Project - Home

Prediktivni modeli u zdravstvu (Predictive Models in Health Care) No: 108-0982560-0257 .

Status projekta - Dio programa "Računarski postupci otkrivanja znanja u znanstvenim primjenama" (Computational Knowledge Discovery Methods in Scientific Applications).
Prioritetno područje istraživanja - Zdravlje
Ustanova nositelj projekta -Medicinski fakultet Sveučilišta u Zagrebu. Dekan, ravnatelj, čelnik ustanove: Prof.dr.sc. Nada Čikeš.
Trajanje projekta - 5 godina

VODITELJ PROJEKTA - Dr.sc. Zdenko Sonicki, dr. med.
Zavod za medicinsku statistiku, epidemiologiju i medicinsku informatiku (Department of Medical Statistics, Epidemiology and Medical Informatics)
Znanstveno područje, polje i grana: Biomedicina i zdravstvo,
Javno zdravstvo i zdravstvena zaštita, Javno zdravstvo
Vrsta istraživanja - Primijenjeno istraživanje

Summary:

Presumptions: In Croatia as well as in the Europe, suicide is the most frequent cause of death among young and middle-aged people. Persons suffering from post-traumatic stress disorder (PTSD) and bipolar affective disorder (BAD) are at high-risk for committing suicide. The situation in Croatia is even worse after the war. Many analyses have been conducted in order to define risks and prevent suicide, most of them in the form of medical guidelines, but few based on prospective statistical studies. Predictive models are a form of healthcare technology and should be studied as well as new drugs or diagnostic devices. Predictive models in healthcare system are tools for medical-decision making. They are useful in prevention of illness.

Goals and hypotheses: The goals are: to determine the most powerful predictors of suicidal behavior for people suffering from PTSD and BAD in a prospective study using statistical methods and data mining techniques; develop suicide prediction model according to methodology standards; choose model with the highest prediction power and the lowest number of predictors; estimate clinical sense and applicability of chosen model; maximize model simplicity and adjust to different target groups; prospectively validate the model; apply methodology for prediction modeling, that was developed to predict suicide intentions, to other healthcare domains. The hypotheses are that prediction model when used in healthcare facilitates prediction of a certain illness, and development of models adjusted to different target groups enables prevention of certain illness, as well as of suicide in people with PTSD and BAD diagnoses.

Expected results: Suicide prevention model for people suffering from PTSD and BAD adjusted to be used by physicians, family and working milieu. Discovering new methodological approaches for development of predictive models in healthcare, and adjusting them to target populations.

Validation: Mathematical and statistical validation will be conducted first, and then prospective validation on new examinees. Results will be published in scientific papers and final report in the scientific book.

Research importance: Suicide predictive model development will decrease suicide incidence among persons suffering from PTSD and BAD. It is expected that results of this study would be applicable to suicide prevention in other risk groups as well as in other countries.

predictive modeling
feature extraction
model customization
suicide prevention
posttraumatic stress disorder
bipolar affective disorder

Ključne riječi

prediktivno modeliranje
određivanje značajki
prilagodba modela
prevencija suicida
posttraumatski stresni poremećaj
bipolarni afektivni poremećaj