Statistics

Confidence Intervals
1. Confidence that the effect from trial would be replicable to entire population
2. Mean difference= if includes 0 then not statically significant
3. Relative risk/Odds ration = 1 not statically significant
  1. Odds ratio= the odds of the outcome occuring with the exposure compared to (divided by) the odds of the outcome occuring without the exposure
    1. = 1 then same odds for both
    2. <1 then outcome is more likely to occur without exposure
    3. >1 then outcome more likely to occur with exposure
4. Width shows precision of estimate
5. 95% CI
  1. probability that population would lie between two values is 0.95
P value
1. Probability that the observed difference between 2 groups in down to chance
2. Needs to be below 0.05 to be statically significant
Number needed to treat is a reciprocal of absolute risk.
1. Absolute risk reduction- difference in probabilitiy that outcome will be achieved in control and intervention groups
  1. Probability that outcome will occur in exposed group Experimental Event Rate EER
  2. Probability that outcome will occur in non-exposed group Control Event Rate CER
    1. Relative risk = EER/CER
  3. Absolute risk reduction = CER-EER
2. Number needed to treat
  1. Number of people needed to treat for one person to benefit from intervention
Screening
1. Primary prevention
  1. Prevents disease before it occurs
    1. Other types of prevention
      1. Secondary
        Pick up disease early
      2. Tertiary
        Limit amount of damage disease does
2. Sensitivity
  1. How likely test is able to detect all cases
  2. If high low false negatives
  3. TP/TP+FN
3. Specificity
  1. How likely test is able to pick up just the disease you want
    1. TN/TN+FP
  2. If high= low false positives (high true negatives)
4. Positive predictive value
  1. TP/TP+FP
5. Negative predictive value
  1. TN/TN+FN
Heterogeneity
1. How similar studies are in meta-analysis
  1. Meta-analysis is pooling of quantitive data
    1. Systematic review
      1. Non- bias summary of data
    2. Displayed on forest plot
2. I2 test- 0-100% <25% little/no heterogeneity
3. Cochrane Q test
  1. P<0.05 heterogeneity= bad
Publication bias
1. Funnel plot test
2. Egger's or Begg's test
  1. smaller p= greater evidence of publication bias
3. Trim-and-fill method
  1. Calculates values with adjustments for publication bias
Fail safe N- number of studies required to change the statistical significance of the overall result
Standardised mortality ratio SMR= observed injuries/expected injuries *100
1. Expected injuries is from reference populatio
2. SMR of reference population =100
3. if >100 then worse than ref. population
Haddon Matrix
1. Approach to prevention
2. Table- pre event, event, post event on y axis
3. Host/agent/physical environment/social environment
Types of study
1. Case control- start with disease, look at exposure
  1. Odds of exposure in those with the disease/ odds of exposure in those without the disease
2. Cohort- start with exposure, look for disease
  1. Relative risk
    1. Risk of disease in exposed group/ risk of disease in unexposed group
Economics
1. Opportunity cost
  1. Benefits of then next best thing which you aren't choosing, hence missing out on
2. Cost of illness studies
  1. How much an illness costs
  2. Kinda useless
3. Economic Evaluation
  1. Compares multiple things on costs
  2. Takes into account costs and consequences (benefits)
  3. Systematic
  4. Types
    1. Cost effectiveness
      1. i.e £ per life saved/ case detected etc
    2. Cost-utility analysis
      1. £ per QALY
    3. Cost-benefit analysis
      1. best
      2. Benefit-cost in £
      3. Can only use if benefit can be measure in £

Media attachments

2552f753-7e55-4f19-82d9-00026ae830b7 (image/png)

Next up

Statistics

Description

Resource summary

Media attachments

Similar

	Created by Danielle Richardson over 8 years ago

	Copied by Pepe pepeone over 8 years ago