Strengths and limitations of this study

To our knowledge, this is the first prospective study to investigate lucidity in advanced dementia using direct observation and audiovisual recording, which will enable direct measurement of features that may characterise lucid episodes (LEs) and facilitate reliability assessment and re-review of data important to the early developmental stage of research on lucidity in dementia.

Longitudinal audiovisual data will enable development of computational linguistic measures and temporally situated (timed-event) measures, allowing for a comparative and quantitative assessment of LEs and their features by establishing individual participant baselines and quantifying magnitudes of shift from baseline with LEs as well as individual trajectories across time.

As there are not yet established definitional criteria for LEs, triangulation of proxy informant review and validation with quantitative measures will support early efforts to identify specific features of an LE which may inform conceptual definitions.

The study does not incorporate physiological measurements to coincide with audiovisual data, which could provide insights into the underlying causes of LEs.

Provided the lack of standard definitional criteria for LEs occurring in the context of advanced dementia, this study will not perform reliability assessments of informant case reviews aimed at providing face validation of LEs and thus has limited capacity to evaluate potential sources of bias in informant appraisal of LEs.

Introduction

Lucid episodes (LEs), described as a spontaneous and transient recovery of abilities in advanced neurodegenerative disease that take place after such abilities were believed to have been irretrievably lost, have been reported across several conditions including advanced dementia.1 2 In the context of dementia, these events are clinically and potentially epidemiologically significant as they are seemingly implausible within our current understanding of disease biology and progression.1 Evidence and research on LEs in dementia had been limited prior to a 2018 National Institute on Aging (NIA) Workshop and subsequent research funding initiative, spurring a small yet growing body of evidence that is providing a preliminary understanding of LEs, their potential prevalence and characteristics.2–7

Findings across ongoing studies demonstrate that LEs are particularly impactful events for caregivers who witness them.3 5 8 Provided the underlying heterogeneity in the presentation of dementia, as well as an individual’s specific cognitive or functional deficits and retained abilities, LEs can be highly variable events, which present unique challenges for their capture, characterisation and validation.9 Therefore, LEs may be most noticeable to observers with close familiarity to the individual with dementia and an understanding of the meaning associated with the content or nature of communicative, expressive or functional behaviours exhibited during an LE.9 Across studies, caregivers who witness LEs have reported positive, neutral and negative responses associated with the event3 4; one survey study reported 12% of caregivers making changes to their care plan following an LE.5 Caregivers often described LEs as shocking, surprising and highly meaningful moments of connectedness, and reported efforts to understand what may have brought about or may trigger further LEs.3 5 Further, findings across qualitative and survey studies suggest that witnessing an LE can also be accompanied by complex feelings of grief.3–5 8

As data on LEs in dementia are still very limited, conceptual definitions for lucidity remain an area of active development.10–12 Existing provisional definitions for LEs emphasise common attributes that may be central to the measurement of lucidity, including spontaneity, meaningfulness, communication and connectedness.1 Descriptions of LEs across case reports emphasise changes in speech, language and communication abilities.13–15 Additionally, other changes, such as recovery of functional abilities around specific care tasks or daily activities, have also been reported.3 A major challenge to advancing our understanding of LEs in dementia is the lack of empirical data capturing these events. Through observational study designs employing audiovisual observation coupled with informant appraisals, there is an opportunity to develop approaches that enable the direct capture, characterisation and validation of LEs.

The overarching objective of this prospective, longitudinal, observational study is to capture, characterise and validate audiovisual data capable of providing insight into features of LEs in advanced dementia. Specific study aims associated with this objective include the following:

Aim 1. Capture audiovisual data to generate initial estimates of mean scores, duration and frequency of specific features of LEs within the study cohort.

Aim 2. Validate potential LEs via informant case reviews incorporating unstructured and structured informant appraisal of specific event attributes (eg, meaningfulness of communication).

Aim 3. Characterise verbal and non-verbal features of LEs via triangulation across informant appraisal and quantitative variables generated via computational linguistics (score-based verbal output measures) and timed-event coding of verbal expressions (eg, laughing, crying), non-verbal communicative actions (eg, greeting gesture, eye gaze) and functional behaviours (eg, sitting, standing).

Methods and analysisStudy setting and design

This prospective, longitudinal, observational study will be conducted from December 2021 through May 2025 at a 12-bed inpatient hospice unit specialising in dementia care. Summaries of LE case reports have suggested that these events predominate near end of life8 16 17 and often occur in the context of advanced dementia,13–15 17 informing our eligibility criteria and the decision to use a hospice facility as the study site to maximise the opportunity to capture and detect LEs. The study site was also chosen due to the patient population of individuals living with Alzheimer’s disease or a related form of dementia (ADRD).

The study will include audiovisual observation of patients with advanced dementia near end of life to enable capture, characterisation and validation of LEs. Multiple mechanisms will support the identification of potential LEs, including field interviews with caregivers/clinicians and detection by the research team during and after data collection. Audiovisual coding procedures will generate quantitative measures of variations in verbal output, verbal expressions, non-verbal communicative actions and functional behaviours. Informant case reviews will provide opportunities for validation of potential LEs and facilitate triangulation with quantitative measures to support characterisation of features of LEs.

Participants and eligibility criteria

The study will enrol three types of participants: (1) patients with advanced dementia, (2) caregivers who are family and friend visitors, and (3) clinicians working on the unit. It is estimated that approximately 155 participants will be enrolled in this study. Patient enrolment will be dependent on variable census of the 12-bed unit and factors such as eligibility criteria and clinical assessment of proximity to death; therefore, the patient participant pool will likely be limited to enrolment of one patient every 1–2 months. Given this, 35 patients, 70 caregivers and 50 clinicians are expected to enrol. It is anticipated that for every patient participant approximately, two caregivers will enrol given people living with advanced dementia often have more than one caregiver involved in their care.

Patients with advanced dementia will be determined eligible to participate if there is (1) a diagnosis of dementia as noted through chart review, not limited to any specific pathology; (2) a score of ≥37 (severe) on the Dementia Severity Rating Scale (DSRS)18; and (3) documentation confirming decisional incapacity and presence of a legally authorised representative (LAR). Family and friend caregivers/visitors of patient participants and clinicians employed on the unit (ie, nurses, nursing assistants, social workers) will be eligible to participate if they are 18 years or older.

Patient and public involvement

Prior to the development of this protocol, the research team interviewed caregivers of people living with dementia and study site clinicians to better understand the acceptability and feasibility of audiovisual observation procedures when studying lucidity in hospice patients living with advanced dementia.3 Caregivers and clinicians indicated that such a study would be both acceptable and feasible, as well as shared their preferences regarding privacy and flexibility in the data collection procedures. Study participants may opt into communications about study updates and findings, and a summary of findings will be distributed to participants at the conclusion of the study.

Recruitment

The LAR of potentially eligible patients will be referred to the research team by clinicians and will have the opportunity to learn more about participation in the study by meeting with the research team and reviewing consent documents. The research team will screen referred patients for eligibility through confirmation of a dementia diagnosis in the medical record, completion of the DSRS with clinicians serving as proxy respondents and confirmation of documentation of a LAR able to provide informed consent. The LAR of eligible patients will meet with research staff to complete informed consent and enrolment. The LAR of enrolled patients will be invited to share informational brochures with individuals they anticipate may visit the patient throughout the course of the study. Additional caregiver participants may be identified and recruited by the research team as patients have visitors while the study is in process. Clinicians will be recruited via staff meetings, emails, referrals and informational postings in staff common areas. Recruitment of clinicians will be ongoing as new staff members are introduced to the unit throughout the course of the study.

Consent and assent procedures

Written informed consent will be obtained from the LAR of patient participants during a meeting with research staff who will explain the study and provide opportunities for questions and clarification. Patients will be unlikely to participate in the consent process given advanced dementia is an eligibility criterion. However, determining assent/dissent from patient participants may be appropriate during enrolment based on individual patient abilities and guidance from LARs. This process will include an informal meeting with the patient, through which trained research staff will work with the LAR to determine an appropriate communication approach and pacing to enhance understanding of study participation. If patients are unable to provide verbal communication, research staff will look for non-verbal cues during this conversation and throughout the study that indicate assent or dissent.19 Recognising that assent/dissent is a fluid process, assessment for indications of dissent will continue throughout study procedures as research staff will round on patients at least every 30 min during audiovisual observations. Indicators of dissent may include a participant becoming upset when study procedures take place or making statements that indicate they are not comfortable with research staff or procedures.

Caregiver and clinician participants will receive written information and discuss study procedures with a research team member prior to providing verbal consent. Both caregivers and clinicians may consent to engage in any or all of the following components of the study: (1) presence during video observations, (2) identification of potential LEs through informal field interviews regarding a significant event they witnessed during observations and (3) participation in case reviews of potential LEs. The research team will also offer an option to participate in observations with increased privacy in the case caregivers or clinicians prefer to have their face or voice removed from audiovisual data.

Data collectionAudiovisual observations

Video observations will occur between 07:00 and 23:00, not exceeding 12 hours per day, for as many days as the LAR and patient participant consent to participate. The amount of data collected will vary for each patient based on their length of stay and LAR preferences regarding the duration and intensity of study participation. Observations will be structured to avoid observing during personal care time or during activities/visits with individuals who have not consented to participate in observations. Paper signs communicating that audiovisual recording is underway will be posted in the environment to alert clinicians and visitors. While most observations take place in private rooms, some will take place in common areas such as dining or activity rooms. Observations conducted in public areas will be positioned in such a way that only the patient and consented individuals are visible.

Observations will use GoPro Hero cameras and Panasonic HC-W580 compact camcorders. These recording devices will be used because of their unobtrusive size at approximately 2.5″×2″×1″ (GoPro) and 2.4″×2.4″×5.1″ (camcorder), which serves as an important safeguard to reduce burden or negative influence of the observation equipment on patient participants. When camcorders are used, they may be affixed at a greater distance from the participant, using the optical zoom function to focus only on the participant, and a small auxiliary audio recorder will be placed closer to the patient to maximise audio quality of participant verbalisations. In all cases, video recording devices will be positioned narrowly on the patient participant in order to maximise data quality for capturing subtle non-verbal communication behaviours, such as eye gaze.

The research team will remain on site during audiovisual observations, which will enable easy coordination of observation timing with both clinicians and visiting caregivers. During observations, research staff will monitor observations every 30 min to assess for signs of dissent or unintended consequences of observations, verify equipment functionality and check for the presence of any unconsented individuals.

Identification of potential LEs

Multiple methods will be used to support the identification of potential LEs, including (1) field interviews with caregiver/clinician witnesses, (2) reports from research staff witnesses and (3) detection by research staff conducting video processing procedures. Visiting caregivers and clinician participants will be made aware of the option to engage in brief field interviews if they witness a potential LE during data collection. Field interviews will be impromptu, informal conversations with research staff at the study site and may be initiated by caregivers, clinicians or researchers. Perspectives of individuals with familiarity and contextual knowledge of the patient will provide opportunities to identify subtle yet significant changes that may evidence potential LEs, as these changes may be most readily detectable by those with personal history with the patient. The purpose of the field interviews is solely to provide an opportunity for caregivers and clinicians to alert researchers of a potential LE. Research staff may identify potential LEs as a result of witnessing significant events during audiovisual observations or may detect them while processing or coding audiovisual data. All identified potential LEs will be reviewed by the research team, who will consider advancement to informant case review based on the significance of the shift from participants’ baseline (ie, the routine degree of fluctuations in communication, expression and functional abilities that changes over time).

Informant case reviews/validation of potential LEs

Caregiver and clinician participants familiar with the patient with dementia will be invited as proxy informants to complete case reviews of video segments evidencing potential LEs. After an informant views the video, research staff will ask questions to elicit the informant’s reaction and to determine if they validate the video segment as an LE. In addition to verbal questions, informants will be asked to complete a short survey using a Likert scale to assess varying attributes thought to be associated with lucidity, including spontaneity, meaningfulness, connectedness, influence of event on observers, insight or meta-awareness, return or recovery of abilities and coherence.9 Case reviews will be recorded, transcribed and deidentified for analysis to facilitate triangulation between informant appraisal and quantitative shifts in verbal and non-verbal event features.

MeasuresParticipant measuresDemographic characteristics

Patient demographics, including age, gender, race/ethnicity, educational background, marital status, previous occupation, years experiencing memory changes and number of languages fluently spoken, will be collected from the LAR via a demographic questionnaire at the time of enrolment. Caregivers and clinicians will also have the opportunity to complete demographic questionnaires.

Medical chart review

Research staff will conduct a structured chart review of patient electronic medical records to obtain medical history, medical diagnoses, and medications or interventions delivered during the period of participation in the study.

Dementia Severity Rating Scale (DSRS)

The DSRS is an informant scale that will be administered by clinicians to determine eligibility and to assess memory, speech and language, recognition of family members, orientation to time, orientation to place, and ability to make decisions, social and community activity, home activities and responsibilities, personal care—cleanliness, eating, control of urination and bowels, and ability to get from place to place.18 The DSRS has been validated against the Clinical Dementia Rating Scale (CDR) and was selected to ease burden on clinicians by reduced training needs and administration time as compared with the CDR.20 21

Family Confusion Assessment Method (FAM-CAM)

During eligibility screening, if delirium superimposed on dementia is suspected (ie, in the case of a recent admission following hospitalisation), research staff will ask the LAR or another individual familiar with the patient to complete the FAM-CAM22 to reduce the risk of enrolling ineligible patients whose DSRS score was impacted by delirium. The FAM-CAM is designed specifically for caregivers and is predictive of the original CAM; both versions assess items such as acute onset and fluctuating course of confusion, inattention, disorganised thinking, altered level of consciousness, disorientation, perceptual disturbances and psychomotor agitation.22–24

American Speech-Language-Hearing Association Functional Assessment of Communication Skills (ASHA FACS)

The ASHA FACS is an informant measure that will be completed by clinicians to assess functional communication of patient participants.25 Specifically, this measure assesses how speech, language and cognitive-communication impairments impact daily life across four domains: social communication; communication of basic needs; reading, writing and number concepts; and daily planning. It is designed for use among adults and has been validated for use in ADRD populations.26

Audiovisual data measures

Coding processes will generate variables from computational linguistic score-based measures of verbal output and temporal measures of verbal expressions (eg, laughing, crying), non-verbal communicative actions (eg, visual engagement, gestures) and functional behaviours (eg, object use, gross motor behaviour). A list of variables that may be generated to characterise features of LEs can be found in table 1.

Table 1

Verbal and non-verbal variables and definitions

Data processing and coding procedures

Audiovisual data will be coded to generate quantitative variables as noted above. Verbal output will be transcribed using the Codes for Human Analysis of Transcripts (CHAT) format.27 Transcriptions will subsequently be coded using the Computerised Language Analysis (CLAN) program, a computational linguistic analysis software, to generate scores for specific domains of verbal output (eg, speech fluency, lexical diversity, semantic content).28 Verbal output scores will be supplemented with timed-event coding of verbal expressions and non-verbal communicative actions and functional behaviours that may also be salient in characterising LEs and distinguishing these events from routine fluctuations. Timed-event variables will be informed by previous literature investigating communication of people living with ADRD.29 30 All timed-event coding will be completed in Noldus Observer XT. Frequency of onset, total duration and the range of duration for each timed-event code in table 1 will then be extracted for statistical analysis. Additionally, counts of co-occurrence between codes and temporal sequencing of codes will be generated.

Given the time-intensive nature of transcription and coding processes, and the large volume of audiovisual data likely to be generated, we anticipate prioritising data for coding procedures in the following ways: (1) the first three audiovisual observations for each participant will be prioritised to establish an individually specific initial baseline, and (2) observations identified as including potential LEs will be prioritised to expedite case review procedures and analysis. Subsequent to these priorities, we anticipate continued reassessment of baseline and review of data to screen for potential LEs.

Data storage and safety

Audiovisual observations will be immediately transferred to an encrypted hard drive after data collection and then transferred to a secure restricted data server for data processing and analysis. Additional demographic and medical record information will be stored on REDCap,31 32 a HIPAA-secure web-based software platform.

Participants will have the opportunity to opt into audiovisual data storage and re-use during the informed consent process. Data will be stored indefinitely on a secure research server. Fully deidentified databases encompassing coded verbal and non-verbal variables generated from video data, deidentified patient medical information and deidentified proxy informant case review transcriptions will be stored on the secure data server.

Data analysis plan

Variables generated through coding processes will be exported into a linked analytical dataset for descriptive and comparative statistical analysis. Descriptive statistics will be used to summarise and compare scores, frequency and distribution of verbal and non-verbal measures during validated LEs as compared with individual participant baseline using R V.3.6.2. Descriptive statistics for timed-event variables will also be calculated in the Generalised Sequential Querier to examine temporal associations before and during validated LEs including the frequency, rate, duration, proportion and patterns in joint frequencies (temporal overlap) between variables. We expect measures of communication such as total words, symbolic gestures or initiating interactions will fluctuate through greater event frequency and duration during LEs compared with baseline time periods; however, we suspect specific verbal and non-verbal variables and the magnitude of shifts within those variables will vary across participants.

Ethics and disseminationEthical considerations

This study protocol (ID 2021-1243) was reviewed and approved by the University of Wisconsin-Madison Institutional Review Board. Given patient participants lack capacity to provide informed consent to this longitudinal study, it is worth noting the importance of the continual assent/dissent procedures to ensure autonomy and respect for participants and their preferences to participate, or not to participate, in research.

Dissemination of findings

Findings will be disseminated in peer-reviewed journals and at national and international conferences. Caregiver and clinician participants can opt to receive a lay-audience report of findings at the conclusion of the study, as well as links to published papers and presentations.

Discussion

This represents the first prospective, longitudinal, observational study focused on investigating lucidity in advanced dementia. As research on this phenomenon is in very early stages, it is likely that we may need to adjust our measurement strategies and other study procedures as findings from other studies provide new insights into defining features of and measurement for LEs.

As we are recruiting patients from a single hospice unit, enrolment will be functionally limited by accrual of new patients to the environment which may limit our ability to recruit the target sample size. Given that estimates regarding the prevalence and incidence of LEs have not yet been established, it is unclear whether the target sample will yield a sufficiently diverse range of LEs; however, findings from preliminary studies suggest that these events are much more common than initially hypothesised.3 5 6 Another potential challenge is a lack of diversity among study participants due to a single site for study enrolment, which may limit generalisability of the cohort and associated findings.

Recent discussions have raised questions regarding the distinction, and/or overlap, between terminal lucidity and LEs in dementia.10 12 33 As a function of focusing on individuals who are enrolled in hospice care, this study by design is more likely to collect LEs that take place near end of life. While it was initially suggested through summaries of case reports of lucid events that LEs predominated near end of life,13 17 emergent findings from new investigations suggest that these events are not limited to this timeframe.3 6 As a result, this study may generate data that are more reflective of LEs that take place near end of life, which could be representative of terminal lucid events.

The presence of a researcher during audiovisual observations may affect participant behaviour and communication, potentially influencing frequency and duration of verbal and non-verbal measures. Similarly, it is unknown if patient-researcher social interactions resulting from standard study procedures may prompt LEs. Future observational studies using remote monitoring procedures could provide insight into the impact of researcher presence on the characteristics and prevalence of LEs.

Our previous collaboration with clinicians and families to develop and refine procedures fostered feasible and acceptable data collection processes that will likely increase the odds of reaching our enrolment targets.3 Finally, this study will lead to the generation of direct observational data, which is essential for overcoming the significant limitations of retrospective reports including recall bias.

Ethics statementsPatient consent for publication

Not applicable.

Acknowledgments

The authors would like to thank study participants, their families and clinical staff at Agrace HospiceCare; co-investigators Drs Nicole Werner, Roger Brown, Ann Kolanowski and Amy Kind for providing expert guidance and review throughout the development of the study; and Megan Botsch for assistance with initial development of timed-event coding variables.