Health tech algorithms fail to account for chronic conditions, Verge reviewer argues
Victoria Song says the industry’s push for bespoke health insights is premature for patients with complex conditions like polyendocrine metabolic ovarian syndrome.
The health technology industry is increasingly marketing personalised health as the ultimate solution for individualised care, yet current algorithms struggle to accommodate complex chronic conditions, according to an analysis by The Verge. Victoria Song, a senior reviewer at the publication, argues that while companies promise recommendations based on unique health metrics, existing wearables and AI features largely ignore critical variables such as metabolic factors, hormonal treatments, and individual variations in insulin resistance.
Song’s critique centres on polyendocrine metabolic ovarian syndrome (PMOS), formerly known as polycystic ovary syndrome (PCOS). The global medical establishment recently renamed the condition to better reflect its nature as a hormonal and metabolic disorder rather than a purely reproductive one. PMOS affects approximately 170 million women worldwide, or one in eight, and is associated with conditions including insulin resistance, Type 2 diabetes, obesity, and cardiovascular disease. The name change, finalised in 2026 after 14 years of discussion involving over 50 professional medical groups, aims to address historical issues with clinical training and research funding that were exacerbated by the previous focus on ovarian cysts.
Despite the condition’s prevalence, Song notes that health tech algorithms are not equipped to handle its complexity. The condition manifests differently in individuals; for instance, while Song experiences insulin resistance and hirsutism, others may present with ovarian cysts or acne. Standard medical advice often includes weight loss, which can be particularly difficult for PMOS sufferers due to insulin resistance triggering a cycle of excess androgen production and abdominal fat storage. Studies suggest those with the condition may also have lower basal metabolic rates and complications in building lean muscle mass, factors that current fitness trackers do not adjust for in their calorie burn estimates or workout recommendations.
The disconnect is further highlighted by reproductive health features, which often fail to account for hormonal birth control, a common treatment for PMOS. Algorithms used to predict fertile windows or track body temperature changes typically cannot factor in the effects of oral contraceptives. Consequently, patients are forced to manually manage their health data and cobbled together ad hoc solutions, rather than receiving the automated, personalised guidance that industry pitches suggest is now possible.
Song, who has managed PMOS for a decade, describes the current state of personalised health tech as requiring significant "elbow grease." Users must train AI coaches, decide which metrics are relevant, and conduct independent research to interpret data correctly. While she remains cautiously optimistic about future developments, such as algorithmic modes for specific diagnoses, she warns that the industry is prematurely presenting personalised health as a simple, background process. For many patients, the reality remains a manual effort to navigate tools that were not built for their specific biological realities.
