| 摘要: | 目的：健保資料庫，由於納保率高且資料完整，已被許多學者廣泛地應用於學術研究及政策制定、分析還有策略規劃之依據，故資料庫資料的完整性及正確性成為當今醫療資訊之重要課題。本研究即在分析某區域醫院膽囊及/或膽道結石病例之診斷與處置編碼品質。方法：擷取92年度某區域醫院住院醫療資料檔之病例共507個案例為研究樣本，首先由二位疾病分類專家就病歷重新再編碼，比對原始疾病分類編碼與專家第二次編碼之一致性，再隨機抽取50本病歷由第三位疾病分類專家進行再測信度測試，資料分析方法有陽性預測值、敏感度、錯誤率百分比、成對T檢定、卡方檢定、及Pearson相關分析等。結果：(1) ICD-9碼為51.22、51.23、及51.41的準確度、敏感度及陽性預測值都達到97.18%以上；(2)整體而言平均編碼錯誤率為13.53%，診斷錯誤率為12.86%，而處置錯誤率為14.38%，診斷與處置錯誤率有顯著性差異(p<0.05)。如檢視類目碼主要診斷及主要處置錯誤率分別為0.95%及0.39%；(3)診斷編碼個數、處置編碼個數及住院日分別與錯誤個數都呈正相關(p<0.001)；(4)開刀房處置與非開刀房處置編碼錯誤有顯著性關聯(X2=115.86 , p<0.05)，在忽略編碼也有顯著性關聯(X2=88.15 , p<0.05)；(5)改變DRG分派之編碼錯誤原因，編錯代碼為58.82%；病歷書寫不完整為11.11%。結論：ICD碼51.22、51.23、51.41、診斷三位碼、處置二位碼，在研究上可靠性高於非開刀房處置碼。病例愈複雜，則錯誤編碼個數愈多，編碼錯誤是DRG分派改變之主要原因。故建議疾病分類人員在病例複雜度較高之病歷編碼更應審慎思考。
Objectives: Health care data in the form of International classification of diseases, 9th version, clinical modification codes are widely used for health care research, but their accuracy remains uncertain. We aimed to evaluate the reliability of coded data regarding the cases of gallbladder and bile duct calculus in a regional hospital. Methods: The subjects were a sample of 507 admissions with cholelithiasis and/ or choledocholithisis from January 1st 2003 to December 31st 2005 in a hospital inpatient claim data. A retrospective audit was carried out, involving comparison of the original coding in the medical records with blindly recording by two senior coders. The main outcome measures were the levels of disagreement between original and auditing coding. We determined the reasons for DRGs discrepancies affer DRG grouping. The descriptive statistics, pair t-test, chi square test, and pearson’s correlation were performed with SPSS 12.0. Results: (1) The accuracy, sensitivity and positive predictive value of ICD-9 codes 51.22, 51.23, and 51.41 were more than 97.18 per cent. (2) Coding errors occurred in 13.53 per cent of all codes, 12.86 per cent of diagnoses and 14.38 procedures. While in cases of coding errors over the categories of the ICD-9 code for principal diagnosis and procedure were 0.95 per cent and 0.39 per cent, respectively. The difference of coding error between diagnosis and procedure was statistically significant (p＜0.05). (3) The frequency of coding error was positively related to LOS, number of diagnosis and procedures codes. (4) Significant associations were found between operating room (OR) and non OR procedure for coding error and undercoded. (5) Coding errors and record’s incomplete documentation accounted for the change in DRG assignment 58.82% and 11.11%, respectively. Conclusions: ICD-9 codes 51.22, 51.23, 51.41, the categories of ICD-9 codes for both diagnoses and procedures were more reliable than ICD-9 codes for non OR procedures for researches. Coding errors were more prevalent in complex cases with multiple codes and the main reason of DRG assignment change was coding errors. Therefore, coders need to work cautiously for coding in more complex cases to ensure that DRG variation is kept to a minimum. |
