'AI' Is Supercharging Our Broken Healthcare System's Worst Tendencies
AI tools are being woven into U.S. healthcare and insurance workflows largely to cut costs, often by automating claim denials and constraining care through opaque algorithms. Commenters argue that because insurers and hospital systems are structurally incentivized to maximize profit, “AI” is mostly amplifying existing problems—bureaucratic barriers, lack of accountability, and widening power imbalances—rather than reducing costs or improving outcomes for patients. Some see potential for AI in clinical support or record-keeping, but contend that without different incentives and regulation, most benefits will accrue to corporations, not patients or frontline providers.
Scope: Insurance Algorithms vs Healthcare AI
- Several commenters note the article is about insurance utilization-review algorithms, not clinical decision-making, but others argue insurance is integral to the “healthcare system,” so the title is fair.
- Some point out the system in question predates modern LLMs and may be simple heuristics marketed as “AI,” reflecting a broader pattern of relabeling existing analytics.
Incentives, Profits, and Claim Denials
- Strong consensus that US insurers use automation primarily to reduce payouts, not improve care, by denying or delaying claims until patients give up or die.
- Lawsuit claims that denials driven by one such tool are overturned on appeal ~90% of the time; commenters infer this is seen internally as acceptable savings because few patients appeal.
- Automation also increases personal unaccountability: decisions can be blamed on “the algorithm,” making it harder to challenge.
Parallels to Electronic Medical Records and “Big Data”
- EMR rollout is cited as a cautionary tale: heavily sold as cost-saving and quality-improving, but in the US mostly increased bureaucracy, vendor power, and hospital consolidation while doing little to lower per‑capita costs.
- Interoperability is described as poor by design (loose standards, customized installs), turning data into costly noise.
- Some in Europe report better EMR use, underscoring that incentives and system design matter more than the tech itself.
Potential Uses and Limited Current Deployment
- Actual AI/LLM deployment seen so far is mostly in documentation and transcription (e.g., auto‑drafting clinical notes, telehealth summaries) to save clinician time.
- There are mentions of promising diagnostic/triage tools and radiology-support startups, but little evidence in the thread of wide clinical use yet.
Risks to Patients and Society
- Many fear AI will “supercharge” existing bad incentives: optimize denial of care, worsen customer service, deepen information asymmetries, and further financialize healthcare.
- Concerns about unsafe AI medical advice vs. possible benefits for people with no access to doctors are debated; one side stresses harm from authoritative-sounding errors, the other argues that, in deprived settings, even imperfect guidance could net-help.
System-Level Debates
- Disagreement over remedies: some argue for public or single‑payer systems; others for a genuinely competitive, transparent market.
- Broad but not universal view that without structural reform and clear liability, AI will amplify current dysfunction rather than fix it.