Decision Making for Rational Drug Use Interventions
Session Guide
SESSION GUIDE DECISION MAKING FOR RATIONAL DRUG USE INTERVENTIONS
Decision Making for Rational Drug Use Interventions
SESSION GUIDE
PURPOSE AND CONTENT
In this session you will be exposed to a range of options and choices that you will need to review. This session will review a process of needs assessment, assessing the present situation in your country, and making plans for future activities.
OBJECTIVES
This session will develop your ability to—
- Review intervention strategies.
- Choose between interventions.
- Develop a plan to undertake an intervention.
PREPARATION
- Read the Session Notes.
- Review the copy of INRUD News, which you have been given.
FURTHER READING
Soumerai SB, McLaughlin TJ, Avorn J. Improving drug prescribing in primary care: A critical analysis of the experimental literature." Milbank Quarterly. 1989; 67(2): 268-317.
Hogerzeil HV, Walker GJA, Sallami AO, Fernando G. Impact of an essential drugs programme on availability and rational use of drugs. Lancet. 1989; (Jan. 21):141-142.
Landgren FT, et al. Changing antibiotic prescribing by educational marketing. Med J Aust. 1988; 149: 595-599.
van de Geest S. Pharmaceuticals in the Third World: the local perspective. Soc Sci Med. 1987; 25(3): 273-276.
Soumerai SB, Ross-Degnan D, Gortmaker S, Avorn J. Withdrawing payment for non-scientific drug therapy. JAMA. 1990; 263: 831-839.
Quick J, Laing R, Ross-Degnan D. Intervention research to promote clinically effective and economically efficient use of pharmaceuticals: The International Network for Rational Use of Drugs. J Clin Epidemiol. 1991: 44(Supp.II):57s-65s.
WEB REFERENCEs
http://www.who.ch/programmes/dap/icium/iciumpage.html
See also the list of ICIUM posters on pp.10-11.
SESSION NOTES
INTRODUCTION
Once a determination has been made that a drug use problem exists in an institution, an area, or a country, action to remedy the problem usually follows. Unfortunately, it is not always clear what intervention will be most effective. To decide which intervention(s) to undertake, preliminary work is required.
First, the drug use problem should be clearly defined. After this, the various motivating factors should be identified and assessed. Then comes the stage of listing interventions. There are usually multiple options for dealing with any specific problem. Once these possibilities have been listed, the difficult task of choosing one or two interventions should occur. When more than one intervention is selected, each should be of a different type (regulatory, managerial, and educational).
When the intervention is undertaken, it is important that there be a control group and that the sample sizes are adequate to detect differences if they exist.
Once the control study has been undertaken, the results should be assessed and follow-up decided. Three outcomes are possible. First, the intervention may be ineffective and should be dropped; second, the intervention might require revision and restudy; third, in a few cases the intervention may be clearly effective, and such interventions can then be translated into national programs.
CHOOSING STRATEGIES TO TEST AND IMPLEMENT
You might find that, in addition to the above techniques, other managerial and regulatory strategies could be considered for your country or program. It is important to choose a small number of strategies likely to succeed, test them on a pilot basis, and then implement the strategy as effectively as possible.
The following factors should be considered in choosing strategies:
• Expected magnitude of impact — If the strategy is successful, what is the likely impact? That is, will it affect only a few drugs, only a few providers, or only save a small amount of money? Or will the impact be great? Obviously, preference goes to strategies likely to have greater impact on priority drug use problems.
• Likelihood of success — All things considered, how likely is success? Will opposition be so great or the task so complex that success is unlikely?
• Unintended effect — What are the unintended effects that might occur? How can these effects, if any, be minimized?
• Political and cultural feasibility — How acceptable is the strategy in the local context? Will political and cultural factors favor development and implementation of the strategy, or will they severely hinder it?
• Technical feasibility — What are the technical requirements of the strategy? Computers? A highly developed information system? How much technical help (people, systems, equipment) will be needed?
• Cost (economic feasibility) — What is the cost, particularly compared to available resources and to the potential benefits of successfully implementing the strategy?
• Potential for donor support — Will donor support be needed? Requested? How likely is it that the donors with whom you work will support the proposed approaches?
If an informal review of possible strategies has reduced the number to relatively few (perhaps two to eight), then a decision matrix could be made using the alternative strategies as one dimension and the above seven selection factors as the other dimension. Whatever strategies are chosen, they should be tested in advance wherever possible and the impact of their implementation should be carefully monitored.
STAGES IN ATTACKING A DRUG USE PROBLEM
Figure 1
FRAMEWORK FOR FORMATIVE AND INTERVENTION STUDIES
In choosing a target, the initial action required is to characterize the drug use situation with a quantitative drug use indicator study. The drug use problem may require clarification, which can be done through follow-up quantitative studies (disease specific or level specific).
Following this stage, it is essential that the motivations of prescribers and the constraints within the system are investigated. This requires qualitative studies.
With the available data, you can then identify the key factors to change. This requires synthesis of data and prioritization of problems. You need to determine if the factors are important, changeable, and feasible.
In choosing the intervention, ensure that they are targeted to identified factors and constraints. Study design should take into account the available resources, both financial and human, and the administrative structure.
Monitor progress of the intervention closely, as unexpected changes may complicate the result of the intervention.
Evaluate the results carefully, obtaining assistance in data analysis if necessary.
Provide feedback on the results of the intervention (positive or negative).
STUDY DESIGNS FOR TESTING INTERVENTIONS
Interventions to improve drug use are designed to bring about changes in behavior. To evaluate whether change has occurred, data must be collected at least twice — before the program begins (baseline) and after it has been conducted (followup). Frequently, results are observed only after an intervention has been completed. This weak postonly study design should never be used, because there is no information about whether behavior truly changed and if the program was the cause.
Evaluations differ in how often they collect data on outcomes. If outcomes are measured once at baseline, and once again after the intervention, a study is said to have a prepost design. If outcomes are measured at a number of points over time — for example, at monthly intervals for twelve months before and after an intervention — the study has a time series design.
With either a prepost or a time series design, the most important feature that sets good studies apart is the use of an appropriate comparison group.
For example, after an educational program to increase the use of ORS by health workers, the percent of diarrhea cases treated with ORS rose from 26 percent to 60 percent. However, use also rose among health workers who did not receive the program, from 29 percent to 54 percent. It was concluded that although health workers with the program used ORS 9 percent more than those without, changes in the supply system and public education were responsible for most of the increases in ORS use.
The comparison group should be as similar as possible to the group receiving the intervention, and their outcomes should be measured in the same way. Ideally, people are put in an intervention or comparison group after they are selected, a process called random assignment. A comparison group can also be selected from a region where the program is not implemented.
A time series intervention looks at activities that occur over time, trying to identify factors which have changed over time. This method is usually descriptive and does not give absolute answers.
A randomized trial starts with a population that is studied over time. One group receives the intervention, while another group remains as a control group. At the end of the study period, both groups are compared to see if there is a difference. Apart from the intervention, each group is treated in the same way.
PRINCIPLES OF GOOD INTERVENTION TESTING
Key principles of good intervention testing include the following:
· Use a relevant comparison group. Wherever possible, randomly assign facilities or prescribers to the intervention and control groups randomly. Collect data on both groups in the same way.
· Measure outcomes at multiple time points. Always measure before and after the intervention. Whenever possible, increase the number of time points. Usually this means measuring before the intervention, one month after the intervention, and six months afterwards. If possible collect enough time points to do a time series analysis.
· Focus on key outcome measures. Identify in advance the key behaviors the intervention aims to change. Develop indicators that can feasibly be used to measure change in this key behavior.
CHOOSING USEFUL OUTCOME MEASURES
Choosing useful outcome measures means considering the following:
· Focus on key behaviors to be changed
· Consider likely substitute behaviors. For example, when anti-diarrheals were banned, use of antiparasitic drugs such as metonidazole and mebendazole increased in Bangladesh.
· Focus on several important outcomes, not all possible changes.
· Choose outcomes that can be clearly defined and reliably measured.
· Measure more than one dimension. For example, measure changes in knowledge, changes in prescribing, and changes in patient knowledge.
USING SAMPLES TO COLLECT DATA
When data are collected to study a drug use problem or to evaluate an intervention, usually some form of sample is chosen in order to save time and effort. Samples can contain many different types of sampling units: geographic locations, health facilities, prescribers, pharmacies, patients, community members, or drug transactions, among others. The unit of analysis must be decided upon before sampling begins. This can be difficult. Frequently prescribers in a facility tend to prescribe like each other. If this occurs, the unit of analysis would need to be the facility, not the individual prescriber. This has implications for sample size selection. The way a sample is chosen can often strongly influence results. The rules for drawing a proper sample are the same whatever the type of sampling unit.
A sample should be typical of the overall group of interest. The best way to ensure this is to follow strictly some process of random selection, examples of which are—
· simple random sampling: selection from all possible units in a random way, for example, using a random number table;
· systematic sampling: all possible units are organized in a list, the total number of units desired in the sample is determined (for example, 25), the list is divided into that number of equal sized blocks, and units are picked starting from a random starting place in the first block according to that interval (in our example, every 25th unit);
· stratified sampling: units are first separated into groups with similar characteristics (for example, geographic areas), then sample units are chosen randomly from each group so that a certain proportion of the sample has the characteristics of the group.
The accuracy of estimates from a sample depends on the sample size. The larger the sample size, the more closely, on average, will a number estimated from the sample resemble the true number (if all units have been sampled). An adequate sample size should be studied to detect significant differences that occur as a result of the intervention. You may need to seek advice on this issue.
If sample units are drawn in clusters, the size of the clusters should be small, and the number of clusters chosen should be large. Members of a group (for example, two patients at a particular health center) tend to be more alike than members from different groups. If only a few groups are selected, each containing many units, the sample can give biased results.
Sometimes people are selected for an intervention because they have extreme values on some measure: for example, because they frequently prescribe a particular drug during a baseline survey. These people will often have scores much closer to average when measured later even without any intervention. When people are selected in this way, they should always be compared to people selected in the same way who received no intervention.
INVOLVING DECISION MAKERS AT DESIGN STAGE
The purpose of an intervention is to eventually change practices and policies. It is far easier to convince decision makers that a change is needed if they have been involved in planning the intervention from the outset. If a policy maker has an opportunity to make an input at the design stage, that person is likely to take ownership of the results. This would help to bring about widespread implementation if the intervention is successful. Decision makers are often busy and are unwilling to sit through long study design meetings. Alternative methods of communicating with them may be necessary. These might include short briefings at different stages of the process, asking the decision maker to choose which problem should be studied first, or asking advice about possible interventions to be tested. Many decision makers have academic training and may welcome the opportunity to be involved if approached in an open way.
PLANNING AN INTERVENTION
When planning an intervention study, it is important to go through a series of steps:
1. Define the problem.
2. Identify the motivations and constraints that affect the problem.
3. List possible interventions that could be undertaken.
4. Choose an intervention or a combination.
5. Decide what sort of study will be used to test the intervention.
6. Define the study and control population.
7. Define how to select a sample and its size.
8. Define the outcome variables to be measured.