Fascinating to hear about the genesis, implementation, & development of the NHS coronavirus dashboard by @Pouriaaa at the #RadStats conference.
@RadicalStats
@RadicalStats
An awful lot of thought goes into how the data are displayed & described...
...it's a continuing process, with user research, &c.
#RadStats
...it's a continuing process, with user research, &c.
#RadStats
Not least accessibility...
(...which they certainly don't approach as a simple box-ticking exercise)
#RadStats
(...which they certainly don't approach as a simple box-ticking exercise)
#RadStats
They are sanguine about the (ab)uses made of the data by others...
...& learn from mistakes.
#RadStats
...& learn from mistakes.
#RadStats
Many data sources are complex to wrangle & reconcile...
...e.g. deaths—50 relevant metrics to consider!
#Radstats
...e.g. deaths—50 relevant metrics to consider!
#Radstats
And people use for for many reasons...
..."It has material impact on people's daily activities"
#Radstats
..."It has material impact on people's daily activities"
#Radstats
Many lessons for the future...
..."More important to do things fast, than to do them nice".
Someone needs to have the final say about how things are done, at the time. Decisions have to be made.
Need expertise in handling large volumes of data in (near) real time.
#Radstats
..."More important to do things fast, than to do them nice".
Someone needs to have the final say about how things are done, at the time. Decisions have to be made.
Need expertise in handling large volumes of data in (near) real time.
#Radstats
"It's just good luck that this happened—a few people in the right place at the right time"
NB Many people have made this possible by contributions.
#Radstats
NB Many people have made this possible by contributions.
#Radstats
@Pouriaaa has shown just how complex the process is of garnering, collating, quality controlling and displaying all this data — and in near real time...
...big shout out to him and @statsgeekclare & team!!
#Radstats
...big shout out to him and @statsgeekclare & team!!
#Radstats
What is 'agent-based modelling'...?
...involves people interacting with their environments, then simulates these different interactions.
#Radstats
...involves people interacting with their environments, then simulates these different interactions.
#Radstats
By June (other events occurred!) and the model began to be used for answering specific questions...
...both by modellers and by local decision-makers.
#Radstats
...both by modellers and by local decision-makers.
#Radstats
By October, more data, impending lockdown meant model could be and needed to be
modified, with better predicition of hospitalisation risks.
#Radstats
modified, with better predicition of hospitalisation risks.
#Radstats
In Novemeber the model became embedded in the Durham council dashboard...
...different asumptions with their outcomes shown
#Radstats
...different asumptions with their outcomes shown
#Radstats
In January, new variant, people's reduced activitiy, & immunisations necessitated adjustments to model.
#Radstats
#Radstats
Reflections:
1) No model is 'right', they do different things.
2) More potential improvements identified
#Radstats
1) No model is 'right', they do different things.
2) More potential improvements identified
#Radstats
NB both direct and indirect decision support are enhanced by this process—e.g. making assumptions explicit
#Radstats
#Radstats
The concept of 'justified stories' creating a coherent narrative around the data & assumptions...
#Radstats
#Radstats
Session 4 starting now
Mike Sandys (Director of Public Health for Leicestershire CC) on How a Director of Public Health has engaged with data during COVID and what issues have been experienced
#Radstats
Mike Sandys (Director of Public Health for Leicestershire CC) on How a Director of Public Health has engaged with data during COVID and what issues have been experienced
#Radstats
Leicestershire's 3 stages of data:
1) national lockdown #1 - little data
2) local restricition - dawn of decent data
3) second lockdown - data +++
#Radstats
1) national lockdown #1 - little data
2) local restricition - dawn of decent data
3) second lockdown - data +++
#Radstats
First phase: only a vague snapshot of the true picture...
...much fear because of lack of data.
Deaths largely located around local care homes.
#Radstats
...much fear because of lack of data.
Deaths largely located around local care homes.
#Radstats
By June, as data started becoming more available, better understanding of the situation developed, allowed better comms...
...but decisions influenced by central govenment's announcements.
#Radstats
...but decisions influenced by central govenment's announcements.
#Radstats
The local population couldn't be briefed about why the decisions were made, hence much restiveness...
...public engagement did have some positive effects
#Radstats
...public engagement did have some positive effects
#Radstats
Soon, as more data became avaialble, the disparity between cases in the community & hospital became clear & could be communicated to the public...
...good data allow pre-emptive comms, which is invaluable to support the public's understanding, engagement & response.
#Radstats
...good data allow pre-emptive comms, which is invaluable to support the public's understanding, engagement & response.
#Radstats
The current situation—data++...
...triangulating many channels of data (sewage data, heat maps, activity/exposure data, &c), both formal & informal.
#Radstats
...triangulating many channels of data (sewage data, heat maps, activity/exposure data, &c), both formal & informal.
#Radstats
Major developments, both in team, locality & public...
..should we teach children better stats!!
#Radstats
..should we teach children better stats!!
#Radstats
Next presentation:
Andrei Morgan on The impact of the Coronavirus pandemic on perinatal activity
#Radstats
Andrei Morgan on The impact of the Coronavirus pandemic on perinatal activity
#Radstats
Re EMBRACE inter-hospital neonatal (& in utero) transfer system, which covers 15 NHS trusts (~20 hospitals)
#Radstats
#Radstats
Trigger - two studies on reduced pre-term births & increased stillbirths during pandemic...
..wanted to investigate this
#Radstats
..wanted to investigate this
#Radstats
However not all data sources available or complete, therefore only option was to approach all the relevant hospitals...
...and the items of data sought were set out.
#Radstats
...and the items of data sought were set out.
#Radstats
Huge difficulty getting the data from trusts...
...black entries on second pic = missing data item
#Radstats
...black entries on second pic = missing data item
#Radstats
Could use multiple imputation (i.e. assume missing data by other data patterns) but may potential pitfalls—allowing bias— with this...
...didn't work!
#Radstats
...didn't work!
#Radstats
When looking only at 2020 in detail, wrt lockdown, no obvious pattern was identified re stillbirths & pre-term births...
#Radstats
#Radstats
Many caveats...
...but other studies from around the worls seem to concur—no obvious impact of pandemic on this issue.
#Radstats
...but other studies from around the worls seem to concur—no obvious impact of pandemic on this issue.
#Radstats
What next...
...more detailed data needed, but:
- prob no great link between SARS-CoV-2
- subtle behaviour changes may be occurring
- will take time
- watch this space!
NB
1) data cleaning vital
2) data collection is hard
3) silos hinder!
#Radstats
...more detailed data needed, but:
- prob no great link between SARS-CoV-2
- subtle behaviour changes may be occurring
- will take time
- watch this space!
NB
1) data cleaning vital
2) data collection is hard
3) silos hinder!
#Radstats
Now Session 5
Ted Schrecker on Health inequality in a post-pandemic world: the real grand challenges
#Radstats
Ted Schrecker on Health inequality in a post-pandemic world: the real grand challenges
#Radstats
Title slide & introduction...
...mainly focusing on US & Uk (& a little on Canada), & alluding to the 'Gate's grandes challenge' in public health
#Radstats
...mainly focusing on US & Uk (& a little on Canada), & alluding to the 'Gate's grandes challenge' in public health
#Radstats
Background - Halimi's 'inequality machine reshaping the planet'...
...and how we might 'build back better' post pandemic.
#Radstats
...and how we might 'build back better' post pandemic.
#Radstats
A schematic of the model envisaged by the speaker, re differential exposures, opportunities, vulnerabilities & consequences to the infection.
#Radstats
#Radstats
More examples...
...augmented by austerity, & drainage of resources from local authorities.
#Radstats
...augmented by austerity, & drainage of resources from local authorities.
#Radstats
This was not un-commented on—see @MichaelMarmot & @richardhorton1, for examples, among others...
#Radstats
#Radstats
Warning signals evident...
...Chicago has the widest intra-city health inequality in USA, & 'deaths of despair'.
#Radstats
...Chicago has the widest intra-city health inequality in USA, & 'deaths of despair'.
#Radstats
Although rescue measures will blunt the pandemic's effects in the short term, the imact in the long term will be massive...
#Radstats
#Radstats
Key concept in this is 'risk contagion' (Thelen, 2019)...
...problems in one domain trigger problems in others.
#Radstats
...problems in one domain trigger problems in others.
#Radstats
1. rapid growth rebound*
2. stagnation*
3. economic/health collapse (like post-USSR)*
4. equity-oriented redevelopment strategy
[*with ⬆ inequality]
#Radstats
2. stagnation*
3. economic/health collapse (like post-USSR)*
4. equity-oriented redevelopment strategy
[*with ⬆ inequality]
#Radstats
Concluding Session starting in about 5 mins...
After COVID – Where now for Radical Statistics?
#Radstats
After COVID – Where now for Radical Statistics?
#Radstats
I have to leave now, but thanks, #Radstats, for a fascinating day's content! 👍🙂
• • •
Missing some Tweet in this thread? You can try to
force a refresh
