Optimising resource allocation in public hospitals through parallel network DEA models

Marta Alexandra dos Santos Abreu

Key information

Authors:

Marta Alexandra dos Santos Abreu (Marta Alexandra dos Santos Abreu)

Supervisors:

Miguel Brau Canadas Alves Pereira (Miguel Alves Pereira)

Published in

December 2, 2025

Abstract

The Portuguese National Health System (SNS) is under increasing pressure due to demographic changes, staff shortages, and rising demand for high-quality care, making hospital efficiency assessment particularly relevant. Recent studies have proposed methods to measure hospital inefficiency by considering both inefficiencies arising from internal processes and those introduced by suboptimal resource allocation decisions. However, detailed information on the quantities of inputs and outputs tied or allocatable to each sub-unit is often unavailable, making it difficult to apply such methods when data are absent. Therefore, this thesis applies an existing Parallel Network Data Envelopment Analysis model and contributes to this field by integrating a Monte Carlo simulation to evaluate the efficiency of Portuguese public hospitals that compose Local Health Units. To estimate the quantities of inputs and outputs used and produced by each process, the simulation relied on aggregated 2024 data related to nine variables retrieved from the Portuguese SNS, together with average proportions derived from previous studies and a few available reports. Although more work is required to definitively prove the suitability of the proposed approach, the results obtained indicate that data scarcity can significantly restrict the discriminatory power and practical applicability of the adopted model. Overall, this study highlights the need for more granular and transparent healthcare data from the government to ensure valid and policy-relevant performance assessments, while opening the discussion on the development and refinement of efficiency evaluation methodologies that may, in the future, need to explicitly account for data scarcity.