PHEBP - Precision in Estimates of Treatment Effects Flashcards
What is Evidence Based Practice?
“integration of best research evidence with clinical expertise and patient values”
Treatment decisions based on:-
*Clinical expertise
*Preferences of the patient
*AND the best available evidence
To highlight and advance clinical effectiveness and
evidence-based practice (EBP) agendas, the Institute of Medicine set a goal that by 2020, 90% of clinical
decisions will be supported by accurate, timely and up-to-date clinical information and will reflect the best available evidence to achieve the best patient
outcomes.”
Where does the “evidence” come from & how good is it?
Information acquired about a new treatment or cause of a disease
- Data on individuals?
- Data on groups of individuals?
- Is it accurate?
- Is it precise
- Where did they come from?
Who are the population of interest?
*If studied everyone we would know the answer.. …
*Very difficult to do!
*Possible if routine sources of health data are
available… still limited
*Not necessary
*Things change over time
*Need updated information
Electronic data is gathered.
What is a world database?
WHO Mortality Database, age-and sex-specific coronary heart disease (CHD) & stroke mortality rates for the world’s most populous countries -1980 and 2010
death registrations & vital statistics
There are records of death and causes in countries. One of the overall trends is that the rates of mortality due to CVD has decreased. Rates are generally higher in men. Proportionally, CHD and strokes have gone down in men than in women - managed poorly in women.
How do you pick a sample?
You need to take a snapshot from the population.
*Most research is done on a snap-shot of population
Sample is drawn from a larger population
of interest
Sample tells us something about the population (ie all people to whom were are trying to make inferences about)
E.g. blood sample to estimate usual level of
cholesterol in a person
Aliquot of blood to represent all the blood in the
body
Aliquots from many people to represent e.g. lipid
level in people who suffer myocardial infarct
compared to those that do not
We use samples to provide estimates of population parameters
From Population to sample –who/how/where/how big?
*How common is CHD (not just CHD death)?
How high are CHD risk factors levels in England?What is the death rate from CHD in England c.w.
Scotland?
Estimate prevalence from random sample of the
population
E.g. Birth cohorts,GP practices information, Health Survey for England
If a “fair” sample of the population, we extrapolate
the findings to the whole population
Birth cohort - recruited people born in a certain period of time that were followed through time. Eg. Massive GP practice data base. Post code snapchat that gets sent a health survey. It needs to be accurate and big enough.
What is sampling bias?
Sample is selected in such a way that the individuals chosen are not representative of the whole study population.
Example:- estimating SBP in middle aged adults in England, but only those that are in employment will underestimate true answer
What is the level of systolic blood pressure (mmHg) in middle aged men in England?
To avoid bias:
Take random sample if possible of the population of interest
e.g.birth cohort individuals born with a certain week in a given year and followed up until adulthood
Or random sample of households by post-code sectors
Q: How do we calculate the 95% reference range?
Q. What does the 95% reference range tell us?
Crude Interpretation: raw values are “on average”
20mmHg away from the mean
- 68% of men will have values between 120 to 160mmHg
- 95% of men will have SBP values between 100 to 180mmHg
Is it possible to calculate a reference range for other estimates such as ……a single proportion, incidence or prevalence of disease, or risk?
- Because these are not continuous numerical data measures on an individual
- All of these estimates are a type of “group average” not a “data measurement” on an individual
- There is no concept of “individual data measurement”
Reference rages cannot be takes for anything that is not continuous data on a scale with units that has meaning. Groups cannot have these taken.
What is the accuracy and precision?
There is uncertainty in everything we measure or estimate how is this captured?
Accuracy is to do with getting the right answer – avoiding bias. Precision is to do with inherent
error associated with any estimate
What is the confidence interval for the mean value of systolic blood pressure in middle aged men in England
*How precise is the estimated average level of systolic BP based on a sample from the population?
*Need to understand that there is uncertainty in everything we estimate from data / measurement, a confidence interval is a measure of statistical uncertainty or “precision”
confidence intervals can be taken for anything in the world.
What are the confidence intervals?
*If we had series of studies, all trying
to estimate the same thing, and obtained the estimate of interest from each study what would the distribution of estimates look like …..
How would sample estimates be distributed?
This applies to ALL estimates of interest e.g.
Average or mean level
Mean value of biomedical measure
Proportion / prevalence
Incidence / rate etc…..
Comparisons between groups
Difference in means between groups
Risk differences
Relative risks
Survival rates ….. etc.
The spread of ‘estimates’ or ‘means of something of
interest’ is defined by the 95% confidence interval
