Catriona: /* Notes: Key concepts for measures of disease and association */

2014-07-25T07:31:20Z

‎Notes: Key concepts for measures of disease and association

Catriona: Created page with "===Notes: Key concepts for measures of disease and association=== This is a refresher on information already learnt during the field epidemiology training course. This will h..."

2014-07-25T07:30:43Z

Created page with "===Notes: Key concepts for measures of disease and association=== This is a refresher on information already learnt during the field epidemiology training course. This will h..."

Halaman baru

===Notes: Key concepts for measures of disease and association===

This is a refresher on information already learnt during the field epidemiology training course. This will help with step 4 (hypothesis testing).

'''Measure of disease frequency'''

It is important to be able to describe the amount of disease (or some other measure) present in a population. Prevalence data is one means of doing this.

Prevalence data measures the amount of disease at a point in time (or defined time interval).

For example if you had a herd of 100 cattle and you sampled them all at one time and found 50 cattle had ''Brucella'' antibodies, the prevalence of ''Brucella'' antibodies would be 0.50 (50%). Your conclusion is that 50% of the cattle had been previously exposed to ''Brucella''. However, you are not sure if this was due to a recent natural infection, a natural infection that occurred a long time ago or due to something else such as vaccination against ''Brucella''.

<math>\mathrm{Prevalence~}\mathrm{of~}\mathit{Brucella~}\mathrm{antibodies}=\frac{\mathrm{numerator}}{\mathrm{denominator}}=\frac{50\mathrm{~antibody~}\mathrm{positive}}{100\mathrm{~in~herd}}=0.50=50%</math>

'''Measures of disease association'''

Measures of association are important tools to help assess the "cause" of disease. (Actually they measure associations; we will discuss the difference in interpretation below). They measure the relationship between hypothesised risk factors and the amount of disease. We will investigate one measure of association here.

''Odds ratio''

Consider the same 100 cow herd that we discussed in the prevalence section (above). Remember there were 50 ''Brucella'' positive cattle in that herd. Assume we wished to determine whether age was a factor in being ''Brucella'' positive and we could tell whether cattle were old or young. We could then calculate an odds ratio that would allow us to answer the question of whether age was associated with the prevalence of ''Brucella''.

Assume 50 cattle were young and that 15 of these young cattle were ''Brucella'' positive. This means that 50 cattle are also old (and 35 of these old cattle are ''Brucella'' positive).

This allows us to establish a contingency table of the data.}}

{| class="wikitable"|center
|-
! !! '''Old''' !! '''Young''' !!
|-
| Brucella positive || 35 || 15 || '''50'''
|-
| Brucella negative || 15 || 35 || '''50'''
|-
| || '''50''' || '''50''' || '''100'''
|}

{{hlbox|We can then calculate the odds of ''Brucella'' in both old and young cattle.

<math>\mathrm{odds~}\mathrm{of~}\mathit{Brucella~}\mathrm{~in~young~}\mathrm{cattle}=\frac{\mathrm{number~}\mathrm{of~}\mathrm{cases}}{\mathrm{number~}\mathrm{of~}\mathrm{non~}\mathrm{cases}}=\frac{15}{35}=0.43</math>

<math>\mathrm{odds~}\mathrm{of~}\mathit{Brucella~}\mathrm{in~old~}\mathrm{cattle}=\frac{\mathrm{number~}\mathrm{of~}\mathrm{cases}}{\mathrm{number~}\mathrm{of~}\mathrm{non~}\mathrm{cases}}=\frac{35}{15}=2.33</math>

The odds ratio is simply the ratio of these two odds:

<math>\mathrm{odds~}\mathrm{ratio}=\frac{\mathrm{odds~}\mathrm{of~}\mathit{Brucella~}\mathrm{in~old~}\mathrm{cattle}}{\mathrm{odds~}\mathrm{of~}\mathit{Brucella~}\mathrm{young~}\mathrm{cattle}}=\frac{2.33}{0.43}=5.44</math>

This is interpreted as the odds of ''Brucella'' in old cattle relative to young cattle is 5.44. This means that the odds of being ''Brucella'' positive is much higher (more than 5 times higher) in older cattle. Off course, calculating a confidence interval for the odds ratio is important for interpretation. The 95% confidence interval is 2.31 - 12.81 and does not cross 1. This indicates that being older is associated with an increased odds for being Brucella antibody positive. If the odds ratio had been close to 1 (and the confidence interval crossed 1) we would assume that age had no effect on being Brucella positive. See Exercise 14 for an example of this.

'''Interpretation of measures of disease association'''

It is important to note that a significant measure of association (i.e. one where the measure of association is not equal to 1 and the confidence interval does not cross 1) indicates an association between a supposed risk factor and the presence of disease.

It does not prove that the risk factor causes the disease. To demonstrate that the risk factor causes disease would require a number of other causal criteria such as biological plausibility, that the exposure happened before disease developed etc. However, a statistical association combined with knowledge of the system and biological common-sense can go a long way to identifying causes of disease or risk factors for disease.}}

@@ Baris 23: / Baris 23: @@
 Assume 50 cattle were young and that 15 of these young cattle were ''Brucella'' positive. This means that 50 cattle are also old (and 35 of these old cattle are ''Brucella'' positive).
-This allows us to establish a contingency table of the data.}}
+This allows us to establish a contingency table of the data.
 {| class="wikitable"|center
@@ Baris 36: / Baris 36: @@
 |}
-{{hlbox|We can then calculate the odds of ''Brucella'' in both old and young cattle.
+We can then calculate the odds of ''Brucella'' in both old and young cattle.
 <math>\mathrm{odds~}\mathrm{of~}\mathit{Brucella~}\mathrm{~in~young~}\mathrm{cattle}=\frac{\mathrm{number~}\mathrm{of~}\mathrm{cases}}{\mathrm{number~}\mathrm{of~}\mathrm{non~}\mathrm{cases}}=\frac{15}{35}=0.43</math>
@@ Baris 52: / Baris 52: @@
 It is important to note that a significant measure of association (i.e. one where the measure of association is not equal to 1 and the confidence interval does not cross 1) indicates an association between a supposed risk factor and the presence of disease.
-It does not prove that the risk factor causes the disease. To demonstrate that the risk factor causes disease would require a number of other causal criteria such as biological plausibility, that the exposure happened before disease developed etc. However, a statistical association combined with knowledge of the system and biological common-sense can go a long way to identifying causes of disease or risk factors for disease.}}
+It does not prove that the risk factor causes the disease. To demonstrate that the risk factor causes disease would require a number of other causal criteria such as biological plausibility, that the exposure happened before disease developed etc. However, a statistical association combined with knowledge of the system and biological common-sense can go a long way to identifying causes of disease or risk factors for disease.

Data Analysis:Notes Measures of Disease - Revision history

Catriona: /* Notes: Key concepts for measures of disease and association */

Catriona: Created page with "===Notes: Key concepts for measures of disease and association=== This is a refresher on information already learnt during the field epidemiology training course. This will h..."