ROC Curve
The aim of this analysis to identify the optimal TTD cut off which best
predicts and represents likely macrolide resistance. The overall
defervescence pattern in patients with MSMP and MRMP are shown in figure
1. The sensitivity and specificity of using different cut off values for
TTD were plotted on a ROC curve as shown in figure 2. The sensitivity
and specificity of TTD using cut off at 24hrs, 48 hrs and 72 hours were
compared. In order to achieve a best indicatory cut-off to confirm
positive resistance, we aim for a higher specificity with regards to TTD
as we aim to avoid over-treating patients due to the potential side
effects of doxycycline. By setting a cut-off at 48 hours (46-49 hours)
81% specificity and 38% sensitivity was achieved. With the cut off at
72 hours (68-74 hours) a high specificity of 96% was achieved with a
sensitivity of 29%. TTD at hours beyond the 72 cut-off produced an even
higher specificity (98-100%) and lower sensitivity (around 10%).
However, the presence of two outliers with prolonged febrile illness,
possibility due to co-infection and may have skewed the results. Also
the average length of stay was 3 days in MSMP and 4 days in MRMP, hence
cut off beyond 86 hours was not considered reliable or practical.
The area under the ROC curve (AUC) for TTD was 0.59 with a 95%
confidence interval lower bound of 0.413 and upper bound of 0.775. This
supports the time to defervescence indicates a fair discriminative
ability to predict likelihood presence of MRMP. For the ease of clinical
practice, the TTD was rounded off to the nearest 24hours which means the
optimal cut off to use is at 72 hours.
As per current literature and recommendations for children diagnosed
with CAP, beta-lactam antibiotics and a macrolide group
antibiotic,[12] clarithromycin, were used as first
choice on admission. During the course of hospitalization, 15 patients
changed the choice of antibiotic treatment to doxycycline and 10 of
those patients were in the MRMP group. In the 5 patients with MSMP, 1
patient switched to doxycycline due to the development pleural effusion.
Two out of the remaining 3 patients had significant shortness of breath
and required oxygen and the remaining patient had prolonged fever of
more than 1 week prior to admission. All patients with MSMP that
switched to doxycycline were older than 12 years old. Four patients with
MRMP treated with doxycycline were excluded from this subgroup analysis
due to co-infection or other febrile conditions mentioned previously.
Patients in both MSMP and MRMP group with doxycycline use achieved rapid
defervescence compatible with previous publications. The mean TTD after
use of doxycycline use in MSMP was 18 hours (±13 hours) and 15 (±12
hours) in MRMP group. However due to limited sample size, this was not
statistically significant. (P=0.61)
A significant difference was noted between the two groups in the length
of stay. The mean length of stay was 3 days (±2 days) and 4 days (±4
days) in patients with MSMP and MRMP respectively (P=0.004).