### Description

Kharbanda et al developed 2 decision rules to identify a child with a low risk of appendicitis. A child at low risk can be managed conservatively with careful observation and does not need a CT scan. The authors are from Children's Hospital Boston.

Patient selection: 3 to 18 years of age

Parameters for first rule (derived by logistic regression):

(1) nausea

(2) history of focal right lower quadrant (RLQ) pain

(3) migration of pain

(4) difficulty walking

(5) rebound tenderness and/or pain on percussion

(6) absolute neutrophil count

 Parameter Finding Points nausea absent 0 present 2 history of focal RLQ pain absent 0 present 2 migration of pain absent 0 present 1 difficulty walking absent 0 present 1 rebound tenderness or pain on percussion absent 0 present 2 absolute neutrophil count <= 6,750 per µL 0 > 6,750 per µL 6

total score for first rule =

= SUM(points for all 6 parameters)

Parameters for second rule (derived by recursive partitioning):

(1) nausea

(2) tenderness maximal in RLQ

(3) absolute neutrophil count

 Absolute Neutrophil Count Nausea Maximal Tenderness RLQ Percent Appendicitis > 6,750 per µL NA NA 53% <= 6,750 per µL no NA 0% <= 6,750 per µL yes no 0% <= 6,750 per µL yes yes 11%

 Absolute Neutrophil Count Nausea Maximal Tenderness RLQ Percent Appendicitis > 6,750 per µL NA NA 44% <= 6,750 per µL no NA 0% <= 6,750 per µL yes no 9% <= 6,750 per µL yes yes 13%

Interpretation:

• minimum score to the first rule: 0

• maximum score to first rule: 14

• A child was considered low risk by the first rule if the score was <= 5.

Performance:

• The first model has a sensitivity of 99% in the derivation set and 96% in the validation set. The specificity was not stated but is 30% in the ROC curve in Figure 2. The NPV was 96% and negative likelihood ratio 0.102.

• The second model has a sensitivity of 100% in the derivation set and 98% in the validation set. The NPV was 97.5% and negative likelihood ratio 0.058.