denials management solution using machine intelligence · case study 13 identified new patterns of...

Denials Management Solution using Machine Intelligence

1

Company Confidential & Proprietary 2

Company Confidential & Proprietary

The Revenue Opportunity

3

__ __ __

1 2 4 TYPEOF BILL

FROM THROUGH5 FED. TAX NO.

a

b

c

d

DX

ECI

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

A

B

C

A B C D E F G HI J K L M N O P Q

a b c a b c

a

b c d

ADMISSION CONDITION CODESDATE

OCCURRENCE OCCURRENCE OCCURRENCE OCCURRENCE SPAN OCCURRENCE SPANCODE DATE CODE CODE CODE DATE CODE THROUGH

VALUE CODES VALUE CODES VALUE CODESCODE AMOUNT CODE AMOUNT CODE AMOUNT

TOTALS

PRINCIPAL PROCEDURE a. OTHER PROCEDURE b. OTHER PROCEDURE NPICODE DATE CODE DATE CODE DATE

FIRST

c. d. e. OTHER PROCEDURE NPICODE DATE DATE

FIRST

NPI

b LAST FIRST

c NPI

d LAST FIRST

UB-04 CMS-1450

7

10 BIRTHDATE 11 SEX 12 13 HR 14 TYPE 15 SRC

DATE

16 DHR 18 19 20

FROM

21 2522 26 2823 27

CODE FROM

DATE

OTHER

PRV ID

THE CERTIFICATIONS ON THE REVERSE APPLY TO THIS BILL AND ARE MADE A PART HEREOF.

b

.INFO

BEN.

CODEOTHER PROCEDURE

THROUGH

29 ACDT 30

3231 33 34 35 36 37

38 39 40 41

42 REV. CD. 43 DESCRIPTION 45 SERV. DATE 46 SERV. UNITS 47 TOTAL CHARGES 48 NON-COVERED CHARGES 49

52 REL51 HEALTH PLAN ID 53 ASG. 54 PRIOR PAYMENTS 55 EST. AMOUNT DUE 56 NPI

57

58 INSURED’S NAME 59 P.REL 60 INSURED’S UNIQUE ID 61 GROUP NAME 62 INSURANCE GROUP NO.

64 DOCUMENT CONTROL NUMBER 65 EMPLOYER NAME

66 67 68

69 ADMIT 70 PATIENT 72 73

74 75 76 ATTENDING

80 REMARKS

OTHER PROCEDURE

a

77 OPERATING

78 OTHER

79 OTHER

81CC

CREATION DATE

3a PAT.CNTL #

24

b. MED.REC. #

44 HCPCS / RATE / HIPPS CODE

PAGE OF

APPROVED OMB NO. 0938-0997

e

a8 PATIENT NAME

50 PAYER NAME

63 TREATMENT AUTHORIZATION CODES

6 STATEMENT COVERS PERIOD

9 PATIENT ADDRESS

17 STAT STATE

DX REASON DX 71 PPS

CODE

QUAL

LAST

LAST

National UniformBilling CommitteeNUBC™

OCCURRENCE

QUAL

QUAL

QUAL

CODE DATE

A

B

C

A

B

C

A

B

C

A

B

C

A

B

C

a

b

a

b

Individual Claim

28%

11% 44%

17%

Macro Level Summary

Scope of Analysis Narrow Broad

Opportunity

Low

High


The Revenue Opportunity

4

__ __ __

1 2 4 TYPEOF BILL


a

b

c

d

DX

ECI

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

A

B

C


a b c a b c

a

b c d




TOTALS


FIRST


FIRST

NPI

b LAST FIRST

c NPI

d LAST FIRST

UB-04 CMS-1450

7


DATE

16 DHR 18 19 20

FROM

21 2522 26 2823 27

CODE FROM

DATE

OTHER

PRV ID


b

.INFO

BEN.

CODEOTHER PROCEDURE

THROUGH

29 ACDT 30

3231 33 34 35 36 37

38 39 40 41



57



66 67 68


74 75 76 ATTENDING

80 REMARKS

OTHER PROCEDURE

a

77 OPERATING

78 OTHER

79 OTHER

81CC

CREATION DATE

3a PAT.CNTL #

24

b. MED.REC. #


PAGE OF


e

a8 PATIENT NAME

50 PAYER NAME



9 PATIENT ADDRESS

17 STAT STATE

DX REASON DX 71 PPS

CODE

QUAL

LAST

LAST


OCCURRENCE

QUAL

QUAL

QUAL

CODE DATE

A

B

C

A

B

C

A

B

C

A

B

C

A

B

C

a

b

a

b

Individual Claim

28%

11% 44%

17%

Macro Level Summary

Scope of Analysis Narrow Broad

Opportunity

Low

High

The Denials Opportunity


Coding

Traditional Analytics Have Hit the Wall

5

Hypotheses Data Analysts Insight

Weeks/Months/Years/Never


Complexity is the Challenge

6

c Dataset

# of possible insights in Dataset is exponential

function of size

# of possible insights in the Dataset after it grows another 40%


The Solution is Machine Intelligence

7

Automated Discovery

Data Insights

Minutes

Decision Maker


How It Works

8

•  Apply algorithms to a data set

•  Statistical •  Geometric •  Machine learning

•  Construct a multi-faceted notion of similarity

•  Cluster similar data points

•  Create similarity maps that reveal mid-level patterns


Similarity Maps Using Machine Intelligence

9

Nodes are groups of similar data points Edges connect similar nodes


Surface Mid-Level Patterns of Denials

10

Discover distinct groups of similar claims that have high concentration of denials

Concentration of Denied Claims

Low High


Identify Hot Spots

11

Uncover the unique attributes of each group to produce rich descriptions of denials patterns

Concentration of Denied Claims

Low High


How Machine Intelligence Reduces Denials

1.  Identifies drivers of rejections and denials for groups of claims

2.  Prioritizes & streamlines work queue for claims resubmission

3.  Informs upstream process changes to prevent future denials

12


Case Study

13

Identified new patterns of denied claims. Identified specific drivers for known patterns of denials, suggesting new options for intervention.

SOLUTION

Despite years of effort, the provider continued to write off tens of millions of dollars annually.

PROBLEM

Solution is projected to address 70% of denied dollars and will continue to identify new denials patterns as they arise.

OUTCOME


Data included: •  UB-04 hospital billing data •  Additional account data from provider’s

financial databases •  Associated 835 remit transactions (denials,

rejections, underpayments, payments) •  Primary payer data •  Personal or family guarantors

__ __ __

1 2 4 TYPEOF BILL


a

b

c

d

DX

ECI

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

A

B

C


a b c a b c

a

b c d




TOTALS


FIRST


FIRST

NPI

b LAST FIRST

c NPI

d LAST FIRST

UB-04 CMS-1450

7


DATE

16 DHR 18 19 20

FROM

21 2522 26 2823 27

CODE FROM

DATE

OTHER

PRV ID


b

.INFO

BEN.

CODEOTHER PROCEDURE

THROUGH

29 ACDT 30

3231 33 34 35 36 37

38 39 40 41



57



66 67 68


74 75 76 ATTENDING

80 REMARKS

OTHER PROCEDURE

a

77 OPERATING

78 OTHER

79 OTHER

81CC

CREATION DATE

3a PAT.CNTL #

24

b. MED.REC. #


PAGE OF


e

a8 PATIENT NAME

50 PAYER NAME



9 PATIENT ADDRESS

17 STAT STATE

DX REASON DX 71 PPS

CODE

QUAL

LAST

LAST


OCCURRENCE

QUAL

QUAL

QUAL

CODE DATE

A

B

C

A

B

C

A

B

C

A

B

C

A

B

C

a

b

a

b

15

Example of data used for analysis Typical claim form


Investigate Hot Spots

Staging DB, Data

Transforms

Generate Similarity Maps

Color by % of

Denials

Denials Management Workflow

16

Provider

Claims Data Output Denials Hot Spot

Report

Identify drivers of rejections Drive process modifications Reduce denials upstream


How It Works

17

•  Use a multi-faceted notion of similarity to cluster similar data points

•  Create similarity maps that reveal mid-level

patterns

•  Nodes are groups of similar data points •  Edges connect similar nodes •  Colors let you see values of interest •  Position of a node on the screen doesn’t matter


Uncovering Characteristics of a Hotspot

18

HOTSPOT ANALYSIS Dataset: denials_100K_wide_150429.tsv Network Parameters: Prin DX Codes + HCPCS + Payor1 (Red = under-‐represented) Summary Medical Necessity Rejec;ons Medical Necessity Denials Hotspot Claims Volume Percent Amount Volume Percent Amount Island A 2947 2590 87.89% $2,602,381 158 5.36% $34,357

Top Payors % Rejected Rejected $ Top Providers % Rejected Rejected $ Top Bill Types % in Group Rejected $ PAYOR 1 89.11% $2,318,982 Clinic 1 45.13% $1,174,455 **131 89.75% $2,335,637 PAYOR 2 1.97% $51,267 Hospital 1 12.76% $332,064 **851 9.30% $242,021 PAYOR 3 1.39% $36,173 Hospital 2 10.72% $278,975 TOTAL 99.05% $2,577,659 Hospital 3 4.21% $109,560 TOTAL 92.47% $2,406,422 TOTAL 72.82% $1,895,054

Revenue Codes Percent HCPCS Descrip;on Code Percent Medical/Surgical Supplies And Devices -‐ Sterile Supply 98.47% Normal saline soluYon infus J7040 54.22% Pharmacy -‐ Drugs With Detailed Coding 98.27% Tissue exam by pathologist 88305 36.34% Recovery Room -‐ General 97.93% Colonoscopy and biopsy 45380 26.87%

Gastro-‐IntesYnal Services -‐ General 94.03% Colonoscopy w/lesion removal 45385 22.87%

Pharmacy -‐ General 81.20% Colon ca scrn not hi rsk ind G0121 22.63% Anesthesia -‐ General 72.62% Colorectal scrn; hi risk ind G0105 17.92%

Laboratory Pathological -‐ Histology 36.41% Colonoscopy w/lesion removal 45384 11.03%

Medical/Surgical Supplies And Devices -‐ General 32.10% Anesthesia -‐ Anesthesia Incident To Other DiagnosYc Services 22.57% Laboratory -‐ Chemistry 18.49%


Recap

1.  Identifies drivers of rejections and denials for groups of claims

2.  Prioritizes & streamlines work queue for claims resubmission

3.  Informs upstream process changes to prevent future denials

19


Facilitating the ICD-10 Transition

20

Code Agnostic

Understand the Long

Tail

Infer New Rules as they

Arise

Automated & Adaptive


Re-Cap, Documents + Q&A

•  The challenge of denials management is complexity – the low-hanging fruit is gone •  Using Machine Intelligence, providers can find

actionable patterns of denials •  This presentation and other docs in Handouts •  To connect with me: [email protected]

or ayasdi.com/healthcare or [email protected]

21