2018/2019 Cohort
The Hopkin’s statistic for 2018/2019 was 0.837. The 2-cluster model was
selected with the highest average silhouette width of 0.27 (Figure 3).
The silhouette plots indicate the possibility of moderate
misclassification of some individuals in Cluster 1
(C181). There were significant differences in underlying
comorbidity between clusters, though there were no major differences in
demographics (Table 2). For the variables included in the PAM algorithm,
those in C181 had significantly higher mean glucose
(minimum mean 157.2mg/dL, SD 82.2) than those in Cluster 2
(C182) (minimum mean 99.7mg/dL, SD 21.4). Additionally,
those in C181 had significantly lower minimum oxygen saturation than
those in C182 (mean 88.5 (SD 6.7) vs. mean 91.6 (SD
3.7)). The C181 group had higher rates of mechanical
ventilation (32.4% vs. 9.0%) and prolonged hospital stay ≥ 8 days
(16.2% vs. 3.9%) than those in C182, as well as a
longer hospital stay (mean 5.3 days (SD 5.7) vs. mean 2.8 days (SD
2.0)).
After adjustment for age group, sex, hospital, continuous CCI, and
influenza vaccination status, those in C181 had 4.9
times the odds of being mechanically ventilated than those in
C182 (95%CI:1.7,14.3; Figure 2), and 4.3 times the odds
of having a prolonged hospital stay (95% CI:1.2,16.4). After
adjustment, those in C181 had a significantly longer
model-adjusted mean hospital length of stay than those in
C182 (mean 2.5 days longer, 95%CI:1.1,3.9). Vaccination
status was not associated with adverse outcomes in the fully adjusted
models.