What counts as evidence in healthcare architecture?

01

Begin with the decision, not the paper

The Center for Health Design defines evidence-based design as basing built-environment decisions on credible research to achieve the best possible outcomes. The definition is deliberately active: evidence is used in a decision process. It is not a catalogue of universally correct room types.

A useful review starts by writing the decision question precisely. “Does nature help patients?” is too broad. “For adult medical-surgical inpatients, is access to a window view associated with measurable differences in pain, length of stay or reported experience?” identifies a population, an environmental exposure and outcomes. “How do nurses experience visibility and interruption at decentralized work points?” is a different question and requires a different form of evidence.

Codes, regulations and adopted standards also matter, but they are requirements rather than empirical findings. Local utilization data, staff observation and patient feedback are evidence about this organization. Expert judgment helps interpret all of it. Responsible planning names each source for what it is instead of allowing every input to borrow the authority of “research.”

02

Match the method to the question

Evidence does not come in one ideal form. A controlled comparison may estimate whether an intervention changes an outcome. An observational study can describe associations in real settings. Interviews and ethnography can reveal mechanisms, meanings and unintended consequences that a metric misses. Simulation can test how assumptions behave under variable demand. A post-occupancy evaluation can compare intention with performance after move-in.

An analysis of 157 research articles published in HERD between 2016 and 2020 found a wide range of study designs, strategies, populations, variables and analytic methods, along with inconsistent methodological terminology. That diversity reflects the field: health environments combine physical features, human behavior, clinical processes and organizational systems.

Evidence hierarchies can still help, but only when tied to the question. The Oxford Centre for Evidence-Based Medicine cautions that its levels should be read with their introductory framework, not as a freestanding ranking. A systematic review of weak or indirect studies does not automatically resolve uncertainty, and a carefully conducted qualitative study may be the strongest evidence for an experience question.

03

Judge confidence with five tests

A study can be relevant and still be unreliable; rigorous and still be inapplicable. Project teams need to judge both validity and fit. Five questions provide a practical screen:

1. Risk of bias: Could selection, measurement, confounding or missing data explain the result?
2. Consistency: Do independent studies point in the same direction, or is the finding isolated?
3. Directness: Are the population, setting, intervention and outcome close to the project decision?
4. Precision: Is the estimate stable enough to distinguish a meaningful effect from noise?
5. Publication bias: Could positive findings be more visible than neutral or negative results?

These tests adapt established evidence-appraisal concepts rather than turning a design team into a guideline panel. The warning is material. A 2025 scoping review of 46 studies connecting inpatient design features to clinical outcomes found that 43% lacked appropriate methods to address residual confounding. It also found that studies reporting positive associations were cited much more often than those reporting negative associations. Visibility is not the same as certainty.

For project use, a simple confidence label is often enough: high when several strong, direct and consistent studies support a claim; moderate when the signal is credible but limited; emerging when evidence is small, indirect or inconsistent; and local hypothesis when the proposition should be tested within the project.

04

Treat a healthcare setting as a complex intervention

A built environment does not act alone. A decentralized supply location may reduce walking only if stocking levels, replenishment routes, staff roles and digital systems support it. A visibility strategy may affect observation, interruption, privacy and teamwork at the same time. After occupancy, outcomes may also change because staffing, technology, policy and patient mix changed with the building.

The UK Medical Research Council framework for complex interventions is useful here. It asks researchers to consider context, program theory, stakeholders, key uncertainties, refinement and economic consequences across development, feasibility, evaluation and implementation. Applied to architecture, a program theory is the explicit causal chain connecting a spatial move to an expected outcome and the conditions required for that chain to operate.

This changes the design conversation. Instead of saying “decentralized supplies improve efficiency,” the team records: “placing frequently used supplies within the care cluster is expected to reduce retrieval travel during peak periods, provided par levels and replenishment responsibilities are reliable.” The second statement can be measured, challenged and revised.

05

Build an evidence register, not a citation pile

A project evidence register keeps research connected to decisions. Each entry should identify the claim, source, study design, population and setting, environmental feature, measured outcome, principal result, limitations, confidence, proposed design implication and local validation method. The record should also identify who accepted the implication and when it should be reconsidered.

The register prevents two common failures. First, a compelling sentence cannot travel into a report without its qualifications. Second, a team can distinguish a repeated citation from independent confirmation. If several reviews rely on the same small group of primary studies, the apparent volume of evidence is larger than the underlying evidence base.

A 2024 evidence-gap map reviewed 406 inpatient built-environment studies. Architectural features dominated the literature, while ambient, interior, social and nature-related features received less attention. Outcomes and methods varied by topic. The finding is a reminder that an empty cell in the evidence base is not evidence that a design issue is unimportant; it may simply be understudied.

06

Translate carefully, then test after occupancy

The final translation should place each design input in one of four categories: requirement, a mandatory rule or adopted policy; supported principle, a direction backed by credible and relevant research; project hypothesis, a plausible proposition that depends on local assumptions; or preference, a stakeholder choice that should be stated honestly. None is automatically unimportant, but each carries a different burden of justification.

Healthcare facility research still leans heavily on observation, surveys, post-occupancy studies, space syntax and retrospective analysis. A systematic review of 65 facility-design studies found operations research and management-science methods comparatively underused, even though they can test layouts and performance earlier. Another review found a consistent understanding of evidence-based design and post-occupancy evaluation, but a less developed understanding of predesign evaluation.

The practical implication is a learning loop: define the intended outcome, record the evidence and assumptions, establish a baseline, test the proposal before commitment where possible, and evaluate the occupied environment. Evidence-based design is strongest when a project contributes new, transparent evidence rather than only consuming old citations.