Multilevel Interventions in Health Care Conference: Presentation by Kurt Stange, MD, PhD

>>>DR. STEVEN CLAUSER: So I'll introduce now Kurt Stange.
Kurt is a family physician. He has been practicing for many
years. He has done a lot of work in family oriented system
design and primary care. He is the editor of the Annals of
Family Medicine. He and his team did an incredible amount
of work preparing for this and we are happy to have him
and it's an honor to have him participating with us today.
>>>DR. KURT STANGE: Thanks, Steve. Erica Bresau,
Allen Dietrich, Russ Glasgow and I contributed equally
to this work. We actually developed a method for the
madness of trying to synthesize this diverse literature.
And so the method started with those of us who were on
the team. We tried to develop a team that would have a
multi-disciplinary perspective to look at the literature from
different angles. And we did multiple iterations of
literature searches that were designed to include the breadth
of the field concentrating on the cancer control continuum,
but also trying to inform the literature search
from other fields of multilevel interventions.
And then we developed a matrix of article types and tried to
make sure we had filled in that matrix.
So there were three dimensions of that.
There were cancer and non-cancer articles;
intervention and observation articles;
and then three levels that went from a theory to methodological
work, to empirical research. And then we spent three and
a half days immersed in this literature, getting it down to
about 45 articles that we thought were worthy of
particular input. And we did this where there were no
distractions except for the overhead noise of airplanes
and really bad food at one of the Chicago O'Hare airports.
And our goal there was to really characterize how
multilevel interventions are currently conceptualized and
implemented in the literature, to describe illustrative
examples and to identify opportunities for multilevel
interventions to advance cancer control.
So I've got a couple of slides to characterize this
literature, three slides on these opportunities,
and then one closing slide of the discussion questions we'd
like you to consider. So we identified three threads,
three large threads in this literature. The first are
a large series of contextualized single level studies.
So these are studies that typically intervene on one
level, often the individual or the practice or provider.
But they actually consider other contextually important levels.
And in some way this is not part of the charge of
the conference. And yet if we think of where the literature
is now, where mostly we have intervention studies that just
look at one level, this would be a huge advance if those
studies that were targeting one level, actually just put
that level in context. Maybe gather a little data on some of
those other levels, put that single level in a
larger theoretical context, and interpreted the findings in
that. So as we're thinking specifically about multilevel
interventions here, there's also an opportunity to think about
single level studies that we actually put in context.
And there were a lot of studies that started to try to do that.
There was a large body of literature of health system
research. And actually the work that Steve Taplin did when he
was at Group Health at Puget Sound, and that many others
here in the HMO were Cancer Research Network and others
have done. These studies typically include interventions
at the individual and the practice level, often at the
system level. And this can be done because these integrated
systems actually really have access to data,
and their business model fosters them thinking about
those levels. And they also often have sufficient community
context that they can involve that level,
or that they consider the local or regional health
policy level. So there was a large body of literature from
these health system intervention studies.
And then finally we saw a lot of studies that were
community-wide studies. And we've heard references to a
number of those as well. These studies actually typically took
a public health perspective, but within that larger frame looked
at health care system family and individual level interventions,
practice level interventions. And because of their public
health perspective, when they looked at individuals sometimes
they were thinking of them as patients, but they also were
able to consider that when we're thinking about things that
affect health it's not just when they're patients in healthcare,
but it's about living the rest of their lives. And within
this we focus primarily on the cancer control studies.
But there really was a large body of learning that can
inform this field from really four generations of
community-based heart studies. So this is a single slide that
characterizes our observations from this literature.
The first observation is that multilevel interventions are
inherently context-dependent, but context is inadequately
reported in most reports. And this might be because often
these multilevel studies are large studies and they're
published over many papers over time.
But reading this literature you don't get enough of a sense of
context often to think about how you might apply these into
different settings, or how you might think about how the
levels interact. Often less than three levels are reported.
We found a large body of literature that were theory
articles. But when we got to the empirical literature,
often theory was briefly mentioned,
but it was really hard to see how it was used to actually
guide the intervention, or to interpret the findings,
or how the empirical findings actually went back to re-inform
and reinterpret the theory. Across the cancer control
continuum, the vast majority of studies we were able to find
really focused on the prevention and early detection end.
A few were on the survivorship end, but very few on the
diagnosis treatment end of life care of the spectrum.
And one question, is that because these multilevels are
less important for these studies or is that because
those literatures are blind to these other levels that might
be affecting the phenomena that they're studying at those
levels. Reporting of temporal issues.
The idea that often there's a life course perspective on the
people in the study, or that the different levels of
environment to change over the course of the study,
those temporal effects, there was very limited evidence of
reporting of those. And across this literature there were lots
of glimmers of hope, that this is really an important area,
that these multilevel interventions are where we need
to go, as Steve has talked about in setting up this conference.
But it's mostly a hope. And when we look at how we
publish things and how we fund things, often those are really
at odds with what we need to do to really advance these
multilevel approaches. So looking at the opportunities
that we can draw from this literature, we organized what
we found in the literature in three areas. One is in design,
next is analysis, and finally dissemination.
So in the design of studies there is a need for multilevel
studies to just consider contextual factors across these
levels and across time, how these things evolve over as
they interact over time. There is a need for designs
that really involved rapid learning over time,
that are dynamic and adaptive, and really open to
the emergent phenomena that happen when you're
looking at the interaction across multiple levels.
There is a need to specify the interfaces among the levels.
Often these were just alluded to, that they
weren't really specified. And I think of the papers we're
going to hear at this conference, Brian Weiner's
paper that really talks about, gives us a way of explicitly
articulating those interactions, is very important.
And then finally, specifying the levels affecting the
phenomena of the understudy, even when they're not
explicitly focused on. So, just in the theory and the framing
of the question and the interpreting of the results,
to acknowledge that there are these other levels that are
affecting the phenomena is important.
In the analysis across the literature we saw a greater
need to evaluate both process and outcomes.
And some of the exemplary studies that we did didn't just
treat all those intermediate steps as a black box and look
at an intervention effect on a distal outcome,
but really looked at how these different levels interacted
over time, how that process evolved over time.
In order to do that, there's a need to capture measures of
central tendency, but to move beyond that.
And one of the ways to do that is with mixed method designs.
So, studies that do quantitative analyses,
quantitative modeling, at levels where there are
sufficient data and sufficient variability;
but then not just ignoring other levels,
those higher up levels where there might not be sufficient
numbers or sufficient variability for modeling.
That's where a mixed methods approach that really uses
qualitative methods can be really helpful.
And the qualitative methods also can be helpful in understanding
the context for how some of these levels interact.
And then Joe Morrissey is going to tell us more about an
opportunity we saw in the literature for more complex
system and simulation modeling. And then finally in
translation, there is a need for more transparency
in the reporting of contextual factors.
And Elizabeth Yano is going to talk to us more about this.
We saw a need in the literature for moving beyond our
reification of fidelity to the intervention to ways of
thinking about interventions, the locally adapted,
that are reinvented, that evolve over time,
and that include participatory approaches.
Really moving beyond this idea that we're trying to discover
interventions that can be sustainable, to discovering
principles and approaches in intervention opportunities
that we can articulate into ways that can evolve
over time and in different situations and settings.
And then to report that in a way that others can use.
I'm going to close with discussion questions that we
hope you'll consider in your groups.
First there's the question of how do we develop multilevel
interventions that create synergy across these multiple
levels, where the whole is greater than the sum of the
parts, that continually pay attention to context,
and that adapt and learn over time.
And another way of saying this is "how do we develop designs
and analytic techniques that reflect the complexity of the
phenomena we're trying to study and trying to influence?"
And then how do we disseminate to foster
informed adaptation or reinvention of local context
and continued evolution and learning. Thank you.
>>>[APPLAUSE]