Chapter 22: Generating Evidence Through Quantitative and Qualitative Research

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement, not replace, the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

You know, there is this, um,

this really comforting illusion when you first start studying clinical practice.

Oh, totally.

You kind of assume everything is just already mapped out for you.

Right.

Like you're following this complex, but you know, super reliable recipe.

You have a patient presenting with a specific condition.

You look up the established protocol and bam, you just execute the step.

Yeah.

Feel safe.

It feels certain.

Exactly.

But that certainty, I mean, it relies on this massive invisible foundation of existing knowledge.

We just blindly trust the protocols because we assume, well, we assume they've been battle tested over decades by all these unseen researchers.

But then you actually step onto the floor as a future clinician and suddenly you just hit a wall.

A very solid wall.

Yeah.

You encounter this totally unique patient scenario or like a bizarre systems issue where that recipe book is just entirely blank.

Right.

We're talking about a clinical landscape that is sometimes just severely lacking in evidence.

It's the absolute definition of clinical uncertainty.

And when you hit that wall, you can't just shrug and rely on old unit traditions, right?

Or like a single provider's personal opinion.

Definitely not.

You have to actively generate your own external evidence to figure out what actually works.

And that is exactly what we are getting into today.

Welcome to this deep dive.

Today, our mission is to completely demystify Chapter 22 of evidence -based practice in nursing and healthcare.

We are stripping down the complexities of EBP.

Yes.

Specifically, how you, the learner, move from a raw clinical observation to actually generating bulletproof evidence.

We are exploring the entire life cycle of solving a medical mystery.

From designing a rigorous quantitative study to finding the right patients without skewing your data.

All the way to crossing over into the really rich, human -centered world of qualitative research.

Exactly.

We're going to unpack how you structure an actual answerable question, why certain experimental designs inevitably fail in the real world, and what you have to do when numbers simply aren't enough to capture a patient's reality.

And for those of you listening who are stepping into your roles in nursing or the health sciences,

I want to emphasize right up front.

You do not have to become a lone wolf scientist.

Oh, please don't.

Right.

When you spot a gap in care on the floor, you don't retreat to a lab by yourself.

You form a transdisciplinary team.

That is so key.

You bring the sharp clinical insight from the bedside and you partner with Ph .D.

prepared researchers.

Right.

The folks who bring the methodological muscle.

And they culminate all of this in a big research design meeting to hammer out the study outline.

And that partnership is absolutely vital because, well, the very first hurdle your team has to clear is identifying the so -what factor.

The so -what factor.

I love that term.

It's shockingly easy to get sidetracked by a research question that is, you know, merely interesting but ultimately kind of useless.

Like wondering if playing classical music at the nursing station lowers the stress levels of the unit secretaries.

Right.

Exactly.

I mean, it might be mildly interesting to you personally.

But does it fundamentally change health care?

Probably not.

No.

A question with the so -what factor has to have high impact potential.

We are talking about moving the needle on things the health care system desperately cares about.

Like reducing patient complications.

Yes.

Or lowering readmission rates, decreasing the length of stay, or, you know, cutting massive costs.

Which makes perfect sense because if you're asking for funding or even just permission to run a study on a busy unit, no one is going to green light a project that doesn't actually improve real -world outcomes.

Exactly.

So to filter out the weak ideas, researchers rely on a stress test called the FINER acronym.

FINER.

Let's break that down from the text.

It stands for Feasible, Interesting, Novel, Ethical, and Relevant.

Let's look at the mechanics of that.

Feasible isn't just about having a good idea.

Right.

It's about reality.

Yeah.

It's about whether you actually have the budget, the time, and the access to the patient population to actually pull it off.

And novel means you aren't just duplicating a study that was definitively answered like ten years ago.

Right.

And relevant ties right back into that so -what factor.

It must impact clinical policy or practice.

Okay.

Wait.

Let me push back on something here.

Let's say I'm on the floor.

I see a patient suffering from a unique complication, and I immediately come up with a finer question that absolutely passes the so -what test.

Okay.

Isn't it my duty to jump straight into testing a solution?

Like why shouldn't I just start experimenting right then and there to fix it?

Well, because jumping straight to testing an intervention without understanding the underlying problem is incredibly dangerous.

Really?

Oh, absolutely.

You have to respect the logical progression of research, which is laid out perfectly in Figure 22 .1 of the text.

Right.

The progression chart.

If a clinical phenomenon isn't well understood yet, you absolutely must start with descriptive research.

You just, you know, observe, measure, and document the baseline reality.

So you literally cannot test a cure for a problem you haven't even clearly described yet.

Precisely.

Once you have that descriptive baseline data, then you move to predictive research.

That's where you look at how variables relate to one another over time.

Okay.

Making connections.

Yeah.

And only after you have those solid building blocks can you ethically and logically design experimental research to actually test a specific intervention.

Because if you skip those early steps, I guess you might test a drug or a therapy that targets completely the wrong mechanism.

Exactly.

You'd be shooting in the dark.

Okay.

That mechanism piece is crucial, which brings us to the 15 specific steps for designing a quantitative study.

Right.

Step one is formulating the question.

Step two is establishing significance.

And step three is synthesizing prior evidence.

But before you even think about handing a patient a new medication, you hit step four.

You need a theoretical or conceptual framework.

You do.

Because interventions do not work by magic.

You need an underlying theory explaining exactly why your proposed solution should create the change you're hoping for.

It shows how the variables relate.

I want to use a specific example from the text to make this concrete for the listeners.

Yeah.

It's the flow chart for the COPE model creating opportunities for personal empowerment.

That's a great example.

It's an intervention designed to help kids and teens.

The intervention itself is based on cognitive behavioral therapy, or CBT.

Right.

But the CBT doesn't just magically fix the patient's outcomes, it acts as a catalyst.

Exactly.

The flow chart in figure 22 .2 shows this perfectly.

The CBT leads to a shift in the patient's knowledge and cognitive beliefs.

Okay.

That internal psychological shift is the critical mechanism of action.

In research terms, we call that the mediating variable.

The mediating variable.

It's like the bridge between the therapy you provide and the final result you want to see.

Without that bridge, the intervention fails.

The framework maps out that once the cognitive beliefs change, that naturally leads directly to the specific patient outcomes.

Things like decreased depression, lower anxiety, a drop in PMI.

Yes, alongside an increase in healthy lifestyle behaviors, self -esteem, and even academic performance.

And having this map, this theoretical foundation, naturally leads right into step five, which is generating hypotheses.

You aren't just guessing anymore.

No, you're making a very predictive statement.

Right.

Like, if I introduce this independent variable, the CBT, it will positively affect the dependent variables, the patient outcomes, because of this specific psychological shift.

And once your hypothesis is locked in, you move to step six, selecting the research design.

You have to design a study strong enough to actually prove it.

So if your goal is to definitively prove cause and effect,

to say without a shadow of a doubt that your intervention caused the healing,

you need the absolute gold standard of quantitative research.

Randomized controlled trial, the RCT.

Let's look under the hood of an RCT.

What separates it from just, like, a regular true experiment?

Well, there are three non -negotiable rules for an RCT.

First, you need an experimental group that actually receives the intervention.

Right.

Second, you need a control or comparison group.

And that group receives either standard care or an attention control placebo.

Makes sense.

And third, and honestly, this is the anchor of the whole design, you must use random assignment to ensure baseline equality.

Flipping a coin, essentially.

Or using a computer algorithm to place subjects into groups.

But why is randomization so sacred?

Like, why not just let patients choose which group they want to be in?

Because human choice introduces bias.

Randomization is the single strongest method to ensure your two groups are perfectly equal at baseline.

Oh, I see.

If you let people choose, maybe all the highly motivated, health -conscious patients pick the intervention group and the less motivated ones end up in the control group.

And then if the intervention group gets better, you have no idea if it was your brilliant therapy or just because they were more motivated to begin with.

Exactly.

Randomization neutralizes those hidden variables.

If the groups are completely equal at the start, and the only difference between them is the intervention you provide, then you can confidently isolate the intervention as the sole cause of any changes.

That makes total sense.

Now, sometimes you want to test more than one thing at a time.

The text describes something called a factorial design in figure 22 .10.

I want to use an analogy for this because it perfectly illustrates the mechanism.

Imagine you are trying to bake the perfect cake, and you want to test two variables, oven temperature and the amount of sugar.

Well, you wouldn't just bake one cake.

Right.

You'd need a grid, like a two -by -two grid.

You bake a high -temp, high -sugar cake, a high -temp, low -sugar cake, a low -temp, high -sugar cake, and a low -temp, low -sugar cake.

And this lets you see not only how temperature acts alone and how sugar acts alone, but how they actually interact together synergistically.

Exactly.

And the text maps this two -by -two grid directly to a clinical example for hypertension.

Right.

Group A gets a new education program, A and D, an exercise regimen.

Group B gets the education only.

Group C gets the exercise only.

And Group D, the pure control, gets neither.

This design is incredibly efficient because it lets you test the combined effects versus the separate effects in a single study.

Okay, but here's where it gets really interesting.

Let me throw a wrench in this perfect randomized world.

What if it is completely unethical or physically impossible to randomly assign people?

Say we want to study the long -term effects of vaping on teenage lung capacity.

You can't legally or ethically take a group of healthy high schoolers and randomly assign half of them to a five -year smoking group.

You absolutely cannot.

And this is the messy reality of clinical research.

Sometimes the gold standard is locked behind an ethical door.

So what do you do?

When you can't randomize, you have to abandon the true experiment and step down to a quasi -experimental or non -experimental design.

How does a quasi -experiment work if you don't have random assignment?

Well, you still introduce an intervention, but you lack random assignment, so you rely on other ways to measure its impact.

A classic approach is the interrupted time series design.

Okay.

Instead of control group, you use time as your baseline.

Say you track hospital -acquired infection rates on a specific ward for two years.

Then you introduce a massive new hand hygiene intervention.

You then continue tracking the infection rates for another two years.

So the long history of data before the intervention acts almost like the control.

You're looking to see if your intervention literally interrupted the established trend.

Precisely.

But in your vaping example, we can't even introduce the intervention.

We can't force teenagers to vape.

That requires a non -experimental design, specifically a correlational study or a retrospective case

You find teenagers who are already vaping of their own free will, and you observe how their lung capacity varies with their habit compared to non -vapors.

But because we didn't control the environment, and we didn't randomize, we lose the ability to strictly claim causation.

We can only say that the two variables cover.

Correct.

It's a trade -off.

We sacrifice the absolute certainty of an RCT to answer a question that would otherwise be impossible to study ethically.

Which brings us to another trade -off.

Pragmatic trials.

Even if you can run a perfect RCT, it's usually done in a highly sanitized, optimal, heavily funded environment.

Pragmatic trials are totally different.

They test how well an intervention actually works in the messy, chaotic routine conditions of everyday clinical practice.

It's the difference between a pristine laboratory and a busy ER floor at two in the morning.

That distinction perfectly highlights the difference between internal and external validity, moving into step seven to ten.

Yes.

Internal validity asks, did the intervention actually cause the change in our specific study or was our data skewed by a flaw in the design?

While external validity asks, can we take these findings and generalize them to the theoretical population out in the messy real world?

Let's talk about the things that destroy internal validity.

The text lists some really memorable threats.

One of them is history.

Right.

Let's say you're running a year -long study testing a violence prevention or mindfulness intervention to reduce anxiety in high school students.

Six months in, a tragic school shooting makes national news.

Suddenly, the anxiety data for every single student in your study skyrockets.

Your intervention didn't necessarily fail, but an external historical event contaminated your data so severely that you can no longer measure the intervention's true effect.

Or the threat of maturation.

If you're testing a cognitive development program on six -month -old infants, those infants naturally accelerate in cognitive development over time anyway.

They mature regardless of your program.

That is exactly why a control group is necessary.

The control group matures at the exact same natural rate, so any difference beyond that is attributable to your intervention.

But the most insidious threat to internal validity.

Oh, attrition.

Yes, patients dropping out.

It ruins more studies than almost anything else, and they rarely drop out randomly.

If your intervention is a brutal, high -intensity workout program, the people who find it too hard are going to quit.

If half your experimental group drops out, the only people left are the highly motivated success stories.

Your final data looks artificially successful, but your internal validity is completely destroyed.

Exactly.

So, assuming your internal validity survives, how do we build a sample to protect external validity?

This is step seven, sampling strategies.

Right.

If you just survey your friends, your results mean nothing.

Probability methods are the best defense.

Random sampling gives everyone an equal chance.

But what if you randomly select 100 nurses, and by pure chance, 95 are female and five are male?

That is when you use stratified sampling.

You divide the population by a key variable, first like biological sex, and then you randomly sample within those divided groups.

Or if you need a massive nationwide sample,

you use cluster sampling.

Randomly selecting whole hospitals, and then sampling the patients inside those clusters.

Okay, but what if your target population is actively hiding?

Hidden populations are a massive challenge.

Figure 22 .16 in the text details an incredible mechanism for this called respondent -driven sampling, or RDS.

It's like mapping out an underground network.

You find one initial person, a seed, maybe someone you've built deep trust with at a local needle exchange program.

But you don't ask them for names of their peers, which would violate their trust.

Instead, you give them unique trackable vouchers to recruit their peers.

That seed recruits friends, creating wave one.

Those friends recruit more friends, creating wave two.

It's a highly effective variation of snowball sampling.

And regardless of who you're sampling, you have to determine your measures, making sure they are reliable and valid, yielding categorical, ordinal, or interval ratio data.

That's step eight.

And before collecting data in step nine, you hit step 10.

You must face the IRB, the Institutional Review Board.

You cannot gloss over this.

The IRB evaluates your study against the three pillars of the Belmont report that every student needs to memorize.

Beneficence, meaning do no harm, respect for human dignity, which mandates voluntary consent, and justice, demanding fair treatment.

You can't just test risky interventions on marginalized populations simply because they're easily accessible.

OK, so we spent all this time talking about control groups, random assignment, measuring variables.

But as anyone who has ever spent five minutes with a patient knows, numbers simply aren't enough to capture the human condition.

Exactly, which requires a profound shift in perspective.

What does this all mean when we cross over into qualitative research?

In this realm, you throw out the numerical data.

You throw out the rigid hypotheses to test.

The entire qualitative philosophy here is to explore the lived experience from the subjective perspective of the patient.

It fundamentally flips the scientific method on its head.

In quantitative research, we use deductive reasoning.

But qualitative research uses inductive reasoning.

Yes, you start at the very bottom.

You move from specific personal quotes from individual patients, and you slowly build up to general themes.

And the researcher isn't some distant objective observer.

No, the researcher is actually the instrument of data collection.

Your empathy, your listening skills, that is your measurement tool.

Let's look at the four specific traditions from table 22 .2.

If I want to understand what it is viscerally like to receive a terminal cancer diagnosis, I use phenomenology.

Rooted in the philosophy of Husserl and Heidegger, it's the deep study of the lived experience.

But what if the problem is cultural?

Say you're a nurse in a rural clinic, and the community refuses your protocols.

You need ethnography.

Drawn from anthropology, it requires prolonged engagement in that culture to understand their specific beliefs.

And if you need to build a theory from scratch about a social process.

Grounded theory, developed by Glazer and Strauss, you let a theory emerge organically from the ground up.

The last one is participatory action research, or PAR, rooted in Lewin.

This is about collaborating with the community for social change, not just observing them.

Because the goals are so personal, qualitative sampling is usually purposive.

You handpick informants.

But when do you stop interviewing?

You keep going until you reach saturation, meaning you're hearing the exact same things over and over.

No new insights are emerging.

And because we aren't using statistics, we don't use the quantitative term validity.

Right.

Qualitative researchers replace that with the concept of trustworthiness established by Lincoln and Guba.

Outlining four criteria.

Credibility, dependability, confirmability, and transferability.

It is a beautiful complement to the rigid numbers of quantitative research.

The numbers tell you if the intervention works, but the qualitative stories tell you why the patient cares.

And that brings the entire EBP loop full circle.

We've summarized steps 11 through 15, implementing the study, analyzing the data, interpreting results, disseminating findings, and ultimately incorporating that evidence into practice.

The text's overarching flow is complete.

Sound methods lead to good evidence, and good evidence changes patient lives.

Before we go, I want to leave you with a final provocative thought.

If qualitative research relies entirely on the researcher as a deeply empathetic instrument, and quantitative research relies on strict objective control,

how might the integration of artificial intelligence and big data challenge or support these traditional ways of generating evidence in your future nursing career?

Wow.

Will AI spot qualitative themes humans can't see, or will empathy keep us at the center?

It's something you'll have to ponder as you step into this rapidly evolving landscape.

You've worked hard to master this material.

Thank you for listening from the Last Minute Lecture Team.