1. A SAS® Format Catalog for ICD-9/ICD-10 Diagnoses and Procedures: Data Research Example and Custom Reporting Example September 20, 2017   Robert Richard Springborn, Ph.D. Office of Statewide Health Planning and Development Information Services Division, Healthcare Analytics Branch, Clinical Data Unit Sacramento, California

2. Motivation for Topic
Office of Statewide Health Planning & Development (OSHPD) provides administrative data containing ICD-9 and ICD-10 diagnosis and procedure codes to users worldwide including: researchers; health care providers; health care insurance providers; and media.
On October 1, ,614 ICD-9-CM diagnosis and procedure codes were replaced by 140,746 ICD-10-CM codes which have many benefits including: More accurate reflections of patients’ clinical complexity and severity of illness; and ability to better identify high-risk patients.
A SAS format catalog which adds ICD-9 and ICD-10 diagnosis and procedure code descriptions to administrative data, would greatly benefit all users and encourage new uses of administrative data by reducing the need for large ICD-9 and ICD-10 look-up manuals.
Today we will examine highlights of SAS program code that creates a SAS format catalog; and a data research example.
Finally we will review a custom reporting example.
READ FULL SLIDE.
Please obtain a copy of the handout or share with your neighbor. The handout is available at SAScommunity.org at SASMickeyMouse.
Springborn (2013) provides the SAS program code for the custom reporting example.
NOTES ONLY
Centers for Disease Control and Prevention (CDC) explains “The U.S. has been using ICD-9-CM since 1979, and it is not sufficiently robust to serve the health care needs of the future. The content is no longer clinically accurate and has limited data about patients’ medical conditions and hospital inpatient procedures, the number of available codes is limited, and the coding structure is too restrictive.” 1
American Health Information Management Association (AHIMA) reports “ICD-10 benefits patients through improving: Patient outcomes and patient safety with better data for analysis and research; Ability to manage chronic disease by better capturing patient populations; More accurate reflections of patients’ clinical complexity and severity of illness; Ability to better identify high-risk patients; Ability to manage population health; Ability to assess effectiveness and safety of new medical technology; Increased patient engagement.” 2

3. A SAS Format Catalog For ICD-9/ICD-10
Create Four MS Excel Files (Display 1; Page 2)
In lines 8-27, SAS reads a data set containing diagnosis code values, text descriptions, and a date time variable.
Statements in lines use a datetime20. format to convert a date time value to a numeric value (e.g. “30SEP2015:00:00:00”DT; “01OCT2015:00:00:00”DT) and create two new SAS data sets containing ICD-9 and ICD-10 diagnosis codes and descriptions.
In lines SAS reads another data set and the same method is used to create two more SAS data sets containing ICD-9 and ICD-10 procedure codes and descriptions (lines 43-44).
SAS Enterprise Guide tasks Filter and Sort read these four data sets and create four MS Excel files.
DIAGNOSIS_ICD9_18APR2017.xls
PROCEDURE_ICD9_18APR2017.xls
DIAGNOSIS_ICD10_18APR2017.xlsx
PROCEDURE_ICD10_18APR2017.xlsx
We begin by creating a SAS format catalog consisting of four SAS formats (ICD-9 diagnoses and procedures, and ICD-10 diagnoses and procedures).
The first step is to create four MS Excel files (Handout; Display 1; Page 2).
AUDIENCE PARTICIPATION; SHOW ME THE HIGHLIGHTS; READ FULL SLIDE

4. A SAS Format Catalog For ICD-9/ICD-10
Create ICD-9 Diagnosis and Procedure SAS Formats (Display 2; Pages 3-4)
SAS program code (Lines 29-70) creates ICD-9 diagnosis format “$dx_icd9f.”
PROC IMPORT reads an MS Excel file containing 15,964 ICD-9 diagnosis codes (Lines 36-39).
Three items are created: start a list of diagnosis codes; label a concatenation of diagnosis code and diagnosis description; and SAS format named “$dx_icd9f” (Lines 41-50).
PROC FORMAT stores format “$dx_icd9f” in library forcat6 (Lines 68-69).
Similarly in lines the above method is repeated to create ICD-9 procedure code format “$px_icd9f” and store it in library forcat6.
The second step is to create SAS formats for ICD-9 diagnoses and ICD-9 procedures (Handout; Display 2; Pages 3-4).
SHOW ME THE HIGHLIGHTS; READ FULL SLIDE.

5. A SAS Format Catalog For ICD-9/ICD-10
Create ICD-10 Diagnosis and Procedure SAS Formats (Display 3; Pages 5-7)
SAS Program code (Lines 28-88) creates ICD-10 diagnosis format “$dx_icd10f.”
SAS Enterprise Guide tasks File, Import Data read DIAGNOSIS_ICD10_18APR2017.xlsx to create WORK.DIAGNOSIS_ICD10_18APR2017 containing 64,945 ICD-10 diagnosis codes and descriptions (lines 46-51).
Three items are created: start a list of diagnosis codes; label a concatenation of diagnosis code and diagnosis description; and SAS format named “$dx_icd10f” (lines 53-63).
PROC FORMAT stores format “$dx_icd10f” in library forcat6 (lines 86-88).
Similarly in lines the above method is repeated. SAS reads PROCEDURE_ICD10_18APR2017.xlsx and creates SAS format $px_icd10f.” and stores it in SAS format catalog forcat6.
The third step is to create SAS formats for ICD-10 diagnoses and ICD-10 procedures (Handout; Display 3; Pages 5-7).
This section very similar to ICD-9 previous slide.
SHOW ME THE HIGHLIGHTS; READ FULL SLIDE.

6. A Data Research Example
A researcher wants to investigate 25 diagnoses, 21 procedures and their descriptions for patients who had coronary artery bypass graft (CABG) surgery in March, June, September, and December of 2015.
The researcher wants the ability to:
Visually examine the results in MS Excel.
Use MS Excel tools for graphical, tabular and statistical analysis.
This data research example has three steps.
Now the SAS format catalog can be used in our data research example.
READ FULL SLIDE.

7. A Data Research Example –Program Details
Compile Macros that Define CABG Surgery (Appendix 1; Pages 8-10)
In lines two macro lists containing ICD-10 procedure codes were created to identify Isolated and Non-isolated CABG surgery.
&cabgicd10 containing 296 procedure codes for CABG surgery (Line 33).
&cabgexclicd10 containing 4,645 procedure codes for CABG surgery exclusions (Line 58).
Isolated CABG surgery means that no other major procedure was performed at the same time as the bypass surgery. Isolated CABG with another major procedure (exclusion) is called Non-isolated CABG. A Non-isolated CABG surgery has an higher risk of mortality.
In lines several macros are compiled. These macros will use ICD-9 procedure codes to identify Isolated and Non-isolated CABG surgery.
The first step is to create several macros (Handout; Appendix 1; Pages 8-10).
SHOW ME THE HIGHLIGHTS; READ FULL SLIDE.

8. A Data Research Example –Program Details
Construct the Cohort Study Population (Appendix 2; Pages 11-17)
The SAS format catalog is read in lines 9-12.
Lines identify Isolated and Non-isolated CABG surgery cases using either ICD-9 codes or ICD-10 codes based on patient discharge date (lines ).
Advantages of using a macro list.
For the ICD-9 group each macro reads one procedure code. One CABG procedure code is read at line 232, and 51 CABG exclusion procedure codes are read in lines Isolated and Non-isolated CABG are identified in line 321.
Two macro lists are used for the ICD-10 group. There are 296 CABG procedure codes read at line 336 using macro list &cabgicd10, and 4,645 CABG procedure exclusion codes are read at line 347 using macro list &cabgexclicd10. Isolated and Non-isolated CABG are identified in line 351.
Patient data for the cohort is selected by choosing CABG cases (icd361=1) in lines 375 to 384.
The second step is to create the cohort study population (Handout; Appendix 2; Pages 11-17).
SHOW ME THE HIGHLIGHTS; READ FULL SLIDE.

9. A Data Research Example –Program Details
Construct Diagnosis and Procedures Code Descriptions (Appendix 2; Pages 17-21)
OSHPD’s administrative data contains ICD-9 and ICD-10 diagnosis and procedure codes. without the decimal point. This program code reconstructs diagnosis and procedure codes with the decimal point and adds the code description using the SAS format catalog.
For example, ICD-9 diagnosis codes have a decimal in position four. In lines twenty-five diagnosis codes were reconstructed with the decimal point, and the ICD-9 diagnosis format $dx_icd9f was applied to create new text variables containing diagnoses code and description.
In lines the same approach is used to create text variables containing code value and description for ICD-9 procedure codes, ICD-10 diagnosis codes, and ICD-10 procedure codes.
Procedure dates were also created for both ICD-9 and ICD-10.
The third step is to reconstruct diagnosis and procedure codes and add the code descriptions (Handout; Appendix 2; Pages 17-21).
SHOW ME THE HIGHLIGHTS; READ FULL SLIDE.
NOTES ONLY.
ICD-9 diagnosis codes have a decimal in position four.
ICD-9 procedure codes which have a decimal in position three are created in lines
ICD-9 procedure dates are created in lines
ICD-10 diagnosis codes which have a decimal in position four are constructed in lines
ICD-10 procedure codes which do not have a decimal and are constructed in lines
ICD-10 procedure dates are created in lines
Hospital names are added to the data set in lines

10. A Data Research Example –Program Details
Send Results to MS Excel (Appendix 2; Pages 21-24)
In lines PROC REPORT send results to MS Excel using Tagset ExcelXP with several options to improve the appearance of the report (Notes).
In lines PROC REPORT defines each column using statements of “define,” “display,” and “style”. Variables include month of patient discharge, hospital name, type of CABG surgery, birth date, admission date, CABG surgery date, discharge date, 25 diagnoses with descriptions; 21 procedures with descriptions; and procedure dates (Notes).
Finally we send results to MS Excel (Handout; Appendix 2; Pages 21-24)
SHOW ME THE HIGHLIGHTS; READ FULL SLIDE.
NOTES ONLY.
Lines provides location, name of MS Excel file, and several Tagset.ExcelXP options to improve the appearance of the final report. Namely “absolute_column_width=” to specific column width; “autofit_height=‘Yes’” enables the height of the rows to adjust to the data; “autofilter=’Yes’” provides a filter and sort utility at the top of each column; “frozen_headers=’1’” fixes the header row as stationary while the scroll bars apply only to data rows; “sheet_name=” provides spreadsheet name; and “embedded_titles=’Yes’” includes titles in the worksheet body.
Lines PROC REPORT defines each column using statements of “define,” “display,” and “style”. For example, in lines we define a column of the report using variable hospitalfull which is the license number and name of the hospital. The variable is displayed in the report with column heading “Hospital Name.” The style attributes of the column include bold font, left justified, and font_size of 1. In addition the background color for the column is “very light moderate blue” with a foreground (font color) of “purple”.

11. A Data Research Example –Final Result
Final Result (Appendix 3; Page 25)
For the time period used, 5,412 patients had CABG surgery. Appendix 3 displays part of the result in MS Excel. Notice 2 of 25 diagnosis codes and descriptions as well as 2 of 21 procedure codes and descriptions with procedure date for both ICD-9 ( ) and ICD-10 ( ).
This result enables the researcher to:
Visually study 25 diagnoses and 21 procedures with descriptions for each patient without using look-up tables.
Use MS Excel tools for graphical, tabular and statistical analysis.
Import the results into SAS Enterprise Guide for further study.
Use macros or a macro list to: modify procedures in the current study cohort of CABG surgery; or use diagnosis codes to construct a new study cohort such as diagnosis of liver disease. An unlimited number of study cohorts are possible.
Great Job! We made it!
The final result is given in MS Excel (Handout; Appendix 3; Page 25).
SHOW ME THE HIGHLIGHTS; READ FULL SLIDE.

12. A Custom Reporting Example
A custom data discrepancy report was created using the SAS format catalog.
Each year OSHPD releases the California Report on CABG Surgery which presents findings from analyses of data collected from California-licensed hospitals that perform adult CABG surgery. This report features risk-adjusted operative mortality used to evaluate hospital and surgeon performance.
To make fair comparisons of care delivered by different healthcare providers, the risk model adjusts for differences in severity of illness (case mix) of patients across providers.
OSHPD reviews data submitted by each hospital for completeness and errors. The custom data discrepancy report (Springborn 2013) compares the CCORP clinical data to OSHPD’s hospital administrative data source, the Patient Discharge Data (PDD). Hospitals are asked to review discrepancies between the two data sources via patient medical chart review to verify that coding for eleven data variables are consistent.
Let’s look at the custom reporting example.
READ FULL SLIDE.
NOTES ONLY.
To make fair comparisons of care delivered by different healthcare providers, it is necessary to adjust for differences in severity of illness (case mix) of patients across providers. CCORP “levels the playing field” by considering the pre-operative condition of each patient. Providers that handle more complex cases receive a larger risk-adjustment weight in the risk model, and providers that handle less complex cases receive a smaller weight. Thus, hospitals and surgeons treating sicker patients are not at a disadvantage when their performance is compared with other surgeons and hospitals.”
OSHPD reviews data submitted by each hospital for completeness and errors. The data discrepancy report compares the CCORP clinical data to OSHPD’s hospital administrative data source, the Patient Discharge Data (PDD). Hospitals are asked to review discrepancies between the two data sources via patient medical chart review to verify that coding for the following eleven data variables are consistent. Data variables include: 1) CABG Surgeries Reported; 2) Isolated CABG Reporting; 3) CABG+Valve Reporting; 4) Resuscitation Prior to CABG Procedure; 5) CABG surgery deaths; 6) Post-Operative Stroke; 7) Dialysis Requirement; 8) Prolonged Ventilation; 9) Reoperation for Bleed; 10) Deep Sternal Wound Infection; and 11) Graft Occlusion.”

13. A Custom Reporting Example –Review Report
The Summary Table (Appendix 5; Page 27) identified 3 “Missing CABG Cases” and 21 “NON-CABG Cases”. Brief explanation of “Missing CABG Cases” and “NON-CABG Cases”.
Appendix 6; Page 28 gives a “Missing CABG Case” (under-reporting) example using ICD Administrative data reported a CABG surgery as ICD-10 procedure code W on March 29, 2016 but the clinical data did not report CABG surgery. The hospital action item is “SHOULD THIS CASE BE ADDED TO CCORP?”
Appendix 7; Page 29 gives a “NON-CABG case” (over-reporting) example using ICD-9. Administrative data did not report a CABG surgery because ICD-9 procedure code 36.1 is missing but the clinical data did report CABG surgery. The hospital action item is “SHOULD THIS CASE BE REMOVED FROM CCORP?”
Hospitals are asked to review discrepancies via patient medical chart review to verify that coding for eleven data variables are consistent.
The data discrepancy report (DDR) begins with a Help Sheet (Handout; Appendix 4; Page 26) which defines eleven data variables. Each data variable corresponds to a section in the DDR. Each section contains: a summary table of data variable discrepancies; and a list of patient hospital records corresponding to the data variable discrepancies.
Let’s examine the data variable section CABG-Reporting.
SHOW ME THE HIGHLIGHTS; READ FULL SLIDE.
Note the use of both diagnoses and procedures complete with text description for both ICD-9 and ICD-10 (Appendices 6-7)!
Great Job!!
This report continues to identify discrepancies for each of the remaining 10 data variables. Every six months this report is generated for each of 125 CCORP hospitals.

14. A Custom Reporting Example –Final Result
This custom reporting example showed how a SAS format catalog was used to develop the report on Coronary Artery Bypass Graft (CABG) surgery. This custom report’s main purpose is to help hospitals correctly report and classify all CABGs, thus ensuring the results are based on the highest quality data possible.
This custom data discrepancy report was created with ODS Report Writing Interface (RWI) and ODS LAYOUT offered in the Output Delivery System (ODS) in SAS® ODS RWI and ODS LAYOUT were designed exclusively to address the custom reporting requirements that go far beyond traditional tabular output produced by SAS procedures (current paper; Springborn (2013)).
Using ODS RWI and ODS LAYOUT with the SAS format catalog enables all users of administrative data to meet and exceed their own custom data reporting needs.
READ FULL SLIDE.

15. Footnotes Cited
1 International Classification of Diseases, (ICD-10-CM/PCS) Transition – Background, Why Change? (2015) Centers for Disease Control and Prevention. National Center for Health Statistics. Retrieved October 26, 2016, from 2 Choose Your Path to ICD-10, ICD-10 Leveraging an Asset Infographic. (2016) American Health Information Management Association. Retrieved October 26, 2016, from

16. Literature Cited
Springborn, R “Set Yourself Free –Use ODS Report Writing Technology in SAS Enterprise Guide Instead of Dynamic Data Exchange in PC SAS Part II SAS Code Revealed” Proceedings of the SAS® Global Forum 2013 Conference. April 2013., Paper San Francisco, CA.

17. Thanks for attending my talk.
Contact Information
Thanks for attending my talk.
Name: Robert Richard Springborn, Ph.D.
Enterprise: OSHPD, Healthcare Analytics Branch
Address: 2020 West El Camino Avenue
City, State ZIP: Sacramento, CA 95833
Work Phone:
Fax:
Web:
Twitter: