Napovedovanje rehospitalizacij za paciente z multiplo sklerozo

Rikanović, Sanja

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Napovedovanje rehospitalizacij za paciente z multiplo sklerozo
Authors:	ID Rikanović, Sanja (Author) ID Povalej Bržan, Petra (Mentor) More about this mentor... ID Flisar, Dušan (Comentor)
Files:	MAG_Rikanovic_Sanja_2016.pdf (888,29 KB) MD5: 53F58932E0BB5DB029B7EB40B9BAF9D7
Language:	Slovenian
Work type:	Master's thesis/paper
Typology:	2.09 - Master's Thesis
Organization:	FZV - Faculty of Health Sciences
Abstract:	V magistrskem delu smo raziskovali nenačrtovane ponovne sprejeme pri pacientih z multiplo sklerozo (MS). Cilj našega raziskovalnega dela je bil sestaviti model, ki bo pri napovedovanju nenačrtovanih ponovnih sprejemov uspešnejši od modelov, ki niso vezani na posamezno diagnozo. Pri pisanju teoretičnega dela naloge smo se opirali na strokovno literaturo o multipli sklerozi ter na raziskave o modelih za napovedovanje nenačrtovanih ponovnih sprejemov. Za empirični del naloge smo uporabili podatke iz podatkovne baze SID (State Inpatient Database) za Kalifornijo, ki je del skupine podatkovnih baz, razvitih v okviru projekta HCUP (Healthcare Cost and Utilization Project). Specializiran napovedni model, zgrajen na osnovi podatkov o pacientih z multiplo sklerozo, se je pri napovedovanju ponovnega sprejema bolnikov z MS v manj kot 30 dneh izkazal kot uspešnejši od globalnega modela, ki je bil zgrajen na osnovi podatkov o vseh pacientih ne glede na diagnozo. Povprečna AUC-vrednost specializiranega modela je znašala 0,708, kar je za 0,042 višje od povprečne AUC-vrednosti globalnega modela (AUC = 0,666). Prav tako smo pri specializiranem modelu zaznali višje povprečne vrednosti diagnostične natančnosti, senzitivnosti, specifičnosti in NPV. Dodaten prispevek specializiranega modela v primerjavi z globalnim modelom se kaže tudi v nižji Brierjevi oceni ter v manjšem številu uporabljenih vhodnih spremenljivk in posledično v manj kompleksnem modelu. Vse našteto govori v prid specializiranemu napovednemu modelu za paciente z MS, zato smo v naslednjem koraku temu modelu dodali še podatke o predhodnih hospitalizacijah in ugotovili, da vključitev zgodovinskih podatkov o hospitalizacijah prav tako pozitivno vpliva na napovedovanje nenačrtovanih ponovnih sprejemov. Za napovedovanje nenačrtovanih ponovnih sprejemov pri pacientih z MS je bolje uporabiti specializiran model kot splošnega. Rezultati magistrskega dela so primerni za nadaljnje proučevanje rehospitalizacij pri pacientih z MS.
Keywords:	multipla skleroza, rehospitalizacija, bolnišnična odpustna pisma, napovedni model, Lasso regresija, ansambelske metode, odločitvena drevesa.
Place of publishing:	Maribor
Publisher:	[S. Rikanović]
Year of publishing:	2016
PID:	20.500.12556/DKUM-59341
UDC:	616.8(043.2)
COBISS.SI-ID:	2219428
NUK URN:	URN:SI:UM:DK:7DGLIMBS
Publication date in DKUM:	01.09.2016
Views:	2546
Downloads:	222
Metadata:
Categories:	FZV
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Secondary language

Language:	English
Title:	Predicting readmission risk for patients with multiple sclerosis
Abstract:	In this master's thesis we were investigating unplanned readmissions of patients with multiple sclerosis (MS). The aim of our research work was to build a prediciton model for unplanned readmissions that would perform better than models that do not depend on specific diagnosis. When writing the theorethical part of the thesis we relied on the scientific literature on multiple sclerosis and the researches of models for predicting unplanned readmission. For the empirical part of the paper we used data from SID (State Inpatient Database) California, which is part of the databases developed within the HCUP (Healthcare Cost and Utilization Project). Specialized predictive model, that was built on the data of patients with multiple sclerosis (MS), has proved to be more successfull at predicting readmission of patients with MS in less than 30 days than the global model, which was built on the basis of all patients irrespective of diagnosis. The average AUC value of a specialized model was 0.708, which is 0.042 higher than the average AUC of a global model (AUC = 0.666). We have also detected higher average values of accuracy, sensitivity, specificity and NPV of a specialized model. An additional contribution of a specialized model in comparison with the global model was also reflected in a lower Brier score and a smaller number of the input variables and, consequently, less complex model. All of this argues in favor of specialized predictive model for patients with MS, that is why we went further and added information about previous hospitalizations and found that the inclusion of historical data on hospitalizations also has a positive impact on the prediction of unplanned readmissions. For predicting unplanned readmissions for patients with MS is preferable to use a specialized model, rather than general model. The results of the master thesis are suitable for further study of readmissions in patients with MS.
Keywords:	multiple sclerosis, readmission, electronic health records, prediction model, Lasso regression, ensemble methods, decision trees

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back