The Preventive Care Landscape in Southeast Asia

Trends and Assessments

---

Preventive Health: An Overview

Preventive health looks to anticipate risk and intervene early, preventing later-stage treatments and complications. The ideal case is to intervene before symptoms emerge.

The framework splits into two parts: information (acquiring, processing, and evaluating sufficient data so a decision can be reached) and action (deciding on the correct course of treatment based on the information available).

Information pipeline:

• Data acquisition — Whether there is sufficient and relevant data about a person's health signals, ranging from physiological metrics (heart rate, glucose levels) to behavioral, environmental, and genomic data. Multimodal, continuous inputs are often necessary for accurate early detection.
• Data quality / continuity — Whether signals are reliable, well-calibrated, and usable for longitudinal modeling. Incomplete, noisy, or fragmented data streams can severely limit predictive accuracy.
• Predictive modelling — The accuracy of prediction or early flagging of risk. Models must not only predict correctly, but do so early enough to allow meaningful intervention. This step often improves through feedback loops from actual outcomes, forming a learning health system.

Action pipeline:

• Actionability — Whether predictions translate into actual decisions or changes in care. This depends not just on accuracy, but on clinical relevance, interpretability, cost-effectiveness, and integration into provider workflows.
• User engagement — Whether people care about and act on predictions. Engagement can be passive (taking a prescribed medication) or active (sustained lifestyle change, continuous data sharing).

Clinical actionability and system integration are the bottlenecks: systems are useless if they can't act.

---

The SEA Preventive Healthcare Landscape

Analysis consolidated across 104 companies.

Across the SEA landscape, three key structural factors impede greater adoption of preventive care:

1. Regulatory bottlenecks leading to poor data pipelines. Clinical-grade risk profiling remains unreachable for most startups — regulatory certification poses a large hurdle. Most platforms capable of clinical-grade predictive modelling serve healthcare providers directly, acting as software aids without anticipating treatment needs at the personal level.

2. Patient-level data is available but not being utilized. There is an easy entry point at data collection, but consolidation and prediction are largely skipped. Companies move directly to engagement — bypassing the hard adoption challenges, but also creating market saturation.

3. Data continuity remains limited at the consumer level. Multi-biomarker data consolidation is used mostly for clinical purposes and is not made available to consumers for self-care and moderation.

---

The Consumer Predictive Modelling Gap

At the final stage of the data pipeline, consumer-oriented innovation falls to zero. This signals a structural absence of tools that:

• Translate data insights into understandable, personal action steps
• Enable behavioral change or intervention without clinical gatekeeping
• Close the loop between data generation and improved outcomes in daily life

Why this gap exists:

1. Incentive misalignment — Most healthcare systems and startups are paid by institutions, not individuals, so they optimise for enterprise use (hospitals, insurers). Consumers often lack the willingness or literacy to pay for health tools they don't understand or trust.

2. Data-action translation is hard — It's relatively easy to collect data (wearables, apps) and run models (via APIs or pre-trained LLMs). The hard part is what to do next: nudging behavior, adapting interventions, and surfacing context-aware insights in a way that is timely, trustworthy, and non-overwhelming.

3. Liability and regulation — Delivering advice or care to consumers requires navigating clinical safety, liability, and compliance (FDA, CE mark). Many teams avoid this final mile because it exposes them to medical device regulation or malpractice concerns.

4. Cognitive overload — Consumers are already overloaded with information. Most apps push more data, but what users need is less noise and more clarity.

---

Preventive vs. Curative Care: The Full Spectrum

Healthcare operates across a spectrum from preventive to curative:

Preventive:

• Lower risks across population — Reduce likelihood of illness via behavior modification or environmental control
• Prediction and early intervention — Identify potential health risks or early signs before symptoms emerge
• Reduce harm in care — Prevent adverse outcomes due to improper or delayed self-care

Curative:

• Diagnosis and stratification — Accurately identify conditions and determine severity or subtype to support treatment decisions
• Intervention and treatment — Deliver targeted therapeutic actions
• Recovery and recurrence prevention — Ensure the patient heals correctly and prevents relapse over time

---

Technology Landscape: Population Disease Risk Mitigation

The intervention technologies map across four risk driver categories:

Genetic / biological (inherited traits and physiological predispositions)

• AI analysis: AI-assisted variant interpretation, polygenic risk scoring

Environmental (physical and built environment, including toxin exposure)

• IoT + hardware: air and water sensors, urban sensing networks
• Urban layout and pollution monitoring

Healthcare access (ability and behavior in seeking and affording health services)

• Telemedicine: AI-driven appointment prediction, risk stratification
• Remote diagnostics, mobile health
• Fintech: insurance models
• Point-of-care (POC) testing

Behavioural (actions and habits that influence health)

• Wearables and digital phenotyping
• Digital health: AI-driven behavioral change, monitoring and recommendations
• Targets: better diet, physical activity, better sleep, medical adherence

All four drivers flow toward the common goal of predictive and early intervention.

---

Drivers Analysis by Category

---

Cognitive Burden vs. Route to Actionability

A key design tension in consumer health: how much effort is required from the user, and how directly does it lead to care?

---

Private vs. Public Primary Care

About 3 in 4 private clinics in Malaysia are operated as solo practices.

Public health clinics are structured with a Director (typically a Family Medicine Specialist) overseeing Medical Officers, Assistant MOs, Nurses, Pharmacists, and Allied Health Professionals. The public sector tends to be stronger in care coordination and comprehensiveness, as shared practices allow interactions across multiple primary care providers.

Private health clinics are typically owned by an individual who also acts as the primary healthcare provider, supported by General Practitioners, Nurses, Pharmacists, and Allied Health Professionals. Private clinics can provide better continuity of care through more personalised service, enabled by smaller practices and different governance structures.

---

Principal Causes of Hospitalisation in Malaysia

Data: MOH 2024 Health Facts (MOH vs. Private sector)

Notable differences: respiratory and digestive diseases are significantly more prevalent in private hospitals, suggesting a different patient mix with more chronic and elective admissions. MOH hospitals handle a disproportionate share of maternity cases.

---

Investment Considerations

On predictive modelling and engagement:

• Without clinical-grade predictive modelling and risk profiling, engagement will always be a challenge — especially for consumer platforms. Lack of effective follow-through limits care delivery effectiveness and pushes companies into highly competitive markets.
• Biomarker-based assessments alone do not add meaningfully to clinical decision-making. Downstream services may complete the value chain but don't necessarily benefit from the upstream complexity.
• Consumers are data generators but aren't equipped to act on it meaningfully. Automating individual-level triage, where possible, represents a major opportunity.

On data strategy:

• Data is plentiful — new entrants should focus on using it, not collecting more. POC tests, wearables, and self-assessment tools already exist in abundance. New entrants need strong distribution and significant resources to enter for marginal gain. (Exception: medical devices requiring proprietary data acquisition for high-level predictive modelling.)
• Gaps remain in data consolidation and quality monitoring — streams remain fragmented. Comprehensive individual-level health profiles for EMRs are still rarely built at scale.

On the role of intermediaries: Platforms must not only aggregate data but own the full journey from data → action → feedback. To move beyond aggregation, they need to:

• Build features that help users act on insights (recommended pathways, automated referrals)
• Provide closed-loop workflows for clinicians and patients (flag → triage → action → follow-up)
• Become active participants in patient journeys, not passive data hosts