The Epidemic Character Of HIV/AIDs And Compariso-n Of Different Prediction Models In Shenzhen

Chapter one:The distribution character of HIV/AIDS in ShenzhenObjective:To analyze the distribution character of HIV/AIDS in Shenzhen city from1992to2010and find the population who need to be focused in the prevention and control of AIDS.Methods:1The source of data:The data of new reported HIV/AIDS and mortality information in Shenzhen city from1992and2010was provided by Shenzhen center for disease control and prevention.2. Data analysis:Excel software was used to compute absolute growth number (cumulative absolute growth number and absolute growth number year by year), development speed (base ratio and chain ratio), growth rate (base ratio and chain ratio), average development speed and average growth rate. SPSS16.0was used to analyze the distribution of AIDS.Results:1. Population distribution character:During the period of1992and December31,2010, the cumulative reported number of HIV/AIDS is5078in Shenzhen.3952of them (77.8%) were males and1126were females (22.2%). The proportion of non registered population was89.5%, and the proportion of registered population was only10.5%. The educational level of most HIV/AIDS was lower, only29.2%of them had received the education in senior high school or college. Sexual transmission and drug abuse transmission was the main route of HIV transmission in Shenzhen, the proportion was53.2%and39.3%respectively. By December31,2010,133deaths from HIV/AIDS were reported.115(86.5%) died from AIDS and18(13.5%) died from HIV.97(72.9%) of them were males and36(27.1%) of them were females.2. Regional distribution character:The reported cases showed phenomenon of area un-balanced. The cases from bao'an district were largest, the number was1466and the proportion was28.9%. The second one was longgang district, the number was1149and the proportion was22.6%.3. The time trend:There was no report of HIV from1992-1994. The average growth rate of new reported HIV from1995to2010was43.1%. The average growth rate of new reported AIDS from1992to2010was32.6%. It was predicted that the number of new reported HIV and AIDS in2011would be1013and278basing on the average growth rate of HIV/AIDS. After2009, the new reported HIV increased slower than before, and the chain ratio growth rate was4.8%in2009and1.3%in2010. Conclusions:1. The average growth rate of new reported HIV is higher than that of AIDS. So there will be more AIDS patients in the future. The designated medical units for AIDS should pay more attention to this trend.2. The epidemic of AIDS in Shenzhen showed increasing trend from1992-2010.3Males, adults, the people with lower educational background and non registered population should be paid more attention to control the epidemic of AIDS in Shenzhen. It's very important to strengthen the surveillance and test of them.4. Three transmission routines of HIV coexit in Shenzhen. Sexual transmission and drug transmission is the main route of HIV transmission in Shenzhen. These bring about heavy threaten to the prevention and control of AIDS. So we should master clearly the number of female sex workers, men having sex with men, bisexual men, and one HIV-positive spouse family, then strengthening the surveillance and test to them. Chapter two:The comparison of different prediction models in the application of epidemic trend of AIDS in ShenzhenAim:To explore good prediction models to predict epidemic trend of AIDS during the rapid growth period in Shenzhen city and test the prediction effect, so as to provide reference basis for setting up the strategy of preventing and controlling AIDS.Methods:Gray theoretical model, curve estimation, exponential smoothing and autoregressive integrated moving average (ARIMA) model were fitted to predict the epidemic trend of AIDS. The data of new reported HIV/AIDS in Shenzhen city from2000to2009were used to fit, and the data of2010was used to test the effect of prediction models. Excel software was used to fit gray model (1,1). SPSS13.0was used to do statistical description and fit curve estimation model, exponential smoothing and ARIMA model. The significant level a was0.05.Results:1. GM(1,1)GM(1,1) could be used to fit the epidemic trend of new reported HIV in Shenzhen during the period of2000and2009and the fitting precision of model was qualified. The standard error of the model was75.49. The average relative error was38.2%. The prediction model was Y(t+1)=702.3le0.205t-686.31. Basing on this prediction model, the new reported HIV in2010would be1012, while the actual number was708. The relative error of prediction was29.7%.GM(1,1) could be used to fit the epidemic trend of new reported AIDS in Shenzhen during the period of2000and2009and the fitting precision of model was qualified. The standard error of the model was13.37. The average relative error was50.5%. The prediction model was Y(t+1)=102.62e0.25t-91.62. Basing on this prediction model, the new reported AIDS in2010would be263, while the actual number was211. The relative error of prediction was22.6%.2. Curve estimationAmong all the models of curve estimation, power function was the best one to fit the epidemic trend of new reported HIV in Shenzhen during the period of2000and2009. The determination coefficient R2of power function was0.974. The standard error of the model was0.211. The average relative error was16.7%. The prediction model was yt=17.769×t1.664. It was predicted there would be961new reported HIV in2010, larger than the actual number253. The relative error of prediction was26.3%.Among all the models of curve estimation, power function was the best one to fit the epidemic trend of new reported HIV in Shenzhen during the period of2000and2009. The determination coefficient R2of power function was0.974. The standard error of the model was0.211. The average relative error was16.7%. The prediction model was yt=17.769×t1.664. It was predicted there would be961new reported HIV in2010,253more than the actual number. The relative error of prediction was26.3%.Quadratic prediction equation was the best model among all the models of curve estimation to fit the epidemic trend of new reported AIDS in Shenzhen during the period of2000and2009. The determination coefficient R2of quadratic prediction equation was0.980. The standard error of the model was14.000. The average relative error was18.7%. The prediction model was yt=7.083+2.219t+1.580×t2. It was predicted there would be223new reported AIDS in2010,12more than the actual number. The relative error of prediction was5.4%.3. Exponential smoothingHolt quadratic exponential smoothing model was better method to fit the epidemic trend of new reported HIV in Shenzhen during the period of2000and2009. The standard error of the model was63.6. The average relative error was51.4%. It was predicted there would be768new reported HIV in2010,60more than the actual number. The relative error of prediction was8.5%.Holt quadratic exponential smoothing model was better method to fit the epidemic trend of new reported AIDS in Shenzhen during the period of2000and2009. The standard error of the model was20.5. The average relative error was73.5%. It was predicted-there would be210new reported AIDS in2010,1less than the actual number. The relative error of prediction was0.5%.4. ARIMA modelSquare sum of residual was142601.0and standard error of residual was119.4of ARMA(1,1) model. Log-likelihood function value was-62.074, Akaike's Information Criterion (AIC)=130.148, Schwarz's Bayesian Criterion (BIC)=131.055. The actual value of new reported HIV in all years was within the range of95%confidence interval of prediction, which meant the new reported HIV in all years was in the normal range. It was predicted there would be688new reported HIV in2010,20less than the actual number. The relative error of prediction was2.8%.Squard sum of residual was9308.7and standard error of residual was30.5of ARMA(1,1) model. Log-likelihood function value was-48.403, Akaike's Information Criterion (AIC)=102.806, Schwarz's Bayesian Criterion (BIC)=103.714. The actual value of new reported AIDS in all years was within the range of95%confidence interval of prediction, which meant the new reported AIDS in all years was in the normal range. It was predicted there would be202new reported AIDS in2010,9less than the actual number. The relative error of prediction was4.3%. Conclusions:1. The epidemic trend of AIDS in Shenzhen will increase in the following years.2. Not only the index of fitting goodness, but also the index of prediction relative error should be taken into account when predict the epidemic trend of AIDS.(1) Power function under curve estimation is the best model to fit the epidemic trend of new reported HIV. Quadratic prediction equation under curve estimation is the best one to fit the epidemic trend of new reported AIDS in Shenzhen during the period of2000and2009.(2) ARMA(1,1) model and Holt quadratic exponential smoothing model have more advantage in prediction the trend of HIV/AIDS, the prediction value is more consistent with the actual value.(3) GM(1,1) can also be used to fit the epidemic trend of new reported HIV/AIDS in Shenzhen during the period of2000and2009. But the effect of fitting and prediction is not perfect. So we don't recommend to use GM(1,1).