The Impact of AI in Modern Healthcare: Transforming Patient Care and Innovation

Introduction

Artificial intelligence (AI) has moved from science fiction into daily clinical practice. Algorithms can now read X‑rays, suggest surgical plans, accelerate drug discovery and even transcribe consultations. Although adoption is uneven, AI promises to make health care more accurate, efficient and accessible. Global market projections reflect this momentum: the healthcare AI market is expected to grow from $11.2 billion in 2023 to $427.5 billion by 2032, a compound annual growth rate of 47.6%, with North America leading early adoptionpmc.ncbi.nlm.nih.gov. At the same time, a 2024 survey of U.S. health systems found that 72 % identified reducing clinician burden as their top reason for deploying AI and 90 % have started using AI in imaging and radiologypmc.ncbi.nlm.nih.gov. Meanwhile, digital therapeutics and AI‑enabled wearables are reaching millions of consumers, offering continuous monitoring and personalized care.

This comprehensive guide explores how AI is transforming modern health care, from diagnostics and surgery to drug discovery and mental health. It synthesizes peer‑reviewed research and market data to reveal practical benefits, real‑world examples, step‑by‑step explanations and the challenges that remain. Where appropriate, it links to related articles on the FrediTech blog for further reading.

{getToc} $title={Table of Contents} $count={Boolean} $expanded={Boolean}

---

How AI Enhances Medical Diagnostics

The rise of AI in radiology

Imaging is one of the most mature applications of AI. Modern algorithms learn from thousands of annotated scans and detect subtle patterns that elude the human eye. In brain imaging, a deep‑learning model distinguished low‑ versus high‑grade gliomas with a remarkable area under the curve (AUC) of 93.2 %, outperforming traditional methods and aiding surgical planningpmc.ncbi.nlm.nih.gov. In breast cancer screening, AI software analyzing more than 22,621 mammograms achieved an AUC of 89.6 %, improving early detection and risk assessment. For chest radiographs, the CheXNeXt model matched radiologists across 10 pathologies, exceeded them in detecting atelectasis (AUC 0.862 vs. 0.808) and read 420 images in 1.5 minutes compared with 240 minutes for human expertspmc.ncbi.nlm.nih.gov.

These improvements are more than statistical. In practice, AI‑augmented radiology shortens the time from scan to diagnosis and helps triage emergencies. During a suspected stroke, for example, AI algorithms can highlight intracranial hemorrhages or large‑vessel occlusions on CT scans within seconds, allowing clinicians to prioritize time‑critical interventions. Many health systems view imaging as low‑hanging fruit for AI deployment: the 2024 Scottsdale Institute survey reported that 90 % of health systems have at least piloted AI for imagingpmc.ncbi.nlm.nih.gov. However, only 19 % of organizations reported high success with AI‑based diagnosispmc.ncbi.nlm.nih.gov, underscoring that workflow integration and validation remain challenges.

Step‑by‑step: AI reading an image

1. Data acquisition – High‑resolution images (X‑rays, CT, MRI or ultrasound) are captured and labeled by radiologists.
2. Model training – Convolutional neural networks (CNNs) or transformer models learn features from thousands of labeled examples, adjusting weights to minimize error.
3. Inference – When a new image arrives, the model processes pixel intensities through successive layers, extracts patterns and outputs probability scores for each condition. For instance, CheXNet outputs probabilities for pneumonia, pleural effusion and 12 other pathologies.
4. Triage and decision support – The AI flags images with high risk for urgent conditions, allowing clinicians to prioritize them. Some systems automatically generate a draft radiology report, which the radiologist reviews and edits.
5. Continuous learning – Feedback from radiologists and new patient outcomes feed back into the algorithm to improve its accuracy over time.

Real‑world example: AI in mammography

In a multicenter study of 22,621 screening mammograms, researchers compared an AI algorithm with radiologist performance and found that the AI achieved an AUC of 89.6 % for detecting malignanciespmc.ncbi.nlm.nih.gov. The algorithm identified subtle micro‑calcifications and architectural distortions that can be early signs of breast cancer, reducing false negatives. When radiologists reviewed cases flagged by the AI, detection rates improved further. This synergy illustrates how AI can serve as a “second reader,” boosting sensitivity without replacing human expertise.

Beyond radiology: predictive analytics and risk stratification

AI’s pattern‑recognition capabilities extend beyond images to electronic health record (EHR) data, allowing predictive analytics. Models can analyze demographics, vital signs, laboratory results and comorbidities to identify patients at high risk of sepsis, heart failure or readmission. According to the same health‑system survey, 67 % of organizations deployed AI for early detection of sepsis and 52 % used it to predict unplanned readmissionspmc.ncbi.nlm.nih.gov. However, only about one‑third of these systems reported strong success. This modest success reflects the “immaturity” of current tools; 77 % of respondents cited lack of AI tool maturity as the top barrier to adoptionpmc.ncbi.nlm.nih.gov.

 Learn more about how digital imaging evolved from analog X‑rays to AI‑enhanced CT and MRI in our post Digital Imaging in Medical Diagnostics.

AI‑Assisted Surgery and Robotics

Precision surgery with robotic platforms

Robotic surgery combines mechanical precision with AI‑driven guidance. Systems like the da Vinci Xi include articulated arms, high‑definition cameras and machine‑learning algorithms that analyze tissue properties and instrument movements. A 2025 review of 25 studies reported that AI‑assisted robotic surgery reduced operative time by 25 %, decreased intraoperative complications by 30 % and improved surgical precision by 40 %pmc.ncbi.nlm.nih.gov. Patients benefit from 15 % shorter recovery times and hospitals see 10 % cost reductions because of fewer complications and shorter stayspmc.ncbi.nlm.nih.gov.

How AI guides surgeons

1. Preoperative planning – AI analyzes patient imaging (CT or MRI) to create 3D models of organs and tumors. Surgeons can simulate the operation on a digital twin to choose optimal incisions and trajectories.
2. Intraoperative navigation – During surgery, real‑time imaging and sensor data are fed into AI algorithms that track instrument position and tissue deformation. The system provides haptic feedback or visual overlays to maintain safe margins.
3. Predictive analytics – Algorithms monitor vital signs and surgical parameters to predict bleeding or other complications. Early warnings prompt surgeons to adjust techniques.
4. Postoperative analysis – AI reviews videos of the procedure to evaluate efficiency and identify skill gaps. Hospitals use these insights for training and quality improvement.

Example: Prostatectomy improvements

Robotic platforms have been particularly transformative in urologic and gynecologic surgeries. In radical prostatectomy, AI‑assisted robotics enable nerve‑sparing techniques that preserve continence and sexual function. Precise dissection guided by real‑time analytics reduces blood loss and hospital stay. The 2025 review estimated that workflow efficiency improved by 20 % and that patients went home roughly one day earlier than with traditional laparoscopypmc.ncbi.nlm.nih.gov.

Limitations and considerations

Despite impressive results, AI‑assisted surgery is expensive and requires specialized training. Data quality is critical; algorithms trained on homogeneous populations may not generalize to diverse patients. Ethical questions arise around autonomy: if an AI recommends a risky maneuver, who is liable? These concerns underscore the need for rigorous validation and regulatory oversight.

AI Accelerating Drug Discovery and Development

Rethinking the pharmaceutical pipeline

Bringing a new drug to market typically takes 10–15 years and costs about $2.8 billion. The failure rate is high because candidate molecules often prove ineffective or unsafe during clinical trials. AI offers tools to streamline every stage of this process, from target identification to clinical trials.

How AI cuts time and cost

Research shows that AI can shorten the drug discovery phase by 1–2 years by predicting drug efficacy, toxicity and optimal molecular structurespmc.ncbi.nlm.nih.gov. Machine‑learning models analyze vast libraries of chemical compounds, identify promising candidates and prioritize those most likely to succeed. This reduces the number of compounds requiring costly laboratory screening and can cut discovery costs by targeting only the most promising molecules. Beyond discovery, AI‑optimized clinical trial designs automate patient recruitment and monitoring; one review found that AI reduced trial duration by 15–30 %. AI‑discovered molecules have also shown 80–90 % success rates in Phase 1 trials, significantly higher than the historical 40–65 % success rates for conventional candidatespmc.ncbi.nlm.nih.gov.

Real‑world examples

• InSilico Medicine: This biotech company used an AI platform to design a drug candidate for idiopathic pulmonary fibrosis in just 18 months, screening billions of molecules to identify a potent inhibitor that moved into preclinical trialspmc.ncbi.nlm.nih.gov.

• Exscientia: In 2023, Exscientia designed a highly selective protein kinase C‑theta inhibitor (EXS4318) in 11 months using generative AI, a task that had eluded large pharmaceutical companiespmc.ncbi.nlm.nih.gov.

• Adaptive trial design: AI models optimize dosing schedules and predict adverse events to shorten trials, as seen in oncology trials where dynamic models adjust therapy based on real‑time tumor responses.

Step‑by‑step: AI‑driven drug discovery

1. Target identification – Genomic and proteomic data are mined to identify disease‑associated genes or proteins. AI helps prioritize targets by predicting their druggability.
2. Molecular design – Generative models such as variational autoencoders and generative adversarial networks (GANs) propose novel molecular structures that fit target binding sites. Predictive models estimate pharmacokinetic and toxicity profiles.
3. Virtual screening – AI rapidly screens millions of compounds, ranking them by predicted binding affinity and safety.
4. Lead optimization – QSAR (quantitative structure–activity relationship) models refine candidates, optimizing potency and minimizing toxicitypmc.ncbi.nlm.nih.gov.
5. Preclinical testing – AI predicts which compounds are most likely to succeed in animal models, guiding experimental priorities.
6. Clinical trial design – Algorithms select eligible patients, predict dropout risk and identify surrogate endpoints; they monitor patient data in real time and adjust dosing strategiespmc.ncbi.nlm.nih.gov.

For more on emerging biomedical innovations, see our article Emerging Medical Innovations: Pioneering the Future of Healthcare.

AI Streamlining Clinical Documentation and Administration

The rise of ambient AI scribes

Administrative burden contributes to clinician burnout. AI‑powered ambient scribes listen to patient‑doctor conversations, extract structured information and generate draft notes for clinician review. A 2024 assessment revealed that about 30 % of physician practices were already using AI scribes and that they cut documentation time by 20–30 %pmc.ncbi.nlm.nih.gov. A quality‑improvement study involving 45 clinicians found that ambient AI scribes reduced documentation time by 2.6 minutes per appointment and decreased after‑hours EHR work by 29.3 %pmc.ncbi.nlm.nih.gov. Allied health professionals reported a 33 % reduction in documentation time. Beyond time savings, AI scribes free clinicians to focus on eye contact and rapport during consultations.

Step‑by‑step: How AI scribes work

1. Audio capture – Microphones record the conversation between clinician and patient (with consent). Noise‑reduction algorithms isolate voices.
2. Speech recognition – Large language models convert speech to text, identify speakers and segment utterances.
3. Information extraction – Natural‑language processing identifies key elements: symptoms, history, assessments and plans. Medical ontologies help map free text to standardized codes (ICD‑10, SNOMED).
4. Draft note generation – The AI composes a note in the clinician’s preferred template, including history, physical exam, assessment and plan. It can pre‑populate orders and follow‑up reminders.
5. Clinician review – The physician reviews, edits and signs the note, ensuring accuracy and legal compliance.

Benefits and caution

AI scribes improve workflow efficiency and job satisfaction. However, the technology remains nascent. Early studies note risks of hallucination, omissions and misinterpretationspmc.ncbi.nlm.nih.gov. Transparency about how notes are generated and robust validation against gold‑standard transcripts are essential. As the Scottsdale survey highlights, lack of tool maturity and regulatory uncertainty are major barriers to broader AI adoptionpmc.ncbi.nlm.nih.gov.

AI in Telemedicine and Remote Monitoring

Continuous monitoring through wearables

Remote patient monitoring pairs wearable sensors with AI algorithms to track vital signs and detect anomalies. Devices such as Apple Watch and KardiaMobile can identify atrial fibrillation with high sensitivity and specificity, while cuff‑less blood‑pressure monitors detect masked or white‑coat hypertension and smart rings identify sleep apneapmc.ncbi.nlm.nih.gov. AI analyzes these streams to flag irregularities and send alerts to clinicians. In telemedicine triage, chatbots and digital assistants screen patient symptoms and prioritize appointments, improving access during physician shortages.

Step‑by‑step: Remote monitoring workflow

Sensor data collection – Wearables measure heart rate, heart rhythm, blood oxygen saturation, respiration and activity levels. Newer devices track blood pressure, glucose and even cardiac output.
Data transmission – Bluetooth or cellular networks transmit data to a secure cloud platform.
AI analysis – Machine‑learning models detect patterns such as arrhythmias or nocturnal desaturation. They compare current readings with baseline trends and population norms to assess risk.
Alerts and interventions – If the AI detects a significant deviation (e.g., potential atrial fibrillation), it sends an alert to the patient and clinician. Some systems automatically schedule telehealth appointments or adjust medication dosing.
Long‑term modeling – Over time, AI builds personalized risk profiles to predict exacerbations of chronic conditions and suggest lifestyle adjustments.

Real‑world impact

In heart rhythm disorders, the combination of AI and wearables enables early detection and reduces hospitalizations. During the COVID‑19 pandemic, remote patient monitoring became standard for chronic disease management. A cross‑sectional survey of U.S. adults found that 36.36 % used healthcare wearables in 2022 (up from 28–30 % in 2019) and that 78.4 % were willing to share data with physicianspmc.ncbi.nlm.nih.gov. Adoption is higher among women and high‑income groups, while cost and digital literacy remain barrierspmc.ncbi.nlm.nih.gov. Importantly, early detection of conditions like atrial fibrillation or sleep apnea empowers patients to seek timely care.

Challenges in telehealth AI

Integrating AI into telemedicine raises privacy and regulatory questions. Data must be encrypted and stored securely. Algorithms must be transparent to avoid biased decision‑making. The integration of AI into telehealth platforms is still limited; a review noted that many tools are in pilot phases and that prospective trials are neededpmc.ncbi.nlm.nih.gov. The digital divide also persists: older adults and low‑income patients may lack access to devices or broadband.

Internal link: To learn more about wearable health devices and adoption trends, read our article Wearable Tech and Health: Transforming Personal Wellness in the Digital Age.

AI in Mental Health and Digital Therapeutics

Explosion of digital therapeutics

Digital mental health tools provide self‑help and clinician‑guided support. According to industry estimates, about 44 million people used a digital therapeutic in 2021; this doubled to 90.2 million in 2022 and is projected to reach 652.4 million by 2025med.uth.edu. Generative AI chatbots such as “Therabot” can deliver cognitive‑behavioral therapy via text or voice. A randomized trial found that Therabot users experienced significantly greater reductions in symptoms of depression and anxiety compared with controlsmed.uth.edu. Digital therapeutics operate 24/7, offering support to individuals who might otherwise wait months for therapy appointments.

Generative AI vs. rule‑based chatbots

Earlier mental health chatbots relied on rigid decision trees. A 2022 review suggested that 96 % of healthcare chatbots used predefined scripts rather than true AIpmc.ncbi.nlm.nih.gov. These systems are predictable but inflexible. Modern large language models (LLMs) like GPT‑4 process free‑text inputs, maintain context across conversations and produce empathetic responses. They can provide psychoeducation, relapse detection, medication guidance and even crisis interventionpmc.ncbi.nlm.nih.gov. However, generative AI also carries risks: in 2023, a generative AI embedded in an eating‑disorder chatbot made harmful statements and was removed. Bias and hallucinations remain concerns.

Responsible implementation

Experts emphasize that digital therapeutics should supplement, not replace, care by licensed mental health professionalsmed.uth.edu. The Therabot trial took place under ideal conditions; real‑world effectiveness is uncertain. AI models are trained on internet data and can absorb societal biases, generating stigmatizing language or inaccurate advice. To mitigate risks, developers must diversify training data, implement real‑time content filters and embed escalation protocols that direct users to emergency services when necessary. Clinical trials with diverse populations and clear evaluation standards are imperative.

If you’re interested in how medical laboratories use microscopes, check out our guide Types of Microscopes Used in Medical Laboratories.

Adoption Trends, Market Growth and Barriers

Health system priorities and success rates

The 2024 Scottsdale Institute survey provides a snapshot of AI adoption in U.S. health systems. Respondents cited their top organizational goals for AI deployment as caregiver burden/satisfaction (72 %), patient safety and quality (56 %) and workflow efficiency (53 %)pmc.ncbi.nlm.nih.gov. Imaging and radiology were the most widely deployed use cases, with 90 % of health systems using AI in at least limited areas. Early detection of sepsis (67 %), ambient clinical documentation (60 %) and risk‑of‑clinical‑deterioration models (56 %) were also common. However, only 19 % of organizations reported high success rates in AI‑powered diagnosis. Clinical documentation enjoyed the highest perceived success (53 %), while revenue‑cycle automation and analytics laggedpmc.ncbi.nlm.nih.gov.

Barriers to adoption

Health systems identified several impediments to AI. The leading barrier was lack of AI tool maturity, cited by 77 % of respondentspmc.ncbi.nlm.nih.gov. Financial concerns came second (47 %), followed by regulatory and compliance uncertainty (40 %). By contrast, clinician adoption and leadership support were less frequently cited barriers (17 % and 7 %, respectively). Survey authors concluded that generative AI products like ambient notes crossed the chasm from early adopters to early majority faster than any prior medical technology, but noted that imaging AI still struggles with workflow integration and cost justification.

Market statistics and adoption by physicians

According to market analyses, 66 % of physicians used AI in 2024—a 78 % increase from 2023—and over 340 AI tools have received FDA clearancedemandsage.com. The global AI healthcare market is projected to reach $110.61 billion by 2030, expanding from $21.66 billion in 2025 (CAGR 38.6 %)demandsage.com. New Jersey leads U.S. states with 48.94 % of hospitals using AI, while New Mexico reported 0 % adoption. On average, the return on investment for AI in healthcare is estimated at $3.20 for every $1 investeddemandsage.com. Though the figures come from industry reports that may overestimate adoption, they demonstrate the growing enthusiasm for AI.

Regulatory landscape

The U.S. Food and Drug Administration maintains a public list of approved AI/ML‑enabled medical devices. A May 2024 update added 191 new devices, bringing the total to 882. Of these, 128 focus on radiology, emphasizing the importance of imaging in AI innovationhealthhq.world. Notably, nearly 80 % of all approved AI devices relate to medical imaging. However, insurance coverage lags: only about 10 AI‑enabled devices are reimbursed by the Centers for Medicare & Medicaid Services, highlighting a gap between regulatory approval and financial viability. Industry leaders securing approvals include Siemens, GE, Philips, Canon, Viz.ai and Aidochealthhq.world.

For a broad overview of laboratory diagnostics—including genetic testing, biopsies and imaging—see Medical Diagnostics: A Comprehensive Guide.

Ethical Considerations, Bias and Equity

Algorithmic bias: a mirror of society

AI systems learn from historical data. If that data reflects inequities in health care, AI may perpetuate or worsen disparities. A Harvard Medical School article explains that biases are inadvertently programmed into AI systems because training datasets often underrepresent minority groups or rely on proxies like cost rather than clinical needlearn.hms.harvard.edu. One widely used risk‑prediction algorithm prioritized healthier white patients over sicker Black patients because it used past health‑care spending (higher for white patients) as a surrogate for need. Such biases can lead to under‑treatment of vulnerable groups.

Addressing bias and inequity

Experts recommend several steps to mitigate bias:

1. Diversify training data – Include comprehensive datasets that reflect demographic diversity, ensuring that AI models learn from varied patient populationslearn.hms.harvard.edu.
2. Continuous monitoring – Regular audits of AI outputs can detect emerging biases and allow corrective action.
3. Interdisciplinary collaboration – Ethicists, sociologists and patient advocates should be involved in AI development to ensure cultural sensitivity.
4. Regulatory standards – Regulatory bodies must establish guidelines for transparency and bias mitigation. As of 2025, frameworks like STANDING Together and the Coalition for Health AI are advocating for shared evaluation networks and common deployment platformspmc.ncbi.nlm.nih.gov.

Data privacy and security

AI relies on large volumes of patient data. Protecting privacy requires robust de‑identification, encryption and compliance with laws like HIPAA. The potential for data breaches and misuse underscores the importance of secure architectures. Patients should give informed consent for data use and have access to explainable AI outputs. In practice, health systems report regulatory uncertainty as a key barrierpmc.ncbi.nlm.nih.gov. Transparent algorithms and accountability frameworks are essential to build trust.

Future Directions and Innovations

Multimodal AI and digital twins

The next frontier involves combining data from multiple sources—imaging, genomics, wearables, social determinants—to build comprehensive digital twins of patients. These virtual replicas will enable clinicians to simulate surgeries, predict disease trajectories and personalize treatments. Early digital‑twin models in cardiology reduced procedural complications by 25 % and improved long‑term outcomes by 15 %pmc.ncbi.nlm.nih.gov. Integration with AI‑driven predictive analytics could yield truly personalized medicine.

Quantum computing and accelerated training

As AI models grow more complex, quantum computing may offer the computational power needed to train models on massive datasets quickly. Researchers are exploring hybrid quantum–classical algorithms for molecular simulation, which could revolutionize drug discovery and imaging reconstruction. These technologies remain experimental but illustrate the rapid pace of innovation.

Human–AI collaboration and education

The greatest potential of AI lies in complementing—not replacing—clinicians. Training programs should equip health‑care professionals with AI literacy so they understand model limitations, validate outputs and use AI to augment clinical reasoning. Many physicians already report feeling more excited than concerned about AI: a 2025 physician survey found that 36 % felt more excited than worried, up from 30 % in 2023demandsage.com.

Policy and reimbursement

For AI to deliver value, payment policies must evolve. Insurance coverage for AI‑enabled diagnostics and therapeutics remains limited. Advocacy groups are urging Congress to create clear pathways for reimbursement of AI toolshealthhq.world. Regulators must also adapt approval frameworks to accommodate adaptive algorithms that learn post‑deployment.

Conclusion

Artificial intelligence is reshaping health care from the radiology suite to the operating theater and the laboratory bench. Deep‑learning models match and sometimes surpass human experts in diagnosing complex conditions, while AI‑assisted robotics cut operative times and improve precision. In drug discovery, AI shortens timelines, reduces costs and boosts success rates. AI‑enabled wearables and digital therapeutics provide continuous monitoring and mental health support. Administrative tools like ambient scribes alleviate clinician burnout. Yet adoption remains uneven, hindered by immature tools, financial barriers and regulatory uncertainty. Algorithmic bias and privacy concerns demand careful oversight.

For AI to realize its promise, health‑care systems must foster multidisciplinary collaboration, invest in robust validation and embrace transparency. Patients and clinicians should view AI not as a replacement but as an ally—a tool that amplifies human expertise while guarding against error. With thoughtful implementation, AI can transform patient care and drive innovation across the health ecosystem.

Frequently Asked Questions (FAQ)

What is the most successful application of AI in health care today?

Imaging is one of the most advanced applications. Deep‑learning models can detect tumors, strokes and other pathologies with high accuracy, sometimes outperforming radiologists. AI also accelerates documentation through ambient scribes and improves surgical precision via robotic platformspmc.ncbi.nlm.nih.gov.

Does AI replace doctors or radiologists?

No. AI is a decision‑support tool. It can triage scans, suggest diagnoses or draft notes, but clinicians remain responsible for interpretation, treatment and patient communication. Studies show that AI works best when combined with human expertisepmc.ncbi.nlm.nih.gov.

Are AI-diagnosed results reliable?

Performance varies by application. In mammography and brain imaging, AI algorithms achieve AUC values above 0.89. However, real‑world success rates depend on training data quality, population diversity and workflow integration. Health systems report only 19 % high success with AI‑powered diagnosispmc.ncbi.nlm.nih.gov.

How does AI reduce drug discovery time?

AI uses machine‑learning models to predict which compounds will bind to disease targets, reducing the number of molecules that need to be synthesized and tested. This shortens the discovery phase by 1–2 years and reduces costs. AI also optimizes clinical trials, cutting their duration by 15–30 %pmc.ncbi.nlm.nih.gov.

What are the risks of AI in mental health?

Generative AI chatbots can provide supportive conversations, but they may produce inaccurate or harmful responses if not carefully designed and monitored. Experts caution that digital therapeutics should enhance—not replace—professional caremed.uth.edu. Safety requires diverse training data, content filters and clinical oversight.

Why isn’t AI widely adopted despite its benefits?

Health systems cite lack of mature tools, high costs and regulatory uncertainty as major barriers. Integrating AI into workflows requires technical expertise and change management. Reimbursement policies are limited, and concerns about data privacy and algorithmic bias persistlearn.hms.harvard.edu.

How can biases in AI be addressed?

Diversifying training data, continuously monitoring outputs and involving ethicists and patient advocates in AI development are key stepslearn.hms.harvard.edu. Regulatory standards and transparent reporting can also help ensure equitable outcomes.