Gilbody Simon, Whitty Paula
Academic Unit of Psychiatry and Behavioural Sciences, University of Leeds, Leeds LS2 9LT, UK.
Br J Psychiatry. 2002 Jan;180:13-8. doi: 10.1192/bjp.180.1.13.
There is an ethical imperative to evaluate service and policy initiatives, such as those highlighted in the recent National Service Framework, just as there is to evaluate individual treatments.
To outline the best methods available for evaluating the delivery and organisation of mental health services.
We present a narrative methodological overview, using salient examples from mental health services research.
Cluster randomised studies involve the random allocation of groups of clinicians, clinical teams or hospitals rather than individual patients, and produce the least biased evaluation of mental health policy, organisation or service delivery. Where randomisation is impossible or impractical (often when services or policies are already implemented), then quasi-experimental designs can be used. Such designs have both strengths and many potential flaws.
The gold standard remains the randomised trial, but with due consideration to the unit of randomisation. Use of quasi-experimental designs can be justified in certain circumstances but should be attempted and interpreted with caution.
评估服务和政策举措，如近期《国家服务框架》中强调的那些举措，与评估个体治疗一样，存在伦理上的必要性。
概述评估心理健康服务提供和组织的最佳可用方法。
我们使用心理健康服务研究中的显著例子，给出一个叙述性的方法概述。
整群随机研究涉及对临床医生、临床团队或医院组而非个体患者进行随机分配，并对心理健康政策、组织或服务提供产生偏差最小的评估。在随机化不可能或不切实际的情况下（通常是在服务或政策已经实施时），则可使用准实验设计。此类设计既有优点也有许多潜在缺陷。
金标准仍然是随机试验，但要适当考虑随机化单位。在某些情况下，使用准实验设计是合理的，但应谨慎尝试和解释。
