Study design | Disadvantages specific to study design | Key difficulty | Outcome |
---|---|---|---|
Recruitment into study by GP followed by randomisation | Self-referral to PAS increases risk of contamination of comparison group | Service is available external to the study | Dilution bias Underestimate of effect |
 | GPs may refer those who they think most in need/most likely to benefit - rather than recruit to the study | Group being evaluated not representative of those using the service |  |
 | This would half the flow of patients being referred to PAS | PAS may not be sustainable |  |
 | Requires high levels of co-operation from GP and PAS | Resource implications for GPs/PAS |  |
Cluster randomisation | Need to identify IB recipients in comparison practices who would be eligible for referral to PAS: it is likely that this would only be around 20% of the total sample | Non-specific criteria for referral to service limits our capacity to identify an appropriate comparison group | Possible selection bias depending on ability to match controls |
 | Cluster level differences need to be accounted for | Requires high levels of collaboration with policy makers well before implementation of pilot | Not possible given that PAS had been rolled out by the time of this evaluation |