NURS3030H NURSING RESEARCH IN PRACTICE MODULE 4 Quantitative research & designs

Janet Rush, RN, PhD, 2010

Objectives:
Review: quantitative methodology Introduce designs where questions are asked about comparisons, associations, risks
Comparing the independent and dependent variables Experimental designs Effectiveness Trials Causation Trials

Quantitative methodology
Emanates from a positivist perspective Has been the predominant biomedical focus Useful in nursing and allied health: programs, interventions, understanding/comparing factors in population health Uses objectivity, logic, experimental/scientific processes to: Compare Establish statistical probabilities Control Define statistical significance Predict Test hypotheses Infer

The research question

Recall PICO and PECO
P = population I or E = intervention or exposure (independent variable) C = comparator group(s) O = the outcome/result (the dependent variable)

PICO for studies of effectiveness

Among (P), what is the effectiveness of (I) versus (C) on (o)?

PECO for studies of causation/risk

What is the risk of (O) among (P) who are/were exposed to (E) versus (C)?

Examples: PICO, PECO

PICO: questions of effectiveness of treatment, programs, interventions
Aim to assess new or existing programs of potential benefit
Among LTC elder women with incontinence, what is the effectiveness of a prompted voiding protocol vs. conventional care on mean, weekly incidence of incontinence over 1 month> Among a rural community, what is the effectiveness of routine PHN visits for firsttime mothers vs. conventional followup on the length of exclusive breastfeeding at 6 months?

PEOC: risk or causation questions

Aim to assess the level or risk or harm

What is the risk of MI among men (40-60 yrs) with Type 2 DM vs. those without Type 2 DM? What is the relative risk of lung cancer among adult women (age 40-70) with a history of smoking vs. those who have never smoked?

Answering quantitative questions

Literature review getting the answer from the existing evidence PICO real time or prospective PECO often retrospective (ethical reasons) Designing studies:
Longitudinal, comparitive/descriptive Retrospective . Prospective

Case/Control

RCT, (quasi) RCT, Cohort analytic

What are the strengths and limitations of retrospective and prospective studies?

Retrospective vs. Prospective

Which design to use - influenced by many factors ($, time, ethics, feasibility) Effectiveness studies best design is prospective randomized controlled trial (aka, experimental), then quasi RCT, then cohort analytic (a prospective study without randomization), then pre/post comparison Causation studies most feasible design is retrospective case control (sometimes called cohort comparison)

Design architexture
Research question
Sample of the population

Intervention/data collection

Group A
Experimental subjects

Group B
Control subjects

Statistical comparisons Results, conclusions, decisions, dissemination

Designs varied, rules for rigor to be followed every step! For strong internal validity of the study (otherwise, bias exists)

Simple designs

Questions re. descriptions,comparisons or correlations

E.g What is the prevalence of Alzheimers disease in Peterborough, ON?

Key considerations: representative sample, accurate outcome assessment/measurement tool/assessors

Sample
Research question (denominator)

Outcome -Yes (the answer) - No

May want to make secondary analyses Think about the factors of interest - Think about other data to collect

Other analyses/observations
Relationships between/among variables Extraneous, confounding variables? Think. determinants of health
Biological

Context .e.g., social

Health

Education

Income

Culture

What will affect The relationships?

Causation
E.g., what is the risk of Alzheimers disease among adults exposed to lower vs. higher income levels?* Find cases and controls those with Alzheimers and those without (Case control design) Gather income data Make comparisons
-Cases (with Alzheimers) -Controls (without Alzheimers)
* From N414 student, A. Clawson

- Low income, yes

- Low income, no

- Low income, yes

- Low income, no

Experimental designs
Can be for PICO or PECO Can be retrospective or prospective Hypothesis testing is undertaken Inferential thinking Inferential statistics
Strength of the associations Finding a difference (vs. the null hypothesis of no difference)

The RCT:Each step has biases and rules

Research question
Study Team Outcome measures Inclusion criteria Approval/Consent Allocation concealment Consecutive subjects Audit because. Follow up

Select the sample & design

R*
Group 1 Intervention Group 2 Control

ANALYSIS
DESCRIBE GROUPS COMPARE GROUPS COMPUTE SIGNIFICANCE

* R = randomization

Decision/Application/Communicate (Knowledge Transfer)

Rules and Bias doing it right SAMPLE

Finding the sample?
Convenience Pilot Random Volunteer Issue of the setting

Similar at baseline?
How do you know? Inclusion/Exclusion Criteria

The issue of Sample Size (SS)

SS is based on Probability theory
Assumption of a random sample

When is SS important?
pilot vs. full trial Rarely use SS calculation for pilot Use for hypothesis testing in inferential stats

Interpreting the articles what to look for?

our SS was based on

Hypothesis testing: need to know

The expected proportion or mean differences between groups Consider the % difference you expect with your intervention & change to proportion ( e.g., 15% is .15)

Sample Size, Website

On-line reference free download

http://biostat.mc.vanderbilt.edu/twiki /bin/view/Main/PowerSampleSize

Once downloaded, the icon will appear on your desktop. When you click it, this screen will come up click continue to start the program

Example
Procedure follow along the screen prompts

For intervention/treatment studies: click dichotomous, independent samples, 2 proportions, uncorrected chisquare test), type in ()alpha .05, () beta .80, p 0 (first proportion), then p1 (the second proportion), then click m = 1 & finally, click calculate

Here, I suggested a 25% change between the groups. That is, usual rate = 50% (.50) & hypothesized rate with a new intervention = 75% (.75)

58 subjects/group

Error
Beta error: rejecting the null hypothesis when it is false (when there IS a difference) & should be rejected about the adequacy of the sample size too small? insufficient power to detect differences between groups if differences do exist (type 2 error). Alpha error: The probability of erroneously concluding there is a difference between groups when there is, in fact, no difference (type 1 error) may be accepting a false positive result usually willing to accept = 0.05

Measurement: Validity / Reliability

Validity Consider an appendicitis scale .
Content validity: scale includes all the aspects, eg, signs, symptoms Construct validity: scale supports or lines up with the theory of the concept. Eg., high WCB with appendicitis Criterion validity: scale has the capacity to diagnose or predict changes., eg., if low score on the scale, then the pt probably does not have appendicitis - - it reflects reality rules in or rules out.

Reliability the extent to which the scales variance is attributed to random error should have sufficient variability but should control for sources of variation should be able to be reproducible, one setting/population to another. Considers test-retest reliability, inter and intra-rater reliability. Measure: Cronbachs alpha (>.8 = acceptable)

Randomization & Allocation

Agreement to be randomized
Issue of informed consent

Random sampling
Computer generated random number table Blocking factor (of 4, 6, 8 etc)

Quasi randomization
Pros and cons?

No randomization?
When to use Cohort analytic trials

Allocation to group(s)
Concept of concealment Not to be confused with blinding

The study groups

Sticking to protocol
Ways of maintaining adherence Involving subjects Auditing

Contamination
Cross-over How many/group is OK? What to do about cross-overs

Co-intervention
Why/when might this happen? What to do?

Follow up
Dropouts, losses to follow up How many/group is OK?

Fitting everything together

Critical appraisal every step!! start with the question getting the right design sampling methods, inclusion/exclusion +++ variables confounders internal validity/data collection & audits tight manoeuvres well explained analysis? decisions from an analysis

Analysis - Using the data Results significant?

P value (<.05) Chi square T-test Relative risk Odds ratio 95% Confidence interval NNT Statistical vs. Clinical Significance Generalizability
Test statistics related to question & data