A PILOT STUDY ON APPLYING PRESCRIPTION INFORMATION TO RISK ADJUSTMENT MODELS IN TAIWAN
Resource
台灣公共衛生雜誌 v.25 n.3 pp.189-200
Journal
台灣公共衛生雜誌,v.25
Journal Issue
n.3
Pages
189-200
Date Issued
2006
Date
2006
Author(s)
HSIEH, MENG-FU
CHANG, RAY-E
Abstract
目標：風險校正可避免論人計酬制度所帶來就醫公平與風險選擇的問題。診斷和處方資訊包含豐富的臨床訊息，利用其進行風險校正近年來國際發展蓬勃。相較於國際，過去研究顯示國內風險校正因子具有相當之預測力及發展性。全民健保藥品費用約佔門診醫療支出的1/3 ，其中慢性疾病藥費又佔相當比率。故本研究欲探討處方資訊對台灣風險校正模型預測力之改善。方法：利用國家衛生研究院全民健保資料承保資料抽樣歸人檔資料，選取於2000年全年在保且2001年至少一個月在保之保險對象共167,518人為研究樣本。分析採加權最小平方迴歸模型，運用前一年人口統計因子、診斷基礎因子、用藥處方因子以及附加處方資訊診斷因子，建立4個風險校正模型，並以跨樣本PR^ 2和預測比評估不同模型之預測力。結果：各模型中，在整體預測力上，人口統計模型偏低，其餘三模型均有大幅提升，而以附加處方資訊診斷模型之預測力最高。對特定群體的預測力，人口統計模型的預測比表現在4個模型中仍是最差，其餘三個模型之預測準確程度又以附加處方資訊診斷模型和TASGs模型表現較佳，而對特定疾病群附加處方資訊診斷模型又較TASGs模型為佳。結論：較多的醫療資訊，可使風險校正模型具有更好的預測能力。增加處方資訊確可更進一步提升風險校正模型預測能力。同時在特定之慢性病族群上，也具有更精確之預測能力，此有助於風險選擇的抑制並達成效率目標。未來應可針對我國用藥型態修正，並與更多診斷資訊整合。 Objective: Using risk adjustment to set premium subsidies can reduce risk selection and ensure the equity of access to care. Diagnosis-based adjusters and prescribed drugs adjusters have attracted the research attention due to their recent rich clinical messages recently and have thus been studied intensively. Results of previous studies in Taiwan demonstrated the outstanding predictability of risk adjusters using Taiwan's NHI data. The medical expense structure of Taiwan's NHI shows that one-third of outpatient medical expenses is spent on prescription drugs, and chronic disease prescriptions account for the main portion. This study intends to investigate the improvement inthe predictability of risk adjustment models in Taiwan by introducing prescription information. Methods: From a database containing a panel of 200,000 individuals randomly selected from the NHI's beneficiaries, 167,518 individuals who had a full twelve months of eligibility in 2000 and were still enrolled on January 1, 2001 were selected as the study sample. Through weighted least square regression, demographic adjusters, diagnosis-based adjusters, prescription adjusters and diagnosis-based adjusters attaching prescription information were employed to establish four risk adjustment models. PR^2 and predictive ratio were employed to evaluate the predictability of various models through cross-sample validation. Results: The predictability of the demographic model was markedly inferior to the other three models. Of the other three models, the model with prescription and diagnosis information outperformed the other two. Although the overall predictability of the model with prescription and diagnosis information is slightly higher than that of the TASGs model , it has more accurate predictions in terms of expenditures of several subgroups, particularly the subgroups with chronic diseases. Conclusions: More clinical information improves the predictability of risk adjustment models, and prescription information does contribute to the improvement on the predictability of Taiwan's risk adjustment models, in particular on the predictions concerning the expenditures of individuals with chronic diseases. This should be of assistance in obtaining the objectives of reducing the selection incentive and improving efficiency. Future studies should incorporate Taiwan's practice patterns of prescribing drugs as well as more integration with diagnosis information.
Subjects
風險校正
處方用藥
診斷
論人計酬
risk adjustment
prescription
Type
journal article