Abstract:
Diagnosis of life-threatening viral diseases, such as Meningitis, Viral Hepatitis, Japanese Encephalitis, Dengue, Leptospirosis (Rat Fever) to name a few, is extremely challenging particularly in low-resource settings, because the clinical presentation of such diseases cannot accurately be differentiated from that of other types of viral fever and laboratory tests need to be done to confirm the diagnosis. Due to limitations on cost or availability of diagnostics, or lack of access to laboratory facilities for specimen testing, it may not be possible to conduct diagnostic testing nationwide on all recorded suspected disease cases. Therefore epidemiologists will select a subset of such suspected cases for further investigation based on a rule of thumb. Thus a classification rule is vital to assist doctors in order to do this selection. In addition to diagnosis, it is also important to determine the prognosis of such patients as the concern is on life threatening diseases. Determining diagnosis and prognosis is often further complicated by the presence of missing values. The major objective of this study was to develop a user friendly Automated Statistical Information System (ASIS) that will output the diagnosis and prognosis of the patient when details regarding risk factors are given. In order to satisfy each of these objectives logistic modeling, survival modeling and Missing value imputation was used. Once the appropriate models were fitted, these models were combined using a Hierarchical Statistical Decision model (HSDM) to aid in developing the ASIS. The methodology developed was illustrated on a dataset of Acute Encephalitis Syndrome (AES) patients. The developed ASIS is applicable to any life threatening viral disease and it will help the epidemiologist to make quick decisions particularly in low income settings where there are low funds for sophisticated diagnostics.
