Two Models of Clinical Informatics
Clinical informatics organizations frequently operate as centralized service providers. A clinician identifies a problem, submits a request, and waits for an analyst, informaticist, or technical team to deliver a solution.
This model serves an essential purpose. Healthcare systems require specialized technical expertise, controlled production environments, validated reporting, and centralized coordination. Many problems genuinely require dedicated teams to solve them.
But a service-delivery model becomes limiting when it is the only mechanism through which clinicians can engage with institutional technology and information.
An alternative is the clinical enablement model: an approach in which the informatics organization develops the infrastructure, governance, tools, and institutional knowledge that allow clinicians to investigate problems and participate directly in improvement.
These models are not mutually exclusive. An effective institution needs both.
The strategic question is whether informatics exists primarily to complete requests or to increase the organization’s capacity to solve problems.
The Difference Between Answering Questions and Enabling Discovery
Consider a physician who wants to understand variation in clinical practice.
In a conventional request-based model, the physician must first formulate a sufficiently specific question. They identify the patient population, define the relevant variables, specify the desired output, and submit a request.
An analyst produces the report.
But meaningful analytical work rarely proceeds so neatly.
An unexpected distribution raises a new question. A subgroup behaves differently than anticipated. A data element proves unreliable. A workflow variation emerges that was not part of the original investigation.
Each discovery changes the next question.
The most productive analysis is often exploratory and iterative. It requires the person who understands the clinical context to interact with the information directly, following observations as they emerge.
When every new question requires another ticket, another explanation, and another waiting period, the process loses momentum. Some questions are never asked at all.
A system designed exclusively to answer predefined questions cannot fully support the discovery of questions that have not yet been imagined.
Low Utilization Is Not Necessarily Low Demand
A common argument against expanding self-service analytical capabilities is that relatively few clinicians use the tools already available.
But observed utilization is not an independent measure of demand.
Adoption depends on usability, discoverability, training, accessibility, workflow integration, and the effort required to obtain a useful result.
A technically capable analytical platform may be functionally inaccessible to a busy physician if it requires extensive training, knowledge of database structures, or substantial time outside clinical responsibilities.
Low utilization may indicate limited interest. It may also indicate that the institution has made the desired activity unnecessarily difficult.
These possibilities require investigation rather than assumption.
The better question is not simply how many clinicians currently use analytical tools. It is how many could contribute meaningful clinical insight if exploration were sufficiently intuitive and accessible.
The objective is not universal adoption. It is to remove avoidable barriers for clinicians whose questions and expertise could produce institutional value.
Governed Self-Service Analytics
Greater analytical independence does not imply unrestricted access to clinical information.
Healthcare data requires rigorous safeguards for privacy, security, appropriate use, and interpretive reliability.
A mature self-service environment should incorporate:
- Role-based access and minimum-necessary permissions.
- Auditability and accountability for data use.
- Standardized definitions and governed clinical concepts.
- Transparent data provenance and limitations.
- Validated datasets and appropriate analytical guardrails.
- Escalation pathways for complex or high-risk analyses.
These controls should be embedded in the analytical environment rather than imposed entirely through manual gatekeeping.
Good governance does not merely restrict what users can do. It creates the conditions under which more people can safely do useful work.
The goal is to make appropriate access easier while preserving centralized oversight of the standards, infrastructure, and safeguards that make the information trustworthy.
The Role of AI in Lowering Analytical Friction
One of the most persistent barriers to clinical analytics is the distance between clinical concepts and enterprise data architecture.
A clinician may know precisely what information they want to investigate without knowing the relevant database tables, field names, source systems, or relationships among data elements.
Traditional analytical environments often expect users to bridge that gap themselves.
Semantic discovery tools offer a different approach.
An institution-aware language model could help users translate clinical questions into candidate data elements, identify relevant sources, understand relationships between concepts, and navigate the institutional data model.
For example, a physician interested in transfusion practices should not need to know the underlying database schema before discovering where transfusion orders, administration records, laboratory results, and clinical outcomes reside.
An effective discovery system could guide that exploration while clearly distinguishing verified metadata from inferred relationships.
The model would not independently authorize access, establish the validity of an analysis, or replace expert review. Its purpose would be to reduce the technical burden of finding and understanding information.
This represents an important application of AI: not replacing clinical judgment, but making institutional information more accessible to the people who possess it.
Informatics as Institutional Infrastructure
The broader implication extends beyond analytics.
A mature clinical informatics organization should not be judged solely by ticket completion, project throughput, or the number of solutions it delivers.
It should also be evaluated by the capabilities it creates throughout the institution.
Can clinicians recognize and investigate workflow problems? Can they participate meaningfully in system design? Can they evaluate the effects of interventions? Can they discover and use institutional information without unnecessary dependence on technical intermediaries?
These capabilities depend on deliberate investment in governance, education, accessible tools, reliable data, and relationships between central informatics teams and clinical domains.
Centralized expertise remains essential. But its highest-value contribution is not necessarily performing every task itself.
It is creating an environment in which distributed clinical expertise can contribute effectively.
Toward a Learning Health System
The distinction between service delivery and clinical enablement ultimately concerns the kind of institution we want to build.
A request-driven organization improves when someone identifies a problem, formulates a solution, and persuades a central team to act.
An enabled organization develops more pathways through which problems can be discovered, investigated, understood, and addressed.
This does not eliminate the need for prioritization, expertise, or coordinated governance. It makes those functions more valuable by connecting them to a broader network of clinical insight.
The long-term objective is a learning health system in which information, clinical expertise, and technological capability reinforce one another.
The highest achievement of clinical informatics is not answering every question clinicians bring to IT. It is building an institution in which clinicians can discover and answer more of those questions themselves.