Risk Adjustment Score (RAF) implementation uses the Hierarchical Condition Category (HCC) and Prescription Hierarchical Condition Category (RxHCC) Risk Adjustment models from the Centers for Medicare & Medicaid Service (CMS). HCC groups similar conditions in terms of healthcare costs and similarities in the diagnosis, and the model uses any ICD code that has a corresponging HCC and RxHCC category in the computation, discarding other ICD codes.
CMS-HCC Risk Score Calculation
This module supports versions 22, 23, 24, 28 and ESRDV21 of the CMS-HCC risk adjustment model and needs the following parameters in order to calculate the risk score:
- ICD Codes (Obtained by, e.g., our pretrained model
sbiobertresolve_icd10cm_augmented_billable_hccfrom theSentenceEntityResolverModel annotator) - Age (Obtained by, e.g., our pretrained model
ner_jslfrom theNerModel annotator) - Gender (Obtained by, e.g., our pretrained model
classifierdl_gender_biobertfrom the ClassifierDLModel annotator) - The eligibility segment of the patient (information from the health plan provider)
- The original reason for entitlement (information from the health plan provider)
- If the patient is in Medicaid or not (information from the health plan provider)
Available Softwares and Profiles
CMS-HCC module supports the following versions and years:
| Version 22 | Version 23 | Version 24 | Version 28 | ESRD V21 |
|---|---|---|---|---|
| Year 2017 | Year 2018 | Year 2019 | Year 24 | Year 19 |
| Year 2018 | Year 2019 | Year 2020 | ||
| Year 2019 | Year 2021 | |||
| Year 2020 | Year 2022 | |||
| Year 2021 | ||||
| Year 2022 |
Usage
The module can perform the computations given a data frame containing the required information (age, gender, ICD codes, eligibility segment, the original reason for entitlement, and if the patient is in Medicaid or not). For example, given the dataset df:
+--------------+---+--------------------+------+-----------+----+--------+----------+
| filename|age| icd10_code|gender|eligibility|orec|medicaid| DOB|
+--------------+---+--------------------+------+-----------+----+--------+----------+
|mt_note_03.txt| 66|[C499, C499, D618...| F| CND| 1| false|1956-05-30|
|mt_note_01.txt| 59| [C801]| F| CFA| 0| true|1961-10-12|
|mt_note_10.txt| 16| [C6960, C6960]| M| CFA| 2| false|2006-02-14|
|mt_note_08.txt| 66| [C459, C800]| F| CND| 1| true|1956-03-17|
|mt_note_09.txt| 19| [D5702, K5505]| M| CPA| 3| true|2003-06-11|
|mt_note_05.txt| 57|[C5092, C5091, C5...| F| CPA| 3| true|1963-08-12|
|mt_note_06.txt| 63| [F319, F319]| F| CFA| 0| false|1959-07-24|
+--------------+---+--------------------+------+-----------+----+--------+----------+
Where column orec means original reason for entitlement and DOB means date of birth (can also be used to compute age). You can use any of the available profiles to compute the scores (in the example, we use version 24, year 2020):
from johnsnowlabs import medical
# Creates the risk profile
df = df.withColumn(
"hcc_profile",
medical.profileV24Y20(
df.icd10_code,
df.age,
df.gender,
df.eligibility,
df.orec,
df.medicaid
),
)
# Extract relevant information
df = (
df.withColumn("risk_score", df.hcc_profile.getItem("risk_score"))
.withColumn("hcc_lst", df.hcc_profile.getItem("hcc_lst"))
.withColumn("parameters", df.hcc_profile.getItem("parameters"))
.withColumn("details", df.hcc_profile.getItem("details"))
)
df.select(
"filename",
"risk_score",
"icd10_code",
"age",
"gender",
"eligibility",
"orec",
"medicaid",
).show(truncate=False)
+--------------+----------+---------------------------------------------+---+------+-----------+----+--------+
|filename |risk_score|icd10_code |age|gender|eligibility|orec|medicaid|
+--------------+----------+---------------------------------------------+---+------+-----------+----+--------+
|mt_note_01.txt|0.158 |[C801] |59 |F |CFA |0 |true |
|mt_note_03.txt|1.03 |[C499, C499, D6181, M069, C801] |66 |F |CND |1 |false |
|mt_note_05.txt|2.991 |[C5092, C5091, C779, C5092, C800, G20, C5092]|57 |F |CPA |3 |true |
|mt_note_06.txt|0.299 |[F319] |63 |F |CFA |0 |true |
|mt_note_08.txt|2.714 |[C459, C800] |66 |F |CND |1 |false |
|mt_note_09.txt|1.234 |[D5702, K5505] |19 |F |CPA |3 |true |
+--------------+----------+---------------------------------------------+---+------+-----------+----+--------+
For more details and usage examples, check the Medicare Risk Adjustment notebook from our Spark NLP Workshop repository.
CMS-RxHCC Risk Score Calculation
This module supports versions 05 and 08 of CMS-RxHCC risk adjustment model and needs the following parameters in order to calculate the risk score:
- ICD Codes (Obtained by, e.g., our pretrained model
sbiobertresolve_icd10cm_augmented_billable_hccfrom theSentenceEntityResolverModel annotator) - Age (Obtained by, e.g., our pretrained model
ner_jslfrom theNerModel annotator) - Gender (Obtained by, e.g., our pretrained model
classifierdl_gender_biobertfrom the ClassifierDLModel annotator) - The eligibility segment of the patient (information from the health plan provider)
- The original reason for entitlement (information from the health plan provider)
- End-stage renal disease indicator (ESRD) of the patient (information from the health plan provider)
Available Softwares and Profiles
CMS-RxHCC module supports the following versions and years:
| Version 05 | Version 08 |
|---|---|
| Year 20 | Year 22 |
| Year 21 | Year 23 |
| Year 22 | |
| Year 23 |
Usage
The module can perform the computations given a data frame containing the required information (age, gender, ICD codes, eligibility segment, the original reason for entitlement, and the ESRD indicator of the patient). For example, given the dataset rxhcc_df:
+--------------+---+------------------------------------------------------------------------------------------+------+--------------+----+-----+----------+
| filename|age| icd10_code|gender| eligibility|orec| esrd| DOB|
+--------------+---+------------------------------------------------------------------------------------------+------+--------------+----+-----+----------+
|mt_note_01.txt| 59| [C50.92, P61.4, C80.1]| F|CE_NoLowNoAged| 0| true|1961-10-12|
|mt_note_03.txt| 66| [C49.9, J18.9, C49.9, D61.81, I26, M06.9]| F| CE_NoLowAged| 1|false|1956-05-30|
|mt_note_05.txt| 57|[C50.92, C50.91, C50.9, C50.92, C50.9, C80.0, T78.40, R50.9, G25.81, P39.9, I97.2, C50.92]| F| CE_LTI| 3| true|1963-08-12|
|mt_note_06.txt| 63| [C45, F31.9, Q40.1, D35.00, R18, D20.1, F31.9, L72.0]| F| NE_Lo| 0|false|1959-07-24|
|mt_note_08.txt| 66| [J90, C45.9, J90, C45, C80.0]| F| NE_NoLo| 1| true|1956-03-17|
|mt_note_09.txt| 19| [D50.8, D57.0, O99.0, D57.02, K55.05, M46.42]| M| CE_LowNoAged| 3| true|2003-06-11|
|mt_note_10.txt| 16| [C69.60, C69.60]| M| NE_LTI| 2|false|2006-02-14|
+--------------+---+------------------------------------------------------------------------------------------+------+--------------+----+-----+----------+
Where column orec means original reason for entitlement, esrd means end-stage renal disease indicator and DOB means date of birth (can also be used to compute age).
from johnsnowlabs import medical
# Creates the risk profile
rxhcc_df = rxhcc_df.withColumn(
"rxhcc_profile",
medical.profileRxHCCV08Y23(
rxhcc_df.icd10_code,
rxhcc_df.age,
rxhcc_df.gender,
rxhcc_df.eligibility,
rxhcc_df.orec,
rxhcc_df.esrd
),
)
# Extract relevant information
rxhcc_df = (
rxhcc_df.withColumn("risk_score", df.rxhcc_profile.getItem("risk_score"))
.withColumn("parameters", df.rxhcc_profile.getItem("parameters"))
.withColumn("details", df.rxhcc_profile.getItem("details"))
)
df.select(
"filename",
"risk_score",
"age",
"icd10_code",
"gender",
"eligibility",
"orec",
"esrd",
).show(truncate=False)
+--------------+----------+---+------------------------------------------------------------------------------------------+------+--------------+----+-----+
| filename|risk_score|age| icd10_code|gender| eligibility|orec| esrd|
+--------------+----------+---+------------------------------------------------------------------------------------------+------+--------------+----+-----+
|mt_note_01.txt|0.575 | 59| [C50.92, P61.4, C80.1]| F|CE_NoLowNoAged| 0| true|
|mt_note_03.txt|0.367 | 66| [C49.9, J18.9, C49.9, D61.81, I26, M06.9]| F| CE_NoLowAged| 1|false|
|mt_note_05.txt|2.729 | 57|[C50.92, C50.91, C50.9, C50.92, C50.9, C80.0, T78.40, R50.9, G25.81, P39.9, I97.2, C50.92]| F| CE_LTI| 3| true|
|mt_note_06.txt|1.703 | 63| [C45, F31.9, Q40.1, D35.00, R18, D20.1, F31.9, L72.0]| F| NE_Lo| 0|false|
|mt_note_08.txt|1.38 | 66| [J90, C45.9, J90, C45, C80.0]| F| NE_NoLo| 1| true|
|mt_note_09.txt|1.677 | 19| [D50.8, D57.0, O99.0, D57.02, K55.05, M46.42]| M| CE_LowNoAged| 3| true|
|mt_note_10.txt|1.645 | 16| [C69.60, C69.60]| M| NE_LTI| 2|false|
+--------------+----------+---+------------------------------------------------------------------------------------------+------+--------------+----+-----+
For more details and usage examples, check the Medicare Risk Adjustment notebook from our Spark NLP Workshop repository.