Delirium is a complex neuropsychiatric syndrome that is common across healthcare settings, occurring in 29% to 64% of medical in-patients [1],[2] with even higher rates among patients in intensive and palliative care settings [3]. It is independently associated with a range of adverse outcomes that include elevated risk of dementia and mortality [4],[5]. However, delirium is often misdiagnosed and under detected in real-world practice [6]-[8] such that clear and concise diagnostic criteria are fundamental to improving detection and management.

The advent of clear diagnostic criteria for delirium in the Diagnostic and Statistical Manual third edition (DSM-III) and subsequent Diagnostic and Statistical Manual third edition revised (DSM-IIIR) and the Diagnostic and Statistical Manual fourth edition (DSM-IV) [9]-[11] versions has supported considerable growth in research activity in the field of delirium [12]. The DSM-IV criteria provide a highly inclusive description of delirium that has become the preferred diagnostic criteria for both clinicians and researchers [13]. However, the essential criteria have been progressively abbreviated [14] and studies indicate considerable disparity in delirium detection when applying these different DSM versions and the International Classification of Diseases - Tenth Edition (ICD-10) [15]-[19].

Moreover, although the gold standard for delirium identification is considered a clinical diagnosis according to DSM-IV criteria, there is widespread recognition among `deliriumologists’ that there is no consensus as to how this should actually be determined in practice. Against this background, and considering the new DSM-5 criteria, we explored how the criteria can be applied when the individual elements are assessed in a systematic and operationalized way, thus allowing for various interpretations of the new criteria to be examined. In order to explore how DSM-5 criteria might differ from those of DSM-IV, we examined a pooled dataset derived from previous prospective phenomenological research exploring delirium in a variety of clinical populations and research sites using standardized assessments.

The aims of this study were: (1) to analyze retrospectively the pooled database to compare features of delirium cohorts identified by the originally applied DSM-IV criteria (identified by the gold standard of a detailed clinical assessment) as well as those identified by applying scores from the Delirium Rating Scale-Revised-98 (DRS-R98) [21] items relevant to criteria for a post hoc application of strict and relaxed interpretations of DSM-5 criteria; (2) to examine whether different interpretations of DSM-5 criteria impact significantly upon delirium identification rates by exploring concordance across these three groups. We thus sought to understand to what degree DSM-IV and DSM-5 were concordant and how the DSM-5 criteria can be best applied to allow for inclusiveness and that the gap between rates of diagnosis by the different systems is not excessively wide; and (3) to examine how DSM-IV and the different interpretations of DSM-5 differ in terms of delirium phenomenology.

The principal purpose of these analyses was to explore the concordance/discordance between DSM-IV and the new DSM-5 criteria for delirium, especially given the differences in wording that could be interpreted differently by different DSM-5 users, thereby leading to potentially widely disparate diagnostic rates with serious implications for patient care and research. We used proxy symptom information captured systematically using a standardized administration of the DRS-R98 to retrospectively generate delirium diagnoses according to two different interpretations of DSM-5 criteria. We then compared the level of concordance between these criteria with each other and with prospectively-determined DSM-IV delirium identified through detailed and comprehensive clinical assessment. This extrapolation was possible because we first tested the proxy method for DSM-IV and found high concordance, supporting the post hoc method as valid. We found that a more relaxed interpretation of DSM-5 criteria detected most of the cases of DSM-IV defined delirium, while a strict interpretation excluded more than two-thirds of such cases. There was thus considerable overlap for the actual cases designated as delirious by DSM-IV and the relaxed DSM-5 approach, suggesting that delirium diagnosis will be minimally changed by the use of DSM-5 criteria if the presence of impaired attention is interpreted as sufficient for criterion A, and either acute/subacute onset or fluctuation symptom course is used to interpret criterion B.

Studies exploring the concordance between DSM IIIR, DSM-IV and ICD-10 [32] diagnostic systems in elderly hospitalized patients indicate that DSM-IV provides a highly inclusive definition of delirium that includes a substantial number of patients who do not have delirium as defined by DSM-IIIR or ICD-10 [15]-[19]. The inclusion of disorganized thinking as a criterion in DSM-IIIR [33] and the wide range of mandatory features in ICD-10 are key factors that underpin this finding. Other work suggests that DSM-IV includes many cases that are of subsyndromal severity according to DRS-R98 cut-off scores, although the distinction between subsyndromal versus `mild’ delirium can be challenging, especially when dimensional approaches are employed [24],[34]. Subsyndromal delirium is addressed under `attenuated delirium syndrome’ in DSM-5.

Our work examines how different interpretations can strongly influence diagnosis rates among a mixed neuropsychiatric population who have undergone detailed and highly consistent assessment for the symptoms that characterize delirium. The concordance between DSM-IV-defined cases and DSM-5 delirium in this dataset varied considerably (30% to 89%) according to the interpretation of DSM-5 criteria (strict versus relaxed), especially in relation to the requirements for acute onset, fluctuating course and role of disorientation. In the absence of convincing evidence to support a major change in the concept of delirium, as well as the desire to maintain the generalizability of existing literature to delirium knowledge, our data suggest that the more relaxed interpretation of the DSM-5 criteria should be used. The relaxed interpretation has considerable overlap with DSM-IV delirium, while also adding the benefit of additional precision to the definition that might allow for more focused research efforts. Importantly, this means ignoring the parenthetical notation to require simple disorientation to the environment in Criterion A, which potentially is duplicative of the `disorientation’ mentioned in Criterion C and inadequate from a phenomenological perspective to provide a surrogate for the complexity and breadth of what likely constitutes awareness of normal consciousness. The strict DSM-5 interpretation was clearly too restrictive when disorientation was also required in Criterion A, and too few patients were diagnosed as delirious. Conversely, another reason to favor the relaxed approach is that DSM-IV patients who were excluded by the relaxed DSM-5 criteria had significantly lower mean DRS-R98 scores below the DRS-R98 diagnostic cutoff score for delirium diagnosis.

The anchoring of awareness to simple testing of orientation in the Criterion A parenthesis is a significant departure from prior DSM versions and ICD, in which inattention is cardinal and an alteration of consciousness is left to the judgment of the observer who can incorporate a more full definition of awareness throughout the bedside interview. `Awareness’ presents a difficult concept to test in objective terms as it relates to the ability not only to accurately perceive and assimilate one’s surroundings but also to the appreciation and understanding of self. Recent work using functional magnetic resonance imaging (fMRI) in delirium [35] found alterations in the resting state default mode neural network, which reflects the quiet internal thinking mode. This reveals that the brain in a delirious person is not working normally and that, specifically, the internal thinking state - not just orientation to the external world - is impaired. Empirically, this is consistent with delirious patients’ thought process and comprehension impairments regarding self, others and situations. We strongly recommend that the guidance notes for DSM-5 advising impaired awareness is `manifested by a reduced orientation to the environment’ not be followed in a strict sense. In addition, the use of `and’ means the identification of disorientation would become crucial to the diagnosis of delirium - a position that is not supported by studies that indicate a frequency that is not sufficiently high to justify a role as a mandatory diagnostic feature [23],[36].

Another challenge relates to the optimal combination of the elements of criterion B since the DSM-5 text indicates that acute onset is `usually’ evidenced by onset over hours or days, while disturbances `tend’ to fluctuate over the course of the day. The implication is that neither description is mandatory, but that a pattern exists whereby either of these elements suffices reflected in the relaxed interpretation of this criterion. Using a strict DSM-5 approach (requiring both be present), this criterion accounted for almost three quarters of DSM-IV-diagnosed cases that were excluded even though the majority had mean DRS-R98 scores that exceeded the diagnostic cutoff score. Therefore, we do not recommend that both aspects of Criterion B be required in order to detect delirium. Similar findings have been described using the CAM algorithm where sensitivity is enhanced when either acute onset or symptom fluctuation (rather than acute onset and fluctuating symptoms) is required [37],[38].

Criterion D addresses the attribution of symptoms to delirium versus other states, particularly dementia and newly introduced in DSM-5 is an exclusion from low arousal states `such as coma’ (although the guidance notes state that non-comatose patients giving even minimal responses to verbal stimulation should be classified as showing `severe inattention’). Operationalizing this aspect of diagnosis is challenging but can be achieved by applying tools with established discriminating capacity in cases complicated by comorbid dementia. For these analyses we used the DRS-R98 which distinguishes delirium from other neuropsychiatric conditions [22],[29] and can thus help to clarify if a diagnosis of delirium should apply where there is evidence for a comorbid neuropsychiatric condition that may complicate clinical presentation.

Our finding of high concordance between clinically-determined DSM-IV diagnosis of delirium and the algorithm-generated DSM-5 relaxed diagnosis supports the usefulness of the DRS-R98 for identifying core diagnostic elements of delirium in an operationalized way. DSM criteria are aimed primarily towards clinicians and designed for use in everyday practice where flexibility and `common sense’ are desirable elements. However, they have also become the research standard and in this context require precise systematic methods for research use. The DRS-R98-based approach described herein overlapped substantially with the DSM-IV criteria identified by detailed clinical review by expert clinician-researchers and can thus assist where more systematic diagnosis is desired.

The concept of delirium described in DSM-5 overlaps considerably with DSM-IV delirium, but with narrower capture of delirium. Depending on the interpretation of criteria that is applied, between 11% and 70% of cases of DSM-IV delirium did not meet the new criteria, which has important implications for case identification in clinical and research activity. Overly-strict adherence for some new text details in DSM-5 criteria would greatly reduce the number of delirium cases diagnosed; however a more `relaxed’ approach renders DSM-5 criteria comparable to DSM-IV with minimal impact on their actual application. We also found that delirium diagnosis based upon relevant DRS-R98 items has substantial overlap with DSM-IV diagnoses from detailed clinical assessment, especially when interpreted that inattention is accompanied by impaired awareness flexibly defined in criterion A and that acute or subacute onset of symptoms with or without a fluctuating course accounts for criterion B of DSM-5. This relaxed interpretation of DSM-5 criteria has greater concordance with DSM-IV and is more inclusive of cases that have substantial delirium symptoms as measured on the DRS-R98. We, therefore, recommend this approach to delirium diagnosis according to DSM-5 criteria as it maintains the perceived strengths of DSM-IV in terms of simplicity and inclusiveness while clarifying how to address issues such as reduced alertness or inability to co-operate with assessments. Further studies can explore the therapeutic and prognostic relevance of different application of these criteria.