Northernlands 2 - Open analytics at the Health Foundation

Description

Why collaboration and working in the open is important in health and beyond

Transcript

This transcript comes from the captions associated with the video above. It is "as spoken".

00:04 Hi my name is Emma Vestesson I'm a senior data analyst at the

00:08 Health Foundation and I'm here to

00:10 share some of the work we're doing about open analytics

00:14 at the Health Foundation.

00:16 The Health Foundation is an independent charity committed

00:18 to bring about better health and health care for people in

00:21 the UK. I'm part of the data analytics team at the Health

00:25 Foundation and over the last year or so, we've done some

00:28 pretty intense work around making our analysis more

00:31 reproducible and open.

00:33 So there are big challenges, facing health and

00:36 healthcare in the UK and this was true even before covid

00:39 happened. But at the same time we're seeing some rapid

00:43 innovation and use of analytics and data-driven tech in the

00:47 healthcare sector. And this carries huge potential to

00:49 address some of these challenges. But it does also

00:52 carry quite substantial risks if we're not careful with how it's

00:56 implemented and our aim is to ensure that data analytics has a

01:00 positive impact in the UK's

01:02 health. And we're working for a future where everyone's health

01:06 and care benefits from analytics and data-driven technology.

01:10 So how are we doing this?

01:14 We want data analytics to have a positive impact in the

01:17 UK's health and there are several ways in which we're

01:21 working to achieve this, so we want to have innovations and data

01:24 that tackle some of the big challenges facing the sector.

01:29 But we also want to influence the national conversation and

01:33 be an independent voice, an expert voice on national policy

01:37 and system design.

01:39 We offer funding and support for innovative work. Because we

01:44 want to support better analytical capability.

01:48 And we want to help create networks, communities and

01:51 partnerships. So what does open analytics mean to us at the

01:56 Health Foundation? So we're very committed to being transparent/open

01:59 when we're doing analysis 'cause we want people to be able

02:03 to look at what we've done, and ask any questions or

02:07 pose any challenges. So we want to actively champion good

02:09 analytical practice across the sector and spread and share what

02:12 we know. So we started doing this by publishing our code on

02:17 GitHub around a year ago and last September we had kind of a

02:21 mini launch of our own GitHub page, so we now make all of our

02:25 analysis code public by default even when it's work-in-progress

02:28 and we're working towards exclusively using only open source

02:31 languages. But we don't think that sharing code is enough, so

02:34 each repo on our GitHub has got description of what the project

02:38 set up to do. Information about the data source, whether or not you

02:41 can access that which you should be able to see if its open data.

02:46 And what else you need to do to reproduce or analysis so other

02:49 things you might install or

02:50 packages you need. So for us working openly is an ongoing

02:53 project, and we're kind of using our own team to experiment a bit

02:57 and test new and more open ways of working which

03:01 sometimes can be a bit challenging, but we feel like we

03:04 want to understand how we can work openly as a team.

03:07 And then start talking to other people about how to do it and

03:11 advise and what they can do.

03:13 So in In addition to putting things on GitHub we're also

03:17 sharing what we did and learned along the way and how

03:21 we overcame some of the barriers to working more openly. So

03:24 we published a blog post on Medium where we describe the

03:27 process we went through as a team to launch our public GitHub

03:31 account. And provide a kind of step-by-step guide for how

03:35 others can do the same thing.

03:38 So the problems we're working

03:39 on are kind of a big and really require a more

03:44 collaborative approach

03:46 to be actually be tackled and we believe that transparency

03:51 and working openly and reproducibly are essential to

03:54 achieving this. And we see open coding and open data as a good

03:59 analytical practice.

04:03 There are so many benefits of working openly. So first of

04:07 all, sharing our code helps us avoid duplicating work within

04:10 the team, but also allows other researchers and analysts to

04:14 review and validate and reuse our code. And this is something

04:17 we've definitely done for covid where I know I use code from an

04:22 anaylst in Nottingham and I know that other people have used our

04:26 code. It also makes it a lot easier to share our code and our

04:30 work internally so we

04:32 know what our colleagues are up to and we can celebrate progress

04:38 and it makes it a lot easier to credit people both internally

04:43 and externally. I think it also ... openness also just

04:47 makes for better science. So there's a permanent record of

04:51 the analysis that we've done, and it's reproducible.

04:54 I think also personally knowing that my code is going on GitHub

04:58 makes me go that extra mile.

05:01 Because I just make sure that there are enough comments to

05:04 explain what the code is doing and I haven't left things

05:06 in there that I don't want anymore and this makes it

05:09 easier for other people when they use my code, but it also

05:12 makes it a lot easier for future me to go back to a project.

05:17 So unsurprisingly, COVID-19 has changed the way we work at the

05:22 Health Foundation, but actually we realized that as we're all working

05:27 remotely, working openly has become even more important.

05:31 We have done a lot of reactive work related to the pandemic using

05:35 open data and this has allowed us to inform the national

05:39 policy debate and highlight the effect of covdi on

05:42 vulnerable groups such as care home residents and you can see

05:46 if you have the examples of some charts that we've created using

05:49 open data at the Health Foundation.

05:53 So even though it was great to use open data for this analysis,

05:57 there were some challenges with using open data. And sometimes

06:01 we were slightly limited in what we could do, and some of those

06:06 problems were related to the data itself, so it might not

06:10 have been granular enough, so we would, for example, not often be

06:14 able to look at both

06:17 sex and age at the same time, and sometimes the data wasn't

06:21 updated as frequently as we would have liked so

06:24 we would produce some analysis, and even when we

06:27 knew that more data would bring value to discussion, we

06:30 weren't able to do that.

06:33 There are also slightly more technical issues that we had.

06:37 So quite often open data isn't shared in a machine readable

06:41 format, meaning that there might be multiple tables in one

06:44 spreadsheet or there will be random text in places where a

06:48 computer wouldn't expect there to be text. It happened a few times

06:52 that I went to a website to download a new version of a data

06:56 set and it just disappeared and it was being published somewhere

07:00 else. And this as an analyst makes it really hard to build a

07:04 good workflow because

07:06 that data just is no longer there any work you've done to put in

07:11 kind of automatic downloads for anything. This is gone, when

07:14 there's no consistency. And it also obviously happened that

07:18 data changed format, which again poses a big problem for

07:22 an analyst because you need to go back and check your code.

07:25 But we do think that the people who are sharing data

07:30 want us to use it.

07:32 Especially when things are moving quickly, it's quite easy

07:35 to not share data in maybe the best way. So we wrote a blog

07:39 with some thoughts on how to collaborate in a time of crisis

07:42 and how to share data in a time of crisis with a focus on how to

07:46 make sure that if you're sharing something that people can

07:49 actually use it. So we wrote our own blog, but we

07:54 also know that there's a lot of good work out there on how

07:58 to share open data. ODI Leeds has got some open data tips.

08:02 The government statistical service has got some

08:04 excellent resources, and the Turin Way, which is slightly

08:07 more academic resources, got excellent advice on how to

08:10 make your research reproducible.

08:14 So a lot of the call of the action is coming from analysts

08:19 to kind of improve how we do analysis. But we also think that

08:23 data controllers could really, really help speed up the work

08:27 and spread open working.

08:29 So from a position as someone who actually does

08:32 this kind of work.

08:34 A lot of us use data managers, Digital ONS and it would just

08:38 be really, really good if

08:41 we could see some processing code, some reference

08:47 data and metadata.

08:48 As this would really help us with our work but also being

08:52 able to be sign-posted to the other people writing code for

08:55 the data that you are looking at would be really useful.

08:59 So I've focussed quite a lot on code and technology today, but

09:03 we don't think that's enough to collaborate effectively.

09:07 Based on the work that we've done internally to work more

09:12 openly, but also some of the work that we do

09:16 to support analysts in the sector, we have a few things

09:20 that we think might be stopping people from working openly.

09:23 And we think that the first thing you can do is ask yourself

09:27 why is my team not working openly?

09:29 And we think a big factor is buy-in for management.

09:33 Analysts really to feel safe sharing data and often that

09:36 needs to come from somewhere higher up to make sure that

09:40 there's that support.

09:41 There also needs to be support for the extra work during setup

09:45 because when you start working openly, at the beginning it is

09:48 a bit of extra work, but it's really worth it and I think in

09:52 the long run you're probably

09:53 saving time. But also having a community is really important,

09:57 so knowing that more people are working the same way provide

10:00 some reassurance that even though it's hard, this is this

10:03 is a good thing to do.

10:05 So if you want to start working more openly and you need to

10:09 knowledge that working this way might require a culture change.

10:14 So what do we do to support analysts? I've talked a lot about

10:20 analysis that we've done but a lot of the work that we do is actually

10:24 focussed on supporting analysts, working the health and

10:27 social care sector. So so far we've funded 43 projects through

10:32 four different rounds - something that we call advancing

10:36 applied analytics.

10:39 And we've had a lot of workshops with these award holders.

10:44 So the support doesn't stop with the funding.

10:48 We have these workshops and people come together and

10:52 troubleshoot issues they've encountered over the course of

10:56 their projects. When covid happened, we also set up a

11:00 repository of resources for analysts working to support

11:03 health and care during the COVID-19 pandemic, and they were

11:07 trying to share different solutions that people have

11:10 suggested that are

11:13 more open.

11:17 So we have a few projects that we just want to highlight.

11:24 But probably the biggest one being the NHS-R community,

11:27 which was, I think probably our big first investment in this area

11:31 and it's an initiative that aims to build a community of

11:34 analysts in the NHS and to spread the

11:37 use of the R language.

11:39 And through this we know that there's interest in the system

11:43 to work more openly, and we're hoping that our funding program

11:47 encourages an open approach to both sharing methodology and code

11:53 I want to end by focussing on where we're going next and have

11:57 a call for action. The last few months we've seen some

12:01 really remarkable innovation with health data, but we haven't

12:04 seen the same thing in social care and this is despite how hard the

12:08 pandemic has hit the social care sector. We're doing a few things.

12:12 First of all, we're teaming up with Future Care Capital

12:16 to do some work around social care. We really think

12:19 that open analytics will play a key role in sharing the learning

12:23 of these projects. And we're therefore launching a funding

12:27 programme in August for projects that can address barriers to

12:31 analytical capability and social

12:32 care? [inaudible] and if you think you can do some

12:38 good work in this area, please join us. Thank you.