基于ARIMA模型的天津地区单中心HPV感染趋势及基因型特征

Trend and genotype characteristics of single-center HPV infection in Tianjin area based on ARIMA model

摘要:

目的 采用自回归移动平均(ARIMA)模型构建时间序列,分析天津地区单中心人乳头瘤病毒(HPV)感染趋势及基因型特征。方法 选择2018年1月—2022年12月某院进行HPV检测的7 236例女性患者,比较2018—2022年天津地区HPV感染情况及基因型分布。建立ARIMA模型时间序列,分析模型拟合。预测2023年HPV感染数,并与实际发生数进行比较,评价模型的预测效果。结果 2018—2022年天津地区HPV感染率为14.41%;HPV感染率在31~40岁年龄段最高,感染率为15.47%。阳性标本中HPV单一型别感染比率最高,占比为73.54%(767/1 043),以高危型HPV为主。低危型感染占比最高的是HPV-6型,为2.59%,高危型感染占比最高的是HPV-16型,为16.06%。建立ARIMA模型,确定最佳模型为ARIMA(0,1,2)(0,1,1)₁₂,其AIC值和BIC值分别为3.877、4.005,经白噪声检验Ljung-Box Q=8.828差异无统计学意义(P>0.05)。利用模型预测2023年HPV感染数,实际值、预测值的总体趋势基本保持一致,模型RMSE、MAPE、MAE分别为6.289、34.149、4.706,提示模型的预测效果较好。结论 天津地区女性人群中,HPV病毒感染类型以单一高危型感染为主,其中HPV-16型感染率最高。天津地区HPV感染存在季节性,ARIMA模型在HPV感染流行趋势的预测中效果较好,适用于短期预测。

Abstract:

Objective To construct time-series by adopting autoregressive integrated moving average (ARIMA) model for analyzing the trend and genotype characteristics of single-center human papillomavirus (HPV) infection in Tianjin area. Methods A total of 7 236 female patients who underwent HPV testing in a hospital from January 2018 to December 2022 were selected. HPV infection status and genotype distribution in Tianjin area from 2018 to 2022 were compared. ARIMA model time-series was constructed, and model fitting was analyzed. The number of HPV infections in 2023 was predicted and compared with the actual occurrence, the predictive performance of the model was evaluated. Results HPV infection rate in Tianjin area from 2018 to 2022 was 14.41%, with the highest rate (15.47%) in the age group of 31-40 years. Among the positive specimens, the proportion of single type HPV infection was the highest, accounting for 73.54% (767/1 043), with high-risk HPV being the main type. The highest infection rates of low-risk and high-risk types were type HPV-6 (2.59%) and type HPV-16 (16.06%), respectively. ARIMA model was constructed, and the optimal model was ARIMA (0,1,2)(0,1,1)₁₂, with akaike information criterion (AIC) and bayesian information criterion (BIC) values of 3.877 and 4.005, respectively. There was no statistical significance in Ljung-Box Q=8.828 showed by white noise test (P>0.05). The number of HPV infection in 2023 was predicted by the model. The overall trend of the actual value and the predicted value was basically consistent, RMSE, MAPE and MAE of the model were 6.289, 34.149 and 4.706, respectively, suggesting that the model had a good prediction effect. Conclusion Among the female population in Tianjin area, HPV infection is mainly caused by single, high-risk type, with HPV-16 having the highest infection rate. There is seasonal variation in HPV infection in Tianjin. ARIMA model has good prediction effect on the prevalence trend of HPV infection, which is suitable for short-term prediction.