Multivariable analysis of alternative clinical prediction rule models 1–4
| Explanatory variable | Model 1* | Model 2† | Model 3‡ | Model 4§ | ||||
| Regression coefficient (95% CI) | Points | Regression coefficient (95% CI) | Points | Regression coefficient (95% CI) | Points | Regression coefficient (95% CI) | Points | |
| Presence of a temperature >38 | 1.14 (–0.68 to 2.97) | 11 | 1.13 (–0.61 to 2.88) | 11 | ||||
| Presence of exudate on tonsils | 0.28 (–0.83 to 1.38) | 3 | 0.06 (–0.98 to 1.09) | 1 | ||||
| Presence of splenomegaly | 1.98 (–1.57 to 5.53) | 20 | 3.50 (–0.19 to 7.18) | 35 | ||||
| Presence of palatal petechiae | 0.21 (–0.87 to 1.30) | 2 | 0.16 (–0.81 to 1.13) | 2 | ||||
| Presence of enlarged/tender posterior cervical lymph nodes | 1.73 (0.70 to 2.76) | 17 | 4.55 (0.42 to 2.68) | 5 | 1.60 (0.66 to 2.55) | 16 | 1.56 (0.69 to 2.44) | 2 |
| Absence of enlarged/tender inguinal lymph nodes | 0.40 (–2.18 to 2.97) | 4 | 0.51 (–2.32 to 3.33) | 5 | ||||
| Presence of enlarged/tender axillary lymph nodes | 0.35 (–2.21 to 2.92) | 4 | 0.49 (–2.17 to 0.15) | 5 | ||||
| Absence of shortness of breath | 1.38 (–0.17 to 2.92) | 14 | 1.18 (–0.18 to 2.53) | 12 | ||||
| Feeling generally unwell | 1.12 (–0.23 to 2.48) | 11 | 0.82 (–0.47 to 2.09) | 8 | ||||
| Clinician impression | 1.92 (0.47 to 3.37) | 19 | 1.67 (0.71 to 2.62) | 2 | ||||
| Presence of exudate on pharynx | 0.69 (–0.51 to 1.89) | 7 | 0.93 (–0.16 to 2.03) | 9 | 0.98 (0.08 to 1.87) | 1 | ||
| Hosmer-Lemeshow goodness of fit | χ2=3.15; p=0.87 | χ2=5.23; p=0.07 | χ2=0.26; p=0.88 | χ2=0.69; p=0.71 | ||||
| AUC (95% CI) | 0.81 (0.75 to 0.85) | 0.73 (0.63 to 0.82) | 0.77 (0.67 to 0.87) | 0.70 (0.62 to 0.79) | ||||
*Explanatory variables had to be of prior clinical importance or be associated with a threshold p value of ≤0.15 in univariate analysis. Score/points assigned to each predictor by multiplying the regression coefficient by 10 and rounding to nearest integer.
†Explanatory variables had to be of prior clinical importance or be associated with a threshold p value of ≤0.15 in univariate analysis, and then a backward selection was used with a p value threshold of 0.05. Score/points assigned to each predictor by rounding the regression coefficient to nearest integer.
‡Explanatory variables had to be of prior clinical importance or be associated with a threshold p value of ≤0.15 in univariate analysis. Clinician impression was not used in this model. Score/points assigned to each predictor by multiplying the regression coefficient by 10 and rounding to nearest integer.
§Explanatory variables had to be of prior clinical importance or be associated with a threshold p value of ≤0.15 in univariate analysis, and then a backward selection was used with a p value threshold of 0.05. Clinician impression was not included in this model. Score/points assigned to each predictor by rounding the regression coefficient to nearest integer.
AUC, area under the receiver operating characteristic curve.