EUVIMEDOpdag EUVIMED
Open access

Horizon-Aware Forecasting of Hypertension Prevalence and Mean Total Cholesterol in Chinese Men: Retrospective Leakage-Controlled Benchmark Study Comparing Statistical, Machine-Learning, and Neural Models

Horizon-aware Prognose der Hypertonieprävalenz und des mittleren Gesamtcholesterins bei chinesischen Männern: retrospektive, leakage-kontrollierte Benchmark-Studie zu statistischen, Machine-Learning- und neuronalen Modellen

Weiguo Lai, Fazli Subhan, Nana Zhou, Jawad Khan

JMIR Medical Informatics · 2026-11-11 · 14 · e99686-e99686

DOI: 10.2196/99686 ↗

Auf einen Blick

Zwei jährliche Reihen zu Hypertonieprävalenz und mittlerem Gesamtcholesterin bei chinesischen Männern wurden mit statistischen, Machine-Learning- und neuronalen Modellen verglichen. Einfache statistische Modelle erzielten die besten RMSE-Werte.

Fragestellung

Wie schneiden statistische, klassische Machine-Learning- und neuronale Prognosemodelle für ASPH und MTC bei chinesischen Männern ab, und bringt Aufmerksamkeit oder mehr neuronale Kapazität einen konsistenten Vorteil?

Methode

Retrospektive Benchmark-Studie mit 2 offiziellen jährlichen Reihen; leakage-kontrollierte expanding-window-Evaluation ab 21 Beobachtungen für 1-, 2- und 5-Jahres-Horizonte; Vergleich mehrerer Modellklassen mit vorab definierten Metriken.

Ergebnisse

Statistische Modelle hatten in allen 6 Ziel-Horizont-Kombinationen die niedrigste RMSE. ARIMA gewann bei ASPH h=1 (0.000178), h=2 (0.000578) und h=5 (0.004324) sowie bei MTC h=1 (0.000400) und h=2 (0.001479); ETS bei MTC h=5 (0.006991).

Einschränkungen

Die Ergebnisse beziehen sich auf 2 Zielreihen, ziel- und horizonenspezifische Auswertungen und kleine Evaluationsstichproben. Vorhersageintervalle waren oft unzureichend kalibriert.

Schlussfolgerung der Autor:innen

Einfache statistische Prognosen blieben in diesen kurzen, retrospektiv geschätzten Jahresreihen konkurrenzfähig. Mehr Komplexität führte nicht sicher zu besseren Punktvorhersagen; Aufmerksamkeit oder größere Kapazität brachten keinen konsistenten Vorteil.

Kernaussagen

  • Statistische Modelle erzielten in allen 6 Zellen die beste RMSE.
  • ARIMA gewann 5 der 6 Ziel-Horizont-Kombinationen.
  • Vorhersageintervalle unterdeckten häufig.

Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.

At a glance

Journal article in JMIR Medical Informatics, published online on 2026-11-11. Written by Weiguo Lai and 3 co-authors. Institution: School of Computer Sciences, Shandong Xiehe University, Jinan, Shandong, China. Major MEDLINE topics: Machine Learning, Hypertension, Cholesterol. 43 references. The full text is freely available in PubMed Central.

Auf einen BlickAbstractSubjectsBibliografiske oplysningerCiteRelated studies

Abstract (original language)

Background

Forecasts of population-level cardiovascular risk factors can support surveillance and planning, but annual country-level health series are often short. With only 30 or 39 annual values per target, model flexibility can outpace the information available for training, making leakage control, repeated temporal evaluation, and comparison with simple baselines essential.

Objective

The aim of this study is to compare statistical, classical machine-learning, and neural forecasting models for age-standardized prevalence of hypertension (ASPH) and mean total cholesterol (MTC) in Chinese men and to test whether attention or substantially greater neural capacity provides consistent predictive benefit.

Methods

The authors analyzed 2 officially sourced annual series: ASPH for Chinese men aged 30 to 79 years (1990‐2019; 30 observations) and MTC for Chinese men aged 18 years or older (1980‐2018; 39 observations). No series was interpolated, extrapolated, imputed, smoothed, spliced, or synthetically extended. Leakage-controlled expanding-window evaluation began after 21 observations and assessed 1-, 2-, and 5-year horizons. Classical machine-learning and neural models used a fixed primary lookback of 3 (L=3), with nested lookback selection disabled; neural models used 10 fixed seeds. Performance was assessed with root mean square error (RMSE), mean absolute error, mean absolute scaled error, root mean squared scaled error, symmetric mean absolute percentage error, directional accuracy, mean error, median error, diagnostic R ², prespecified neural-ablation inference, empirical forecast-interval coverage, and computational-cost reporting.

Results

Statistical models had the lowest RMSE in all 6 target-horizon cells. Autoregressive integrated moving average won for ASPH at h=1 (RMSE 0.000178), h=2 (RMSE 0.000578), and h=5 (RMSE 0.004324), and for MTC at h=1 (RMSE 0.000400) and h=2 (RMSE 0.001479); error-trend-seasonal state space won for MTC at h=5 (0.006991). Classical machine-learning and primary neural models won no cells. Removing attention reduced RMSE in 5 of 6 ablation cells, while the compact architecture had lower RMSE than Original23K in 5 of 6 cells; formal differences were detected for both contrasts at ASPH h=1 and h=2, whereas ASPH h=5 (n=5) was descriptive only. Predictive intervals often undercovered. The benchmark completed all 7242 planned output-producing fits.

Conclusions

Simple statistical forecasting remained highly competitive in these short, retrospectively estimated annual series. Greater complexity did not ensure better point prediction, and neither attention nor larger capacity conferred a consistent advantage. Point accuracy did not ensure well-calibrated predictive uncertainty; findings were conditional on the 2 targets, target-specific horizons, and small evaluation samples.

Subjects

MeSH terms (MEDLINE)

Machine LearningHypertensionCholesterolHumansMaleChinaAdultForecastingAgedMiddle AgedRetrospective StudiesPrevalenceNeural Networks, ComputerBenchmarkingPredictive Learning ModelsPrediction AlgorithmsEast Asian People

Author keywords

annual time seriescholesterolexpanding-window evaluationhypertensionleakage controlmachine learningmodel comparisonneural forecastingreproducibility

Substances

Cholesterol

Authors and institutions

  • School of Computer Sciences, Shandong Xiehe University, Jinan, Shandong, China.
  • School of Information and Artificial Intelligence, Yangzhou University, Yangzijin Campus, No. 196 Huayang West Road, Yangzhou, Jiangsu, 225127, China, 86 17851975652.

Cite

APA

Lai, W., Subhan, F., Zhou, N., & Khan, J. (2026). Horizon-Aware Forecasting of Hypertension Prevalence and Mean Total Cholesterol in Chinese Men: Retrospective Leakage-Controlled Benchmark Study Comparing Statistical, Machine-Learning, and Neural Models. JMIR Medical Informatics, 14, e99686-e99686. https://doi.org/10.2196/99686

Vancouver

Lai W, Subhan F, Zhou N, Khan J. Horizon-Aware Forecasting of Hypertension Prevalence and Mean Total Cholesterol in Chinese Men: Retrospective Leakage-Controlled Benchmark Study Comparing Statistical, Machine-Learning, and Neural Models. JMIR Medical Informatics. 2026;14:e99686-e99686. doi:10.2196/99686

BibTeX
@article{Lai2026,
  author = {Lai, Weiguo and Subhan, Fazli and Zhou, Nana and Khan, Jawad},
  title = {Horizon-Aware Forecasting of Hypertension Prevalence and Mean Total Cholesterol in Chinese Men: Retrospective Leakage-Controlled Benchmark Study Comparing Statistical, Machine-Learning, and Neural Models},
  journal = {JMIR Medical Informatics},
  year = {2026},
  volume = {14},
  pages = {e99686--e99686},
  doi = {10.2196/99686}
}
Start for free →