Chronic diseases account for most healthcare spending and most of the disease burden in Australia. Diabetes, cardiovascular disease, COPD, chronic kidney disease—these conditions affect millions of Australians and consume enormous healthcare resources.

AI is often proposed as part of the solution. Remote monitoring. Risk prediction. Personalised treatment. Automated care coordination. The promises are compelling.

How much of this is real, and how much is marketing?

Where AI Is Actually Helping

Risk stratification. AI systems that analyse clinical data to identify patients at highest risk of complications or hospitalisation are working in some settings. These enable targeted intervention—intensive support for high-risk patients while routine care serves lower-risk populations.

Several Australian health services have implemented risk stratification tools. When combined with care coordination programs, they’ve shown reductions in emergency department presentations and hospital admissions.

The key: risk stratification only helps if it triggers action. Identifying high-risk patients without intervention resources just creates a list.

Retinal screening. I’ve mentioned this before—AI for diabetic retinopathy screening is one of the most validated applications. For diabetes management specifically, this enables screening in primary care settings that lack ophthalmology access.

Medication optimisation. AI decision support for medication management in complex chronic disease patients helps identify drug interactions, dosing issues, and deprescribing opportunities. This is particularly valuable for patients with multiple comorbidities on extensive medication regimens.

Remote monitoring analytics. For patients using home monitoring (blood pressure, blood glucose, pulse oximetry), AI can analyse patterns to identify concerning trends before patients become acutely unwell.

Where AI Is Overpromised

Lifestyle modification. AI-powered apps promise to change patient behaviour—diet, exercise, medication adherence. The evidence for lasting behaviour change is thin. Apps work for motivated patients who would probably succeed anyway. For the harder cases, digital interventions have limited impact.

Virtual care management. Chatbots and virtual assistants for chronic disease support exist. Some patients find them useful for information. But they don’t replace human care relationships, and for complex patients, they’re inadequate.

Predictive accuracy. Vendor claims about predicting hospitalisations or exacerbations are often inflated. Real-world performance is typically lower than published studies suggest. Populations differ. Data quality varies. The clean conditions of research don’t reflect messy clinical reality.

Cost savings. AI for chronic disease is often sold on cost savings. These savings are hard to demonstrate. Yes, preventing hospitalisations saves money—but proving that AI prevention actually happened (rather than spontaneous improvement) is difficult.

Implementation Realities

Having worked on chronic disease AI implementations, some practical observations:

Data quality limits everything. Chronic disease management requires longitudinal data. If your data is fragmented, incomplete, or poorly coded, AI won’t perform as advertised.

Primary care integration is essential. Most chronic disease management happens in primary care. AI that isn’t integrated with GP workflows won’t be used. Australian general practice IT is fragmented, making integration challenging.

Patient engagement varies wildly. Some patients embrace digital tools. Many don’t. Assuming patients will use apps, portals, and monitoring devices doesn’t match reality for much of the chronic disease population—older, less digitally literate, dealing with multiple challenges.

Care coordination is the hard part. AI can identify patients who need attention. Actually providing coordinated care across multiple providers is the difficult work that AI doesn’t solve.

What I’d Recommend

For healthcare organisations considering AI for chronic disease:

Start with risk stratification if you have care coordination capacity. Knowing who needs attention is only valuable if you can provide it. If you have care coordinators, nurse navigators, or integrated care programs, AI risk stratification can help direct their effort.

Implement diabetic retinopathy screening if you serve diabetes populations with limited ophthalmology access. This is straightforward and evidence-based.

Be sceptical of behaviour change claims. Apps and digital therapeutics have value for some patients. They’re not transformational for most chronic disease populations.

Invest in data integration before AI analytics. If your chronic disease data is scattered across systems, fix that first. AI analytics on fragmented data produces poor results.

Measure outcomes that matter. Hospitalisations. ED presentations. Disease control measures. Quality of life. Not app engagement metrics or AI accuracy statistics.

The Primary Care Challenge

Most chronic disease AI discussion assumes a controlled health system context—integrated care, shared records, coordinated providers. Australian primary care is largely independent practices with variable IT, limited care coordination capacity, and fee-for-service incentives that don’t reward chronic disease management time.

AI solutions designed for integrated health systems don’t translate easily to Australian general practice.

This is a structural problem, not an AI problem. Until primary care models change, AI for chronic disease will have limited impact in community settings.

Where AI might help in this context:

• Tools embedded in GP software that don’t require workflow change
• Population health analytics for PHNs managing regional chronic disease programs
• Remote monitoring that feeds into GP clinical systems automatically

What’s unlikely to work:

• Standalone AI platforms that require separate access
• Patient-facing tools that expect high digital literacy
• Complex decision support requiring significant GP time

Looking Forward

I expect gradual progress rather than transformation. AI will help at the margins—better risk identification, improved screening access, enhanced decision support. The fundamental challenges of chronic disease management—fragmented care, behaviour change, social determinants—aren’t AI problems.

Healthcare organisations should implement AI where evidence supports it, but maintain realistic expectations. The chronic disease burden won’t be solved by algorithms. It requires system redesign, care model innovation, and sustained investment in workforce and services.

AI is a tool. A useful one in some applications. But only a tool.

---

Dr. Rebecca Liu is a health informatics specialist and former Chief Clinical Information Officer. She advises healthcare organisations on clinical AI strategy and implementation.