The Study For Health Service: Statistical Rapid Survey And Linear Modelling

Health service managers in the army need timely and accurate information on which to base their assessment and decision. Yet the foundation of heath service survey in our army is not well established. Although there are many epidemiological and statistical which used in health service survey, many of these traditional methods are not very well suited to an environment with extremely limited financial resources and with few people skilled in data collection and analysis. On the other hand, the traditional linear models, such as ANOVA and ANCOVA, are not suited to analyze the survey data with clustering. In this thesis, some concepts of health service survey are defined and new rapid survey methods and linear models which suited to collect and analyses the health service data are studied. The conception of heath service and heath service survey is compared with that for army. Both of them have same research method except army health service only for a special community-the army. In this thesis, some basic concept is defined and necessity for sampling method in health service survey is discussed.Rapid assessments, appropriately used and carried out, can result in accurate appraisals of the situation in relatively short time. As the recommendation by World Health Organization (many WHO programmers have been actively involved in developing and using these techniques) , the rapid survey methods are widely applied in the world. Those methods include KAP method, community diagnosis, rapid rural appraisal, rapid assessment procedures, rapid evaluation method, rapid epidemiological assessment. Mostly , new survey method is compared with Simple Random Sampling by calculating Deff(design effect) and computer simulating to show its efficacy. In practice, it is shown that above mentioned rapid survey methods are quick and reliable. Those methods have been used in our country and some results have been gotten.EPI developed a rapid method for cluster sampling that is well suited to settings in which sampling frames are not readily available. But EPI dose not consider the size of the chosen cluster, a modified EPI sampling (MES) are developed in this thesis. For MES, the sample size of 30 chosen cluster is determined bywhere N_i is the ith cluster size, N_min= min{N_i} . Computer simulation and application show that it is better than EPI and SRS.