The real cost of not translating: what health outcomes in CALD communities show

During the Delta wave of the pandemic, Australians born overseas died from COVID-19 at close to four times the rate of people born in Australia (ABS). Over 70% of Victorians who died from COVID-19 during that wave were born overseas (Victorian Department of Health).

Not every part of that disparity is about communication. Housing density, occupation and access to care all played their part. But a measurable share of it is about communication, and the pandemic did something useful for anyone who has ever had to argue for a translation budget: it made the cost of not communicating in-language visible and countable.

This article assembles the Australian evidence. It's written to be forwarded.

What the outcome data shows

The pattern isn't confined to the pandemic. Across screening, chronic disease and hospitalisation, the same gap appears wherever the data can separate people by language or country of birth.

• Cancer screening. After adjusting for age, breast screening participation is 26% lower for women who speak a language other than English at home than for women who speak only English at home (AIHW, BreastScreen Australia monitoring report 2025). Around 290,000 women who speak a language other than English participated, a rate of 40%. Screening is the mechanism that finds cancers while they're small and survivable, so a 26% participation gap shows up later as a survival gradient.
• Cancer outcomes. Around 25% of Victoria's total cancer cases occur in people born in non-English-speaking countries, and people from CALD backgrounds diagnosed with cancer tend to have poorer outcomes and poorer quality of life than non-CALD groups (Cancer Council Victoria).
• Preventable hospitalisations. People born in North Africa and the Middle East, and in Oceania, have higher potentially preventable hospitalisation rates than the Australian-born population. For people born in Syria, Somalia, Sudan and Samoa, the rate is over double the average (Victorian Department of Health, Multicultural Health Action Plan 2023-27). Potentially preventable hospitalisations are, by definition, the ones that primary care and health information were supposed to prevent.
• Chronic disease. Higher rates of dementia, heart disease, stroke, diabetes and kidney disease are found among Australians born in Polynesia, South Asia and the Middle East, and higher chronic disease rates are associated with low English proficiency combined with more than 10 years of settlement in Australia (Victorian Department of Health). The risk concentrates in people who have been here longest without the system reaching them in their language.
• The next generation. Children from a language background other than English are more likely to be developmentally vulnerable than children from English-only backgrounds, 25.3% compared with 20.8% (Victorian Department of Health). Early childhood services communicate with parents. When that communication doesn't land, the effect shows up in the child's data, not the parent's.

None of these gaps is explained by communication alone. All of them are widened by it. Language is the one contributing factor a communications team can actually move this financial year.

The pandemic made the cost countable

COVID-19 produced the clearest natural experiment in Australian history on what happens when official information doesn't reach people in a language they trust.

The mortality figures above are the headline. The finding underneath them is more uncomfortable for anyone who signed off on a translation invoice in 2020 and considered the job done. A University of Sydney survey of 708 people from CALD communities in Sydney, 31% with low English proficiency, asked how hard it was to find easy-to-understand COVID-19 information. Participants rated the translated materials slightly harder to understand than the English ones: 4.36 out of 10 versus 4.13 (Ayre et al., 2022).

Translated material existed. It was funded and published. And the people it was made for found it less usable than the English content it was translated from. Communities responded rationally: the same research programme found people turning to family, social media and overseas sources when local in-language information was missing, delayed or unreadable.

That's the distinction the outcome data keeps pointing at. The cost isn't only in the content you never translated. It's in the content you translated badly, literally rendered, checked by nobody from the community it was meant for, and published into a vacuum of trust.

Why "we translated it" isn't the same as "it worked"

Translation quality in community-facing health material fails in three ascending ways, and only the first one is visible from inside the organisation:

1. The words are wrong. Mistranslations, machine output published unreviewed, the failures that make the news. Rare, embarrassing, and the smallest part of the problem.
2. The words are right but the meaning doesn't carry. Health literacy levels, reading age, terminology that exists in English bureaucratic practice but has no natural equivalent in the target language. The Sydney survey result lives here: technically accurate material that scores worse than English for the people it targets.
3. The meaning carries but the message was never theirs. Content built on assumptions about family structure, authority, stigma or health beliefs that don't transfer. It reads fine. It just doesn't change behaviour.

Fixing level 1 is proofreading. Fixing level 2 requires the content to be fitted to the audience's actual context. Fixing level 3 requires people from the community in the review loop before publication, not in the complaints queue after it. Cheap translation buys you level 1 and calls it done. The outcome data above is what levels 2 and 3 cost when they're skipped.

Putting a budget frame around it

The health economics here don't need heroic assumptions:

• A potentially preventable hospitalisation is, on any state's figures, orders of magnitude more expensive than the information campaign that could have contributed to preventing it. For four countries of birth, those hospitalisations run at more than double the population average (Victorian Department of Health).
• A cancer found through screening is cheaper to treat and more survivable than one found late. The screening participation gap for women who speak a language other than English at home is 26% (AIHW).
• For NFPs and community health services, the same evidence runs the other way: it's the strongest available argument that in-language funding is core service delivery rather than an engagement extra. The communities with the highest preventable hospitalisation rates are the ones settlement and community health services already hold relationships with.
• The logic extends past health. An insurer's hardship provisions, a utility's disconnection notices and a bank's scam warnings all carry duty-of-care weight, and all fail the same three ways health information does.

The question a budget holder should ask isn't "what does translation cost". It's "which of my catchment's languages carry the largest outcome gaps, and what does it cost to keep not reaching them".

Answering that question with data instead of instinct

Answering it requires three numbers most organisations don't have in one place: how many speakers of each language are in the service area, how many of them have low English proficiency, and what in-language material peers already provide.

The first two are ABS Census data, and the platform holds both at LGA and suburb level across 551 LGAs and 4,288 suburbs. The third is what a gap analysis is for: your published in-language resources, matched against your catchment's demographics and your peers' coverage. The output is the sentence a communications manager can put in a briefing: which languages, how many people, what the peers do, what it costs to close.

The evidence in this article is the "why". The census data is the "where". What happens next is a budget decision, and it's better made with both.