# Building Trust in Multi-Institutional Research Collaboration Most multi-institutional collaborations that lose trust did not lose it over the science. They lost it over governance. The single most common cause of failed trust in multi-institutional research collaboration is the absence of a written, shared agreement about who decides what, who owns which data, and who gets credit for what. When those questions stay implicit, each institution fills the silence with its own assumptions, and the first disagreement over authorship or data access turns into a referendum on whether the partnership is fair. The first thing to check, before commissioning a new assay or renegotiating a budget, is simple: does the collaboration have a document that states its mission, its leadership structure, and its rules for data and credit? Large consortia that work well almost always do. The PROSPR-Lung consortium, a multi-site lung cancer screening research program, listed "developed and reinforced a shared mission, vision, and goals" and "maintained a transparent and representative leadership structure" as the first two practices it applied from the Science of Team Science, and it named building a culture of trust as an explicit, deliberate objective rather than a byproduct [1]. If that document does not exist, creating it is the fix. If it exists but nobody has read it since the grant was submitted, updating and re-ratifying it is the fix. This article walks through the causes of trust failure from the most common to the rarest, explains the mechanism behind each one, gives a diagnostic sequence you can follow, and then lays out the solutions, tools, and prevention practices that keep scientific collaboration across institutions productive. ## Why Trust Breaks Down in Multi-Institutional Collaboration Trust between institutions is not a feeling. It is a prediction: each partner predicts that the others will behave as agreed under conditions where nobody is watching every move. Trust fails when that prediction fails, and the reasons cluster into seven recurring patterns. ### Cause 1: No Shared Mission (Most Common) The most frequent cause is also the least dramatic. Institutions join a collaboration for different reasons. A university medical center may want publications, a community hospital may want access to expertise, a funder may want measurable population impact. None of these motives is wrong, but when they are never reconciled into one shared mission, every later decision becomes a negotiation from first principles. Partners begin to suspect that others are optimizing for their own goals at the group's expense, and suspicion is the seed of distrust. The mechanism is well documented. Evaluations of long-running partnerships consistently find that a common agenda and mutual goals are the foundation on which everything else sits. The University of Missouri-Kansas City Health Care Quality and Patient Safety Consortium, which spans six affiliated hospitals and four university-based health sciences schools, explicitly established a governance structure whose members "share a common agenda and mutual goals," and measured its success against published frameworks for educational and patient outcomes [2]. Without that alignment step, trust has nothing to rest on. ### Cause 2: Opaque Leadership and Credit Allocation The second most common cause is ambiguity about power and credit. Who is the principal investigator of record? Whose institution's IRB reviews the protocol? Who appears on the byline, in what order, and who decides? In single-institution science these questions have local answers. Across institutions, they have no default answer at all, so silence becomes a dispute waiting to happen. Credit disputes are especially corrosive because they are zero-sum and public. A junior collaborator who believes their contribution was erased does not simply leave. They tell colleagues at their home institution, and the reputational shadow extends to every future partner that institution might recruit. Trust damage from credit disputes is institutional, not individual. ### Cause 3: Data That Cannot Be Trusted The third cause is technical but produces social damage. When Site A sends data that Site B cannot reproduce, parse, or reconcile with its own records, Site B does not usually conclude "the data model is inconsistent." It concludes "their team is careless." Scientific collaboration depends on each partner's data being trustworthy, and multi-institutional data are structurally vulnerable because every institution has its own local terminology, its own coding habits, and its own quality culture. The problem is solvable, and the solutions themselves build trust. One framework implemented across three hospitals within a shared healthcare data platform standardized local terms to common clinical terminologies and evaluated quality along six dimensions: validity, uniqueness, completeness, consistency, timeliness, and accuracy. Inter-institutional harmonization minimized semantic differences between sites, and expert review with adjudication resolved issues in every dimension [3]. When partners can see that a framework caught format errors, removed duplicates, and bridged mapping gaps, they stop wondering whether the pooled dataset is reliable. The antidepressant safety literature offers a concrete demonstration of what becomes possible when standardization succeeds. A [cohort study](/blog/guides/cohort-studies-a-practical-guide-to-design-conduct-and-interpretation) drew on electronic medical records from 15 hospitals standardized into the Observational Medical Outcomes Partnership Common Data Model, a shared structure that let the investigators build comparable patient cohorts at every site and pool results with fixed-effect meta-analysis [4]. No partner would have accepted pooled conclusions from fifteen hospitals' raw, non-standardized records. Standardization is what made the pooled result credible to all of them. ### Cause 4: Privacy Fears and Data-Sharing Anxiety The fourth cause is the fear that sharing data will expose an institution to privacy violations, regulatory action, or reputational harm. This fear is rational, not paranoid. Health data contain identifiers, and quasi-identifiers (combinations of ordinary fields such as age, sex, and ZIP code that together can re-identify an individual) are easy to underestimate. The technical answer is the distributed research network (DRN), an architecture in which each institution standardizes its data into a common data model and shares only analysis results, never raw records. Validation studies have quantified how de-identification choices change privacy risk in such networks by comparing limited and safe-harbor datasets and measuring the "trust differential gap" between privacy policies [5]. For a collaboration, the significance is institutional: when a partner can verify that raw data never leaves its walls, the negotiation over sharing shifts from "do we trust you with our patients' records" to "which analysis results can we release," a far easier conversation. ### Cause 5: Capacity and Resource Asymmetry The fifth cause is inequality between partners. When one institution brings most of the funding, most of the analytical staff, or most of the publication infrastructure, smaller partners can end up contributing patients, specimens, or community relationships while receiving proportionally little of the credit, training, or revenue. Over time the weaker partners conclude the arrangement is extractive, and the stronger partners conclude the weaker ones are underperforming. Both conclusions destroy trust. Long-lived partnerships solve this deliberately. The Meharry-Vanderbilt-TSU Cancer Partnership, funded through the National Cancer Institute's program to build multi-institutional partnerships addressing cancer health disparities, is the longest continuously funded site in that program. A qualitative evaluation that interviewed 43 stakeholders, including faculty, students, and community advisory board members, found four categories of impact: increased funding and financial return on investment, institutional collaboration, educational and career advancement, and community outreach, education and engagement [6]. The design distributed benefit across all three institutions rather than concentrating it at the best-resourced one, which is precisely why it survived long enough to become the program's longest-running example. ### Cause 6: Missing Project Management Infrastructure The sixth cause is mundane and preventable: the collaboration has no operational backbone. Meetings are scheduled through personal email, files live in individual accounts, action items from monthly calls are never tracked, and each partner's staff spends hours reconstructing what was decided. Every missed handoff looks to the other side like disrespect or disorganization, and accumulated small failures read as untrustworthiness. Consortia that run well treat infrastructure as a trust technology. PROSPR-Lung described employing strong research support systems, providing efficient and effective [data management](/blog/guides/data-management-basics-principles-processes-and-best-practices), and detailing "specific project management tools and processes used to drive collaboration, efficiency, and scientific productivity" as core practices rather than administrative afterthoughts [1]. ### Cause 7: Crisis and Shock (Rarest) The rarest cause is an external shock that outruns the partnership's existing agreements. The COVID-19 pandemic is the clearest recent example. A rapid qualitative assessment conducted through the Hospital Medicine Reengineering Network (HOMERuN) held five focus groups with 33 individuals from 24 institutions, representing 52 percent of the network's sites, and found that workforce adaptations shifted between the first wave of COVID patients and subsequent waves, with three global themes emerging: adaptability and comfort with dynamic change, the importance of local context, and the strain of repeated replanning [7]. Partnerships without flexible governance and rapid communication channels found their trust tested not by bad behavior but by impossible conditions. The lesson is that trust built only for stable conditions is fragile trust. ## Troubleshooting Table Use this table to move from an observed symptom to a confirmed cause and a concrete fix. | Symptom | Likely cause | How to confirm | Fix | |--|--|--|--| | Partners agree in meetings, then act inconsistently afterward | No written shared mission or charter | Ask each partner independently to state the collaboration's goals and decision rules in writing, then compare the answers | Draft or refresh a governance charter, have every institution sign it, and revisit it annually [1] | | Repeated arguments over authorship, author order, or presentation slots | Opaque credit allocation | Check whether any document specifies authorship criteria, or whether decisions are made ad hoc by the most senior voice | Adopt authorship criteria modeled on the ICMJE recommendations at project kickoff and apply them before the first manuscript [1] | | One site's records cannot be reconciled with another site's dataset | Inconsistent local terminology and data capture | Run a data quality audit across sites against defined dimensions such as validity, completeness, and consistency [3] | Adopt a common data model and a centralized terminology master, with scheduled QA and expert adjudication [3] | | A partner refuses to share data or delays indefinitely | Privacy fear or unclear data-use terms | Ask directly what privacy or legal concern is blocking release, and what the institution's counsel requires | Move to a distributed research network model where only analysis results are shared, and document the privacy risk assessment [5] | | Smaller partners miss meetings, stop volunteering sites or patients, or quietly disengage | Perceived extractive arrangement | Review where funding, authorship, training, and community benefit have actually flowed over the past two years | Restructure so education, career advancement, and funding flow to all sites, as demonstrated by long-running academic partnerships [6] | | Deadlines slip, decisions are relitigated, staff duplicate work | No project management backbone | Inventory where project files, decisions, and action items currently live, and count how many systems hold them | Consolidate on shared project management tools with a named coordinator, following consortium best practices [1] | | Trust collapses suddenly after an external crisis | Governance designed only for stable conditions | Ask whether the charter specifies who can decide quickly when normal processes are impossible | Add contingency and rapid-decision provisions, and hold structured debriefs after each crisis, as surge-planning networks did [7] | | Long-standing partners drift apart with no visible conflict | Complacency, no feedback loop | Survey partners anonymously on whether the collaboration still serves their institutional goals | Institute periodic structured evaluation, including qualitative interviews across partner types [6] | ## A Diagnostic Sequence You Can Follow ![Six-step diagnostic flow from document check through leader interviews, data audit, benefit review, stakeholder mapping and trust measurement.](/article-images/illustrations/trust-in-multi-institutional-research-collaboration-diagnostic-sequence.webp) When trust is already strained, do not begin with a trust-building retreat. Begin with evidence, in this order. 1. **Document check.** Locate the charter, memorandum of understanding, data-use agreements, and authorship criteria. Note which documents exist and which are missing. Missing documents are themselves findings. 2. **Governance interview.** Ask the leaders of each partner institution, separately and in the same words, what the collaboration's mission is, who decides what, and how credit is allocated. Divergent answers locate the fault line. 3. **Data audit.** Take one small, recent data exchange and trace it end to end. Check whether the terminology mappings, quality checks, and timeliness of updates were defined and performed [3]. Most "their data are sloppy" complaints resolve into specific, fixable defects once traced. 4. **Benefit review.** List what each institution has received over the partnership's life: funding, authorship positions, trainees mentored, community impact, infrastructure. Asymmetry here predicts disengagement before it becomes visible. 5. **Stakeholder mapping.** Enumerate every actor who touches the collaboration, including staff and community partners, not only principal investigators. A systematic approach developed for the Global Polio Eradication Initiative showed how to define a target "universe" of actors and enumerate a source population for structured assessment, a method that surfaced the full range of contributors whose perspectives were otherwise missed [8]. 6. **Trust measurement.** Where the stakes justify it, measure trust directly. A multi-institutional pilot across three Clinical and Translational Science Award hubs used concept mapping, a mixed-methods technique in which participants brainstorm contributing factors and then sort and rate them to reveal how the factors relate, with three stakeholder groups: community, health care, and academic partners. The work treated trust as "social capital" and produced an initial set of measures that can track change over time and distinguish positive from negative aspects of trust [9]. Applying such a measure before and after an intervention turns "we feel better about each other" into evidence. The sequence matters because each step constrains the next. A document check that reveals no charter explains divergent governance interviews. A governance interview that reveals unclear decision rights reframes the data audit, because data-delay disputes are often authority disputes in disguise. ## How to Fix Trust in Multi-Institutional Research Collaboration The solutions map directly onto the causes. None of them is exotic. All of them require someone to actually do the work. ### Write and Ratify a Governance Charter The charter should state the mission, the decision-making structure, the rules for adding and removing sites, data ownership and sharing terms, authorship criteria, publication review timelines, and conflict-resolution steps. Keep it short enough that people read it. The consortium literature is consistent that governance is not overhead, it is the mechanism by which partners verify each other's predictability. ### Make Leadership Transparent and Representative A leadership structure in which every partner can see that its interests have a voice prevents the quiet accumulation of grievance. The UMKC consortium built governance around leaders who "collectively, represent the diverse members and stakeholders" of the partnership [2]. Representation is not symbolic. It determines whose concerns reach the table before decisions are final. ### Build a Shared, Auditable Data Infrastructure ![Decision path for sharing patient data, branching on keeping raw records local, common data models and published quality audits.](/article-images/illustrations/trust-in-multi-institutional-research-collaboration-data-sharing-path.webp) Converge on a common data model, such as the OMOP Common Data Model, so that every site's records use the same structure and standard vocabularies [4]. Then maintain quality actively. A practical pattern is a centralized terminology master assessed continuously against the six quality dimensions of validity, uniqueness, completeness, consistency, timeliness, and accuracy, with inter-institutional harmonization to minimize semantic drift between sites [3]. Publish the audit results to all partners. Visible quality assurance converts data trust from an act of faith into a verified property. A data-safety note belongs here. Any collaboration handling human data carries real hazards, including re-identification risk, regulatory exposure, and harm to participants if identifiers leak. Before establishing any sharing arrangement, complete a privacy risk assessment appropriate to your data types, prefer architectures in which raw records remain at the source institution, and treat de-identification as an ongoing QA task rather than a one-time export step [5]. ### Manage Credit Before the First Manuscript Decide authorship criteria, data-access rights for secondary analyses, and intellectual property terms before results exist. Retroactive credit negotiation after a finding is announced is the single most trust-destroying meeting in scientific collaboration. ### Use Structured Consensus Methods for Shared Decisions When partners must agree on a shared protocol, template, or definition, consensus panel methodology beats open debate. The PACT-UK project developed a standardized radiological reporting template for pancreatic cancer through a three-stage process: a core stakeholder group of oncologists, radiologists, and surgeons identified the current landscape and ran a blinded validation study, consensus panel methodology produced a provisional draft, and then every UK hepatopancreatobiliary unit trialed the template in clinical practice with feedback driving the final version [10]. The design made every participant a co-author of the standard, which is exactly why they trusted it. ### Invest in the Weaker Partners If one institution has less analytical capacity, less grants administration, or less publication experience, budget for the gap. Training pathways, co-mentoring, and deliberate inclusion of junior faculty and students from every site convert asymmetry from a source of resentment into a source of shared success [6]. ### Measure Trust Periodically Treat trust as a tracked variable, not a vibe. Concept-mapping-derived measures developed across CTSA hubs can be applied to track changes over time, compare how stakeholder groups conceptualize trust, and identify the characteristics that maintain it [9]. A partnership that measures trust can detect decline while it is still cheap to reverse. ## Research Collaboration Tools That Support Trust Tools do not create trust, but the right tools make trustworthy behavior easy and untrustworthy behavior visible. - **Governance and charter documents.** A signed, versioned charter is the foundation. Store it where every partner can access the current version and the change history. - **Shared project management platforms.** Consortia running at scale credit specific project management tools and processes with driving collaboration, efficiency, and scientific productivity. The key properties are a single source of truth for decisions, tracked action items with owners and dates, and shared calendars and agendas visible to all sites [1]. - **Common data models.** The OMOP Common Data Model and comparable standards give every institution the same data structure and vocabulary, which is the precondition for pooling results across sites [4]. A quality assurance framework with periodic audits against six quality dimensions keeps the shared model healthy over time [3]. - **Distributed research networks.** A DRN standardizes multi-institutional data into a common data model and shares only analysis results, protecting patient privacy while enabling pooled science [5]. - **Standardized reporting templates and structured consensus workflows.** Templates built through staged consensus, trial, and refinement, such as the PACT-UK radiology reporting proforma, reduce ambiguity within and between institutions [10]. - **Structured sampling and stakeholder frameworks.** When a collaboration needs to understand its own actors, systematic approaches that define the target population, enumerate a source population, and survey it in a repeatable way prevent the loudest voices from standing in for the whole partnership [8]. - **Qualitative and concept-mapping methods.** Focus group networks and concept mapping give partners a structured, comparable way to surface concerns across institutions, as demonstrated by multi-institution qualitative work across more than twenty hospital sites [11]. ## What Successful Multi-Institutional Collaboration Looks Like Concrete examples anchor the abstract advice, and the literature offers several across different scales and disciplines. At the largest scale, the Global Polio Eradication Initiative coordinated actors across the globe and seven study countries, and the lessons-learned effort required a systematic sampling method simply to identify the full range of people involved. The scale of coordination is the point. Initiatives of that size prevented millions of paralyzes, and their operational complexity shows why governance and systematic stakeholder identification are not optional [8]. In surgery, a multi-institutional collaboration pooled outcomes from 4011 patients treated by 119 surgeons at 22 institutions between 1997 and 2022 to study how surgical experience relates to kidney function after partial nephrectomy. Nearly a quarter of the surgeons contributed learning-curve cases to both laparoscopic and robotic approaches, which is only possible when institutions trust each other enough to share long-term operative histories [12]. In pathology and artificial intelligence, a multi-institutional retrospective study trained a deep-learning model on 5835 whole slide images from the National Cancer Institute, the Children's Brain Tumor Network, and the Digital Brain Tumour Atlas, then tested it on 5516 samples from four additional institutions in the United States and United Kingdom across 52 central nervous system tumor types. Training data and test data came from different institutions precisely so the test would be independent, an arrangement that requires every partner to trust the others' labeling and processing standards [13]. In psychiatry, a Clinical Research Center for Neuromodulation in Psychiatry linked a leading mental health institute with two other institutions to build interventional psychiatry capacity tailored to resource-limited settings, with translational studies and training programs spanning all partners [14]. In adult congenital heart disease, a partnership between two research networks designed a cross-sectional pilot enrolling 500 young adults with moderate or severe complexity disease across 14 centers in North America, divided into four diagnostic groups of 125 participants each [15]. In nursing research, a partnership between five clinical sites and the national research center of a healthcare technology organization showed that multisite studies can succeed outside traditional academic consortia when strategies for success, obstacles, and leadership expectations are discussed explicitly from the start [16]. And in hospital medicine, the HOMERuN network repeatedly convenes dozens of institutions for rapid qualitative work, first on COVID-19 surge planning across 24 institutions [7] and later on diagnostic excellence across 23 institutions with 36 participants including patient representatives [11]. A network that has run high-trust, rapid-turnaround studies across so many sites has institutionalized the practices this article describes. ## Common Mistakes and Limitations - **Mistaking politeness for trust.** Harmonious meetings can conceal partners who have already disengaged. Anonymous structured measurement, not meeting tone, reveals the true state. - **Treating governance as paperwork for the grant, then ignoring it.** A charter that no one consults during a conflict is worse than none, because it signals that the partnership's written word does not bind anyone. - **Centralizing everything at the strongest partner.** Even when efficient, concentration of analysis, data, and credit breeds the perception of extraction. Distributed designs and shared data models exist precisely to avoid this. - **Assuming data standardization is a one-time setup.** Terminology mappings degrade as coding systems and local practice change. Without scheduled maintenance against current standard releases, quality decays silently [3]. - **Fixing credit disputes after publication rather than before.** Once a manuscript is submitted, the options are bad and worse. Front-loaded authorship agreements cost an hour and prevent months of damage. - **Limitation of the evidence base.** Much of what is known comes from consortia in healthcare, clinical research, and public health. Teams collaborating on bench science, engineering, or the humanities should expect to adapt rather than copy these practices, because regulatory and data-sharing constraints differ. - **Expecting trust to generalize.** Trust is specific to task and context. A partner who is reliable in data sharing may be unreliable in protocol adherence. Diagnose per dimension rather than issuing a global verdict on a partner's character. ## Prevention Practices 1. **Charter before the first data point moves.** Write the mission, decision rights, data terms, and authorship rules first, sign them, and review them annually. 2. **Adopt a common data model and publish QA results.** Standardize vocabulary and structure at every site, audit continuously against defined quality dimensions, and share the audit trail with all partners [3][4]. 3. **Default to architectures that minimize raw data movement.** Distributed designs in which only analysis results cross institutional boundaries reduce both privacy risk and the trust burden on every partner [5]. 4. **Build representative governance.** Ensure every partner class, including community representatives where research touches communities, holds a seat with real authority [2][9]. 5. **Use structured consensus and trial stages for shared standards.** Draft, trial in real settings at all sites, refine, and finalize, so every partner shapes the standard they must live with [10]. 6. **Budget for capacity building at smaller sites.** Route training, mentorship, and career advancement through all institutions deliberately [6]. 7. **Maintain project management infrastructure from day one.** One platform, one decision log, one owner per action item [1]. 8. **Stress-test governance before a crisis.** Define who can decide rapidly when normal processes fail, and debrief after every shock so the partnership adapts rather than erodes [7]. 9. **Measure trust on a schedule.** Apply validated trust concepts and measures at fixed intervals so decline is caught early [9]. 10. **Map stakeholders systematically as the collaboration grows.** Redo stakeholder enumeration at each major expansion so no contributing group becomes invisible [8]. ## Frequently Asked Questions ### What is the most common cause of trust failure in multi-institutional research collaboration? The most common cause is the absence of a written shared agreement on mission, decision rights, data ownership, and credit. When those terms stay implicit, every later disagreement reopens the whole partnership, and partners begin assuming the worst about each other's motives. ### How do you fix low trust between partner institutions? Start with evidence rather than sentiment. Check whether a charter exists, interview leaders separately to find where their accounts diverge, audit recent data exchanges, and review how benefits have actually been distributed. Then fix the specific defect found, usually by writing or refreshing governance agreements, standardizing data, or rebalancing credit and benefit. ### What tools help scientific collaboration across institutions? The most useful tools are a signed governance charter, shared project management platforms with a single decision log, common data models that give every site the same data structure, distributed research networks that share only analysis results, and standardized templates built through structured consensus processes. ### How can institutions share patient data without breaking privacy rules or trust? Use a distributed research network architecture in which each institution keeps raw records and shares only standardized analysis results. Pair this with documented de-identification choices and a quantified privacy risk assessment so partners can verify the protections rather than take them on faith. ### How do you handle authorship and credit disputes in a consortium? Prevent them. Adopt written authorship criteria at project kickoff, before any results exist, and apply data-access and intellectual property terms from the same moment. Retroactive negotiation after a finding is announced is the most trust-destroying meeting a collaboration can hold. ### How do you measure trust in a research partnership? Treat trust as social capital and measure it with structured methods. Concept mapping across community, healthcare, and academic stakeholder groups produces a set of trust factors that can be tracked over time, compared between groups, and used to distinguish the characteristics that build trust from those that erode it. ### Are multi-institutional collaborations worth the overhead for a small lab? Yes, when the science genuinely requires scale. Multi-site designs access patient populations, specimen volumes, and independent validation that no single institution can provide, and small partners in well-structured collaborations gain training, authorship, and funding they could not reach alone. The overhead is real, so invest the fixed cost of a charter and project infrastructure once rather than paying recurring costs in conflict. ### Can a collaboration recover after a serious breach of trust? Recovery is possible when the breach has an identified, specific cause and the partnership responds with structural change rather than apology alone. Reratify the charter, correct the harmed partners visibly, implement the missing safeguard, and then measure trust at intervals to confirm the repair held. ## Sources 1. [Application of team science best practices to the project management of a large, multi-site lung cancer screening research consortium.](https://pubmed.ncbi.nlm.nih.gov/37456270/) 2. [Establishing a Multi-Institutional Quality and Patient Safety Consortium: Collaboration Across Affiliates in a Community-Based Medical School.](https://pubmed.ncbi.nlm.nih.gov/32590471/) 3. [Implementing a quality assurance and maintenance framework for standardized clinical terminology data in a multi-institutional setting.](https://pubmed.ncbi.nlm.nih.gov/42286653/) 4. [Antidepressants and the risk of hyponatremia: A multi-institutional cohort study using observational medical outcomes partnership-Common Data Model.](https://pubmed.ncbi.nlm.nih.gov/41531184/) 5. [Evaluation of the Privacy Risks of Personal Health Identifiers and Quasi-Identifiers in a Distributed Research Network: Development and Validation Study.](https://pubmed.ncbi.nlm.nih.gov/34057426/) 6. [Understanding and evaluating the impact of a multi-institutional academic partnership to reduce cancer health disparities.](https://pubmed.ncbi.nlm.nih.gov/42464333/) 7. [Adaptability on Shifting Ground: a Rapid Qualitative Assessment of Multi-institutional Inpatient Surge Planning and Workforce Deployment During the COVID-19 Pandemic.](https://pubmed.ncbi.nlm.nih.gov/35319085/) 8. [Sampling method for surveying complex and multi-institutional partnerships: lessons from the Global Polio Eradication Initiative.](https://pubmed.ncbi.nlm.nih.gov/32787958/) 9. [Trust in Community-Engaged Research Partnerships: A Methodological Overview of Designing a Multisite Clinical and Translational Science Awards (CTSA) Initiative.](https://pubmed.ncbi.nlm.nih.gov/30612444/) 10. [PACT-UK (PAncreatic Cancer reporting Template-UK): a cross-specialty multi-institutional consensus panel development of a standardised radiological reporting proforma for pancreatic cancer.](https://pubmed.ncbi.nlm.nih.gov/39886489/) 11. [A multi-institutional rapid qualitative assessment of factors supporting diagnostic excellence in hospital medicine.](https://pubmed.ncbi.nlm.nih.gov/42429159/) 12. [Surgical Experience and Functional Outcomes after Laparoscopic and Robot-Assisted Partial Nephrectomy: Results from a Multi-Institutional Collaboration.](https://pubmed.ncbi.nlm.nih.gov/39408076/) 13. [Classification accuracy of a hierarchical molecular inference-based deep-learning system for CNS tumour diagnosis: a multi-institutional, retrospective study.](https://pubmed.ncbi.nlm.nih.gov/41643698/) 14. [Clinical Research Center for Neuromodulation in Psychiatry: A Multi-Center Initiative to Advance Interventional Psychiatry in India.](https://pubmed.ncbi.nlm.nih.gov/34789940/) 15. [Design of a multi-institutional neurocognitive discovery study in adult congenital heart disease (MINDS-ACHD).](https://pubmed.ncbi.nlm.nih.gov/37084934/) 16. [Innovative Collaborative Partnerships to Conduct Multisite Research: Strategies for Success.](https://pubmed.ncbi.nlm.nih.gov/30973428/)