1 / 39

Big Data: Why Big is Beautiful Jill Pell

Big Data: Why Big is Beautiful Jill Pell. Scotland: a research laboratory. Population - 5.1 M Stable - net migration <10k Devolved health service Vast majority of healthcare from sole provider (NHS) Comprehensive, high quality data

knisley
Télécharger la présentation

Big Data: Why Big is Beautiful Jill Pell

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Big Data: Why Big is Beautiful Jill Pell

  2. Scotland: a research laboratory Population - 5.1 M Stable - net migration <10k Devolved health service Vast majority of healthcare from sole provider (NHS) Comprehensive, high quality data Significant investment in health informatics infrastructure, governance and capacity: Farr, ADRC, Urban Big Data Centre

  3. Data, data everywhere ….. Maternity records Breastfeeding Immunisation Growth Cause of death The data “life-cycle” Surgical procedures Development Investigations Screening Hospital admission GP / OPD attendance Medicines

  4. Health Sector Databases

  5. Trials & Cohort Studies • The gold standards • Increasing in number • Low hanging fruit harvested • Increasing in size • Therefore cost increasing Published RCTs per annum

  6. Since 1990, one new cohort per annum • ½ followed up for >20yrs • Includes: • Million Women Study • UK Biobank • £28M pa on the 34 largest, ongoing UK population cohort studies

  7. Using routine data as anadjunct to traditional methods • Increase VFM • Reducing costs • Providing additional information • Selection • Stratification • Exposures, confounders • Follow-up

  8. Randomised controlled trials Long-term follow-up of WOSCoPS All-cause mortality CHD mortality

  9. UK Biobank

  10. Using routine data to conduct studies Systematic review of 98 clinical trials: • 29% of potential subjects excluded by trials • A further 29% refuse to participate Melberg HO, Humphreys K. Ineligibility and refusal to participate in randomised trials of treatments for drug dependence. Drug Alcohol Rev 2010;29(2):193-201.

  11. Natural experiments “Back to sleep” campaign and SIDS Smoke-free legislation and preterm deliveries Smoke-free legislation and stroke: cerebral infarction vs intracerebral haemorrhage

  12. Health Service Research: Hospital volume of throughput and MACE 30 days post PCI Heart 2006;92:1667-72.

  13. Time of delivery and neonatal death

  14. Scottish Veterans Cohort Study • Retrospective cohort study • 57,000 veterans in Scotland born 1945-1985 • 173,000 individuals with no record of military service, matched for age, sex and postcode sector • Service entry/exit flagged on primary care records • Linked to acute/psychiatric hospitalisations, cancer registrations and deaths Peptic ulcer by length of service

  15. Individual and inter-generational linkage

  16. Inter-generational record linkagePregnancy complications and maternal/grandmaternal IHD

  17. New datasets Prescribing Information System (PIS) • Picture Archiving and CommunicaitonsSysem (PACS)

  18. Prescribing Information System (PIS) • Pharmaco-epidemiology; pharmaco-vigilence; precision medicine • Information on medicines prescribed, dispensed and reimbursed • Covers all NHS prescriptions dispensed in the community • Includes prescribed by GP, practice nurse, dentist and hospital • Does not cover hospital prescriptions dispensed in hospital • Covers whole Scottish population (5.2 million) • Includes prescriptions issued in England but dispensed in Scotland • 1993 - set up for budgeting/reimbursement – provided aggregated data • 2009 - achieved 100% CHI coverage – individual level data • Linkable to other databases • Currently (1993-2014) data on: • 507 million medications prescribed • 344 million medications dispensed • InFormation on • Prescriber • Dispenser • Medication – manufacturer, formulation, stregth, dose

  19. Potential uses • Drug as intervention • Outcomes • Clinical / cost effectiveness • Adverse outcomes / ADRs • Precision medicine • Interactions / sub-groups • Health services research • Patterns of usage; health inequalities • Factors associated with uptake/compliance • Drug as proxy of disease

  20. Wider determinants (and outcomes) of health

  21. Individual-level data EDUCATION PENSIONS HOUSING WORK SOCIAL SUPPORT CRIMINAL JUSTICE BENEFITS HEALTH

  22. Aggregated data • Pollution • Climate • Green space • Public transport • Leisure facilities • Tobacco/alcohol/fast-food outlets

  23. Education linkage • Annual pupil census • Record of special education need • Cause • Absenteeism / exclusion • Free school meals • SVQ – exam results • School leaver destination – job, HEI, unemployed

  24. + SMR 1/4/6 Mother SMR2 SMR11 CHS-P/PS SMR 1/4/6 SMR 1/4/6 ScotXeD Child + SCI/DC PIS PIS PIS

  25. Gestation of delivery and SEN Population attributable percentage Prevalence Odds ratio • Gestation overall 10% • Preterm 3.6% • Early term 5.5%

  26. Breech presentation

  27. Month of conception

  28. Results • 774,079 pupils across 2009-2013 • 3,363 (0.43%) pupils treated for diabetes • 46,403 (6.24%) asthma • 5,374 (0.69%) epilepsy • 7,488 (0.97%) ADHD • 5,386 (0.72%) depression

  29. Results - Education

  30. Subgroup analyses - Exclusions

  31. Big Data offers scope for Innovation using new methods to do the same things Innovation “plus” Using new methods to do new things

More Related