Screening Administrative Data

1
Screening Administrative Data To Assess the
Accuracy Of Present-on-Admission Coding Michael
Pine, M.D., M.B.A. Michael Pine and Associates,
Inc. Chicago, Illinois 773-643-1700 mpine_at_aol.com
2
Overview

• Rationale for Development of POA Screens
• Developmental Database and Selection of Cases
• Description and Aggregate Performance of 12
  Screens
• Evaluation of Coding By Individual Hospitals
• Computation of Composite Scores for Hospitals

3
Rationale for Development of POA Screens

• POA Code Identifies Hospital-Acquired
  Complications
• Important in Computing Rates of Adverse Outcomes
• Important in Risk-Adjusting Performance Measures
• Accurate Coding Requires Expertise and Teamwork
• Inaccurate Coding
• Affects Assessments of Clinical Quality
• Affects Reimbursement
• Chart Reviews to Detect Coding Errors Are
  Expensive
• Well-Designed Screens Can Detect Problems
  Efficiently

4
Developmental Database

• New York State SPARCS Data from 2003 through 2005
• 8,388,179 Discharges from 246 Hospitals
• Secondary Diagnosis Codes Have POA Modifiers
• 1 Present on Admission
• 2 Hospital-Acquired
• 9 Status on Admission Unknown

5
Selection of Cases for Screening

• High-Risk Conditions By Principal Diagnosis
• 33 Categories (e.g., septicemia, respiratory
  failure)
• Mortality 9.2 70 of Deaths 22 of
  Discharges
• Elective Admissions for Selected Surgical
  Procedures
• 7 Procedures (e.g., hysterectomy, knee
  replacement)
• Principal Diagnosis Consistent with Procedure
• Operation During First 2 Days of Hospitalization
• Inpatient Childbirth By Diagnosis or Procedure
  Codes

6
Diagnoses Almost Always Present on Admission

• 231 Diagnosis Groups (e.g., malignancy,
  osteoporosis)
• Analyzed for Each of the 3 Sets of Cases Screened
• Aggregate Data for Each Set

7
Complications in High-Risk Conditions

• Chronic Diagnoses with and without Acute
  Components
• 21 Pairs (e.g., hernia with and without
  obstruction)
• Rates At Which Coded As Hospital-Acquired
• Chronic without Acute 1.06 of 1,612,079
  Diagnoses
• Chronic with Acute 3.34 of 222,641 Diagnoses
• Diagnoses Frequently Hospital-Acquired (e.g.,
  anuria)
• 3 Categories Based on Frequency Hospital-Acquired
• 27 Diagnosis Groups in Category A 59 in B 54 in
  C
• Category A - 63.5 of 172,472 Codes
  Hospital-Acquired
• Category B - 34.7 of 469,970 Codes
  Hospital-Acquired
• Category C - 24.8 of 772,049 Codes
  Hospital-Acquired

8
Mortality with Hospital-Acquired Complications

• Only for High-Risk Conditions
• Mortality Greater When Diagnosis
  Hospital-Acquired
• 3 Categories Based on Ratio of Mortality Rates
• 66 Diagnosis Groups in Category A 54 in B 64 in
  C
• Aggregate Data for Each Category

9
Complications in Elective Surgical Admissions

• Diagnoses Frequently Hospital-Acquired
  Complications
• 64 Diagnosis Groups (e.g., septicemia, shock)
• Of 138,655 Codes, 68.3 Hospital-Acquired
• Chronic Diagnoses with and without Acute
  Components
• 21 Pairs (e.g., asthma with and without
  exacerbation)
• Rates At Which Coded As Hospital-Acquired
• Chronic without Acute 0.39 of 187,453 Diagnoses
• Chronic with Acute 18.72 of 2,174 Diagnoses

10
Risk-Adjusted Post-Op Lengths of Stay

• High Rates of Prolonged LOS in Uncomplicated
  Cases
• Develop Predictive Equations for Routine Post-Op
  LOS
• Compute Observed Minus Predicted Post-Op LOS
• For All Live Discharges at Each Hospital
• Create XmR Control Charts of OBS minus PRED LOS
• Remove Outliers with Prolonged Post-Op LOS
• Repeat Process Until No Further Outliers
  Identified
• Set Upper Bound at Median Outlier Rate for All
  Hospitals
• Repeat Process Using Only Uncomplicated Cases
• Compute Outlier Rates for Each Hospital
• Identify Hospitals with Rates Greater Than Upper
  Bound

11
Risk-Adjusted Post-Op Lengths of StayLive
Discharges with and without Reported Complications
12
Risk-Adjusted Post-Op Lengths of StayLive
Discharges without Reported Complications
13
Complications in Obstetrical Admissions

• Diagnoses Usually Present on Admission
• 7 Diagnosis Groups (e.g., multiple gestation)
• Of 448,242 Codes, 5.19 Hospital-Acquired
• Fifth Digit Codes Incompatible with Inpatient
  Delivery
• 737,125 Inpatient Deliveries
• Fifth Digit 0 or 3 or 4 in 0.27
• Inpatient Post-Partum Complications
• 74,669 Cases with Obstetrical Fifth Digit 2
• No Diagnosis Coded As Hospital-Acquired in 36.5

14
Initial Analyses of Hospital Coding

• 226 Hospitals Screened with One or More Measures
• 22 Hospitals Have More Than 10 Unknowns
• Diagnoses Almost Always Present on Admission
• Less Than 2 of Diagnoses Hospital-Acquired

15
Hospital Coding for High-Risk Conditions

• Chronic Diagnoses with Acute Components
• Hospital-Acquired Rate Greater Than 2 AND
  Greater Than Twice Rate for Chronic Codes
• Of 145 Hospitals, 71.7 Met Criteria
• Diagnoses Frequently Hospital-Acquired
• Hospital-Acquired Rate Greater Than 15 for
  Category B Diagnoses AND Rate Monotonically
  Decreasing from Category A to Category C
• Of 181 Hospitals, 83.4 Met Criteria

16
Hospital Mortality Rates for High-Risk Conditions

• Compute Predicted Mortality Rates
• Indirect Standardization within Each Category
• Based on Rates for Diagnoses Present on Admission
• Odds Ratio of Observed to Predicted Mortality
  Rates
• Greater Than 1.60 for All Diagnoses OR
• Greater Than 1.30 for All Diagnoses AND Greater
  Than 1.60 for Diagnoses in Categories A and B
• Of 184 Hospitals, 82.6 Met Criteria

17
Hospital Coding for Elective Surgical Admissions

• Diagnoses Frequently Hospital-Acquired
  Complications
• Hospital-Acquired Rate Greater Than 65
• Of 175 Hospitals, 61.1 Met Criterion
• Chronic Diagnoses with Acute Components
• Compute 2 Standard Deviation Lower Bounds for
  Hospital-Acquired Rates
• Hospital-Acquired Rate Greater Than 12 AND
  Greater Than Three Times Rate for Chronic Codes
  OR
• Lower Bound Greater Than Twice Rate for Chronic
  Codes
• Of 93 Hospitals, 96.8 Met Criteria

18
Prolonged Risk-Adjusted Post-Op Length of Stay

• Median Outlier Rate for All Live Discharges
  5.36
• Outlier Rates for Uncomplicated Cases Less Than
  Upper Bound
• In 81.5 of 178 Hospitals
• In 98.4 of 64 Reference Hospitals
• In 71.9 of 114 Remaining Hospitals

19
Hospital Coding for Obstetrical Admissions

• Diagnoses Usually Present on Admission
• Hospital-Acquired Rate Less Than 3
• Of 134 Hospitals, 63.4 Met Criterion
• Fifth Digit Codes Incompatible with Inpatient
  Delivery
• Less Than 0.5 of Obstetrical Codes Incompatible
• Of 134 Hospitals, 87.3 Met Criterion
• Cases with Inpatient Post-Partum Complications
• Less Than 20 without Hospital-Acquired Diagnosis
• Of 123 Hospitals, 41.5 Met Criterion

20
Composite Hospital Scoring

• Range of Points Assigned to Each Measure
• Range from 1 to N with N 4, 5, 8, or 10
• Score Only for 204 Hospitals with Adequate Data
• Score Measure Only When Volume Criteria Met
• For Each Hospital, Compute
• Total of Points Scored for Each Measure
• Maximum and Minimum Possible Points
• For Each Measure, Compute Average of Points
  Scored
• Obtain Final Adjusted Hospital Scores By
  Interpolation

21
Final Adjusted Hospital Scores
22
Screening and Improvement of POA Coding