Hurdle Model in Multi-Attribute Utility Data on Epidural Analgesia in Labor

• Main Authors: Cheng-Hsi Chang, 張正熹
• Other Authors: Hsiu-Hsi Chen
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/90206271730725394041

碩士 === 國立臺灣大學 === 流行病學與預防醫學研究所 === 100 === Background Clinical decision making is always affected by a constellation of factors. The elucidation of the relationships of these factors to the final decision is of paramount importance. Multi-attribute utility (MAU) theory has been successfully used to model the decision process of parturients in Taiwan about EA, and a hierarchical questionnaire on the basis of MAU theory can predict pre-labor decision and final decision of parturients. However, the two-stage property of MAU data needs specific statistical methods, such as hurdle model for analysis. Objectives The current study aims to develop MAU-based hurdle model, to assess factors related to whether to agree and what extent of agreement/disagreement to each question answered by clients, and to apply the developed MAU-based hurdle model to the data from a previously developed 20-item multi-dimensional questionnaire on attitude toward labor EA underpinning MAU theory. Materials and methods We used data collected in a previous study in a medical center in Taipei. Study participants enrolled from January to April 2006, were of mixed parity, and were native speakers of Chinese with uncomplicated singleton pregnancy. All eligible parturients during the study period were invited to participate after delivery. The exclusion criteria were elective cesarean delivery, emergency cesarean delivery without sufficient time to consider whether to use epidural analgesia for labor, contraindications to epidural analgesia (e.g., coagulopathy, local infection), and history of psychiatric disorders or substance abuse. Of 167 parturients responding to the MAU-based questionnaire, 151 participants who completed all questions were enrolled in this study. The response to each question in this questionnaire can be divided into three parts: (1) agree or disagree to each question (binary outcome), (2) the intensity of agreement on a 1 to 10 scale, (3) the intensity of objection on a 1 to 10 scale. Then the standard hurdle model was modified to examine the interactive effect of basic characteristics of parturients on the three parts of each question using Bayesian Markov Chain Monte Carlo methods. Partial and complete hurdle models were used to predict individual response to this MAU-based questionnaire and the final decision of parturients. Receiver operating characteristics (ROC) curves were also used to assess predictive validity. We recruited another 101 parturients in the same medical center to do external validation. Result Of 151 participants (75 EA and 76 non-EA groups), parturients in the EA group had significantly higher education level (rate of university or above: 63% vs. 44%, P = 0.001), higher family income per month (rate of more than sixty thousand NT: 63% vs. 46%, P = 0.04), and higher proportion of support and lower proportion of discourage from family members or friends (P < 0.001). There were also more primiparae in the EA group compared with non-EA group (76% vs. 46%, P < 0.001). The first part of our modified hurdle model showed only 16 out of 20 questions could be modeled by basic characteristics. Basic characteristics had significant effects on second part or third part consisted of 15 questions. Significant net effect of basic characteristics between second and third part could be identified in 11 questions. The area under ROC curve equaled to 0.73 (95% confidence interval = 0.65, 0.81) for partial hurdle model, and 0.75 (95% confidence interval = 0.68, 0.83) for complete hurdle model. External validation leads to the area under ROC curve of both partial and complete hurdle models equal to 0.64 (95% confidence interval = 0.53, 0.75). Conclusion We applied a hurdle model to analyze a MAU-based questionnaire originally developed to measure attitude toward labor EA. This novel method is able to highlight differential effects of basic characteristics on parturients who accept and refuse EA. This analysis enables us to have a better understanding of how basic characteristics influence the response to each question and the final decision.