大型醫院建築維護成本分析－－以國立台灣大學醫學院附設醫院為例

Analysis of maintenance cost for large hospital buildings – a case study of National Taiwan University Hospital

醫院建築逐漸朝大型化與複雜化發展，其維護與修繕的迫切度也遠高於一般建築，而醫院的維持良好運作高度仰賴著硬體設備的維護管理。然而，在每年國家補助預算遞減的情況下，即便如國家醫學中心都必須自負盈虧。因此，如何在有限的預算之下，尋找出關鍵維護項目，進而控管大宗維護金額花費，以更有效率的管理醫院建築，是本研究所探討的課題。本研究將維護成本依其類型分為定期維護、損壞修繕與需求變更等三大類，分別探討台大醫院東西址院區在此三大類型中的關鍵維護項目與細項，並比較不同生命週期階段的維護行為差異。本研究以灰色預測模型為架構，整合指數級數與自組織映射網路分別進行規則殘差與不規則殘差修正，建立有效的醫院建築維護預算預測模型，預估未來一段時間內的預算需求，以提供醫院工務室於預算編列之參考。

With the increasing demands on healthcare facilities and services, hospital buildings have gradually matured to become large capacity and numerous complex facilities. A hospital is a more complex system of interacting environments, in comparison with other kinds of buildings. The performance of hospital buildings depends highly upon the efficiency of maintenance execution. However, with ever-growing demands and decreasing budgets, facilities managers of hospitals must ensure facilities properly maintained without compromising their performance. In order to provide an appropriate environment in such a high-complex building, key maintenance items should be carefully analyzed and executed as well as decisions for proper maintenance budget allocation. This study classifies the maintenance cost into three categories, including periodic maintenance, repair, and demand change, and identifies the key maintenance items and their sub-items of the National Taiwan University Hospital (NTUH). Besides, this study also compares the differences between NTUH and literatures based on the maintenance cost unit. This study not only identifies the key maintenance items but also establishes an effective prediction model for hospital building maintenance, evaluating the tendency of maintenance cost to provide important information for the facility managers. In the aspect of prediction model, this study adopts the grey forecasting model as the main structure and then integrates the exponential series and self-organizing mapping to recognize the regular and irregular residual errors, respectively. Results show that the proposed prediction model can catch the tendency of building maintenance cost effectively.