Time to revisit school decile groupings in health research?

Decile determinations are based on Census data on five indicators aggregated at the small area (meshblock) level: household income, household crowding, parental educational qualifications, proportion of parents on income support benefits and occupational skill level of employed parents. Grouping school decile into low, medium and high is problematic because: * it hides potentially important variation in risk factors and health or behavioural outcomes within decile groups * it may mislead policymakers, funders and those designing interventions as to appropriate targeting or resourcing (eg, using resources to target Decile 4 schools to the same extent as Decile 1 schools) * it may result in inadequate adjustment and residual confounding if broad decile groups are used in regression analyses to adjust for confounding (where more detailed categorisation would be expected to provide better adjustment). The same problems may apply to other youth health topics that are strongly patterned by socio-economic status (eg, food insecurity,7 residential mobility,8 oral health9 and private health insurance coverage9), where we might expect exposures or outcomes to be heavily concentrated at one end of the school-decile spectrum rather than being more evenly distributed. [...]we urge researchers to carefully consider the most appropriate school-decile grouping for their particular analysis, rather than using low, medium and high by default.