Exploring Dosage in Cost-Effectiveness Using Machine Learning Methods: A Case Study Using the Clinical Data of a Hospital in Suzhou
Author(s)
Yao Zhang, PhD1, Lian Tang, Master2.
1PHD, China Pharmaceutical University, Nanjing, China, 2Suzhou Municipal Hospital, Suzhou, China.
1PHD, China Pharmaceutical University, Nanjing, China, 2Suzhou Municipal Hospital, Suzhou, China.
OBJECTIVES: To explore the cost-effectiveness and heterogeneity in the cost-effectiveness of precise drug administration compared with standard dosage according to instruction manual for severe and critical infection inpatients in China.
METHODS: Using data from clinical real-world dataset, we explored heterogeneity at subgroup levels using the CEAforests approach. The dataset included inpatients in a Chinese hospital from Jiangsu Province. Patients were randomized to receive either precise drug administration or standard dosage according to instruction manual.
RESULTS: In base case, precise drug administration had a total incremental costs of -5,048 CNY (95% CI: -12,041 CNY to 1,944 CNY), incremental QALYs of -0.009 (95% CI: -0.017 to -0.001), INB of 4,873 CNY (95% CI: -2,123 CNY to 11,869 CNY) and ICER of 593,520 CNY per QALY compared with standard dosage. Using a WTP threshold of 95,749 CNY per QALY gain, the result was not so cost-effective. In sensitivity analysis, the result showed that precise administration has approximately a 64% probability of being cost-effective. However as the study conducted heterogeneity analysis with CEAForest algorithm, after the bootstrapping, estimates showed substantial individual heterogeneity with a pronounced benefits. It was observed that younger patients (age ≤78.5 years) demonstrated an 82% probability of cost-effectiveness compared to only 47% in older patients(age >78.5 years), which reveals that the advantage of precise administration may not be that notable in the all-age dataset, however for subgroup analysis, younger group identified statistically significant benefit from precise drug administration (INB=11,579 CNY; 95% CI: 101 CNY, 23,057 CNY vs. INB=9,975 CNY; 95% CI: 104 CNY, 19,846 CNY).
CONCLUSIONS: In the context of severe and critical infections—diseases closely related to patient age and physiological function or so on—machine learning can facilitate the identification of distinct patient subgroups based on clinical features. These subgroups often demonstrate markedly different cost-effectiveness profiles, potentially rendering previously non-cost-effective interventions economically justifiable.
METHODS: Using data from clinical real-world dataset, we explored heterogeneity at subgroup levels using the CEAforests approach. The dataset included inpatients in a Chinese hospital from Jiangsu Province. Patients were randomized to receive either precise drug administration or standard dosage according to instruction manual.
RESULTS: In base case, precise drug administration had a total incremental costs of -5,048 CNY (95% CI: -12,041 CNY to 1,944 CNY), incremental QALYs of -0.009 (95% CI: -0.017 to -0.001), INB of 4,873 CNY (95% CI: -2,123 CNY to 11,869 CNY) and ICER of 593,520 CNY per QALY compared with standard dosage. Using a WTP threshold of 95,749 CNY per QALY gain, the result was not so cost-effective. In sensitivity analysis, the result showed that precise administration has approximately a 64% probability of being cost-effective. However as the study conducted heterogeneity analysis with CEAForest algorithm, after the bootstrapping, estimates showed substantial individual heterogeneity with a pronounced benefits. It was observed that younger patients (age ≤78.5 years) demonstrated an 82% probability of cost-effectiveness compared to only 47% in older patients(age >78.5 years), which reveals that the advantage of precise administration may not be that notable in the all-age dataset, however for subgroup analysis, younger group identified statistically significant benefit from precise drug administration (INB=11,579 CNY; 95% CI: 101 CNY, 23,057 CNY vs. INB=9,975 CNY; 95% CI: 104 CNY, 19,846 CNY).
CONCLUSIONS: In the context of severe and critical infections—diseases closely related to patient age and physiological function or so on—machine learning can facilitate the identification of distinct patient subgroups based on clinical features. These subgroups often demonstrate markedly different cost-effectiveness profiles, potentially rendering previously non-cost-effective interventions economically justifiable.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
EE456
Topic
Clinical Outcomes, Economic Evaluation, Methodological & Statistical Research
Disease
Infectious Disease (non-vaccine), No Additional Disease & Conditions/Specialized Treatment Areas, Personalized & Precision Medicine, Respiratory-Related Disorders (Allergy, Asthma, Smoking, Other Respiratory)