SlideShare a Scribd company logo
1 of 19
# LS-120:
A Novel Approach to Calculating
Medicare Hospital 30-Day
Readmissions for the SAS® Novice
Karen E. Wallace
Centene® Corporation
Agenda
• Overview – Medicare readmissions
• SAS® Readmissions Roadmap & Methodology
• Comparing:
– SAS® DATA Step
– PROC SQL
• Lessons Learned
& Conclusion
• Questions
Why Care about Medicare
Readmissions?
• Affects quality of life: patient and caregiver
• Exposes gaps in Inpatient prospective payment
system (IPPS)
• Key indicator for measuring success - or failure
• $17 Billion spent in Medicare expenditures annually
Retrieved from: http://www.hhs.gov/blog/2016/02/24/reducing-avoidable-hospital-readmissions.html#
Brief History
• Affordable Care Act added section 1886(q) to the
Social Security Act
• Established Hospital Readmissions Reduction
Program (HRRP)
• Reduce payments to IPPS facilities with discharges
starting October 1, 2012
• Began with 3 conditions:
– Acute myocardial infarction (AMI)
– Congenital heart failure (CHF)
– Pneumonia (PNE)
Retrieved from: http://www.hhs.gov/blog/2016/02/24/reducing-avoidable-hospital-readmissions.html#
Brief History (continued)
• FY 2015 added 2 additional conditions:
– Chronic obstructive pulmonary disease (COPD)
– Total hip & knee arthroplasty complications
(THA/TKA)
• Maximum penalty increased to 3%
• Penalty calculation involves:
– Claims grouped upon DRGs
– Risk adjustments based on predicted vs. expected
readmissions per disease state
Retrieved from: http://www.besler.com/2015-readmissions-penalties/
Brief History (continued)
Retrieved from: https://www.cms.gov/medicare/medicare-fee-for-service-payment/acuteinpatientpps/readmissions-reduction-
program.html
More than 75%
had penalties
1.00 (no penalty) Readmissions adjustment factor .970 (3% max penalty)
But it’s not so simple…
• World Health Organization’s International
Classification of Diseases (ICD) transitioned from
version 9 to 10 starting October 1, 2015
– 19x more procedure codes in ICD-10-PCS vs. ICD-9-CM
– 5x more diagnosis codes in ICD-10-CM vs. ICD-9-CM
– GEMS: No definitive crosswalk between ICD-9 and ICD-10
Retrieved from: http://www.cms.org/uploads/ICDLogicGEMSWhitePaper.pdf
5 digits => 7 digits
Data Supports Improved
Outcomes
Decrease ~ 3.5 %
565,000 reduction in
extra hospital stays
Methodology
• 2 Data sets: June 2015 and January 2016
• Minimum necessary data elements:
– Medical record number (MRN)
– Index date
– Primary disease state
– Payer upon admission
– Discharge date
– Discharge disposition
DATA Step Processing
1. De-identify MRN
PROC SORT DATA=MIDAS_TEST_TO_DEIDENTIFY_V3 OUT=PT_SORTED;
BY MRN Index_date DC_Date;
RUN;
DATA PT_CLEAN;
SET PT_SORTED;
BY MRN Index_date DC_Date;
RETAIN PTID 0;
IF FIRST.MRN THEN PTID = PTID + 1;
DROP MRN;
RUN;
Using RETAIN Statement and
FIRST variable, PTID auto-
increments based upon each
new “MRN” encountered
DATA Step Processing (continued)
2. Sequence readmits & categorize disease states
DATA DATA_TEST;
SET PT_CLEAN;
BY PTID Index_date DC_Date;
RETAIN N;
IF FIRST.PTID THEN N = 0;
N = N + 1;
IF NOT MISSING(First_Dx) THEN IF SUBSTR(First_Dx,1,3) = '410' OR (SUBSTR(First_Dx,1,3) = 'I21') OR
(SUBSTR(First_Dx,1,3) = 'I22') OR (SUBSTR(First_Dx,1,3) = 'I23') THEN Dx1 = 'AMI';
ELSE IF (SUBSTR(First_Dx,1,3) = '402') OR (SUBSTR(First_Dx,1,3) = '404')
OR (SUBSTR(First_Dx,1,3) = '428') OR (SUBSTR(First_Dx,1,3) = 'I50') THEN Dx1 = 'CHF';
ELSE IF (SUBSTR(First_Dx,1,3) = '480') OR (SUBSTR(First_Dx,1,3) = '481')
OR (SUBSTR(First_Dx,1,3) = '482') OR (SUBSTR(First_Dx,1,3) = '483') OR
(SUBSTR(First_Dx,1,3) = '485') OR (SUBSTR(First_Dx,1,3) = '486') OR
(SUBSTR(First_Dx,1,3) = '487') OR (SUBSTR(First_Dx,1,3) = '488') OR (SUBSTR(First_Dx,1,3) = 'J18') THEN Dx1 =
'PNE'; FORMAT Index_Date DC_Date YYMMDD10.;
LABEL <steps omitted for brevity>
DROP DUMMY SeqNo;
RUN;
Using RETAIN Statement and FIRST
variable to sequence readmission
visits, then use SUBSTR function to
“bucket” disease states
DATA Step Processing (continued)
3. Determine gaps between sequential readmits
DATA DATA_TEST2;
SET DATA_TEST;
BY PTID Index_date DC_Date;
REF_DATE = LAG(DC_Date);
GAP =(Index_Date - REF_DATE);
IF FIRST.PTID THEN DO;
Ref_Date =.;
Gap =.;
Flg =.;
Readmissions =.;
END;
IF 0 <= Gap <=30 then Flg=1;
IF (Gap = '.' OR Gap le 0) THEN DELETE;
Readmissions + Flg;
IF ((Payer ne 'MEDICARE') OR (Payer ='')) THEN DELETE;
<steps omitted for brevity>
LAG function IDs previous
discharge date where applicable
and a “Gap” variable is assigned
the length between visits
DATA Step Processing (continued)
4. Calculate LOS
LOS = (DC_Date - Index_date) + 1
ICD9_DESC = UPCASE(Dx_Desc);
FORMAT INDEX_DATE REF_DATE YYMMDD10.;
LABEL Dx1 = 'ICD9/10 Grouping'
Gap = 'Gap to Readmission (d)'
LOS = 'Length of Stay (d)'
Ref_Date = 'Index Reference Date';
DROP Flg Dx_Desc;
LOS quantifies hospital length
Alternative Method: PROC
SQL
1. Combining several DATA steps into one block of code*
OPTIONS MSGLEVEL=I;
PROC SQL;
CREATE TABLE PATIENTS_SQL AS
SELECT a.PTID LENGTH = 4 LABEL = "Patient ID"
,a.N LENGTH = 3 LABEL = "SeqNo"
,a.DC_Date LENGTH = 8 LABEL = "Discharge Date"
,b.Index_Date LENGTH = 8 LABEL = "Index Date"
,(b.Index_Date - a.DC_Date) as Gap LENGTH = 3 LABEL = "Gap to
Readmission (d)"
,(b.DC_Date - b.Index_Date) + 1 as LOS LENGTH = 3 LABEL = "Length of Stay (d)"
,a.Payer
,1 as Flg
,b.First_Dx LENGTH = 6 LABEL = "ICD9/10 Primary Diagnosis"
,b.Dx1 LENGTH = 6 LABEL = "ICD9 Grouping"
,a.First_DC_Disp LENGTH = 155 LABEL = "Discharge Disposition“
FROM DATA_TEST as a
INNER JOIN DATA_TEST as b Inner join creates a table almost
identical to that of the DATA Step
Alternative Method : PROC
SQL (continued)
Use of an inner join to achieve similar output*
ON a.PTID = b.PTID
WHERE (b.N = a.N + 1) AND
(b.Index_Date - a.DC_Date) BETWEEN 1 AND 30
AND a.Payer = 'MEDICARE'
AND (a.First_DC_Disp NOT IN('I/P ACUTE HOSPITAL TRANSFER'
,'O/P ACUTE HOSPITAL TRANSFER'
,'DISCHARGED AGAINST MED ADVICE'
,'I/P EXPIRED'))
AND b.DX1 ne ''
ORDER BY a.PTID, a.N;
QUIT;
* LEAD function would be handy…
Lessons Learned
• Many ways in SAS® may achieve desired result
• Future steps:
– Combine separate DATA steps to make code
more “elegant” based upon accumulating SAS®
knowledge
– Automate, automate, automate!
Conclusion
• Medicare readmissions history
• Although calculating readmissions is not a new
SAS® concept, incorporates ICD-10
• Reproducible “use case” – other variables to add
• Build predictive model for those most “at risk”
• Proactive clinical interventions
Thank you.
Questions?

More Related Content

Similar to SESUG2016_PowerPoint_template v1

Working With Large-Scale Clinical Datasets
Working With Large-Scale Clinical DatasetsWorking With Large-Scale Clinical Datasets
Working With Large-Scale Clinical DatasetsCraig Smail
 
Healthcare deserts: How accessible is US healthcare?
Healthcare deserts: How accessible is US healthcare?Healthcare deserts: How accessible is US healthcare?
Healthcare deserts: How accessible is US healthcare?Data Con LA
 
Teknis _ Use Case Vital Sign, Procedure, Diet (2022-08-09).pdf
Teknis _ Use Case Vital Sign, Procedure, Diet (2022-08-09).pdfTeknis _ Use Case Vital Sign, Procedure, Diet (2022-08-09).pdf
Teknis _ Use Case Vital Sign, Procedure, Diet (2022-08-09).pdfBayuSugiarno1
 
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...Mohamad Syafiq
 
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...Mohamad Syafiq
 
PageRank for Anomaly Detection
PageRank for Anomaly DetectionPageRank for Anomaly Detection
PageRank for Anomaly DetectionDataWorks Summit
 
Page rank for anomaly detection - Big Things meetup in Israel
Page rank for anomaly detection - Big Things meetup in IsraelPage rank for anomaly detection - Big Things meetup in Israel
Page rank for anomaly detection - Big Things meetup in IsraelOfer Mendelevitch
 
Predictive model for falls poster v3
Predictive model for falls poster v3Predictive model for falls poster v3
Predictive model for falls poster v3Marmi Le
 
Health Science Data and Metadata: Trends and Needs
Health Science Data and Metadata: Trends and NeedsHealth Science Data and Metadata: Trends and Needs
Health Science Data and Metadata: Trends and NeedsLynne Frederickson
 
Financing Healthcare (Part 2) Lecture A
Financing Healthcare (Part 2) Lecture AFinancing Healthcare (Part 2) Lecture A
Financing Healthcare (Part 2) Lecture ACMDLearning
 
Predicting Hospital Readmissions
Predicting Hospital ReadmissionsPredicting Hospital Readmissions
Predicting Hospital ReadmissionsDerek Christensen
 
Medical Cost Reimbursement of Medical Devices
Medical Cost Reimbursement of Medical DevicesMedical Cost Reimbursement of Medical Devices
Medical Cost Reimbursement of Medical DevicesDickson Consulting
 
Study Planner Site Selector Master (Abridged) 070809
Study Planner  Site Selector Master (Abridged)   070809Study Planner  Site Selector Master (Abridged)   070809
Study Planner Site Selector Master (Abridged) 070809kesre
 
PREDICTION OF DIABETES (SUGAR) USING MACHINE LEARNING TECHNIQUES
PREDICTION OF DIABETES (SUGAR) USING MACHINE LEARNING TECHNIQUESPREDICTION OF DIABETES (SUGAR) USING MACHINE LEARNING TECHNIQUES
PREDICTION OF DIABETES (SUGAR) USING MACHINE LEARNING TECHNIQUESIRJET Journal
 
1Big Data Analytics forHealthcareChandan K. ReddyD.docx
1Big Data Analytics forHealthcareChandan K. ReddyD.docx1Big Data Analytics forHealthcareChandan K. ReddyD.docx
1Big Data Analytics forHealthcareChandan K. ReddyD.docxaulasnilda
 
Standardization of “Safety Drug” Reporting Applications
Standardization of “Safety Drug” Reporting ApplicationsStandardization of “Safety Drug” Reporting Applications
Standardization of “Safety Drug” Reporting Applicationshalleyzand
 

Similar to SESUG2016_PowerPoint_template v1 (20)

Working With Large-Scale Clinical Datasets
Working With Large-Scale Clinical DatasetsWorking With Large-Scale Clinical Datasets
Working With Large-Scale Clinical Datasets
 
Healthcare deserts: How accessible is US healthcare?
Healthcare deserts: How accessible is US healthcare?Healthcare deserts: How accessible is US healthcare?
Healthcare deserts: How accessible is US healthcare?
 
Teknis _ Use Case Vital Sign, Procedure, Diet (2022-08-09).pdf
Teknis _ Use Case Vital Sign, Procedure, Diet (2022-08-09).pdfTeknis _ Use Case Vital Sign, Procedure, Diet (2022-08-09).pdf
Teknis _ Use Case Vital Sign, Procedure, Diet (2022-08-09).pdf
 
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
 
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
Appointment and Booking System for Clinics - Mohamad Syafiq Bin Mohd Razadi (...
 
PageRank for Anomaly Detection
PageRank for Anomaly DetectionPageRank for Anomaly Detection
PageRank for Anomaly Detection
 
Data Leveraging
Data Leveraging Data Leveraging
Data Leveraging
 
Page rank for anomaly detection - Big Things meetup in Israel
Page rank for anomaly detection - Big Things meetup in IsraelPage rank for anomaly detection - Big Things meetup in Israel
Page rank for anomaly detection - Big Things meetup in Israel
 
Predictive model for falls poster v3
Predictive model for falls poster v3Predictive model for falls poster v3
Predictive model for falls poster v3
 
Health Science Data and Metadata: Trends and Needs
Health Science Data and Metadata: Trends and NeedsHealth Science Data and Metadata: Trends and Needs
Health Science Data and Metadata: Trends and Needs
 
Financing Healthcare (Part 2) Lecture A
Financing Healthcare (Part 2) Lecture AFinancing Healthcare (Part 2) Lecture A
Financing Healthcare (Part 2) Lecture A
 
Predictive Medicine
Predictive Medicine Predictive Medicine
Predictive Medicine
 
Cardiology Coding Alert White Paper
Cardiology Coding Alert White PaperCardiology Coding Alert White Paper
Cardiology Coding Alert White Paper
 
Predicting Hospital Readmissions
Predicting Hospital ReadmissionsPredicting Hospital Readmissions
Predicting Hospital Readmissions
 
Medical Cost Reimbursement of Medical Devices
Medical Cost Reimbursement of Medical DevicesMedical Cost Reimbursement of Medical Devices
Medical Cost Reimbursement of Medical Devices
 
Study Planner Site Selector Master (Abridged) 070809
Study Planner  Site Selector Master (Abridged)   070809Study Planner  Site Selector Master (Abridged)   070809
Study Planner Site Selector Master (Abridged) 070809
 
PREDICTION OF DIABETES (SUGAR) USING MACHINE LEARNING TECHNIQUES
PREDICTION OF DIABETES (SUGAR) USING MACHINE LEARNING TECHNIQUESPREDICTION OF DIABETES (SUGAR) USING MACHINE LEARNING TECHNIQUES
PREDICTION OF DIABETES (SUGAR) USING MACHINE LEARNING TECHNIQUES
 
scientific paper on emr
scientific paper on emrscientific paper on emr
scientific paper on emr
 
1Big Data Analytics forHealthcareChandan K. ReddyD.docx
1Big Data Analytics forHealthcareChandan K. ReddyD.docx1Big Data Analytics forHealthcareChandan K. ReddyD.docx
1Big Data Analytics forHealthcareChandan K. ReddyD.docx
 
Standardization of “Safety Drug” Reporting Applications
Standardization of “Safety Drug” Reporting ApplicationsStandardization of “Safety Drug” Reporting Applications
Standardization of “Safety Drug” Reporting Applications
 

SESUG2016_PowerPoint_template v1

  • 1. # LS-120: A Novel Approach to Calculating Medicare Hospital 30-Day Readmissions for the SAS® Novice Karen E. Wallace Centene® Corporation
  • 2. Agenda • Overview – Medicare readmissions • SAS® Readmissions Roadmap & Methodology • Comparing: – SAS® DATA Step – PROC SQL • Lessons Learned & Conclusion • Questions
  • 3. Why Care about Medicare Readmissions? • Affects quality of life: patient and caregiver • Exposes gaps in Inpatient prospective payment system (IPPS) • Key indicator for measuring success - or failure • $17 Billion spent in Medicare expenditures annually Retrieved from: http://www.hhs.gov/blog/2016/02/24/reducing-avoidable-hospital-readmissions.html#
  • 4. Brief History • Affordable Care Act added section 1886(q) to the Social Security Act • Established Hospital Readmissions Reduction Program (HRRP) • Reduce payments to IPPS facilities with discharges starting October 1, 2012 • Began with 3 conditions: – Acute myocardial infarction (AMI) – Congenital heart failure (CHF) – Pneumonia (PNE) Retrieved from: http://www.hhs.gov/blog/2016/02/24/reducing-avoidable-hospital-readmissions.html#
  • 5. Brief History (continued) • FY 2015 added 2 additional conditions: – Chronic obstructive pulmonary disease (COPD) – Total hip & knee arthroplasty complications (THA/TKA) • Maximum penalty increased to 3% • Penalty calculation involves: – Claims grouped upon DRGs – Risk adjustments based on predicted vs. expected readmissions per disease state Retrieved from: http://www.besler.com/2015-readmissions-penalties/
  • 6. Brief History (continued) Retrieved from: https://www.cms.gov/medicare/medicare-fee-for-service-payment/acuteinpatientpps/readmissions-reduction- program.html More than 75% had penalties 1.00 (no penalty) Readmissions adjustment factor .970 (3% max penalty)
  • 7. But it’s not so simple… • World Health Organization’s International Classification of Diseases (ICD) transitioned from version 9 to 10 starting October 1, 2015 – 19x more procedure codes in ICD-10-PCS vs. ICD-9-CM – 5x more diagnosis codes in ICD-10-CM vs. ICD-9-CM – GEMS: No definitive crosswalk between ICD-9 and ICD-10 Retrieved from: http://www.cms.org/uploads/ICDLogicGEMSWhitePaper.pdf 5 digits => 7 digits
  • 8. Data Supports Improved Outcomes Decrease ~ 3.5 % 565,000 reduction in extra hospital stays
  • 9.
  • 10. Methodology • 2 Data sets: June 2015 and January 2016 • Minimum necessary data elements: – Medical record number (MRN) – Index date – Primary disease state – Payer upon admission – Discharge date – Discharge disposition
  • 11. DATA Step Processing 1. De-identify MRN PROC SORT DATA=MIDAS_TEST_TO_DEIDENTIFY_V3 OUT=PT_SORTED; BY MRN Index_date DC_Date; RUN; DATA PT_CLEAN; SET PT_SORTED; BY MRN Index_date DC_Date; RETAIN PTID 0; IF FIRST.MRN THEN PTID = PTID + 1; DROP MRN; RUN; Using RETAIN Statement and FIRST variable, PTID auto- increments based upon each new “MRN” encountered
  • 12. DATA Step Processing (continued) 2. Sequence readmits & categorize disease states DATA DATA_TEST; SET PT_CLEAN; BY PTID Index_date DC_Date; RETAIN N; IF FIRST.PTID THEN N = 0; N = N + 1; IF NOT MISSING(First_Dx) THEN IF SUBSTR(First_Dx,1,3) = '410' OR (SUBSTR(First_Dx,1,3) = 'I21') OR (SUBSTR(First_Dx,1,3) = 'I22') OR (SUBSTR(First_Dx,1,3) = 'I23') THEN Dx1 = 'AMI'; ELSE IF (SUBSTR(First_Dx,1,3) = '402') OR (SUBSTR(First_Dx,1,3) = '404') OR (SUBSTR(First_Dx,1,3) = '428') OR (SUBSTR(First_Dx,1,3) = 'I50') THEN Dx1 = 'CHF'; ELSE IF (SUBSTR(First_Dx,1,3) = '480') OR (SUBSTR(First_Dx,1,3) = '481') OR (SUBSTR(First_Dx,1,3) = '482') OR (SUBSTR(First_Dx,1,3) = '483') OR (SUBSTR(First_Dx,1,3) = '485') OR (SUBSTR(First_Dx,1,3) = '486') OR (SUBSTR(First_Dx,1,3) = '487') OR (SUBSTR(First_Dx,1,3) = '488') OR (SUBSTR(First_Dx,1,3) = 'J18') THEN Dx1 = 'PNE'; FORMAT Index_Date DC_Date YYMMDD10.; LABEL <steps omitted for brevity> DROP DUMMY SeqNo; RUN; Using RETAIN Statement and FIRST variable to sequence readmission visits, then use SUBSTR function to “bucket” disease states
  • 13. DATA Step Processing (continued) 3. Determine gaps between sequential readmits DATA DATA_TEST2; SET DATA_TEST; BY PTID Index_date DC_Date; REF_DATE = LAG(DC_Date); GAP =(Index_Date - REF_DATE); IF FIRST.PTID THEN DO; Ref_Date =.; Gap =.; Flg =.; Readmissions =.; END; IF 0 <= Gap <=30 then Flg=1; IF (Gap = '.' OR Gap le 0) THEN DELETE; Readmissions + Flg; IF ((Payer ne 'MEDICARE') OR (Payer ='')) THEN DELETE; <steps omitted for brevity> LAG function IDs previous discharge date where applicable and a “Gap” variable is assigned the length between visits
  • 14. DATA Step Processing (continued) 4. Calculate LOS LOS = (DC_Date - Index_date) + 1 ICD9_DESC = UPCASE(Dx_Desc); FORMAT INDEX_DATE REF_DATE YYMMDD10.; LABEL Dx1 = 'ICD9/10 Grouping' Gap = 'Gap to Readmission (d)' LOS = 'Length of Stay (d)' Ref_Date = 'Index Reference Date'; DROP Flg Dx_Desc; LOS quantifies hospital length
  • 15. Alternative Method: PROC SQL 1. Combining several DATA steps into one block of code* OPTIONS MSGLEVEL=I; PROC SQL; CREATE TABLE PATIENTS_SQL AS SELECT a.PTID LENGTH = 4 LABEL = "Patient ID" ,a.N LENGTH = 3 LABEL = "SeqNo" ,a.DC_Date LENGTH = 8 LABEL = "Discharge Date" ,b.Index_Date LENGTH = 8 LABEL = "Index Date" ,(b.Index_Date - a.DC_Date) as Gap LENGTH = 3 LABEL = "Gap to Readmission (d)" ,(b.DC_Date - b.Index_Date) + 1 as LOS LENGTH = 3 LABEL = "Length of Stay (d)" ,a.Payer ,1 as Flg ,b.First_Dx LENGTH = 6 LABEL = "ICD9/10 Primary Diagnosis" ,b.Dx1 LENGTH = 6 LABEL = "ICD9 Grouping" ,a.First_DC_Disp LENGTH = 155 LABEL = "Discharge Disposition“ FROM DATA_TEST as a INNER JOIN DATA_TEST as b Inner join creates a table almost identical to that of the DATA Step
  • 16. Alternative Method : PROC SQL (continued) Use of an inner join to achieve similar output* ON a.PTID = b.PTID WHERE (b.N = a.N + 1) AND (b.Index_Date - a.DC_Date) BETWEEN 1 AND 30 AND a.Payer = 'MEDICARE' AND (a.First_DC_Disp NOT IN('I/P ACUTE HOSPITAL TRANSFER' ,'O/P ACUTE HOSPITAL TRANSFER' ,'DISCHARGED AGAINST MED ADVICE' ,'I/P EXPIRED')) AND b.DX1 ne '' ORDER BY a.PTID, a.N; QUIT; * LEAD function would be handy…
  • 17. Lessons Learned • Many ways in SAS® may achieve desired result • Future steps: – Combine separate DATA steps to make code more “elegant” based upon accumulating SAS® knowledge – Automate, automate, automate!
  • 18. Conclusion • Medicare readmissions history • Although calculating readmissions is not a new SAS® concept, incorporates ICD-10 • Reproducible “use case” – other variables to add • Build predictive model for those most “at risk” • Proactive clinical interventions

Editor's Notes

  1. Why doing this? I want to walk an individual relatively unfamiliar with the nuances of the readmission process as well as advanced SAS functionality, through the methodology, coding and validation techniques to prepare a clean data set with descriptive statistics. Although many papers have been written about calculating hospital readmissions, this paper includes updated code for ICD-10 (International Classification of Diseases) as well as a novel and comprehensive approach using SAS DATA step and PROC SQL
  2. Unfortunate circumstances of a friend or relative being readmitted, patient satisfaction drops, poor quality care – improve discharge planning and ensure pts have fu appts
  3. Unfortunate circumstances of a friend or relative being readmitted, patient satisfaction drops, poor quality care – improve discharge planning and ensure pts have fu appts
  4. To calculate the penalty, CMS reviews the claims for five readmission measures for excess readmissions. The claims are grouped by measure based on diagnosis and procedure codes. Excess readmission ratio = risk-adjusted predicted readmissions/risk-adjusted expected readmissions
  5. CMS publishes a readmissions adjustment factor for each affected hospital to indicate the level of its penalty, which ranges from 0.9700 (reflecting the maximum 3% penalty for FY 2015) to 1.00 (indicating no penalty). The penalty is assessed against Medicare base operating DRG payments for all discharges at a penalized hospital. However, CMS does not publish an estimated penalty for individual hospitals.
  6. GEM – General equivalency mapping
  7. And the data show it’s working. Between April 2010 and May 2015, we estimate that approximately 565,000 readmissions were prevented across all conditions, compared to the readmission rate in the year prior to the passage of the Affordable Care Act (April 2009 to March 2010). That’s 565,000 times that a patient didn’t have to experience an extra hospital stay. A new study by Department of Health and Human Services researchers published today in the New England Journal of Medicine shows that readmissions fell sharply following enactment of the Affordable Care Act2. The study found that, as shown in Figure 1, readmission rates fell more sharply for conditions that were targeted by the Hospital Readmissions Reduction Program, including heart attack, heart failure, and pneumonia, than for other conditions requiring hospitalizations, such as surgeries and diabetes. Conditions that weren’t targeted by the Hospital Readmissions Reduction Program probably saw spillover benefits from actions hospitals took in response to new incentives. For both the targeted conditions and other hospitalizations, the drop in readmissions mostly occurred during the period between the enactment of the Affordable Care Act in March 2010 and the start of the Hospital Readmissions Reduction Program in October 2012, when hospitals would have taken action to avoid facing penalties. Although penalties began in October 2012, hospitals had incentives to improve their performance well before the payment adjustment occurred. 
  8. Given that framework, creating a roadmap to approach hosp readmissions
  9. Pre and post implementation –  Are discharged alive  Are Medicare recipients at time of admission  Discharge date is not the same as the index admission date  Patients are not discharged to another acute care hospital (ACH)  Patients were not discharged against medical advice
  10. The RETAIN statement allows values to be kept across observations enabling complex data manipulation. Mention of HIPAA for sharing data, de-id MRN
  11. The RETAIN statement allows values to be kept across observations enabling complex data manipulation.
  12. The RETAIN statement allows values to be kept across observations enabling complex data manipulation.
  13. The RETAIN statement allows values to be kept across observations enabling complex data manipulation.
  14. a limitation to using PROC SQL with either Enterprise Guide or University Edition is that the variable “Ref_Date” cannot be generated. In BASE SAS®, the LEAD function is akin the LAG function, though the latter is not recognized in PROC SQL.
  15. * Note on coding difference for Gap variable LEAD function ~ LAG function to calculate gaps
  16. The RETAIN statement allows values to be kept across observations enabling complex data manipulation.
  17. The RETAIN statement allows values to be kept across observations enabling complex data manipulation.