A Literature Review of Empirical Studies of Recommendation Systems

Abstract

OBJECTIVES: Health economics research aims to maximise health gain within a given budget constraint, and is predominantly achieved through the maximisation of the quality-adjusted-life-year (QALY). Historically, the QALY assumes that individuals derive utility only from the consequences of processes and not from the processes themselves. This assumption of narrow consequentialism suggests health care consumers gain no benefit from the consumption of health, but that its value is determined exclusively through health outcomes. There is growing interest, however, in being able to differentiate between treatments with different processes, such as delivery methods, in quantitative terms. This study aimed to identify and review published empirical studies which include a measure of process utility (PU) which could be incorporated into the QALY calculation. METHODS: A literature review was performed in Medline using the search term “process utility”. Additional searches were performed to identify studies not using this specific term, for example: “treatment-related attributes” AND (utilities OR “utility measurement”). RESULTS: Fifteen studies were identified. A variety of approaches were used to detect and measure PU: 4 studies used standard-gamble techniques; 5 studies used time-trade-off techniques; 2 studies used conjoint analysis; 1 study used SF-36 data; 3 studies used waiting-trade-off techniques. Eight studies evaluated treatments: 5 in diabetes and the remainder in gallstone disease, HIV and pain treatments; 6 studies evaluated testing/screening procedures and 1 study evaluated preventative care. All studies identified the presence of PU. Utility decrements ranged from 0.03-0.14 for different drug delivery methods, and 0.0005-0.12 for different dosing strategies. CONCLUSIONS: Although there is no universally accepted methodology for the detection and measurement of PU, evidence to date does support its existence. The wide range in values and approaches suggests that standardisation of methods around the most valid approach is highly desirable. Without this type of consensus, comparability and validity of results will be limited.