/MD.0000000000003072
  26. Khan BA, Zawahiri M, Campbell NL, Fox GC, Weinstein EJ, Nazir A, et al. Delirium in Hospitalized patients: implications of current evidence on clinical practice and future avenues for research —a systematic evidence review. J Hosp Med. 2012;7(7):580–9. [PubMed] DOI: 10.1002/jhm.1949
  27. Mu DL, Wang DX, Li LH, Shan GJ, Li J, Yu QJ, et al. High serum cortisol level is associated with increased risk of delirium after coronary artery bypass graft surgery: a prospective cohort study. Crit Care. 2010;14(6):R238. [PubMed] DOI: 10.1186/cc9393
  28. Brown KE, Mirrakhimov AE, Yeddula K, Kwatra MM. Propofol and the risk of delirium: exploring the anticholinergic properties of propofol. Med Hypotheses. 2013;81(4):536-9. [PubMed] DOI: 10.1016/j.mehy.2013.06.027
  29. Zaal IJ, Devlin JW, Hazelbag M, Klein Klouwenberg PM, van der Kooi AW, Ong DS, et al. Benzodiazepine-associated delirium in critically ill adults. Intensive Care Med. 2015;41(12):2130-7. [PubMed] DOI: 10.1007/s00134-015-4063-z
  30. Nelson S, Muzyk AJ, Bucklin MH, Brudney S, Gagliardi JP. Defining the role of dexmedetomidine in the prevention of delirium in the intensive care unit. Biomed Res Int. 2015;2015:635737. [PubMed] DOI: 10.1155/2015/635737
  31. Kwizera A, Nakibuuka J, Ssemogerere L, Sendikadiwa C, Obua D, Kizito S, et al. Incidence and risk factors for delirium among mechanically ventilated patients in an african intensive care setting: an observational multicenter study. Crit Care Res Pract. 2015;2015:491780. [PubMed] DOI: 10.1155/2015/491780
Figure 1: Flow diagram

4-OA-F1
There were 174 patients screened during the study period, 12 refused to participated in the study, 25 were excluded because of cognitive impairment before operation, visual and (or) hearing impairment, taking antipsychotic drugs in recent six months, cerebral operations, or operation time less than one hour. As a result, 137 patients enrolled and 3 dropout. In the end, 134 patients were eligible and were analyzed.

Figure 2: The Receiver Operating Characteristic (ROC) Curve shows good network performance 13 risk factors for delirium in critically ill elders in
surgical ICU after non-cardiac surgery

4-OA-F2
 The chart shown here displays two curves, one for the category No (Non-Delirium) and one for the category Yes (Delirium). The curves are symmetrical about a 45-degree line from the upper left corner of the chart to the lower right.

 

Table 1: Rank of 13 important pathophysiological risk factors based on MIV
Factors Delirium Non-Delirium MIV
X / n SD / (%) X / n SD / (%)
pH  <  7.35 3 18.750 18 15.254 5.834
≥ 7.35 13 81.250 100 84.746  
Body Temperature 36.625 0.949 36.181 0.986 -0.335
Na+ (mmol/L) 141.000 3.596 140.305 4.359 -0.283
Alb (g/L) 27.187 6.299 30.362 4.678 0.142
Child-Pugh Score 6.438 0.629 6.305 0.965 0.120
Sedatives Yes 14 87.500% 98 83.051% 0.118
No 2 12.500% 20 16.949%  
PaCO2 (mmHg) 38.750 10.260 40.851 7.027 0.109
Hb (g/L) 99.313 19.815 105.779 20.012 0.107
Remifentanil (mg) 0.544 0.731 0.437 0.875 -0.088
Propofol (g) 0.440 0.899 0.218 0.321 0.071
Age 82.375 7.108 76.322 7.137 0.071
PA (g/L) 95.750 50.316 131.183 57.088 -0.070
K+ (mmol/L) 3.888 0.503 3.800 0.506 0.067
Note: MIV= mean impact value, ANN= artificial neural net work, SD= standard deviation

Table 2: Classification table of training and testing cases
Sample Observed Predicted Percent Correct
0 1
Training 0 76 5 93.8%
1 1 11 91.7%
Overall Percent 82.8% 17.2% 93.5%
Testing 0 29 8 78.4%
1 2 2 50%
Overall Percent 75.6% 24.4% 75.6%