Skip to main content

Individual Patient Risk Prediction

Marryat Stevens
Updated

MWDH has developed a number of predictive analytics using data from large cohort in Scotland and revalidated in London to cover a wider ethnic mix. These predict major macrovascular and microvascular complications of diabetes and other events. The number of models available will increase over time as the system develops and we will continue to revalidate models in wider data sets / geographies of deployment going forward.

The models may perform slightly differently to other models that you are familiar with like Q RISK (which can be accessed from the medication response page given they use different inputs and endpoints. We encourage you to continue to use these additional tools if a guideline specifically recommends that.

As with other decision support, these predictions are an aid and should not replace clinical decision making, which should make use of all the information available (noting that models only work off a limited data set). 

**Also note for amputation and blindness predictions – these are rare end points, so prediction percentages may be low; do not be falsely reassured that the patient is not at risk of developing progressing towards earlier events such as retinopathy or foot ulcers.

Model approach and input features

For information, the complications prediction models use a ‘metamodel’ approach, making use of these input features:

  1. Time to the beginning of the prediction window
  2. Time to the end of the prediction window
  3. Age at diagnosis
  4. Duration of diabetes
  5. Ethnicity
  6. Male/Female
  7. Current (most recent) Body Mass Index
  8. HbA1c: mean of all recorded values in the last 2 years
  9. HbA1c: current (most recent) value
  10. Systolic Blood Pressure: mean of all recorded values in the last 2 years
  11. Systolic Blood Pressure: current (most recent) value
  12. Current (most recent) HDL cholesterol
  13. Current (most recent) LDL cholesterol
  14. Current ratio of total to HDL cholesterol
  15. Current (most recent) estimated glomerular filtration rate
  16. Current (most recent) urine albumin-creatinine ratio
  17. History of cardiovascular disease
  18. History of peripheral vascular disease
  19. History of blindness in either eye
  20. History of congestive heart failure
  21. History of ischemic heart disease
  22. History of micro or macro albuminuria
  23. History of amputation
  24. History of stroke
  25. Current smoking status
  26. Past smoking status
Powered by Zendesk