Abstract
Hierarchical Cluster Analysis on the distribution of the EQ-5D-5L in different groups of perinatal people
APHA 2022 Annual Meeting and Expo
OBJECTIVE: Pregnant and postpartum persons are particularly susceptible to pandemic-induced anxiety/depression, which can adversely affect maternal and infant health. We sought to evaluate whether unbiased hierarchical cluster analysis could identify distinct factors impacting the Health-Related Quality-of-Life (HRQoL) measures during the perinatal period and to explore the distribution of EQ-5D-5L profiles.
METHODS: Individuals who were pregnant any time since January 2020 (i.e., the beginning of the pandemic) were invited to participate in a national online survey between May and June 2021 (n=3,359, EuroQol grant: 260-2020RA). Variables collected including respondents’ personal experiences with the COVID-19 as well as the experiences or diagnoses of their family members, friends, and people they know in other social circles. HRQoL was measured by the EQ-5D-5L and other HRQoL instruments. We used unbiased hierarchical cluster analysis to define and characterize mutually exclusive groups.
The distributions of the EQ-5D-5L utilities and the EQ-VAS scores in each group were compared using the standard t-test.
RESULTS: Among the 3,359 pregnant and postpartum participants, 14.62% reported they had COVID-19 themselves, 14% reported that their partners had COVID-19, 28% reported that their close family members had COVID-19 and 39% were concerned about being pregnant during the pandemic. The hierarchical cluster analysis classified participants into 3 optimally distinct groups. Group 1 (n=971) consisted of people who were impacted by their partner and/or family’s diagnosis with COVID-19. Group 2 (n=736) were distinguished by race and their social circle’s hospitalizations with COVID-19. This group also showed they had other health conditions including depression, gestational diabetes, high blood pressure and others. Group 3 (n=1652) were impacted by their family’s death due to COVID-19 and/or hospitalized with COVID-19. Although there was no strong evidence of clustering of EQ-5D-5L utility values, we did find that many participants in Group 2 had slightly lower utility values compared to other groups.
CONCLUSIONS: The results suggest that the EQ-5D-5L and EQ-VAS are superior in that they are less likely to generate artefactual clusters since they do not seem to be driven by the clusters. Statistical learning algorithms may allow for improved classification of factors impacting the pregnant and postpartum women during the pandemic. This classification can be replicated and validated in other prospective cohorts.