# Early Mortality Following Systemic Anticancer Therapy in Lung Cancer: A Bayesian Spatiotemporal Multilevel Analysis

## TL;DR

This study identifies geographic and temporal patterns in early death after lung cancer treatment, highlighting areas and patient groups needing targeted care.

## Contribution

The study introduces a Bayesian spatiotemporal multilevel model to analyze early mortality after SACT, reducing ecological bias and enhancing statistical power.

## Key findings

- Early mortality after SACT varies significantly across regions and over time in Victoria.
- Advanced cancer stages and poor patient health are strongly linked to higher early mortality.
- Multidisciplinary meetings and supportive care reduce early mortality risk.

## Abstract

Despite advances in lung cancer treatment, early mortality following Systemic Anticancer Therapy (SACT) remains a major concern. Identifying patients at high risk of early mortality may inform treatment decision‐making, particularly where SACT may offer limited benefit. We aimed to quantify spatiotemporal patterns of early mortality following SACT and to examine the relative contributions of individual‐ and area‐level risk factors.

We conducted a secondary analysis of data from the Victorian Lung Cancer Registry. Bayesian spatiotemporal multilevel models were used to assess associations between individual‐ and area‐level factors and early mortality following SACT. Adjusted Odds Ratios (AORs) with 95% Credible Interval (Crl) were reported to indicate statistical significance.

Substantial spatiotemporal variation in early mortality following SACT was observed across Victoria. Factors associated with increased odds of early mortality included age ≥ 60 years (AOR = 1.06, 95% Crl: 1.01–1.22), clinical stage II (AOR = 1.15, 95% Crl: 1.01–1.59), stage III (AOR = 1.28, 95% Crl: 1.01–2.05), stage IV (AOR = 3.19, 95% Crl: 2.12–5.07), comorbidity (AOR = 1.12, 95% Crl: 1.01–1.34), and poor performance status (AOR = 3.09, 95% Crl: 2.24–4.24). Presentation at a multidisciplinary meeting (AOR = 0.66, 95% Crl: 0.52–0.88) and receipt of supportive care screening (AOR = 0.53, 95% Crl: 0.41–0.69) were associated with reduced odds of early mortality.

Marked spatiotemporal variation in early mortality following SACT highlights the need for targeted, risk‐informed treatment strategies, particularly for patients with advanced disease, comorbidities, poor performance status, and those residing in high‐risk areas.

This study enhances statistical power and addresses ecological biases by incorporating joint cross‐classifications of individual‐level covariates within a Bayesian spatiotemporal multilevel model. After accounting for both spatial and temporal random effects using the extended Hierarchical Related Regression framework, significant geographic variation in early mortality following Systemic Anticancer Therapy was observed in Victoria. These reported maps indicate potential areas for further research and targeted interventions.

Geographic and temporal variation in early mortality following Systemic Anticancer Therapy (SACT) identifies regions with disproportionately high mortality. Characterizing these patterns enables targeted interventions and more effective allocation of healthcare resources to populations at greatest risk. Spatiotemporal analysis also provides insights into underlying drivers of early mortality, treatment effectiveness, and opportunities to improve patient care and supportive services.

Methodologically, this study enhances statistical power and reduces ecological bias by jointly modeling individual‐level covariates within a Bayesian spatiotemporal multilevel framework. By incorporating spatial and temporal random effects using an extended Hierarchical Related Regression approach, we demonstrate substantial spatiotemporal variation in early mortality following SACT across Victoria. The resulting risk maps highlight priority areas for further investigation and targeted interventions.