Big data, AI, data science and public health

Outline

Digital, big data, data science and machine learning/ AI might impact on public health

• Data and analysis
• Evidence synthesis
• Decision making tools
• Digital interventions
• Policy and priority
• Forecasting
• Evaluation and impact assessment
• “Precision public health”

Big data

Formal definition

95% unstructured

Bigness

• GBD 38 billion estimates
• Fingertips 20 million rows of data
• PHE data lake 10 billion rows of data
• Active 10 5Gb data per day
• 1:2 ngram document term matrix of corpus of DPH reports - ~ 1 billion columns

Types of big data for public health

Source: Mooney SJ, Pejaver V. Big Data in Public Health : Terminology , Machine Learning , and Privacy. Annual review of public health 2018;1–18. doi:29261408

Drivers

• Data growth (exabytes) (big data)
• Data variety
• Compute power - Moore’s Law, quantum computing
• Data science in govt.
• GDS
• Digital by design
• Data and technology standards and codes of practice
• Technology service manual
• Reproducible analytical pipelines
• Mainstreaming data science tools
• Precision public health
• Open data and democratisation of data
• Industrial strategy
• HDR-UK

Changing face of PH analysis

Pubmed search: 16000 most recent abstracts for ”machine learning/ ai public health” 

Data science processes

Exposomes and phenomes

Precision public health

https://rpubs.com/jflowers/428618

Various definitions but all encompass a set of consistent ideas: * Use of big data and new technology to improve health
* Data used to give more precise descriptions of populations and individuals
* Application of new techniques and methods
* Speed, accuracy and scale
* “the application and combination of new and existing technologies, which more precisely describe and analyse individuals and their environment over the life course, to tailor preventive interventions for at-risk groups and improve the overall health of the population.” (Weeramanthri et al. 2018)
* “improving the ability to prevent disease, promote health, and reduce health disparities in populations by applying emerging methods and technologies for measuring disease, pathogens, exposures, behaviors, and susceptibility in populations; and developing policies and targeted implementation programs to improve health” (Khoury and Galea 2016)
* “requires robust primary surveillance data, rapid application of sophisticated analytics to track the geographical distribution of disease, and the capacity to act on such information” (Dowell, Blazes, and Desmond-Hellmann 2016)
* “Precision public health is characterized by discovering, validating, and optimizing care strategies for well-characterized population strata” (Arnett and Claas 2016)

Precision

Modernising PHI

Modernising PHI

• Yesterday |

• Today |

• Tomorrow |

Modernising PHI

Current focus in PHE

• Data acquisition and consolidation
  • Surveys
  • Primary care
  • App data/ marketing
• Profile and indicator rationalisation
• R package strategy
  • fingertipsR
  • fingertipscharts
  • odsr
  • PHEindicatormethods
• Automation
• DHIP
• PHISy
• ICT challenges
• Data fluency

Machine learning examples: local health cluster analysis

• Local health contains ~ 60 small area indicators - many of which are highly correlated

• By using unsupervised machine learning techniques such as k-means analysis we can cluster areas into similar groups based on all the indicator data

• This shows distinct small area clustering in inner and outer London, coastal areas, northern industrial areas, affluent rural areas an so on (map)

• This is similar to ONS classification but is completely data driven

Scaled up

Machine learning examples: area obesity rates

• Can we predict area obesity rates from Health Profile data
• Use supervised machine learning - a range of algorithms to try and predict area obesity
• Penalised linear models make reasonable predictions (r2 ~ 0.84; rmse ~ 2% - estimates area obesity to within 2% on average ~ 8% error)

AI in literature searches and reviews

• AI is impacting on literature searching, review and synthesis…

http://rpubs.com/jflowers/441656

Novel data uses: Google street maps

## [1] "Certain features of this site make use of javascript. For maximum benefit it is strongly advised that you \r\n          switch on javascript before continuing."
## [2] "Use of search engine’s imagery forms part of the era of ‘data abundance’, according to the UK’s national statistician"                                          
## [3] "Credit: Byrion Smith/CC BY 2.0"

## [1] "(Credit: Shutterstock)"

App data