|
Abstract
|
The current and loadable trend in the practice of medicine is reliance on sound empirical evidence as a prerequisite for the formulation of treatment protocols. Evidence-based medicine, however, is as much a requirement for prevention as it is for treatment for, at least, two reasons: (1) prevention is applicable to the totality of the population, whereas any particular treatment concerns a certain sector of this population; and (2) the cost/eectiveness ratio of the various preventive measures depends on a multitude of variables, the values of which have powerful sociodemographic and cultural determinants (for example, smoke detectors are more cost-eective where wood rather than stones or bricks are used for house construction).
Paraphrasing from Miettinens commentary The modern scientic physician [1], it may be more appropriate in the area of injury prevention to use the term scientically based public health interventions rather than evidence-based interventions, since logic is as powerful a component of the science of prevention as empirical evidence is. For instance, several measures of passive prevention in the automobile industry rely on laws of physics and deductive reasoning rather than on formal empirical research in humans.Primary prevention of injuries relies on identication of causes, whereas secondary prevention of injuries presupposes some understanding of the pathophysiologic and mechanistic considerations of the induction of injury. Identication of causes and pathogenetic mechanisms may be accomplished through various processes of logical argumentation, through experiments in articial non-human settings or through human studies, which can be either of informal clinical nature or formal epidemiological design.
Knowledge of causes, however, cannot be easily and directly translated into potential for effective prevention. As in other elds of prevention, theoretical effectiveness should be distinguished from population (community) eectiveness and it is the latter that matters for public health professionals [2, 3].There are several rings in the chain between theoretical and population eectiveness and the chain can be broken at any one of them.
To take a simple example, the eectiveness of seatbelts in preventing serious injuries has been known for more than five decades. Nevertheless, their theoretical eectivenesswas compromised either because individuals did not appreciate their potential, because seatbelts were not available in all cars or because the states did not have or were not willing to enforce respective legislation.
Whereas logic and laws of physics may be suficient for estimating theoretical eectiveness only studies in humans can reliably ascertain population eectiveness of injury prevention measures.Evaluation of population eectiveness can focus either on intermediate (process) or outcome measures. Thus, a campaign to increase helmet use aims at reducing head injuries but it can be evaluated through ascertainment of the prevalence of helmet users before and after the intervention, under the very reasonable assumption that after the crash a helmet will reduce the likelihood of a head injury [4].
Focusing on process measures greatly increases statistical power because the intermediate state is far more common than the nal outcome and the relevant latency is also far shorter. Implicit, however, is the realization that it would be extremely dicult to document through observational studies an actual reduction of serious head injuries through helmet use.
Why has empirical research, including randomized control trials, been generally unable to document substantial reductions in preventing injuries following community intervention programs [5]? This is because injuries represent a wide range of etiological and manifestational entities, each of which is responsible for a small fraction of the whole spectrum. Injury prevention interventions focus on no more than a handful of practices and only a couple of them can be realistically expected to be eective in any particular setting. It is an unusual situation when the prevalence of the targeted exposure is suciently high or the relative risk for an injury associated with this exposure is considerably elevated that the effectiveness of the intervention can be documented with sucient statistical power. And the eectiveness of the intervention itself may depend on a number of factors (intensity, duration, repetition, etc), which may or may not be optimally reached in any particular setting.Focusing on high-risk group interventions, such as prevention of falls among elderly in nursing homes makes the intervention amenable to outcome evaluation because the outcome is not very rare and injuries from falls represent the principal injury among616the elderly in the otherwise protected environment of the nursing home [6].
In such instances, actually outcome evaluation may also be more convenient than process evaluation, because there may be no obvious link between the process and the outcome states or events.Injury prevention projects are worthwhile whether they are community based or focusing on high-risk groups. The evaluation of projects focusing on highrisk groups can be done on the basis of outcome measures, because high risk implies considerable potential for reduction and thus, the relevant studies may have sucient statistical power. Communitywide interventions may have multiple objectives addressing small relative risks associated with widespread exposures or a single objective concerning rare outcomes. In these instances process evaluation may be the only feasible option, e.g. interventions for bicycle helmets.
Although outcome evaluation is certainly preferable to process evaluation, this should not prejudice public health professionals away from community-based interventions.
Language: en |