Health star rating system's design flaws persist despite algorithm tweaks: Australian study

 

By Jag Jivan 

A comprehensive new study from Deakin University in Australia has delivered a sobering verdict on the Health Star Rating (HSR) system: after more than a decade of operation and multiple algorithm "improvements," the voluntary front-of-package labelling scheme continues to classify a significant proportion of ultra-processed foods and discretionary foods as healthy.

The research, published in Public Health Nutrition, analysed 4,720 food and beverage products displaying HSR on the Australian market between November 2020 and June 2023 – a period spanning the implementation of updated algorithm criteria intended to bring the scheme more in line with dietary guidelines.

The findings reveal that 73.7% of ultra-processed foods (UPF) and 58.2% of discretionary foods received a rating of 2.5 stars or higher – considered a "healthy pass" score. The median HSR for UPFs was 3.5 out of 5 stars, unchanged from studies conducted before the algorithm updates.

Algorithm 'Tweaks' Fail to Address Core Problem

The updated algorithm, implemented over a two-year period from November 2020, introduced stricter criteria for sugar content, automatic high ratings for minimally processed fruits and vegetables, and other technical adjustments. Yet the study found that the proportion of UPFs displaying "healthy" HSR scores actually increased during the implementation period – from 60.2% in 2020 to 78.5% in 2023.

For discretionary foods, the proportion rose from 47.0% to 62.5% over the same period.

The researchers found that agreement between HSR and the Nova food processing classification system was "none to slight" (kappa = 0.09), while agreement with the Australian Dietary Guidelines was only "fair" (kappa = 0.38).

Dr. Arun Gupta, Founder of the Breastfeeding Promotion Network of India (BPNI), Convenor of Nutrition Advocacy for Public Interest (NAPi), and former Member of the Prime Minister's Council on India's Nutrition Challenges, offered a sharp critique of the findings:

"This new study from Australia confirms what many of us have been saying for years: the Health Star Rating system has a fundamental design problem that no amount of algorithmic tweaking can fix. Despite claims that the HSR algorithm has improved over the past 12 years, it is still enabling many ultra-processed foods to be marketed with at least 2.5 stars – in fact, this study found the median number of stars on UPFs was 3.5 out of 5."

Dr. Gupta emphasised two key takeaways from the research:

"The paper's relevance is two-fold. First, it shows that the core problem continues to be with the HSR's design – and mandating its implementation will not solve this design problem. Second, despite claims that we should trust algorithm tweaks to improve the system, we now have clear evidence that such tweaks are not improving its performance."

How Manufacturers Game the System

The study identified specific strategies manufacturers use to obtain higher ratings. Analysis of carbonated soft drinks – products that are normally major sources of added sugars – found that 77.1% of those receiving "healthy" HSR scores contained non-nutritive sweeteners (NNS), allowing them to bypass stricter sugar penalties in the updated algorithm.

Similarly, "nutritional drinks and other beverages" – including protein shakes, meal replacements, and weight loss shakes – achieved a median HSR of 5.0 stars, despite being classified as both discretionary foods and UPFs. These products boost scores by increasing protein and fibre content while using NNS to reduce sugar.

The researchers note that most such "special-purpose foods" are not even permitted to display HSR under Australian regulations, highlighting a significant regulatory blind spot.

The Reductionist Trap

The study's authors argue that the fundamental problem lies in the HSR's nutrient-based approach, which assesses foods based on isolated "risk nutrients" (saturated fat, sodium, sugar) and "beneficial components" (fibre, protein, fruit and vegetable content). This reductionist framework cannot account for the degree or purpose of food processing – a limitation shared with Europe's Nutri-Score system.

The researchers cite evidence that 71.5% of Nutri-Score A products in European studies were UPFs, suggesting this is not merely an Australian anomaly but a global pattern.

Only 24.2% of new or updated products during the study period displayed an HSR, indicating that manufacturers selectively apply the label to products likely to receive favourable scores. Most UPFs and discretionary foods are packaged, making HSR easier to display on these products than on minimally processed foods – creating what researchers call a "health halo" effect.

Dr. Gupta warned about the broader implications: "This is not just an Australian issue. Countries around the world considering front-of-package labelling systems must learn from this evidence. A nutrient-focused approach that ignores food processing will inevitably be exploited by industry. Consumers are being misled into thinking UPFs are healthy choices, and that is a profound failure of public health policy."

Alternative Approaches Exist

The study notes that an integrated FOPL model – one that first considers the degree and purpose of processing, then profiles risk and beneficial nutrients – has been recommended by some researchers. A novel system incorporating these criteria was recently modelled by Dickie and colleagues and was shown to classify a higher proportion of food categories in line with healthy dietary patterns.

As countries worldwide implement front-of-package labelling to combat diet-related non-communicable diseases, the Australian experience with HSR offers a cautionary tale. After more than a decade, multiple algorithm updates, and considerable resources allocated to "improving" the system, ultra-processed foods continue to receive healthy ratings at nearly the same rate as before.

The study's authors conclude that "further changes are required" – but also acknowledge the possibility that "it may not be possible to adjust existing criteria to adequately account for UPF and discretionary foods" given the HSR's reductionist foundation.

For Dr. Gupta, the message is clear: "We need to stop tinkering around the edges and fundamentally rethink how we assess the healthiness of foods. Processing matters. Industry has shown it can game any system focused solely on nutrients. The science has moved beyond reductionism – policy must follow."