{"id":807,"date":"2016-06-15T10:26:34","date_gmt":"2016-06-15T08:26:34","guid":{"rendered":"https:\/\/site.uit.no\/trec\/?p=807"},"modified":"2016-06-15T11:08:10","modified_gmt":"2016-06-15T09:08:10","slug":"prediction-models-a-modern-version-of-crystallomancy-just-better","status":"publish","type":"post","link":"https:\/\/site.uit.no\/trec\/2016\/06\/15\/prediction-models-a-modern-version-of-crystallomancy-just-better\/","title":{"rendered":"Prediction models: A modern version of crystallomancy (just better)"},"content":{"rendered":"

By S\u00f8ren Beck Jensen<\/a>, postdoc TREC<\/p>\n

<\/a><\/p>\n

An anthropologist friend of mine once told me about a field study where he attempted to decipher the religious thoughts of an Inuit tribe. The Elder who had always lived on the edge of the world was very reluctant to give any answers about this, to him, taboo topic but finally gave one piece of information: We do not believe. We fear.<\/p>\n

We, humans, feel comfort in knowing the future. Degrees of uncertainty triggers excitement that becomes unease and then fear. Traditionally, a gaze into the crystal ball was a means to predict the future. Surely, most people would feel at unease if the doctor were to look into the crystal ball to decide on treatment and to give estimates on health and disease. Medical science today make use of prediction models to give estimates on health and disease and to balance benefits against harm for a treatment. A recent article published in The Lancet [1] describes the development of a prediction model for the risk of bleeding complications in connection with anticoagulant treatment of atrial fibrillation. The model makes use of clinical information as well as a set of biomarkers to improve the predictive power over existing models.<\/p>\n

Initially, the study ranked a set of candidate predictors to identify five measures that gave the most information on bleeding risk. The best predictors: Age, Biomarkers, and Clinical history of bleeding were then used to design\u00a0the ABC-prediction model. Hereafter, the ABC-model was validated with another group of patients and was found superior to older prediction models. The increased predictive power should be useful in decision support for anticoagulant treatment in patients with atrial fibrillation. A main concern in atrial fibrillation patients is the risk of stroke, which may be estimated in another model [2] . We look to the past to predict the future. We teach it to our kids, like the ABC: Today was good. Today was fun. Tomorrow is another one (Dr. Seuss). The Inuit Elder looks down.<\/p>\n

References:<\/p>\n

    \n
  1. Hijazi, Z., et al., The novel biomarker-based ABC (age, biomarkers, clinical history)-bleeding risk score for patients with atrial fibrillation: a derivation and validation study.<\/a><\/em> Lancet, 2016.<\/li>\n
  2. Hijazi, Z., et al., The ABC (age, biomarkers, clinical history) stroke risk score: a biomarker-based risk score for predicting stroke in atrial fibrillation.<\/em> <\/a>Eur Heart J, 2016. 37<\/strong>(20): p. 1582-90.<\/li>\n<\/ol>\n