Introduction Class Part 1 Flashcards
What does prediction modelling aim to do?
Predict future (?) events or outcomes based on the pattern of available data using an mathematical algorithm.
What does prediction modelling involve?
Creating, testing, validating and evaluating a model to best predict an event or response.
- Clinical prediction models are tools to do what?
- What does this mean?
- “Aid health care providers [and service users] in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision making.” - Collins G
- This means that prediction models do not predict future events alone given that in diagnostic models we predict present events.
What does evidence based medicine focus upon?
Using randomized clinical trials (RCTs) to establish the best treatment for the average patient
What is considered to be the ‘highest level of evidence’ and why?
Randomized clinical trials (RCTs) as prevents
selection bias through random allocation
What are two caveats of randomised clinical trials?
Ignores statistical heterogeneity of patients
A new treatment is applied to all patients
What does precision medicine assume?
Assumes that reality is not homogeneous!
What is precision medicine defined as?
“An emerging approach for treatment and prevention that takes into account each person’s variability in genes, environment and lifestyle” -> Need to predict the best treatment for each patient
What is precision medicine made possible by?
Provision of “new” data: Imaging, Omics, patient records and sensors in smartphones and wearables and internet and smart phone usage
High performance computing technology that effectively collects, processes, stores and analyses “big data” (Health Informatics)
What do clinical prediction models aim to improve?
Patients outcome by enabling a more precise, personalized approach to health care and risk prevention
What do clinical prediction models inform health care providers and patients about?
- The risk of developing an disease
- The risk of the presence of a disease
- Future course of an illness
based on currently available information about the patient.
What are the 3 main types of clinical prediction models?
- Risk prediction models
- Diagnostic models
- Prognostic Models
a) Prognostic models (untreated)
b) Prognostic or Prediction models (treated) -> predicts likely benefit of treatment (Personalized medicine)
What do risk prediction models predict?
Risk to develop a major medical disorders.
What do Diagnostic models provide?
Risk estimates for the presence of disease (classification)
What do prognostic models provide information on?
The likely outcome of the disease in an untreated individual.
What do prognostic models predict?
Outcome of untreated patients
This helps identifying subpopulations of patients who are most likely to benefit from a given therapy or to assign the most likely best treatment to a patient (Personalized/Stratified medicine)- sometimes called prediction model
What are examples of risk calculators used by NHS?
QRISK®3 - calculates a person’s risk of developing a heart attack or stroke over the next 10 years.
People at high risk need to be recalled and assessed in more detail to reduce their risk of developing CVD and may be prescribed Statins
https://qrisk.org/three/
QDiabetes®- Estimates the risk of developing Type 2 Diabetes over the next ten years.
Based on the results, recommendations of prevention are recommended by GP
https://qdiabetes.org/index.php
Predict - Online tool that helps patients and clinicians see how different treatments for early invasive breast cancer might improve survival rates after surgery. https://breast.predict.nhs.uk/
Very few risk tools are implemented for mental health problems
True or false
True
Traditionally, clinical decision making follows after what?
Disease diagnosis
What does modern health care use?
Prognostic information, which asks whether a decision will affect an individual patient’s future outcome
What would be even more effective than traditional and modern clinical decision making?
Avoiding diseases by predicting the risk of developing it and offer prevention programmes and monitor health status.
In health care, what is prognosis the forecast of?
The course and future outcomes of a person’s disease, injury or other health conditions.
What is often distinguished between?
Prognosis of the untreated person and the treated
i.e. person should not get drugs with severe side effects if this person recovers without this treatment
What are two types of biomarkers?
A prognostic biomarker predicts the outcome independent of treatment (untreated)
A predictive biomarker predicts the benefit of a treatment, helps to select best treatment (treated)
Can biomarkers be both prognostic and predictive?
Yes
What are biomarkers?
Often mutations or polymorphisms within the genome, gene expression, blood measures, etc
What are examples of a prognostic biomarker?
Mutations within the genes BRCA1,BRCA2,ATMandP53 (whose products participate in DNA repair) predispose the patients to an increased risk of developing breast cancer.
What are examples of a predictive biomarker?
The analysis of theHER2gene amplification is the basic genetic test used for the treatment of patients with breast cancer with either trastuzumab or lapatinib.
A biomarker can be any…
“substance or biological structure that can be measured in the human body and may influence, explain or predict the incidence or outcome of disease”
What study designs does prediction modelling use?
Study designs from epidemiology - observational
Clinical trials
We want to develop a model which tells us how likely it is that a person diagnosed as ARMS (“At risk of mental state”) will develop a psychosis within 2 years.
How would you design a study for each question?
Which statistical analyses approach would you use for each study? (Think about regression models)
Cohort study: people classified as CARMS are followed prospectively and subsequent status evaluations with respect to a psychosis are conducted to determine which initial participants exposure characteristics (prognostic factors) are associated with it.
Analyses: Logistic regression with predictor variables collected at baseline and developed psychoses (yes/no) as binary outcome
We want to develop a model to predict which treatment, Cognitive Behavioural Therapy or Graded Exercise, will be more effective for a patient diagnosed with CFS.
How would you design a study for each question?
Which statistical analyses approach would you use for each study? (Think about regression models)
Randomized controlled trial:
Patients are randomly assigned to CBT or GE treatment.
Severity of CF score and potential predictors of treatment success are collected before randomization and CFS severity outcome after treatment.
Analyses:
Linear regression with moderators of treatment outcome (interaction between treatment arm and baseline variables) as independent variables and severity of symptoms as outcome
i.e. Treatment arm + Age + Interaction between arm and age - severity of symptoms
According to Kramer et al .2002 what do treatment moderators specify?
“ for whom or under what conditions the treatment works.”
What do treatment moderators inform clinicians of?
Which of their patients might be most responsive to the treatment and for which patients other, more appropriate, treatments might be sought.
-> Personalized/stratified medicine
