A critical look into artificial intelligence and healthcare disparities

Frontiers in Artificial Intelligence, Год журнала: 2025, Номер 8

Опубликована: Март 6, 2025

Artificial intelligence (AI) has permeated many aspects of daily life, including medicine, in recent years. As 2021, 343 AI-enabled medical devices had been approved by the United States Food and Drug Administration, with more development (1). Most notable thus far AI's ability to assist every step radiology workflow: it can determine appropriateness imaging, recommend most appropriate imaging exam, predict wait times or appointment delays, interpret much potential utilizations (2). The World Health Organization proposed that AI tools be integrated into healthcare improve efficiency achieve sustainable health-related (3). reduce costs administrative burdens, waiting for patients receive care, diagnostic abilities patient facilitate data management, expedite discovery (4,5).However, advancement comes unique drawbacks. For example, security privacy are at risk must improved, as may readily unknowingly provide consent covert collection methods (6,7). Use seriously reconsidered if poses a confidentiality, non-negotiable healthcare. With rapidly gather analyze large amounts data, controlling scope its use becomes challenge: these progress collect disclose without direct investigator oversight (5). In addition, healthcare-based research conducted non-clinical settings, rolling out certain clinical settings result non-evidence-based practice (6). clinicians feel tempted tasks beyond their validation, training not adequately represent scenarios encounter (8). fact, studies on have administered (5).That is say should used It does, however, require immense consideration how designed why utilized. Some contended goal developing minimize health disparities make system equitable (1,9). Yet, characteristics this difficult achieve. such, there growing body literature discusses role both closing perpetuating inequalities (10)(11)(12). directly proportional quality sets used, authors addressed concerns regarding bias datasets lack diversity teams ultimately resulting AI-driven care (5,(13)(14)(15). This article draws from existing add ongoing conversation about implications disparities. Specifically, we discuss economic implications, explainability systems, importance compassionate care. Ultimately, while indeed confer benefits system, remains may, instead, backfire.One essential any kind social disparity economics. notorious having highest expenditure globally, costing $3.5 trillion, 17.9% Gross Domestic Product (16). Any measure decrease burden -either US internationally -may attractive. save billions annual (17). greatly streamline workflow, even tasks. An automated alleviate burdens such scheduling patients, estimating times, billing insurance companies (2,17,18). Such workflow optimization cost delivery cutting intermediaries typically handle mundane turn, patients' financial responsibility related reduced.On side, screen diagnose conditions, stratify disease risk, devise treatment plans significantly errors factors associated adverse outcomes (4). Eventually, technology advances, perform procedures, given deemed ethical, safe evidence-based. While seem like simply perk those practicing physician-rich areas, they could become indispensable areas affected shortages professionals (19). Urban rural communities bear brunt inequity, struggling access primary specialty (20). estimated 2030, shortage up 104,900 physicians implementation underserved populations help challenges Furthermore, assistants physician burnout therefore (21).These advantages conferred only proper development, installation maintenance systems. requires investment. One model an glaucoma screening tool Changjiang county China fifteen-year accumulated incremental using was $434,903.20 approximately 2000 (22). arguably worth early detection reduced progression, impractical roll larger populations. institutions wealthy countries easily But what countries? Community hospitals limited government funding? Practices less purchasing power? Even analyses demonstrate saved long run, upfront investment too obstacle (16).Once developed, purchased, installed, another issue.Software updates, advanced computing technologies, ever-increasing cloud storage requirements evolving cybersecurity needs protect information create further barriers widespread application (23). These all-around nuanced than mere implement practice. Inevitably, algorithms higher lower levels sophistication, infrastructures robust, measures stronger weaker. choose will closely tied status. Of course, then leave behind under-resourced communities.Currently, "explainable" play decision making (24). other words -exactly do technologies work? How decisions? questions developers themselves cannot answer; know work, yet nobody fully explain how. "black box" holds important worldwide. Machine learning (ML) component which involves based (25). Detecting correcting biases ethical prerequisite justice AI-and ML-based decision-making words, explainable enables identify correct set-based currently skew (10,13).The discussion additional considerations. Explainable models keep accountable accountability precedes error concern compounded fact who literate likely ask seek (26). Since prepared participate shared making, challenge questionable decisions (27).AI treated support decision-making, one independently. prescription systems developed aid prevent human (28,29). recommendation conflicts judgement. arise trained treat, thereby generating recommendations poorly aligned realities particularly relevant minority historically under-studied (15). Healthcare providers critically assess context experience preferences. Institutions establish clear policies accept reject suggestions maintain care.Justice also transparent foster trust system. Unexplainable, opaque models, hand, exacerbate mistrust already pervades prevalent socially economically marginalized (30). A key underprivileged patient's comfort physicians' personal involvement (31). see unexplainable black box ML -if handled correctly -would certainly concerns. Lack explanation impersonal, alienate vulnerable population widen disparities.Even elucidate box, ever replace physician-patient relationship delivering empathic care? Currently, seems unlikely -one study demonstrated chatbots empathetic sympathetic responses lowered perception authenticity (32). contrast, empathy sympathy expressed did induce negative effect perceived undermine subjective satisfaction but objectively worsen outcomes. some provided sound biomedical diabetes overlooked psychosocial components necessary glycemic control (33). Algorithms A1c goals, calculate medication dosages, send prescriptions optimize However, tailored disproportionately affect greater barriers. Continuing stand-alone case diabetes, significant include afford healthy food, free time follow-up visits literacy understand (34). Now combine slew medications, unemployment ailing family member. Surely, manage countless different ways. There no path. Regardless, imperative -human AI-based -address compassion.Palliative emphasizes relieving suffering optimizing life end-oflife field compassion (35). risks depersonalizing cases lacking when families need most. Death dying often rooted culture, beliefs, spirituality. deeply each Whereas encourage open communication death, others uncomfortable it; whereas value life-prolonging regardless prognosis, so (36). Palliative imposing "one-size-fits-all" aWestern dataset Once again, understudied cultural minorities fall "understanding" -or thereof -of values.Society large, regulators, policy makers, companies, carefully consider incorporating practices business medicine. Regulators raised over regulation well generalizability (37). Another area several stakeholders regulators legal fear exists scenario made conversely accusations negligent AI. Physicians were neither nor agreed assuming AI, believed liable since medicine Each side felt understood "part whole" highlighting Appropriate makers needed ensure promote mitigated informing involved (38).Certain narratives pitted rival skills education physicians, claims day (38). solely setting final being human. Rhetoric continues pit against hinder incorporation Patients benefit replacing avoiding altogether.In discussing disparities, low-and middle-income (LMICs). where resources personnel scarce, workload (39). especially available (40). Disease outbreaks predicted earlier allow mobilization areas. severity failure illnesses malaria, tuberculosis dengue fever LMICs face implementing electronic records limiting factor input high income (HICs) reflect When applying LMICs, updated applied to. Failure reinforce (40).Gaining integration problem future. small interviewing perspectives GP, subjects mixed feelings (41). common amongst participants sharing wanted assurance would obtained prior anonymization used. survey 203 public opinion yielded results, near 50/50 split asked physician's diagnosing conditions (42). same study, majority trusted culturally biased decision. positive outlook towards future 25% respondents believe next 10 years nearly half 50 (42).Similarly, lacks intelligence, rather wisdom -the sense intuition accumulate (43). Can develop time? mimic brain synthesizing decades' making? simple cases, can. complex story -risks intervention weighed complications predicted, all easilydigestible manner. Yet layer nuance added shared-decision introducednow, desires uncertainties level incorporate recommendations. Moreover, remain maker, spend meaningful conversations facilitating (37,44). alike.While increase bridge gaps healthcare, inclusive avoid worsening hinges optimal course action, execute plan appropriately. Particularly communities, critical process building maintaining proved absence pose success delivery. Both alike wish standard interaction. Instead, serve adjunct reducing chance error. Collaboration among

Язык: Английский

---

Attitudes of older patients toward artificial intelligence in decision-making in healthcare

Journal of Medicine Surgery and Public Health, Год журнала: 2025, Номер unknown, С. 100193 - 100193

Опубликована: Апрель 1, 2025

Язык: Английский

---

Leveraging Artificial Intelligence to Improve Health Insurance Access and Address Disparities in Rural Africa

Journal of Medicine Surgery and Public Health, Год журнала: 2024, Номер unknown, С. 100172 - 100172

Опубликована: Дек. 1, 2024

Язык: Английский

---

Leveraging Artificial Intelligence/Machine Learning Models to Identify Potential Palliative Care Beneficiaries: A Systematic Review

Journal of Gerontological Nursing, Год журнала: 2024, Номер 51(1), С. 7 - 14

Опубликована: Дек. 30, 2024

The current review examined the application of artificial intelligence (AI) and machine learning (ML) techniques in palliative care, specifically focusing on models used to identify potential beneficiaries services among individuals with chronic terminal illnesses.

Язык: Английский

---