Healthcare Design - March 2015 - 50

R RESEARCH | 03.15
The wrong way to conduct EBD research
The validity of evidence-based design depends on some fundamental research strategies-
strategies that can be easy to overlook
By Upali Nanda
Jason Schroer, director of the Houston office of HKS,
posed an interesting question recently that made me
think deeply about research in healthcare design and,
in particular, evidence-based design (EBD). He drew
a parallel between our field and the popular Animal
Planet show "Finding Bigfoot."
The program chronicles a team of three researchers who've spent decades tracking evidence
of Big Foot in different parts of the world. They all
believe in the existence of Big Foot and claim to
have had their own personal encounters. A fourth
member of the team is a field biologist who is a
skeptic but accompanies the team on its investigations, providing logical and rational explanations to
the rustles, bumps, and whoops that the others see
as evidence.
The big question that one has to ask is, does all of
the evidence they find point to Big Foot because they
believe in Big Foot? In other words, is the hypothesis
forcing them to accept this single explanation and
reject other potential explanations?
This analogy holds surprisingly true for EBD. Are
we so hypothesis-driven that we're rejecting all other
explanations for improved outcomes, and zooming in
on the ones that support our hypothesis for design?
Let's say we argue that the physical environment reduces infections. In our studies, are we only
looking at data that supports design as the solution
as opposed to other possible reasons for reduced
infections? Are we, in a way, seeking to find "Big Evidence," simply because we believe or want to prove
its existence?
The power of the null hypothesis
It's a question many skeptics pose for EBD and many
researchers ask themselves. It's very easy to go down
the slippery slope of testing a hypothesis and finding
evidence to support it-but this approach is, in fact,
fundamentally contrary to scientific method.
If you take a statistics class, one of the first things
you learn is that you never set out to prove the hypothesis. In fact, you set out to disprove the opposite.
The null hypothesis is the assumption of innocence in
a scientific experiment and corresponds to a default
or neutral state, assuming an absence of relationship
between key elements studied.
The null hypothesis provides a neutral starting
point. If we're interested in exploring, say, the relation50

HCDmagazine.com 03.15

ship between flooring and falls, the starting assumption
is that no relationship exists. An alternative hypothesis is
that some relationship exists and that types of flooring
can either increase or decrease falls. This fundamental
approach of starting from a neutral ground is key: It ensures that we're not busy proving what we hope to find.

Are we so hypothesis-driven that we're
rejecting all other explanations-beyond
design-for improved outcomes?
Correlation is not causality
The other key component of conducting good research
is the distinction between causality and correlation. Just
because two things happened simultaneously doesn't
mean that one caused the other.
For example, a new replacement hospital opened
and its HCAHPS scores went up-two events that
happened simultaneously. We can go back and study
if the increase in HCAHPS scores coincided with the
new hospital and whether it had a sustained effect.
If it did, we can argue for a correlation. We cannot,
however, claim without a shadow of doubt that the
facility design of the new hospital caused improved
HCAHPS scores.
There could be many factors that affected the score,
such as Lean process improvements that were implemented, a change in management, shifts in staffing
models, etc. We can speak of the potential relationship
between these concurring changes and see some patterns emerge. But when so many factors change simultaneously, we have to be careful before we state "our
new hospital resulted in improved HCAHPS scores."
It's more accurate to say that the HCAHPS scores
increased by a meaningful margin, that the effect has
been sustained over the last three years since opening,
and that based on an analysis of the data, one can
argue that the building design was a contributing factor
(not a cause) of this improvement.
It's critical that we make the distinction, so designers and researchers stay credible during conversations with a healthcare community that understands
scientific methods very well.
Once a pattern emerges in the HCAHPS scores-for
example, ratings for "quiet at night" went up after the
new hospital was opened-one can go back and do
an exploratory study. If good baseline information was

Upali Nanda

http://www.HCDmagazine.com

Healthcare Design - March 2015

Table of Contents for the Digital Edition of Healthcare Design - March 2015