Cognitive disturbance is a defining feature of delirium. However, there is a paucity of literature characterizing the acute neurocognitive profile of delirium. Textbooks of neuropsychology either fail to describe delirium as its own syndrome or provide a cursory description of the differences between delirium and dementia (Watt, in press). Although this is an important distinction, very little guidance is provided beyond noting that attention is more impaired in delirium and memory is more impaired in dementia. There is an emerging interrelationship between delirium and Alzheimer disease, whereby each may increase the risk for the other through pathophysiological changes in brain structure and function (such as alterations of the blood–brain barrier through the pro-inflammatory cytokine release), that is largely overlooked or missed in texts and clinical training (Di Bona et al.; Fong et al., 2009). In the majority of empirical studies, a brief screening instrument is used to assess cognition in delirious patients, such as the Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975). Although the MMSE is clearly useful in the quick assessment of medical patients, it does not provide detailed information about which cognitive domains are affected. Such information has obvious utility for differential diagnosis of other conditions (i.e., delirium vs. aphasia), for making recommendations based on cognitive strengths and weaknesses and for the detection of milder cases of delirium, as even subsyndromal delirium is associated with a poorer long-term outcome (Cole, You, McCusker, Ciampi, & Belzile, 2008). Subsyndromal delirium represents a kind of borderline zone whereby patients experience milder, though still abnormal, symptoms or fewer symptoms than would be needed for a delirium proper diagnosis and highlights the important role of detailed neuropsychological assessment in tandem with thorough medical workup for possible early intervention (von Gunten & Mosimann; Watt, in press).

A small number of studies have more fully explored the cognitive changes associated with delirium. In a study comparing older medical/surgical patients with delirium to those with Alzheimer's disease and controls, Brown and colleagues (2009) found evidence of impaired visual perception (as well as general cognition) in the delirious patients. Using the cognitive test for delirium (CTD) in 100 palliative care inpatients, Meagher and colleagues (2007) found that cognition is globally reduced, with most notable impairments in attention and vigilance, comprehension, and visuospatial skills. Leonard and colleagues (2008) found the same areas of impairment using the CTD; importantly, they found that greater cognitive impairment on the CTD was more indicative of irreversable delirium (i.e., delirium preceeding death) than other “classic signs” of delirium such as psychomotor agitation and psychosis, further indicating the importance of cognitive assessment in delirious patients. Some older studies have identified language abnormalities in delirium, including confabulation and writing impairments (Chedru & Geschwind, 1972; Wallesch & Hundsalz, 1994). Fann and colleagues (2005) examined the clinical features and course of delirium in hematopoietic stem cell/bone marrow transplantation (BMT) patients, where delirium is a common complication (Beglinger et al., 2006; Fann, Roth-Roemer, Burington, Katon, & Syrjala, 2002). Patients undergoing BMT probably have multiple predisposing conditions leading to increased vulnerability for delirium, such as the use of opiates and anticancer drugs, infections and their medicinal treatments post-transplantation, and sleep deprivation during hospitalization. Fann and colleagues (2005) found three factors representing delirium symptoms: Psychosis behavior, cognition, and mood consciousness. Importantly, changes in cognition, specifically attention, memory, and working memory, began 4 days prior to the onset of delirium, increased for 7–10 days after delirium onset and then resolved.

Consistent with prior research (Fann, Roth-Roemer, Burington, Katon, & Syrjala, 2002), we found the timing of delirium onset to be, on average, almost 12 days after transplantation. This time frame suggests several possible etiological factors underlying delirium in this population. First, there are nadir effects from the conditioning therapy that patients undergo prior to BMT, like GI mucositis with medicinal interventions to counteract the mouth pain, esophageal reflux, and diarrhea. Second, cytokine release from tissue injury like tumor necrosis factor alpha and interleukin 1B may cause inflammation which predisposes the patient (or lowers their threshold) for delirium (van Gool, van de Beek, & Eikelenboom, 2010). Third, increased transfusion requirements occur during this time frame, with most transfusions performed at night, with pre-medications and interrupted sleep blocks causing sleep deprivation. Relatedly, sleep medications may last into the next day and interact or overlap causing cognitive impairment. Finally, multiple antibiotics are given, most in a prophylactic mode, that can cause symptoms of delirium (e.g., a common one, Vfend, causes visual hallucinations with some frequency). Many of the above factors may be percolating several days before delirium onset and cause the types of cognitive impairment we identified up to a week before delirium was identified. With awareness of the timing of these subtler symptoms of dysfunction, treatment staff may be able to identify and treat potentially reversible causes of delirium, such as sleep-related issues.

Previously published studies have identified specific areas of neuropsychological dysfunction associated with delirium, specifically reduced attention and working memory, visuoperceptual and language deficits (Chedru & Geschwind, 1972; Fann et al., 2005; Meagher et al., 2007; Wallesch & Hundsalz, 1994). However, the prior studies have relied on limited measures of cognitive performance, such a DRSs or mental status exams. Our results, using a neuropsychological screening battery, confirm prior findings of the importance of attention and working memory. Over the entire study period, patients with delirium were most different from healthy comparisons on Semantic Fluency, List Learning, List Recognition, and Trails A. For Trails A, List Recognition, and Fluency, comparison performance improved over time while the performance of delirium patients declined; this is especially apparent between the first two visits. We see a practice effect for the healthy comparisons on every test administered during the study. Performances in the delirium group declined between the first two visits, then remained flat or are variable. An exception is in the List Learning and Recall plots; both the BMT no delirium and delirium groups had increasing performance suggesting good recovery of immediate and delayed memory over time. In a subset of this sample, we previously found that RBANS List Learning and List Recall were two of the domains of greatest recovery by 100 days post-BMT (Beglinger et al., 2007). These findings indicate that memory is an area of early recovery both post-BMT and post-delirium. In the patients who did not experience delirium, Trails A, Trails B, Fluency, and Coding, performance was relatively stable for the first few visits before trending up at the third repeat visit. List Learning and Recall performance for these subjects continually increased from one visit to the next. These plots should be viewed cautiously, however, as the number of patients experiencing delirium differs from one visit to the next, depending on when in the study participants had delirium onset.

In addition to differing from healthy comparisons, the patients with delirium also performed below patients undergoing BMT who did not experience delirium, an important distinction in determining whether the cognitive impairments found in the delirium group are accounted for solely due to transplant- and cancer-related factors. Differences between the delirium and nondelirium groups are evident on seven of eight tasks in Fig. 1. There were statistically significant differences in the trends on the List Learning and List Recognition plots. In each case, the nondelirium group had a faster rate of increase than the delirium group (p = .013 and .013, respectively); on List Recognition, the delirium group actually declined over time. In contrast, performances on Semantic Fluency were unexpected. The delirium group outperformed the nondelirium group at four of the five visits and the comparison group at the first two visits. It is unclear why this task did not separate the two BMT groups, but it was the least impaired task in the delirium group and the only speeded task without a graphomotor component. Measures with a writing component have been sensitive in delirious patients in previous research (Chedru & Geschwind, 1972) and may be more taxing and thus better able to distinguish subtler defects compared with measures that only have an oral component. Additionally, the RBANS fluency task measures semantic fluency, which may be less sensitive to delirium group differences than a measure of phonemic fluency. Taken together, these findings converge with prior research demonstrating that attention, working memory, memory, and language are the critical features of the neuropsychological profile of delirium. We also show that the measures of psychomotor speed and learning are abnormal, two areas that have not been traditionally highlighted in the literature.

Examination of the plots in Fig. 2 representing the course of delirium provide information about which cognitive domains may be most sensitive to underlying cerebral dysfunction. At the visit in which delirium was first identified, mean scores on four tasks were abnormal: Trails B (z= −1.8), Coding (z= −1.4), Trails A (z= −1.1), and List Recall (z= −1.0). This suggests that complex attention, processing speed, visual scanning, writing, and immediate memory are deficient during delirium. The other three neuropsychological tasks were below normal (List Learning, Recognition, and Semantic Fluency all at z= −0.7). Four of the tasks showed a decline up to two visits, or 5–7 days, before patients met the threshhold for delirium: Trails B, List Recall, Coding, and Trails A. These results indicate that measures of attention, learning, psychomotor speed, and scanning may be useful indicators of impending delirium. This is consistent with the view that delirium represents multifactorial cognitive dysfunction and is not simply a disorder of attention. More recently, delirium and other disorders of impaired consciousness have been conceptualized as a network of distributed and interrelated neuropsychological processes (e.g., attention, executive functions), working as an integrated system through subcortical gating (Schiff & Plum, 2000). If one of the primary processes is abnormal, downstream effects may be observed in other neuropsychological domains. The above results suggest that these primary areas of dysfunction may be detected with the standard neuropsychological tests. Measures of memory and language, while abnormal during delirium, may not be sensitive to the prodromal phase. Two visits after delirium identification, an interesting second dip in performance occurred on several measures (Trails B, List Learning and Recall, Coding) indicating that deficits may persist or worsen in these domains. Fann and colleagues (2005) also found continuing decline in the measures of attention and working memory for 7–10 days after delirium onset in their BMT sample. These converging findings about the persistence of delirium should alert the treatment team to the need for extra monitoring in the 2 weeks following delirium identification, a time which may hold clinical significance (e.g., discharge planning).

The three delirium measures (DRS Total, DRS Severity, and MDAS) were all sensitive to impending delirium, with mild increases in scores shown at each of the three visits prior to the identified delirium, with the largest increase between the visits adjacent to delirium onset (i.e., visits −1 and 0). However, the delirium measures also showed an equal decline (i.e., “recovery”) in symptoms in the visit immediately following delrium onset (+1) and a near return to baseline by three visits post-delirium. The neuropsychological profile suggests that the delirium postdrome conferred more cognitive morbidity than these scales indicate. Semantic Fluency, Trails B, and Coding all remained substantially lower after delirium than they had in the prodromal phase, suggesting that more detailed cognitive testing may be necessary as delirium is resolving to tease out the full extent of cognitive deficits. For example, Coding remained between 1.3 and 2 SD below average at all post-delirium visits. Coding is both a quick and a complex test of multiple skills (attention, memory, writing, processing speed) making it a good candidate for delirium screening. Based on our findings, we recommend adding a version of Coding to delirium screenings and assessments for both research and clinical evaluation.

Although it was not an aim of this study to characterize neuropsychological performance in patients undergoing a BMT who did not experience delirium, our data also shed light on the cognitive changes that occur immediately after BMT. There is emerging literature on the cognitive sequelae associated with BMT. It is not surprising that patients who undergo the intensive preparative regimens associated with BMT (e.g., total body irradiation, high-dose chemotherapy) would have cognitive morbidities. Several studies have demonstrated mild impairments in BMT patients both before the transplantation and up to many years after (Meyers et al., 1994; Parth, Dunlap, Kennedy, Ordy, & Lane, 1989; Peper et al., 2000; Syrjala, Dikmen, Langer, Roth-Roemer, & Abrams, 2004). We have previously shown that BMT is associated with mild, diffuse cognitive impairments pre-transplantation, particularly on the Trail-Making Test and List Recall (Beglinger et al., 2007). The larger sample presented here confirms our earlier findings. The BMT-no delirium group remained consistently between 0.5 and 1SD below the normative average on Trails B, Coding, and Semantic Fluency throughout the whole study. Importantly, despite using the same test form, there was a lack of practice effect in the BMT group across the five test sessions—a different trajectory from the healthy comparisons indicating that these are treatment-related impairments. The implication of these results is that patients undergoing BMT may display processing speed and executive deficits even without the presence of delirium, which could be important for patient care and recommendations (e.g., patients may need assistance following complicated medical plans or may need additional time to process informed consent documents).