SPECIAL ARTICLE Advancing the Neurophysiological Understanding of Delirium †‡ †‡§ Mouhsin M. Shafi, MD, PhD,* Emiliano Santarnecchi, PhD,* Tamara G. Fong, MD, PhD, || † Richard N. Jones, ScD, # Edward R. Marcantonio, MD, SM, ** Alvaro Pascual-Leone, †‡ †§ MD, PhD,* and Sharon K. Inouye, MD, MPH ** annual costs of $164 billion in the United States.1,2 Delir- Delirium is a common problem associated with substantial ium is associated with cognitive decline, loss of functional morbidity and increased mortality. However, the brain independence, and increased mortality.1 Therefore, identi- dysfunction that leads some individuals to develop delir- fying individuals at risk for developing delirium, differenti- ium in response to stressors is unclear. In this article, we ating episodes of delirium likely to lead to lasting briefly review the neurophysiologic literature characteriz- deleterious consequences, and minimizing the impact of ing the changes in brain function that occur in delirium, delirium have assumed heightened importance. One central and in other cognitive disorders such as Alzheimer’s dis- limitation preventing the achievement of these clinical ease. Based on this literature, we propose a conceptual goals is that current clinical definitions, such as the widely model for delirium. We propose that delirium results from used Confusion Assessment Method (CAM3), are based on a breakdown of brain function in individuals with impair- cognitive tests and bedside observations, and it is not com- ments in brain connectivity and brain plasticity exposed to pletely clear how these signs and symptoms relate to the a stressor. The validity of this conceptual model can be underlying brain dysfunction. Thus, a critical step is to tested using Transcranial Magnetic Stimulation in combi- understand, at a neurophysiologic level, why delirium hap- nation with Electroencephalography, and, if accurate, pens in specific individuals. To this end, we begin by could lead to the development of biomarkers for delirium briefly reviewing the current literature on neurophysiologic risk in individual patients. This model could also be used investigations of delirium and other conditions that affect to guide interventions to decrease the risk of cerebral dys- cognition, such as Alzheimer’s disease (AD). We present a function in patients preoperatively, and facilitate recovery conceptual model advancing the hypothesis that delirium in patients during or after an episode of delirium. JAm is due to a breakdown of normal brain function reflecting Geriatr Soc 2017. impairments in brain connectivity and plasticity. We explain how this model can be tested using the combina- Key words: delirium; electroencephalography; transcra- tion of Transcranial Magnetic Stimulation (TMS) and elec- nial magnetic stimulation; connectivity; plasticity troencephalograpy (EEG), and discuss some of the clinical implications of the model. Neurophysiological Investigations of Delirium to Date Neurophysiology during delirium has traditionally been studied using EEG, which measures the electrical fields elirium is a common and costly problem, affecting up produced by synchronized synaptic activity of cortical neu- to 50% of hospitalized older adults, with estimated D rons. EEG activity is often divided into different spectral frequency bands, and changes in spectral band power (sig- From the *Berenson-Allen Center for Noninvasive Brain Stimulation, Beth † ‡ nal strength) have been reported in different disease states. Israel Deaconess Medical Center; Harvard Medical School; Department § of Neurology, Beth Israel Deaconess Medical Center; Aging Brain Center, The traditional spectral frequency bands evaluated during Institute for Aging Research, Hebrew SeniorLife, Boston, Massachusetts; EEG recordings include delta (1–4 Hz), theta (4–8 Hz), || Department of Psychiatry and Human Behavior, Brown University alpha (8–13 Hz) and beta (13–30 Hz) bands; alpha activ- # Warren Alpert Medical School; Department of Neurology, Brown ity is the most prominent rhythm in the resting awake University Warren Alpert Medical School, Providence, Rhode Island; and **Department of Medicine, Beth Israel Deaconess Medical Center, state (eyes-closed). More recently, techniques have been Boston, Massachusetts. developed to assess statistical correlations in the EEG sig- Address correspondence to Mouhsin M. Shafi, Department of Neurology, nals recorded from different electrodes, that indicate con- Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, nectivity between different brain regions. Typically, these MA 02215. E-mail: mshafi@bidmc.harvard.edu measures fall into two broad categories: measures of func- DOI: 10.1111/jgs.14748 tional connectivity, which identify correlations in the JAGS 2017 © 2017, Copyright the Authors Journal compilation © 2017, The American Geriatrics Society 0002-8614/17/$15.00 2 SHAFI ET AL. 2017 JAGS statistical properties of brain signals from two or more A Conceptual Neurophysiological Model of Delirium regions, or measures of effective connectivity, which attempt to identify causal interactions between regions. Based on the above results, we propose a conceptual Changes in EEG spectral power and connectivity have model (Table 1; Figure 1) that delirium is the consequence been identified in patients with diseases that affect cognition of the breakdown in brain network dynamics induced by such as Mild Cognitive Impairment (MCI) and AD. In these insults or stressors in individuals with baseline low brain conditions, for example, loss of frontoparietal EEG connec- resilience due to low connectivity and/or deficient mecha- tivity is correlated with cognitive test results and disease nisms of neuroplasticity, such as may be present in AD. progression over time.4 Extending this work in MCI and Neuroplasticity, defined as the brain’s ability to reorganize AD, EEG measures may be useful in characterizing cerebral itself by forming new neural connections throughout life, dysfunction in patients at risk for delirium. EEG is also use- allows the brain to compensate for injury and disease, and ful in understanding the physiologic changes that occur dur- is often considered necessary for neurologic resilience (the ing delirium, when the most consistent neurophysiological ability to accommodate to or recover from a stressor).14 abnormality is a relative slowing of resting-state EEG Relevant brain stressors can include major surgery, general rhythms, with abnormally decreased background alpha anesthesia, systemic inflammation, infections, and psy- power and increased theta- and delta-frequency activity.5,6 chotropic drugs. Health conditions that might result in The degree of EEG slowing in delirium correlates with impaired plasticity include pre-existing neurodegenerative decline in performance on cognitive tests,6 and both EEG disorders, such as MCI and AD,15 and comorbid condi- slowing and cognitive dysfunction normalize when meta- tions such as diabetes16 or renal impairment. This model bolic derangements leading to delirium (e.g., hypoxemia, predicts that when individuals are confronted with acute hypoglycemia) resolve.5 More recently, increased spectral insults, these stressors will alter brain connectivity (e.g., variability,7 decreased complexity of EEG activity,7 and within the dorsal attention network17) and/or brain net- decreased EEG connectivity in the alpha band8 have also work dynamics (e.g., the relationship between dorsolateral been reported during delirium. EEG changes can differenti- prefrontal cortex activity and posterior cingulate activ- ate individuals with delirium from those without delirium ity12), resulting in symptoms of delirium such as inatten- with an estimated sensitivity of 83.3% and specificity of tion. This impact will be greater in individuals with 77.8% for visual analysis of EEG features,9 up to a sensitiv- pre-existing deficits in brain connectivity, specifically in the ity of 100% and specificity of 96% for a quantitative mea- brain networks involved in resilience—which are linked to sure of EEG spectral power.10 Notably, EEG features may the construct of cognitive reserve.18 Such alterations in help differentiate patients with delirium and dementia from brain connectivity and dynamics will be inadequately com- those with dementia alone with up to 83% accuracy.11 pensated in a brain with impaired plasticity, manifesting as However, in these studies the specific measures were retro- the clinical syndrome of delirium. Supporting this model is spectively identified. Prospective studies validating these the finding of altered connectivity and impaired plasticity measures against reference standard delirium ratings are in AD, which is established as a major risk factor for delir- needed. In current clinical practice, EEG is used to distin- ium.1 This validity of this conceptual model can be guish delirium from nonconvulsive status epilepticus or an assessed using the combination of Transcranial Magnetic underlying psychiatric condition. Stimulation (TMS) and EEG. Another method useful for analysis of functional con- nectivity is resting-state functional magnetic resonance Testing the Conceptual Model of Delirium with TMS- imaging (rs-fMRI). In this method, brain activity is mea- EEG sured while the subject sits in the MRI in a resting state (not doing any tasks, in contrast to active functional neu- EEG and functional MRI passively record brain activity, roimaging); different brain regions that show correlated and therefore are limited in their capacity for