Risk Factor Stratification for Postoperative Delirium: A Retrospective Database Study

Susana Vacas, Tristan Grogan, Drew Cheng, Ira Hofer
doi: https://doi.org/10.1101/2022.03.06.22271982

Abstract

Background Postoperative Delirium (POD) is a disturbing reality for patients and their families. Absence of easy-to-use and accurate risk scores prompted us to retrospectively extract data from the electronic health records (EHR) to identify clinical factors associated with POD. We seek to create a multivariate nomogram to predict the risk of POD based upon the most significant clinical factors.

Methods The EHR of patients >18 years of age undergoing surgery and had POD assessment were reviewed. Patient characteristics and study variables were summarized between delirium groups. We constructed univariate logistic regression models for POD using each study variable to estimate odds ratios (OR) and constructed a multivariable logistic regression model with stepwise variable selection. In order to create a clinically useful/implementable tool we created a nomogram to predict risk of delirium.

Results Overall, we found a rate of POD of 3.7% across our study population. The Model achieved an AUC of the ROC curve of 0.83 (95% CI 0.82-0.84). We found that age, increased ASA score (ASA 3-4 OR 2.81, CI1.49-5.28, p<0.001), depression (OR 1.28, CI1.12-1.47, p<0.001), postoperative benzodiazepine use (OR 3.52, CI3.06-4.06, p<0.001) and urgent cases (Urgent OR 3.51, CI2.92-4.21, p<0.001; Emergent OR 3.99, CI3.21-4.96, p<0.001; Critically Emergent OR 5.30, CI3.53-7.96, p<0.001) were associated with POD.

Conclusions We were able to distinguish the contribution of individual risk factors to the development of POD. We created a clinically useful easy-to-use tool that has the potential to accurately identify those at high-risk of delirium, a first step to prevent POD.

Introduction

Each year approximately 52 million Americans undergo surgical procedures1, characterized by immense physiologic stress. Despite a fundamental goal of returning patients to optimal health, harmful cognitive changes, such as postoperative delirium (POD), are a common and well-established occurrence after surgery.2 POD is the most common surgical complication in older adults3 and is estimated to cost $32.9 billion per year in the US alone.4–6

POD doubles a patient’s risk of post-discharge institutionalization, enhances all-cause mortality by 10-20% for every 48 hours of delirium,7 and is linked to long-term cognitive decline, including a 8-fold increased risk of dementia.8–10 It is estimated that 30-40% of POD cases (and ultimately long-lasting deficits) are preventable.2,11,12

Currently, perioperative prevention strategies are resource and personnel intensive and general prevention for all surgical patients is not cost-effective and probably unnecessary. A more realistic solution is risk stratification, whereby patients who are at high risk for POD can be targeted for preventative measures. Recognizing which patients are at heightened risk for developing POD continues to be an active area of research. Although several risk factors have been identified to be predictors of POD, most studies are limited by small sample size and/or are targeted for specific procedures or patient groups.13,14–18 By 2014, a systemic review and metanalysis identified 37 different risk prediction models for POD, but most were not externally validated.15,19 Approximately 80% of these tools focused on cardiac and orthopaedic surgery,19,20 while some were limited to a single type of pathology.13 Unfortunately, many commonly-used POD risk stratification instruments show no superior advantage in application and performance than random chance,14,15,21 which may be indicative of the small-sized studies on which these instruments were developed.

Over the past several years, the introduction of electronic health records (EHR) has increased the feasibility of obtaining a number of data points across a large patient population. In particular, our group has demonstrated the ability to extract complex medical phenotypes from a variety of EHR data.22,23 The ability to extract these data from the EHR allows for the possibility of expanding on current work on POD risk stratification to include larger populations and a wider variety of surgical procedures.

In this manuscript we seek to leverage our expertise in complex data extraction from the EHR to retrospectively identify clinical factors associated with the risk of POD. As a secondary outcome we then seek to create a multivariate nomogram to predict the risk of POD based upon the most significant clinical factors.

Methods

Data Extraction

The University of California Los Angeles (UCLA) Institutional Review Board approved and waived the requirement for informed consent (IRB#15-000518) for this single-center retrospective observational study. All patients aged 18 years and older undergoing surgery or procedures under anesthesia between April 2013 and July 2020 at the UCLA Medical Center and affiliated surgery centers were considered eligible for inclusion. We excluded from the analysis patients aged less than 18 years old, and cases that spent less than one night in the hospital.

All data for this study were extracted from the Perioperative Data Warehouse (PDW), a custom-built robust data warehouse containing all patients who have undergone surgery at the UCLA Health since the implementation of our EHR (EPIC Systems, Madison, WI) in March 2013. We have previously described the creation of the PDW, which has a 3-stage design.24,25 Briefly, in the first stage, data are extracted from EPIC’s Clarity database into 29 tables organized around 3 distinct concepts: patients, surgical procedures, and health system encounters. These data are then cleaned and used to populate a series of 4000 distinct measures and metrics such as procedure duration, readmissions, admission International Classification of Diseases (ICD) codes, and postoperative outcomes.

Definition of Outcome Measure: Postoperative Delirium

We defined the primary outcome as POD using the Confusion Assessment Method (CAM) score.24 The CAM is a commonly used tool that identifies the presence/absence of POD and is consistently applied and documented in the EHR of surgical patients at UCLA.24 It consists of a series of yes/no questions meeting the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) criteria for delirium: acute onset and/or fluctuating course during the day of disturbance of attention, awareness or cognition not explained by another preexisting or established neurocognitive disorder.25,26 The responses to the individual metrics are stored in the EHR at the time of documentation. These data were extracted from the EHR and the individual metrics grouped at the level of the calendar day. For that calendar day, a positive response to any of the CAM questions was taken as the patient having POD on that day. For our study, a patient was considered to have POD if they had a single positive CAM score in the seven calendar days following surgery. Because not all elements were always documented contemporaneously, in the even that only some elements were charted the most recent values for the other elements were used to compute the score.25,26 Patients who were never screened for delirium were excluded from the study.

Perioperative characteristics

For the purposes of this study, preoperative patient characteristics were chosen based on current literature and included age, sex, height, weight, American Society of Anesthesiologists (ASA) status, surgical service, procedure type, inpatient versus outpatient status, elective versus emergent/critically emergent procedure, metabolic equivalents of task, SPICES score and comorbidities such as congestive heart failure, peripheral vascular disease, pulmonary history including Chronic Obstructive pulmonary Disease, kidney and endocrine disease, obstructive sleep apnea, neuropsychiatric history including anxiety and depression, history of tobacco or alcohol use, as well as imaging (echocardiogram) and laboratory values (glomerular filtration rate).

Intraoperative characteristics included type of anesthetic, duration of anesthesia, administration of benzodiazepines and intraoperative hemodynamic monitoring data, such as time spent with mean arterial pressure < 55 or <65 mmHg.

Statistical Analysis

Patient characteristics and study variables were summarized between delirium groups using frequency (%) or mean (SD) and compared between groups using the chi-square test or t-test as appropriate. We then constructed univariate logistic regression models for delirium using each study variable to estimate odds ratios (OR) with 95% confidence intervals. Next, we constructed a multivariable logistic regression model with stepwise variable selection to get a parsimonious model. Finally, in order to create a more clinically useful/implementable tool we created a nomogram from this model to predict risk of delirium both visually and as a point system score. The overall discrimination/prognostic ability of the model was assessed using the area under the ROC curve (AUC). P-values <0.05 were considered statistically significant and statistical analyses were performed using IBM SPSS V26 (Armonk, NY) and R V4.1.0 (www.r-project.org, Vienna, AU).

Results

Patient Demographics and Univariate Analysis

Table 1 shows the overall demographics of the cohort and the rates of the covariates studied. We identified 32,734 patients who met inclusion criteria. The overall incidence of POD was 3.7%. The cohort consisted of 7,315 (22.3%) patients aged 65-75 years and 5,401 (16.5%) patients age >75 years old. The majority (20,199 (62.7%)) of patients were ASA physical status class 3 or 4 and the most common surgical services were Orthopedics (7,200), General Surgery (6,126), and Neurosurgery (3,345). Most cases (19,950 (60.9%)) were elective.

View this table:
Table 1.

Overall cohort demographics and univariate analysis of risk factors and delirium. P-values listed are those of the logistic regression for individual risk factors. Values presented as frequency (%) unless otherwise noted; *values presented as mean / SD

Overall, in univariate analysis a wide variety of factors were directly associated with POD. Of note, patient age (45-65 years old, OR 2.18, 95% CI 1.74-2.74, p <0.001; 65-75 years old, OR 2.60, 95% CI 2.05-3.30, p <0.001; >75 years old, OR 8.07, 95% CI 6.49-10.05, p <0.001), ASA physical status (ASA 3-4 OR 6.40, 95% CI 5.22-7.85, p <0.001), female sex, surgical service (Cardiac surgery OR 7.01, 95% CI 3.16-15.54, p <0.001; Gastroenterology OR 12.56, 95% CI 5.77-27.35, p <0.001; General Surgery OR 5.29, 95% CI 2.48-11.32, p <0.001; Liver Transplant OR 18.60, 95% CI 8.50-40.69, p <0.001; Neurosurgery OR 13.62, 95% CI 6.40-29.01, p <0.001; Orthopaedics OR 11.59, 95% CI 5.47-24.52, p <0.001; Otolaryngology OR 3.58, 95% CI 1.60-8.01, p =0.002; Plastic Surgery OR 3.88, 95% CI 1.61-9.32, p =0.002; Surgical Oncology OR 5.63, 95% CI 2.32-13.64, p <0.001; Thoracic Surgery OR 6.66, 95% CI 2.88-15.41, p <0.001; Urology OR 3.82, 95% CI 1.75-8.34, p <0.001; Vascular Surgery OR 8.51, 95% CI3.72-19.47, p <0.001) and case urgency (Urgent OR 5.40, 95% CI 4.61-6.33, p <0.001; Emergent OR 5.28, 95% CI 4.36-6.38, p <0.001; Critically Emergent OR 7.60, 95% CI 5.35-10.80, p <0.001; Inpatient OR 6.05, 95% CI 5.10-7.18, p <0.001; Transplant OR 2.02, 95% CI 1.43-2.85, p <0.001) were associated with POD. In addition, BMI<18.5 (OR 2.16, 95% CI 1.75-2.66, p <0.001), BMI 30-39.99 (OR 0.71, 95% CI 0.61-0.83, p <0.001), BMI >40 (OR 0.42, 95% CI 0.28-0.62, p <0.001), depression (OR 1.45, 95% CI 1.28-1.64, p <0.001) and postoperative benzodiazepine use (OR 4.08, 95% CI 3.61-4.60, p <0.001) were associated with POD; however, premedication with benzodiazepines was not associated with POD, nor was anesthesia type. Further variables study can be found in Table 1.

Multivariate Model

Table 2 shows the odds ratios for the features included in the multivariate analysis of POD. Overall, the Model achieved an AUC of the ROC curve of 0.83 (95% CI 0.82-0.84). Age above 45 was consistently associated with POD, with age greater than 75 years having a OR 8.26, 95% CI 6.39-10.68 of POD. In addition, increased ASA physical status score (ASA 3-4 OR 2.81, 95% CI 1.49-5.28, p <0.001), history of depression (OR 1.28, 95% CI 1.12-1.47, p <0.001), postoperative benzodiazepine use (OR 3.52, 95% CI 3.06-4.06, p <0.001) and more urgent cases (Urgent OR 3.51, 95% CI 2.92-4.21, p <0.001; Emergent OR 3.99, 95% CI 3.21-4.96, p <0.001; Critically Emergent OR 5.30, 95% CI 3.53-7.96, p <0.001) were associated with POD. Of interest premedication with benzodiazepines was protective of POD.

View this table:
Table 2.

Multivariate odds ratios for postoperative delirium and p-values for each category

Creation of a score for POD risk stratification

In order to facilitate prospective risk stratification Figure 1 contains a nomogram that can be used to convert the odds ratios from the multivariate model into a “score” for POD prediction. The points associated with the various sections of the nomogram can be found in supplementary table 1.

Figure 1.
Figure 1.

Nomogram Point to predict Individual risk for Postoperative Delirium (service legend a) Obstetrics and Gynecology, b) Gastroenterology, c) cardiac Surgery, d) Surgical Oncology, e) Otolaryngology, f) Other, g) Urology, h) General Surgery, i) Vascular Surgery, j) Thoracic Surgery, k) Plastic Surgery, l) Orthopaedics, m) Liver Transplant, n) Neurosurgery)

Discussion

This is (to our knowledge) the largest retrospective cohort analysis of patients studied for POD to date. We examined nearly 33,000 patients aged >18 years old with an overall incidence of POD of 3.7%. The final multivariate model, containing nine features achieved an AUC of 0.85 (0.8-0.86) and was thus highly discriminating for POD. While many of our findings are similar to those of previous work in smaller studies on more focused populations, we report several differences. Like others, we found that POD was highly associated with advancing age, increased patient co-morbidities (ASA physical status), male gender, depression, high BMI, emergent cases, and invasive surgical procedures. However, we found no differences regarding type of anesthesia or if patients where premedicated with midazolam. The surgical services most likely where patients were more likely to develop POD where orthopaedics, liver transplant and neurosurgery.

Preoperative identification of high-risk patients for postoperative delirium is essential to allocate time and resources to implement targeted preventive strategies. As noted, this manuscript is not the first to perform a retrospective analysis of risk factors for POD.27–29 Medical literature contains many risk scores designs, each attempting to distill the likelihood of one of a variety of different outcomes. 13,14–18 However, to be easily calculated, were based on a limited number of highly predictive features. These scores evaluate disease status in broad strokes30–33 and often lack precision at the patient level – making them better suited to risk adjustment for populations than individualized risk prediction.34 Most of these studies were on smaller populations or focused on specific types of surgeries, or patient groups. For example, one important risk score tool for POD based on a large population (approximately 10,000 patients), recently developed by Whitlock et al14 accounts for precipitating factors solely by classifying surgical risk as moderate or high, and their risk score is unweighted and additive (i.e. summing the points for each risk factor), while clinical risk comes from the complex interplay of multiple factors. Our study differs in several ways. Firstly, rather than providing broad surgical groupings (low, medium and high risk) we stratified by surgical services while incorporating urgency as its own feature. Secondly, by directly harnessing EHR data we were able to incorporate a wider range of clinical co-morbidities hypothesized to be associated with POD including depression and BMI. Of note, the ability to increase the amount of data included in our model resulted in an AUC substantially higher than those reported previously. This gives further credence for the use of advanced modeling techniques, such as machine learning, that has been used to predict other conditions.35–38

In developing our model, we attempted to extract a wide variety of relevant data from the EHR to include as potential co-variates. Nonetheless, not all information was necessarily readily available. Important predictors, such as detailed information on cognition and sensory testing was not available. Further, previous studies (including several by our group) have repeatedly shown that EHR data is sparely populated and thus potentially inaccurate.23 We attempted to account for this by leveraging various types of EHR data to make more complete clinical phenotypes, however for patients less known to our system we may not have been sufficiently sensitive.

As important as the features that were associated with POD, are findings of features that were not associated with POD. In particular the use of midazolam as a preoperative anxiolytic has been the topic of much discussion for several years. In our cohort, this was actually protective of POD. However, it is critical to call out the limits of associations in a retrospective analysis such as this one. Given the awareness of theoretically associations between preoperative benzodiazepine use and POD, it is possible that there was significant selection bias in which patients were given midazolam preoperatively thereby affecting the results. Thus, these results must be interpreted carefully given the inability to address all possible confounders.

Importantly, we only included those patients who were screened for POD at least once during their hospitalization. Although, all patients admitted to the hospital should undergo delirium screening, it does not always happen. It is highly probable that patients at higher risk of delirium were screened more frequently introducing a selection bias to our cohort. In addition, previous work has shown that hyperactive delirium is more likely to be recognized by providers than hypoactive delirium. Our healthcare system has trained staff to recognize signs and symptoms of delirium. At least, for those patients that developed hyperactive delirium, it is highly likely that this information was introduced in the chart. Nonetheless, the exact rates reported in this manuscript may not be completely accurate and it is possible that the exact risk calculated based on our nomogram is not completely generalizable to patients who were not screened for POD. However, these broader trends and associations are likely correct and thus informative in creating a risk score.

POD is a multifactorial problem and until real-time clinical implementation of machine learning models become accurate and feasible, POD prediction that relies on weighted risk factors with an easy-to-use tool has the tremendous potential to improve predictive capacity and deployment of targeted interventions.

Data Availability

All data produced in the present work are contained in the manuscript

Authors’ contributions

SV: This author helped in study conception and design, data acquisition, data validation and interpretation and with manuscript drafting and revision. The author has approved the final version of the manuscript.

DC: This author helped in study design, data acquisition and data validation. The author has approved the final version of the manuscript.

TG: This author helped as methodology advisor, with statistical analyses, and with manuscript drafting and revision. The author has approved the final version of the manuscript.

IH: This author helped in study conception and design, data acquisition, data validation and interpretation, and with manuscript drafting and revision. The author has approved the final version of the manuscript.

Declaration of interests

Ira Hofer is the founder and President of Extrico health a company that helps hospitals leverage data from their electronic health record for decision making purposes. Dr. Hofer receives research support and serves as a consultant for Merck.

Funding

This work was supported by the National Institutes of Health, National Institute of General Medical Sciences under Award Number K23GM132795 (Vacas), National Institutes of Aging under award number R21AG070269 (Vacas) and National Heart, Lung and Blood Institute under Award Number K01HL150318 (Hofer). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health

Acknowledgements

Emily Walters and Eilon Gabel for their assistance in data acquisition.

References

  1. 1.↵
    Centers for Disease Control and Prevention - National Hospital Discharge Survey, 2010 - Number of all-listed procedures for discharges from short-stay hospitals, by procedure category and age: United States. Accessed February 2020,
  2. 2.↵
    Inouye SK, Westendorp RGJ, Saczynski JS. Delirium in elderly people. Lancet (London, England). 2014;383:911–22. doi:10.1016/S0140-6736(13)60688-1
    OpenUrlCrossRefPubMedWeb of Science
  3. 3.↵
    Postoperative delirium in older adults: best practice statement from the American Geriatrics Society. J Am Coll Surg. Feb 2015;220(2):136–48.e1. doi:10.1016/j.jamcollsurg.2014.10.019
    OpenUrlCrossRefPubMed
  4. 4.↵
    Leslie DL, Inouye SK. The importance of delirium: economic and societal costs. Journal of the American Geriatrics Society. Nov 2011;59 Suppl 2:S241–3. doi:10.1111/j.1532-5415.2011.03671.x
    OpenUrlCrossRefPubMed
  5. 5.
    Leslie DL, Marcantonio ER, Zhang Y, Leo-Summers L, Inouye SK. One-year health care costs associated with delirium in the elderly population. Archives of internal medicine. Jan 14 2008;168(1):27–32.
    OpenUrlCrossRefPubMedWeb of Science
  6. 6.↵
    Gou RY, Hshieh TT, Marcantonio ER, et al. One-Year Medicare Costs Associated With Delirium in Older Patients Undergoing Major Elective Surgery. JAMA Surg. May 1 2021;156(5):430–442. doi:10.1001/jamasurg.2020.7260
    OpenUrlCrossRef
  7. 7.↵
    Whitlock EL, Vannucci A, Avidan MS. Postoperative delirium. Minerva anestesiologica. Apr 2011;77(4):448–56.
    OpenUrlPubMedWeb of Science
  8. 8.↵
    Davis DH, Muniz Terrera G, Keage H, et al. Delirium is a strong risk factor for dementia in the oldest-old: a population-based cohort study. Brain : a journal of neurology. Sep 2012;135(Pt 9):2809–16. doi:10.1093/brain/aws190
    OpenUrlCrossRefPubMedWeb of Science
  9. 9.
    Aranake-Chrisinger A, Avidan MS. Postoperative delirium portends descent to dementia. British journal of anaesthesia. 2017;119(2):285–288. doi:10.1093/bja/aex126
    OpenUrlCrossRef
  10. 10.↵
    Sprung J, Roberts RO, Weingarten TN, et al. Postoperative delirium in elderly patients is associated with subsequent cognitive impairment. British journal of anaesthesia. Aug 1 2017;119(2):316–323. doi:10.1093/bja/aex130
    OpenUrlCrossRef
  11. 11.↵
    Inouye SK, Bogardus ST, Jr.., Charpentier PA, et al. A multicomponent intervention to prevent delirium in hospitalized older patients. The New England journal of medicine. Mar 4 1999;340(9):669–76. doi:10.1056/nejm199903043400901
    OpenUrlCrossRefPubMedWeb of Science
  12. 12.↵
    Marcantonio ER, Flacker JM, Wright RJ, Resnick NM. Reducing delirium after hip fracture: a randomized trial. Journal of the American Geriatrics Society. May 2001;49(5):516–22. doi:10.1046/j.1532-5415.2001.49108.x
    OpenUrlCrossRefPubMedWeb of Science
  13. 13.↵
    Choi NY, Kim EH, Baek CH, Sohn I, Yeon S, Chung MK. Development of a nomogram for predicting the probability of postoperative delirium in patients undergoing free flap reconstruction for head and neck cancer. Eur J Surg Oncol. Apr 2017;43(4):683–688. doi:10.1016/j.ejso.2016.09.018
    OpenUrlCrossRef
  14. 14.↵
    Whitlock EL, Braehler MR, Kaplan JA, et al. Derivation, Validation, Sustained Performance, and Clinical Impact of an Electronic Medical Record-Based Perioperative Delirium Risk Stratification Tool. Anesth Analg. Dec 2020;131(6):1901–1910. doi:10.1213/ANE.0000000000005085
    OpenUrlCrossRef
  15. 15.↵
    Jansen CJ, Absalom AR, de Bock GH, van Leeuwen BL, Izaks GJ. Performance and agreement of risk stratification instruments for postoperative delirium in persons aged 50 years or older. PloS one. 2014;9(12):e113946. doi:10.1371/journal.pone.0113946
    OpenUrlCrossRefPubMed
  16. 16.
    Marcantonio ER, Goldman L, Mangione CM, et al. A clinical prediction rule for delirium after elective noncardiac surgery. Jama. 1994;271:134–139.
    OpenUrlCrossRefPubMedWeb of Science
  17. 17.
    Litaker D, Locala J, Franco K, Bronson DL, Tannous Z. Preoperative risk factors for postoperative delirium. Gen Hosp Psych. 2001;23:84–87.
    OpenUrl
  18. 18.↵
    Kim EM, Li G, Kim M. Development of a Risk Score to Predict Postoperative Delirium in Patients With Hip Fracture. Anesthesia and analgesia. Jan 2020;130(1):79–86. doi:10.1213/ane.0000000000004386
    OpenUrlCrossRef
  19. 19.↵
    van Meenen LC, van Meenen DM, de Rooij SE, ter Riet G. Risk prediction models for postoperative delirium: a systematic review and meta-analysis. Journal of the American Geriatrics Society. Dec 2014;62(12):2383–90. doi:10.1111/jgs.13138
    OpenUrlCrossRefPubMed
  20. 20.↵
    Rudolph JL, Jones RN, Levkoff SE, et al. Derivation and validation of a preoperative prediction rule for delirium after cardiac surgery. Circulation. Jan 20 2009;119(2):229–36. doi:10.1161/circulationaha.108.795260
    OpenUrlAbstract/FREE Full Text
  21. 21.↵
    Lindroth H, Bratzke L, Purvis S, et al. Systematic review of prediction models for delirium in the older adult inpatient. BMJ Open. 2018;8(4):e019223. doi:10.1136/bmjopen-2017-019223
    OpenUrlAbstract/FREE Full Text
  22. 22.↵
    Hofer IS, Cheng D, Grogan T. A Retrospective Analysis Demonstrates That a Failure to Document Key Comorbid Diseases in the Anesthesia Preoperative Evaluation Associates With Increased Length of Stay and Mortality. Anesth Analg. Jan 15 2021;doi:10.1213/ANE.0000000000005393
    OpenUrlCrossRef
  23. 23.↵
    Hofer IS, Cheng D, Grogan T, et al. Automated Assessment of Existing Patient’s Revised Cardiac Risk Index Using Algorithmic Software. Anesthesia and analgesia. May 2019;128(5):909–916. doi:10.1213/ANE.0000000000003440
    OpenUrlCrossRef
  24. 24.↵
    Inouye SK, van Dyke CH, Alessi CA, Balkin S, Siegal AP, Horwitz RI. Clarifying confusion: the confusion assessment method. Annals of internal medicine. 1990;113:941–948.
    OpenUrlCrossRefPubMedWeb of Science
  25. 25.↵
    Evered L, Silbert B, Knopman DS, et al. Recommendations for the Nomenclature of Cognitive Change Associated With Anaesthesia and Surgery-2018. Anesthesia Analgesia. Nov 2018;127(5):1189–1195. doi:10.1213/ane.0000000000003634
    OpenUrlCrossRef
  26. 26.↵
    Association AP. American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition: DSM-5. American Psychiatric Association; 2013.
  27. 27.↵
    Ansaloni L, Catena F, Chattat R, et al. Risk factors and incidence of postoperative delirium in elderly patients after elective and emergency surgery. British Journal of Surgery. 2010;97(2):273–280. doi:10.1002/bjs.6843
    OpenUrlCrossRefPubMed
  28. 28.
    Benoit AG, Campbell MI, Tanner JR, et al. Risk factors and prevalence of perioperative cognitive dysfunction in abdominal ancurysm patients. J Vasc Surg. 2005;42:884–890. doi:10.1016/j.jvs.2005.07.032
    OpenUrlCrossRefPubMedWeb of Science
  29. 29.↵
    Moller JT, Cluitmans P, Rasmussen LS, et al. Long-term postoperative cognitive dysfunction in the elderly ISPOCD1 study. ISPOCD investigators. International Study of Post-Operative Cognitive Dysfunction. Lancet. Mar 21 1998;351(9106):857–61. doi:S0140673697073820 [pii]
    OpenUrlCrossRefPubMedWeb of Science
  30. 30.↵
    Le Manach Y, Collins G, Rodseth R, et al. Preoperative Score to Predict Postoperative Mortality (POSPOM): Derivation and Validation. Anesthesiology. Mar 2016;124(3):570–9. doi:10.1097/ALN.0000000000000972
    OpenUrlCrossRefPubMed
  31. 31.
    Sobol JB, Wunsch H. Triage of high-risk surgical patients for intensive care. Review. Crit Care. 2011;15(2):217. doi:10.1186/cc9999
    OpenUrlCrossRefPubMed
  32. 32.
    Vincent JL, Moreno R, Takala J, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med. Jul 1996;22(7):707–10. doi:10.1007/bf01709751
    OpenUrlCrossRefPubMedWeb of Science
  33. 33.↵
    Knaus WA, Wagner DP, Draper EA, et al. The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults. Chest. Dec 1991;100(6):1619–36. doi:10.1378/chest.100.6.1619
    OpenUrlCrossRefPubMedWeb of Science
  34. 34.↵
    Hofer IS, Halperin E, Cannesson M. Opening the Black Box: Understanding the Science Behind Big Data and Predictive Analytics. Anesthesia and analgesia. Nov 2018;127(5):1139–1143. doi:10.1213/ANE.0000000000003463
    OpenUrlCrossRef
  35. 35.↵
    Hill BL, Brown R, Gabel E, et al. An automated machine learning-based model predicts postoperative mortality using readily-extractable preoperative electronic health record data. British journal of anaesthesia. Dec 2019;123(6):877–886. doi:10.1016/j.bja.2019.07.030
    OpenUrlCrossRef
  36. 36.
    Hofer IS, Burns M, Kendale S, Wanderer JP. Realistically Integrating Machine Learning Into Clinical Practice: A Road Map of Opportunities, Challenges, and a Potential Future. Anesth Analg. May 2020;130(5):1115–1118. doi:10.1213/ANE.0000000000004575
    OpenUrlCrossRef
  37. 37.
    Hofer IS, Lee C, Gabel E, Baldi P, Cannesson M. Development and validation of a deep neural network model to predict postoperative mortality, acute kidney injury, and reintubation using a single feature set. NPJ Digit Med. 2020;3:58. doi:10.1038/s41746-020-0248-0
    OpenUrlCrossRef
  38. 38.↵
    Lee CK, Hofer I, Gabel E, Baldi P, Cannesson M. Development and Validation of a Deep Neural Network Model for Prediction of Postoperative In-hospital Mortality. Anesthesiology. Oct 2018;129(4):649–662. doi:10.1097/ALN.0000000000002186
    OpenUrlCrossRefPubMed
Back to top
Posted March 08, 2022.
Download PDF
Data/Code
Email
Risk Factor Stratification for Postoperative Delirium: A Retrospective Database Study
Susana Vacas, Tristan Grogan, Drew Cheng, Ira Hofer
medRxiv 2022.03.06.22271982; doi: https://doi.org/10.1101/2022.03.06.22271982
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area