University of Groningen
Feasibility Study of Assessing the Preclinical Alzheimer Cognitive Composite (PACC) Score via Videoconferencing
Seghezzo, Giulia ; Van Hoecke, Yvonne ; James, Laura; Davoren, Donna; Williamson, Elizabeth ; Pearce, Neil; McElvenny, Damien; Gallo, Valentina
Published in:
Journal of Neurology
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Early version, also known as pre-print
Publication date: 2021
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Seghezzo, G., Van Hoecke, Y., James, L., Davoren, D., Williamson, E., Pearce, N., McElvenny, D., & Gallo, V. (2021). Feasibility Study of Assessing the Preclinical Alzheimer Cognitive Composite (PACC) Score via Videoconferencing. Journal of Neurology.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Journal of Neurology
Feasibility study of assessing the Preclinical Alzheimer Cognitive Composite (PACC)
Score via videoconferencing
--Manuscript
Draft--Manuscript Number: JOON-D-20-02865R1
Full Title: Feasibility study of assessing the Preclinical Alzheimer Cognitive Composite (PACC) Score via videoconferencing
Article Type: Original Communication
Corresponding Author: Valentina Gallo University of Groningen NETHERLANDS Corresponding Author Secondary
Information:
Corresponding Author's Institution: University of Groningen Corresponding Author's Secondary
Institution:
Corresponding Author E-Mail: v.gallo@rug.nl;v.gallo@qmul.ac.uk
First Author: Giulia Seghezzo
First Author Secondary Information:
Order of Authors: Giulia Seghezzo
Yvonne Van Hoecke Laura James Donna Davoren Elizabeth Williamson Neil Pearce Damien McElvenny Valentina Gallo Order of Authors Secondary Information:
Funding Information: Drake Foundation
(EPMSZO61) Prof Neil Pearce
Abstract: Background
The Preclinical Alzheimer Cognitive Composite (PACC) is a composite score which can detect the first signs of cognitive impairment, which can be of importance for research and clinical practice. It is designed to be administered in person; however, in-person assessments are costly, and are difficult during the current COVID-19
pandemic. Objective
To assess the feasibility of performing the PACC assessment with videoconferencing, and to compare the validity of this remote PACC with the in-person PACC obtained previously.
Methods
Participants from the HEalth and Ageing Data IN the Game of football (HEADING) Study who had already undergone an in-person assessment were contacted and re-assessed remotely. The correlation between the two PACC scores was estimated. The difference between the two PACC scores was calculated and used in multiple linear regression to assess which variables were associated with a difference in PACC scores.
Findings
Of the 43 participants who were invited to this external study, 28 were re-assessed.
The median duration in days between the in-person and the remote assessments was 236·5 days (7·9 months) (IQR 62·5). There was a strong positive correlation between the two assessments for the PACC score, with a Spearman correlation coefficient of 0·75 (95% CI 0·56, 0·95). The multiple linear regression found that the only predictor of the PACC difference was the time between assessments.
Interpretation
This study provides evidence on the feasibility of performing cognitive tests online, with the PACC tests being successfully administered through videoconferencing. This is relevant, especially during times when face-to-face assessments cannot be performed. Response to Reviewers: We thank the reviewers for their valuable comments, and we enclose in the document
our responses, and the data relating to the registration of the protocol for this study. PROTOCOL
The protocol for this study was published, as requested by the journal:
Seghezzo, G., Van Hoecke, Y., James L., Davoren, D., Williamson, E., Pearce, N., McElvenny, D., Gallo, V. “Feasibility study of assessing the Preclinical Alzheimer Cognitive Composite (PACC) Score via Videoconferencing” Protocol Exchange (2020) DOI: 10.21203/rs.3.pex-1256/v1
REVIEWER 1 COMMENTS
1. It is very unclear why a Spearman rather than a Pearson correlation was used. At the very least either the Pearson or both should be presented.
Spearman was used due to small sample size, and due to non-normal distribution of the two PACC scores, however, its limitation is that it only assesses if there is an association between the two values, it cannot determine the distribution of the data, which Pearson can. The Pearson correlation has now replaced the previous Spearman correlation.
2. There seems no real need for the figures-they do not add anything to the evaluation.
- Figure 2: Scatterplots show the correlation between the two PACC scores and their distribution;
- Figure 3: The Bland-Altman plot shows the agreement between the two scores, and how there are no differential biases between the scores. It is difficult to convey, in words, the bias present;
-Figure 4: This is a visual representation of the possible effect of time between assessments on the difference between the PACC scores. This figure aids in the interpretation of the beta coefficient of the linear regression.
Any of these could be moved to the supplementary material if all reviewers agree, however we believe they aid the interpretation of results.
3. While the authors are aware their sample is biased to those who are more computer literature, this point needs to be expanded.
There is some evidence that our sample might be bias towards participants who are more computer literate, as it is based on participants who were willing to participate to a second assessment, remotely. This makes it difficult to generalize our results to the general public including a less computer literate population. Therefore, while it does not affect the association found, this will hinder the external validity of our results, meaning the same association might not be found when we include a wider range of computer literacy into the sample. These considerations have been added in the discussion, providing also some reference to substantiate the argument.
4. The authors need to note that the education level of those who refused the study was significantly less than those who accepted. A chi-square comparison is appropriate (those who are above versus those at GSC or below)
There is a statistically significant difference between the education level of those who participated in the remote assessment to those who did not, with those participating in the remote assessment having a higher educational qualification (Chi-square p=0.024). This has now been made explicit in the text. Again, while hampering generalizability of results, this has no impact on their internal validity. Such considerations have been added to the discussion
5. A discussion of the problems of evaluating those with more advanced symptoms
than seen here.
This study only provides evidence of the remote administration of the PACC in a substantially cognitive integer population. It is difficult to extrapolate the same considerations to a more compromised population. However the PACC is designed to assess subtle cognitive changes detectable before any clinical diagnosis of cognitive impairment, therefore our populations represents a typical population to be assessed using this score test. We have now made explicit in the discussion that generalizing the present data to a cognitively impaired population requires further evidence.
Reviewer #4 Comments
This is a timely manuscript on an important topic, but unfortunately there are some major problems with the research design.
The main difficulty is the test-retest interval and the use of a self-selected, convenience sample. The sample size is small on the participation rate is only 65%. The patients that declined participation may have incurred changes in mental status during the long test-retest interval. To conclude that the remote assessment is feasible is not a strong enough conclusion for publication.
The original HEADING sample was selected randomly in batches from a list of all members in the PFA database meeting the study requirements. No bias is present in this initial invitation into the main HEADING study. The subsample of this sub-study initially comprised all HEADING participants assessed at the time the COVID-19 lockdown measures were introduced. The data already available from the participants who had participated in-person previously were used to validate the remote
assessments. Therefore the original sample was either convenience nor self-selected. Nonetheless, response rate is less than ideal. Also, there is some evidence suggesting the presence of a selection bias: participants assessed remotely were more educated and younger. This hampers generalizability of results which has been now discussed extensively in the discussion.
The time between assessments was long and varying, ranging from 3 months to 9 months. While a long time had passed between assessments, original research for the PACC score found that a significant difference in scores was found at earliest 12 months [7]. The participants that did not participate in the remote assessment had a median time between assessments of 266 days (q25, p75: :206, 268), which was about a month longer than the average time between assessments of 236.5 days (p25, p75: 194.5, 257) of those who did participate in the remote assessment.
We indeed found a difference in PACC among those re-assessed and those who declined. This information has been added in Table 1, and discussed in the discussion. The correlation between the in-person and remote testing is statistically significant but far from ideal. It accounts for 56% of the variance. Should be compared with the PACC reliability as originally established.
The correlation is high, and the large confidence interval is due to the low power this study had and is considered with the other limitations presented in the manuscript. The original research of the PACC does not provide a reliability of their composite score, however, there is research on the individual components of the PACC. The DSST has the highest reliability of 0.86 [.85, .87], followed by the LMDR 0.65 [.63, .67] and the MMSE with 0.49 [.47, .52] [8]. This study uses the FNAME-12, while the FNAME-16 found a reliability of 0.62 (p<0.001) [9]. However, reliability of a study is dependent of many methodological and personal factors, including the cognitive function of a participant along with the time between assessments, potential learning effects, measurement error, and random error, including regression to the mean [10]. It should be noted that these studies mentioned had a longer time between assessments (average, 12 months) which introduces more variability into the sample the longer the assessment, as well as performing different tests [10]. Therefore, given these considerations, the reliability of the individual tests cannot be immediately compared, however it does reflect similarly to the correlation found in this study.
If the question is whether the remote test can supplant the in-person assessment, the design would ideally be a random allocation to each testing condition at baseline and a 2-3 week follow-up. The design would be counterbalanced, ½ getting each format first. We agree that this would be the ideal study design to test if the remote PACC could entirely replace the in person one. In fact, we had already prepared a protocol for such
a study as an add-on to the BRAIN/HEADING studies. Unfortunately, however, with the sudden implementation of the lockdown measures, such a study was not feasible any more for some time, and would have been delayed. We decided therefore to leverage on the resources we had available at the time of lockdown to starting explore this concept. For this reason we are only sharing our consideration on the feasibility of administering the PACC remotely and if this would produce similar results to those already obtained in-person. We have noted in the conclusion that such a study will be eventually needed.
It would be helpful to know what the test-retest reliability of the PACC is in the basic psychometric validation research.
See above. We have added considerations on the PACC reliability in the manuscript, and fully responded to the reviewer previous question.
The platform used by telemedicine is important as each one has different size of images presented. What platform was used is only cursorily described. I wonder how the visual stimuli could possibly be presented on a smart phone.
This is a good point, and we agree that the videoconferencing device could have made a difference (e.g. the Face-Name test where participants are shown a face with name and occupation pairings, as the level of detail can be different given the size of the device). Given that this was a pragmatic study, we preferred to include any
participants, irrespective of their access to different devices. Despite we have collected information on software used, unfortunately we did not collect information on the device used. Retrospectively we should have collected this information as well, adjusting for software and device. Stillerova et al. have performed video assessments using multiple software and various devices, and did not find any difference among modes, however we cannot confirm the same for our study.
I am troubled by the mailing of test materials as opposed to a strictly screen sharing procedure as has been utilized in the literature - and is becoming routine in this pandemic environment. For example, the DSST would be presented on the screen as the examiner commences timing, rather than patients, some of whom are cognitively impaired, essentially timing themselves.
As described in the text, in order to ensure maximal comparability with the in person test, the participants did not time themselves, the examiner commenced the time when telling the participant to begin, after completing the practice portion of the DSST and explaining the test to the participants. Giving the key of the DSST on the screen would imply that the participant as to look at the screen, then look at their sheet to write their response, this will slow them down, and alter the DSST results which assesses psychomotor speed.
There are online versions of the DSST, however since the original in-person DSST was administered on paper, we aimed at keeping it as similar as possible to minimize variability. Furthermore, in keeping the assessments as similar as possible, minimal computer skills were implied in the assessment, besides screen sharing for the FNAME. The test was in a sealed envelope, with clear instructions not to open prior to instruction by the examiner, and once complete, the participants held up their DSST worksheet to the camera, so the examiner could mark how far they completed the DSST. The DSST is then immediately placed in the return envelope in front of the assessor.
I note that some patients had interrupted internet signal and confusion in setting up. This should be expanded upon.
The interrupted internet signal was adjusted for in analysis, as this was recorded during each videoconference (by recording the internet speed (mbps)). All participants were set a simple step by step instruction guide of how to use Skype or Zoom if they were not familiar with any videoconferencing software. Once on the call, the
participants did not have to do anything, as the research assistants facilitated the call with screensharing when necessary.
There are copyright protections for tests (eg MMSE, WMS) to prevent patients having access to test materials when there is no professional oversight. What is to prevent a patient from sharing this information with others? Did the authors have permission from test publishers to mail the stimuli?
We have checked with the London School of Hygiene and Tropical Medicine copyright team for permission. There are no issues with the MMSE worksheet, as all that was sent was a blank page with space for a sentence to be written, and a picture of two overlapping pentagons, which hold no intellectual property. The WMS was only given orally, and the participants were watched to ensure they did not write anything, which they did not, along with the FNAME as well. The DSST worksheet was sent to the participants in a sealed envelope with clear instructions not to open prior to the assessment and was immediately returned after the completion of the remote assessment. Nonetheless we requested a retrospective license for using the test remotely, we appreciate the concern, and we will add a note in the future warning participants to not photocopy the worksheets and to return then as they were received. Given all of these potential confounds it is surprising that a correlation of 0.75 was found. If the sample size was sufficient and the design counterbalanced the conclusion would have greater credibility.
Of course, a larger sample size and counterbalanced study design would provide greater credibility to our study. The small sample size does decrease the power to detect significant results in this instance, increasing the chance for an overestimation of the true effect estimate. However, Figure 2 and Figure 3 show there is an agreement between the two assessments. Given the COVID restrictions the researchers could not perform an appropriate counterbalanced design, however, think sharing the results found are still worthwhile in the field. See conclusions
Author Comments: Prof Valentina Gallo Campus Fryslan University of Groningen Leeuwarden The Netherlands email: v.gallo@rug.nl https://www.rug.nl/staff/v.gallo/research Journal of Neurology Editor-in-chief Prof Roger Barker
January 8th, 2021
Dear Roger,
Thanks for the interesting and insightful comments we have received from the Journal of Neurology reviewers.
As requested by the editorial board, we have registered the protocol for this study; moreover we have addressed all the reviewer comments in the present revised version.
We have been invited by more than one source to share our positive experience on continuing our research during the lockdown and the COVID-19 restrictions. We are therefore looking forward to quote this paper outlining our efforts in doing so, including the discussion of its strengths and limitations.
Please find enclosed a revised version of the manuscript entitled, “Feasibility study of assessing the Preclinical Alzheimer Cognitive Composite (PACC) Score via
videoconferencing” alongside the rest of the material.
We thank you in advance for considering our paper once more.
Yours faithfully, Prof Valentina Gallo
We thank the reviewers for their valuable comments, and we enclose in the document our responses, and the data relating to the registration of the protocol for this study.
PROTOCOL
The protocol for this study was published, as requested by the journal:
Seghezzo, G., Van Hoecke, Y., James L., Davoren, D., Williamson, E., Pearce, N., McElvenny, D., Gallo, V. “Feasibility study of assessing the Preclinical Alzheimer Cognitive Composite (PACC) Score via Videoconferencing” Protocol Exchange (2020) DOI: 10.21203/rs.3.pex-1256/v1
REVIEWER 1 COMMENTS
1. It is very unclear why a Spearman rather than a Pearson correlation was used. At the very least either the Pearson or both should be presented.
Spearman was used due to small sample size, and due to non-normal distribution of the two PACC scores, however, its limitation is that it only assesses if there is an association between the two values, it cannot determine the distribution of the data, which Pearson can. The Pearson correlation has now replaced the previous Spearman correlation.
2. There seems no real need for the figures-they do not add anything to the evaluation.
- Figure 2: Scatterplots show the correlation between the two PACC scores and their distribution;
- Figure 3: The Bland-Altman plot shows the agreement between the two scores, and how there are no differential biases between the scores. It is difficult to convey, in words, the bias present;
-Figure 4: This is a visual representation of the possible effect of time between assessments on the difference between the PACC scores. This figure aids in the interpretation of the beta coefficient of the linear regression.
Any of these could be moved to the supplementary material if all reviewers agree, however we believe they aid the interpretation of results.
3. While the authors are aware their sample is biased to those who are more computer literature, this point needs to be expanded.
There is some evidence that our sample might be bias towards participants who are more computer literate, as it is based on participants who were willing to participate to a second assessment, remotely. This makes it difficult to generalize our results to the general public including a less computer literate population. Therefore, while it does not affect the association found, this will hinder the external validity of our results, meaning the same association might not be found when we include a wider range of computer literacy into the sample. These considerations have been added in the discussion, providing also some reference to substantiate the argument.
Authors' Response to Reviewers' Comments Click here to access/download;Authors' Response to Reviewers' Comments;JOON review comments_response.docx
4. The authors need to note that the education level of those who refused the study was significantly less than those who accepted. A chi-square comparison is appropriate (those who are above versus those at GSC or below)
There is a statistically significant difference between the education level of those who participated in the remote assessment to those who did not, with those participating in the remote assessment having a higher educational qualification (Chi-square p=0.024). This has now been made explicit in the text. Again, while hampering generalizability of results, this has no impact on their internal validity. Such considerations have been added to the discussion
5. A discussion of the problems of evaluating those with more advanced symptoms than seen here.
This study only provides evidence of the remote administration of the PACC in a substantially cognitive integer population. It is difficult to extrapolate the same considerations to a more compromised population. However the PACC is designed to assess subtle cognitive changes detectable before any clinical diagnosis of cognitive impairment, therefore our populations represents a typical population to be assessed using this score test. We have now made explicit in the discussion that generalizing the present data to a cognitively impaired population requires further evidence.
Reviewer #4 Comments
This is a timely manuscript on an important topic, but unfortunately there are some major problems with the research design.
The main difficulty is the test-retest interval and the use of a self-selected, convenience sample. The sample size is small on the participation rate is only 65%. The patients that declined participation may have incurred changes in mental status during the long test-retest interval. To conclude that the remote assessment is feasible is not a strong enough
conclusion for publication.
The original HEADING sample was selected randomly in batches from a list of all
members in the PFA database meeting the study requirements. No bias is present in this initial invitation into the main HEADING study. The subsample of this sub-study initially comprised all HEADING participants assessed at the time the COVID-19 lockdown measures were introduced. The data already available from the participants who had participated in-person previously were used to validate the remote assessments. Therefore the original sample was either convenience nor self-selected.
Nonetheless, response rate is less than ideal. Also, there is some evidence suggesting the presence of a selection bias: participants assessed remotely were more educated and younger. This hampers generalizability of results which has been now discussed extensively in the discussion.
The time between assessments was long and varying, ranging from 3 months to 9 months. While a long time had passed between assessments, original research for the
PACC score found that a significant difference in scores was found at earliest 12 months [7]. The participants that did not participate in the remote assessment had a median time between assessments of 266 days (q25, p75: :206, 268), which was about a month longer than the average time between assessments of 236.5 days (p25, p75: 194.5, 257) of those who did participate in the remote assessment.
We indeed found a difference in PACC among those re-assessed and those who
declined. This information has been added in Table 1, and discussed in the discussion. The correlation between the in-person and remote testing is statistically significant but far from ideal. It accounts for 56% of the variance. Should be compared with the PACC reliability as originally established.
The correlation is high, and the large confidence interval is due to the low power this study had and is considered with the other limitations presented in the manuscript. The original research of the PACC does not provide a reliability of their composite score, however, there is research on the individual components of the PACC. The DSST has the highest reliability of 0.86 [.85, .87], followed by the LMDR 0.65 [.63, .67] and the MMSE with 0.49 [.47, .52] [8]. This study uses the FNAME-12, while the FNAME-16 found a reliability of 0.62 (p<0.001) [9]. However, reliability of a study is dependent of many methodological and personal factors, including the cognitive function of a participant along with the time between assessments, potential learning effects, measurement error, and random error, including regression to the mean [10]. It should be noted that these studies mentioned had a longer time between assessments (average, 12 months) which introduces more variability into the sample the longer the assessment, as well as performing different tests [10]. Therefore, given these considerations, the reliability of the individual tests cannot be immediately compared, however it does reflect similarly to the correlation found in this study.
If the question is whether the remote test can supplant the in-person assessment, the design would ideally be a random allocation to each testing condition at baseline and a 2-3 week follow-up. The design would be counterbalanced, ½ getting each format first.
We agree that this would be the ideal study design to test if the remote PACC could entirely replace the in person one. In fact, we had already prepared a protocol for such a study as an add-on to the BRAIN/HEADING studies. Unfortunately, however, with the sudden implementation of the lockdown measures, such a study was not feasible any more for some time, and would have been delayed. We decided therefore to leverage on the resources we had available at the time of lockdown to starting explore this concept. For this reason we are only sharing our consideration on the feasibility of administering the PACC remotely and if this would produce similar results to those already obtained in-person. We have noted in the conclusion that such a study will be eventually needed.
It would be helpful to know what the test-retest reliability of the PACC is in the basic psychometric validation research.
See above. We have added considerations on the PACC reliability in the manuscript, and fully responded to the reviewer previous question.
The platform used by telemedicine is important as each one has different size of images presented. What platform was used is only cursorily described. I wonder how the visual stimuli could possibly be presented on a smart phone.
This is a good point, and we agree that the videoconferencing device could have made a difference (e.g. the Face-Name test where participants are shown a face with name and occupation pairings, as the level of detail can be different given the size of the device). Given that this was a pragmatic study, we preferred to include any participants, irrespective of their access to different devices. Despite we have collected information on software used, unfortunately we did not collect information on the device used. Retrospectively we should have collected this information as well, adjusting for software and device. Stillerova et al. have performed video assessments using multiple software and various devices, and did not find any difference among modes, however we cannot confirm the same for our study.
I am troubled by the mailing of test materials as opposed to a strictly screen sharing procedure as has been utilized in the literature - and is becoming routine in this pandemic environment. For example, the DSST would be presented on the screen as the examiner commences timing, rather than patients, some of whom are cognitively impaired, essentially timing themselves.
As described in the text, in order to ensure maximal comparability with the in person test, the participants did not time themselves, the examiner commenced the time when telling the participant to begin, after completing the practice portion of the DSST and explaining the test to the participants. Giving the key of the DSST on the screen would imply that the participant as to look at the screen, then look at their sheet to write their response, this will slow them down, and alter the DSST results which assesses
psychomotor speed.
There are online versions of the DSST, however since the original in-person DSST was administered on paper, we aimed at keeping it as similar as possible to minimize variability. Furthermore, in keeping the assessments as similar as possible, minimal computer skills were implied in the assessment, besides screen sharing for the FNAME. The test was in a sealed envelope, with clear instructions not to open prior to
instruction by the examiner, and once complete, the participants held up their DSST worksheet to the camera, so the examiner could mark how far they completed the DSST. The DSST is then immediately placed in the return envelope in front of the assessor.
I note that some patients had interrupted internet signal and confusion in setting up. This should be expanded upon.
The interrupted internet signal was adjusted for in analysis, as this was recorded during each videoconference (by recording the internet speed (mbps)). All participants were set a simple step by step instruction guide of how to use Skype or Zoom if they were not
familiar with any videoconferencing software. Once on the call, the participants did not have to do anything, as the research assistants facilitated the call with screensharing when necessary.
There are copyright protections for tests (eg MMSE, WMS) to prevent patients having access to test materials when there is no professional oversight. What is to prevent a patient from sharing this information with others? Did the authors have permission from test publishers to mail the stimuli?
We have checked with the London School of Hygiene and Tropical Medicine copyright team for permission. There are no issues with the MMSE worksheet, as all that was sent was a blank page with space for a sentence to be written, and a picture of two
overlapping pentagons, which hold no intellectual property. The WMS was only given orally, and the participants were watched to ensure they did not write anything, which they did not, along with the FNAME as well. The DSST worksheet was sent to the participants in a sealed envelope with clear instructions not to open prior to the assessment and was immediately returned after the completion of the remote assessment. Nonetheless we requested a retrospective license for using the test remotely, we appreciate the concern, and we will add a note in the future warning participants to not photocopy the worksheets and to return then as they were received. Given all of these potential confounds it is surprising that a correlation of 0.75 was found. If the sample size was sufficient and the design counterbalanced the conclusion would have greater credibility.
Of course, a larger sample size and counterbalanced study design would provide greater credibility to our study. The small sample size does decrease the power to detect
significant results in this instance, increasing the chance for an overestimation of the true effect estimate. However, Figure 2 and Figure 3 show there is an agreement between the two assessments. Given the COVID restrictions the researchers could not perform an appropriate counterbalanced design, however, think sharing the results found are still worthwhile in the field. See conclusions
5
Feasibility study of assessing the Preclinical
1
Alzheimer Cognitive Composite (PACC)
2
Score via videoconferencing
3
Giulia Seghezzo1, Yvonne Van Hoecke2, Laura James1, Donna Davoren2, Elizabeth Williamson2, Neil
4
Pearce2, Damien McElvenny2,3,4, Valentina Gallo1,2,5
5
6
1. Institute of Population Health Sciences, Queen Mary University of London, London, UK
7
2. London School of Hygiene and Tropical Medicine
8
3. Institute of Occupational Medicine, Edinburgh, UK
9
4. Centre for Occupational and Environmental Health, University of Manchester, Manchester, UK
10
5. Campus Fryslân, University of Groningen, Leeuwarden, the Netherlands
11 12 13 14 Corresponding author 15
Prof Valentina Gallo 16 Campus Fryslân 17 University of Groningen 18 Leeuwarden 19 The Netherlands 20 v.gallo@rug.nl 21 Click here to access/download;Manuscript;JOON_revised_clean.docx Click here to view linked References
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
Abstract
22Background
23
The Preclinical Alzheimer Cognitive Composite (PACC) is a composite score which can detect 24
the first signs of cognitive impairment, which can be of importance for research and clinical 25
practice. It is designed to be administered in person; however, in-person assessments are costly, 26
and are difficult during the current COVID-19 pandemic. 27
Objective
28
To assess the feasibility of performing the PACC assessment with videoconferencing, and to 29
compare the validity of this remote PACC with the in-person PACC obtained previously. 30
Methods
31
Participants from the HEalth and Ageing Data IN the Game of football (HEADING) Study who 32
had already undergone an in-person assessment were re-contacted and re-assessed remotely. The 33
correlation between the two PACC scores was estimated. The difference between the two PACC 34
scores was calculated and used in multiple linear regression to assess which variables were 35
associated with a difference in PACC scores. 36
Findings
37
Of the 43 participants who were invited to this external study, 28 were re-assessed. The median 38
duration in days between the in-person and the remote assessments was 236·5 days (7·9 months) 39
(IQR 62·5). There was a strong positive correlation between the two assessments for the PACC 40
score, with a Spearman correlation coefficient of 0·75 (95% CI 0·56, 0·95). The multiple linear 41
regression found that the only predictor of the PACC difference was the time between assessments. 42
Interpretation
43
This study provides evidence on the feasibility of performing cognitive tests online, with the PACC 44
tests being successfully administered through videoconferencing. This is relevant, especially 45
during times when face-to-face assessments cannot be performed. 46
47
Key Words:
48
Telemedicine, Cognitive Testing, Cognitive Decline, Mild Cognitive Impairment 49 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
Declarations
50Acknowledgements:
We are grateful to all the HEADING study participants for the time,51
interest, and commitment they have shown in contributing to the data collection. Thank you to
52
the Professional Footballers’ Association for their continued support with recruitment; We would
53
like to thank Prof Carol Brayne who has provided very valuable input on each of the phases of
54
the HEADING study by chairing the HEADING study Independent Oversight Committee (IOC).
55
Thanks also to the IOC Members for their invaluable advice and guidance throughout; Bill
56
Treadwell, Simon Jones, Dr. Collette Griffin, Professor Sinead Langan, Tim Lindsay, Lauren
57
Pulling, Tim Stevens, John Bramhall Richard Jobson and Charlotte Cowie. We would like to
58
thank Kirsty Lu and Sebastian Crutch as UCL assisted in facilitation and interpretation d the
59
neuropsychological tests. We are grateful to Ms Saba Mian who trained G Seghezzo, L James
60
and Y van Hoecke in administering the PACC.
61 62
Availability of data and material
: Dr Gallo had full access to all of the data in the study and 63takes responsibility for the integrity of the data and the accuracy of the data analysis. She 64
declares that this manuscript is honest, accurate, and transparent account of the study being 65
reported; that no important aspects of the study have been omitted. All co-authors had full access 66
to the data and can take responsibility for the integrity of the data and the accuracy of the data 67
analysis. 68
69
Funding
: This study was funded by the Drake Foundation as part of the BRAIN study funded 70to London School of Hygiene and Tropical Medicine (EPMSZO61) in collaboration with Queen 71
Mary University of London and the Institute of Occupational Health. The funder had no role in 72
the preparation of the manuscript, which has been written by the co-authors completely 73
independently. 74
75
Ethics Approval:
The HEADING Study was approved by the London School of Hygiene & 76Tropical Medicine’s Ethical Committee (16282). Written informed consent was obtained from 77
the participants, with further verbal consent to be re-assessed remotely. 78
79
Authors’ contributions
: 80Study concept and design: V Gallo, D McElvenny 81
Analysis and interpretation of data: G Seghezzo, D McElvenny, E Williamson 82
Drafting of the manuscript: G Seghezzo, Y van Hoecke 83
Data collection: G Seghezzo, Y van Hoecke, L James, D Davoren 84
Critical revision of the manuscript for important intellectual content: V Gallo, N Pearce 85
86
Conflict of interest statement
: We declare that we have no conflict of interests. 87 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 655
Introduction
88Dementia is a growing public health challenge, with an estimated 40-50 million people living with 89
this condition globally [1]. Worldwide, the prevalence of dementia more than doubled from 1990 90
to 2016, mainly due to the ageing population; it is now the 5th leading cause of death globally [1,2]. 91
Dementia onset is usually preceded by Mild Cognitive Impairment (MCI), with population-based 92
studies finding up to 22% of people with MCI developing dementia [1]. Currently, there is an 93
increasing interest in the early diagnosis of dementia, to allow potential screening programs, as 94
well as clinical trials testing disease-modifying drugs early on in the neuropathological process 95
[3]. In this context, assessing patients at very early stages of MCI is important. 96
97
The Preclinical Alzheimer Cognitive Composite (PACC) is a composite score which combines 98
tests that assess episodic memory, timed executive function, and global cognition, and it has been 99
shown to be able to detect the first signs of cognitive decline, before clinical signs of MCI manifest 100
[4]. The PACC score is increasingly used in epidemiological studies to assess an association 101
between exposures and early changes in cognitive function [5,6]. The PACC is designed to be 102
administered in person, by a trained research psychologist or nurse. However, in epidemiological 103
studies, in-person assessments are costly, often require extensive travelling, and are difficult in the 104
current pandemic situation. Assessing cognitive function in older adults may be possible via 105
videoconferencing, but there have been calls for further validation studies [7]. Some early studies 106
have shown that remote video assessments are feasible on cognitively normal participants, as well 107
as those with Alzheimer’s disease, dementia and Parkinson’s disease [8-12]. However, no study 108
has assessed the feasibility of performing videoconference assessments in the participant’s home 109
with their own equipment, with the majority of studies assessing feasibility by performing video 110
assessments in clinics [7,8, 11-13]. 111
112
The HEalth and Ageing Data IN the Game of football (HEADING) Study is an ongoing study 113
assessing the relationship between concussions and repetitive sub-concussive head injuries in 114
retired football players, and cognitive function as measured with the PACC score. In March 2020, 115
due to the COVID-19 pandemic, and the imposed lockdown by the UK government, the 116
HEADING Study could no longer assess its participants in-person, prompting a need to find other 117
modes of assessment. The aim of this study is to assess the feasibility of performing the PACC 118 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
score via videoconferencing and comparing the validity of the remote PACC score with the in-119
person PACC score obtained previously, by recalling participants of the HEADING study who 120
had already been assessed for a new remote assessment. 121
Methods
122Source population
123
Participants in the HEADING Study were selected from the Professional Footballer’s Association 124
(PFA) member database, a union for current and former professional football players of the English 125
Premier League. Any male member over the age of 50 with an address in England was sent an 126
invitation in the mail regarding the study, and a request to contact the study team to schedule an 127
appointment. Appointments were held in clinics in London or Manchester, or at the participant’s 128
home. The in-person assessment included a lifestyle questionnaire, exposure assessment 129
questionnaire, and cognitive tests, in addition to some physical measures, the assessment protocol 130
is similar to that of the BRAIN Study[5], apart from the addition of repetitive sub-concussive head 131
injuries to the exposure assessment. 132
133
The HEADING Study recruitment was ongoing when, on 23/03/2020 due to the COVID-19 134
pandemic, a lockdown in the UK was announced and in-person assessment were no longer 135
possible. 136
137
All participants who had already completed the in-person assessment for the HEADING Study 138
between July 2019 and March 2020 were contacted by telephone and/or by email, requesting their 139
voluntary participation in an additional remote assessment. Participants were asked if they had the 140
capability to perform video calls, by having access either a computer, tablet or smartphone with a 141
camera. Step-by-step instructions and over the phone support were offered to participants if they 142
were not familiar with downloading or using any videoconferencing software (instructions 143
provided for Skype and Zoom). If the participant agreed and met the requirements of joining a 144
video call, an appointment was scheduled with a video-conferencing software the participant was 145
most familiar with (Zoom®, Skype®, Microsoft Teams®, Facebook Portal® or FaceTime®). 146 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
The HEADING Study was approved by the London School of Hygiene & Tropical Medicine’s 147
Ethical Committee (16282). Participants were not involved in the design of the study nor of the 148 present sub-study. 149 150
In person assessment
151The PACC score used in the HEADING Study is based on that used in the British 1946 Birth 152
Cohort [6,14] and in the BRAIN Study [5], and consists of the following: 153
The Mini Mental State Examination (MMSE) total score (0-30 points): used to assess 154
multiple cognitive domains including orientation to time and place, attention and 155
calculation, recall, language, writing, visuospatial function, and executive function. 156
The total score of the 12-item Face-Name Associative Memory Test (F-NAME 12A) (0-157
96 points): used to assess the ability of the participant to recall names and occupation of a 158
number of people showed in pictures. 159
The delayed recall score on the logical memory IIa subtest from the Weschler Memory 160
Scale (0-25 story units): used to assess the ability to freely recall a short story. 161
The Digit Symbol Substitution Test (DSST) score from the Weschler Adult Intelligence 162
Scale Revised (0-93 symbols): used to assess attention and psychomotor speed. 163
Each of the four component scores was divided by the standard deviation (SD) of that component 164
to form standardised z-scores. The mean of these z-scores was then calculated to form the 165
composite score [4]. A complete PACC score for this study was defined as of having the MMSE 166
and at least two other tests completed [5]. 167
168
Remote Video Assessments
169
Prior to the assessment participant packs were posted to the participant address which contained 170
all materials necessary for their assessments. These included: (i) cover letter; (ii) blank paper (for 171
MMSE commands involving grabbing paper with right hand, folding paper and placing it on lap) 172
(iii) MMSE worksheets (draw pentagons and write sentence); (iv) DSST worksheet; (v) Post 173
Assessment Interview; (vi) stamped return envelope. 174
175
As the DSST was a timed task, the worksheet was enclosed in a sealed envelope within the 176
participant pack, with the following sentence ‘Please do not open until you are told to.’ Participants 177 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
then opened the sealed envelope when instructed to do so by the research assistant during the 178
remote assessment. Additional material available to each of the remote assessors included: (i) timer 179
for the MMSE, F-NAME, Logical memory test, and DSST; (ii) wristwatch for the MMSE; (iii) 180
Stimulus card for the MMSE; (iv) PowerPoint file for the F-NAME; (v) hard copy of the narrative 181
of the Logical Memory test; (vi) hard copy of the worksheet for the DSST; and (vii) pen for scoring 182
the tests and taking notes. 183
184
The order of the tests was changed slightly from the order of the in-person assessment to fit with 185
time restrictions required for the tests, as well as to ensure that the remote assessment was short 186
and did not include too many gaps between tests (Box 1): there was a 20-minute delay between 187
the Immediate Recall Logical Memory Test and the Delayed Recall Logical Memory Test; 188
similarly, there was a 30-minute delay between the Cued Face Name Associative Memory Test 189
and the Delayed Face Name Associative Memory Test. The tests were scheduled to take 60 190
minutes in total. 191
192
In addition, during the remote assessment, the participants were asked to check and report their 193
internet speed, using a website (www.fast.com). At the end of the call they were asked to complete 194
a post-assessment interview to record how they felt about the two assessments compared (in-195
person and remote). The post-assessment interview comprised of three questions, the first two 196
were ratings on a scale of 1-5 assessing how comfortable the participants felt with the in-person 197
and remote assessments (1 being very uncomfortable and 5 being very comfortable). The third 198
question was open-ended for the participants to give their opinion on the two assessments and if 199
they believed they were comparable. At the end of the assessment, the participant was asked to 200
place in the stamped return envelope the MMSE writing and drawing sheet, the DSST worksheet 201
and the post-assessment interview including a unique ID given to them to be identifiable to the 202
researchers, before they were returned. 203
204
A database was created on Excel which included the participant ID, the assessor from the in-person 205
assessment and the assessor for the remote assessment (different assessors were used for the in-206
person and remote assessments), internet speed of the participant, video software used, date of in-207
person and remote assessment, completed HEADING lifestyle questionnaire from the in-person 208 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
assessment, in-person PACC test scores and remote PACC test scores, and the post-assessment 209 interview. 210 211
Data Analysis
212Descriptive statistics were produced including means and medians and graphical displays of 213
distributions using histograms and scatterplots. Participants were included in the analysis if they 214
had a complete PACC score for both the remote and in-person assessments. Those who 215
participated in both assessments were compared to those who only performed in-person 216
assessments with descriptive statistics, such as mean and medians, Chi-square and t-test and 217
scatterplots. Since the PACC score is bases on the standardized test results of a sample, the PACC 218
score for the in-person assessment was calculated twice, first with all the participants who 219
completed only the in-person assessments, then again for the sample who completed both 220
assessments, the second to be used for the difference measure. The correlation between the two 221
PACC scores was estimated, and the difference between the two PACC scores was calculated. A 222
positive difference implies the remote PACC score is higher than the in-person score, and a 223
negative difference represents a higher in-person score. This difference measure was then used in 224
a multiple linear regression to assess the role of variables potentially associated with a difference 225
in PACC scores. The time elapsed between the two tests was modelled as both a categorical and 226
continuous variable in order to explore a possible effect of time. Continuous variables (age, time 227
and internet speed) were centered on the mean for the regression analysis. To better interpret the 228
results of the regression, a marginal effect plot was explored on the mean PACC difference in the 229
sample by varying time between assessments. Differences in scores of the individual tests 230
comprising the PACC were also analyzed separately with the same approach. Agreement of the 231
two measures was further assessed with a Bland-Altman plot. All analyses were further run without 232
an identified outlier. 233
Results
234As of March 13th, 2020, 45 participants had been assessed for the HEADING Study. These 45 235
participants were invited to take part in the remote assessment; 31(69%) agreed to participate in 236
the remote assessment. At the time of this analysis, there were 30 participants who had completed 237
a virtual PACC assessment and had data available for the in-person assessment (67% of the total). 238
Of the 30 participants in the feasibility study, two participants completed only two of the four tests, 239 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
with one participant suffering from aphasia and another having recently undergone hand surgery 240
inhibiting their ability to do the DSST and half the MMSE. Therefore, leaving 28 participants 241
(Figure 1). The median age was 60 years old (IQR 16) (Table 1), with 57% of the participants 242
being educated up to GCSE standard. Only 10% of the participants had ever smoked, and 75% of 243
the participants drank alcohol. A comparison between the original sample, and the participants 244
included in the remote assessment is reported in Table 1. Participants who accepted to be re-245
assessed were on average younger (p=0.03), more educated (p=0.02) and had a higher PACC score 246
(p=0.05). 247
248
The shortest time between assessments spanned 103 days, while the longest time between 249
assessments was 293 days (3·4 and 9·6 months, respectively). The median duration in days 250
between the in-person and the subsequent remote assessments was 236.5 days (or 7.9 months) 251
(IQR 62·5). When the time between assessments variable was categorized, 5 (17%) participants 252
had less than 149 days (4·9 months) between assessments, three (10%) between 150-199 days (4·9 253
– 6·5 months), 12 (40%) ranged between 200-249 days (6·6 – 8·2 months) and ten (13%) over 250 254
days (8·2 months) between assessments. Most of the remote assessments (80%) were performed 255
using applications Skype and Zoom. 256
257
The PACC scores for the two assessments are plotted in Figure 2. There was a strong positive 258
correlation between the two assessments for the PACC score, with a Pearson correlation 259
coefficient of 0·82 (95% CI 0·66, 0·98). Summary statistics for the PACC scores and the PACC 260
difference are shown in Table 2. A Bland-Altman plot was further used to assess agreement 261
between the two PACC scores (Figure 3). This suggests that the difference between in-person and 262
remote assessments are not detected differentially in those with higher or lower scores. 263
264
The multiple linear regression included the age of participants, highest educational qualification, 265
internet speed, in-person and remote assessors, and time between assessments for 27 participants 266
due to a missing value for internet speed (Table 3). The regression produced a constant value of -267
0·17 (95% CI -0·54, 0·19), meaning the expected PACC difference between the in-person PACC 268
was 0·17 points higher than the remote PACC for an ‘average’ person in the sample (specifically, 269
a person aged 61 years, with GCSE education level, assessed by video assessor 1, person assessor 270 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
4, with an average time of 218 days between assessments and with an internet speed of 31mbps). 271
The time between assessments, as a continuous variable, was identified to be associated with the 272
PACC difference. Time between assessment predicted a decrease in PACC difference by -0·004 273
points (β=-0·004 95% CI: -0·007, -0·00008) with increasing time between assessments (in days). 274
This means that the difference in PACC scores between each increasing day will differ on average 275
by 0.004 points. The analysis of marginal effects showed that when the two tests were administered 276
relatively closer in time, the mean difference was positive (the subsequent remote test performance 277
was better), but the difference became negative (in-person performance better) with increasing 278
time difference between the assessments (Figure 4). When analyzing the individual tests, time 279
between assessments was also identified with the Logical Memory test (data not shown). The 280
analysis run after removing the outlier did not change the results, as shown in Table 3. The 281
responses of the post-assessment questionnaire are displayed in Supplemental Table 1. 282
283
Overall, the participants reacted well to the remote assessment, with 22 (78%) participants 284
responding that they felt extremely comfortable performing the remote assessment on the post-285
assessment interview, with 3 participants scoring the remote assessment worse than the in-person 286
assessment (Supplemental Table 1). Participants mentioned that although there are more problems 287
with remote assessment (interrupted internet signal, confusion in setting up) the testing process 288
felt equal to that already performed in-person. 289
Discussion
290This study provides evidence of the feasibility of administering the tests comprising the PACC 291
score via videoconferencing: administering the tests sending the participants some material via the 292
post, in advance was shown to be feasible. The differences between the in-person and remote-293
administered PACC were overall very small. There weren’t systematic differences between the 294
two PACC scores, arguing against a potential bias introduced by the remote assessment: those who 295
performed worse presented similar in-person/remote differences compared to those who 296
performed best. 297
298
The only variable which predicted the PACC score differences was the time between assessments; 299
its association with the PACC difference being very small. The marginal effect plot showed a 300 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
tendency of better performance at the remote assessment, on average, when the time difference 301
between the two was shorter, but this association reversed as the time between assessments 302
increased. This could be interpreted as a potential learning effect - given that all the participants 303
had already undergone the face-to-face assessment prior to the remote assessment- which wears 304
off over time [15,16]. A learning effect involves increases in repeated test scores due to factors 305
such as memory for specific test items, learned strategies for problem solving or general experience 306
and comfort with testing. This learning effect was observed despite the average time between tests 307
was 7·2 months in this study, which is longer than the two to three months other studies have used 308
when considering learning effects [7,10] suggesting that the learning effect in substantially 309
cognitively intact people could be longer than previously recognized. This difference between the 310
two measures could also be interpreted as participants being in a more comfortable setting in their 311
home performing better [17]. 312
313
The fact that the learning effect wears off over time is predictable, less easy to interpret is a 314
tendency toward a reverse effect (remote assessment worse than in person). One possible 315
interpretation is that it is a detection of very subtle cognitive decline over time among the 316
participants. The BRAIN study has shown that PACC scores decrease with age among retired 317
rugby players (manuscript under review) [5]. Given the long test-retest interval between the two 318
assessments, this could be a possibility. The PACC was originally established to measure cognitive 319
decline over time, being administered every 6 months over the course of 36 months. However, an 320
average of seven months between the assessments may not be long enough to detect a change in 321
cognitive function, as seen by Donohue et al., where the earliest detectable change in PACC scores 322
was at 12 months [4]. Finally, it is not possible to rule out completely that different assessors for 323
the in-person and remote assessments might have had an effect on the scoring, although this was 324
also adjusted for in the analysis. 325
326
Importantly, the effect of time on the scoring difference is no relevant for the HEADING and many 327
other epidemiological studies which use the remote PACC assessment solely in person, or 328
remotely. The high correlation between the two sets of tests and the absence of a clear bias 329
affecting disproportionally people performing less well, suggest that this is a valid method to be 330 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
used in epidemiological studies on a given population with high computer literacy and cognitively 331
relatively integer. 332
333
Participants accepted well the remote cognitive testing finding it comfortable to be assessed 334
remotely, nonetheless all of them had already undergone the assessment, so they knew what to 335
expect. It remains to be explored if participants assessed remotely only would self-rate the 336
assessment as comfortable as well. 337
338
Limitations
339
The response rate to reassessment was not ideal (65%), introducing a potential for selection bias, 340
as those who did not participant were different to those who did. For example, the sample who 341
agreed to participate declared to be more adept in using technology and have videoconferencing 342
devices and internet connection available to them. Of those who declined to participate, two 343
mentioned they either did not have a device for videoconferencing, or they had no internet 344
connection. Studies found that participants who are less computer literate have increased computer 345
anxiety, which could affect scores of computer tests [9, 11]. Therefore, this makes our results not 346
immediately generalizable to a less technologically confident population as the same correlation 347
may not be found. Likewise, participants who did not agree to participate had a longer interval of 348
time between assessments, with a median of 266 days (q25, q75: 206, 268). It is unlikely that 349
cognitive status changed over the course between the in-person assessment and the retest, however 350
this is a possibility. The median PACC of the in-person assessments was lower in those who did 351
not agree to participate in the remote assessment, compared to those who did (p=0.05). This would 352
potentially introduce a bias towards a more cognitively able population in the results. 353
354
Contextualization of results
355
The present results are in line with previous studies comparing face-to-face and virtual assessments 356
[7-13, 18-20]. Telemedicine is a growing field, becoming more relevant particularly with regards 357
to assessing cognitive function, as it has an advantage of reaching more participants and reducing 358
the burden of lengthy travelling, cutting time, cost and making the participant feel more 359
comfortable in their own home [7,11,19,20]. Moreover, telemedicine can be used to reduce in-360
person contact to abide by recent government guidelines by increasing the possibility of social 361
distancing as well as the security of being able to reach at risk participants during these times. In 362 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
5
this study, the use of videoconferencing was chosen over telephone assessment because the PACC 363
assessment involved the need to see the participant performing tasks, as well as sharing the screen 364
for the Face-Name Association Task. Nonetheless, videoconferencing can be seen as more 365
insightful because it captures non-verbal cues that can’t be done in telephone interviews such as 366
facial expressions and attentiveness [7]. 367
368
Conversely, compared to face-to-face assessments, remote assessments have some disadvantages, 369
such as loss of attention due to surroundings as well as the potential for participants writing down 370
answers or looking at a calendar. For instance, this study noted one participant being distracted 371
during the video call by their surroundings at home, while another participant received a phone 372
call during the digit symbol substitution test. The analysis accounted for internet speed to adjust 373
for potential connection problems that could have interfered with the assessment. Furthermore, the 374
analysis took into account the software used for the videoconferencing, however what was not 375
taken into account was the device used, as the visual cues, particularly with the FNAME, could be 376
altered on a smartphone compared to on a computer, as the stimuli would be smaller. Stillerova et 377
al. assessed remote testing with different software and different devices and found no difference 378
among modes used [9]. The marking of the overall scores can also be adjusted for video 379
assessment, such as Timpano et al. lowering the cut off for the virtual MMSE, to account for poor 380
internet speed and other factors that could influence the assessment [8]. Other potential problems, 381
such as writing down questions and changing answers, were addressed by ensuring that 382
participants showed their responses for the MMSE writing and drawing tasks and for the digit 383
symbol substitution test. Besides the disadvantages that may arise from remote assessments, this 384
study also had a limitation of a small sample size, thus introducing variability in the results and 385
reducing statistical power, limiting the ability for clear interpretation of the results. This reduction 386
in power is also denoted by the large confidence intervals seen in the correlation coefficient, 387
preventing the reliability of the results to be generalized to a broader audience. Further non-linear 388
trends of the effect of time could therefore not be explored given the small sample size. A 389
sensitivity analysis removing the outlier did not affect the results. 390 391 392 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65