Random Notes

Theories

By definition Theories obscure some phenomena and exalt a few others. This is just how it is. So the correct choice of theory is of a lot of importance!

Also: words used in theories do not map to real world things. “Decision-making” is different for example: in some theory, this means that there was some optimal decision to begin with; i.e. it prioritizes a normative view. Another is “sensemaking”. This is not what you think it is. There’s a rigorous definition in cognitive science. You need to be aware of this.

Frameworks and theories: there are cases where either is a subset of either.

“Unanticipated consequences” is an example of concept synthesis. This was used to describe the negative impact of EHR on clinical practice.

Don’t just start with theories. Start with a study. Understand people and the data before you start theorizing.

Some Motivation Theories (“Why don’t people come in to work more often?”)

• Bandura
• Self-Motivation Theory

Hutchins and cockpits: Pilots converted from Cognitive → Perceptual Tasks. Led to theory of Distributed Cognition: Cognition doesn’t simply exist in someone’s mind; they offload cognition to objects in the world around them. (Person + Objects in World) = Cognitive system.

TODO:

• Constructs, Concepts, Variables, Operationalization
• Propagation of representations (think of how patient data flows through a system; how is it represented?)
• “Cognition in a Vial” - Hutchins

Activity Theory

If you saw some tribesman beating a bush in isolation that would seem like a pointless activity. But if you zoomed out and saw that he’s helping his mates hunt wild boar? Aha! In AT, the unit of analysis is an activity where you focus on the structure of the activity: who are the subjects, objects, and community/context? What are the relationships?

Situation Awareness

S.Awareness(Perceive → Comprehend → Project) → Decide → Perform

You want to prevent people from getting hit in the head at the stadium? What’s an informatics intervention to design? This theory can help you break down the problem.

Effs’es

Efficacy and Effectivness are two different but related things. The former deals with intervention outcomes under some ‘ideal’ or controlled conditions (think RCTs). The latter is when the intervention is deployed into the world. The gap between the two is a well-known thing.

The efficacy-effectiveness (EE) gap describes the differences in survival seen in clinical trials and routine clinical practice, where patients in real-world practice often have inferior outcomes compared to trial populations.

Sidenote: Experimental studies allow us to deal with confounders better than observational studies.

Funding

• DARPA and DOD: not risk averse. They expect a 5% success rate!
• NIH: require big teams that cover all competencies. Significance, innovation, scientific approach. Pretty risk averse.
• NSF: They value transformative research. More risk taking. Emphasis on students and training.
• AHRQ: Focus on comparative effectiveness, quality and patient safety.
• PCORI: Patient-centered outcomes research.

Two types of applications: You can apply unsolicited, not in response to any priority or call. NSF works almost exclusively like this. FOudnations are different. They have a mandate and strategy and issue calls for proposals.

Intramural support gives you the smallest amount of money! $20-40K. Faster process tho. Often pilot studies.

NIH is the largest funder of research. 21 institutes (4 may be nixed). Emphasis on specific diseases. NLM funds most the innovation in informatics (and runs PubMed).

• R21 (Small R01) - Exploratory research
• R01
• RFAs are Requests for Applications. There are phases. E.g. precision nutrition.
• SF424 R&R (Research and Related)
  • Split intp Specific Aims, Research Straegy (Significance, Innovation, Approach), ???

NIH: High Risk and High Reward stuff. They like this. Preliminary data helps with feasibility: THIS IS YOUR STUFF! Related Work is something other people have done.

Significance is about logical flow.

Aims can be split into Primary, Secondary, Exploratory.

AS a reviewer

From the perspective of the reviewer: there’s now a “Simplified Review Framework”. Smush significance and innovation.

You win points on significance and innovation you lose them on approach.

Session 5 - Qualitative Research I

Chicago School of Social Sciences.
Read papers

Design is a consideration! Remember that EHR was initially meant to support billing and not decision support lol.

Historically, people would just armchair it when it came to studying people.

Identifying stakeholders: can you start with who is everyone who touches a system or idea or question of interest? Layer One. Layer Two+ (transitive)?

Think very carefully and learn to delineate between users and stakeholders!

Think very carefully about permission versus buy-in! These are two different things! You can have permission to conduct a study of nurses but you may not have buy-in from the actual nurses. You can start by asking nursing management and not the attending.

Important differences in qualitative methods

• Observation: You see but you cannot ask why. You don’t get motivations.
• Interviews: Ask for motivations, attitudes, perspectives. Hard to interview about implicit practices (how to ride a bicycle, what do you do before you go to bed, ask nurses how they care for a patient).
• Hybrids

Conduct, Analyze, Scale: These things are hard to do with qualitative studies. You typically use hybrid approaches.

People do different things at different times! Understand this.

People are different (duh): Residency is a job interview. They are used to being interrogated all the time and doing this in large groups. Nurses are not used to this. Things to keep in mind.

“The Grand Tour Question”

Learn to ask questions. “How do you use EHR?” is a shit question. “Tell me about how you ordered labs yesterday” better. Summarization is your job.

Analyzing Results

Sources of Error

Random Error: Measures would be distorted in either direction because of stuff you cannot help (“The World is a Random Place”). This will affect precision. Remember that Precision is PPV (Positive Predictive Value). Given the outcome, how well did your model/approach do with identifying the truth? $P(Y = 1 | \hat{Y} = 1)$

One way to get around this: You could just increase the sample size 🤷‍♂️. So things (confounders, noise) cancel out. But this is not always feasible.

Systematic Error: This affects accuracy.

Postmodernists

• Disillusionment with modernist thinking Not to be redundant, but it’s worth stating again. Postmodernists are unsettled by the modernist’s inability to make strides in achieving peace and advancement in society, so they challenge the conventional way of operating.
• Opposition to traditional authority Overall, authority is dangerous and not to be trusted. Authority figures are to be opposed because of their loyalty to the establishment and rigid beliefs in moral truths.
• Truth is relative There are no objective truths. They have been defined by people and groups who use them to attain power. One person’s perception of reality doesn’t always match another person’s perception of reality. For example, although you may view a person of the opposite sex as in a relationship based on their seemingly conventional practices that indicate they’re taken, they may not see themselves as in a relationship.
• Facts are worthless Facts can change from day-to-day or they may be outright lies. This comes with believing that truth is subjective and it’s something that shouldn’t be widely acknowledged.
• Rationalization Opinions are what matter in postmodern thinking; after all, facts are ignored. As long as the opinion is rationalized, it can be accepted as the best explanation. This means that science is rejected because there is no objectivity.
• Morality is relative There isn’t a moral system that’s right for everybody. Traditional beliefs that have been accepted for centuries fall by the wayside when people accept that truth is relative. This is where many Christians clash with postmodern thinkers.
• Each religion is legitimate Because there’s no objective truth, postmodernists believe there isn’t one religion that’s “right.” Instead, all of them are legitimate. If you subscribe to a faith based on personal experiences and it’s compatible to you, then it’s accepted by postmodernists.
• Belief in internationalism Postmodernists reject the idea of nationalism because it drives nations into conflict with one another, deterring healthy human progress. Internationalism ensures that we’re looking out for the greater good, uniting as opposed to dividing.
• Collective ownership Staying true to the belief that the greater good must be considered, postmodernists protest the idea of individual ownership. It would be most fair if we divided and administered goods together. For the most part, individualism isn’t a characteristic they value.
• Equality Postmodernists hold many socially liberal views, believing in equality for all in spite of race, sexual preference, class or creed. Of course, they believe no one should impose their views of right and wrong on others, so there isn’t one right way to live.