Social determinants of health among culturally and linguistically diverse people in Australia

People with low English proficiency (speaking English not very well or not at all) or who spoke ‘English only’ at home, were more likely to report long-term health conditions than those with high English proficiency. 

Modelling the odds of selected long-term health conditions from proficiency in spoken English

Logistic regression modelling was used to explore the association between proficiency in spoken English with each long-term health condition. For this CALD variable, the high English proficiency group (those who spoke English very well or well) was used as the reference category to estimate odds ratios in all the models. Full outputs from all regression models are provided in the supplementary tables (Tables S1.1–S1.4).

Overall, the results observed from the set of regression models indicated that age was the strongest factor for the observed associations between the level of English proficiency and the reported long-term health conditions, with the exception of asthma and mental health conditions. Detailed findings from the modelling are presented below.

1. Unadjusted models

The unadjusted logistic regression models showed that the low English proficiency group (i.e. those who did not speak English well or at all) was associated with higher odds of reporting almost all reported long-term health conditions (except asthma) compared with the high English proficiency group. 

2. Adjusting separately for age and social determinants

When the results were adjusted for the effects of age and social determinants of health, age appeared to have the largest effect on reported long-term health conditions, with the exception of asthma and mental health conditions. Controlling for age alone substantially changed the estimated odds ratio for measuring the association between English proficiency and reported long-term health conditions, other than for asthma and mental health conditions, particularly for people with low English proficiency. 

However, the impacts of adjusting for most social determinants of health were much less apparent for all reported long-term health conditions regardless of the level of English proficiency, with the most notable effect generally observed when adjusting separately for labour force status or occupation, followed by income or marital status. For example, the unadjusted odds ratios for reporting one or more long-term health condition(s) was 2.2 higher for both males and females with low English proficiency than those with high English proficiency. When the same outcome was modelled using proficiency in spoken English and age or a single social determinant, these odds ratios dropped to:

• 1.4 for males and 1.5 for females after adjusting only for labour force status
• 1.7 for males and 1.8 for females after adjusting separately for income
• 1.9 for males and 1.8 for females after adjusting only for marital status.

These drops in odds ratios between the unadjusted models and those adjusting for another social factor were also apparent when considering some other reported long-term health conditions including multimorbidity, arthritis, diabetes, or heart disease. 

3. Fully adjusted models

In the fully adjusted models, for both males and females, the odds ratios for the reported long-term health conditions generally remained similar to the age adjusted odds ratios. For example, when unadjusted, the odds of females with low English proficiency reporting diabetes were 3.1 times higher than the odds of females with high English proficiency. This odds ratio dropped to 1.5, when the effects of age were taken into account. This further dropped to 1.1, in the fully adjusted model – that is when the results were adjusted for the effects of education, labour force status, income housing suitability, tenure, citizenship status, remoteness and marital status (Table 1 below). When adjusting for the effects of age and the social determinants of health in the fully adjusted model, compared with those with high English proficiency:

• those who spoke ‘English only’ had higher odds of reporting all analysed long-term health conditions except diabetes, which had lower odds
• those with low English proficiency had similar or higher odds of reporting all long-term health conditions.

Notes

1. Results are from the fully-adjusted model which included proficiency in spoken English language, age, education, income, employment, tenure, housing suitability, remoteness, citizenship, marital status and occupation. Analysis excluded overseas visitors, people who live in non-private dwellings or Migratory, offshore and shipping SA1s, non-classifiable households or Visitor only households. 
2. Analyses included 7,751,459 males and 8,065,784 females aged 15 and over living in Australia in occupied private dwellings on Census Night, who were not overseas visitors and provided a valid response to the 2021 Census questions on proficiency in spoken English language, age, the selected determinants of health, and the long-term health conditions. 
3. ‘Very well or well’ proficiency in spoken English was selected as the reference category, when calculating the odds ratios for the levels of proficiency in spoken English language.
4. OR (95% CI) refers to odds ratio and the 95% confidence interval.
5. All odds ratios are rounded to two decimal places.

Source: AIHW analysis of PLIDA, 2021.