Data vs Dementia
Maxine Mackintosh on early intervention to prevent cognitive decline and the One HealthTech community.
Maxine Mackintosh is a PhD student at University College London working at the intersection of data science and dementia. Her PhD involves mining medical records for new predictors for dementia. She is passionate about understanding how we might make better use of routinely collected data to improve our cognitive health. Alongside this, she is the co-founder of One HealthTech – a community which champions and supports underrepresented groups, particularly women, to be the future leaders in health innovation. Her professional work has led her to the Royal Society, Roche, L’Oreal, Department for International Development, and NHS England. She is part of a number of communities and committees including the World Economic Forum’s Global Shapers, the British Computer Society (Health Exec) and the Google DeepMind Health’s Independent Review Panel.
You’re currently working towards you PhD in data science and dementia at the Institute of Health Informatics at UCL. Could you tell us a bit about your research project?
My work involves mining medical records (from CALIBER) for early predictors of dementia. I take primary care data (ie from your GP), linked to secondary care data (i.e. hospital data), linked to death certificate information for over 10 million patients, and do exploratory analyses into all the medications, symptoms, diseases, signs and procedures that happen in the lead up to dementia diagnosis with the view of broadening our understanding of what the manifestation of dementia are, beyond merely often subtle and nuanced symptoms of cognitive decline. I have a great bank of supervisors: A data supervisor, Spiros Denaxas, a clinical supervisor, Martin Rossor, and a statistics supervisor, Arturo Gonzalez-Izquierdo.
Why is studying the 20 years before someone is diagnosed with dementia so important to your work?
Dementia has an enormous latency period, in other words, there is believed to be up to 20 years delay between the neuronal cell death in the brain and becoming symptomatic, such as forgetting things, or your personality changing. If there is any way that we could pick people up earlier during that period of decline then that might be the difference between giving someone a drug that could stop of even reverse the damage caused thus far, or that person being beyond help because too much damage has already occurred. We don’t even know what dementia 15 years earlier looks like, so how can we think about intervening early? The reality is that cognitive decline has global effects on the body, and that’s what I am trying to find out.
You’re working with thousands of diagnostic codes – what kind of data are they made up of?
All the data I use is structured data (in rows and columns). We do not have any free text, which can be seen as a massive disadvantage because a huge amount of detail about mental health conditions is recorded in a patient narrative. The data literally comes to me in big xml files, with scrambled patient IDs, dates of event, demographic information and diagnostic codes. I then spend a lot of my time reshuffling columns, cutting, merging, reorganising data so that it is clean and in a format that I can then start my analyses. That process took me months and months.
You mention that there is other data, such as consumer data, that might be helpful to identify markers for dementia – what are the barriers to accessing this?
Your postcode is a better predictor of your health than your genetic data right? If we think about all the types of information that are captured in the “social determinants of health” domains, it’s information about where you live, who you are meeting, what you eat, where you work. All that information is captured by commercial, consumer organisations. But these organisations are not healthcare companies, and for many of them, it hasn’t occurred to them to work in healthcare, or it’s too controversial to pivot, or, in most cases, consumers have not and would not consent to have their data used for such personal and sensitive applications as healthcare. Information governance around this is also usually a nightmare.
” there is believed to be up to 20 years delay between the neuronal cell death in the brain and becoming symptomatic, such as forgetting things, or your personality changing. ”
Are there many ethical issues that you need to be aware of when working with sensitive data?
Yes, plenty! Issues range from anonymisation, to storage, to dissemination. It can get pretty philosophical pretty quick; at what point is something sufficiently anonymised? At what point does safe storage hinder research? An example being that in the Safe Haven we use (a secure environment where we analyse sensitive data), it doesn’t connect to the internet, so if you are using a really niche package on R, you have to send an email requesting that package to be installed which can take anything from a few hours to action, to a few weeks. On the dissemination front, something that is very pertinent is feeding back sensitive information to people, how it’s done, and whether they want to. My work looks at early detection of dementia, but do you really want to be told you are 20 years off getting diagnosed when there is nothing that can be done abut it? Some people do, but lots won’t. I think the two key things to remember in all of this are that the ethical issues vary a lot according to who the person or stakeholder is and also, we must never presume any knowledge. Less than 20% of people know that researchers like me are using their sensitive EHR data for research, so whilst you think someone may well be concerned with how the data is stored, it’s more likely that they want to know why and how on earth you have access in the first place.
What has moved you to choose this path and why is it so important?
I was studying a Health Economics and Health Policy Masters when I started getting interested in tech and data. We never once discussed digital health, personalized medicine or genomics, which felt odd as lots of these technologies were fundamentally challenging all the structures, processes and regulations we were studying. So I started attending various healthtech Meetups to learn more. I found the community so warm, friendly and interesting I started teaching myself to programme, attended various on-the-side coding classes and the rest is history!
” We never once discussed digital health, personalised medicine or genomics, which felt odd as lots of these technologies were fundamentally challenging all the structures, processes and regulations we were studying “
You’ve setup One HealthTech to support people of diverse backgrounds and women access opportunities and develop their careers in tech. What sort of work do you do?
The bulk of One HealthTech activities are local Meetups around the country, usually evening or breakfast events, with some interesting talks around a specific healthtech topics, usually with an all-female panel, followed by networking. Then behind the scenes there are a huge amount of introductions and connections that happen, from internship opportunities, to finding mentors, to finding commercial partners for projects. We are in the process of making all of those really valuable, behind-the-scenes activities more formalised, transparent and accessible for individuals.
One HealthTech has a range of members, from NHS staff to SME employees and academics – how do you see the platform developing over the next few years?
There are two main areas of growth for us, one is launching international hubs, of which there are many about to get up and running, we just need to finally pin down some processes (and we have an amazing new community manager who is doing 99% of this work). Secondly, the chaotic, bubbling noisy community Meetups are amazing, but we are going to develop programmes that sit on top of the community so that when people experience common inflexion points, be it academia into industry, early stage startup, scaling up, going freelance etc, we can help people more efficiently, opposed to supporting individuals on a case-by-case basis.
What advice would you give to someone who is interested in pursuing a tech career but unsure how to go about it or where to look for opportunities?
Join One HealthTech! (or any other related or relevant Meetup-based community). It opens up your world to people who can help you, are further down the line to you and have all the tips, and it provides you with that support network that when you are starting out you don’t have. The good thing about Meetup-based groups is they are also so informal, you leave boring, stale work-you at the door, so it’s easier to be honest about being stuck and lost about where to go.Stay up to date