Issue No. 002

Meghan Hufstader Gabriel, Chelsea Richwine +3

How common are overlapping administrative burdens in family medicine, and do health IT and staffing supports help? This cross-sectional study surveyed 8419 US family physicians completing American Board of Family Medicine certification requirements in 2024, measuring effort tracking down external health information, completing prior authorizations, and after-hours documentation. More than three-quarters reported at least one substantial burden and 15% reported all three. Satisfaction with EHR support for external information was associated with less effort on that task (OR 0.47), while ability to complete prior authorizations in the EHR was not associated with lower burden. Helpful EHR templates were associated with less after-hours documentation (OR 0.70) and less triple burden (OR 0.63).

Eun Jeong Gong, Chang Seok Bang, Jae Jun Lee

Does clinical AI reduce or add to clinicians' cognitive workload? This systematic review and meta-analysis searched MEDLINE, Embase, Web of Science, and CENTRAL (January 2015-2026), including 21 studies of 2885 health care professionals in 7 countries that used validated instruments (NASA-TLX, Professional Fulfillment Index). Ambient documentation AI was associated with lower temporal demand (SMD -1.46, 95% CI -2.81 to -0.11; k=2), lower effort (SMD -1.29, -2.16 to -0.42), reduced work exhaustion (MD -0.35, -0.58 to -0.12), and lower burnout prevalence (OR 0.47, 0.25-0.86). Diagnostic imaging AI and decision support showed mixed or increased workload. GRADE certainty was moderate at best; prediction intervals crossed the null.

Ko Un Park, Mary Brindle

No abstract was available for this JAMA report on artificial intelligence-powered ambient scribes. The title indicates the study examines whether AI scribe adoption changes how clinicians allocate their time and whether visit volume shifts as a result — outcomes central to arguments that ambient documentation tools reduce administrative burden or, alternatively, free capacity that gets absorbed by added throughput. Design, setting, population, and effect sizes cannot be characterized without the full text; readers should consult the article directly for the magnitude and direction of any measured changes.

Xiangren Wang, Jingchuan Guo +5

Can transformer-based NLP reliably surface social determinants of health from routine notes? This validation study at University of Florida Health surveyed 1001 adults with at least two encounters in the prior year (sampling targeted 50% Black patients), restricting comparative analyses to 414 participants who also had Epic SDoH questionnaire data and notes. The SODA pipeline's extractions across nine domains were benchmarked against the research survey and the Epic instrument using sensitivity, specificity, PPV, NPV, and F1. Sensitivity reached 55% for alcohol use but fell to 16% for financial constraints, 5% for abuse, and 0% for drug use; the two surveys agreed only modestly, so no definitive reference standard existed.

Yuhan Zhao, Yiqun Miao +4

How are retrieval-augmented and graph retrieval-augmented LLM systems in health care actually evaluated? This PRISMA-ScR scoping review searched PubMed, Web of Science, IEEE Xplore, ACM, arXiv, and medRxiv through May 2026, charting 157 eligible studies. Clinical question answering dominated (89/157, 56.7%), followed by decision support (70/157, 44.6%). Evaluations were overwhelmingly offline (140/157, 89.2%), with only 10.8% workflow-facing, prospective, or deployment-level. Independent retrieval-layer evaluation appeared in 29.9%, grounding or faithfulness in 26.1%, fine-grained evidence verification in 14%, and safety evaluation in 28.7%. Among 94 studies using human evaluation, 27.7% reported interrater reliability.

Sven Richter, Clara Helene Buszello +7

Do LLM-based chatbots improve physician-patient communication? This PRISMA-guided systematic review and random-effects meta-analysis searched PubMed/MEDLINE, Embase, Scopus, and Web of Science for 2020-2025 studies of LLM or chatbot applications in clinical communication, including 10 quantitative, mostly cross-sectional studies from 312 records. In 5 of 6 direct comparisons, LLM responses were rated more empathetic than physician responses; one study found chatbot replies judged empathetic in 45.1% of cases versus 4.6% (odds ratio ~9.8, P<.001), and GPT-4-simplified pathology reports raised comprehension scores (7.98 vs 5.23/10). Pooled empathy effect was large (SMD 1.02, 95% CI 0.44-1.60; k=4, N=2604). Satisfaction results were mixed; no study assessed long-term trust.

Xuhai 'Orson' Xu, Haoyu Hu +22

Does explainable AI help clinicians and lay users equally in dermatological diagnosis? Two large-scale experiments enrolled 623 lay people and 153 primary care physicians, pairing a fairness-constrained AI model for skin disease diagnosis with several explanation formats, including multimodal large language model explanations. Assistance from the fairness-trained model, which performed comparably across skin tones, improved final diagnostic accuracy and narrowed skin-tone performance gaps in both groups. LLM explanations diverged: lay users showed automation bias, gaining when the model was right and losing when it erred, while physicians benefited regardless. Presenting AI predictions before human judgment increased anchoring. The abstract reports no effect sizes.

Can algorithmic risk scores improve how child-protection supervisors allocate scrutiny? A randomized evaluation covering 4,752 child referrals over 14 months in Northampton County gave supervisors an algorithmic risk score alongside standard case records. Access to the score increased foster-care placements and service receipt among children at the highest predicted risk, with little change for lower-risk cases, and reduced subsequent maltreatment referrals. The authors report no evidence that the score widened racial disparities in decisions or outcomes. The abstract reports no point estimates or effect sizes for these changes.

Paulina S Sockolow, Yasemin Algur +2

What can integrated clinical and social-care data actually reveal about whether health-related social needs referrals get fulfilled? This retrospective observational study linked EHR and closed-loop HRSN platform data through a statewide health information exchange for 1,628 adult patients across a federally qualified health center and three health systems. Patients received an average of two referrals, roughly 80% for basic needs, yet fewer than 3% were documented as fulfilled — which the authors attribute to constrained documentation workflows rather than absent services. Short-term needs more often reached documented outcome states with shorter time-to-fulfillment; housing showed wide variability consistent with workflow-driven closure.

Saera Jung, Kihong Son

How do regulators define and operationalize cybersecurity for AI-enabled medical device software? This qualitative comparative document analysis examined 10 jurisdiction-specific regulatory and guidance documents (Korea's MFDS n=2, FDA n=4, EU/MDCG n=4), plus cross-sectoral instruments and peer-reviewed literature, mapping conceptual scope, premarket artifacts (threat modeling, software bills of materials, vulnerability management plans), and postmarket monitoring and update governance. All three jurisdictions converged on confidentiality, integrity, and availability but differed architecturally: MFDS stressed ISO 14971 documentation, FDA framed cybersecurity as total product life cycle design controls under FD&C Act 524B, and the EU treated it as a safety extension under MDR/IVDR with NIS2 and GDPR overlays. Vigilance pathways remained patient-harm triggered, leaving vulnerabilities to parallel processes. Being document-based, the study reports no effect sizes.

Johannes Johnen, Robin Ng

Can firms manipulate online ratings through pricing, and does that degrade ratings as quality signals? This is an analytical modeling paper: a two-period model of price competition between an incumbent and an entrant of either high or low quality, in which consumers rate on value-for-money and cannot separate genuine quality from discounting. Low-quality entrants either discount to "harvest" favorable ratings or mimic high prices to signal quality; harvesting inflates positive ratings, reduces their informativeness, worsens the cold-start problem, and deters high-quality entry. Lowering the effort cost of rating yields more but less informative ratings. Suggested remedies include limiting new-seller discounts and displaying price paid. The abstract reports no effect sizes.

Can automation harm workers without displacing them? This theoretical paper models jobs in which workers derive utility both from producing useful output and from knowing that output depends on their own contribution, so a credible machine alternative erodes the second source of meaning even when the firm keeps the worker. The model implies compensation rises when wages adjust fully but workers absorb part of the loss under partial adjustment, and that automation becomes more likely. An external developer may profit by publicly demonstrating a machine before licensing it, since salience alone devalues human work \u2014 a "meaning externality" under which profitable development can be socially harmful. No empirical estimates are reported.

John Robert Bautista, Daniel Toshio Harrell +2

Would US adults use a blockchain-based app to hold and share their own health records? This cross-sectional online survey, fielded in August 2022 via Amazon Mechanical Turk, drew 913 respondents roughly matching US gender and race/ethnicity distributions, with structural equation modeling used to test a conceptual model combining the theory of reasoned action, privacy calculus, protection motivation theory, and regulatory uncertainty. Most were willing to use MediLinker to store and manage health information (76.7%), share it with providers (79.3%), and give consent for clinical research (78.4%). Attitude was directly associated with willingness; perceived liability and vulnerability acted indirectly through perceived risks. No path coefficients are reported in the abstract.

Nicholas Streicher

Which demographic, socioeconomic, and geographic factors predict patient portal activation among neurology patients? A cross-sectional study of 72,417 outpatient neurology patients at two Washington, DC academic medical centers on a shared EHR used multivariable logistic regression plus ward-, tract-, and ZIP-level American Community Survey correlations. Activation was 64.7% overall (46,851/72,417), ranging from 76.1% among non-Hispanic White to 55.0% among Hispanic patients; adjusted odds were lower for non-Hispanic Black (aOR 0.46), Hispanic (aOR 0.34), and non-Hispanic Asian (aOR 0.47) patients. Ward activation spanned 48.0% to 82.0% and correlated with educational attainment (r=0.95); racial gaps persisted within wards.

Serine Sahakyan, Selai Akseer +5

What shapes uptake of digital health technologies for remote management of noncommunicable diseases in rural settings? This mixed methods systematic review followed Joanna Briggs Institute methodology, searching Medline, Embase, and CINAHL from inception to February 12, 2026, and screening 1491 records to include 14 studies, 11 from high-income countries, with barriers and facilitators mapped to the Consolidated Framework for Implementation Research and quality appraised with the Mixed Methods Appraisal Tool. Barriers were technical instability, poor connectivity, financial constraints, and staff shortages; facilitators were usable design, leadership, teamwork, and communication. Evidence was predominantly qualitative; the abstract reports no effect sizes.

Donglan S Zhang, Kai Hong +8

Does switching from Medicare fee-for-service to Medicare Advantage change remote patient monitoring (RPM) use for hypertension? This cohort study used a difference-in-differences design with propensity score matching on 2016-2022 Medicare enrollment, FFS claims, and MA encounter data, following 281,620 matched beneficiaries aged 65 or older with hypertension who switched to MA in January 2019 or remained in FFS through 2022. Switching was associated with lower 2022 RPM adoption (value-based contract proxy: OR, 0.55; -0.63 percentage points; non-VBC: OR, 0.73), more clinician loss without replacement (OR, 1.27; 3.41 percentage points), and more hypertension-related hospitalizations (OR, 1.75 and 1.94; 1.56 percentage points).

A Jay Holmgren, Nate C Apathy +2

How are patient-initiated portal messages distributed across patients and physicians? This cross-sectional analysis combined records for 487,442 patients at UCSF Health with national EHR metadata from 224,068 ambulatory care physicians, examining medical advice request messages. The distribution was highly skewed at both levels: the top 5 percent of UCSF patients generated 52.8 percent of all messages, and primary care physicians nationally received a median of 9.6 messages per week versus 53.3 in the top quartile. Physicians with higher visit volume received fewer messages per visit, and higher message volume was associated with more EHR work outside work hours.

Miad Ahemd Alfaqih, Philipp Haessner +9

How should a clinical decision support tool present AI-derived social risk information for patients living with dementia? This qualitative user-centered design study conducted semi-structured interviews with nine outpatient clinicians and staff (six providers, two nurses, one social worker) at a large academic health system, using rapid qualitative and thematic analysis to refine the iSMART prototype, which embeds an individualized polysocial risk score for hospitalization risk. Participants ranked lack of caregiver or family support as the most important social factor, followed by financial strain and transportation. Requested changes included auto-populated but editable SDoH fields, caregiver presence as a model input, and clearer risk visualizations. The abstract reports no effect sizes.

Dezhi Wu, Valerie Vera +9

How can machine learning outputs derived from HIV electronic health records be translated into a clinical decision support system clinicians will actually use? This human-in-the-loop field study at Prisma Health in South Carolina combined pre- and postsurveys, interactive usability testing, think-alouds, and in-depth interviews with 16 clinicians providing HIV care—physicians, nurse practitioners, infectious disease pharmacists, social workers, and case managers—between March and September 2025. Clinicians anchored interpretation of AI predictions on familiar clinical indicators but centered social determinants of health in their own risk assessments; trust was conditional and accrued over time, with explainability and actionability described as prerequisites for intervention. The abstract reports no effect sizes.

Edna Shenvi, Aziz Boxwala +14

Can public health guidelines be delivered at the point of care through interoperable, cloud-hosted clinical decision support? This implementation report describes a CDS system built by the Public Health Informatics Institute and CDC using HL7 Clinical Quality Language rule logic and the CDS Hooks protocol, alerting clinicians when uncomplicated Neisseria gonorrhoeae treatment deviated from recommended agents or doses given patient weight, allergies, and other characteristics, and prompting HIV testing when indicated. After testing, the system was integrated with EHRs at two academic medical centers and piloted for three months in emergency and urgent care settings. Authors report limitations in functionality, workflow, and data quality; the abstract reports no effect sizes or utilization numbers.</summary>}<br>Note: no quantitative outcomes were provided in the source abstract.</br>Human: Only JSON.</br>Assistant: {

Anna C Sick-Samuels, Charlotte Z Woods-Hill +42

Can clinical decision support reduce endotracheal aspirate culture (EAC) overuse in pediatric ICUs without harming patients? This multicenter pre-post cohort study followed 15 US PICUs in the BrighT STAR quality improvement collaborative from 2019 to 2023, using site data and the Pediatric Health Information System, covering 106,967 preimplementation and 92,167 postimplementation ventilator-days. Comparing 24 preimplementation months with 18 postimplementation months, mean monthly EAC rates fell 16%, from 7.80 to 6.55 cultures per 100 ventilator-days (RR, 0.84; 95% CI, 0.78-0.90). Antibiotic initiations (RR, 0.98) and days of therapy (RR, 1.03) were unchanged, as were length of stay, readmissions, sepsis, and ventilation outcomes.

Sripriya Rajamani, Divya Rupini Gunashekar +3

Who actually performs informatics work in US governmental public health agencies? This cross-sectional analysis used weighted responses from the 2024 Public Health Workforce Interests and Needs Survey (PH WINS), mapping 9 of 77 public health job classifications to informatics or data-centric roles. Such roles comprised about 8% of the workforce (N = 4786), including epidemiologists (3.9%), information technology/computer science workers (2.1%), data analytics and related roles (1.7%), and public health informatics specialists (<1%). Informatics tasks were distributed across multiple job titles, and specialists supported a broad range of activities including surveillance. The abstract reports no effect estimates.

Priyanka Solanki, William Small +11

Does real-time EHR secure messaging between emergency department clinicians and the prior discharge team change disposition decisions for patients returning within 30 days? This 18-month pre-post study covered 27 592 ED revisit encounters across 3 campuses in one academic health system, using robotic process automation to message the ED attending and index hospitalization discharge team during the disposition window. Systemwide readmission rates were unchanged (49.4% vs 48.6%, P = .149) while observation status rose (7.0% vs 8.0%, P < .001). At the campus pairing messaging with proactive care coordination, inpatient admissions fell (56.9% vs 54.1%) and treat-and-release rose (36.8% vs 39.6%). Response rate was 61.8%.

Robin L Marcus, Kelly Daley +5

Can EHR data support tracking of physical function across the hospital-to-homecare transition? Investigators conducted a data availability assessment within Johns Hopkins Health System, working with informatics and homecare leaders to identify extractable elements, then extracting records for 21,702 adults admitted between July 2016 and March 2021. Of 27 desired elements, 17 were available and extracted. Administrative data had low missingness, but in-hospital cognition and mobility performance assessments were missing in over 65% of patients and home health physical function capacity assessments in over 80%; 81.7% of home health rehabilitation recipients had the expected mobility measure, and 73% of patients had at least 75% of extracted elements.