Predicting Adolescent Intervention Non-responsiveness for Precision HIV Prevention Using Machine Learning

Study on the prediction performance of AIDS monthly incidence in Xinjiang based on time series and deep learning models

OBJECTIVE

AIDS is a highly fatal infectious disease of Class B, and Xinjiang is a high-incidence region for AIDS in China. The core of prevention and control lies in early monitoring and early warning. This study aims to identify the best model for predicting the monthly AIDS incidence in Xinjiang, providing scientific evidence for AIDS prevention and control.

METHODS

Monthly AIDS incidence data from January 2004 to December 2020 in Xinjiang were collected. Six different models, including the ARIMA (2,1,2) model, ARIMA (2,1,2)-EGARCH (2,2) combined model, ARIMA (2,1,2)-TGARCH (1,1) combined model, ETS (A, A, A) model, XGBoost model, and LSTM model, were used for fitting and forecasting.

RESULTS

All models were able to capture the overall trend of the monthly AIDS incidence in Xinjiang. In terms of RMSE and MAE, the ETS (A, A, A) model performed the best, achieving the smallest values. For the MAPE metric, the ARIMA (2,1,2)-TGARCH (1,1) model performed the best. Considering RMSE, MAE, and MAPE together, the ETS (A, A, A) model was the best-performing model in this study. The LSTM model also showed good predictive performance, while the XGBoost model and ARIMA (2,1,2) model performed relatively poorly.

CONCLUSION

The ETS (A, A, A) model is the best model for predicting the monthly AIDS incidence in Xinjiang. Deep learning models (such as LSTM) have significant potential in time series forecasting. The XGBoost model and ARIMA (2,1,2) model may have limitations when handling time series data, and future improvements or combinations could enhance prediction performance.

Special Section on Digital Health for Precision in Prevention: Notable Papers that Leverage Informatics Approaches to Support Precision Prevention Efforts in Health Systems

OBJECTIVE

To identify notable research contributions relevant to digital health applications for precision prevention published in 2023.

METHODS

An extensive search was conducted to identify peer-reviewed articles published in 2023 that examined ways that informatics approaches and digital health applications could facilitate precision prevention. The selection process comprised three steps: 1) candidate best papers were first selected by the two section editors; 2) a diverse, international group of external informatics subject matter experts reviewed each candidate best paper; and 3) the final selection of four best papers was conducted by the editorial committee of the Yearbook. The section editors attempted to balance selection by authors' global region and areas with clinical medicine and public health.

RESULTS

Selected best papers represent studies that advanced knowledge surrounding the use of digital health applications to facilitate precision prevention. In general, papers identified in the search fell into one of the following categories: 1) applications in precision nutrition; 2) applications in precision medicine; and 3) applications in precision public health. The best papers spanned several disease targets, including Alzheimer's disease, HIV, and COVID-19. Several candidate papers sought to improve prediction of disease onset, whereas others focused on predicting response to interventions.

CONCLUSION

Although the selected papers are notable, significant work is needed to realize the full potential for precision prevention using digital health. Current data and applications only scratch the surface of the potential that information technologies can bring to support primary and secondary prevention in support of health and well-being for all populations globally.

Predicting Beta-Lactam Target Non-Attainment in ICU Patients at Treatment Initiation: Development and External Validation of Three Novel (Machine Learning) Models

In the intensive care unit (ICU), infection-related mortality is high. Although adequate antibiotic treatment is essential in infections, beta-lactam target non-attainment occurs in up to 45% of ICU patients, which is associated with a lower likelihood of clinical success. To optimize antibiotic treatment, we aimed to develop beta-lactam target non-attainment prediction models in ICU patients. Patients from two multicenter studies were included, with intravenous intermittent beta-lactam antibiotics administered and blood samples drawn within 12-36 h after antibiotic initiation. Beta-lactam target non-attainment models were developed and validated using random forest (RF), logistic regression (LR), and naïve Bayes (NB) models from 376 patients. External validation was performed on 150 ICU patients. We assessed performance by measuring discrimination, calibration, and net benefit at the default threshold probability of 0.20. Age, sex, serum creatinine, and type of beta-lactam antibiotic were found to be predictive of beta-lactam target non-attainment. In the external validation, the RF, LR, and NB models confirmed good discrimination with an area under the curve of 0.79 [95% CI 0.72-0.86], 0.80 [95% CI 0.73-0.87], and 0.75 [95% CI 0.67-0.82], respectively, and net benefit in the RF and LR models. We developed prediction models for beta-lactam target non-attainment within 12-36 h after antibiotic initiation in ICU patients. These online-accessible models use readily available patient variables and help optimize antibiotic treatment. The RF and LR models showed the best performance among the three models tested.