Article type
Year
Abstract
Background:
Medicine, including obstetrics, is increasingly focused on risk-based or personalized medicine. Consequently, there has been a steady increase in the development and publication of prediction models for various maternal and fetal outcomes, but an overview is lacking and their clinical applicability is unknown.
Methods:
Through MEDLINE we identified all peer-reviewed journal articles on prediction models in obstetrics published up to July 2012. After title and abstract screening in duplicate, four reviewers extracted for each prediction model, the study design, description of participants and outcome, sample size, type of model, model presentation, type of validation including model performance and guidance for use in clinical practice.
Results:
From 10,152 hits, we identified 177 papers that reported the development of 263 prediction models for 40 different outcomes. The number of published models increased each year (Figure 1). The most frequently predicted outcomes were pre-eclampsia (n = 69), preterm delivery (n = 63), mode of delivery (n = 22), gestational hypertension (n = 11) and small for gestational age neonates (n = 10). Internal and external validation was reported for 21.7% (57/263) and 8.7% (23/263) of the prediction models, and discrimination and calibration for 62.7% (165/263) and 17.5% (46/263), respectively. Two-thirds of the papers (164/263) presented the model in such a way that application in other populations was possible and the clinical applicability was discussed in 11.0% (29/263). The performance of models published more recently was often not better than previously published, older models.
Conclusions:
Many prediction models for (adverse) pregnancy outcomes are developed, but very few models are validated, and their impact on clinical practice remains unknown. Efforts should be directed towards evaluating the performance of existing models rather than developing new models. The role of existing models in guiding management of pregnant women in clinical practice is rarely described and the impact of models on clinical practice and on clinical outcomes remains unknown.
Medicine, including obstetrics, is increasingly focused on risk-based or personalized medicine. Consequently, there has been a steady increase in the development and publication of prediction models for various maternal and fetal outcomes, but an overview is lacking and their clinical applicability is unknown.
Methods:
Through MEDLINE we identified all peer-reviewed journal articles on prediction models in obstetrics published up to July 2012. After title and abstract screening in duplicate, four reviewers extracted for each prediction model, the study design, description of participants and outcome, sample size, type of model, model presentation, type of validation including model performance and guidance for use in clinical practice.
Results:
From 10,152 hits, we identified 177 papers that reported the development of 263 prediction models for 40 different outcomes. The number of published models increased each year (Figure 1). The most frequently predicted outcomes were pre-eclampsia (n = 69), preterm delivery (n = 63), mode of delivery (n = 22), gestational hypertension (n = 11) and small for gestational age neonates (n = 10). Internal and external validation was reported for 21.7% (57/263) and 8.7% (23/263) of the prediction models, and discrimination and calibration for 62.7% (165/263) and 17.5% (46/263), respectively. Two-thirds of the papers (164/263) presented the model in such a way that application in other populations was possible and the clinical applicability was discussed in 11.0% (29/263). The performance of models published more recently was often not better than previously published, older models.
Conclusions:
Many prediction models for (adverse) pregnancy outcomes are developed, but very few models are validated, and their impact on clinical practice remains unknown. Efforts should be directed towards evaluating the performance of existing models rather than developing new models. The role of existing models in guiding management of pregnant women in clinical practice is rarely described and the impact of models on clinical practice and on clinical outcomes remains unknown.