finding and appraising evidence for PT interventions

boolean operators

conjunction words used to connect keywords when implementing search

used to broaden or narrow searches (or, and)

others: parentheses (nesting) for synonyms within same concept (PubMed, GS); asterisks to include variations of same word stem (PEDro, GS); “ “ to search for exact phrases (GS only),

important considerations for GS and PEDro searches

not too many keywords

search terms in PubMed don’t always work in GS and PEDro, but can extrapolate same strategy/terms

consider PEDro for higher level study designs (SR, MA, RCT, CPG) for interventions

consider GS for SRs/MAs for diagnosis and prognosis

common study designs for intervention studies

true experiments/RCTs

quasi-experiments

non-experimental (observation)

true experiment/RCT

random allocation

best for controlling bias; gold standard for establishing causation

excessive control may overshadow clinical relevance

quasi-experiments

non-randomized, but may/may not be controlled

bias may be reduced thru baseline matching in groups (sex, age)

more vulnerable to bias but may be more clinically relevant and common in pragmatic research designs: particularly usefull for effectiveness studies where randomization not possible

non-experimental/observational

most vulnerable to bias

may be clinically relevant and common in pragmatic research approaches

cross sectional studies and case series are least useful in evaluating intervention effectiveness, findings from these studies should be interpreted with a lot of caution

parallel group RCT

participants randomly assigned to either intervention or control group (could be multiple groups) that run simultaneously but independently

groups are monitored in parallel—no switching interventions, peer controlled

most common RCT design

unit of randomization: individual participants

cluster RCT

entire groups/subgroups are randomized to receive intervention/control (parallel)

commonly used in public health research to avoid contamination btwn groups

unit of randomization: group/subgroup/cluster of participants

crossover RCT

ethical delivery of intervention

participants receive intervention and control in sequential periods, decided by randomization w/ washout period in btwn to eliminate carryover effects

design allows participants to serve as own control

unit of randomization: individual participants

washout period is necessary

waitlist RCT

ethical/equitable delivery of intervention

staggered implementation: mostly applicable to community based (cluster) intervention eval that is expected to be beneficial and of low risk

washout not needed, participants are standard controls

types of RCTs based on study aims/comparisons

superiority trial: is new tx better than existing tx/placebo

equivalence trial: is new tx is equally effective as standard tx within a specified margin

non-inferiority trial: new tx is not significantly worse than standard within specified margin

exploratory/pilot trial: smaller scale trials to investigate effects, feasibility, refine methods, identify issues before larger, full scale trial

hybrid trial: combines efficacy testing w/ research on implementation factors (scalability, acceptability)

methods to determine intervention effect

within: repeated measures (pre-post)

within and between: peer controlled: measure each participant within a group, compare overall measurements between groups

within and between: self controlled (crossover)

randomization techniques

single (unrestricted): randomly allocates participants into groups, may not work when sample size is small and group sizes may be uneven

block (restricted): for small sample sizes, ensures equal group size

stratified: ensures speicifc potential confounder (age, sex) equally distributed

• stratification into subgroups before randomization

cluster: pre existing sub groups forming units of randomization

blinding

conceals allocation info so that whoever doesn’t know who is in int/con groups

types: single, double, triple

intention to treat/ITT analysis

outcome data analyzed regardless of how much/how well pt completes intervention

reduces bias by preserving benefits of randomization; presents conservative estimates that reflect real world

dilutes rx effects if high non-compliance

per protocol analysis

aka as treated

only includes those who complied with int

highly controlled

necessary to understand max potential effect (ex. dose response effects), less external validity