Background: The implementation of the Evidence Based Prevention (EBP) requires a better definition of rules to be applied for using and pooling non-experimental evidence, e.g. etiologic studies and studies on effectiveness of Public Health interventions. Methods: The best tool for summarizing the evidence is Systematic Review (SR). Choosing the study design for inclusion of studies is a difficult task. For studies on risk exposure cohort and case-control studies are the most reasonable choice. For studies on health impact assessment the choice depends on the “unit of administration”: 1. intervention administered at the individual level: RCT is the best design; 2. intervention administered at group level (classes or schools): a Cluster RCT is the most appropriate; 3. intervention provided uniformly within a vast area (e.g. the fluoridation of water): Controlled Before-After study; 4. intervention consisting in a policy or a law: the only design providing results with low level of bias is the Interrupted Time Series. To be pooled producing estimates with low level of bias, the rules of SR of RCTs have to be translated for SR of non-randomized studies: i) the search strategy has to be standardized and comprehensive; ii) unpublished studies have to be traced, to reduce the risk of publication bias; iii) exclusion criteria have to be defined, including the adjustment for confounding factors, and the conditional misclassification of exposure and outcomes; iv) a quality appraisal of the included studies have to be performed. v) data extraction has to be performed by pooling individual data or by calculating adjusted absolute numbers from adjusted RRs; vi) study heterogeneity has to be evaluated and dealt with, e.g. stratifying or performing sensitivity analyses. Conclusions. In spite a number of systematic reviews has been conducted using observational studies, a general agreement on the rules to apply has not been reached yet. To implement Evidence Based Prevention, this rules have to be defined.
Can data from observational studies be pooled?
VIGNA-TAGLIANTI, Federica
2005-01-01
Abstract
Background: The implementation of the Evidence Based Prevention (EBP) requires a better definition of rules to be applied for using and pooling non-experimental evidence, e.g. etiologic studies and studies on effectiveness of Public Health interventions. Methods: The best tool for summarizing the evidence is Systematic Review (SR). Choosing the study design for inclusion of studies is a difficult task. For studies on risk exposure cohort and case-control studies are the most reasonable choice. For studies on health impact assessment the choice depends on the “unit of administration”: 1. intervention administered at the individual level: RCT is the best design; 2. intervention administered at group level (classes or schools): a Cluster RCT is the most appropriate; 3. intervention provided uniformly within a vast area (e.g. the fluoridation of water): Controlled Before-After study; 4. intervention consisting in a policy or a law: the only design providing results with low level of bias is the Interrupted Time Series. To be pooled producing estimates with low level of bias, the rules of SR of RCTs have to be translated for SR of non-randomized studies: i) the search strategy has to be standardized and comprehensive; ii) unpublished studies have to be traced, to reduce the risk of publication bias; iii) exclusion criteria have to be defined, including the adjustment for confounding factors, and the conditional misclassification of exposure and outcomes; iv) a quality appraisal of the included studies have to be performed. v) data extraction has to be performed by pooling individual data or by calculating adjusted absolute numbers from adjusted RRs; vi) study heterogeneity has to be evaluated and dealt with, e.g. stratifying or performing sensitivity analyses. Conclusions. In spite a number of systematic reviews has been conducted using observational studies, a general agreement on the rules to apply has not been reached yet. To implement Evidence Based Prevention, this rules have to be defined.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
