Lecture 13:Revision

1

Epidemiology

• Lecture 13Revision

2
Learning outcomes

• Apply epidemiological principles to evaluating
  health services
• Revise key features of epidemiological studies
  (study design, error, measures)

3
Evaluating health services

• What aspects of health services require
  evaluation?
• New drugs, programs, procedures, technology
• Changes in health care provision staff levels,
  in-hospital vs community care
• Problems that might arise in the system eg
  outbreaks, medication errors
• If services have the right focus or new services
  are needed

4
Scenario Evaluating a drug

• We have been asked to evaluate the introduction
  of a new drug for the hospitals drug committee.
• The drug is called tripolol and is used to
  treated glaucoma preventing intra-ocular
  pressure build up within the eye.

5
Scenario

• What is the intervention?
• Tripolol
• What is the outcome?
• Intraocular pressure
• What relationship are we seeking to identify?
• Causation Does tripolol reduce intraocular
  pressure?
• How will we measure the outcome?
• Tonometry

6
Epidemiological Study Designs
7
Epidemiological Study Designs
Target Population All people with glaucoma
Exposure/Study factor Intervention TRIPOLOL
Intervention
Intra-occular Pressure
Study Population/ Sample
No Intervention
8
Scenario

• What study designs could we use to evaluate the
  benefits and potential harms of the drug?
• RCT
• Cohort
• Before and After
• Case control

9
Randomised Controlled Trials

• An experiment in which subjects in a population
  are randomly allocated into groups
• to receive an experimental preventative or
  therapeutic procedure or intervention
  (intervention group) or
• to receive a placebo, no intervention or usual
  care (control group) and the outcomes are
  compared

10
Randomised Controlled Trials

• What are the key features of RCTs?
• All subjects free from outcome factor/disease at
  the commencement of the study
• Study pop defined geographically, temporally and
  by other exclusion criteria RCTs
• Subjects randomly allocated to an intervention
  group (study factor) or non-intervention (control
  placebo or established therapy)

11
Advantages Disadvantages

• Best evidence for causality
• ensures that individuals are allocated to the
  intervention or control groups without prejudice
  
• Minimizes/ eliminates unequal distribution of
  factors that influence clinical outcome between
  groups
• Facilitates statistical analysis

• Expensive (time and money)
• Organisationally difficult
• Difficult to recruit healthy professionals to
  participate
• Not always generalisable
• Sometimes ethical problems

12
Other important study designs

• Cohorts
• Case control studies
• Before and after studies

13
RCTs

• What problems/bias can occur?
• Selection bias
• Confounding
• Measurement bias
• Contamination

14
Bias

• systematic distortion or deviation in the study
  results from the true value
• ? overestimation or underestimation of the effect
  due to a deficiency in the design or execution of
  the study
• Results from systematic flaws in study design,
  data collection or analysis of results

15
Selection Bias

• Error due to the way the study participants are
  selected

16
Selection Bias occurs when

• Poor selection of subjects from the study
  population
• non-random selection
• ill defined populations
• failure to locate or unwillingness to
  participate
• loss due to health outcome (healthy worker
  effect)
• Non-random assignment of exposure
• Omission of subjects from analysis
• loss to follow-up

17
Confounding

• occurs when the measurement of the effect of an
  exposure (study factor) is distorted because of
  the association of exposure with other factor(s)
  that influence the outcome under study.
• presence of confounding ? mixing of effect of
  the study factor (exposure) with that of another
  factor(s)
• ? overestimate or underestimate of the true
  association between exposure and outcome

18
Confounder
Exposure
Outcome/Disease
Confounder

• Is an independent risk factor for the
  disease/outcome
• If removed ? changes association between exposure
  and outcome/disease
• Is NOT an intervening variable (ie not on the
  causal pathway between exposure outcome)

19
Information/Measurement Bias

• Systematic error in the measurement of
  information about exposure or outcome
• Error due to the incorrect classification of
  exposure or disease status

20
Types of Information Bias

• Recall bias (or subject error)
• Instrument error
• Follow-up or surveillance bias group with known
  exposure or outcome may be followed more closely
  or longer than the comparison group
• Hawthorne effect - people act different if they
  know they are being watched
• Observer bias observers may have preconceived
  expectations of what they should find in an
  examination (intraobserver interobserver bias)
• Interviewer bias - an interviewers knowledge
  may influence the structure of questions and the
  manner of presentation, which may influence
  response

21
Types of Information Bias

• Misclassification bias errors are made in
  classifying disease or exposure status
• non-differential misclassification error
• subjects are misclassified with respect to
  exposure in a non-systematic way.
• This weakens the association observed between
  outcome exposure, if a real association exists

22
Contamination

• that people in control group will receive part or
  all of intervention that is used for the
  intervention group
• ? reduces any differences between the two groups
• ? decreases the likelihood of identifying these
  differences.

23
Contamination

• may be caused by
• service providers or trials inadvertently
  applying trial interventions to the control
  group
• individual participants seeking additional care
  from providers outside of the trial
• other influences on usual care which are out of
  the control of the trial organisers

24
RCTs

• How can these problems/errors/bias be avoided?
• Selection bias
• Confounding
• Measurement bias
• Contamination

25
Controlling Selection Bias

• Choice of subjects from the target
• Random selection
• Have clear definition of the population eg area
  of residence, occupation, place of employment
• Encourage participation
• Consider how the health outcome may have affected
  the population (eg healthy worker effect)
• Use a control group
• Randomisation of exposure
• Ensure follow up of all participants and
  inclusion in analysis
• Intention to treat

26
Random Allocation of subjects

• Randomisation avoids bias by
• Ensures unpredictability of next assignment
• Reduces differences in risk between treatment and
  control groups.
• should make both groups similar in terms of the
  distribution of risk factors (potential
  confounders)
• larger the randomised groups, the greater the
  probability of equal baseline risks.

27
Intention to Treat

• Analyse the RCT data using the original groups to
  which participants were randomly allocated
• Maintains the original design of the study
• Preserves the external validity of the study

28
Reducing the effect of confounders

• In the design and conduct of the study by
• Randomisation
• Restriction (Allow only those into the study who
  fit into a narrow band of a potentially
  confounding variable)
• Matching in case control studies
• (Match cases and controls on the basis of the
  potential confounding variables especially age
  and gender)

29
Reducing the effect of confounders

• In the analysis of data
• Stratification
• Adjustment Statistical modeling
• eg Multiple Linear Regression, Logistic
  Regression, Proportional Hazards Model

30
Controlling information bias

• Blinding (or masking)
• Knowledge of whether the participant is in a
  treatment or control group can influence
  behaviour
• Standardised methods of data collection (staff
  training, calibrated instruments, standard
  procedures)

31
Blinding (or masking)

• Single blinding subject (participant) not given
  any information about whether allocated to
  treatment or comparison group
• usually via a placebo (inert agent usually
  indistinguishable from the active treatment)
• Double Blinding - neither subject or observer
  have any information about allocation of subject
  to treatment or comparison groups
• Minimises bias during assessment and care
• Triple blinding neither subject, observer or
  person analysing the data have any information
  about allocation of the subject to treatment or
  comparison groups
• Unblinded or open label studies no attempt at
  blinding

32
Measures of Association (Measures of Effect)

• An effect is the difference in disease/outcome
  occurrence between two groups who differ with
  respect to an exposure
• For example difference in occurrence of
  intraocular pressure between people prescribed
  Tripolol (exposed) and people not prescribed
  tripolol (unexposed)

33
Measures of Association

• Examine the association between exposure and
  development or incidence of outcome/disease
• To assess whether differences in incidence are
  related to differences in exposure to a factor

34
Measures of Association

• relative measures
• give an estimation of how much more or less a
  person exposed to the study factor is at risk of
  developing the outcome relative or compared to
  the people not exposed to the study factor

35
Calculating Measures of Association

• Two-way tables a convenient way of ordering
  categorical data

36
Two way Table
Whos in cell a cell b cell c cell d?
37
Relative Risk /Risk Ratio (RR)

• Estimates the magnitude of association between
  exposure and outcome
• Indicates likelihood of developing
  disease/outcome in the exposed group relative to
  people not exposed

38
Two way Table

• CI exposed 10
• CI unexposed 30
• Relative Risk CI exposed /CI unexposed
• 10/30 0.33
• People taking tripolol have one third the risk
  of developing IOP compared with those taken
  placebo

39
Relative Risk
Risk is lower in the exposed
Risk is higher in the exposed
RR
1.0 no association
40
Odds Ratio

• Another measure of association that can be used
  in case-control studies, cohort studies and RCTs
• is an estimate of the relative risk in studies
  where we cannot calculate incidence

41
Odds Ratio (Relative Odds)

• Is the odds of exposure in the diseased divided
  by the odds of exposure in the non-diseased.
• When a disease is rare,
• the odds ratio relative risk

42
Odds Ratio

• Odds Ratio ad
• bc

43
Odds Ratio

• Odds Ratio ad 21000 1.5 bc 14000
• P

44
Measuring the precision of the results

• Tests for Statistical Significance
• show you the precision of the results of a study
  by examining the confidence intervals or the p
  values help you to evaluate whether the study was
  statistically significant.

45
Confidence intervals

• show a range within which the true effect of the
  intervention is likely to be.
• a confidence interval that includes the value of
  no effect (e.g. RR1 or RRR0) shows that the
  intervention group is not statistically
  significantly different from the control.

46
Confidence intervals

• Where the confidence interval does not include
  the no effect value this shows that there is a
  statistically significant different between the
  intervention and control group.
• Statistical significance is usually measured
  using a 95 confidence interval, meaning that if
  the study is repeated multiple times, 95 of the
  studies will have result within that range.

47
p-value

• reflects the degree of certainty about the
  existence of a true effect.
• based on the supposition that the null hypothesis
  is true i.e. that there is no true difference
  between the intervention and control groups
• Statistical significance is usually set at
  plt0.05 or plt0.01.

48
Statistical and practical significance

• When study results are not statistically
  significant you will usually decide that there
  is no association
• But in some cases this may not be true
• eg the intervention may have a real effect
  (practical significance) as judged by the size of
  the effect but the sample size was too small to
  be statistically significant.

49
Scenario Conducting a needs assessment

• You have been asked by the Area Health Board to
  determine whether a new diabetes service should
  be set up in Hobbitown.

50
What do you need to know?

• Incidence
• Prevalence
• Mortality
• Availability of services now
• Risk factors

51
To find out whether a community is healthy or
unhealthy or needs a service

• first measure one or more indicators of health
• Incidence
• Prevalence
• Mortality
• Risk factors
• Availability of services now
• compare the results with another community or
  group.

52
When measuring health indicators it is important
to define

• What is being measured
• Person ie individuals included
• Place or location of the study population
• Time period of the study

53
Incidence

• Incidence Number of new cases or events in a
  population, over a defined period of time.

54
Cumulative Incidence

• Cumulative incidence (CI) is the proportion of
  people in a population who became diseased or ill
  or experienced an event during the specified
  period of time.
• CI No new cases of disease or events
  during time period Total population at risk at
  the beginning of the time period

55
Cumulative Incidence

• Two assumptions when calculating CI
• entire population at risk has been followed from
  the beginning of the study till the end
• all participants are at risk of the outcome of
  interest

56
Incidence Rate (Incidence Density)

• The incidence rate or incidence density is the
  number of new cases in a population divided by
  the total time units each individual in the
  population at risk was observed.
• Incidence Rate
• No new cases of disease/events during the
  specified time period
• Sum of the length of time during which each
  person in the population is at risk

57
Incidence Rate

• can be presented in many different ways.
• 10 cases/1000 person-years
• 1 case/100 person-years
• 0.1 cases/10 person-years
• 0.01 cases/1 person-year
• These are all the same

58
Incidence Rate

• In many circumstances, you can assume
• that entry and exit from the population occurs
  evenly over the time period, or
• you only know the average population at risk,
  an approximate incidence density rate can then be
  estimated as
• Incidence Rate
• No new cases of disease/events during the
  specified time period
• (Initial population at risk final population)
  /2 in the time period

59
Mortality Rate

• Crude Mortality Rate the incidence of deaths
  from all causes (all cause mortality rate) for
  the Australian population in one year
• Crude All Cause Mortality Rate 2001
• No. new deaths during 2001
• Total Aust. population at risk
• midyear 2001

60
Comparison of crude death rates for indigenous
Australians and all Australians in 1995-97

• Are there differences between the crude deaths
  rates for indigenous Australian compared with all
  Australians?

(AIHW, 2001)
61
Comparison of crude death rates for indigenous
Australians and all Australians in 1995-97

• However we know
• death is closely related to age
• the age structure of the Indigenous population is
  very different to that of the total population
• Therefore we need to adjust for the effect of the
  age structure to make a meaningful comparison.
• This can be done using standardization

62
Direct Standardisation

• is used to compare large populations
• uses a standard reference population to compare
  both populations
• applies the age-specific disease/death rates of
  the population of interest to the standard
  population
• allows us to compare death rates, by calculating
  what their death rates would be if the
  populations of interest had the same age
  population structure as the reference population.

63
Key points to remember for Direct Standardisation

• Select/Identify a standard population
• Calculate the age-specific rates for each of the
  population of interest
• Calculate the expected deaths for each age group
  of the populations of interest by multiplying
  their age-specific death rates by the age-group
  population of the standard population
• Sum the expected deaths for each population of
  interest
• The age-standardised death rate
• sum of expected number of deaths
• (population of interest )
• total population (standard population)

64
Comparison of all death rates for indigenous
Australians all Australians in 1995-97
65
Prevalence

• Prevalence is the proportion of a defined
  population with the disease/event of interest at
  a specified time period.
• Prevalence is usually established by
  cross-sectional surveys
• An incident case becomes a prevalent case and
  remains a prevalent case until recovery or death.
• Where a population is in a steady state,
  prevalence depends on incidence and duration of
  disease.
• ? prevalence of a disease may increase when
  incidence remains stable but survival of cases
  improves

66
Factors influencing prevalence rate
67
Prevalence

• Point Prevalence
• Total number of the population with the
  disease/event at a particular time
• Total population at that time

68
Period Prevalence

• Period Prevalence
• Number of the population with the disease/event
  at any time during a specified period
• Total population during that period

69
Use of Incidence and Prevalence

• If one wishes to look at a change in disease (eg
  studies of causality, acute conditions or events,
  outbreak investigation) ? use incidence.
• For example Looking at the change in the
  incidence of cancer is important to know whether
  current prevention, screening and treatment
  activities are working.

70
Use of Incidence and Prevalence

• Prevalence is used when looking at the magnitude
  of existing diseases usually chronic disease like
  diabetes where change does not occur rapidly
• Often both measures are used

71
Where do you get the data?

• Surveillance systems already operating
• Individual data sources
• Collect the data

72
What happens when you get the data?

• Analyse the data
• Interpret the data

73
Trends in Diabetes Prevalence, US (CDC)
74
Diabetes Prevalence, US (CDC)
75
Diabetes Trends Among Adults in the
U.S.,(Includes Gestational Diabetes) BRFSS,
1990,1995 and 2001
Source Mokdad et al., Diabetes Care
2000231278-83 J Am Med Assoc 200128610.
76
Obesity Trends Among U.S. AdultsBRFSS, 1985
Source Mokdad A H, et al. J Am Med Assoc
199928216, 200128610.
77
Obesity Trends Among U.S. AdultsBRFSS, 2001
Source Mokdad A H, et al. J Am Med Assoc
199928216, 200128610.
78
Other information

• Deaths due to diabetes
• Comparison of treatment or outcomes for people
  with diabetes
• Amputation rates
• Medical Compliance with diabetes guidelines

79
Investigating Outbreaks

• Occurrence of more cases of disease than
  expected in a given area among a specific group
  of people over a particular period of time
• Two or more linked cases of the same illness
• equivalent to epidemic
• Outbreaks are most frequently associated with
  communicable diseases

80
Investigating Outbreaks

• Endemic constant or habitual presence of a
  disease within a given geographical area or the
  usual prevalence of a given disease within such
  area
• eg malaria in many areas of Africa
• Pandemic worldwide epidemic
• eg Plague, influenza

81
Outbreak

• Endemic constant or habitual presence of a
  disease within a given geographical area or the
  usual prevalence of a given disease within such
  area
• eg malaria in many areas of Africa
• Pandemic worldwide epidemic
• eg Plague, influenza

82
Steps in an Outbreak Investigation

• Verify outbreak and confirm diagnosis
• Develop case definitions
• Identify cases and obtain information
• Collect data, analyse and appraise
• Formulate and test hypothesis
• Introduce control measures
• Conduct special studies
• Communication, including outbreak report

83
Attack Rate

• Attack rate is a type of cumulative incidence
  applied to a narrowly defined population observed
  for a limited period of time, such as during an
  epidemic.
• Attack rate
• No new cases of illness during a specified
  time period
• Total population at risk during that
  specified period

84
Outbreak Management - key objectives

• Anticipation - in order to prevent an epidemic
  occurring
• Preparation - for quick and effective response
• Early detection - to know when there is a problem
• Rapid Investigation - to describe the event and
  identify interventions
• Effective Response - to implement appropriate
  control measures
• Evaluation - to identify what achievements and
  failures before and during the outbreak for
  future prevention and more effective response.

85
Effective Outbreak Management
Anticipation/Prediction
Evaluation
Preparedness
Coordinated effective investigation and response
Early warning/detection
86
Outbreak Detection and Response
CASES
DAY
87
Outbreak Detection and Response
CASES
DAY
88
Scenario evaluating a screening program

• Screening
• Aims to reduce morbidity and mortality from that
  disease among persons being screened
• Is the application of a relatively simple,
  inexpensive test, examinations or other
  procedures to people who are asymptomatic, for
  the purpose of classifying them with respect to
  their likelihood of having a particular disease
• a means of identifying persons at increased risk
  for the presence of disease, who warrant further
  evaluation

89
Screening

• Screening test is not used for diagnosing illness
  
• Those who test positive to the screening test are
  sent for further evaluation with one or more
  diagnostic tests to determine if they have the
  disease

90
Natural History of Disease
Preclinical Phase
Clinical Phase
Progression of the disease
Biological Onset
Symptoms Begin
Death
S
D
O
Y
Disease detectable via screening
Detectable Preclinical Phase
91
Validity

• A screening test should provide a good
  preliminary indication of which individuals
  actually have the disease and which do not.
• Two components
• Sensitivity and
• Specificity

92
Sensitivity

• Proportion of persons with pre-clinical disease
  who screen positive
• The probability of screening positive if the
  disease is truly present
• a/(ac)

93
Specificity

• Proportion of persons without pre-clinical
  disease who screen negative
• The probability of screening negative if the
  disease is truly absent
• d/(bd)

94
Predictive value positive

• Probability that a person actually has the
  disease, given the results of the screening test
• Proportion of persons with a positive screening
  test who have pre-clinical disease
• PV a
• a b

95
Predictive value negative

• Probability that a person is truly disease free,
  given the negative results of the screening test
• Proportion of persons with a negative screening
  test who do not have preclinical disease
• PV- d
• c d

96
Reliability

• Consistency of results when repeat examinations
  are performed on the same person under the same
  conditions
• Sources of variability are due to
• Biological variation
• Reliability of the instrument
• Intra-observer variation
• Inter observer variation

97
Beware of Bias when evaluating Screening Programs

• Lead time bias
• Length bias
• Volunteer bias

98
Epidemiology Examination

• Week 15 You must confirm time and date in your
  exam timetable at
• http//sas1.fhs.usyd.edu.au/Students.htmlexmt
• You must bring a non programmable calculator
• See approved calculators at
• http//sas1.fhs.usyd.edu.au/Approved.html