Analiza vremenskih serija obolevanja od moždanog udara

Abstract

Dosadašnja istraživanja ukazala su da obolevanje od moždanog udara (MU) nije slucajno rasporeeno u vremenu vec da postoje jasni dnevni, nedeljni i sezonski obrasci. Podaci koji se koriste za ovu vrstu analiza cesto predstavljaju vremenske serije (u vremenu ureeni niz podataka) i zahtevaju posebne metode za analizu vremenskih serija. Znacajan izvor podataka o obolevanju od moždanog udara predstavljaju izveštaji o osobama lecenim u stacionarnim uslovima, tzv. hospitalni registri. Ciljevi ovog istraživanja bili su: provera validnosti podataka u korišcenom hospitalnom registru, ispitivanje demografskih karakteristika hospitalizovanih pacijenata, analiza dnevnog i nedeljnog variranja broja prijema obolelih od razlicitih tipova MU, identifikacija ARIMA modela mesecnih i kvartalnih vremenskih serija, analiza sezonskog variranja broja prijema obolelih od razlicitih tipova MU na osnovu definisanih modela i ispitivanje uticaja meteoroloških faktora na sezonske varijacije u broju hospitalizovanih pacijenata obolelih od moždanog udara na osnovu modela funkcije prenosa. Istraživanje je dizajnirano po tipu nepopulacione deskriptivne studije. Osnovni izvor podataka bio je elektronski registar pacijenata Klinike za neurologiju, Klinickog centra u Nišu. Ovim istraživanjem su obuhvaceni hospitalni prijemi pacijenata koji su kao završnu dijagnozu imali subarahnoidnu hemoragiju (SAH), intracerebralnu hemoragiju (ICH), akutni ishemijski moždani udar (AIMU) ili nespecifikovani tip moždanog udara (NMU). U slucaju analize dnevnih i nedeljnih varijacija broja prijema uzorak je stratifikovan prema polu, obrazovanju i starosti. Za analizu dnevnih i nedeljnih varijacija podaci su grupisani i podeljeni na cetiri šestocasovna perioda dana, sedam dana u nedelji i na dane vikenda i radne dane. Korišcen je 2 c test za uporeivanje registrovanog i ocekivanog broja hospitalizovanih pacijenata. Izdvojeni su dani u nedelji odnosno periodi dana sa najvecim i najmanjim brojem hospitalnih prijema. Sezonske varijacije u obolevanju od moždanog udara su ispitivane metodama za analizu vremenskih serija. Vremenskim agregiranjem su formirane prekidne vremenske serije broja hospitalizacija na mesecnom i kvartalanom nivou, posebno za svaki tip MU. Vremenske serije su modelovane korišcenjem Box Jenkinsovog metoda izgradnje ARIMA (autoregresioni integrisani proces pokretnih proseka) modela...

Available studies have shown that stroke onset is not accidentally distributed in time, but rather that there are clear daily, weekly and seasonal patterns. Data used for this type of analysis often take the form of time series (a sequence of measurements ordered in time) and they require specific methods of time series analysis. A relevant source of data on stroke onset may be found in institutional reports on in-patients, socalled hospital registers. The goals of the present research were: checking the validity of data used in the hospital register, studying the demographic characteristics of hospitalized patients, analysis of daily and weekly variation in the number of admissions of patients suffering from various types of stroke, identification of ARIMA models of monthly and quarterly time series, analysis of seasonal variation in the number of admissions based on defined models and studying the influence of meteorological factors on seasonal variations in the number of hospitalized stroke-affected patients, based on the transfer function models. The research was designed as a non-population descriptive study. The principal data source was the electronic patient register of the Clinic for Neurology, Clinical Center, Niš. The present research encompassed hospital admissions of patients whose final diagnosis was subarachnoid hemorrhage (SAH), intracerebral hemorrhage (ICH), acute ischemic stroke (IS), or unspecified stroke type (US). In the analysis of daily and weekly variations, the sample was stratified by sex, education, and age. In this analysis, data were grouped and classified into four six-hour periods in the day, seven days in the week, and weekends and workdays. The 2 c test was used for comparing the registered and expected numbers of hospitalized patients. Days in the week and periods of the day with the highest and lowest numbers of hospital admissions were separated. Seasonal variations in stroke onset were tested by time series analysis methods. Time aggregation was used to form interrupted time series of monthly and quarterly admission numbers, separately for each stroke type. Time series were modeled using the Box Jenkins ARIMA (autoregressive integrated moving average) model...