Quasiexperimental studies

The term 'quasi-experiment' was first used by Campbell and Stanley(59) to refer to a situation in which the decision about whether an individual does or does not receive the intervention to be evaluated is not under the investigator's control. Random allocation of patients is not made, so selection bias may occur. In other respects, a quasi-experiment aims to apply the logic of randomized controlled trials to the study design, and the researcher tries to reduce this bias by making the study units, in the groups being compared, as alike as possible in terms of the most important characteristics. This approach is known as matching. Characteristics chosen for matching are those expected to influence the outcomes. Therefore, the overriding aim of matching is to reduce the contribution made by the matched variables to the selection bias, although this method is inferior to randomization in that it cannot reduce the selection bias from all other variables.

There are two main approaches to matching. Paired matching consists of selecting individuals for the comparison group (or groups) who have closely similar characteristics to those included in the experimental group, for example in terms of age, gender, and occupation. This form of prestratification will need to be taken into account at the data analysis stage. A less rigorous variant is group matching, which only ensures that there are similar overall proportions of people, for both the experimental and comparison groups, in the various age bands, occupational groups, or other predefined strata used for the variables chosen for matching. (60)

An example of this design is the study by Dean et al.,(6!) who compared a range of outcomes for people (aged from 16 to 65 years) treated for severe acute psychiatric disorders by a community-based service (n = 69) or by a traditional hospital-based service (n = 55) in two separate catchment areas with similar sociodemographic characteristics in Birmingham. The study hypotheses were not clearly stated. Criteria for admission to the study were clinical and were not imposed by the investigators; therefore the patients can be considered fairly representative of those 'requiring 24-hour availability of speciality services' (the definition of serious illness used in this study) treated in the two areas. Patients in both groups were assessed before the study intervention began and again after 1 year.

Post hoc analysis showed that there were no differences between the two groups in terms of age, marital status, previous admissions to hospital, number of compulsory admissions, and ethnic origin. The main results were that there no differences between groups for social or psychiatric status, although both groups improved. There were no differences in objective burden to carers, but the families of patients in the experimental group experienced less distress and were more satisfied with the service. Patients in the experimental group spent significantly fewer days in hospital over the study period and remained in contact with psychiatric services more often.

This study is an example of a reasonable design applied to a comparison of services for two whole catchment areas. Although it cannot exclude, as can a non-randomized design, all possible sources of selection bias, it does illustrate a compromise between research rigour and pragmatic feasibility. Another example of this design is the PRiSM Psychosis Study,(35) which compared two types of community mental health team (a single generic team against a two-team model) in closely matched geographical sectors in South London and followed up patients over a 2-year period. It showed that the outcomes on the majority of variables were similar for the two models, and that both were superior to the baseline hospital-based model of care. Apart from the value of the results of such studies, they can also generate specific hypotheses to be tested in future randomized controlled trials.

In fact the randomized controlled trial paradigm is most suited to short-term single intervention studies, such as psychopharmacological trials, in which 'blindness' can more easily be maintained. It is more often the case that health service research evaluations (i) take place in routine settings, in which control over a critical parameter, such as 'blindness', is more difficult, (ii) need longer term follow-up, and (iii) assess the effects of multiple simultaneous inputs on a range of outcome measures. This is because the specific active ingredients, and their effect size, are not yet known.

Breaking Bulimia

Breaking Bulimia

We have all been there: turning to the refrigerator if feeling lonely or bored or indulging in seconds or thirds if strained. But if you suffer from bulimia, the from time to time urge to overeat is more like an obsession.

Get My Free Ebook

Post a comment