{
  "id": 13487687,
  "title": "Why most health apps are biased even before their first lines of code are written",
  "url": "https://urgent.news/2026/10/10/why-most-health-apps-are-biased-even-before-their-first-lines-of-code",
  "topic": "health",
  "section": "Health & Medicine",
  "published": "2026-10-10T18:00:01.000Z",
  "source": {
    "name": "Medical Xpress",
    "slug": "medical-xpress",
    "url": "https://medicalxpress.com/news/2026-10-health-apps-biased-lines-code.html"
  },
  "original_language": "en",
  "account": "Many of us rely on digital health tools daily, such as period trackers, fitness apps, and online quizzes for mental health assessments. These apps and devices, often powered by artificial intelligence (AI), are constructed based on the underlying data and the perspectives of their creators. However, the datasets feeding these digital tools frequently fail to represent a diverse population. Consequently, the resulting health technologies risk causing significant harm to underrepresented groups. Since the 2010s, we have been aware that such algorithms can lead to inaccurate diagnoses, particularly for individuals with darker skin tones. This issue does not remain static; a biased AI model can influence healthcare delivery, subsequently affecting the development of future algorithmic models and exacerbating health inequities over time.\n\nOur recent research, involving collaborators from ten countries, offers a potential solution. For digital health tools to be inclusive, equitable, and truly beneficial, crucial conversations must commence even before any lines of code are written. These tools are predominantly shaped by Western, Eurocentric approaches that may disregard diverse notions of health and well-being. Additionally, they are typically built on hidden assumptions about what constitutes good health, which bodies are deemed 'normal,' and which forms of knowledge hold value. Technology companies, funders, and research institutions wield considerable power in this process, often determining which issues to prioritize and which knowledge to acknowledge. They frequently overlook the broader social and historical contexts that shape people's health and well-being, inadvertently causing harm. For instance, an algorithm used in U.S. hospitals to identify patients requiring additional care for complex medical conditions only flagged fewer than 18% of the assigned cases as Black, despite the figure being over 46%. This disparity arises because the algorithm used health expenditure as a risk score, yet Black patients generally spend less on healthcare due to lower socioeconomic status and systemic mistrust, which stems from historical racial bias. This demonstrates how marginalized communities are frequently excluded from discussions about new health technologies or are only consulted after most decisions have been made. Common assumptions include that these communities lack digital literacy or will not utilize the technology. Furthermore, there exists a conflict between the priorities of funders and developers (such as technological advancement and profit) and the needs and values of communities. Moreover, questions of data ownership arise, particularly for Indigenous and other marginalized groups globally. Their data may represent a cultural and economic asset tied to identity, sovereignty, and collective rights. Before any coding begins, it is essential to scrutinize Western, Eurocentric assumptions about health and healthcare. By asking these questions, we can integrate community knowledge, often passed down orally and through lived experience, alongside dominant Western-produced information to shape more equitable health technologies. For example, Indigenous communities often view health as multifaceted, encompassing connections between the body, mind, spirit, community, and environment. By incorporating Indigenous understandings of natural environments, climate, and human health into AI models, we can uncover risks and resilience patterns that may otherwise remain undetected. When executed properly, culturally sensitive digital technologies can transcend Western notions of health, as seen in mental health programs for young Indigenous people. These programs foster cultural identity, honor traditional knowledge, improve well-being, and promote resilience. Innovators must also actively engage with community values and ensure that the technology's values align with those of the communities they serve. For instance, assistive technologies designed for Indigenous communities across Canada, Australia, and the U.S. yielded better adoption rates when they emphasized enhancing family and community involvement in care and strengthening connections with healthcare providers. Users preferred tools that promoted interdependence over independence. Crucially, communities must have control over how their data are collected, analyzed, linked, interpreted, shared, and managed—a principle known as data sovereignty. Transforming digital health requires more than merely creating accessible or technically precise technologies. It demands a fundamental shift in how we perceive health, knowledge, power, and value. This transformation is more urgent than ever as digital health technologies, especially AI, become embedded in healthcare systems and everyday health decisions.",
  "summary": "Many of us use digital health tools every day, such as period trackers, fitness apps or online quizzes to check our mental health.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "The Conversation AU",
        "title": "Why most health apps are biased even before their first lines of code are written",
        "url": "https://urgent.news/2026/10/07/why-most-health-apps-are-biased-even-before-their-first-lines-of-code",
        "published": "2026-10-07T18:28:23.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}