Background
Phenotypes
Delirium
Subsyndromal delirium
Aims
Term | Definition | Potential application to delirium |
---|---|---|
Phenotype | A set of clinical features in a group of patients who share a common syndrome or condition. | Altered cognition Inattention Altered awareness Disorientation |
Subphenotype | A set of features in a group of patients who share a phenotype. Includes shared risk factors, traits, diagnostic features, expression markers, mortality risk, or treatment response—which distinguishes the group from other patients with the same phenotype. | Clinical Shared risk quantification Shared precipitants Specific symptoms, e.g. inattention, agitation Delirium duration Diagnostic features Defined by pathophysiology Prominent mechanism Inflammatory/non-inflammatory Melatonin levels Neurotransmitter presence Network connectivity extent Presence of oxidative stress |
Endotype | A distinct biological mechanism of disease which is often associated with an anticipated clinical course, shared by a patient subgroup. | Associations between biological putative pathways of delirium and the clinical symptoms which occur as a result |
Treatable traits | Subgroup characteristics which may be successfully targeted by an intervention. | Decisions and development of the best course of action for treatment- Treating symptoms Treating the mechanisms which express the symptoms A combination of both |
Subphenotypes of delirium
Clinical subphenotypes
Precipitant subphenotypes
Risk factor subphenotypes
General medicine | Additional operative risks | Additional ICU risks | |
---|---|---|---|
Predisposing risk factors | Older age Low daily activity levels Immobility Sensory impairment Low levels of education Malnutrition Pre-existent cognitive impairment Frailty Comorbidities Alcohol consumption Visual/Hearing impairment | Cerebral disease Chronic diseases: renal, cardiac, hepatic, or pulmonary Alcohol/sedative-hypnotics addiction History of delirium/functional psychosis Depression Vitamin deficiency Seizures or porphyria | Higher illness severity Unexpected hospital admission |
Precipitating risk factors | Acute medical illness Fractures Head injury Trauma Surgery Psychological stress Drug use/withdrawal Urinary catheterisation Longer hospital stay | Drug intoxication/anaesthesia Metabolic disturbance Hemodynamic disturbance Respiratory disorders Infection Acute cerebral disorder Alcohol/sedative withdrawal Intraoperative/post-operative: Sleep deprivation Immobilisation Restraints | Mechanical ventilation (and duration of ventilation) Sepsis Opioids Polypharmacy Circadian rhythm disruption Deep sedation Organ failure |
Mechanisms
Translation of subphenotypes into clinical practice
Challenges in understanding delirium, multimorbidity, and comparison in subphenotyping
Challenge/limitation | Explanation | Suggestions and examples |
---|---|---|
Understanding of pathophysiology | Knowledge of the pathophysiological mechanisms of delirium remains largely hypotheses. There is a lack of understanding in how the biomarkers are regulated and interact. However, it is believed that the mechanisms are complementary or co-dependant, so they may be impossible to isolate. | Large studies using unbiased analytical approaches are required to establish causality of delirium models in individual patients. Samples of specific biomarkers from patients who have been subphenotyped should be analysed and their profiles compared Identification of the putative pathways, and the resultant treatable traits, may be achieved by unsupervised clustering analyses of large delirium datasets recording biological variables. Example: Methods used in the successful endotyping of asthma [1] Clinically relevant parameters were identified, and a threshold was established to include or omit patients from each endotype. Cluster analysis was then used to identify each endotype. These methods may be applied to delirium by identifying the most suitable parameters for patient categorisation. |
Study of relevant patient populations | Studies in specific medical specialities often cannot be generalised across populations due to their environmental heterogeneity, particularly in the ICU. ICU patients are often heavily sedated and difficult to assess, which sometimes leads to small sample sizes in studies, weakening power of results. Due to the high prevalence of delirium in the older population, loss of study follow-up due to attrition is common [118]. | Recruitment of representative clinical populations into large, global research studies. |
Logistics, global data sharing and research cooperation | Practical and logistical challenges present limitations: Complications, time restraints, and opinions associated with data sharing may also limit the use of globally generated datasets, therefore preventing appropriate research development and the robust findings. In critical care, resources, time, and staffing are occasionally limited, so the completion of regular diagnostic delirium assessments or sample retrieval may not always be possible, especially during the COVID-19 pandemic. Research waste must be overcome. | Technology must be adopted appropriately to allow ease of data sharing and collaboration. Increased levels of training and employment will alleviate staffing issues. To allow direct comparability between studies, future work should use core outcome sets. |
Overcoming heterogeneity | The existing delirium literature contains a high level of heterogeneity, therefore identifying the ‘correct’ delirium subphenotypes poses a challenge. The large range of screening techniques could lead to heterogeneity in diagnostic success, and some tests may not distinguish subsyndromal delirium. Mild cases may also be missed during screening, and earlier studies may include conditions other than delirium due to wider definitions which existed [119]. | Reproduction of latent class analysis in many large patient cohorts will highlight the extent of the heterogeneity problem. Comparison of results between cohorts of similar characteristics, for example, similar clinical settings, age groups, risk factors, or precipitants, may allow for the identification of subphenotypes suited exclusively to individual groups or, alternatively, show their reproducibility in differing groups. Common techniques must be adopted in appropriate populations for comparability, and consensus reached on a sensitive, robust, diagnostic technique. |
Comparison of subphenotypes | It is plausible that independent subphenotypes of delirium cannot be identified, due to its transient nature and risk factor interaction. | Description of subphenotype interactions or creating hybrids between multiple subphenotypes would facilitate increased understanding of the syndrome’s expression in individuals, which is the aim. |
Executing analysis correctly | Different results may be generated during cluster analysis, depending on variables included in the analysis and the specific method used [111]. | Example: Successful subphenotyping in acute kidney injury and ARDS Accumulation of pre-existing vulnerabilities and insult(s) results in acute kidney injury which may be categorised as pre-renal, renal, post-renal, or a combination [112]. Subphenotyping of ARDS has been completed using latent class and cluster analysis, which identified hypoinflammatory, hyperinflammatory, uninflamed, and reactive ARDS [109, 113]. Similar methods of identification must be implemented in delirium for improved and more efficient identification, with methods reported in detail for comparison and replication. |
Subphenotype Stability | It is unknown whether subphenotypes and underlying putative pathways are constant throughout. Subphenotype stability across settings and age groups is also unknown. The transient and fluctuating nature of delirium may be problematic in the process of identifying delirium subphenotypes [2]. There is also limited understanding of how short-term phenotypes translate into long-term outcomes. This knowledge limit persists in ARDS despite the progression of subphenotyping in this condition [120]. | Symptom fluctuations may be tracked by consistent delirium monitoring using validated assessment methods, at short time intervals, alongside recording individual patient characteristics. Collection of blood and CSF samples consistently at short time intervals and subsequent biomarker analysis will allow an increased understanding of mechanism stability. Stability of the clinical and mechanism-driven subphenotypes should be compared to expand understanding of the interrelationship between delirium symptoms and their biological pathways. Example: Tracking of subphenotype stability in two ARDS
subphenotypes using latent class and latent transition models [114]. Subphenotype identification may be feasible in the clinical trial context. Methods for stability tracking may be established after the identification of delirium subphenotypes. |
Speed of subphenotype assignment | The speed by which subphenotypes of delirium may be identified is crucial in determining their viability. The current method of categorisation by psychomotor subtype allows for quick assessments without leaving the patient bedside. | Fast subphenotype assignment may be achievable for the possible clinical subphenotypes, with robust recording of aetiologies, comorbidities, and response to treatments. Subphenotypes defined by underlying mechanism currently require more time dedicated to assessment of blood and CSF. Development of point of care testing would aid in alleviating this issue. |
Multimorbidity | Multimorbidity presents as a problem in delirium, and the range of terms used to report comorbidities in studies increases heterogeneity [54]. | Tools such as the Charlson comorbidity index increase ease of comparison between cohorts and may allow a degree of multimorbidity adjustment in analyses [115]. |
Treatment response | The possibility exists that the subphenotypes which are identified in patient cohorts are not prognostic for a treatment response. | In this event, the findings will rule out the suggested characterisation methods and allow for development of further novel research plans to improve delirium categorisation. |
Methods for subphenotype validation
Recommendations for future studies
Research recommendation | Actions required |
---|---|
Establishment of subphenotype reproducibility and overlap | Use of large prospective studies with heterogeneous patient cohorts to validate subphenotypes and compare similar subphenotypes. |
Establishment of subphenotype stability across clinical settings and patient demographics | The prospective studies must exist of heterogeneous patient cohorts across multiple clinical settings. The studies should be repeated to assess subphenotypes which differ by severity and duration, and studies should also be repeated at multiple time points. Subphenotype-related biomarkers should be compared across various settings where appropriate. |
Validation of subphenotyping strategies | Sharing of large datasets and algorithms between investigators, ideally by making data open access. Greater levels of cybersecurity required. |
Reduction in research competitivity | Emphasis on collaboration and involvement in publications. |
Conclusion
Acknowledgements
Declarations
Ethics approval and consent to participate
Not applicable.
Consent for publication
Not applicable.