ModifiedMedicarePolicy N/A

HHS-Developed Risk Adjustment Model Algorithm 'Do It Yourself (DIY)' Software Instructions for the 2018 Benefit Year

Instructions and algorithm for HHS-developed 'Do It Yourself' software to calculate enrollee risk scores under the HHS risk adjustment models for the 2018 benefit year; intended for issuers of individual and small group market plans and others operating the HHS models.

Policy Summary

PayerMedicare

PolicyHHS-Developed Risk Adjustment Model Algorithm 'Do It Yourself (DIY)' Software Instructions for the 2018 Benefit Year

Policy CodePolicy N/A

Change TypeModel version update and RXC additions

Effective DateApr 6, 2018

Next Review DateN/A

Key ActionCreate the required person, diagnosis, NDC, and HCPCS input files and generate RXC and HCC indicators per the updated tables and hierarchies.

SourceLink

POLICY UPDATE CHANGES

Model classification updated from V04 to V05 HHS-HCC with HCC 37 split into HCC 37_1 (Chronic Viral Hepatitis C) and HCC 37_2 (Chronic Hepatitis, Except Chronic Viral Hepatitis C).

Prescription Drug Categories (RXCs) and RXC interactions were added to the adult models for the 2018 benefit year.

Table 2 updated to add 2018 CPT/HCPCS codes used for diagnosis filtering and includes 2018 quarterly updates as of January 1, 2018.

Table 3 ICD-10 to CC Crosswalk updated to include V05 HCC assignments and removed FY2017/CY2017 MCE columns; FY2018 ICD-10 and MCE edits should be used.

Table 9 Model Factors updated using MarketScan data (2013-2015) to include V05 HCC variables and new RXCs/RXC interactions for adult models.

5concurrent models

Trek Health ingests and normalizes Transparency in Coverage data and payer policy updates to give provider organizations a clear view of how commercial reimbursement behaves across markets, payers, and services. Our platform transforms raw payer disclosures into structured intelligence that supports contract evaluation, payer negotiations, and service line strategy. By combining market benchmarks with ongoing policy visibility, Trek helps teams identify variability, risk, and opportunity in commercial reimbursement. The result is faster insight, stronger negotiating positions, and more informed financial decisions.

years of MarketScan data

60%high-cost risk pool %

Tables 1-11supporting tables

4required input files

Inclusion and Algorithm Criteria

Input file inclusion criteria

Rules for including diagnoses and claims in the diagnosis-level input file for CC/HCC creation:

Include only ICD-10-CM diagnosis codes listed in Table 3; codes not listed may be present in the diagnosis-level file but are ignored by the algorithm.

User is responsible for creating the diagnosis-level file and must evaluate each claim/encounter to determine whether diagnoses meet the allowable-source rules before input to the algorithm.

Professional claims: For professional records, include diagnoses only from records that have at least one line item with an acceptable CPT/HCPCS code from Table 2; if at least one acceptable line exists, use all header diagnoses; if no acceptable service lines, exclude all diagnoses from that record.

Professional line-item diagnoses: For professional records, if a line item has an acceptable CPT/HCPCS code, include diagnoses from that line item.

Inpatient facility claims: For inpatient facility records, include all header diagnoses when facility bill type equals 111 or 117 (inpatient admit through discharge or inpatient replacement of prior claim); no procedure screen is applied.

Outpatient facility claims: For outpatient facility records, restrict to facility bill type codes 131,137,711,717,761,767,771,777,851,857 and include header diagnoses only if at least one acceptable CPT/HCPCS service line is present; otherwise exclude diagnoses.

Apply fiscal year code-validity checks using Table 3: for this preliminary 2018 software, include ICD-10 diagnoses with service dates 2018-01-01 through 2018-09-30 only if Table 3's Code Valid FY2018 column = Y; codes outside that window or without a Y should be excluded until Table 3 is updated for later dates.

Detect and, where possible, split bundled mother-newborn claims into separate mother and newborn records (separate enrollee ID, sex, and age). If split is not possible, assign diagnoses according to enrollee age/sex and apply MCE age/sex edits as described in the algorithm documentation.

NDC-level file requirement: include normalized 11-digit NDCs (HIPAA format) from pharmacy or inpatient pharmacy claims within the benefit year; NDCs not listed in Table 10a are ignored by the RXC creation step.

HCPCS-level file requirement: include HCPCS codes from inpatient, outpatient, and professional claims within the benefit year; HCPCS codes not listed in Table 10b may be present but will be ignored by the algorithm for RXC creation.

CC/HCC and RXC creation rules

Rules to create CCs, HCCs, and RXCs from input files

Create CC and HCC 0/1 indicator variables for each CC/HCC listed in Tables 3 and 4; initialize CCs and HCCs to 0 for each person before processing diagnoses.

Apply Medicare Code Edits (MCEs) for age and sex using AGE_AT_DIAGNOSIS and the FY2018 MCE Age/Sex Condition columns in Table 3; only if MCE conditions are met set CCs per the CC Age Split/CC Sex Split and Additional CC logic using AGE_LAST where specified.

After all diagnoses are processed and CC variables created, create HCCs using Table 4 hierarchies: set HCC to 1 when corresponding CC = 1, and apply hierarchy exclusions by setting lower-priority HCCs to 0 when a higher-priority HCC (as listed in Table 4 columns B/C) is present.

Create 12 RXC 0/1 indicator variables (RXC_01..RXC_12); set RXC to 1 when a person's NDCs (Table 10a) or HCPCS codes (Table 10b) map to that RXC, and initialize RXCs to 0 prior to mapping.

Apply RXC hierarchies from Table 11: when a person has an RXC listed in column A, set RXCs listed in column B to 0 to enforce exclusions; RXC interactions are created in a later step.

Algorithm implementation criteria

Implementation steps required to produce model variables and scores:

Create 12 RXC binary indicators using Tables 10a (NDC) and 10b (HCPCS); set RXC_X = 1 when mapped NDCs/HCPCS are present for a person; then apply Table 11 hierarchies to zero out excluded RXCs.

Create demographic age-sex indicator variables using Table 5 and AGE_LAST; assign each person to one of 28 mutually exclusive age-sex variables appropriate to adult, child, or infant models.

For adults, create SEVERE_V3 by setting it to 1 if the adult has at least one of the 8 specified HCCs (Table 6); then create 17 adult HCC group indicators (GXX) per Table 6 and set group=1 if any member HCC=1.

For children, create 17 child HCC group indicators per Table 7 and set each group indicator to 1 if the child has at least one HCC in that group.

Create interaction variables: construct the severe-illness interaction variables (16), aggregate them into INT_GROUP_H and INT_GROUP_M per predicted expenditures rules (INT_GROUP_H = 1 if any high-cost interactions present; INT_GROUP_M = 1 if any medium-cost interactions present and INT_GROUP_H = 0). Also create 15 RXC x HCC interaction variables for adults using RXC_01..RXC_12 and specified HCCs.

Code Crosswalks, Validity, and Formats

Risk adjustment code crosswalksmixed

Table 2 CPT/HCPCS list	CPT/HCPCS codes used to filter diagnosis sources for risk adjustment (2017 and 2018 columns).
Table 3 ICD-10 to CC crosswalk	ICD-10 codes mapped to Condition Categories, use FY2018 code-valid info for service dates Jan–Sep 2018.
Table 10a NDC list	NDCs to be used for RXC creation (11-digit HIPAA format).
Table 10b HCPCS list	HCPCS codes to be used for RXC creation.

Code crosswalks and formatsmixed

ICD-10-CM	Diagnosis codes used for CC and HCC creation; 7-character, no periods, left-justified; age/sex edits applied.
NDC (normalized 11-digit)	NDCs mapped to RXCs per Table 10a; HIPAA 11-digit format; source RxNorm recommended; store as string to preserve leading zeroes.
HCPCS (5-character)	HCPCS drug codes mapped to RXCs per Table 10b; included from inpatient/outpatient/professional claims with dates in benefit year.

Drug and procedure crosswalks for RXCsmixed

NDC	Used to set RXC indicators via Table 10a; NDCs must be in 11-digit, no-dash HIPAA format; NDCs not in Table 10a are ignored by the algorithm.
HCPCS	Used to set RXC indicators via Table 10b; HCPCS codes not in Table 10b are ignored by the algorithm.

inv-07: FY2018 code validity window

Inclusion windowInclude ICD-10 diagnosis codes with service dates 2018-01-01 through 2018-09-30 only if the FY2018 Code Valid (Table 3, column E) = Y.

Source tableUse Table 3 (ICD-10 to CC crosswalk) FY2018 Code Valid column to determine validity for Jan–Sep 2018.

Ignored codesICD-10 codes not listed in Table 3 may be present in the diagnosis-level file but will be ignored by the algorithm unless marked valid in FY2018.

inv-08: Code validity for benefit year

ICD-10 validityUse the FY2018 'Code Valid' column in Table 3 (column E) to determine which ICD-10 codes are valid for the 2018 benefit year.

NDC/HCPCS inclusion

Required Actions and Operational Notes

Billing Rule

Use limitations

CMS provides supplemental risk adjustment software (SAS and HHS 'Do It Yourself' simulation tools) to help issuers simulate 2018 benefit-year risk scores. This software is optional and is not required to submit claims to the EDGE server. Do not use the DIY software to pre-filter or exclude claims prior to EDGE submission — the EDGE server may apply additional operational rules and the DIY tool is solely a simulation aid. Use the software cautiously for estimating transfers because risk adjustment transfers depend on data submitted by all issuers in the State market risk pool.

Do not filter or remove claims using DIY software prior to EDGE submission.
DIY software is a simulation tool only, not a substitute for EDGE submission or required processing.
Estimate transfers using additional data sources due to dependence on market-wide submissions.

Note

HCPCS file creation and use

Key Variables and Concepts

inv-15: HHS-Hierarchical Condition Categories (HCC) and Condition Categories (CC) used in HHS risk adjustment models

DefinitionHHS-Hierarchical Condition Categories (HCC) and Condition Categories (CC) used in HHS risk adjustment models; see Table 4 for HCC hierarchies and Table 3 for ICD-10 to CC mapping.

ScopeHCC/CC lists used by the adult, child, and infant models; models reference HCCs up to HCC254 as applicable.

Hierarchy applicationApply HCC hierarchies (Table 4) during variable creation so excluded HCCs are set to 0 when higher-priority HCCs are present.

inv-16: Prescription Drug Category used in adult models; mapped from NDC and HCPCS using Tables 10a and 10b

RXC mappingPrescription Drug Categories (RXC_01–RXC_12) are mapped from NDCs (Table 10a) and HCPCS (Table 10b) crosswalks for adult models.

RXC usage

Policy Summary

PayerMedicare

PolicyHHS-Developed Risk Adjustment Model Algorithm 'Do It Yourself (DIY)' Software Instructions for the 2018 Benefit Year

Policy CodePolicy N/A

Change TypeModel version update and RXC additions

Effective DateApr 6, 2018

Next Review DateN/A

Key ActionCreate the required person, diagnosis, NDC, and HCPCS input files and generate RXC and HCC indicators per the updated tables and hierarchies.

SourceLink

On This Page

Input file formats and required fields: PERSON file must include ENROLID, SEX, DOB, AGE_LAST, METAL, CSR_INDICATOR, ENROLDURATION; DIAG file must include ENROLID, DIAG (ICD-10), DIAGNOSIS_SERVICE_DATE, AGE_AT_DIAGNOSIS; NDC file must include ENROLID and normalized 11-digit NDCs; HCPCS file must include ENROLID and HCPCS codes. The user must create these files; the algorithm does not generate them.

HCPCS codes not listed in Table 10b may be included in the HCPCS input file but will be ignored by the algorithm and will not contribute to RXC indicators for adult risk score calculations.

The algorithm outputs demographic variables, CCs, HCCs, RXCs, model-specific enrollment and maturity variables, score variables for adult/child/infant models, and CSR-adjusted scores; final unadjusted and CSR-adjusted scores depend on enrollee metal level and CSR indicator.

Use the FY2018 Code Valid column in Table 3 to validate ICD-10 codes for the benefit year; only ICD-10 codes with 'Y' in FY2018 Code Valid should be used for constructing CCs/HCCs for the preliminary 2018 software release.

Derive ENROLDURATION (1-12) for each person in the PERSON file and create ED_1..ED_11 mutually-exclusive enrollment-duration indicators for adult models: set ED_X = 1 when ENROLDURATION = X; if ENROLDURATION = 12, set all ED indicators = 0.

For infants, create 5 severity-level indicators and 5 maturity-category indicators, apply hierarchies so each infant has one assigned severity and one maturity category, reassign male age 0 without newborn HCC to age 1, and create 25 maturity x severity interaction indicators.

Compute unadjusted model scores by metal level using Table 9 factors: for each person sum factors for demographic variables, HCCs, HCC groups, severe-illness interaction group (if any), enrollment-duration indicator (if any), RXCs 1-10, and RXC interactions to produce SCORE_* by metal level; select the enrollee's metal level (Table 1) to assign SCORE_* as the person's model score.

Create CSR-adjusted scores per metal level using directions in Tables 6-8 and Table 9: produce CSR_ADJUSTED_SCORE_* variables for each metal level and then select the CSR-adjusted score corresponding to the enrollee's metal level to produce CSR_ADJUSTED_SCORE for adult/child/infant models.

Only NDCs listed in Table 10a and HCPCS codes listed in Table 10b are used to create RXCs; other codes may be present in input files but are ignored by the algorithm for score calculations.

Format requirementNDCs must be in normalized 11-digit, no-dash HIPAA format to match Table 10a formatting and preserve leading zeroes.

inv-09: Enrollment duration categorization

ENROLDURATION valuesENROLDURATION is an integer 1–12 representing months enrolled in the benefit year (1=1–31 days ... 12=336–366 days).

ED indicators creationENROLDURATION creates 11 mutually exclusive adult enrollment-duration indicators ED_1 through ED_11; each ED_X = 1 when ENROLDURATION = X.

Full-year ruleIf ENROLDURATION = 12 (12-month enrollment), all ED_1..ED_11 are set to 0.

Create a HCPCS-level input file (HCPCS file) for use in RXC creation. Include HCPCS drug codes from inpatient, outpatient, and professional claims with discharge or through dates within the benefit year. The HCPCS-level file must be created by the user prior to running the algorithm; the algorithm does not generate this data set. Evaluate each claim to determine whether its HCPCS codes are included. HCPCS codes not listed in Table 10b may be present in the file but will be ignored by the algorithm and will not produce RXC flags for adult risk score calculations.

Include HCPCS codes from inpatient, outpatient, and professional claims with discharge/through dates in the benefit year.
User must create the HCPCS-level file; the algorithm does not create it.
Evaluate each claim to determine inclusion; codes not in Table 10b are ignored by the algorithm.

Billing Rule

RXC creation and hierarchy

Create RXC indicator variables from the NDC-to-RXC and HCPCS-to-RXC crosswalks and apply hierarchical exclusions. Start by creating 12 RXC 0/1 indicator variables (RXC_01 through RXC_12) and set them initially to 0 for each person. Set RXC flags to 1 when acceptable NDCs (using Table 10a crosswalk) or acceptable HCPCS codes (using Table 10b crosswalk) are present for the enrollee. Then apply the hierarchies in Table 11: for each hierarchy, set RXCs in column B to 0 when the person has the RXC in column A set to 1. RXC interactions will be created later.

Create 12 RXC binary indicators: RXC_01 ... RXC_12, initialize to 0.
Set RXC to 1 if acceptable NDCs (Table 10a) are present for the person.
Set RXC to 1 if acceptable HCPCS codes (Table 10b) are present for the person.
Apply Table 11 hierarchies: if RXC in column A = 1, zero out RXC(s) in column B.
Do not create RXC interactions here; those are created in a subsequent step.

Note

Variable creation steps

Follow the variable creation steps in order to build the inputs required for score calculation. Create demographic age-sex variables (use AGE_LAST) to place each person into one of the model-specific age-sex buckets. For adults, create the SEVERE_V3 indicator (set to 1 if at least one of the eight severe HCCs is present). Build HCC group variables for adults and children (set to 1 if any HCC in the group is present). Create adult HCC interaction variables and aggregate them into INT_GROUP_H and INT_GROUP_M per Table 6. For adults create enrollment duration indicators ED_1 through ED_11 based on ENROLDURATION. Use RXC_01–RXC_12 together with HCC flags to create the 15 RXC interaction variables listed in Table 6 (set to 1 when both the RXC and the specified HCC(s) are present). For infants, create maturity and severity category indicators (IHCC_ prefixes), apply hierarchies so each infant has a single severity and a single maturity flag, reassign male age-0 infants without newborn HCCs to Age 1 as specified, and create the 25 mutually exclusive maturity-by-severity interaction variables. After creating these variables, use Table 9 factors and the instructions in Tables 6–8 to compute unadjusted and CSR-adjusted scores by metal level for adults, children, and infants.

Demographics: create 28 mutually exclusive age-sex variables using AGE_LAST (adult/child/infant model distinctions).
Adult severe illness: create SEVERE_V3 = 1 if any of the 8 severe HCCs (Table 6) are present.
HCC groups: create adult (17) and child (17) HCC group indicators set to 1 if any constituent HCC is present.
HCC interactions: create specified adult interactions and aggregate into INT_GROUP_H and INT_GROUP_M per Table 6.
Enrollment duration: create ED_1–ED_11 for adults based on ENROLDURATION; EDs all zero if ENROLDURATION=12.
RXC interactions: create 15 RXC x HCC interaction variables (set to 1 when both RXC and HCC criteria met).
Infant variables: create 5 severity and 5 maturity indicators, apply hierarchies so each infant has a single severity and single maturity flag, reassign male age 0 without newborn HCCs to Age 1, and create 25 maturity-by-severity interaction variables.
Scoring: use factors from Table 9 and the created variables to compute unadjusted scores (by metal level) and CSR-adjusted scores using the detailed directions in Tables 6–8.

RXCs 1–10 are included as stand-alone variables and interactions in adult models; RXCs 11–12 are included only via interactions.

Input sourcesNDCs and HCPCS used for RXC creation must come from pharmacy or inpatient claims (NDC) and inpatient/outpatient/professional claims (HCPCS) within the benefit year.

inv-17: Input file of allowable ICD-10-CM diagnosis codes (per Table 3) prepared from claims/encounters

Source of codesDiagnosis-level input file must include only ICD-10-CM diagnosis codes listed in Table 3 (ICD-10 to CC crosswalk).

Record evaluationUsers must evaluate each claim/encounter to determine whether its diagnoses are allowable and include eligible diagnoses in the DIAG file for CC/HCC creation.

Ignored codesICD-10 codes not listed in Table 3 may be present but are ignored by the algorithm for CC/HCC creation.

inv-18: Four required input data files: person-level, diagnosis-level, NDC-level, and HCPCS-level

Required filesFour input files required: person-level, diagnosis-level, NDC-level, and HCPCS-level. Users are responsible for creating these files.

PERSON file contentsPERSON file must include ENROLID, SEX, DOB, AGE_LAST, METAL, CSR_INDICATOR, and ENROLDURATION among other demographic/enrollment fields.

DIAG/NDC/HCPCS filesDIAG file must include ENROLID, DIAG, DIAGNOSIS_SERVICE_DATE, AGE_AT_DIAGNOSIS; NDC/HCPCS files must include ENROLID and normalized codes as specified in Tables 10a/10b.

inv-19: Person-level enrollment duration variable with allowable integer values 1-12 representing months enrolled in benefit year; used to create adult-model enrollment duration indicators (ED_1 - ED_11).

Variable descriptionENROLDURATION is a person-level enrollment duration variable with allowable integer values 1–12 representing months enrolled in the benefit year.

UsageAlthough used only in adult models, it is recommended to create ENROLDURATION for all enrollees for consistency; values are ignored for child/infant models.

Derived indicatorsENROLDURATION is used to create adult-model enrollment duration indicators ED_1 through ED_11 where ED_X = 1 if ENROLDURATION = X; ENROLDURATION = 12 => all EDs = 0.

inv-20: Person-level cost-sharing reduction indicator (values 0-13) mapping enrollee to CSR plan variations and RA factors; non-CSR or unknown = 0.

Indicator valuesCSR_INDICATOR is a person-level variable with allowable integer values 0–13 mapping enrollees to CSR plan variations and RA factors (1–13 = CSR variants; 0 = non-CSR/unknown).

ExamplesExamples include 1 = 94% AV Silver plan variation (CSR RA factor 1.12), 3 = 73% AV Silver (RA factor 1.00), up to indicator 13 for premium assistance variants.

Usage in scoringCSR_INDICATOR determines which CSR-adjusted score factors to apply when computing CSR-adjusted risk scores by metal level.

inv-21: Medicare Code Edits applied to diagnoses to verify appropriateness by AGE_AT_DIAGNOSIS and sex before assigning CCs.

PurposeApply Medicare Code Edits (MCE) Age and Sex conditions to each diagnosis using AGE_AT_DIAGNOSIS and diagnosis sex condition before assigning CCs.

Table 3 fieldsTable 3 includes FY2018 MCE Age Condition (column F) and FY2018 MCE Sex Condition (column G); diagnoses must meet these conditions for CC assignment.

Assignment effectIf MCE age/sex conditions are not met, the CC for that diagnosis should not be assigned.

inv-22: Prescription Drug Category indicator variables (RXC_01 - RXC_12) created from NDC and HCPCS crosswalks with hierarchy exclusions applied per Table 11.

Indicator setCreate 12 RXC binary indicator variables RXC_01 through RXC_12 from NDC (Table 10a) and HCPCS (Table 10b) crosswalks.

Hierarchy exclusionsApply RXC hierarchies from Table 11 to set excluded RXCs to 0 when a higher-priority RXC is present.

Interaction readinessRXC indicators are used later to create 15 RXC x HCC interaction variables for adult models per Table 6.

inv-23: Severe illness indicator for adults (set to 1 if adult has >=1 of 8 specified HCCs)

Severe illness ruleFor adults, set SEVERE_V3 = 1 if the adult has at least one of the eight specified HCCs listed in Table 6 (rows 5-12 of column E); otherwise set to 0.

Creation orderSEVERE_V3 is created before adult HCC group variables and severe-illness interactions as part of the adult variable creation sequence.

Use in interactionsSEVERE_V3 and related interactions are aggregated into INT_GROUP_H and INT_GROUP_M for use in score calculations.

inv-24: IHCC_ prefix variables representing infant severity levels and maturity categories used to create maturity x severity interactions

Prefix and purposeIHCC_ prefix variables represent infant severity levels and maturity categories used to create maturity x severity interaction variables for infant models.

Severity and maturity setsCreate five IHCC_ severity level indicators and five IHCC_ maturity category indicators, apply hierarchies so each infant has one severity and one maturity set to 1.

InteractionsUse the IHCC_ severity and maturity indicators to create 25 mutually-exclusive maturity x severity interaction variables for infants.

inv-25: Cost-Sharing Reduction adjusted risk score computed per metal level using Tables 6-8 directions

ComputationCompute CSR-adjusted risk scores per metal level using directions in Tables 6–8 and model factors in Table 9; create CSR_ADJUSTED_SCORE_* variables for the enrollee's metal level.

Adult/child/infantCSR-adjusted scores are created separately for adult, child, and infant models based on the enrollee's metal level (Table 1) using instructions in Tables 6–8.

Factor applicationCSR-adjusted score uses CSR indicator mapping and metal-level factors from Table 9 to adjust the unadjusted score as directed in the tables.