jpad journal

AND option

OR option

A UK-WIDE STUDY EMPLOYING NATURAL LANGUAGE PROCESSING TO DETERMINE WHAT MATTERS TO PEOPLE ABOUT BRAIN HEALTH TO IMPROVE DRUG DEVELOPMENT: THE ELECTRONIC PERSON-SPECIFIC OUTCOME MEASURE (EPSOM) PROGRAMME

 

S. Saunders1, G. Muniz-Terrera1, S. Sheehan2, C.W. Ritchie1,3, S. Luz2

 

1. Centre for Clinical Brain Sciences, University of Edinburgh, UK; 2. Usher Institute of Population Health Sciences and Informatics; Molecular, Genetic and Population Health Sciences, University of Edinburgh, UK; 3. Brain Health Scotland, UK

Corresponding Author: Stina Saunders, University of Edinburgh, Centre for Clinical Brain Sciences, UK, stina.saunders@ed.ac.uk

J Prev Alz Dis 2021;
Published online June 9, 2021, http://dx.doi.org/10.14283/jpad.2021.30

 


Abstract

BACKGROUND: It is important to use outcome measures for novel interventions in Alzheimer’s disease (AD) that capture the research participants’ views of effectiveness. The electronic Person-Specific Outcome Measure (ePSOM) development programme is underpinned by the need to identify and detect change in early disease manifestations and the possibilities of incorporating artificial intelligence in outcome measures.
Objectives: The aim of the ePSOM programme is to better understand what outcomes matter to patients in the AD population with a focus on those at the pre-dementia stages of disease. Ultimately, we aim to develop an app with robust psychometric properties to be used as a patient reported outcome measure in AD clinical trials.
Design: We designed and ran a nationwide study (Aug 2019 – Nov 2019, UK), collecting primarily free text responses in five pre-defined domains. We collected self-reported clinical details and sociodemographic data to analyse responses by key variables whilst keeping the survey short (around 15 minutes). We used clustering and Natural Language Processing techniques to identify themes which matter most to individuals when developing new treatments for AD.
Results: The study was completed by 5,808 respondents, yielding over 80,000 free text answers. The analysis resulted in 184 themes of importance. An analysis focusing on key demographics to explore how priorities differed by age, gender and education revealed that there are significant differences in what groups consider important about their brain health.
Discussion: The ePSOM data has generated evidence on what matters to people when developing new treatments for AD that target secondary prevention and therein maintenance of brain health. These results will form the basis for an electronic outcome measure to be used in AD clinical research and clinical practice.

Key words: Clinically meaningful change, electronic patient reported outcome measures, Alzheimer’s disease, outcome measures, brain health.


 

 

Introduction

Attempts to develop disease modifying therapies for Alzheimer’s disease (AD) started over 20 years ago with little success to date. A recent estimate of the costs of AD was US$818B, which is equivalent to the combined GDP of Indonesia, The Netherlands, and Turkey (1).
The lack of progress in finding a pharmacological treatment for AD is however at odds with a rapid development in the understanding of the pathology of AD suggesting that clinical trial design and delivery may partially account for a lack of progress with insensitive outcome measures lacking clinical meaningfulness also playing a part in this lack of progress. It has been shown that the disease process starts long before an individual becomes symptomatic or eventually, the dementia syndrome manifests (2, 3). Increasingly, we are exploring AD processes at earlier disease stages through examining at-risk populations in mid-life which helps identify the earliest manifestation of declining brain health. In the absence of pharmacological interventions, it is estimated that approximately 40% of dementia cases could be prevented by targeting epidemiologically derived modifiable risk factors (4). Changes occurring years earlier than dementia develops have been observed in at-risk populations using exploratory and sensitive computerised tests assessing e.g. allocentric and egocentric spatial processing (5). These test results correlate with brain imaging findings in hippocampal subfields known to be sensitive to amyloid derived neurotoxicity (6); as well as in changes to brain β-amyloid in at risk populations aged between 63-81 years old who did not have dementia (7).
Whilst there are global initiatives focusing on dementia prevention through risk factor modification (8, 9), there remains a major and complementary need for effective AD pharmacological interventions. Irrespective of the type of intervention to reduce incident dementia rates, the fact is that these studies will engage at risk populations who will be, to the most part, in mid-life and healthy. Currently, there are 31 AD drugs being tested in Phase III clinical trials (19 of which are disease modifying) (10). We argue that using outcome measures assessing clinical symptoms and functioning in earlier disease stages is less valid than biological measures of disease and what the individual considers personally meaningful from a treatment. A treatment’s success should therefore be determined not only by the impact on the individual’s disease (as evidenced by biomarker change) but also by its effect on related well-being (as measured by patient reported outcomes).
To this end, whilst it is currently proposed by regulators that AD trials measure cognition as the primary outcome, as trials move to an earlier disease stage it could be argued that many commonly used (cognitive) measures lack ecological validity and are not sensitive enough to detect changes in the earlier stages of the AD continuum where the ideal intervention should take place (11). Moreover, it is recommended by both the US Food and Drug Administration (FDA) (12) and European Medicines Agency (EMA) (13) that AD trials incorporate measures which capture clinically meaningful results to the individual. Patient reported outcome measures (PROMs) are developed for the incorporation of the person’s own perspective regarding their treatment, though these measures are currently not used in AD clinical trials (14). PROMs reflect an individual’s view on what they define as an effective treatment and consider a meaningful change. Notably, PROMs are already more widely used in other disease areas. For example, a recent study of nearly 100,000 clinical trials published on clinicaltrials.gov found that a PROM had been used in 27% of all trials, primarily in oncology (15).
In light of the drive towards early detection, looking at younger at-risk populations and the main regulators’ recommendation for clinically meaningful outcome measures, we have established the electronic Person Specific Outcome Measure (ePSOM) development programme. As the target population in dementia prevention research is an at-risk population, our group took the view that what matters to people when developing new treatments for AD is approached by way of maintenance of brain health (16, 17). The ePSOM programme consists of four sequential steps, ultimately aiming to employ new technology to create an outcome measure to be used in AD clinical research and practice. This will be in the form of an outcome app used on any screen-based device which will assess aspects specific to the individual using it. At the start of the programme, we reviewed literature around PROMs in AD clinical trials which informed our focus group study with people with memory concerns, healthy volunteers and health care professionals (18). The focus group study yielded five domains of importance for what matters to people about brain health. These domains formed the basis for the next stage of the ePSOM development programme. In this paper, we report on a large UK-wide population-based study to understand what matters to people when developing new treatments for Alzheimer’s disease. We consider the respondents to the ePSOM study a representative population of individuals who may be enrolled in dementia primary and secondary prevention clinical trials and characterise what matters to people about brain health focusing on key demographic groups.

 

Methods

We designed and ran a UK-wide population-based online study collecting primarily free text answers (see Appendix 1). The study built on a previously run focus group study which yielded five domains of importance about brain health. The study obtained ethics approval from the ACCORD Medical Research Ethics Committee in Edinburgh, Scotland. The ePSOM study ran from Aug 2019 – Dec 2019 and was divided into sections, starting with an introductory video and informed consent.
Free text answers were collected across five pre-defined domains. These answers were clustered, leading to specific themes of what matters to people about brain health

The study was open to anyone over the age of 18 and was launched primarily via Alzheimer’s Research UK media channels through e-mails to individuals registered on their mailing lists and a social media campaign (with social media support from other dementia related organisations). We collected key sociodemographic and clinical data such as having been seen by a doctor because of any brain health issues though the primary method of the survey used a qualitative approach. Respondents were presented with the five domains derived from the earlier focus group work to orientate and channel free text responses: [1] Everyday functioning; [2] Sense of Identity; [3] Relationships and Social Connections; [4] Enjoyable Activities and [5] Thinking problems. They were then asked to provide free text answers on what they would like to retain or keep being able to do in those domains if their brain health got worse. At the end of the study, respondents were asked to identify five answers across all the answers they had given which they consider the most important. We used Natural Language Processing (NLP) techniques to analyse the free text data (see Figure 1).

Figure 1. Natural Language Processing techniques used to analyse the survey data

 

Free text answers were collected across five pre-defined domains. These answers were clustered, leading to specific themes of what matters to people about brain health

Step 1: Natural Language Processing to create clusters

We used NLP to create clusters of semantically similar free text answers. These clusters were then manually annotated with appropriate labels. We refer to the finally labelled clusters derived from this stepped NLP-manual annotation process as “themes”.
NLP employed word embeddings trained on vast amounts of text data to achieve fine-grained representation of semantic regularities in text. We were thus able to build robust representations of free text answers. To begin, “stop words” (i.e. words that occur very frequently and contribute little to semantic content) and punctuation were removed from the free text answers. The resulting texts were then converted to numerical vector representations by using GloVe vectors (19) to generate sentence embeddings. These vectors encode semantic relationships between words and can be used to cluster semantically similar text segments. This allowed us to use automated methods to identify words, and thus answers, of a similar “theme” or meaning. The K-means clustering algorithm was used to cluster the answer embeddings within each of the five domains. The K parameter, that is, the desired number of automatic clusters per domain, was determined analytically. The goal was to generate fine grained clusters which contain semantically similar answers while avoiding overfitting or creating so many clusters that important themes are not revealed. We found when the number of important items in the largest cluster changes by less than 10, between each of the previous five increments of K, that the majority of the clusters also exhibited minor changes in the number of important items. Using this criterion, we chose a value of 151 clusters across all five domains. This method resulted in a total of 755 clusters of free text answers, or 151 clusters for each of the five domains.

Step 2: Manual Annotation to create themes

The clusters that emerged within each of the five domains were reordered so that semantically similar clusters appeared close together. This was achieved using hierarchical clustering on the cluster centroids. We used the reordered clusters for manual annotation in each of the five domains. Each cluster was represented by the 200 most frequent unique answers, after punctuation and stopwords were removed. The annotation goals were to combine any clusters which fit together, exclude uninterpretable clusters and label the final clusters thus deriving the final themes. Six authors of the current paper annotated two domains each, ensuring two separate people analysed a single domain, which helped ensure inter-rater reliability between domains. Finally, two of the authors did quality control across the five domains and homogenised the labels across domains.

Statistical analyses

In this paper, we focus our analyses on key demographic groups: age (up to the age of 64 / age 65 and older); gender (men / women) and education (no degree / degree and higher). We present the largest themes for each of these demographic groups as well as themes which were identified as particularly important most often in the final question on the study forms. For both of these analyses, we report percentages for each theme by key demographic groups. As the demographic groups are unbalanced in terms of the number of respondents we use percentages rather than the absolute number of answers in the statistical analyses. The percentages are derived by dividing the count of answers within the demographic group by the total number of answers in that demographic group, thus providing proportions for comparison when dealing with imbalanced demographics. It should be noted that respondents were not bounded by a minimum or maximum number of free text answers they could give in each domain.
Finally, we conducted a Chi-squared test to analyse whether the differences in percentages between demographic groupings’ answers within each theme were statistically significant. A p-value of <0.01 was used in statistical significance testing.

 

 

Results

The study was completed by 5,808 people from across the UK. They provided a total of 82,514 free text answers. These were clustered using automated NLP techniques resulting in 151 clusters in each of the five pre-defined domains, a total of 755 automated clusters across all domains, as described. Subsequent analysis reduced the number of clusters to 334 (due to a cluster being represented in two or more domains) which were all manually annotated by the research team. Many of the same themes emerged from different domains (e.g., the theme of Walking in the “Enjoyable activities” domain as well as the “Everyday activities” domain). After merging themes with the same label in different domains, the final number of unique themes was 184. Some respondents used more generic language (e.g., “Maintaining independence”) whereas others were more specific (e.g., “Driving”). Using NLP methods for free text analysis means that, in this example, the “Maintaining independence” theme contains 1100 answers, most containing either the word “independent” or “independence”. Analytically, this is therefore not a general theme for answers which relate to the concept of independence, but a cluster of answers in which the respondents are directly referencing the word independence as something which is important for them to maintain. This has therefore resulted in themes which are more or less specific but directly reflect the language used by the respondents.
Pre-defined answers: Characteristics of the ePSOM survey sample

The characteristics of the 5,808 respondents are presented above (see Table 1).

Table 1. ePSOM survey respondent characteristics

We used NLP techniques and manual annotation to group individual free text answers into clusters and then themes respectively. The most frequent themes across all demographics were reading, driving, friendships and following a storyline (Figure 2). We also calculated the proportion of answers within each key demographic, expressed as percentages of the total answers given by that demographic group.

Figure 2. What matters to people about brain health? The survey received 82,514 free text answers which were clustered into 184 themes

This figure shows themes which were mentioned the most, broken down by key demographics. Full figure of the survey themes in Appendix 2.

At the end of the survey, respondents were asked to identify the five most important answers to them across all their answers. We used this metric to rank the themes in terms of being selected as particularly important and observed a subtle difference between the largest themes (themes mentioned the most frequently) and themes which are identified as the most important. The 5 top important themes across all demographics were family connections, driving, socializing, reading and friendships (Figure 3).

Figure 3. What matters to people about brain health?

This figure shows themes with the highest number of answers selected as particularly important by key demographics. Full figure of themes with the most important answers in Appendix 3.

 

Cross-Tabulations of key demographics

The following tables show statistically significant proportional differences in theme sizes (Table 2) and identifying themes as particularly important (Table 3), focusing on demographic group dyads (younger vs older; men vs women; individuals with no degree vs individuals with a degree or higher).
Table 2 Top 10 themes selected as particularly important which had the highest Chi square values representing greater differentiation between demographic groups by age (younger/older); gender (male/female) and education (no degree / degree and higher). A full list of particularly important themes which were significantly different across key demographics can be found in Appendix 4.

Table 2. Top 10 themes selected as particularly important which had the highest Chi square values representing greater differentiation between demographic groups by age (younger/older); gender (male/female) and education (no degree / degree and higher)

Full list of particularly important themes which were significantly different across key demographics in Appendix 4.

Table 3. Top 10 largest themes which had the highest Chi square values representing greater differentiation between demographic groups by age (younger/older); gender (male/female) and education (no degree / degree and higher)

Full list of largest themes which were significantly different across key demographics in Appendix 5.

 

Discussion

Building on the scientific foundation provided by previous stages of the ePSOM research programme, we designed and ran a nationwide study with open ended questions to derive free text answers exploring what matters to people about maintaining their brain health within five focus group-derived domains. To our knowledge, this is the first study collecting free and systematically analysing text responses from a very large number of respondents on what is important to them about brain health. The themes and granularity derived from our study are in line with the FDA’s guidance for capturing aspects relevant to AD research participants “e.g., [assessing] facility with financial transactions, adequacy of social conversation” (12).
As AD drug development moves to an earlier phase of the neurodegenerative disease spectrum and clinical research targets an earlier, younger population, it is crucial any outcomes are meaningful and relevant to that trial population. Additionally, as upcoming AD treatments are hoped to be disease modifying rather than reducing symptoms, the cognitive domains which respond to the medication may not be the same as with symptomatic treatments measured at a later disease stage (20). We also know from a recent review that lifestyle factors may influence brain health in midlife (21) so it is apposite to examine what matters to people about brain health including lifestyle dependant factors as this will be increasingly relevant in Brain Health Clinics which are developing throughout the UK (16) and Europe (17).
There has been other work collecting evidence on important outcomes focusing on the point of view of people living with dementia (22). The focus of the ePSOM programme though is the maintenance of brain health. As the majority of the individuals in our study had not received a diagnosis of neurodegenerative disease, the findings from our study provide evidence for what matters to people about brain health in normal lived experience which may include people at the earlier (asymptomatic) stages of disease rather than once the dementia syndrome develops. Our findings are supported by literature recognising that AD trials currently do not measure outcomes which are relevant to the patient themselves. Tochel et al. (23) carried out a literature review extracting data from studies where participants described outcomes which matter to them. Their review concluded by demonstrating an array of outcomes which are not commonly captured in clinical trials of new treatments (23).
Changes at the early stages of the AD continuum are currently detected by biomarker assessments, with functional measures used increasingly towards the more symptomatic and advanced stage of the continuum where ultimately impairment is evidenced in basic activities of daily living. However, dementia prevention cohorts have found differences in more than just biomarker assessed pathology, e.g. there is evidence that middle-aged adults at risk of dementia have poorer cognitive performance, principally in visuospatial functions (24) and memory (25). Lau et al. (26) concluded that observing early functional limitations at baseline in the at-risk population had prognostic value in identifying older adults at risk for developing functional disability a few years later (26).
A recent review also concluded that in the pre-dementia stages of AD, executive functions (such as inhibitory abilities), attentional and visuospatial functions can already be impacted (27). A PROM therefore could be viewed as an ecologically valid instrument for cognitive assessment measures which are proxies for what matters to people, especially if the PROM relates to a cognitive process affected early in the course of AD (e.g. activities requiring planning, judgement or navigation/orientation like confidence driving). The key questions here is: if an individual’s score changes on a particular domain using a cognitive assessment measure, does this correlate with a change of score in a PROM and is therefore a change meaningful (by definition) to the patient? While functional or Activities of Daily Living scales measure a more direct or practical effect a drug may have, these measures have limitations such as poor psychometric properties (28) and as evidenced by the analysis of key demographic groups in the ePSOM survey, what matters to people about brain health and their function is different depending on age, sex and education levels. By capturing data specific to the individual who in effect derives their own outcome measure, the ePSOM app in development would present an outcome measure for clinical trials that captures changes noticed by and meaningful to the person themselves and therefore more likely to be correlated to their own specific functional outcomes than generic outcomes which were derived by homogenising population level data. Ultimately, employing more meaningful, ecologically valid and sensitive measures will facilitate more drugs to be approved by regulatory bodies which will actually impact on well-being and not just impact on cognition and function ‘on average’ between groups (29). Moreover – ePSOMs are immune to cultural, educational and language variability as each outcome is unique to that individual and bears no reference to an external ‘population norm’.
We used an online study design as it was important to allow for free text answers and reach a large number of people. However, this is also a limitation in the study leading to inevitable sampling bias of individuals who are able to access an online survey. There was also a demographic imbalance among the survey respondents with reference to the UK population as a whole, but appropriate analysis focusing on proportions rather than absolute values of this relatively large sample mitigates the effects of the imbalances in the data. The main strength of the study was collecting free text answers and using NLP techniques in the data analysis. Employing NLP techniques to gather evidence for what outcomes matter in AD drug development is unique and we are not aware of any similar studies. Free text answers offer insights which go beyond rating themes on a scale which have been predefined as important by the researchers and are culturally biased and limited. Moreover, the open character of the questions may motivate respondents to reveal more (31). In some regards, our study results may be considered comparable to hundreds of focus group studies, though by using NLP techniques, we are able to extract patterns in answers by key demographic at a scale and level of detail not feasible using traditional qualitative methodologies.

 

Conclusion

There is a growing consensus that PROMs should be used in AD trials so that the patient can assess if they observe a change in their well-being which is meaningful and specific to them. Including the patient’s perspective is also recommended by regulatory bodies such as the EMA with whom we collaborated in the initial phases of this project, and the FDA. In our study, we included a large number of people collecting free text responses to understand what matters to people about their brain health – our analyses focussed on key demographic groups. This approach is novel in so much as it uses NLP approaches to create a range of outcomes from a theoretically limitless range of possible responses and then can apply these into quantifiable and ecologically valid outcomes. The main criticism and in many ways fatal flaw of current approaches to PROMs is that they are derived at a population level and therefore have to incorporate the characteristics of the population they were derived from. These populations will hold certain language, cultural and ethnic characteristics making their use in other limited in other populations. The ePSOM app will ultimately be used by people in earlier stages of neurodegenerative disease before dementia develops in populations across the globe, in clinical trials with seamless translation into clinical practice.

 

Acknowledgement: We thank Alison Evans from Alzheimer’s Research UK for her intellectual contribution to this study. We also thank the individuals who took part in the ePSOM study.

Conflicts: The authors declare no conflict of interest.

Funding Sources: The ePSOM survey was funded by Alzheimer’s Research UK.

Declarations of interest: none (all authors).

Ethical standards: The study obtained ethics approval from the ACCORD Medical Research Ethics Committee in Edinburgh, Scotland.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

 

SUPPLEMENTARY MATERIAL1

SUPPLEMENTARY MATERIAL2

SUPPLEMENTARY MATERIAL3

SUPPLEMENTARY MATERIAL4

SUPPLEMENTARY MATERIAL5

 

References

1. Wimo A, Guerchet M, Ali G-C, Wu Y-T, Prina AM, Winblad B, et al. The worldwide costs of dementia 2015 and comparisons with 2010. Alzheimer’s & dementia : the journal of the Alzheimer’s Association. 2017;13(1):1-7. doi: 10.1016/j.jalz.2016.07.150
2. DeTure MA, Dickson DW. The neuropathological diagnosis of Alzheimer’s disease. Molecular Neurodegeneration. 2019;14(1):32. doi: 10.1186/s13024-019-0333-5
3. Association As. Alzheimer’s disease facts and figures. Alzheimer’s & Dementia. 2020;16(3):391-460. doi: 10.1002/alz.12068
4. Livingston G, Huntley J, Sommerlad A, Ames D, Ballard C, Banerjee S, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet. 2020;396(10248):413-46. doi: 10.1016/s0140-6736(20)30367-6
5. Ritchie K, Carrière I, Howett D, Su L, Hornberger M, O’Brien JT, et al. Allocentric and Egocentric Spatial Processing in Middle-Aged Adults at High Risk of Late-Onset Alzheimer’s Disease: The PREVENT Dementia Study. Journal of Alzheimer’s Disease. 2018;65:885-96. doi: 10.3233/JAD-180432
6. Brien JT, Firbank MJ, Ritchie K, Wells K, Williams GB, Ritchie CW, et al. Association between midlife dementia risk factors and longitudinal brain atrophy: the PREVENT-Dementia study. Journal of Neurology, Neurosurgery &amp;amp; Psychiatry. 2020;91(2):158. doi: 10.1136/jnnp-2019-321652
7. Martikainen IK, Kemppainen N, Johansson J, Teuho J, Helin S, Liu Y, et al. Brain β-Amyloid and Atrophy in Individuals at Increased Risk of Cognitive Decline. AJNR Am J Neuroradiol. 2019;40(1):80-5. doi: 10.3174/ajnr.A5891
8. Ritchie CW, Muniz-Terrera G, Kivipelto M, Solomon A, Tom B, Molinuevo JL, et al. The European Prevention of Alzheimer’s Dementia (EPAD) Longitudinal Cohort Study: Baseline Data Release V500.0. J Prev Alzheimers Dis. 2020;7(1):8-13. doi: 10.14283/jpad.2019.46
9. Cummings J, Aisen P, Barton R, Bork J, Doody R, Dwyer J, et al. Re-Engineering Alzheimer Clinical Trials: Global Alzheimer’s Platform Network. J Prev Alzheimers Dis. 2016;3(2):114-20. doi: 10.14283/jpad.2016.93
10. Devenish SRA. The current landscape in Alzheimer’s disease research and drug discovery. Drug Discov Today. 2020;25(6):943-5. doi: 10.1016/j.drudis.2020.04.002
11. Snyder PJ, Kahle-Wrobleski K, Brannan S, Miller DS, Schindler RJ, DeSanti S, et al. Assessing cognition and function in Alzheimer’s disease clinical trials: do we have the right tools? Alzheimers Dement. 2014;10(6):853-60. doi:10.1016/j.jalz.2014.07.158
12. FDA UFaDA. Early Alzheimer’s Disease: Developing Drugs for Treatment Guidance for Industry. 2018.
13. EMA EMA. Guideline on the clinical investigation of medicines for the treatment of Alzheimer’s disease. Committee for Medicinal Products for Human Use (CHMP). 2018:1-36.
14. Saunders S, Muniz-Terrera G, Watson J, Clarke CL, Luz S, Evans AR, et al. Participant outcomes and preferences in Alzheimer’s disease clinical trials: The electronic Person-Specific Outcome Measure (ePSOM) development program. Alzheimers Dement (N Y). 2018;4:694-702. doi:10.1016/j.trci.2018.10.013
15. Vodicka E, Kim K, Devine EB, Gnanasakthy A, Scoggins JF, Patrick DL. Inclusion of patient-reported outcome measures in registered clinical trials: Evidence from ClinicalTrials.gov (2007-2013). Contemp Clin Trials. 2015;43:1-9. doi:10.1016/j.cct.2015.04.004
16. Ritchie CW, Russ, T. C., Banerjee, S., Barber, B., Boaden, A., Fox, N. C., . . . Burns, A. The Edinburgh Consensus: preparing for the advent of disease-modifying therapies for Alzheimer’s disease. Alzheimers Res Ther. 2017;9(1):85. doi:doi:10.1186/s13195-017-0312-4
17. Frisoni GB, Molinuevo JL, Altomare D, Carrera E, Barkhof F, Berkhof J, et al. Precision prevention of Alzheimer’s and other dementias: Anticipating future needs in the control of risk factors and implementation of disease-modifying therapies. Alzheimers Dement. 2020;16(10):1457-68. doi:10.1002/alz.12132
18. Watson J, Saunders S, Muniz Terrera G, Ritchie C, Evans A, Luz S, et al. What matters to people with memory problems, healthy volunteers and health and social care professionals in the context of developing treatment to prevent Alzheimer’s dementia? A qualitative study. Health Expect. 2019;22(3):504-17. doi:10.1111/hex.12876
19. Pennington J, Socher, R., Manning, C. Glove: Global vectors for word representation Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP). 2014:1532–43. doi: 10.3115/v1/D14-1162
20. Hendrix SB. Measuring clinical progression in MCI and pre-MCI populations: enrichment and optimizing clinical outcomes over time. Alzheimers Res Ther. 2012;4(4):24-. doi:10.1186/alzrt127
21. Topiwala H, Terrera GM, Stirland L, Saunderson K, Russ TC, Dozier MF, et al. Lifestyle and neurodegeneration in midlife as expressed on functional magnetic resonance imaging: A systematic review. Alzheimer’s & Dementia: Translational Research & Clinical Interventions. 2018;4(1):182-94. doi:10.1016/j.trci.2018.04.001
22. Morbey H, Harding AJE, Swarbrick C, Ahmed F, Elvish R, Keady J, et al. Involving people living with dementia in research: an accessible modified Delphi survey for core outcome set development. Trials. 2019;20(1):12. doi:10.1186/s13063-018-3069-6
23. Tochel C, Smith M, Baldwin H, Gustavsson A, Ly A, Bexelius C, et al. What outcomes are important to patients with mild cognitive impairment or Alzheimer’s disease, their caregivers, and health-care professionals? A systematic review. Alzheimers Dement (Amst). 2019;11:231-47. doi:10.1016/j.dadm.2018.12.003
24. Ritchie K, Carrière I, Su L, O’Brien JT, Lovestone S, Wells K, et al. The midlife cognitive profiles of adults at high risk of late-onset Alzheimer’s disease: The PREVENT study. Alzheimer’s & Dementia. 2017;13(10):1089-97. doi: 10.1016/j.jalz.2017.02.008
25. Caselli RJ, Locke DEC, Dueck AC, Knopman DS, Woodruff BK, Hoffman-Snyder C, et al. The neuropsychology of normal aging and preclinical Alzheimer’s disease. Alzheimer’s & dementia : the journal of the Alzheimer’s Association. 2014;10(1):84-92. doi:10.1016/j.jalz.2013.01.004
26. Lau KM, Parikh M, Harvey DJ, Huang CJ, Farias ST. Early Cognitively Based Functional Limitations Predict Loss of Independence in Instrumental Activities of Daily Living in Older Adults. J Int Neuropsychol Soc. 2015;21(9):688-98. doi:10.1017/s1355617715000818
27. Guarino A, Favieri F, Boncompagni I, Agostini F, Cantone M, Casagrande M. Executive Functions in Alzheimer Disease: A Systematic Review. Front Aging Neurosci. 2019;10:437-. doi:10.3389/fnagi.2018.00437
28. Weintraub S, Carrillo MC, Farias ST, Goldberg TE, Hendrix JA, Jaeger J, et al. Measuring cognition and function in the preclinical stage of Alzheimer’s disease. Alzheimer’s & dementia (New York, N Y). 2018;4:64-75. doi:10.1016/j.trci.2018.01.003
29. Sabbagh MN, Hendrix S, Harrison JE. FDA position statement “Early Alzheimer’s disease: Developing drugs for treatment, Guidance for Industry”. Alzheimer’s & dementia (New York, N Y). 2019;5:13-9. doi:10.1016/j.trci.2018.11.004
30. Ashour HM, Elkhatib WF, Rahman MM, Elshabrawy HA. Insights into the Recent 2019 Novel Coronavirus (SARS-CoV-2) in Light of Past Human Coronavirus Outbreaks. Pathogens. 2020;9(3). doi:10.3390/pathogens9030186
31. Rohrer JM, Brümmer M, Schmukle SC, Goebel J, Wagner GG. “What else are you worried about?” – Integrating textual responses into quantitative social science research. PLOS ONE. 2017;12(7):e0182156. doi:10.1371/journal.pone.0182156s

FROM BRAIN DISEASE TO BRAIN HEALTH: PRIMARY PREVENTION OF ALZHEIMER’S DISEASE AND RELATED DISORDERS IN A HEALTH SYSTEM USING AN ELECTRONIC MEDICAL RECORD-BASED APPROACH

 

A.M. Fosnacht1, S. Patel1, C. Yucus1, A. Pham1, E. Rasmussen1, R. Frigerio1, S. Walters2, D. Maraganore1

 

1. NorthShore Neurological Institute, NorthShore University HealthSystem, Evanston, IL, USA; 2. Research Institute, NorthShore University HealthSystem, Evanston, IL, USA

Corresponding Author: Demetrius M. Maraganore, MD, Ruth Cain Ruggles Chairman, Department of Neurology, Medical Director, NorthShore Neurological Institute, NorthShore University Health System, 2650 Ridge Ave., Evanston, IL, 60201, USA, Tel: 1-847-570-1678, Fax: 1-847-733-5565, Email: dmaraganore@northshore.org

J Prev Alz Dis 2017;4(3):157-164
Published online January 31, 2017, http://dx.doi.org/10.14283/jpad.2017.3

 


Abstract

Background: Alzheimer’s disease and aging brain disorders are progressive, often fatal neurodegenerative diseases. Successful aging, modern lifestyles and behaviors have combined to result in an expected epidemic. Risks for these diseases include genetic, medical, and lifestyle factors; over 20 modifiable risks have been reported.
Objectives: We aim to primarily prevent Alzheimer’s disease and related disorders through electronic medical record (EMR)-based screening, risk assessments, interventions, and surveillance.
Design:  We identified modifiable risks; developed human, systems and infrastructural resources; developed interventions; and targeted at-risk groups for the intervention.
Setting:  A Community Based Health System.
Participants: In year one (June 2015 to May 2016), 133 at-risk patients received brain health services with the goal of delaying or preventing Alzheimer’s disease and related disorders.
Measurements: We created mechanisms to identify patients at high risk of neurodegenerative disease; EMR-based structured clinical documentation support tools to evaluate risk factors and history; evidence-based interventions to modify risk; and the capacity for annual surveillance, pragmatic trials, and practice-based and genomic research using the EMR.
Results:  This paper describes our Center for Brain Health, our EMR tools, and our first year of healthy but at-risk patients.
Conclusion: We are translating research into primary prevention of Alzheimer’s disease and related disorders in our health system and aim to shift the paradigm in Neurology from brain disease to brain health.

Key words: Neurodegeneration, brain health, primary prevention, risk assessments, surveillance, electronic medical record.


 

Introduction

Alzheimer’s disease (AD) is an aging-related neurodegenerative disorder characterized by progressive accumulation of beta-amyloid protein plaques and tangles of the protein tau in and around neurons the brain. It is the most common type of dementia, accounting for well over half of cases (1).   Early presentation includes lapses in memory regarding conversations, recent events and names, followed by psychological manifestations and then by impaired behavioral and motor functions.  It is the 6th leading cause of death in the US (1).
Social phenomena such as the aging of Baby Boomers, “successful aging”, and growth of the oldest-old segment of the population have forecasted epidemics of aging-related disorders (1).  Alzheimer’s disease is of particular concern due to high prevalence, limitedly efficacious pharmaceuticals, and disappointing clinical trials (2).   Today, 5.4 million Americans live with AD and prevalence expectations for the year 2050 range from 13 to 16 million (1).   Prevalence patterns are mirrored in developed nations and AD has been identified as a priority by the G8 nation’s recently formed World Dementia Council (3).
Alzheimer’s disease is expensive. The combined costs of AD total $236 billion per year (1), fueled by a number of factors including extraordinary cost of nursing homes, which can exceed $92,378 per year (4), and hospital stays, which are longer for patients with AD regardless of reason for hospitalization (1).   By mid-century, costs are expected to exceed $1 trillion. Survey research reveals heavy burden to unpaid/family caregivers, who report serious work and career-related sacrifices, emotional stress, physical pain, and financial struggle (1).
The lifetime risk for AD is approximately 1 in 5 for women and 1 in 10 for men (1); however, knowledge of a variety of health and genetic factors allows personalization of risk estimates. In 1993, the Duke Alzheimer’s Disease Research Center Group published 3 papers describing their discovery that variations in the gene Apolipoprotein E (APOE) are associated with different risks of non-familial AD (5).   Depending on one’s APOE genotype, risk may be 40% less than referent or fifteen times more. In 2010, the US National Institutes of Health issued an Independent State-Of-The-Science Report that named diabetes, smoking, and depression as having reliable evidence for increasing risk for AD.  Evidence for other factors was lacking, and the authors called for more rigorous and higher quality research (6).    Five years later, a meta-analysis by the Alzheimer’s Association reported a robustly stronger body of literature supporting modifiable risk factors for cognitive decline and dementia (3). Research also supports a variety of APOE gene-environment interactions and highlights the particularly beneficial implications of risk mitigation for APOE ε4 carriers (7-9).
In 2015 the results of the first randomized-controlled trial of a multi-domain intervention among at-risk community-dwelling elderly were published, which demonstrated that multi-modal lifestyle intervention could improve or maintain cognitive function even in elderly at-risk adults (10). Accordingly, calls for action concerning models and initiatives in primary prevention of AD have been published.  Statistical estimates of what is possible for prevalence reduction in the US range from 30% to 50% (11, 12).
Based on the ability to risk-stratify, the weight of the evidence on modifiable risk factors, and available resources, the Neurology Department at NorthShore University HealthSystem aimed to build mechanisms to identify cognitively healthy but at-risk individuals years before a possible diagnosis of AD, Parkinson’s disease (PD) or chronic traumatic encephalopathy (CTE), to build practice-based interventions to manage risk factors, and surveillance mechanisms to monitor brain health with the goal of primary prevention of neurodegenerative disease.  Uniquely, we have built in the capacity for quality improvement and practice-based research using the electronic medical record (EMR), including creation of a biobank (“clinomics”), a Neurology Practice Based Research Network (NPBRN), and informatics tools to conduct pragmatic trials using subgroup based adaptive designs (13).   This paper describes how we built the Center for Brain Health, describes our EMR tools and first year of patients, and characterizes our opportunity for risk mitigation and prevention.

 

Methods

Step 1: Identification of risk and protective factors

We identified as having strong evidence in the literature the following factors that increase risk for AD and related disorders: genetic (family history, APOE ε4, other susceptibility genes) (14-17), metabolic (cholesterol, diabetes, midlife obesity) (3, 5, 6, 11, 12, 17), vascular (cardiovascular disease, midlife hypertension, stroke) (3, 12), infectious and inflammatory (periodontitis, others) (18), head trauma (3),  diet (homocysteine, standard American diet, nutrient deficiencies) (17, 19),  habits (smoking, alcohol abuse) (3,6,11,12,17,20),  sleep (poor quality, disorders) (3, 7, 8), depression (3, 6, 10, 16),  early menopause (natural or surgical) (21),  sedentary lifestyle (11, 12) and certain medications (22, 23) (Table 1). We identified as having strong evidence in the literature as decreasing risk for AD and related disorders the following factors: genetic (APOE ε2) (5, 15), social (education, income, engagement) (3, 24), lifestyle (physical, mental exercise) (3, 6, 17, 25),  diet (Mediterranean) (3, 17),  vitamins (B6/B12/Folate, A, C, D, E) (17, 26),  medications (NSAIDS, statins, early hormone replacement therapy, antihypertensives, antidiabetics) (17, 27, 28). (Table 1)

 

 

Table 1. Survey of the literature reveals at least 20 modifiable risk factors for Alzheimer’s disease and related disorders

Table 1. Survey of the literature reveals at least 20 modifiable risk factors for Alzheimer’s disease and related disorders

 

 

Step 2:  Resource mobilization

The Center for Brain Health exists within the NorthShore Neurological Institute of NorthShore University HealthSystem and utilizes its staff, offices, equipment and infrastructure.  NorthShore University HealthSystem is a comprehensive, fully integrated healthcare delivery system serving the North Chicago region and includes 4 hospitals and 2,100 affiliated/employed physicians (29). It was amongst the first health systems to adopt the EMR in the US and was amongst the first to demonstrate “meaningful use” of EMR technology in ways that translate to improved quality, safety and efficiency for patients.  Downstream from its EMR, NorthShore maintains an enterprise data warehouse (EDW) that fosters health-related data analytics and enables broad capacity for increasingly longitudinal health-related research and informatics.
Concept development and planning began via the invitation of physicians and internal professionals to join one of five working groups that met monthly for up to 24 months to develop the concept vis-à-vis their areas of expertise. All groups were lead by the Chairman of Neurology (Director, Center for Brain Health).
The Research and Development Working Group consisted of experts in biomedical research informatics, epidemiology, genomics, molecular medicine, neurology, neuropathology, neuroradiology, nutrition, primary care, and animal models of neurodegeneration. They were tasked with deliberating over gaps in knowledge regarding at-risk populations, risk assessments, interventions, and addressing these gaps through point-of-care research utilizing the EMR.
The Community Engagement Working Group consisted of experts in community relations, marketing, patient engagement, philanthropy, and public health. They were tasked with building strategic community partnerships, community resource leveraging, and identifying events for participation to expand the reach of screening.
The Targeted Populations Working Group consisted of experts in genomics, healthcare administration, health information technology (HIT), marketing, neurology (memory disorders and movement disorders), public health, primary care, reproductive endocrinology, and sports concussion. They were charged with describing high-risk populations and recruitment mechanisms.
The Risk Assessments and Surveillance Working Group consisted of experts in healthcare administration, neurology (memory disorders and movement disorders), neuropsychology, neuroradiology, nuclear medicine, primary care, and public health. Their goal was to identify reliable, valid and pragmatic measures of brain health for patient assessments, and mechanisms for surveillance and capturing outcomes.
The Interventions Working Group consisted of experts in cognitive and physical therapy, healthcare administration, HIT, integrative medicine, medical social work, neurology (memory disorders and movement disorders), nutrition, and public health.  Their goal was to develop a suite of evidence-based interventions for modifiable risk factors for AD, PD and CTE and strategize how to operationalize these within the framework of the health system.
Key practitioners recruited for Center for Brain Health include three neurologists (PD, memory, sleep and integrative), a registered and research dietitian, a medical social worker, a physician assistant, and therapists from cognitive and physical therapy.  Non-clinical staff includes a practice manager, a senior clinical research associate, research assistants, a statistician, and HIT programmer analysts.

Step 3: Build Structured Clinical Documentation Support (SCDS) Toolkits into the Electronic Medical Record

As part of an initiative funded by the Agency for Healthcare Research and Quality to improve the quality of neurology clinical practice and facilitate point-of-care practice-based research using the EMR, the Center for Brain Health utilizes a SCDS toolkit in the EMR for all patient encounters (13). Center for Brain Health neurologists met biweekly for three months to standardize Brain Health office visit types according to evidence-based medicine (toward Best Practices).  We developed consensus on: definitions of AD, PD and CTE; outcomes of interest to clinicians and patients; valid and feasible outcome measures for point-of-care assessments; and factors known to influence the outcomes and measures. The neurologists met with members of NorthShore’s EMR optimization team biweekly over three months to develop and test the SCDS toolkit which navigates care, writes progress notes, provides clinical decision support, and electronically captures structured data.
The brain health SCDS toolkit is utilized by the care team (medical assistant, nurse, neurologist, research assistant when appropriate).  The tools include a custom navigator (index of electronic forms), electronic forms (documentation flow sheets, including cascading data elements, auto-scoring and interpreting and other “smart form” features), Best Practice Advisories (pop-up alerts), and order sets. The content of the electronic forms includes several score test measures (Appendix A) and customized fields that discretely document: chief complaints, patient information (ancestry, special diets, caffeine use, exercise habits, health maintenance, toxin and medication exposures), past medical history (specific to aging brain disorders), family history (specific to aging brain disorders), traumatic brain injury, prior treatments (nootropic, dopaminergic, nutraceutical), prior diagnostics (brain imaging modalities, electroencephalography, polysomnography, neuropsychological, genetic, and cerebrospinal fluid testing), and blood tests.  These tools electronically capture 400+ fields of data.  Some examples of screenshots of the toolkit are shown in Figure 1, with additional screenshots shown in Appendix B.

Figure 1. Screenshots of structured clinical documentation support tools that we have built into the electronic medical record (EPIC) that assess known risks to Alzheimer’s and related disorders. These tools electronically capture large amounts of clinical and diagnostic data. Data capture includes: Mediterranean diet and Readiness Assessment (both shown below). Additional screenshots are shown in Appendix A. © 2015 EPIC Systems, used with permission

Figure 1. Screenshots of structured clinical documentation support tools that we have built into the electronic medical record (EPIC) that assess known risks to Alzheimer’s and related disorders. These tools electronically capture large amounts of clinical and diagnostic data. Data capture includes: Mediterranean diet and Readiness Assessment (both shown below). Additional screenshots are shown in Appendix A. © 2015 EPIC Systems, used with permission

 

Step 4: Identification of at-risk patients

To identify at-risk patients currently engaged with our health system, we sent mailings to Primary Care Providers (PCPs), posted articles to internal websites and newsletters, and gave presentations to NorthShore committees (Council of Chairmen, Medical Group Primary Care Committee, and Medical Group Specialty Care Committee) and departments (Internal Medicine, Neurology, Obstetrics and Gynecology).
To identify at-risk individuals in the community, we sponsored activities and gave presentations at community events. We authored blogs and webinars for local chapters of aging and dementia-related national organizations. The Center for Brain Health initiative has been represented at national scientific conferences and featured in media interviews. We advertised on television, radio, the internet, billboards, newspapers, and magazines.
We developed informatics tools to identify targeted populations: a web- and paper-based Brain Health Quiz (Appendix C), an EMR-based Alzheimer’s Risk Score Algorithm, and EMR-based flags.  The Brain Health Quiz is an un-scored self-screening tool of evidence-based risk factors for AD, PD, and CTE by which individuals can learn about their risk factors and self-refer.  We distribute the paper-based quiz at events and post it electronically on NorthShore intranet and internet sites.
The Alzheimer’s Risk Score is an algorithm-based clinical decision support tool that predicts mild cognitive impairment (MCI), dementia, or AD in the next five years for patients aged 60+.  It will be built into the EMR as a widget (on-demand score) and as a Best Practice Advisory (pop-up notification for high-risk patients).  To construct the algorithm: we utilized data stored in the EDW for patients ages 60+ who had visited their primary care physician in 2009 and again in 2014.  We excluded patients who in 2009 had a diagnosis of MCI, dementia, or AD.  We included as independent variables any of the 23 factors listed in the Brain Health Quiz that were captured by the EMR by 2009.  We included as dependent variables diagnosis of MCI, dementia, or AD by 2014.  Using stepwise statistical model building, we identified variables in the patients’electronic records that contributed to higher risk.  Details regarding the model will be published separately.  Additionally, we will flag patients in the EMR with a documented family history of AD or PD, multiple concussions or more-severe brain injuries, rapid eye movement sleep behavior disorder, or early-unopposed menopause, for primary care physicians to consider referral to the Center for Brain Health.

Step 5: Develop personalized medicine and interventions

We implement personalized, evidence-based interventions to mitigate risk factors and maximize protective factors, and evidence-based interventions with the potential to benefit all patients.  These interventions included lifestyle and behavioral changes, medications, and management of diseases associated with increased risk of AD.  We maintain compliance-focused follow-up via interval visits with a physician assistant.  Annual follow-up visits with the neurologist focus on updating risk profiles and modifying interventions according to proximal outcomes and the latest evidence.  In the event that patients develop aging brain disorders despite our best efforts, they are transitioned at the earliest point to relevant neurology subspecialty practices.
The diagnostic testing and interventions are defined by an order set built into the EMR.  The “smart set” includes frequently ordered labs tests (e.g., metabolic panel, complete blood count with differential, cardiac risk, and vitamin D-25, Vitamin B12, Folic acid blood levels), imaging tests (MRI, CT), additional lab tests (e.g., glucose test, homocysteine level), genetic tests (e.g., APOE, early onset Alzheimer’s evaluation), additional procedures (e.g., cerebrospinal fluid examination), frequently ordered medications (e.g., Folic Acid-Pyridoxine-Cyancobalamin), consults (e.g. dietitian, physical therapy), diagnoses, and billing codes.
We built into the EMR Best Practice Advisories that prompt neurologists to enroll patients into a DNA biobank (each is genotyped for one million single nucleotide polymorphism markers), or to complete a mental health order set when patients were severely anxious or depressed and not taking an anxiolytic or an antidepressant and without documentation of a visit with a mental health practitioner in the prior year.

Step 6: Community Engagement

To include our community in defining our clinical services, we created a Community Advisory Council whose membership consisted of patients with aging brain disorders, caregivers, municipal leaders, public health experts, and professionals representing our partnering organizations. We met bimonthly to provide progress updates and receive feedback.

Statistical Methods

We generated from the Center-for-Brain-Health-specific data mart in the enterprise data warehouse a descriptive cohort report (medians and ranges, means and standard deviations, frequencies, overall and in men and women separately), visualized the data using box plots, bar graphs, and normal Q-Q plots, and performed pairwise correlations (without and with adjustments) and principal component (PC) analysis of the scored tests (without and with Varimax rotations).  A statistician (SW) performed the analyses using SAS 9.3 (Cary, NC) and R software.

 

Results

We created mechanisms to identify patients at high risk of AD, PD, and CTE; the SCDS tools to evaluate risk factors and history; the evidence-based interventions to modify risk; and the capacity for annual surveillance, pragmatic trials, and practice-based and genomic research using the EMR.
In our first year (June 2015 to May 2016), we saw 133 patients. Fifteen were found to have MCI and were referred to the Memory Disorders Clinic for care.  The remaining 118 patients were cognitively normal; median age was 59.5 (range 31-81); 81 were women. The median number of risk factors reported via the Brain Health Quiz was 5 (range 1-10).  The median body mass index (BMI) for women was 26 and for men 27.  The median PREDIMED score (a validated 14-item Mediterranean diet questionnaire) was 7 for both genders (range, 1-12); only 3.4% of patients had “strongly adherent to the Mediterranean diet” scores.  Our patients were highly educated with 100% having finished high school, 83% completing 4 years college, 51% completing 2 years graduate school and 21% receiving a post-graduate education. A complete descriptive cohort report is provided in Appendix D.
Women reported with higher frequencies family histories of dementia, PD, depression, sleep apnea, prior head injuries, prior use of NSAIDs, exposure to pesticides, and alcohol use. Women scored more often in the clinical insomnia and depression ranges.  Men were more likely to report caffeine use and sedentary activity.  Men scored more often in the moderate to severe anxiety range. None of these gender differences reached statistical significance.  Men were more likely to report statin use, the only gender difference reaching statistical significance (p <0.05).
Table 2 provides results of pairwise correlations; Figure 2 illustrates results of the PC analysis.  The analyses included 118 patients with complete data for a Mediterranean diet questionnaire (PREDIMED); Center for Epidemiological Studies-Depression scale (CED-D); Generalized Anxiety Disorder 7-item scale (GAD-7); Insomnia Severity Index (ISI); 9-hole peg test, dominant hand (9-hole dom); 9-hole peg test, non-dominant hand (9-hole non-dom); 25-foot walking test (25ft walk); Short Test of Mental Status (STMS); Unified Parkinson’s Disease Rating Scale-Motor scale (UPDRS); The Eight-Item Informant Interview to Differentiate Aging and Dementia (AD8); a Parkinsonism screening questionnaire (PARK); Body-mass index (BMI); and a Brain Health Readiness Assessment (Readiness) which measures patients’ willingness to engage in behavior change specific to risk and protective factors.  Several pairwise correlations were statistically significant even accounting for multiple comparisons.  None of the measures were over-correlated (rho < 0.8 or > -0.8).  We performed PC analyses restricting to validated score test measures (9-hole peg tests dominant and non-dominant, 25-ft walk, AD-8, CES-D, GAD-7, ISI, PARK, PREDIMED, STMS, UPDRS), and also including additional continuous trait measures (age at study, BMI, Readiness).  For the PC analyses that included all measures, the 9-hole peg test (dominant hand) loaded to the first PC most heavily. Two-factor maps (PCs 1 and 2), without and with Varimax rotation (Figure 2), revealed clustering of ISI, BMI, AD-8, PREDIMED, and Readiness. Another cluster included STMS, 9-hole peg test (dominant, non-dominant), 25-ft walk, age at study, PARK, and UPDRS.  GAD-7 and CES-D formed a 3rd distal cluster.  Inspection of the PCs after Varimax rotation and restricting to factors with eigen values >1 revealed a three factor solution, where the score tests with component loadings >0.40 on PC1 included the 9-hole peg test non-dominant and the 9-hole peg test dominant and age at study, and the score tests with component loadings >0.40 on PC2 included the GAD-7 and CES-D, and the score tests with component loadings >0.40 on PC3 included AD-8 and Readiness.  See Appendix E for a complete score test analytic report.
56 patients were referred to the dietitian.  28 patients opted for APOE genotype analysis.  Of the 118 in our cohort, 101 were eligible for enrollment in our blood DNA and plasma biobank, and 88 (87.1%) of the eligible subjects participated.

Table 2. Pairwise correlations of score test measures at initial visits for 118 patients in the cohort. Table shows pairwise correlations adjusted for age and gender. Correlation coefficients (rho) are shown in the cells only for significant correlations, p-value < .05/n where n = 91, the number of tests. Thus significance is established for p-value < 5.5 x 10-4 (after Bonferroni correction for multiple testing). Note that score tests compared to themselves (x and y axis of the table) are perfectly correlated (rho = 1.0)

Table 2. Pairwise correlations of score test measures at initial visits for 118 patients in the cohort. Table shows pairwise correlations adjusted for age and gender. Correlation coefficients (rho) are shown in the cells only for significant correlations, p-value < .05/n where n = 91, the number of tests. Thus significance is established for p-value < 5.5 x 10-4 (after Bonferroni correction for multiple testing). Note that score tests compared to themselves (x and y axis of the table) are perfectly correlated (rho = 1.0)

PREDIMED (Mediterranean diet questionnaire); CES-D (Center for Epidemiological Studies-Depression); GAD-7 (Generalized Anxiety Disorder 7-item scale); ISI (Insomnia Severity Index); 9-hole dom (9-hole peg test, dominant hand); 9-hole non-dom (9-hole peg test, non-dominant hand); 25ft walk (25-foot walking test); STMS (Short Test of Mental Status); UPDRS (Unified Parkinson’s Disease Rating Scale-Motor); AD8 (The Eight-Item Informant Interview to Differentiate Aging and Dementia); PARK (Parkinsonism screening questionnaire); BMI (Body-mass index); Readiness (Brain Health Readiness Assessment).

Figure 2. Principal component analyses for score test measures at initial visits for 118 patients in the cohort

Figure 2. Principal component analyses for score test measures at initial visits for 118 patients in the cohort

A) Cascade figure, demonstrating the proportion of the variance between the measures explained stepwise by each of the principal components. B) Illustrates the principal component mappings of first principal component (PC1) versus second principle component (PC2) loadings for each of the score tests and their spatial relationships.  C) Illustrates the principal component mappings of PC1 and PC2 loadings for each of the score tests, after Varimax rotation (to maximize the distance between measures).  PREDIMED (Mediterranean diet questionnaire); CES-D (Center for Epidemiological Studies-Depression); GAD-7 (Generalized Anxiety Disorder 7-item scale); ISI (Insomnia Severity Index); 9-hole dom (9-hole peg test, dominant hand); 9-hole non-dom (9-hole peg test, non-dominant hand); 25ft walk (25-foot walking test); STMS (Short Test of Mental Status); UPDRS (Unified Parkinson’s Disease Rating Scale-Motor); AD8 (The Eight-Item Informant Interview to Differentiate Aging and Dementia); PARK (Parkinsonism screening questionnaire); BMI (Body-mass index); Readiness (Brain Health Readiness Assessment)

 

Discussion

In its first year, the Center for Brain Health initiative identified and engaged individuals at increased risk for Alzheimer’s disease and related disorders, explored gender differences, and identified opportunities for mitigation of risk at the individual, system, and community levels. The majority of our patients are women; they scored higher on our Readiness Questionnaire indicating that we may need to work harder to identify and engage men.  On the other hand, as the lifetime risk of Alzheimer’s disease is double in woman versus men (1), this gender bias in referrals may be appropriate.
Only four patients (3.4%) were strongly adherent to the Mediterranean diet, and the median BMI for both genders was above the threshold of normal weight-for-height. As midlife obesity increases risk for AD (3, 11) and strong adherence to the Mediterranean diet reduces risk (3), these data serve as examples that characterize and describe the opportunity we have for risk mitigation in our ever expanding cohort   While long term outcomes are beyond the scope of this descriptive paper, we aim to publish data regarding patient compliance and outcomes as the cohort matures and when change can be measured.
Identification of high-risk individuals within a population, risk mitigation, and surveillance are building blocks of primary prevention initiatives. We demonstrate that this is possible in a large health system and that leveraging the EMR and analytics can automate efforts and create learning opportunities.  Not long ago, the adage that Alzheimer’s disease cannot be prevented was ubiquitous.  But literature disputing that adage is growing, as are calls for action on primary prevention and public health approaches to the AD epidemic. In Baumgart’s 2015 meta-analysis (3), the authors state that it’s no longer acceptable to linger in academic discussion; that the evidence is too strong to warrant inaction. Similarly, Norton and others urge the undertaking of a population health approach (12). And, we may have proof that risk factor management will lower dementia prevalence in the real-world.  This year, Satizabal and others reported a decline of dementia incidence among Framingham Heart Study participants of 44% over 3 decades (30). While the factors responsible for this reduction remain unclear, it is noteworthy that during those 3 decades, the level of education rose and most vascular risk factors declined (30).
We are not alone in this venture; similar initiatives have been developed around the country (Weil Cornell Alzheimer’s Prevention Clinic by New York Presbyterian, the Alzheimer’s Prevention Program at Cedars-Sinai, the Alzheimer’s Risk Assessment and Intervention Program at the University of Alabama at Birmingham) but we are unique in many ways including 1) our scope: we aim to primarily prevent not only Alzheimer’s disease, but also Parkinson’s disease and chronic traumatic encephalopathy; 2) our use of informatics: we are utilizing SCDS tools built into the EMR, and are building into the EMR an Alzheimer’s Risk Score to assist patients and physicians to identify and define risk; 3) our commitment to genomics and clinomics: we are biobanking DNAs from consenting subjects, genotyping the samples for 1 million genomic markers, and associating the genotypes with electronically captured clinical data.  We anticipate referrals from NorthShore’s system-wide Genomic Health Initiative that will identify thousands of APOE E4 carriers; 4) our community engagement: through the Community Advisory Board, we partner to expand the scope of our services; 5) our collaborations: through the Neurology Practice Based Research Network (13) we are sharing our EMR tools and data, which will vastly increase our ability to improve the quality of the care we provide, to make new discoveries relating to brain health, to achieve better outcomes, and ultimately to reduce the burden of brain disorders in the communities we serve.

 

Acknowledgments: The authors acknowledge the generous support of the Auxiliary of NorthShore University HealthSystem with respect to the initial building of electronic medical record (EMR) toolkits, and thank the medical assistants, nurses, neurologists, EMR Optimization and Enterprise Data Warehouse programmers, administrators, and research personnel at NorthShore University HealthSystem who contributed to the quality improvement and practice-based research initiative using the EMR. The authors thank Vimal Patel, PhD for his assistance with editing, formatting, and submitting the manuscript for publication. Finally, the authors thank the neurology patients who inspire us to improve quality and to innovate our clinical practice every day.

Funding: The authors also acknowledge funding support from the Agency for Healthcare Research and Quality (R01HS024057).

Ethical standards: This is a cross-sectional description of historical patients seen in a clinical practice. There is no enrollment.

 

References

1.    Alzheimer’s Association. 2016 Alzheimer’s Disease Facts and Figures. Alzheimer’s & Dementia 2016;12(4). https://www.alz.org/documents_custom/2016-facts-and-figures.pdf. [Accessed September 1, 2016]
2.    Schneider LS, Mangialasche F, Andreasen N, et al. Clinical trials and late-stage drug development for Alzheimer’s disease: an appraisal from 1984 to 2014. J Intern Med. 2014;275(3):251-83.
3.    Baumgart M, Snyder HM, Carrillo MC, Fazio S, Kim H, Johns H. Summary of the evidence on modifiable risk factors for cognitive decline and dementia: A population-based perspective. Alzheimers Dement. 2015;11(6):718-26.
4.    Genworth Financial, Inc. Genworth 2016 Cost of Care Survey. Home Care Providers, Adult Day Health Care Facilities, Assisted Living Facilities and Nursing Homes. https://www.genworth.com/about-us/industry-expertise/cost-of-care.html. [Accessed September 1, 2016]
5.    Roses AD. On the discovery of the genetic association of Apolipoprotein E genotypes and common late-onset Alzheimer disease. J Alzheimers Dis. 2006;9(3 Suppl):361-6.
6.    Daviglus ML, Bell CC, Berrettini W, et al. National Institutes of Health State-of-the-Science Conference statement: preventing alzheimer disease and cognitive decline. Ann Intern Med. 2010;153(3):176-81.
7.    Burke SL, Maramaldi P, Cadet T, Kukull W. Associations between depression, sleep disturbance, and apolipoprotein E in the development of Alzheimer’s disease: dementia. Int Psychogeriatr. 2016;28(9):1409-24.
8.    Lim AS, Yu L, Kowgier M, Schneider JA, Buchman AS, Bennett DA. Modification of the relationship of the apolipoprotein E ε4 allele to the risk of Alzheimer disease and neurofibrillary tangle density by sleep. JAMA Neurol. 2013;70(12):1544-51.
9.    Morris MC, Brockman J, Schneider JA, et al. Association of Seafood Consumption, Brain Mercury Level, and APOE ε4 Status With Brain Neuropathology in Older Adults. JAMA. 2016;315(5):489-97.
10.    Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomized controlled trial. Lancet. 2015;385(9984):2255-63.
11.    Norton S, Matthews FE, Barnes DE, Yaffe K, Brayne C. Potential for primary prevention of Alzheimer’s disease: an analysis of population-based data. Lancet Neurol. 2014;13(8):788-94.
12.    Barnes DE, Yaffe K. The projected effect of risk factor reduction on Alzheimer’s disease prevalence. Lancet Neurol. 2011;10(9):819-28.
13.    Maraganore DM, Frigerio R, Kazmi N, et al. Quality improvement and practice-based research in neurology using the electronic medical record. Neurol Clin Pract. 2015;5(5):419-429.
14.    Foley SF, Tansey KE, Caseras X, et al. Multimodal Brain Imaging Reveals Structural Differences in Alzheimer’s Disease Polygenic Risk Carriers: A Study in Healthy Young Adults. Biol Psychiatry. 2016;pii: S0006-3223(16)31114-3.
15.    Jansen WJ, Ossenkoppele R, Knol DL, et al. Prevalence of cerebral amyloid pathology in persons without dementia: a meta-analysis. JAMA. 2015;313(19):1924-38.
16.    Scarabino D, Gambina G, Broggio E, Pelliccia F, Corbo RM. Influence of family history of dementia in the development and progression of late-onset Alzheimer’s  disease. Am J Med Genet B Neuropsychiatr Genet. 2016;171B(2):250-6.
17.    Xu W, Tan L, Wang HF, Jiang T, et al. Meta-analysis of modifiable risk factors for Alzheimer’s disease. J Neurol Neurosurg Psychiatry. 2015;86(12):1299-306.
18.    Ide M, Harris M, Stevens A, et al. Periodontitis and Cognitive Decline in Alzheimer’s Disease. PLoS One. 2016;11(3):e0151081.
19.    Hu Q, Teng W, Li J, Hao F, Wang N. Homocysteine and Alzheimer’s Disease: Evidence for a Causal Link from Mendelian Randomization. J Alzheimers Dis. 2016;52(2):747-56.
20.    Mukamal KJ, Kuller LH, Fitzpatrick AL, Longstreth WT Jr, Mittleman MA, Siscovick DS. Prospective study of alcohol consumption and risk of dementia in older adults. JAMA. 2003;289(11):1405-13.
21.    Rocca WA, Shuster LT, Grossardt BR, et al. Long-term effects of bilateral oophorectomy on brain aging: unanswered questions from the Mayo Clinic Cohort Study of Oophorectomy and Aging. Womens Health (Lond). 2009;5(1):39-48.
22.    Badiola N, Alcalde V, Pujol A, et al. The proton-pump inhibitor lansoprazole enhances amyloid beta production. PLoS One. 2013;8(3):e58837.
23.    Risacher SL, McDonald BC, Tallman EF, et al. Association Between Anticholinergic Medication Use and Cognition, Brain Metabolism, and Brain Atrophy in Cognitively Normal Older Adults. JAMA Neurol. 2016;73(6):721-32.
24.    Caamaño-Isorna F, Corral M, Montes-Martínez A, Takkouche B. Education and dementia: a meta-analytic study. Neuroepidemiology. 2006;26(4):226-32.
25.    Erickson KI, Voss MW, Prakash RS, et al. Exercise training increases size of hippocampus and improves memory. Proc Natl Acad Sci U S A. 2011;108(7):3017-22.
26.    Miller JW, Harvey DJ, Beckett LA, et al. Vitamin D Status and Rates of Cognitive Decline in a Multiethnic Cohort of Older Adults. JAMA Neurol. 2015;72(11):1295-303.
27.    Kantarci K, Lowe VJ, Lesnick TG, et al. Early Postmenopausal Transdermal 17β-Estradiol Therapy and Amyloid-β Deposition. J Alzheimers Dis. 2016;53(2):547-56.
28.    Tully PJ, Dartigues JF, Debette S, Helmer C, Artero S, Tzourio C. Dementia risk with antihypertensive use and blood pressure variability: A cohort study. Neurology. 2016;87(6):601-8.
29.    NorthShore University HealthSystem. http://www.northshore.org/. [Accessed September 1, 2016]
30.    Satizabal CL, Beiser AS, Chouraki V, Chêne G, Dufouil C, Seshadri S. Incidence of Dementia over Three Decades in the Framingham Heart Study. N Engl J Med. 2016;374(6):523-32.

APPENDIX

Appendix A

Appendix B

Appendix C

Appendix D

Appendix E