Computer Assisted Evaluation of Clinical Data Quality Nordic

  • Slides: 30
Download presentation
Computer Assisted Evaluation of Clinical Data Quality Nordic Casemix Conference 4. 6. 2010 Olafr

Computer Assisted Evaluation of Clinical Data Quality Nordic Casemix Conference 4. 6. 2010 Olafr Steinum, Sequelae AB Seppo Ranta, Datawell Oy

2 Introduction Ø Reported health care information is widely used by authorities • •

2 Introduction Ø Reported health care information is widely used by authorities • • Ø Ø For health care planning For quality analyses For reimbursement For science and research It is of paramount importance that the reported health data are correct and valid Quality assurance is needed • • Coding audits Automatized controls

3 Datawell DRG QA for data quality assurance Ø DRG QA is a Datawell

3 Datawell DRG QA for data quality assurance Ø DRG QA is a Datawell product Ø Uses three different logics for data quality evaluation 1. 2. 3. Ø The • • • Rules and reference databases (e. g. diagnosis codes) used for Nord. DRG grouping Evaluation of the order of diagnoses (which one is the primary diagnosis, which are secondary diagnoses) inferrred from large statistical database (Normalization) Clinical Validation Rulebase (CVRB) created and maintained by Sequelae AB software can be used as part of encoder system for immediate feedback of coding quality standalone system for evaluation of organization data quality benchmarking system comparing several peer organizations

4 Datawell DRG QA Ø DRG QA is a Datawell product **** Ø Uses

4 Datawell DRG QA Ø DRG QA is a Datawell product **** Ø Uses three different logics for data quality evaluation 1. 2. 3. Sequelae AB Rules and reference databases (e. g. diagnosis codes) used for Nord. DRG grouping Is a joint colloboration between Evaluation of the order of diagnoses (which one is the primary diagnosis, which are secondary diagnoses) inferrred from large statistical database (Normalization) Emendor Consulting AB, (Staffan Bryngelsson) Clinical Validation Rulebase (CVRB) created and maintained by Sequelae AB • software can be used as and Olafr part of encoder system for Steinum immediate(dia. Qualos feedback AB) of coding quality Gunnar Henriksson (DRG Henriksson standalone system for evaluation of organization data. AB) quality • benchmarking system comparing several peer organizations Ø The •

5 Datawell DRG QA Ø DRG QA is a Datawell product Ø Uses three

5 Datawell DRG QA Ø DRG QA is a Datawell product Ø Uses three different logics for data quality evaluation 1. 2. 3. Ø The • • • Rules and reference databases (e. g. diagnosis codes) used for Nord. DRG grouping Evaluation of the order of diagnoses (which one is the primary diagnosis, which are secondary diagnoses) inferrred from large statistical database (Normalization) Clinical Validation Rulebase (CVRB) created and maintained by Sequelae AB software can be used as part of encoder system for immediate feedback of coding quality standalone system for evaluation of organization data quality benchmarking system comparing several peer organizations

6 DRG QA – An Example of Indicator Calculation Logic Input data set Pat.

6 DRG QA – An Example of Indicator Calculation Logic Input data set Pat. Id Dg-a Dg-d Pr 13213 LOS Age Dischg Sex 3 134 HOME M 43242 H 10. 1 J 80 WX 101 1 54 HOME N 43242 F 0289 E 756 GD 1 BD 6 3 HOSP N 64243 V 02. 0 2 41 HOME N 34212 O 75. 7 4 34 HOME M MAF 00 Validations Missing principal diagnosis Pat. Id Age not within acceptable limits Local procedure code Dg-a Dg-d Pr 13213 Erroneous ICD-10 code LOS Age Dischg Sex 3 134 HOME M 43242 H 10. 1 J 80 WX 101 1 54 HOME N 43242 F 0289 E 756 GD 1 BD 6 3 HOSP N 64243 V 02. 0 2 41 HOME N 34212 O 75. 7 4 34 HOME M External cause code as principal diagnosis MAF 00 Mismatch of diagnosis and gender

7 DRG QA – An Example of Indicators in DRG QA Database Pat. Id

7 DRG QA – An Example of Indicators in DRG QA Database Pat. Id Dg-a Dg-d Pr 13213 LOS Age Dischg Sex FDG 3 134 HOME M 110 FPR SEXDG FAGE 0 2 43242 H 10. 1 J 80 WX 101 1 54 HOME N 100 110 0 0 43242 F 0289 E 756 GD 1 BD 6 3 HOSP N 100 0 0 64243 V 02. 0 2 41 HOME N 120 0 0 34212 O 75. 7 4 34 HOME M 100 3 0 MAF 00 100 Errors in Diagnosis coding Errors in Age coding Errors in Procedure coding Mismatch of diagnosis and gender

8 DRG QA Pilot Benchmark Database Ø Seven • • DRG QA Database contains

8 DRG QA Pilot Benchmark Database Ø Seven • • DRG QA Database contains patient cases from the Ecomed KPP databases from 2008 Data source: Ecomed KPP used in the 7 hospitals Ø Three • • Hospital Districts in Finland County Councils in Sweden DRG QA Database contains all patient visits and stays from 2008 Data source: Patient Administrative Systems in corresponding county councils Ø Number • • of patient cases Finland n = 4. 928. 113 Sweden n = 4. 332. 206

9 DRG QA Database Formation Process Hospital Districts’ Ecomed KPP databases (FI), or similar

9 DRG QA Database Formation Process Hospital Districts’ Ecomed KPP databases (FI), or similar data retrieval from Patient Administrative Systems (SE) District A District B District C County Council A County Council B County Council C etc. Datawell DRG QA ETL Datawell DRG QA Indicator Calculation Ecomed DRG QA Database Ecomed Analyzer Reporting Analysis of Data Quality • Data format transformations: hospital code common code mappings • Calculation of DRG grouping indicators • DRG normalization • Calculation of CVRB matching • Includes refence population data (1 -year intervals) for standardization

10 Data Analysis for the Current Presentation Ø For this presentation purporses we produced

10 Data Analysis for the Current Presentation Ø For this presentation purporses we produced a sample of the benchmarking database Ø Rules • • • Included hospital stays (coded as ward stays in the source data) Length of stay (LOS) over zero days Excludes psychiatry Ø After • • formation of the present sample applying above mentioned filters the analysis sample consists of 493 689 ward stays in seven Finnish Hospital Distrcits’ hospitals, and 444 255 ward stays in three Swedish County Councils’ hospitals

11 Background Information: Number of Hospital Stays Finnish University Hospitals Swedish County Councils Non-university

11 Background Information: Number of Hospital Stays Finnish University Hospitals Swedish County Councils Non-university (Fi) n = 493 674 (FI); n = 444 255 (SE)

Background Information: Number of Stays in Different Age Groups n = 493 674 (FI);

Background Information: Number of Stays in Different Age Groups n = 493 674 (FI); n = 444 255 (SE) 12

13 Background information: Average Length of Hospital Stays Finnish University Hospitals Swedish County Councils

13 Background information: Average Length of Hospital Stays Finnish University Hospitals Swedish County Councils Non-university (Fi) n = 493 689 (FI); n = 444 255 (SE)

Background information: Average Number of Diagnosis Codes per Stay Non-university (Fi) Finnish University Hospitals

Background information: Average Number of Diagnosis Codes per Stay Non-university (Fi) Finnish University Hospitals n = 493 689 (FI); n = 444 255 (SE) Swedish County Councils 14

Background information: Average Number of Procedure Codes per Stay Finnish University Hospitals Non-university (Fi)

Background information: Average Number of Procedure Codes per Stay Finnish University Hospitals Non-university (Fi) Swedish County Councils n = 493 689 (FI); n = 444 255 (SE) 15

17 DRG CC Cases Finnish University Hospitals Swedish County Councils Non-university (Fi) n =

17 DRG CC Cases Finnish University Hospitals Swedish County Councils Non-university (Fi) n = 120 060 (FI); n = 147 740 (SE)

18 Diabetes as Principal or Secondary Diagnosis Finnish University Hospitals Swedish County Councils Non-university

18 Diabetes as Principal or Secondary Diagnosis Finnish University Hospitals Swedish County Councils Non-university (Fi) n = 20 713 (FI); n = 40 085 (SE)

19 Number of Surgical Procedure Codes (OR property in Nord. DRG) Finnish University Hospitals

19 Number of Surgical Procedure Codes (OR property in Nord. DRG) Finnish University Hospitals Non-university (Fi) n = 493 689 (FI); n = 444 255 (SE) Swedish County Councils

20 Surgical Procedure Codes after Stays in Operative Wards n = 283 633 (FI);

20 Surgical Procedure Codes after Stays in Operative Wards n = 283 633 (FI); n = 203 684 (SE)

21 The classification of diagnosis (ICD-10) ØA • • complex system for collecting data

21 The classification of diagnosis (ICD-10) ØA • • complex system for collecting data for statistics Many axes Many rules • Explicit rules • Rules expressed in the Tabular volume in connection to code categories • Rules assumed, but not explicitely expressed Ø Clinical • validation rule base - CVRB A collection of identified rules

Rate of Z 51. 1 Chemotherapy session as Principal or Secondary diagnosis. Swedish county

Rate of Z 51. 1 Chemotherapy session as Principal or Secondary diagnosis. Swedish county councils 2008 22 Principal dx Secondary dx Data from Swedish National Patient Registry

23 CVRB Rules 2009 1 Not to be used 2 Ought not to be

23 CVRB Rules 2009 1 Not to be used 2 Ought not to be used 17 Not to be used as single code 18 Ought not to be used as single code 3 Not to be used for children < 15 years 4 Ought not to be used for children < 15 yrs 19 Not to be used in combination with other dagger code than code 2 20 Ought not to be used in combination with other dagger code than code 2 5 Not to be used in inpatient care 6 Ought not to be used in inpatient care 7 Not to be combined with code 2 8 Ought not to be combined with code 2 9 Not to be used as principal dx 10 Ought not to be used as principal dx 21 Must be combined with code 2 22 Ought to be combined with code 2 23 Must be combined with an external cause code (chapter XX) for toxic agent 24 Ought to be combined with external cause code (chapter XX) for toxic agent 11 Not to be used as secondary dx 12 Ought not to be used as secondary dx 25 Must be used as secondary dx only in combination with principal dx = code 2 26 Ought to be used as secondary dx only in combination with principal dx = code 2 13 Not to be used as principal dx in combination with code 2 14 Ought not to be used as principal dx in combination with code 2 27 Must be secondary dx when combined with code 2 28 Ought to be secondary dx when in combination with code 2 15 Not to be used as secondary dx in combination with principal dx code 2 16 Ought not to be used as secondary dx In combination with principal dx code 2 © Sequelae AB 29 Must be principal dx in combination with code 2 30 Ought to be principal dx in combination with code 2 31 Rare code 32 Rare code inpatient care 33 Rare code outpatient care

Frequency of CVRB violation in a test database from 10 provinces n = 493

Frequency of CVRB violation in a test database from 10 provinces n = 493 689 (FI); n = 444 255 (SE) 24

CVRB Violation Profile in Finland Sweden, Ward Stays, LOS > 0 days n =

CVRB Violation Profile in Finland Sweden, Ward Stays, LOS > 0 days n = 493 689 (FI); n = 444 255 (SE) 25

26 Distribution of CVRB violation in test database (10 provinces) CVRB violation rule

26 Distribution of CVRB violation in test database (10 provinces) CVRB violation rule

27 CVRB Violation rules Not to be used Ought not to be used Not

27 CVRB Violation rules Not to be used Ought not to be used Not to be used in inpatient care Not to be combined with code 2 Ought not to be combined with code 2 Not to be used as principal dx Ought not to be used as principal dx Not to be used as secondary dx Not to be used as principal dx in combination with code 2 Ought not to be used as secondary dx in combination with principal dx code 2 Ought not to be used as single code Not to be used in combination with other dagger code than code 2 Must be combined with code 2 Ought to be combined with external cause code (chapter XX) for toxic agent Ought to be used as secondary dx only in combination with code 2 Must be secondary dx when in combination with code 2 Rare code inpatient care

28 Example of CVRB Violation rules result in percent of inpatient stays CVRB 10

28 Example of CVRB Violation rules result in percent of inpatient stays CVRB 10 CVRB 11 1 2 3 4 5 Finland 6 7 8 9 10 Sweden CVRB 10 Ought not to be used as principal dx 16 code categories, mainly unspecified codes and codes marking sequelae CVRB 11 Not to be used as secondary dx 273 codes from Ch 21 (Z-codes) only to be used as Reason for admission

Information Process and the Identified Sources of Quality Failure Usability and maintenance of national

Information Process and the Identified Sources of Quality Failure Usability and maintenance of national code systems (ICD, NCSP, DRG etc. ) Code systems Entry of data Processing of data Utilization of data and information Human-Computer interface Feeding of structured information into the PAS Processing rules and logics of the information systems Current transversal study of the information process 29

30 Information Process and Benefits of Datawell DRG QA Code systems Entry of data

30 Information Process and Benefits of Datawell DRG QA Code systems Entry of data Processing of data Utilization of data and information Immediate feedback of coding results to coding personnel Information on organization data quality for focusing education and other corrective actions. Benchmarking data quality with peer organizations. Reports of data quality incorporated with other reporting

31 Data which nobody is using has a quality that nobody wants Thank you!

31 Data which nobody is using has a quality that nobody wants Thank you!