全民健康保險研究資料庫開發與應用研討會 全民健康保險研究資料庫在急診醫療利用分析之應用 翁瑞宏 嘉南藥理科技大學醫管系暨碩士班助理教授 黃金安 台中榮民總醫院急診醫學科主任 2009 年 9 月 3 日
Outline Jin-An Huang, Rhay-Hung Weng, Wen-Chen Tsai, Wei-Hsiung Hu and Dar-Yu Yang, Analysis of Emergency Department Utilization by Elderly Patients under National Health Insurance, Kaohsiung Journal of Medical Science, 2003; 19(3):113-120. Jin-An Huang, Rhay-Hung Weng, Chi-Shiun Lai, Jer-San Hu, Exploring Medical Utilization Patterns of Emergency Department Users, Journal of the Formosan Medical Association, 2008; 107(2): 119-128. 全民健康保險資料庫的特點與限制 對未來研究者的建議 2
3 Analysis of Emergency Department Utilization by Elderly Patients under National Health Insurance
Introduction It is projected that by 2011, 10% of the total population will be over the age of 65 years Elderly persons are thought to use emergency departments (EDs) disproportionately While there is an increasing demand for emergency medical services by elderly persons, we still know little about the specifics of the use of emergency medical care by elderly patients in Taiwan 4
研究問題 老人急診醫療服務利用的特性為何? 老人 (65 歲以上 ) 與非老人 ( 年輕的成人,15-64 歲 ) 對於急診醫療服務利用是否有差異? 5
Methods Data sources and processing NHIRD for 2000 基本資料檔 中的 醫事機構基本資料檔 原始資料檔 中的 門診處方及治療明細檔 Statistical analysis Frequency distribution analysis χ2 test Student s t-test 6
Methods Subjects: ED visits of 12 medical center in Taiwan Five medical centers that did not claim their physician fees by triage categories and individuals whose data were unclear or missing were excluded A total of 519,003 ED visits A total of 394,954 ED users 7
Data Sources and Processing 醫事機構代號 (HOSP_ID) ( 加密 ) 醫事機構基本檔 (HOSB) 特約類別 (HOSP_CONT_TYPE) 縣市區碼 (AREA_NO_H) 醫事機構代號 (HOSP_ID) ( 加密 ) 身分證統一編號 (ID) ( 加密 ) 出生日期 (ID_BIRTHDAY) Age 姓別 (ID_SEX) Outpatient Visits 案件分類 (CASE_TYPE) ED Visits 門診處方及治療明細檔 (CD) 特定治療項目代號 ( 一 )~( 四 ) (CURE_ITEM_No.1~No.4) 部分負擔代號 (PART_NO) 就醫日期 (FUNC_DATE) 治療結束日期 (TREAT_END_DATE) Chronic Illness Major Illness Co-Payment ED stay 國際疾病分類號 (ACODE_ICD9_1) 8 用藥明細金額小計 (DRUG_AMT) 藥事服務費 (DSVC_AMT) 診察費 (DIAG_AMT) 合計金額 (T_AMT) Drug Charge Physician Charge Emergency Triage Total Charge Treatmen t Charge
9 Results
Characteristics of ED Visitors by Patient Group Items Non-elderly Elderly p value (n = 298,567) (n = 96,387) n (%) n (%) Age, mean (years) 36.3 74.7 Sex < 0.001 C Female 157,136 (52.6) 37,051 (38.4) Male 141,431 (47.7) 59,336 (61.6) Chronic illness < 0.001 C No 274,788 (92.0) 84,924 (88.1) Yes 23,779 (8.0) 11,463 (11.9) Major illness < 0.001 C No 286,404 (95.9) 89,894 (93.3) Yes 12,163 (4.1) 6,493 (6.7) Co-payment < 0.001 C No 17,371 (5.8) 37,103 (38.5) Yes 281,196 (94.2) 59,284 (61.5) Frequent users of outpatient services < 0.001 C No 269,139 (90.1) 69,041 (71.6) Yes 29,428 (9.9) 27,346 (28.4) Total ED visits 371,305 (71.5) 147,698 (28.5) < 0.001 t Average visits per year 1.24 1.53 < 0.001 t C Chi-squared test; t Student s t-test. 10
Triage Category and Length of ED Stay by Patient Group Non-elderly Elderly p value* (n = 371,305) (n = 147,698) n (%) n (%) Triage category < 0.001 1 20,226 (5.4) 18,056 (12.2) 2 142,212 (38.3) 70,601 (47.8) 3 205,683 (55.4) 58,366 (39.5) 4 3,184 (0.9) 675 (0.5) ED stay < 0.001 < 1 day 353,771 (95.3) 127,079 (86.0) 1 day 17,534 (4.7) 20,619 (14.0) *Chi-squared test; 1 = immediately life-threatening emergency; 2 = potentially life-threatening emergency; 3 = urgent; 4 = non-urgent. 11
ED Charge per Visit by Patient Group Charge category Non-elderly Elderly p value* (n = 371,305) (n = 147,698) Physician 373 ± 97 NT$ 436 ± 212 NT$ < 0.001 Treatment 1,836 ± 3,400 NT$ 3,667 ± 4,841 NT$ < 0.001 Drug 528 ± 4,054 NT$ 656 ± 2,163 NT$ < 0.001 Total charge 2,779 ± 5,533 NT$ 4,814 ± 6,046 NT$ < 0.001 *Student s t-test. 12
Ranking of the 10 Leading Diagnoses for the Elderly and Non-Elderly Elderly (Total = 147,698) Non-elderly (Total= 371,305) Diagnosis n (%) Diagnosis n (%) Injury 14,314 (9.7) Injury 77,890 (21.0) Cancer 10,564 (7.2) Acute upper respiratory tract infection 27,416 (7.4) Ischemic heart disease 6,649 (4.5) Abdominal pain 26,072 (7.0) Cerebrovascular disease 6,500 (4.4) Acute gastroenteritis 23,945 (6.4) Abdominal pain 6,371 (4.3) Cancer 11,638 (3.1) Hypertensive disease 5,210 (3.5) Gastritis and duodenitis 8,941 (2.4) Chronic obstructive pulmonary disease 4,686 (3.2) Urinary tract infection 8,695 (2.3) Urinary tract infection 4,625 (3.1) Urolithiasis 7,893 (2.1) Pneumonia 4,164 (2.8) Nonspecific chest pain 7,250 (2.0) Diabetes mellitus 3,671 (2.5) Headache 6,588 (1.8) 13
Conclusion ED utilization by the elderly differs significantly from that of younger adults. Elderly patients present to the ED more frequently than their proportion of the population The elderly seeking emergency care have a higher level of acuity, spend more time in the ED and use more ED resources than the non-elderly. The principle diagnoses are strikingly different between elderly and non-elderly patients 14
15 Exploring Medical Utilization Patterns of Emergency Department Users
Introduction NHI in Taiwan covers 99% of citizens and is contracted with 91.86% of medical institutions Frequent ED users still comprise 3.5% of the total number of ED patients. Frequent ED users might have a regular source of care but prefer to use the ED 16 Frequent ED users were more likely to visit primary physicians, visit the OPD, and be admitted to the hospital; that is, frequent use of the hospital ED is an indicator of high use of other health care services(hansagi et al., 2001; Huang et al., 2003).
Introduction The past literature cannot provide a complete picture of the comprehensive utilization patterns of ED patients 17
研究問題 急診醫療服務利用和醫院門診醫療 醫院住院醫療以及基層醫療照護服務利用之間, 是否存在某種關係? 經常利用急診醫療者的固定就醫場所是否只有急診部門? 抑或他們也高度利用了其它醫療服務? 急診醫療服務使用者是否有不同的醫療利用型態? 18
Methods Data sources and processing NHIRD for 2004 基本資料檔 中的 醫事機構基本資料檔 抽樣歸人檔 中的 門診處方及治療明細檔 與 住院醫療費用清單明細檔 Statistical analysis χ2 test Multiple logistic regression analysis Two-stage cluster analysis One-Way ANOVA 19
Methods Subjects: A total of 6,775 ED users Patients were categorized into four ED E1=1 ED visit, E2=2 ED visits, E3=3 ED visits, and E4=4 or more ED visits. ED class E4 is the so-called frequent ED users class. High health care use was dichotomized arbitrarily on the basis of the 75th percentile as follows: 11 or more hospital OPD visits 1 or more hospital admissions 13 or more physician visits in primary care 20
Data Sources and Processing 基本資料檔 醫事機構基本檔 (HOSB) 醫事機構代號 ( 加密 ) (HOSP_ID) 特約類別 (HOSP_CONT_TYPE) 醫事機構評鑑等級 醫事機構代號 ( 加密 ) (HOSP_ID) 門診處方及治療明細檔 (CD) 身分證統一編號 ( 加密 ) (ID) 案件分類 (CASE_TYPE) 急診就醫次數 門診就醫次數 抽樣歸人檔 出生日期 (ID_BIRTHDAY) 姓別 (ID_SEX) 部分負擔代號 (PART_NO) 特定治療項目代 ( 一 )~( 四 ) (CURE_ITEM_No.1~No.4) 年齡 重大傷病 慢性病 住院醫療費用清單明細檔 (DD) 醫事機構代號 ( 加密 ) (HOSP_ID) 21 身分證統一編號 ( 加密 ) (ID) 住院就醫次數
22 Results
Patient Characteristics by ED Class 23
Logistic Regression Models High use of hospital OPD Hospital Admissions High use of primary care ED class OR 95% CI p value OR 95% CI p value OR 95% CI p value E1 (1) 1.00 - - 1.00 - - 1.00 - - E2 (2) 1.95 1.66-2.28 <0.001 2.15 1.85-2.50 <0.001 1.34 1.15-1.55 <0.001 E3 (3) 4.08 3.14-5.31 <0.001 3.33 2.61-4.26 <0.001 1.85 1.45-2.37 <0.001 E4 ( 4) 10.30 7.53-14.10 <0.001 4.90 3.74-6.43 <0.001 1.51 1.14-1.99 0.004 *adjusted for age (10-year intervals) and gender 24
Hierarchical Cluster Analysis Cluster variable profiles Clustering variable mean values Cluster ED visits (n) Hospital OPD Hospital Primary care Cluster visits (n) admissions (n) visits (n) size 1 1.38 7.20 0.34 8.74 6,479 2 2.18 50.30 1.30 5.08 109 3 2.19 32.90 1.17 35.60 144 4 2.00 15.91 1.02 65.35 43 Significance testing of differences between cluster centers Variable Cluster mean Degrees of Error mean square Freedom square df F value ED visits (n) 58.488 3 0.767 6,771 76.247* Hospital OPD visits (n) 96,488.007 3 61.778 6,771 1,561.863* Hospital admissions (n) 70.408 3 0.689 6,771 102.199* Primary care visits (n) 79,326.734 3 90.501 6,771 876.525* *p<0.001 25
Nonhierarchical Cluster Analysis Clustering variable mean values Cluster ED visits (n) Hospital OPD Hospital Primary care Cluster visits (n) admissions (n) visits (n) size Low health care users 1.29 4.83 0.24 3.97 4,195 Hospital fans 1.85 30.26 1.13 8.30 808 Primary care favorers 1.39 6.10 0.29 18.59 1,437 High health care users 1.77 12.14 0.60 44.94 335 Statistical significance of cluster differences F value 113.183 3,983.042 289.881 8,289.843 p value <0.001 <0.001 <0.001 <0.001 26
Profile of Four Medical Utilization Patterns of ED Users* Low health care users Hospital fans Primary care favorers High health care users Total Total 4,195 (61.9) 808 (11.9) 1,437 (21.2) 335 (4.9) 6,775 (100.0) Age (y) Gender 0-14 570 (13.6) 26 ( 3.2) 438 (30.5) 76 (22.7) 1,110 (16.4) 15-44 2,493 (59.4) 152 (18.8) 514 (35.8) 53 (15.8) 3,212 (47.4) 45-64 773 (18.4) 239 (29.6) 293 (20.4) 75 (22.4) 1,380 (20.4) 65 359 ( 8.6) 391 (48.4) 192 (13.3) 131 (39.1) 1,073 (15.8) Male 2,386 (56.9) 358 (47.6) 631 (43.9) 166 (49.6) 3,568 (52.7) Female 1,809 (43.1) 423 (52.4) 806 (56.1) 169 (50.4) 3,207 (47.3) Frequent ED use 82 ( 2.0) 93 (11.5) 53 ( 3.7) 32 ( 9.6) 260 ( 3.8) High OPD visits 556 (13.3) 808(100.0) 305 (21.2) 154 (46.0) 1,823 (26.9) Hospital admissions 727 (17.3) 447 (55.3) 288 (20.0) 123 (36.7) 1,585 (23.4) High PC visits 0( 0.0) 191 (23.6) 1,248 (86.8) 335(100.0) 1,774 (26.2) Chronic illness 398 ( 9.5) 117 (14.5) 108 ( 7.5) 23 (6.9) 646 ( 9.5) Major illness 80 ( 1.9) 174 (21.5) 39 ( 2.7) 25 ( 7.5) 318 ( 4.7) *Data are presented as n (%); as designated by the Bureau of National Health Insurance, Taiwan. 27
Conclusion Frequent ED users did not lack a regular source of primary care, nor did they identify the ED as regular source of care Frequent ED users were also high users of other health care services Frequent ED users were about 10 times more likely to be high users of hospital OPD services ( 11 visits), five times more likely to receive in-hospital care, and 1.5 times more likely to be high users of primary care ( 13 visits) (reference group: 1 ED visit). 28
Conclusion There is an intercategory relationship between emergency medical services and other medical services ED users have four distinctive medical utilization patterns: low health care users, hospital fans, primary care favorers, and high health care users 29
全民健康保險資料庫的特點 就醫資訊 30
全民健康保險資料庫的限制 以健保申報為主的資料庫 申報資料品質大不同 與臨床資訊未必一致, 應用於 outcome research 時較為受限 人口學或社經資訊較甚為缺乏 僅能呈現醫療利用型態, 無法顯示病患在醫療決策或利用過程中的動機 偏好等心理資訊 未包含非健保特約的醫事機構與未納保的民眾 31
全民健康保險資料庫的限制 申報金額非 醫事機構醫療成本 用於醫事機構營運分析之研究較為受限 加密資料, 無法與外部其他資料庫串檔 二代加密後更加困難 32
對未來研究者的建議 創意研究主題更勝艱深統計分析 方向明確, 才不會迷路 妥善清檔再分析, 避免一直做白工 醫療與管理專業的結合," 質 量 " 皆可提升 實務現況多了解, 資料使用才正確 33
Appreciate your listening! Appreciate your listening! Thank you with all my heart Thank you with all my heart 34