Sunday, April 19, 2009

Truth in Diagnosis. A simple guide to important statistics.

Truth in Diagnosis: An evidence based perspective

Essentials made simple.


It is becoming increasingly important for therapists to understand the tests they are using on a daily basis. Terms like sensitivity , specificity and likelyhood ratio are essential terms to understand. Below is, hopefully, a simple guide ot their use and importance.

1. 2x2 contingency tables



1. Sensitivity and Specificity

Sensitivity= True positive rate ( A/ A+C )

Specificity= True false rate ( D/ B+D )


2. SnOUT and SpIN

When sensitivity is high you can use the test to rule things out. When specificity is high you can use the test to rule things in.

Another way of remembering is Sn N OUT (Sensitivity high and test Negative rule OUT), Sp P IN (Specificty high and test Positive Rule IN)


3. Positive and negative predictive values (PPV & NPV )

The positive predictive value is the proportion of subjects with a positive test result who actually have the condition.
( A/ A+B )

The negative predictive value is the proportion of subjects with a negative result who do not have the condition. ( D/ C+D)


“Despite their apparent usefulness, predictive values can be deceptive because they are highly dependent on the prevalence of a condition of interest in the sample. “ ( Fritz 2001)


4. Likelihood ratio

+ Likelihood ratio= Sensitivity / 1- specificity

- Likelihood ratio= (1-sensitivity)/ specificity

Likelihood ration is independent of the prevalence in a population.


+LHR
1-2: minimal changes in odds

2-5: small shifts in probability
5-10: moderate shifts in probability
10+: large shifts in probability



-LHR

0.5-1: minimal changes in probability
0.2-0.5: small shifts in probability
0.1-0.2: moderate shifts in probability

<0.1: color="#000000">5. Pre and post test odds.

Pre test odds is the prevalence in your practice.

Post test odds= Pretest odds x likelihood ration

Or we can use a “Nomogram” to determine the effect of a test on the pretest odds, if we know the likelihood ratio of our test.


Nomogram for calculating post test odds.


Practical example

Clinical Diagnosis of an Anterior Cruciate Ligament Rupture: A Meta-analysis.Anne Benjaminse, PT1 ,Gokeler, PT2 Cees P. van der Schans, PT, PhD3 ,J Orthop Sports Phys Ther • Volume 36 • Number 5 • May 2006



Sn Sp +LR -LR

Anterior drawer test, without anesthesia, whole group.

Pooled 55 92 7.3 0.5

Lachman test, without anesthesia, whole group.

Pooled 85 94 10.2 0.2

Pivot shift, without anesthesia, whole group.

Pooled 24 98 8.5 0.9

Anterior drawer test, with anesthesia, whole group.

Pooled 77 87 5.9 0.4

Lachman test, with anesthesia, whole group.

Pooled 97 93 12.9 0.1

Pivot shift, with anesthesia, whole group.

Pooled 74 99 20.9 0.3


Remember Sensitivity (Sn) and Specificity (Sp) only describe how abnormality, in this case a rupture of the ACL, predicts a test result.(Deeks 2004)

However, for clinical practice it is essential to know how a test result predicts abnormality.

Likelihood ratios provide this information.

According to a prospective investigation from Noyes et al (1980)
44.0% of patients with acute knee injuries (defined as individuals who suffered a traumatic knee injury associated with a rapid onset of swelling [ie, a hemarthrosis] with or without the sensation of a pop) have a completely torn ACL.

In other words, the pretest probability of having an ACL rupture in these patients is 44.0%.

From above, in acute cases the LR+ of the Lachman test is 9.4 (95% CI, 0.4-210.0). From this value the posttest probability
can be calculated, which is 88.1%. See nomograph below.


It can be seen that using Lachman’s test in an acute knee with sweeling can take your pre test odds of 44% and increase them to
88.1%.

If you do the pivot shift test under anesthetic the LR is 20.9, pre test odds are again 44% but post test odds increase to 95%.

From this example you can see the usefulness of knowing the pre test odds and likely hood ratio for your tests in musculoskeletal medicine.


Using the nomogram you can apply the information above. Line A is the pre and post test odds for Lachman's test with out anesthetic.



References:


Bogduk N. Truth in Musculoskeletal medicine Truth in Diagnosis- Validity. Australasian Musculoskeletal Medicine. May1999

Deeks JJ, Altman DG. Diagnostic tests 4: likelihood ratios. BMJ.;329:168-169. 2004


Fritz J , Wainner RS. Examining Diagnostic Tests:
An Evidence-Based Perspective. Physical Therapy . Volume 81 . Number 9 . September 2001

Noyes FR, Bassett RW, Grood ES, Butler DL.Arthroscopy in acute traumatic hemarthrosis of the knee. Incidence of anterior cruciate tears and other injuries. J Bone Joint Surg Am;62:687-695, 757. 1980