ACMT Board Review Course Population Health and Assessments

1

ACMT Board Review
Course Population Health and Assessments

• Jeffrey Brent, M.D., Ph.D.
• Toxicology Associates
• University of Colorado
• School of Medicine
• and
• Colorado School of Public Health

2
Topics for this Lecture

1. Exposure monitoring and sampling
2. PPE
3. Study designs and measures of association
4. Statistical concepts
5. Bias and confounding
6. The Hill criteria
7. Sensitivity, specificity, predictive values

3
Exposure monitoring and sampling

• Exposure monitoring
• Environmental sampling
• Wipe sampling
• Water sampling
• Air sampling
• Breathing zone measurements are best for
  inhalational exposures
• Biological monitoring e.g.
• Blood Pb
• Urine mercury

4
Personal protective equipment

• Respiratory
• Chemically protective clothing

5
(No Transcript)
6
Respiratory protection

• Classification by size
• Quarter face
• Half face
• Full face
• Classification by function
• Air-purifying
• Uses chemical specific cartridges
• Supplied air
• SCBA

7
Protection Factor

• The factor by which exposure is reduced by use of
  a respirator
• Ambient/protection factor exposure
• For example
• Ambient of 100 PPM
• Protection factor of 10
• Exposure 100/10 10 PPM
• Protection factors range form 5 10,000
• The goal is to get exposure to below safe limits

8
(No Transcript)
9
Chemically protective clothing

• Simple protection
• Ex. Aprons, boots, gloves
• Nonencpasulating suits
• 1 or 2 pieces, for example
• 1 piece hooded coveralls
• Hooded jacket chem protective pants
• Encapsulating suit
• Highest level of protection

10
Chemically protective clothing is usually
designated by the EPA rating system

• Level A Max protection
• Encapsulating suit
• SCBA
• Level B
• Supplied air respirator (or SCBA)
• Non-encapsulating garment
• Level C
• Air purifying respirator
• Non-encapsulating garment
• Level D standard work clothes

11
Study design and measures of associationStatistic
al concepts

• NEXT

12
Types of human data

• Anecdotal
• Case-reports and series
• Controlled observational
• Controlled epidemiological studies
• Controlled interventional
• Trials

13
Controlled observational studies

• Cohort
• Cross-sectional
• Mortality
• Case-control
• Ecologic
• For all epi studies
• Groups should be matched for relevant variables
  (e.g.)
• Age
• Sex
• Anything else that can affect results

14
Cohort studies

• Compares exposed group to an unexposed group
• Can be retrospective or prospective
• Can assess incidence rates
• Incidence Rate of new cases
• (e.g. Cases/100,00/yr)
• Prevalence Number of cases in the population
• (e.g. cases/100,000)
• Results expressed as Relative Risk (aka risk
  ratio or rate ratio)

15
Cross-sectional studies

• Compares exposed group to an unexposed group at
  one snapshot in time
• Provides prevalence data
• Example Prevalence of drug abuse in medial
  toxicologists taking the board exam v those that
  are not
• Results expressed as Relative Risk (aka risk
  ratio or rate ratio)

16
Mortality Studies

• Typically a variation of a cohort study
• Assesses diagnoses at time of death
• Results expressed as mortality rates corrected
  for relevant factors (Standardized mortality
  rates)
• Usually expressed as a percentage
• (Mortality rate of exposed/rate in unexposed) X
  100 SMR)
• Thus an SMR of 100 no difference btw exposed
  and unexposed

17
Case-control studies

• Compares individuals with a specific condition
  with individuals that do not have that condition
  and compares exposures (or other risk factors)
• For example Comparing medical toxicologists with
  alcohol abuse (the cases) with those w/o this dx
  (controls) to see if there is a higher likelihood
  of alcoholic abuse if preparing to take the
  boards.
• Thus assesses risk factors (e.g. exposures)
  related to specific conditions
• Recall bias major problem
• Results expressed as Odds Ratios

18
Ecologic Studies

• Assesses population numbers, not individuals
• Example Rate of admission for asthma
  exacerbations in a city with high airborne PM10
  compared to a city with low PM10
• Results expressed as Relative Risk (aka risk
  ratio or rate ratio)
• Ex Snows study of cholera rates in London
  districts

19
Assessment of results of epi studies

• By convention a result is statistically
  significant if the likelihood that it is chance
  result is lt 5 or approx 2SDs from the mean

20
Interpretation of EPI Data

• You can never assess the degree of association
  based only on the magnitude of the RR, OR or SMR
• These values have an inherent uncertainty that is
  determined by the nature of the data
• In modern epi this uncertainty is expressed as
  Confidence Intervals

21
The 95 Convention

• In science the uncertainty in a result is
  expressed as that range of data in which there is
  a 95 likelihood that the real value exists
• CIs express this range
• Ex RR 1.6 (CI 0.7 2.4)

22
The importance of confidence intervals

• If RR 1.6 (0.7 2.4)
• Than there is a 95 likelihood that the real
  value lies between 0.7 2.4
• If the real RR is
• gt1 association
• 1 Non-association
• lt1 negative association (protective effect)
• The 95 rule defines statistical significance
• Thus, in order to be a statistically significant
  result the CI must not include 1

23
What about p values?

• p Values are an older way of describing
  statistical significance
• P lt 0.05 means a result is statistically
  significant
• OR 1.6 (0.7 2.4) OR 1.6 p gt 0.05
• OR 1.6 (1.1 2.1) OR 1.6 p lt 0.05

24
Now the Bad NewsA statistical relationship never
a priori means a causal relationship
25
It is not the falling of the leaves that causes
winter to come

• There are many more statistical associations in
  toxicology than there are causal relationships

26
How to get from association to causation

• Requires specific rigorous methodology
• Stems from Doll and Hills observation of an
  association between smoking and lung cancer

27
Hills Viewpoints

• To be applied if a statistical association is
  shown to exist
• Does not account for quality of studies showing
  such an association

28
Hills Viewpoints

• Strength of association
• Consistency
• Specificity
• Biological gradient
• Temporal precedence
• Coherence
• Plausibility
• Experimental support
• Analogy
• Also must consider the quality of the study

29
Bias and confounding

• Bias systematic error
• Ex You are doing a study on childhood bl Pb
  concentrations and behavior. However, your lab
  technique inflates blood lead values by 20 a
  bias.
• Confounding uncontrolled for factor affecting
  results.
• Ex You are doing a retrospective cohort study on
  chronic exposures to phosgene in laboratory
  workers and the incidence of lung cancer but you
  do not control for smoking.
• Smoking is a confounder

30
A little tip -Know how to calculate sensitivity,
specificity, and predictive values
31
Sensitivity

• The likelihood of a test being positive if the
  condition is present
• Ex Being under 16 yrs old has 100 sensitivity
  for the detection of childhood Pb poisoning.
• Good for screening (few false negatives (FN))
• Sensitivity True positives (TP)/(TP FN)
• Sensitivity is often expressed as a
• In example above if screen 100 individuals and 10
  had Pb poisoning Sens 10/(100) 10/10 1
  (or 100)

32
Another example

• To determine the sensitivity of a terminal R in
  lead AVR for the detecting of Na channel
  antagonist toxicity in all OD patients.
• Screen 1,000 EKGs of OD patients, 100 had ODd on
  Na channel blockers and 80 had a terminal R
  wave (TPs). 50 had a terminal R wave but did not
  OD on these agents.
• TP 80
• FN 20
• Sens TP/(TP FN) 80/(80 20)
• 80/100 0.8 (80)

33
Specificity

• The likelihood of the unaffected individuals
  correctly having a negative test
• Test using criteria of being under 16 for dx of
  childhood Pb poisoning.
• N 100
• 10 with Pb poisoning - the other 90 are false
  positives (FP)
• Specificity True neg (TN)/(TN FP) 0/090 0
  

34
The second experiment

• Screen 1,000 EKGs of OD patients, 100 had ODd on
  Na channel blockers and 80 had a terminal R
  wave. 50 others had a terminal R wave but did not
  OD on these agents (FPs).
• TN 850
• FP 50
• Sp TN/(TNFP) 850/(85050)
• 850/900 0.94 (94)

35
Comparison btw Sensitivity and specificity

• Both True/(True False)
• Sens TP/(TPFN)
• Specificity is the mirror image
• Spec TN/(TNFP)
• For both the trues in the numerator and
  denominator terms are the same.
• The other denominator term is the complete
  opposite

36
Positive predicative value

• PPV likelihood that the test will correctly Dx
  the condition
• Test using criteria of being under 16 for dx of
  childhood Pb poisoning.
• N 100
• 10 with Pb poisoning (TP) - the other 90 are
  false positives (FP)
• PPV TP/(TPFP) 10/(10 90) 0.1
• So 10 PPV

37
PPV the second experiment

• Screen 1,000 EKGs of OD patients, 100 had ODd on
  Na channel blockers and 80 had a terminal R
  wave (TP). 50 others had a terminal R wave but
  did not OD on these agents (FPs).
• TP 80
• FP 50
• PPV TP/(TP FP) 80/(8050) 80/130
• 0.6

38
Negative predicative value

• The likelihood that the disease is not present if
  the test is negative
• Test using criteria of being under 16 for dx of
  childhood Pb poisoning.
• N 100
• 0 are TN
• 0 are FN
• NPV TN/(TNFN) 0/(00) 1 (100)

39
NPV a more rational study

• Screen 1,000 EKGs of OD patients, 100 had ODd on
  Na channel blockers and 80 had a terminal R
  wave. 50 others had a terminal R wave but did not
  OD on these agents (FPs).
• NPV TN/(TN FN)
• TN 850
• FN 20
• NPV 850/(85020) 850/870 0.97

40
Predicative values - summary

• PPV uses only positive terms
• PPV TP/(TPFP)
• NPV uses only negative terms and is exactly
  opposite of the PPV
• NPV TN/(TNFN)

41
If, when you are studying, this you have any
questions call me (24/7) _at_ 303-765-3800 or e-mail
me at Jeffrey.Brent_at_ucdenver.edu