1. +
An Ethnographic Approach to Studying
Technology Use: Evaluating a Diagnostic Kiosk at
Four California Emergency Departments
Sara L. Ackerman, PhD, MPH
2. + Research at the crossroads…
“Variables-centred (experimental and quasi-experimental)
approaches and ethnographic approaches to the study of
information and communication technologies in healthcare have
developed as distinct research traditions with remarkably little
dialogue between them.”
- Greenhalgh & Swinglehurst, 2011:
3. +
Outline
The kiosk
Ethnographic methods
Evaluation results
Theory
Conclusion
rethinking design, implementation & evaluation of technology
4. +
Improving efficiency in overcrowded
acute care settings
more people use emergency departments for non-life
threatening medical problems
75% of urban EDs are at or above capacity
increasing wait times, especially for most acute conditions
5. +
The UTI diagnostic kiosk
UCSF Urgent Care Clinic
Rapid diagnosis & treatment of UTIs
Women aged 18-64 with history of UTI
Touchscreen and audio
Spanish and English
40-50% eligibility
high patient & clinician satisfaction
6. +
2008-2011: Implementing kiosks
at four emergency departments
- implementation/study plan
- staff resistance
- few patients eligible for kiosk-expedited care
- UTI program never adopted at one site
- Kiosk eventually abandoned at two sites
- Retained at one site (home institution)
8. +
Research team perspective:
human failure
Focus on triage nurses:
ineffective site champions;
“failure of leadership on the nursing side”
“culture of independence”
9. +
Ethnography
Study of people in natural settings
Immersion
Context, relationships, processes
photo credit: http://hilobrow.com/2011/12/16/margaret-mead/
photo credit: http://www.farmpd.com/Farm-Blog/?Tag=user%20research
10. +
Ethnographic methods
Focus
perceptions
practices
knowledge
Activities
direct or participation observation in real-life settings
interviews and focus groups
analysis of texts, video ethnography
Result
“thick” description
11. +
The evaluation
Four hospital EDs:
Two county, two public university-
affiliated
Bay Area and Central Valley (agricultural
region of CA)
35 interviews with ED staff, site
research staff, and research/design
team
20 hours of ED observations
Archival analysis of funding proposals;
research team meeting notes; study-
related manuscripts and publications
12. +
RESULTS:
The kiosks’ journey through four
emergency departments…
13. +
Institutional approval:
“Lots and lots of politics”
Urban safety net hospital
Hospital vs. kiosk model of triage
Politics of research:
favoritism
territory
UTI program denied approval
14. +
Locating the kiosk
Spatial
proximity to registration and triage
easy access and privacy for patients
Social
Medical tool or a patient device?
Socioeconomically & culturally appropriate?
Material
hardware and software breakdowns
unexpected uses
Institutional
ongoing changes to triage systems
15. +
Rendering patients (in)eligible: “It’s
like a winning lottery ticket”
“The kiosk is supposed to be easy and reduce
wait time, like at an airport, but everyone we sent
to the kiosk was ineligible!” (nurse)
Kiosk as detour in otherwise complex, collaborative, time and
space-sensitive process in which mistakes can have dire
consequences.
16. +
Different patient populations
Why were eligibility rates so different between
urgent care clinic and EDs?
More severe reported symptoms: “Our patients are really
sick…”
Patient populations
- health status
- familiarity with/perception of technology
Different contexts different kiosk
17. +
Triaging the kiosk
“UTIs are an urgent, not an emergent, problem.” (nurse)
“When it’s slow in the ED, I think that the nurse wants to just
room the patient. They don’t want to send them back out into
the waiting room [to the kiosk].” (site research coordinator)
18. +
Kiosk as virtual clinician:
“That’s why we’ll always have jobs”
low rates of eligibility for kiosk-expedited care
= kiosk’s diagnostic inferiority
diagnosis is clinical and social practice
Legal/regulatory context of emergency medicine
19. +
Revised implementation strategies
1. Behavior change strategies to increase nurses’ “buy in”
2. New modules: chlamydia screening & reproductive health
3. “Expanded criteria” algorithm
20. +
Engaging theory
Traditional system design
technologies have stable, a priori properties and capabilities that
determine their use and impact
assumed algorithmic connection between plan (kiosk program)
and action (referral/expedited service)
Actor-Network Theory
Rejects technological determinism, i.e.”technology” and “human
work” are not conceived as separate categories with fixed
properties.
Characteristics of humans and technologies emerge through their
interaction
21. +
Success or failure:
contested types of evidence
On-the-ground evidence:
“It would be better if you could say that it will expedite a patient’s
care, but this just isn’t true.” (nurse)
“The juice wasn’t worth the squeeze, as they say.” (ED physician)
Research evidence:
“I’d have to say that it was a research success, because we
basically got out of it exactly what was planned. We made a grant.
We said this is what we were going to do. We did it. We got the
data. It’s published.” (researcher)
published data: women who received kiosk-expedited care spent
less time (40-57 minutes, on average) in the ED than women who
received standard care
22. +
Conclusion
Technology as “rational”, “neutral” tool
or…
Artifacts/systems with inscribed understandings of and goals
for medical work, but whose actual characteristics emerge in
specific social contexts, without fixed, predictable course.
23. +
Recommendations for Future Health
IT Projects
Identify “users”/engage in participatory design
Understand technology implementation as process of
institutional and social change
Negotiate success/failure criteria with all stakeholders
Rethink implementation strategy & evaluation design;
incorporate ethnographic research at all stages
I hope that my talk will contribute to a greater appreciation of the value of ethnographic methods and analysis for implementation research.Trisha Greenhalgh & Deborah Swinglehurst, 2011:This is due in large part to differences in ontology (assumptions about the nature of reality), epistemology (how we can know that reality), methodology (what counts as robust study designs), and axiology (what is of value).”
– promising device falters when transferred from one setting to another
kiosk rationalelower wait times; free up staff and clinician to attend to more severely ill patients
2004walk-in, same day appointment clinicreasons for ineligibility: extreme symptoms; possible complications – kidney infection, etc.Questions about symptoms; printout; prescriptionSatisfaction: reduced wait times for patients; clinicians can see more patients in less time
Research team followed implementation roadmap; buy-in from ED staff/promise that kiosk would improve quality and efficiency of services.
research team knew that implementation hadn’t proceeded as anticipated for multiple reasons, but they didn’t know how these factors were interconnected or what they could have done differently; asked me to evaluate the implementation process during the last two months of the study
one thing was clear up front: triage nurses had agreed to refer all women with UTI symptoms but in reality didn’t refer women at the rate expected: 13-84%; the site specific numbers of women eligible per study quarter ranged from about 30 to close to 200”.“[ED nurses] have very different perceptions and attitudes about whether or not part of their mission is to participate in new knowledge translation and quality improvement…when you do dissemination work and you start spreading this stuff out, it’s going to fail at places that don’t care or want to change their behavior.”though the kiosk was not presented as an optional device to the nurses, they ultimately exercised agency in decided when to use it. Nurses have more decision-making authority in the ED than the researchers had anticipated, and than the registration desk staff at SACC. “In emergency medicine, the doctors are particularly beholden to the nurses, as opposed to other kinds of places like the operating room, where traditionally doctors kind of are the bosses..we’re more of a team place…so the nurses kind of run the show.”
Rather than take this assessment at face value, I wanted perspectives of all stakeholders. So I used an ethnographic approach.where they live and workethnographer strives for immersionGoal is to understand and describe context, relationships, processesUsing ethnog to study tech is premised on the assumption that Technology is best understood within the social, material and institutional context of its useMead: http://hilobrow.com/2011/12/16/margaret-mead/User research: http://www.farmpd.com/Farm-Blog/?Tag=user%20research
what people saywhat people do; behavior is understood as social actionhow knowledge is produced/sustained, including formal and informal: embodied knowledge, tacit rules 4. Activities5. In short, social & material worlds in which people live, work, and invest meaning“Thick” vs. “thin” description: beyond a superficial or reductive understanding
interviews: including contextual, ad hoc interviews with ED staff while on shift
fairness & equity: values instantiated in triage practice; came into conflict with model of triage represented in kiosk (instrumental idea of efficiency)“It was hard to get management to do the UTI module because of ethical reasons…if you have a UTI it’s hard to push you up [the queue] versus somebody that’s homeless that has a broken arm or something like that. Because we deal with a vulnerable population.”“At that time, management was also trying to work on ED flow. So if this module decreases ED wait time considerably then somebody from [outside] gets all the credit…So they would like the solution to come from within?...Right…lots and lots of politics.”Hospital’s response seemed irrational to the research team, since the kiosk had been designed to free up staff time and enable the hospital to serve more patients in less time. However, the kiosk continued to serve as a lightning rod for institutional politics, particularly contested definitions of good patient care and economies of professional credit. After a protracted approval process, the controvery was resolved by the hospital’s decision not to approve the UTI program.
Kiosk was both a clinical and social device, and as such locating it in the ED was problematic.Spatial: Ideal home for kiosk couldn’t be found; multiple movesSocial: ambiguous status as clinical-diagnostic tool designed to be used in spaces not typically designated for medical service delivery: in waiting room, more public qualities were apparent – more prone to vandalism, eavesdropping, unwanted use. triage: small, crowded, less accessible to patients and easily forgotten by nurses“You have some Spanish-speaking person who doesn’t have a computer in her house–maybe goes to an ATM once a week. And the nurse, who’s got this cultural barrier from her, says, “Oh, tienes mal de orine? Okay, vamosalla a estamaquina.” And she’s like, “What?” [Then] she has to make her way through this lobby and all these people are looking at her. I mean, it’s just not going to work well. Material: Intermittent breakdowns interfered with triage, which is a complex, collaborative, time-sensitive sequences of tasks in which mistakes can have dire consequences.Institutional: Finally, ongoing Institutional changes affected how and when the kiosk could be used and its relevanceTherefore, kiosk was not just a clinical intervention, but a material and social one as well, and these properties manifested differently in ED than at urgent care
The kiosk also did not perform as expected. Far fewer patients who completed the kiosk program were found eligible for expedited care than at the urgent care clinic (40-50% vs. 3-20%). Note: the study involved the randomization of patients into a control and treatment group, which further reduced the % of women receiving expedited care). For researchers, eligibility rates were simply a form of kiosk “output”. For nurses, low output was not a mere absence, but was disruptive, even jeopardizing the quality of care. Illustrates how technology operates as an active participant in medical work even when it appears to be passive or dormant; in the midst of complex collaborative work, the kiosk became an unpredictable, even incompetent colleague.
And why did they vary between different EDs?2. “…we rarely have any straightforward case in the ER.”. Socioeconomic differences between sites: at the urgent care clinic, “a woman who comes in with a bladder infection knows she has a bladder infection probably at least half the time, right? In an ED, on the other hand, “there’s never really one complaint.” Kiosk: inflexible diagnostic capability
In effect, the triage itself was submitted to a kind of triage.UTIs are not seen very frequently in most EDs and as a low acuity condition they are not top priority for triage nurses trained to identify and prioritize life-threatening conditions. Since kiosk took time, and was unlikely to perform as promised, it was often bypassed. Nurses’ decision to ignore the kiosk also speaks to their status and authority in the ED, which was underestimated by the physician-led research team.In addition, UTI program design represented triage as linear sequence of tasks that could be easily re-engineered. This proved unrealistic.
Physicians also weighed in. For them, the UTI program was not just seen as algorithm that worked or didn’t work. Rather, the device prompted a re-affirmation of emergency room diagnosis as a social practice that cannot replaced by machines. “We can look at them and do our physical and then decide do we believe that that is kidney pain or is it something else? I think with the computer you have a yes/no answer, then that’s the end of it.” “I think that’s why we’ll always have jobs – the complexity of the decision-making…we don’t even understand how complex it is.Legally-required medical exam for all ED patients shaped clinician expectations and practices related to kiosk. “We’re trained to think the worst-case scenario on a patient.”
incentives, competitions, reminders. Attempt to change user, not technology. Failed to account for how triage routines were altered by kiosk, how the kiosk fit/didn’t fit into conceptions of good emergency medicine, or the process by which nurses made decisions around kiosk use/non-use. “It just decays the moment there’s not someone [nurse champion or research staff] around, and yet this is supposed to be this self-sustaining kiosk that just works…” (ED physician)2. only served to intensify nurses’ resistance to kiosk, both because it was categorized as preventive medicine and outside the scope of EM, and because kiosk became more disruptive of nurse-patient interaction. “They just make a face”3. researchers were aware that low eligibility was affecting use of the kiosk; attempted to change the kiosk rather than the ED; too little too late/couldn’t overcome kiosk’s status as disruptor
Why now? Ethnography tends to be more inductive than deductive. Theory doesn’t guide method, but informs interpretation.Our analysis is informed by actor-network theory, an approach to understanding human-technology relations that arose from science & technology studies1. challenge assumption that a given technology will result pre-defined outcome regardless of context; kiosk designers followed this line of thinking in assuming that kiosk inputs and outputs could be predicted and controlled2. rather, impact of technology is always contingent on social processes in interaction with the properties of the technologies themselves2. Technologies shape humans by making some actions possible, others socially or materially difficult. Humans shape technologies by (re)configuring them & using them in unexpected ways. This approach suggests that the kiosk was an active participant in the contexts where it was adopted, resisted, and abandoned. Next: success/failture
What does this data not account for? How did commitment to data collection interfere with implementation?
Enhancing efficiency and improving care were capabilities thought to be built into the kiosk. Mismatch between what kiosk was expected to be when introduced to ED, and the kind of device it emerged as in relation to its surroundings and human users. I would suggest that the kiosk’s electronic representation of triage and diagnosis was more standardized and formal than actual ED work, which is always flexible and contingent. Moving between the requirements of the kiosk and those of actual triage and diagnosis required clinicians to “articulate the inflexible demands of technology to the practical requirements of the ongoing work” To some extent this was achieved, particularly at the home hospital, where nurses referred women to the kiosk in order to support research, even though they believed that it didn’t work. However, the articulation between tech and human activity broke down in midstwhen nurses and physicians rejected the prescribed set of behaviors that was written into the kiosk for them– a process that was largely tacit and whose causes eluded the research team.
Information systems cannot be designed independently of empirical knowledge of workplaces.Identify “users” and study their individual and team-based practices before designing/disseminating new health IT devices/systemsEnlist users at all phases of design, implementation and evaluation. (Standard practice in high tech industry – UX research)Approach technology implementation as an institutional and social change process, not just a technical project that requires individual acceptance of “buy in” and assessing “fit”Conduct ongoing, real-time qualitative evaluations of technology design and implementation processesUnderstand that “success” and “failure” do not have uniform criteria among all stakeholders; success criteria should be continually negotiated in the context of implementationEval design. We need to engage more explicitly with questions of epistemology; tech interventions in complex social settings may not be best evaluated by experimental research.
Successful technology is web, ensemble of artifacts, skills, applications, infrastructure that constitute technical “systems” that evolve in specific sociopolitical contexts.Mechanisticinstrumental rationalitymanagerialtechnological determinismsocial determinismdichotomousprocedural