ARIMA models to predict new-diagnosing cases of pneumoconiosis in Nanjing / 中华劳动卫生职业病杂志

ABSTRACT

<p><b>OBJECTIVE</b>The primary goal of our study was to evaluate and predict the epidemiological trend of pneumoconiosis.</p><p><b>METHODS</b>We established a new database on new-diagnosing cases of pneumoconiosis in Nanjing during 1955-2007. The database was analyzed by using SAS9.1.3 statistical software, Data sequence was stabilized by using the process of differencing. Goodness of Fit Test verified that the residual-error sequence was white noise sequence. We determined the Autoregressive Integrated Moving Average models (ARIMA models) as an appropriate model. By Taking advantage of back-substitution model, we predicted new cases of pneumoconiosis during 2008-2012. We then compared the predicted value with the actual value to test and verify the predicting function.</p><p><b>RESULTS</b>We finally had chosen out ARIMA (2, 1, 0) models to fit the original sequence, which led to the results that the observed values are basically comparable with the predicted values. The past 5-year predicted-value was similar to the actual value. We then built a new model by new cases during 1955-2012, predicting that the trend of pneumoconiosis in the next 5 years will tend to approach plateau with approximately 10∼13 new cases per year.</p><p><b>CONCLUSION</b>ARIMA models is suitable for fitting large sample series of new diagnosed pneumoconiosis over the years and for the predicting the incidence of pneumoconiosis.</p>