Chapter 2: Research Methods

ratio score

Dissociating different cognitive processes can also be done by comparing the performance in two different conditions. The score in a more complex condition is corrected for the score in a more simple condition. A ratio score is calculated, indicating how the more complex condition relates to the simple condition

ecological validity solutions

Sometimes, gap between test and everyday behaviour (ecological validity) - sometimes no impairments are found during a test despite presence of limitations in everyday life

-> questionnaires and observation lists (no standardisation)

-> vr and ar (ecologically valid, standardised and safe :D)

outcome measures

- important that they're ecologically valid

• Can be classified at different levels: level of functions, level of activities, level of participation; which level is most appropriate depends on the RQ

remote outcome measurement

= questionnaires, online tasks

Good: participants need not travel, national/international inclusion, data collection at different specific moments - higher reliability

Bad: no control over circumstances, don't know whether instructions were understood, can only make limited behavioural observations, sometimes people with impaired cognitive function can't, selection bias (computer literacy, resources)

drop-out solutions

• Per protocol analysis: those who do not complete the study are omitted from the analysis

• Intention-to-treat analysis: those who do not complete the study are included in the analysis - preferable, but difficult to interpret if too much data is missing

• (alternative: analyse characteristics of dropouts to infer whether dropout was related to certain patient characteristics)

crossover design

= type of RCT where participants receive multiple interventions

• AB/BA design - first treatment A, wash-out period, then treatment B (or vice versa)

• controls for differences between participants, fewer participants needed

• Investigates the specificity of interventions (whether they affect a single function)

• Only suitable for stable conditions (chronic diseases)

• Hard to determine wash-out period

• Duration is longer, statistical analysis is more complex

multiple baseline design

more stable baseline, improvements can be more unambiguously attributed to treatment

• Helpful when impossible to include a control group (rare condition) or its unethical to withhold treatment

• Can be used to find optimal moment for intervention

• Spontaneous recovery is taken into account, placebo/test-retest effects aren't, need a control

longitudinal design

Cohort study: specific group of people (cohort) is followed for a longer period of time; test-retest effects

cross-sectional design

measurements made at a single point in time (prevalence, gender/age groups etc), no test-retest effects; only correlational

case control study

group of people with a certain condition are compared to controls without that condition

• How often certain risk factors occur in both groups - relationship between risk factors and the condition

• Must choose proper control

• Descriptive, no variables are manipulated

• Allows investigation of rare disorders, multiple risk factors can be investigated simultaneously, takes little time

case study

one person (single-case-study) or several people (case-series), performance is compared to an existing norm group or a new control group.

• Not generalisable, info about unique individuals and groups -> give meaning to existing data/lead to new generalised hypotheses

dissociation

selective loss of cognitive functioning - can show that subprocesses are independent of each other because they are affected separately

• Single dissociation: impaired on task A but not task B - does not imply that task A and B are independent of each other; dissociation can also be due to the hierarchy of functions.

• Double dissociation: patient 1 is impaired on task A but not task B, patient 2 is impaired on task B but not task A - proves two independent cognitive processes

• The differences between the performances must differ significantly from each other

heterogeneity

differing from the population mean?

• Group studies study population mean, desirable to opt for a homogenous group;

• effectiveness of a treatment, match sample with the target group, often a heterogenous group

• Important in case series - insight into why two similar patients don't both develop a disorder

open science

science must be transparent and accessible

• Reproducibility: whether another researcher with the same dataset would get the same results

• Replicability: whether another dataset will provide the same results

 

PICO method for finding relevant research: Population, Intervention, Control, Outcome