Unless certain study designs are used for data collection when the interventions are implemented, the data will not tell us whether the interventions are effective or not. If we use inappropriate study designs, we will not be able to judge whether observed changes in medicine use are due to our intervention or due to some other factor (a confounder). A detailed description of study designs is beyond the scope of this manual. However, all acceptable study designs include some form of control group, where the intervention is not implemented. The effectiveness of an intervention is judged by comparing medicine use in intervention and non-intervention groups. There are three acceptable study designs:
• randomized controlled trial
• before-after study with control group
• time series.
Sample sizes, as described in section 6.3, i.e. number of prescriptions per prescriber or facilitator and the number of facilities, are important in all designs (except in the case of time series - see below).
|
BOX 7.3 RECOMMENDED APPROACHES TO PROMOTING MORE RATIONAL USE OF MEDICINES
• Establish procedures for developing, implementing and updating standard treatment guidelines
• Establish procedures for developing, implementing and updating a formulary list based on standard treatment guidelines or treatments of choice and using only registered drugs
• Establish a DTC and define its responsibilities with regard to monitoring drug use, feeding back drug use data to prescribers, and undertaking other activities to promote quality use of medicines
• Employ adequate numbers of licensed staff in roles suitable to their qualifications
• Provide unbiased independent drug information
• Require staff to attend a regular in-service continuing education, which should be face-to-face, targeted, problem-based, and based on standard treatment guidelines, with the input, if possible, of professional societies, universities and the ministry of health
• Stimulate a group process among health providers and/or consumers to review and apply information about appropriate use of medicines
• Train pharmacists and dispensers, including drug sellers to be active members of the health-care team and to offer useful advice to consumers about health and drugs
• Avoid perverse financial incentives.
Adapted from Laing et al. (2001) and WHO (2002c)
|
|
BOX 7.4 A COMBINED INTERVENTION STRATEGY IN UGANDA
A randomized controlled trial to test the impact of STGs plus training and supervision on rational prescribing was carried out in Uganda. Prescribing quality, as judged by the percentage of prescriptions (Px) conforming to STGs, did not improve when only guidelines were disseminated, but greatly improved if dissemination of the guidelines was accompanied by training and supervision.
|
Group |
% Px conforming to STG |
Control group - no intervention |
24.8 → 29.9 (+5.1%) |
Dissemination of STG |
24.8 → 32.3 (+7.5%) |
STG plus on-site training in therapeutic problems |
24.0 → 52.0 (+28.0%) |
STG plus on-site training in therapeutic problems plus 4 supervisory visits in 6 months |
21.4 → 55.2 (+33.8%) |
Source: Kafuko et al. (1994) |
7.6.1 Randomized controlled trial
This study design is the scientific gold standard. Here the target population (for example patients, health facilities, prescribers) is assigned randomly to receive, or not to receive, the intervention. Data should preferably be collected before and after the intervention from both those who received the intervention and those who did not (control group). Randomization should be performed in a manner which prevents the researchers knowing which treatment subjects will receive (or interventions facilities will receive) as this prevents selection bias (concealed randomization). The assumption is that if the target population is truly chosen randomly, then before the intervention there will be no inherent differences between the intervention and control (non-intervention) groups. Therefore, any observed differences will be due to the intervention and nothing else. This type of study design is often used for clinical trials of drug efficacy, but may also be used for intervention studies to promote rational use of drugs. It is not always possible to conduct a randomized controlled trial for reasons of logistics (for example where the intervention is a regional policy) or ethics (for example non-treatment in the control group).
7.6.2 Non-randomized before-after study with control
In this study design the target population is assigned non-randomly into intervention and control groups and data is collected both before and after the intervention. By taking into account in the analysis any differences between the groups before the intervention, we judge that any differences between the groups observed after the intervention are due to the intervention. The assumption is that any external factors (apart from the intervention) that may have influenced drug use will have influenced both groups equally. This type of study design is often used when evaluating interventions to promote the rational use of drugs in a number of different facilities or areas where a randomized controlled trial is not suitable.
7.6.3 Time series study
This study design involves data collection over a period of time and may or may not include a control group. Ideally data should be collected at least six times before and six times after an intervention, but for pragmatic reasons the minimum number of data collections is often put at four (twice before and twice after the intervention). The assumption is that if observed changes occur in a sustained and consistent manner after an intervention, then they are likely to be due to the intervention. In fact, the baseline trend acts as a ‘control’ for comparison with the post-intervention ‘trend’. This type of study design is used when it is not possible to have a formal control group. In DUE in one hospital, it may be difficult to have a control group since it would be difficult to expose some staff and not others to the intervention. Also, if a DTC is responsible for fewer than 20 health facilities (10 each for intervention and control groups) the number of facilities is insufficient to allow the kind of statistical analysis used in the other study designs.
Chargement en cours...
