全民健康保險糖尿病醫療給付改善方案五年影響評估

Abstract

目的： 中央健保局於2001年10月推行「全民健康保險糖尿病醫療給付改善方案試辦計畫」嘗試從醫療過程面指標，以增加給付的方式，鼓勵醫療提供者給予更好的照護品質。計畫實施後的成效評估目前以短期及區域性比較居多，因此本研究主要目的是分析病醫療給付改善方案實施五年後，對糖尿病患的醫療利用和費用是否改善，以實證資料來分析其中長期影響，提供試辦計畫未來政策改善之參考。 方法： 本研究使用中央健康保險局1999~2006年特別需求醫療申報檔，為次級資料分析。以準實驗研究設計，定義2002年第一次加入試辦計畫之糖尿病患2,134人為研究組，用傾向分數配對出(propensity score matching)從未加入計畫之對照組8,536人，再利用差異中之差異法(difference in difference)及複迴歸以比較前測期(2001年)及追蹤各年(2002~2006年)之糖尿病七項相關檢驗檢查次數變化及門診、急診與住院次數和費用變化情形。本研究再將研究組個案再分為兩組，分別為僅有2002年當年納入計畫者(單一介入組) 581人與連續五年皆納入者(持續介入組)共1,393人。比較其納入試辦計畫時間的長短，對糖尿病患醫療利用和費用影響的差異。 結果： (1)糖尿病患在加入試辦計畫後，完成糖尿病相關七項檢驗檢查的種類數從2001年的3.51次，提高到平均皆有5次以上。但尿液及眼底檢查方面，卻都未達平均每年一次的標準。(2)在差異中的差異分析上，控制了兩組前測的差異後，門診次數、急診次數和急診費用其成長幅度，研究組皆小於對照組。(3)多變項迴歸分析上，研究組除了2006年外，急診次數及急診費用在追蹤各年度都顯著比對照組低；而住院日數及費用研究組各年皆比對照組顯著的低。(4)比較納入計畫時間長短，發現單一介入組造成各項檢驗檢查完成種類數的增加、急診次數及費用的降低、住院日數及費用的降低，這些效果都只能維持當年度或下年度；每年持續納入計畫者，才能維持有其影響。 結論： 本研究發現糖尿病醫療給付改善方案能持續為糖尿病患者帶來糖尿病相關之急診及住院利用的改善，且相關費用的成長得到控制。同一病患在不同年度是否能持續納入試辦計畫，才是影響試辦計畫是否持續發揮效果的關鍵。包裹給付之尿液及眼底檢查率，五年間仍持續未達標準，未來計畫在修訂時應考慮對於未持續進步的指標，在支付上修正或訂立懲罰辦法，同時應加強糖尿病自主管理教育(Diabetes Self-management Education) 落實和獎勵。未來研究可朝病患臨床檢驗值分析及計畫之成本效益分析上進行。

Background: The Bureau of the National Health Insurance (BNHI) implemented 'The Pay for Performance Program for Diabetes' in October 2001. This program aimed to encourage medical provider to supply better quality of health care through additional incentive payments. The effects of this program are mostly evaluated on short-term and regional levels. The main objective of this study is to investigate whether there is continuous improvement on patient’s medical utilization and expenses after doing this program for five years. This study can also provide the long-term effect of this program via evidence-based data and give suggestions for the government in the future. Methods: This study used BNHI claim data during 1999 to 2006 and a quasi-experimental design was used to assess the impact of the program. Diabetic patients who participated in this program fist time in 2001 were identified as the study group and diabetic patients who never participated in this program were defined as the control group. A propensity score method was used to match cases and controls. Finally, 2,134 cases and 8,536 controls were recruited in this study. Difference in difference method and multiple regression models were used to analyze the difference of healthcare utilization and expenses between baseline data in 2001 and follow-up data during 2002 to 2006. Furthermore, the case group was separated into two sub-groups. 581 cases who just participated in the program in 2002 were classed as the single intervention group and 1,393 cases who participated in every year from 2002 to 2006 were classed as the continuous intervention group. We evaluated the program’s effect on medical utilization and expenses in diabetes patients by comparing the length of intervention. Results: Several significant findings of this study were summarized as follows. (1) Patients in the case group tented to have higher frequency in completeness of the 7 important examinations for diabetes. The frequency was from 3.51 times in 2001 to more than about 5 times in following year. However, Urinalysis and Ophthalmoscopic examination did not reach the standard which should be examined at least one time per year. (2)After controlling the baseline difference in the model, we found that patients in the case group had smaller growing slope on the number of outpatient visits、the number of emergency visits and emergency expenditure by using difference in difference analysis.(3)The multiple regression model analysis showed that the case group had fewer emergency visits and emergency expenditure than the control group during observation period except 2006. Similarly, the number of inpatient visits and inpatient expenditure were fewer in the case group than the control group during observation period. (4) We found that the program effect on increasing frequency of completing the 7 important examinations for diabetes, decreasing of emergency visit and expenditure, decreasing of inpatient length and expenditure of single intervention group maintained for one or two years. Oppositely, program effect on continuous intervention group has been preserved during observation period. Conclusions: This study suggest that the 'The Pay for Performance Program for Diabetes' continually have improvement on diabetes-related emergency and inpatient utilization, and make the growing expenditure under control. The key point of the program sustained its effect on patients is that whether the same patient could enroll the program in the successive years or not. We suggest that program should adjust the payment on some indicators, Urinalysis and Ophthalmoscopic examination, which didn’t make progress to the standard over years. The program can be revised on providing incentives to promote the implement of Diabetes Self-management Education. Further study can focus on analyzing the program effect of patient’s clinical examination value and cost-benefit analysis for the program.