Categorization

They developed a tool which stratifies patients into 3 groups: high, intermediate, and low risk.

Overall 8% of the sample (n=486) proved to have an ectopic pregnancy.

In the high risk group  12 of 44 (27%) had an ectopic

In the moderate risk group 26 of 350 (7%) had an ectopic

In the low risk group 1 of 92  (1%) had an ectopic
 

Discrimination: What was the sensitivity and specificity of the rule for detecting or excluding an ectopic?

For differentiating high risk from low or intermediate risk patients the rule derived includes two criteria: 
1) presence of abdominal peritoneal signs 
2) unequivocal cervical motion tenderness. 
They report a sensitivity of 31% (17-48) and specificity of 93% (90-95).

In English, if they have either of these findings they have a high probability of ectopic (good specificity, specificity=spin="specificity, rulesin.

However, absence of both of these signs does not exclude an ectopic (poor sensitivity, sensitivity=snout=sensitivity rulesout).

For differentiating low from moderate risk groups the rule requires 4 additional tests 
1) fetal heart tones 
2) tissue extruding from an open cervical os
3) any abdominal pain other than suprapubic mid line menstrual like cramping 
4) any cervical motion, uterine or adnexal tenderness.

Thus if fetal heart tones are present, the pain is within the confines described, cmt is absent, and no adnexal tenderness present,  the patient falls into the low risk group.

For differentiating low from moderate risk patients (after excluding the high risk group) the derived decision tool was 96% (81-100) sensitive and it was 22% (18-26) specific.  Not bad, but not perfect, for determining low risk patients.

What were the likelihood ratio's

For the entire sample:
The likelihood ratio for the presence of a high risk criteria was 4.3
The likelihood ratio for absence of a high risk criteria (patients falling in the moderate or low risk group combined)  was 0.74

Excluding patients who fell in the high risk group
The likelihood ratio for the presence of any moderate criteria (LR+)was 1.2
The likelihood ratio for the absence of any moderate risk group criteria (low risk group. LR-) was 0.17

What do these LR's mean?

Calibration: We have no calibration data. 

Precision:
Confidence intervals for sensitivity, specificity's and predictive values are available in the article.  However, CI's on the LR's are not presented.  For low risk vs moderate risk patients we calculate CI's as
                       LR+   1.2 (95% CI 1.1 - 1.4)
                        LR-    0.17 (95% CI   0.02 - 1.2)

See an example of the contingency table used to calculate these values (MORE)
 

Did the rule have a beneficial effect on patient care? This is a rule derivation study.  The effect, or potential effect, on care was not evaluated