Description

Dobbs developed a scoring system for predicting the presence of Group A streptococcal throat infection. Points are assigned for clinical findings based on the Bayesian probability and are called "B" scores. The system was found to perform better in predicting positive throat cultures than physician opinion. The author is from the Irish College of General Practitioners in Ireland.


 

Population: The patients are stated to be 4 years of age or older, but it is unclear if there is an upper limit for age.

Parameter

Finding

B Points

time of year

autumn (October to December)

1

 

other

-1

age of patient

< 11 years of age

2

 

>= 11 years of age

-1

duration < 3 days

< 3 days

1

 

>= 3 days

-2

very sore throat

present

1

 

absent

-2

throat too sore to swallow

present

1

 

absent

-3

bad smell

present

2

 

absent

-1

ears sore

present

-3

 

absent

0

cough

present

-5

 

absent

1

fever

present

1

 

absent

-2

muscle aches

present

1

 

absent

-1

flushed

present

1

 

absent

-1

glands (lymphadenopathy)

present

1

 

absent

-2

exudate

present

1

 

absent

-1

mouth red and/or ulcerated

present

1

 

absent

-1

 

total B score =

= SUM(points for all 14 findings) + (points for prevalence in the population)

 

where:

• The threshold temperature for a fever is not given.

• I am confused by the scoring of the population prevalence. The frequency for a positive streptococcal culture in the study population was 35%. On the bottom of page 462 it says that "a score of –2 was added to each total for the population prevalence." My guess is that the population prevalence is the pretest probability and can be converted to an odds which can then be used to generate a B score (see below).

 

Interpretation:

• minimum total score with 35% prevalence: -28

• maximum total score with 35% prevalence: 15

• A total B-score > -3 had a sensitivity of 71% and specificity of 71% for Group A streptococcal throat infection.

 

odds for infection =

= 10^ ((B score) * LOG10(2) / 2)

 

percentage probability of infection =

= (odds for infection) / ((odds) + 1) * 100%

 

For calculating the B score for the prevalence in the population (based on re-arranging the equations in Appendix 2 on page 464):

 

prevalence as decimal fraction =

= (odds) / ((odds) + 1)

 

odds =

= (prevalence) / (1 – (prevalence))

 

B score for prevalence =

= LOG10((prevalence) / (1 – (prevalence))) * 2 / LOG10(2)

 


To read more or access our algorithms and calculators, please log in or register.