Research methods

Question 1

Describe the key features to think about to reduce clinical errors?

Answer 1

Clinical risk management: improve quality of healthcare and maintain safe systems of care

Features of clinical risk management:
-Strategic commitment
-Agreed case definition (what counts as a clinical error?)
-Accountability - clear mechanism for reporting errors (near miss) - blame free culture
-Detection mechanisms for identifying errors
-Integration with corporate governance
-Organisational culture
-Error reduction strategies

Question 2

Describe the approach for a health needs assessment

Answer 2

Definition of health needs assessment: assessing the health needs of x group in x setting over x time

Considerations:
1. Epidemiological (getting the data for which you are assessing the need)
-Description of the problem
-Incidence / prevalence
-Availability of services / interventions
-Possible models of care
-Outcome measures

2. Corporate (is the service doing what it is supposed to do?)
   -assessment of stakeholder perceptions, which includes patients / public groups
3. Comparative
   -comparative study of the services / service models

Question 3

Describe a systematic review and comment on strengths and weaknesses?

Answer 3

Attempt to identify, appraise and synthesize all the empirical evidence that meets pre-specified eligibility criteria to answer a specific research question

Strengths:
-use of explicit method to reduce bias and produce reliable findings
Limitations:
-time consuming to conduct

Question 4

Describe heterogeneity in systematic reviews?

Answer 4

-Studies brought together in a systematic review will differ in method and approach termed heterogeneity
-heterogeneity can be in participants, interventions and outcomes (clinical heterogeneity)
-Assessment can be qualitative and quantitative:
1. assess if results can be combined statistically (not possible if measuring different outcomes)
2. Where possible, meta analyses should investigate heterogeneity
Funnel plot: compares summary measures of effect size against sample size (smaller studies should vary more)

Question 5

Describe when to use a cluster RCT?

Answer 5

-For use when intervention is delivered at the area level e.g. health promotion message
-suitable when there might be contamination of groups
-intervention aimed at the cluster (e.g. reorganisation of services)

ANALYSIS IS AT THE INDIVIDUAL LEVEL (I.E. NOT AGGREGATED BUT THE INTRA CLASS CORRELATION, MEANS LARGER SAMPLE SIZE NEEDED)

The design effect= 1 + (m-1) x p, where p is the ICC (intra cluster correlation) and m the number of individuals in a cluster - used to adjust required sample size for cluster sampling

Question 6

What are the disadvantages of cluster RCTs?

Answer 6

-May be difficult to assess generalisability (generalisable to individuals, clusters or both?)
-requires larger sample size
Statistical issues:
-usual method for sample size (estimating the difference between intervention and control grouping may be used but inflated, allowing for interdependence of outcome data between individuals - usually sample size needs to be about 30% higher

Question 7

Describe the systematic review process

Answer 7

Designed to reduce bias and provide the best evidence

1. Identify clear research question (defined population, exposure, outcome)
2. Develop detailed and comprehensive protocol and search strategy a priori
3. Select studies through 3 stage process (title review, abstract review, full paper review)
4. Combine and interpret data (may include meta analyses to quantify overall effect size)
5. Assess/appraise study quality
6. Extract data from identified / selected studies

Notes:
Maybe refer to cochrane
Not all reviews relate to interventions

I2: measure of heterogeneity, 50% = moderate

Question 8

Describe the key features of a time series study design?

Answer 8

1. Baseline measurement
2. Multiple measurements at different time points (preferable several before and several after)
3. Geographic and temporal controls
4. Many time series studies are ecological / observational conducted as repeated cross-sectional studies at population level

Limitations to time series design:
1. underlying trend
2.Seasonal variation
3.concurrent intervention

Need to think about regression to the mean

Question 9

how do you calculate attributable risk proportion?

Answer 9

(exposed-unexposed)/exposed