The Digital Medicine Society (DiMe)

The Digital Medicine Society (DiMe) appreciates the opportunity to comment on FDA’s discussion paper, Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback (FDA-2026-N-7874).

DiMe recommends a total product lifecycle approach connecting technical benchmarking, clinical confirmation, intended users, conditions for safe use, and postmarket monitoring. Technical performance alone cannot establish safety and effectiveness. Outcomes also depend on how users interpret and act on outputs, whether safeguards are adequate, and how performance changes over time. Expectations should be proportionate to risk while preserving manufacturer accountability.

DiMe’s recommendations draw on its work in digital health validation, meaningful health measurement, real-world evidence, AI implementation, and governance.

1. Expand risk assessment to account for intended users and conditions of use. Assess the consequences of incorrect outputs alongside error detectability and reversibility, opportunities for intervention, time for review, and foreseeable user reliance. Manufacturers should evaluate intended populations, realistic interactions, and potential failure modes, then communicate conditions for safe use through labeling.

2. Evaluate patient-facing systems through downstream consequences. Assess how users interpret and act on outputs, including whether they reach appropriate care in time. For care navigation, measure under-escalation and over-escalation separately and by clinical acuity. Human-AI team performance should be central when lay users interpret outputs, device functions direct action, or users are unlikely to detect errors. Compare results with realistic alternatives, including decisions made without the device.

3. Connect benchmarking to performance in intended use environments. Complement benchmarking with clinical confirmation and usability testing involving representative users. Sponsors should demonstrate that benchmark results predict performance in clinically representative conditions, justify reference measures based on rigor, and show whether controlled findings generalize to intended environments. Evidence should distinguish technical performance, clinical relevance, and usability.

4. Make postmarket monitoring actionable. Monitoring plans should specify performance signals, thresholds, accountable parties, responses, and reassessment triggers. Monitoring frequency should reflect how quickly performance may change and the consequences of failure. Plans may include re-benchmarking, qualified clinician review of real-world interactions, and patient-generated or other real-world data. Manufacturers should track relevant third-party model changes and assess their impact on the marketed device.

5. Enable shared learning without diffusing accountability. Standardized, current Foundation Model Master Files could document model versions, limitations, evaluation results, and safety-relevant changes. Agentic systems require assessment of authorized actions, external tool access, traceability, and safeguards. FDA-affiliated collaborative communities, including DATAcc by DiMe, can support shared evaluation methods and detection of emerging risks without replacing manufacturers’ responsibility for device safety and effectiveness.

Labeling should communicate conditions for safe use, including limitations, expected human oversight, and escalation pathways, so deploying organizations can determine whether they can meet them. Postmarket monitoring and change management should help establish whether safety and effectiveness are maintained.

DiMe supports a rigorous, risk-proportionate, and adaptable approach that connects benchmarking, clinical confirmation, labeling, monitoring, and shared learning to advance patient safety, public trust, and responsible innovation.

Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

October 19, 2026

Dr. Michelle Tarver
United States Food & Drug Administration (FDA)
Center for Devices and Radiological Health (CDRH)
Office of the Center Director
10903 New Hampshire Avenue
WO66-5431
Silver Spring, MD 20993-0002

Subject: Considerations for the Regulation of Generative AIEnabled Medical Devices: Discussion Paper and Request for
Feedback (FDA-2026-N-7874)
Dear Center Director Dr. Michelle Tarver:

The Digital Medicine Society (DiMe) is a global non-profit dedicated to advancing the
use of digital approaches to redefine healthcare and improve lives. We appreciate the
opportunity to comment on “Considerations for the Regulation of Generative AIEnabled Medical Devices: Discussion Paper and Request for Feedback.”

DiMe’s Thesis for Comment

Generative artificial intelligence (GenAI)-enabled medical devices should be evaluated
across their total product lifecycle with their intended users and in their intended
use environments. Technical performance alone cannot establish safety and
effectiveness when outcomes also depend on whether intended users can correctly
interpret and act on outputs and whether the device's safeguards are adequate.
DiMe's responses to FDA's discussion questions are organized around five
recommendations that apply this principle:

1. Expand risk assessment to account for intended users, conditions of use, and
downstream consequences (Questions 1-3 and 5).
2. Evaluate patient-facing systems based on their effects on users' decisions and
care, measuring under- and over-escalation separately (Questions 3, 6, 14-15).
3. Connect benchmarking to clinical confirmation with intended users in intended
use environments (Questions 7-11).
4. Make postmarket surveillance actionable by linking defined signals to
responsible parties and appropriate responses (Questions 18-24).
5. Enable shared learning across the ecosystem while preserving manufacturer
accountability (Questions 21 and 25-26).

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

Why DiMe Is Commenting

DiMe draws on three complementary bodies of work:

● Evidence frameworks for medical product development. DiMe's V3+
Framework describes verification, analytical validation, usability validation, and
clinical validation as distinct components of evidence that a digital health
technology is fit for purpose.1,2 DiMe's framework for selecting reference
measures provides a hierarchy for justifying reference measures in analytical
validation, including whether validation can occur in the intended use
environment.3 The Digital Measures That Matter Framework anchors
measurement in meaningful aspects of health: what patients want to improve,
prevent, or keep from worsening.4 The sensor-based DHT (sDHT) Adoption
Navigator maps those resources and many more out in a practical roadmap to
practically implement sDHTs into clinical trials. Much of this work is led
through DATAcc by DiMe, a collaborative community with FDA's CDRH.
● Real-world evidence and postmarket oversight. DiMe's work on patientgenerated health data (PGHD) examines how PGHD and other real-world data
can strengthen oversight of AI-enabled medical devices after deployment.
● AI implementation, patient interaction, and governance. DiMe’s Playbook:
Implementing AI in Healthcare and the Scaling Trusted, High-Impact AI Care
Navigation and Operationalizing AI Governance in Healthcare programs provide
evidence and best practices informing how intended users interact with and
rely on AI outputs in real use environments.

Together, this work informs DiMe's recommendations on how FDA can connect
benchmarking, clinical confirmation, and postmarket monitoring for GenAI-enabled
medical devices, and specify the conditions of use on which that evidence depends.

Core Principle: Device Performance Depends on Intended Users
and Use Environments

GenAI risk and performance cannot be fully characterized through technical
evaluation alone. FDA recognizes that premarket or retrospective evaluations may not
predict performance in dynamic real-world environments, where patient populations,
clinical sites, inputs, and conditions of use can change.5,6 Emerging research on large
language models (LLMs) in healthcare further underscores the need to evaluate
performance within clinical workflows and human-AI interactions.7,8 Risk may also
vary with intended use, user characteristics and reliance, human oversight,
safeguards, interaction patterns, and changes to the underlying model or surrounding
system. Although DiMe's V3+ Framework is specific to sDHTs, its core organizing

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

principles equally apply here: evidence is specific to a context of use, including the
intended population and use environment, and evidence generated in one context
may not transfer to another.1,2

DiMe recommends that FDA connect premarket evidence with explicit conditions for
safe use and clear responsibilities across the device lifecycle. Manufacturers should
specify the intended users, required safeguards and human oversight, and
information needed to support safe deployment. Manufacturers should communicate
these conditions through labeling in a form that allows deploying organizations to
determine whether they can meet them. Premarket evaluation provides the baseline,
while postmarket monitoring and change management help determine whether safety
and effectiveness are maintained as technology and conditions of use evolve.

Recommendation 1: Expand Risk Assessment to Account for
Intended Users and Conditions of Use
FDA Questions Addressed: 1, 2, 3, and 5.

DiMe agrees with FDA that risk assessment should consider both a device’s function
and the potential consequences of an incorrect output. 5,6 DiMe recommends using
contextual factors to modify risk assessment rather than treating them as
independent risk categories. Relevant factors include error detectability and
reversibility, opportunities for human intervention, time available for review, and
safeguards that limit harm. Manufacturers should characterize foreseeable risks and
conditions for safe use and communicate them in labeling.

Risk also depends on who receives and acts on outputs and how much they rely on
them. Age, cognitive or developmental abilities, clinical expertise, and decisionmaking authority can affect how outputs are interpreted. Research on pediatric AI
governance highlights developmentally appropriate consent and assent, stakeholder
roles, and accountability,9 while research on automation bias shows that trust,
experience, workload, and time pressure can influence reliance on automated
decision support.10 Manufacturers should account for intended populations and
foreseeable reliance in premarket evaluation.

Risk must also be assessed across use environments and over time. FDA recognizes
that changes in patient populations, clinical practices, data acquisition, and
deployment sites can affect safety and effectiveness.11 Manufacturers can apply a
care-journey perspective in use-related risk analysis to identify how an output
prompts action, who is expected to respond, and the consequences if it is wrong.
The same technology may pose different risks in routine information-seeking,
transitions between care settings, or acute situations requiring timely intervention.

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

GenAI risk can also change during a conversation. FDA recognizes that outputs may
shift from informational to action-directing over multiple turns.5,6 DiMe agrees with
FDA that directiveness depends on the substance, specificity, and context of outputs
rather than isolated wording. Manufacturers should test realistic interaction
trajectories, including changes in context, increasingly directive outputs, and potential
failure points. Laban et al. found an average 39% performance decrease across multiturn exchanges compared with fully specified single-turn prompts in an evaluation of
15 LLMs.12 This supports evaluating realistic interactions rather than assuming singleturn performance predicts behavior in extended conversations.

Patient-facing applications require evaluating how people interpret and act on
outputs, while recognizing their potential to improve engagement and access to
information. In a randomized study of 1,298 members of the public, Bean et al. found
that LLM performance on medical assessment tasks did not reliably predict how well
people using those models identified conditions and chose appropriate actions.13 As
part of clinical confirmation, manufacturers should evaluate whether intended users
reach appropriate care in a timely manner and whether safeguards address the
consequences of inappropriate or delayed action. Human oversight should count as a
safeguard only when the person can reasonably detect, interpret, and act on an
incorrect output before harm occurs. These considerations support a risk-based,
least-burdensome approach that matches premarket evidence and ongoing controls
to the risks of the intended use (Table Below).

Illustrative Table of Premarket Characterization, Conditions of Use, and Postmarket
Monitoring

Factors Manufacturers Can Characterize Conditions of Use Specified by the
Premarket Manufacturer and Verified Through
Postmarket Monitoring

Intended User and Patient Population Availability of Human Intervention

Intended Use and Degree of Directiveness Local Workflow and Time Pressure

Expected Degree of User Reliance Site-Specific Safeguards and Escalation
Pathways

Detectability of Errors by Intended Users Staffing and Availability of Appropriate
Expertise

Traceability of Outputs and Actions Actual Workflow Integration and Use
Conditions

Foreseeable Failure Modes and Real-World Performance and Emerging

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

Consequences Failure Modes

Required Conditions for Safe Use Changes in Use, Workflow, or Reliance

The important distinction is what the manufacturer can reasonably establish versus
what only becomes knowable in a particular care environment.
● Premarket: Characterize the GenAI-enabled device, its intended use, foreseeable
risks, users, and conditions needed for safe use.
● Labeling: Communicate those conditions so deploying organizations can
determine whether they can meet them.
● Postmarket: Monitor whether those conditions hold in practice and whether
performance or risk changes over time.

Recommendation 2: Evaluate Patient-Facing Systems Through
Downstream Consequences and Trust
FDA Questions Addressed: 3, 6, 14, and 15.

For patient-facing GenAI-enabled medical devices, evaluation should assess not only
whether a system produces an appropriate response, but how patients interpret and
act on it and how those actions affect care. A care navigation tool, for example, may
influence whether someone seeks care, delays treatment, uses emergency services,
pursues an unnecessary evaluation, or follows a recommended care pathway. These
consequences depend on the patient’s circumstances, understanding of the output,
degree of reliance, and available safeguards. As evidence discussed in
Recommendation 1 illustrates, model performance alone may not predict the quality
of human-AI decision-making.13 DiMe therefore recommends evaluating patient-facing
systems at the level of the decisions and care pathways they are intended to
influence. Consistent with FDA's recognition that comparators may involve the
clinician and device working together as a human-AI team,5,6 and in response to
Question 14, DiMe recommends that human-AI team performance, rather than
device-alone performance, serve as the primary basis for evaluation when lay users
are expected to interpret outputs, when the function directs action, or when
intended users are unlikely to detect an incorrect output.

Evaluation should compare device-supported decisions with the course of action
likely to occur without the device, as FDA raises in Question 15.5,6 The appropriate
comparator will depend on the intended use and decision point and may include
usual care, unaided judgment, or another realistic alternative. For care navigation and
escalation functions, manufacturers should measure under-escalation and overescalation separately, rather than combining them into a single accuracy measure.
Under-escalation may delay necessary care, while over-escalation can cause
avoidable anxiety, unnecessary evaluations or emergency department visits, increased

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

resource use, and loss of trust. FDA’s Question 6 asks how these competing errors
should be characterized and weighted.5,6

DiMe recommends reporting under- and over-escalation rates separately and
stratifying results by clinical acuity or relevant points in the care journey. Acceptance
thresholds should reflect the consequences of each error, with less tolerance for
under-escalation when delayed care could cause significant harm. Consistent with
DiMe's Digital Measures That Matter Framework, clinical confirmation outcomes
should reflect aspects of health that matter to patients, such as receiving the right
level of care at the right time, rather than agreement with a reference answer alone.
Where feasible, evaluation should also measure whether patients reach the
appropriate level of care, whether care is timely, and how device-supported decisions
compare with the no-device alternative. This approach would help FDA distinguish
systems that perform well on benchmarks from those that improve patient decisions
and care without introducing unacceptable downstream harms.

Recommendation 3: Connect Benchmarking to Performance with
Intended Users in Intended Use Environments
FDA Questions Addressed: 7, 8, 9, 10, and 11.

DiMe supports FDA’s proposed competency-based approach of device benchmarking
followed by clinical confirmation, with the depth and type of evidence proportionate
to intended use and risk.5,6 Higher-risk devices should require more rigorous
benchmarking, representative testing, and clinical confirmation. Evaluation should
focus on the final device as configured for its intended use, rather than the
underlying foundation model alone.

Evidence from healthcare demonstrates why technical performance must be
confirmed in practice. Wong et al. externally validated the Epic Sepsis Model and
reported an area under the curve (AUC) of 0.63, substantially below developerreported performance. The model generated alerts for 18% of hospitalizations while
failing to identify 67% of patients with sepsis in the study population.14 Although this
study evaluated a conventional predictive model rather than GenAI, it illustrates the
importance of independent validation under real-world conditions. DiMe's V3+
Framework,1,2 developed for sDHts, offers a tested model for structuring this
evidence. It treats evidence as modular: verification and analytical validation
establish technical performance against prespecified criteria and references, clinical
validation establishes relevance to a meaningful clinical state in a defined context of
use, and usability validation establishes that intended users can use the technology
as intended. FDA's competency-based approach follows a similar logic, with
benchmarking corresponding to analytical validation, clinical confirmation to clinical
validation, and testing with representative users to usability validation, consistent
with FDA's human factors expectations.15 Although V3+ was not designed for GenAI,

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

unless a GenAI component is part of an sDHT, DiMe recommends that FDA use a
similar modular structure to specify what each evidence component must show and
which changes, such as a new foundation model version or intended population,
require it to be repeated.1,2

DiMe recommends that benchmark validity be demonstrated by showing that
benchmark results predict performance under clinically representative conditions.
This could include comparing benchmark and clinical confirmation results across
representative users, patient populations, workflows, and use conditions, and
documenting where the results diverge. Where feasible, sponsor-developed
benchmarks should also be assessed against independent or externally validated
measures to reduce the risk of optimizing performance to the test itself. DiMe's
analytical validation work offers a practical model: reference measures are ranked by
rigor, sponsors must justify using a lower-ranked reference when a higher-ranked one
is available, and studies must show whether results obtained in controlled settings
generalize to the intended use environment.3 DiMe recommends that FDA ask
sponsors to justify benchmark references and clinician-panel comparators against a
similar hierarchy (Questions 10 and 14).

Benchmark performance does not, by itself, establish that intended users can use
the device safely and effectively in intended use environments. Consistent with FDA's
human factors framework,15 manufacturers should characterize intended users,
conditions for safe use, expected human oversight, relevant limitations, and
foreseeable patterns of reliance. Evidence may include usability testing with
representative users and evaluation of whether the device’s limitations and
uncertainty are communicated effectively.

DiMe recommends that FDA clarify these manufacturer responsibilities through
premarket evidence and labeling. FDA's draft guidance on AI-enabled device software
functions already recommends that labeling describe intended users, degree of
automation, known limitations, workflow integration, and user-performed
monitoring.15 For GenAI-enabled devices, DiMe recommends adding a "conditions for
safe use" element stating the human oversight, escalation pathways, and staffing or
expertise that the manufacturer's evidence assumes, so deploying organizations can
determine whether they can meet them. The level of evidence should reflect the
device's intended use and risk.

Recommendation 4: Make Postmarket Monitoring Operationally
Actionable
FDA Questions Addressed: 18, 19, 20, 21, 22, 23, and 24.

Postmarket monitoring should establish how emerging safety and effectiveness
signals will be identified and acted upon. For GenAI-enabled medical devices,

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

performance may change as models, user behavior, deployment environments, or
interaction patterns evolve. Monitoring is meaningful only when the manufacturer's
plan defines which signals require action and how it will respond.

Greater reliance on postmarket monitoring may be appropriate when emerging risks
can be detected reliably and quickly, with predefined thresholds, accountable
owners, escalation pathways, and interventions. It is less appropriate when failures
may go undetected before harm occurs or meaningful intervention cannot happen in
time. DiMe recommends tying any reduction in premarket evidence to a documented
monitoring plan, with the level of premarket evidence remaining proportional to
intended use and risk.

Because DiMe's work on operationalizing AI governance in healthcare identifies
unclear accountability as a risk that can limit effective oversight, DiMe supports
applying the performance monitoring plan recommendations in FDA's draft AIenabled device software functions guidance to GenAI-enabled devices, with
predefined action thresholds and a named response for each signal. 15 Depending on
device risk, these plans could include periodic re-benchmarking, sample-based
review of real-world interactions by qualified, independent clinicians under the
manufacturer's monitoring plan, and monitoring for changes associated with new
inputs, populations, models, or deployment conditions. Routine monitoring frequency
should reflect how quickly meaningful changes could occur. Material model updates,
unexpected performance signals, changes in use, or clinically significant adverse
events should trigger reassessment when warranted. Patient-generated health data
and other real-world data can strengthen these plans by providing more timely and
continuous signals than traditional sources. DiMe's work in this area identifies shared
methods for detecting meaningful safety and performance signals as a near-term
priority for postmarket surveillance of AI-enabled products.16

Automated supervisory agents may help scale monitoring, but should not introduce
an unvalidated layer between a safety signal and human action. When used for
safety-relevant monitoring, these agents should be evaluated for task-specific
reliability, sensitivity, specificity, and failure modes. Critical signals should have
defined escalation pathways and should not depend solely on an automated agent’s
judgment.

DiMe agrees with FDA that change-control requirements should scale with a
modification's potential impact on safety and effectiveness, and that a
Predetermined Change Control Plan (PCCP) can accommodate bounded changes. For
third-party foundation models, manufacturers should establish mechanisms to
identify relevant upstream changes, such as version monitoring, change notifications,
access to relevant documentation, and predefined testing triggers. Manufacturers
should remain responsible for assessing how those changes affect the marketed
device and taking appropriate action.

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

Recommendation 5: Enable Shared Learning Without Diffusing
Accountability
FDA Questions Addressed: 21, 25, and 26.

GenAI-enabled medical devices operate within an ecosystem where manufacturers,
foundation-model developers, healthcare organizations, clinicians, patients and
caregivers, researchers, and regulators may each hold information relevant to safety
and effectiveness. Sharing performance evidence, emerging risks, and implementation
experience can strengthen evaluation and governance, provided that responsibility for
decisions remains clear.

DiMe sees potential value in voluntary Foundation Model Master Files if they provide
standardized, current, safety-relevant information that manufacturers can reference
during premarket and postmarket evaluation. Useful content could include model
versions, training-data provenance where appropriate, known limitations and
healthcare-relevant failure modes, subgroup evaluation results, safety guardrails,
update notifications, and audit-log availability.1,2 Standardized content expectations,
mechanisms for updating information, and notification of material changes would
help maintain the value of such a resource. It could reduce duplicative evaluation,
but should not replace the device manufacturer’s responsibility to demonstrate the
safety and effectiveness of the final product.

Agentic GenAI-enabled devices warrant evaluation of the actions they take, not just
the outputs they generate. Evaluation should address authorized actions, access to
tools and external systems, conditions for acting, human intervention, traceability,
and safeguards against unsafe or unauthorized actions. Acceptance criteria and
oversight should reflect the potential consequences of those actions, with
postmarket monitoring assessing changes in agent behavior and use of external tools.

DiMe recommends that FDA support mechanisms for sharing relevant information
about performance changes, emerging risks, system modifications, and mitigation
strategies across the AI-enabled medical device ecosystem. These mechanisms may
be particularly valuable when safety depends on third-party models or other external
components. FDA-affiliated collaborative communities, such as DATAcc by DiMe,
offer an existing precompetitive model for developing shared evaluation methods. 17
Shared learning should help information reach the parties best positioned to respond,
while preserving the manufacturer's responsibility for the safety and effectiveness of
the device.

In Summary

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA
Docusign Envelope ID: D30FB67B-7098-8B7F-833B-F5749BE411E4

The regulation of GenAI-enabled medical devices should reflect the reality that safety
and effectiveness depend on more than technical performance alone. They are
shaped by intended use, how patients and clinicians interpret and act on outputs, the
conditions of use and safeguards the device relies on and how risks change over
time. To put this into practice, DiMe recommends that FDA: (1) treat reversibility,
detectability, and time available for human intervention as modifiers within its twoaxis risk framework; (2) expect separate, acuity-stratified reporting of under- and
over-escalation, with human-AI team performance as the primary basis for evaluation
when lay users interpret outputs; (3) specify what each evidence component must
show and ask sponsors to justify benchmark references against a reference
hierarchy; (4) add a "conditions for safe use" element to labeling and apply its draft
performance monitoring plan recommendations to GenAI-enabled devices; and (5)
use collaborative communities to develop shared evaluation and signal-detection
methods. Expectations should be proportionate to intended use and risk, supporting
patient safety and public trust while remaining adaptable to evolving technologies
and care settings. DiMe appreciates the opportunity to comment and looks forward
to continued engagement with FDA as its approach to GenAI-enabled medical devices
develops.

Sincerely,

Joshua R. Sparks, PhD
Program Lead
Digital Medicine Society (DiMe)

Ian Miller
Associate Program Director
Digital Medicine Society (DiMe)

Ali Reed
Delivery Lead
Digital Medicine Society (DiMe)

Sarah Averill Lott, MPH, CPH
Scientific Director
Digital Medicine Society (DiMe)

Benjamin Vandendriessche, PhD
Chief Executive Officer (CEO)
Digital Medicine Society (DiMe)

Digital Medicine Society (DiMe) | www.dimesociety.org
90 Canal Street, 4th Floor, Boston, Massachusetts, 02114, USA