糖尿病病患健康狀況差異與糖尿病醫療給付改善方案初步成效之分析

Abstract

全民健康保險為引導醫療提供者提供糖尿病患完整性照護服務正式實施「糖尿病醫療給付改善方案」，以支付額外獎金予醫療提供者並連結其績效作為支付依據。此方案希望建立以整體性與連續性為原則之新支付制度，朝簡單、可行性高之方向設計支付誘因，鼓勵院所參與共同照護網認證及運作，同時建立品質監控機制，使支付項目與品質監控指標相扣連。 醫療提供者選擇病患的行為多發生於前瞻性支付制度下，醫療提供者感到成本控制的壓力時，醫療提供者常藉由選擇資源耗用較低之病患以降低其服務成本，自然會出現病患選樣偏差的情形，本研究有兩個目的:1.在各層級別此方案下，收案病人是否較健康且醫療費用或利用較少。2.探討此方案對於病患在後續的病情控制及醫療費用或利用上變化的影響。 本研究之資料來源為全民健康保險研究資料庫，研究對象為2001年至2003年參與糖尿病醫療給付改善方案，收案人數超過30人之醫療院所下收案及未收案的病人。總共有226間醫療院所，收案組為77,482人，未收案組為128,359人。研究時間點為樣本收案前一年、收案後一年和收案後二年，以Logistic迴歸分析收案前一年疾病合併症嚴重度(CCI)、急診費用、住院費用和住院兩次以上對收案與否的影響。研究結果為:CCI越高、急診費用越高及住院次數越多者越不易加入此方案。各層級中只有地區醫院沒有此顯著現象。 再以廣義估計方程式(GEE)分析各層級醫療院所收案後其收案組和未收案組病患在CCI、總醫療費用、門診費用、急診費用、住院費用、住院日數上的變化的差異。研究結果為: 整體而言在收案後一年收案者在門診次數、急診費用、住院日數、住院機率的成長比未收案者少，門診費用成長比未收案者多。收案後二年收案者在CCI、門診次數、急診費用、住院費用、住院日數、住院機率成長比未收案者少，門診費用成長比未收案者多。 醫學中心在收案後一年實驗組的病人在CCI、門診次數、急診費用、住院機率的成長比對照組成長的幅度小。收案後二年收案者之門診次數、住院費用和住院利用的成長較未收案者少。 區域醫院在收案後一年收案者的門診次數、急診費用、急診次數、住院費用、住院日數、住院機率成長比未收案者少。收案後二年收案者的CCI、住院費用、住院日數、住院機率成長比未收案者少。 地區醫院收案後一年和後二年收案者只在住院機率的成長較未收案者少，其他指標皆無差異。 基層診所在收案後一年和後二年之急診次數和住院機率的成長較未收案者多，其他指標則無明顯差異。 由本研究可看出，「糖尿病醫療給付改善方案」在不同層級的醫療院所實施的成效並不一致，一方面是因為一開始收案之病患其健康狀況就不同，另一方面為各層級醫院面對此方案之積極程度也不一。本研究建議其支付模式應考慮各層級醫療院所之醫療資源不同而有不同的支付模式，或是支付模式應引入適當之風險校正模式，中央健保局應繼續且積極推動全國醫療機構加入，以期給予糖尿病病人更完善的照護，讓更多病患受惠。

Pay-for-performance program for diabetes has been put into practice sinece November 1,2001. Goal of the program conducts health care providers to supply the integrity of diabetes care services, thereby enhancing quality of medical care and control the disease. Method of the program pay bonus to health care providers, and link their performance as a basis for payment. Health care providers may convern about the profit under the prospective payment. The physicians possible select patients with lower expenditure and better health statuse. This situation will seriously distort the intention of this programme for enhancing the quality of care and disease control.The purpose of this research is to evaluate the selection bias of P4P programe for diabetes and to measure the chage of healthe statuse, expenditure and utilization of enrollees , compared with non-enrollees, by 2 years data after enrollment. The samples were from National Health Insurance Research Database of claimed data .This study group were the diabetes who were enrolled in the programe during 2001 to 2003, and the control group were the non-enrolled diabetes during 2001 to 2005, received medical care for diabetes at enrollment hospitals.Logistic regression was used to analyse the correlation between the possibility of enrollment and prior CCI, emergency expenditure, inpatient expenditure and utilization. Generalized estimating equation (GEE) model was used to meature the tend of post-enrollment CCI, totel expenditure, outpatient expenditure, outpatient visits, emergency expenditure, emergency visits, inpatient expenditure, inpatient visits, the length of stay and the probability of inpatient stay. All of the measurments were analysed by each level of Taiwan medical institutions. The main findings of this study: 1. After controlling hospital’ and patients’ characteristics,logistic regression analysis indicated favorable selection associated increase as prior CCI decreases, emergency expenditure decreases , and more than 1 inpatient services were. There was no significant favorable selection in the local hospitals. 2.The GEE model indicated that the increase of outpatient visit, emergency expenditure, length of stay and the probability of inpatient stay among enrollees were less than non-enrollees in the first year postenrollment.In the second year posenrollment the increase of CCI, outpatient visit, emergency expenditure, inpatient expenditure, length of stay and the probability of inpatient stay among enrollees were less than non-enrollees. 3.In the medical centers, the increase of CCI, outpatient visit, emergency expenditure and the probability of inpatient stay among enrollees were less than non-enrollees in the first year postenrollment. In the second year postenrollment, the increase of outpatient visit, inpatient expenditure and inpatient utilization among enrollees were less than non-enrollees. In the regional hospitals, the increase of outpatient visit, emergency expenditure and visit, inpatient expenditure, leghth of inpatient stay and the probability of inpatient stay among enrollees were less than non-enrollees in the first year postenrollment. In the second year postenrollment, the increase of CCI, inpatient expenditure and inpatient utilization among enrollees were less than non-enrollees. In the district hospitals, the increases of the probability of inpatient stay were less among enrollees than non-enrollees in the first and second year postenrollment. In the clinics, the increases of emergency visit and the probability of inpatient stay among enrollees were more than non-enrollees. Based on the results of this study, a few suggestions were proposed as the following: 1. To suggest this program use a severity of illness adjusted model. 2. To limit the enrollment date in one year could avoid the bias from the environment. 2. 3 year may be too short of study period to gauge the inpact of the closures.The longer study period, the more will be found.