Gestalt & Go
MDM
Box content
-
Accordion Title
Key is to understand that all signs and symptoms are not equal.
So how do we know how much weight to place on each sign or symptom?
Most attending’s do this based on experience. They use their intuition.
But there is another way…you can build Gestalt
There are some important terms you need to understand in order to build a gestalt
- Sensitivity & Specificity
- Predictive Values (Positive and Negative)
- Pretest & Post Test probabilities
- Pretest & Post Test Odds
- Likelihood ratio’s
There is a hierarchy of statistics in terms of utility.
Sensitivity & Specificity
- SPIN = Specificity Rules IN: A positive test makes the disease likely
- SNOUT = Sensitivity Rules OUT: A negative test makes disease un- likely
Issues with SPIN & SNOUT:
- With sensitivity and specificity, we use a cutoff point to divide the test into (only) two results: positive or negative.
- In real life, diseases present in gradations of severity
- By limiting a test result to “positive” or “negative”, we stand to lose important diagnostic information.
-
Thinking in terms of “Yes” or “No” is a good way to miss the diagnosis.
- Sensitivity and specificity are independent of disease prevalence
- Sensitivity and specificity are dependent on disease severity
- In early disease, it is difficult to differentiate between health and illness (sensitivity/specificity decrease)
- Sensitivity and Specificity increase in severe disease (as the difference between health and disease become more clear).
- So, reported sensitivity/specificity for a disease may not always reflect the sensitivity/specificity for the individual patient (because they may be early in the disease)
- Sensitivity and Specificity are tough to apply to the individual patient
-
Predictive Values
Predictive Values
Positive Predictive Values
- Probability of patient having disease when test is positive
- In other words, it is the percentage of patients with a positive test who actually have the disease
Negative Predictive values
- Probability of patient not having disease when test is negative
- It is the percentage of patients with a negative test who do not have the disease
Problem with Predictive Values
- Disease prevalence has a significant impact on the positive predictive value (PPV) and negative predictive values (NPV).
- Therefore, despite a high sensitivity and specificity of a given test, the PPV will be very low in a disease with a very low prevalence
- PPV & NPV can only be applied to population that was in the study.
- You can not apply these to individual patients, and thus, does not help you build gestalt
Love the Likelyhood
-
Big Picture
Likelihood ratios are the best STAT for building gestalt and making decisions. To use likelihood ratio’s you need to use incorporate pre test probability and post test probability. Below will discuss these in more detail.
-
Before testing
Pre-test Probability
This is the probability of disease prior to testing
- If you don’t know the Pre-Test Probability for a given disease….
- PRE-TEST PROBABILITY = DISEASE PREVALENCE
- Once the history and examination are completed. The pretest probability may remain the same, decrease, or increase
- When experienced clinicians evaluate a patient they intuitively place a pre-test probability on a decision bar
You can think of probability of disease as a ruler. This is the decision bar.
On the left is 0% probability of disease, on the right (10 cm) is 100% probability of disease. Each millimeter is a 1% probability.
Now we will set the rule out threshold which will determine the rule out zone.
Next we will set our Rule In Threshold, which will determine our Rule In Zone
The area between the rule out threshold and the rule in threshold is the testing zone.
Once the zones are in place, the next step is to place a pretest probability on the ruler.
The pre-test probability can be obtained in many ways. Clinical scoring systems (i.e. Wells score) can give you the pre-test probability. However if you don’t know this, just use the disease incidence as the pre-test probability.
-
After Testing
The post test Probability
- This is the Probability of disease after testing is resulted
Or your post test probability could be somewhere in this area (does not cross threshold)
Next I will show you how to adjust your pre-test probability to get a post test probability
-
How to use it
Next I will show you how to adjust your pre-test probability to get a post test probability
Likelihood ratio’s
- Most useful STAT
- Unlike Sensitivity, Specificity, PPV & NPV; Likelihood ratio’s CAN be applied to individual Patient
- Best Representation of Accuracy of Test
- Can alter pre-test to create post test probability
- Signs and Symptoms can also have associated LR ratios
- Can multiply LR’s
- NOT affected by Disease Prevalence or Disease Severity
A likelihood ration is essentially a number telling you how much weight you should give each sign, symptom & test.
-
Math Behind Likelyhood ratio’s
It is the ratio of probability of test result in patient with disease to the probability of same test result in patient without the disease
- LR > 1 means that a positive test result is more likely happen in a patient with the disease (i.e. can suggest that the disease is present)
- LR < 1 means that a positive test result is more likely to happen in a patient without the disease (i.e. can suggest that the disease is not present)
The further away from 1, the stronger the influence on disease probability
There are 2 types of likelihood ratio’s: LR + and LR –
- LR+: This is the change in pre-test probability caused by a + test result. The higher the number, the more likely the disease.
- LR + of 1-5 is NOT helpful
- LR + of 5-10 is moderately useful
- LR+ of > 10 can change decision making
- LR-: This is the change in pre-test probability caused by a – test result
- LR – of 0.5 to 1 is NOT helpful
- LR – of 0.1 to 0.5 is moderately useful
- LR – < 0.1 is very useful
Pre-test probability and LR’s can be used to calculate a POST TEST probability. You need to convert the probabilities to ODDS for calculations with Likelihood ratio’s.
- Change Pre-test Probability into Pre-test Odds
- Multiply Pre-Test Odds x LR which will give you Post Test odds
- Change Post test odds into Post Test probability
NOTE: You can combine LR’s to get a combined post test probability
- Posttest Odds = Pretest Odds × LR1 × LR2 × LR3 … × LRn
- Then turn this cumulative Post TEST odds into a POST TEST probability
-
End Result
If post test probability does not cross either line, don’t do the test
PE Example
-
Background
- 30 year old female just got off an airplane. She just returned from Europe. She presents complaining of chest pain.
- She now has pleuritic chest pain
- She is mildly tachycardic but not hypoxic
- VS are stable
- You have basic lab services (including a D-dimer) but no imaging at your hospital
- She is on birth control
-
Step 1: Set Pre- Test probability
- Wells criteria gives you a score of 3.0
- This is a Pre-test probability of 20% but since she is on OCP use let’s raise this to 30%
-
Step 2: Pick Discard & Accept Thresholds
-
STEP 3: Get Post Test probabilities
We can use the Fagen NOMOGRAM
- This uses PRE TEST PROBABILITIES and the LR’s to give you a POST TEST PROBABILITY
Or we can turn the pre-test probability into pretest odds and calculate the post test probability:
- Pre-test Probability > Pre-test Odds
- Pre-Test Odds x LR = Post Test odds
- Post test odds > Post Test probability
-
The old DDimer test in action