How differing methods of ascribing ethnicity and socio-economic status affect risk estimates for hospitalisation with infectious disease
Authors: Hobbs MR et al.
Summary: This study was conducted within the Growing Up in
New Zealand (GUiNZ) longitudinal birth cohort, which enrolled 6,822
pregnant women in the Auckland, Counties-Manukau and Waikato
District Health Board areas due to deliver in 2009 or 2010. The
GUiNZ cohort includes representative proportions of Māori, Pacific,
and socioeconomically deprived families. This analysis of GUiNZ data
reports rates of hospitalisation for an infectious disease (ID) within the
first 5 years of life, an outcome that is known to be linked to greater
socioeconomic deprivation. Using only cross-sectional data at 4.5 years
of child age, this analysis included 5,602 children and compared different
methods of ethnicity and socioeconomic deprivation, to determine the
most accurate measures. Primary caregivers assigned the children
to either a self-prioritised ethnicity (i.e. the main ethnic group with
which their child identified), a total response ethnicity (i.e. the ethnic
group or groups their child belonged to), or single-combined ethnic
groups (i.e. assigning the child to a single or combination ethnic group
matching their combination of ethnicities). Socioeconomic status was
measured using household income, census-derived and survey-derived
deprivation indices. The easiest category of child ethnicity to analyse
was the self-prioritisation group. Total response was complicated by
mixed ethnicity allowing overlap between groups. The researchers had
to aggregate small groups for the single-combined ethnicity cohort
to maintain statistical power, but this measure offered greater detail.
Whichever method was used to ascribe ethnicity, Māori and Pacific
children were at greater risk of ID hospitalisation, as were children in
the most socioeconomically deprived households. The magnitude of
these effects differed according to whichever methodology was used.
Household income was affected by non-random missing data. The
census-derived deprivation index offered a high level of completeness
with some risk of multicollinearity and concerns regarding the ecological
fallacy (i.e. making assumptions about individuals within a given decile).
The survey-derived index contained an additional 8 questions but
proved acceptable to participants and provided individualised data, so
largely avoided the ecological fallacy concerns affecting the censusderived
deprivation index.
Reference: Epidemiol Infect. 2018;147:e40
Abstract