Issue 970: reverse percentiles for AMI and life expectancy (#1018)

* switching to low

* fixing score-etl-post

* updating comments

* fixing comparison

* create separate field for clarity

* comment fix

* removing healthy food

* fixing bug in score post

* running black and adding comment

* Update pickles and add a helpful notes to README

Co-authored-by: Shelby Switzer <shelby.switzer@cms.hhs.gov>

data/data-pipeline/README.md

@@ -309,7 +309,7 @@ If you update the score in any way, it is necessary to create new pickles so tha
 
 It starts with the `data_pipeline/etl/score/tests/sample_data/score_data_initial.csv`, which is the first two rows of the `score/full/usa.csv`.
 
-To update this file, run a full score generation and then update the file as follows:
+To update this file, run a full score generation, then open a Python shell from the `data-pipeline` directory (e.g. `poetry run python3`), and then update the file with the following commands:
 ```
 import pickle
 from pathlib import Path
@@ -322,6 +322,8 @@ score_initial_df = pd.read_csv(score_csv_path, dtype={"GEOID10_TRACT": "string"}
 score_initial_df.to_csv(data_path / "data_pipeline" / "etl" / "score" / "tests" / "sample_data" /"score_data_initial.csv", index=False)
 ```
 
+Now you can move on to updating inidvidual pickles for the tests. Note that it is helpful to do them in this order:
+
 We have four pickle files that correspond to expected files:
 - `score_data_expected.pkl`: Initial score without counties
 - `score_transformed_expected.pkl`: Intermediate score with `etl._extract_score` and `etl. _transform_score` applied. There's no file for this intermediate process, so we need to capture the pickle mid-process.

data/data-pipeline/data_pipeline/etl/score/constants.py

@@ -81,7 +81,7 @@ TILES_SCORE_COLUMNS = [
     field_names.DIABETES_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.ASTHMA_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.HEART_DISEASE_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
-    field_names.LIFE_EXPECTANCY_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
+    field_names.LOW_LIFE_EXPECTANCY_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.TRAFFIC_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.FEMA_RISK_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.ENERGY_BURDEN_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
@@ -89,7 +89,7 @@ TILES_SCORE_COLUMNS = [
     field_names.LEAD_PAINT_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.DIESEL_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.PM25_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
-    field_names.MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD
+    field_names.LOW_MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD
     + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.POVERTY_LESS_THAN_200_FPL_FIELD
     + field_names.PERCENTILE_FIELD_SUFFIX,
@@ -115,7 +115,7 @@ TILES_SCORE_FLOAT_COLUMNS = [
     field_names.DIABETES_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.ASTHMA_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.HEART_DISEASE_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
-    field_names.LIFE_EXPECTANCY_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
+    field_names.LOW_LIFE_EXPECTANCY_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.TRAFFIC_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.FEMA_RISK_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.ENERGY_BURDEN_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
@@ -123,7 +123,7 @@ TILES_SCORE_FLOAT_COLUMNS = [
     field_names.LEAD_PAINT_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.DIESEL_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.PM25_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
-    field_names.MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD
+    field_names.LOW_MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD
     + field_names.PERCENTILE_FIELD_SUFFIX,
     field_names.POVERTY_LESS_THAN_200_FPL_FIELD
     + field_names.PERCENTILE_FIELD_SUFFIX,
@@ -137,7 +137,6 @@ DOWNLOADABLE_SCORE_INDICATOR_COLUMNS_BASIC = [
     field_names.DIABETES_FIELD,
     field_names.ASTHMA_FIELD,
     field_names.HEART_DISEASE_FIELD,
-    field_names.LIFE_EXPECTANCY_FIELD,
     field_names.TRAFFIC_FIELD,
     field_names.FEMA_RISK_FIELD,
     field_names.ENERGY_BURDEN_FIELD,
@@ -149,11 +148,11 @@ DOWNLOADABLE_SCORE_INDICATOR_COLUMNS_BASIC = [
     field_names.TOTAL_POP_FIELD,
 ]
 
-# For every indicator above, we want to include percentile and min-max normalized variants also
+# For every indicator above, we want to include percentile also.
 DOWNLOADABLE_SCORE_INDICATOR_COLUMNS_FULL = list(
     pd.core.common.flatten(
         [
-            [p, f"{p} (percentile)"]
+            [p, f"{p}{field_names.PERCENTILE_FIELD_SUFFIX}"]
             for p in DOWNLOADABLE_SCORE_INDICATOR_COLUMNS_BASIC
         ]
     )
@@ -165,8 +164,15 @@ DOWNLOADABLE_SCORE_COLUMNS = [
     field_names.COUNTY_FIELD,
     field_names.STATE_FIELD,
     field_names.SCORE_G_COMMUNITIES,
+    # Note: the reverse percentile fields get moved down here because
+    # we put the raw value in the download along with the *reversed* percentile.
+    # All other fields we put in f"{field_name}" and
+    # f"{field_name}{field_names.PERCENTILE_FIELD_SUFFIX}", which doesn't work for the
+    # reversed percentile fields.
     field_names.MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD,
-    field_names.MEDIAN_INCOME_AS_PERCENT_OF_STATE_FIELD
+    field_names.LOW_MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD
     + field_names.PERCENTILE_FIELD_SUFFIX,
+    field_names.LIFE_EXPECTANCY_FIELD,
+    field_names.LOW_LIFE_EXPECTANCY_FIELD + field_names.PERCENTILE_FIELD_SUFFIX,
     *DOWNLOADABLE_SCORE_INDICATOR_COLUMNS_FULL,
 ]

data/data-pipeline/data_pipeline/etl/score/etl_score.py

@@ -404,9 +404,7 @@ class ScoreETL(ExtractTransformLoad):
             field_names.POVERTY_LESS_THAN_150_FPL_FIELD,
             field_names.POVERTY_LESS_THAN_200_FPL_FIELD,
             field_names.AMI_FIELD,
-            field_names.MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD,
             field_names.MEDIAN_INCOME_FIELD,
-            field_names.LIFE_EXPECTANCY_FIELD,
             field_names.ENERGY_BURDEN_FIELD,
             field_names.FEMA_RISK_FIELD,
             field_names.URBAN_HEURISTIC_FIELD,
@@ -439,7 +437,6 @@ class ScoreETL(ExtractTransformLoad):
             field_names.CENSUS_UNEMPLOYMENT_FIELD_2010,
             field_names.CENSUS_POVERTY_LESS_THAN_100_FPL_FIELD_2010,
             field_names.CENSUS_DECENNIAL_TOTAL_POPULATION_FIELD_2009,
-            field_names.CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009,
             field_names.EXTREME_HEAT_FIELD,
             field_names.HEALTHY_FOOD_FIELD,
             field_names.IMPENETRABLE_SURFACES_FIELD,
@@ -468,7 +465,19 @@ class ScoreETL(ExtractTransformLoad):
             ReversePercentile(
                 field_name=field_names.READING_FIELD,
                 low_field_name=field_names.LOW_READING_FIELD,
-            )
+            ),
+            ReversePercentile(
+                field_name=field_names.MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD,
+                low_field_name=field_names.LOW_MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD,
+            ),
+            ReversePercentile(
+                field_name=field_names.LIFE_EXPECTANCY_FIELD,
+                low_field_name=field_names.LOW_LIFE_EXPECTANCY_FIELD,
+            ),
+            ReversePercentile(
+                field_name=field_names.CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009,
+                low_field_name=field_names.LOW_CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009,
+            ),
         ]
 
         columns_to_keep = (
@@ -505,10 +514,6 @@ class ScoreETL(ExtractTransformLoad):
 
             max_value = df_copy[numeric_column].max(skipna=True)
 
-            logger.info(
-                f"For data set {numeric_column}, the min value is {min_value} and the max value is {max_value}."
-            )
-
             df_copy[f"{numeric_column}{field_names.MIN_MAX_FIELD_SUFFIX}"] = (
                 df_copy[numeric_column] - min_value
             ) / (max_value - min_value)

data/data-pipeline/data_pipeline/ipython/compare_two_score_files_for_differences.ipynb

@@ -8,7 +8,10 @@
    "outputs": [],
    "source": [
     "import IPython\n",
+    "import os\n",
     "import pandas as pd\n",
+    "import pathlib\n",
+    "import sys\n",
     "\n",
     "module_path = os.path.abspath(os.path.join(\"../..\"))\n",
     "if module_path not in sys.path:\n",
@@ -28,12 +31,8 @@
    "outputs": [],
    "source": [
     "# Load\n",
-    "path_to_score_file_1 = (\n",
+    "path_to_score_file_1 = DATA_DIR / \"compare_two_score_csvs/usa (pre 970).csv\"\n",
-    "    DATA_DIR / \"score\" / \"csv\" / \"full\" / \"usa1.csv\"\n",
+    "path_to_score_file_2 = DATA_DIR / \"compare_two_score_csvs/usa (post 970).csv\"\n",
-    ")\n",
-    "path_to_score_file_2 = (\n",
-    "        DATA_DIR / \"score\" / \"csv\" / \"full\" / \"usa2.csv\"\n",
-    ")\n",
     "\n",
     "score_1_df = pd.read_csv(\n",
     "    path_to_score_file_1,\n",
@@ -68,11 +67,16 @@
    "source": [
     "# List rows in one but not the other\n",
     "\n",
-    "if len(score_2_df[ExtractTransformLoad.GEOID_TRACT_FIELD_NAME]) != len(score_1_df[ExtractTransformLoad.GEOID_TRACT_FIELD_NAME]):\n",
+    "if len(score_2_df[ExtractTransformLoad.GEOID_TRACT_FIELD_NAME]) != len(\n",
+    "    score_1_df[ExtractTransformLoad.GEOID_TRACT_FIELD_NAME]\n",
+    "):\n",
     "    print(\"Different lengths!\")\n",
     "\n",
     "print(\"Difference in tract IDs:\")\n",
-    "print(set(score_2_df[ExtractTransformLoad.GEOID_TRACT_FIELD_NAME]) ^ set(score_1_df[ExtractTransformLoad.GEOID_TRACT_FIELD_NAME]))\n"
+    "print(\n",
+    "    set(score_2_df[ExtractTransformLoad.GEOID_TRACT_FIELD_NAME])\n",
+    "    ^ set(score_1_df[ExtractTransformLoad.GEOID_TRACT_FIELD_NAME])\n",
+    ")"
    ]
   },
   {
@@ -83,7 +87,12 @@
    "outputs": [],
    "source": [
     "# Join\n",
-    "merged_df = score_1_df.merge(score_2_df, how=\"outer\", on=ExtractTransformLoad.GEOID_TRACT_FIELD_NAME, suffixes=('_1', '_2'))\n",
+    "merged_df = score_1_df.merge(\n",
+    "    score_2_df,\n",
+    "    how=\"outer\",\n",
+    "    on=ExtractTransformLoad.GEOID_TRACT_FIELD_NAME,\n",
+    "    suffixes=(\"_1\", \"_2\"),\n",
+    ")\n",
     "merged_df"
    ]
   },
@@ -95,13 +104,31 @@
    "outputs": [],
    "source": [
     "# Check each duplicate column:\n",
+    "# Remove the suffix \"_1\"\n",
     "duplicate_columns = [x[:-2] for x in merged_df.columns if \"_1\" in x]\n",
     "\n",
-    "for duplicate_column in duplicate_columns:\n",
+    "columns_to_exclude_from_duplicates_check = [\n",
-    "    print(f\"Checking duplicate column {duplicate_column}\")\n",
+    "    \"Total threshold criteria exceeded\"\n",
-    "    if not merged_df[f\"{duplicate_column}_1\"].equals(merged_df[f\"{duplicate_column}_2\"]):\n",
+    "]\n",
-    "        print(merged_df[f\"{duplicate_column}_1\"].compare(merged_df[f\"{duplicate_column}_2\"]))\n",
+    "\n",
-    "        raise ValueError(f\"Error! Different values in {duplicate_column}\")"
+    "columns_to_check = [column for column in duplicate_columns if column not in columns_to_exclude_from_duplicates_check]\n",
+    "\n",
+    "any_errors_found = False\n",
+    "for column_to_check in columns_to_check:\n",
+    "    print(f\"Checking duplicate column {column_to_check}\")\n",
+    "    if not merged_df[f\"{column_to_check}_1\"].equals(\n",
+    "        merged_df[f\"{column_to_check}_2\"]\n",
+    "    ):\n",
+    "        print(f\"Error! Different values in {column_to_check}\")\n",
+    "        print(\n",
+    "            merged_df[f\"{column_to_check}_1\"].compare(\n",
+    "                merged_df[f\"{column_to_check}_2\"]\n",
+    "            )\n",
+    "        )\n",
+    "        any_errors_found = True\n",
+    "\n",
+    "if any_errors_found:\n",
+    "    raise ValueError(f\"Error! Different values in one or more columns.\")"
    ]
   }
  ],

data/data-pipeline/data_pipeline/score/field_names.py

@@ -60,11 +60,15 @@ MEDIAN_INCOME_FIELD = "Median household income in the past 12 months"
 MEDIAN_INCOME_AS_PERCENT_OF_STATE_FIELD = (
     "Median household income (% of state median household income)"
 )
-MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD = "Median household income (% of AMI)"
 PERSISTENT_POVERTY_FIELD = "Persistent Poverty Census Tract"
 AMI_FIELD = "Area Median Income (State or metropolitan)"
-
 COLLEGE_ATTENDANCE_FIELD = "Percent enrollment in college or graduate school"
+MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD = (
+    "Median household income as a percent of area median income"
+)
+LOW_MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD = (
+    "Low median household income as a percent of area median income"
+)
 
 # Climate
 FEMA_RISK_FIELD = "FEMA Risk Index Expected Annual Loss Score"
@@ -105,7 +109,6 @@ ENERGY_BURDEN_FIELD = "Energy burden"
 DIABETES_FIELD = "Diagnosed diabetes among adults aged >=18 years"
 ASTHMA_FIELD = "Current asthma among adults aged >=18 years"
 HEART_DISEASE_FIELD = "Coronary heart disease among adults aged >=18 years"
-LIFE_EXPECTANCY_FIELD = "Life expectancy (years)"
 CANCER_FIELD = "Cancer (excluding skin cancer) among adults aged >=18 years"
 HEALTH_INSURANCE_FIELD = (
     "Current lack of health insurance among adults aged 18-64 years"
@@ -113,6 +116,8 @@ HEALTH_INSURANCE_FIELD = (
 PHYS_HEALTH_NOT_GOOD_FIELD = (
     "Physical health not good for >=14 days among adults aged >=18 years"
 )
+LIFE_EXPECTANCY_FIELD = "Life expectancy (years)"
+LOW_LIFE_EXPECTANCY_FIELD = "Low life expectancy"
 
 # Other Demographics
 TOTAL_POP_FIELD = "Total population"
@@ -130,9 +135,6 @@ OVER_64_FIELD = "Individuals over 64 years old"
 
 # Fields from 2010 decennial census (generally only loaded for the territories)
 CENSUS_DECENNIAL_MEDIAN_INCOME_2009 = "Median household income in 2009 ($)"
-CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009 = (
-    "Median household income as a percent of territory median income in 2009"
-)
 CENSUS_DECENNIAL_POVERTY_LESS_THAN_100_FPL_FIELD_2009 = (
     "Percentage households below 100% of federal poverty line in 2009"
 )
@@ -141,7 +143,10 @@ CENSUS_DECENNIAL_UNEMPLOYMENT_FIELD_2009 = (
     "Unemployed civilians (percent) in 2009"
 )
 CENSUS_DECENNIAL_TOTAL_POPULATION_FIELD_2009 = "Total population in 2009"
-
+CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009 = (
+    "Median household income as a percent of territory median income in 2009"
+)
+LOW_CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009 = "Low median household income as a percent of territory median income in 2009"
 # Fields from 2010 ACS (loaded for comparison with the territories)
 CENSUS_UNEMPLOYMENT_FIELD_2010 = "Unemployed civilians (percent) in 2010"
 CENSUS_POVERTY_LESS_THAN_100_FPL_FIELD_2010 = (
@@ -265,7 +270,10 @@ ASTHMA_LOW_INCOME_FIELD = (
 )
 HEART_DISEASE_LOW_INCOME_FIELD = f"At or above the {PERCENTILE}th percentile for heart disease and is low income"
 
-LIFE_EXPECTANCY_LOW_INCOME_FIELD = f"At or above the {PERCENTILE}th percentile for life expectancy and is low income"
+LOW_LIFE_EXPECTANCY_LOW_INCOME_FIELD = (
+    f"At or above the {PERCENTILE}th percentile "
+    f"for low life expectancy and is low income"
+)
 
 # Workforce
 UNEMPLOYMENT_LOW_HS_EDUCATION_FIELD = (
@@ -288,9 +296,9 @@ LOW_READING_LOW_HS_EDUCATION_FIELD = (
     " and has low HS education"
 )
 
-MEDIAN_INCOME_LOW_HS_EDUCATION_FIELD = (
+LOW_MEDIAN_INCOME_LOW_HS_EDUCATION_FIELD = (
-    f"At or below the {PERCENTILE}th percentile for median income"
+    f"At or below the {PERCENTILE}th percentile for low median household income as a "
-    "  and has low HS education"
+    f"percent of area median income and has low HS education"
 )
 
 # Not currently used in a factor

data/data-pipeline/data_pipeline/score/score_l.py

@@ -44,6 +44,8 @@ class ScoreL(Score):
         robustness over 1-year ACS.
         """
         # Create the combined field.
+        # TODO: move this combined field percentile calculation to `etl_score`,
+        #  since most other percentile logic is there.
         # There should only be one entry in either 2009 or 2019 fields, not one in both.
         # But just to be safe, we take the mean and ignore null values so if there
         # *were* entries in both, this result would make sense.
@@ -169,7 +171,7 @@ class ScoreL(Score):
     def _climate_factor(self) -> bool:
         # In Xth percentile or above for FEMA’s Risk Index (Source: FEMA
         # AND
-        # Low income: In 60th percentile or above for percent of block group population
+        # Low income: In Nth percentile or above for percent of block group population
         # of households where household income is less than or equal to twice the federal
         # poverty level. Source: Census's American Community Survey]
 
@@ -225,7 +227,7 @@ class ScoreL(Score):
     def _energy_factor(self) -> bool:
         # In Xth percentile or above for DOE’s energy cost burden score (Source: LEAD Score)
         # AND
-        # Low income: In 60th percentile or above for percent of block group population
+        # Low income: In Nth percentile or above for percent of block group population
         # of households where household income is less than or equal to twice the federal
         # poverty level. Source: Census's American Community Survey]
 
@@ -268,7 +270,7 @@ class ScoreL(Score):
         # or
         # In Xth percentile or above traffic proximity and volume (Source: 2017 U.S. Department of Transportation (DOT) traffic data
         # AND
-        # Low income: In 60th percentile or above for percent of block group population
+        # Low income: In Nth percentile or above for percent of block group population
         # of households where household income is less than or equal to twice the federal
         # poverty level. Source: Census's American Community Survey]
 
@@ -315,7 +317,7 @@ class ScoreL(Score):
         # or
         # In Xth percentile or above for housing cost burden (Source: HUD's Comprehensive Housing Affordability Strategy dataset
         # AND
-        # Low income: In 60th percentile or above for percent of block group population
+        # Low income: In Nth percentile or above for percent of block group population
         # of households where household income is less than or equal to twice the federal
         # poverty level. Source: Census's American Community Survey]
 
@@ -363,7 +365,7 @@ class ScoreL(Score):
     def _pollution_factor(self) -> bool:
         # Proximity to Risk Management Plan sites is > X
         # AND
-        # Low income: In 60th percentile or above for percent of block group population
+        # Low income: In Nth percentile or above for percent of block group population
         # of households where household income is less than or equal to twice the federal
         # poverty level. Source: Census's American Community Survey]
 
@@ -410,7 +412,7 @@ class ScoreL(Score):
     def _water_factor(self) -> bool:
         # In Xth percentile or above for wastewater discharge (Source: EPA Risk-Screening Environmental Indicators (RSEI) Model)
         # AND
-        # Low income: In 60th percentile or above for percent of block group population
+        # Low income: In Nth percentile or above for percent of block group population
         # of households where household income is less than or equal to twice the federal
         # poverty level. Source: Census's American Community Survey]
 
@@ -441,7 +443,7 @@ class ScoreL(Score):
         # or
         # In Xth percentile or above for low life expectancy (Source: CDC Places)
         # AND
-        # Low income: In 60th percentile or above for percent of block group population
+        # Low income: In Nth percentile or above for percent of block group population
         # of households where household income is less than or equal to twice the federal
         # poverty level. Source: Census's American Community Survey]
 
@@ -449,8 +451,7 @@ class ScoreL(Score):
             field_names.DIABETES_LOW_INCOME_FIELD,
             field_names.ASTHMA_LOW_INCOME_FIELD,
             field_names.HEART_DISEASE_LOW_INCOME_FIELD,
-            field_names.HEALTHY_FOOD_LOW_INCOME_FIELD,
+            field_names.LOW_LIFE_EXPECTANCY_LOW_INCOME_FIELD,
-            field_names.LIFE_EXPECTANCY_LOW_INCOME_FIELD,
         ]
 
         diabetes_threshold = (
@@ -475,24 +476,14 @@ class ScoreL(Score):
             >= self.ENVIRONMENTAL_BURDEN_THRESHOLD
         )
 
-        healthy_food_threshold = (
+        low_life_expectancy_threshold = (
             self.df[
-                field_names.HEALTHY_FOOD_FIELD
+                field_names.LOW_LIFE_EXPECTANCY_FIELD
                 + field_names.PERCENTILE_FIELD_SUFFIX
             ]
             >= self.ENVIRONMENTAL_BURDEN_THRESHOLD
         )
 
-        life_expectancy_threshold = (
-            self.df[
-                field_names.LIFE_EXPECTANCY_FIELD
-                + field_names.PERCENTILE_FIELD_SUFFIX
-            ]
-            # Note: a high life expectancy is good, so take 1 minus the threshold to invert it,
-            # and then look for life expenctancies lower than that (not greater than).
-            <= 1 - self.ENVIRONMENTAL_BURDEN_THRESHOLD
-        )
-
         self.df[field_names.DIABETES_LOW_INCOME_FIELD] = (
             diabetes_threshold & self.df[field_names.FPL_200_SERIES]
         )
@@ -502,11 +493,8 @@ class ScoreL(Score):
         self.df[field_names.HEART_DISEASE_LOW_INCOME_FIELD] = (
             heart_disease_threshold & self.df[field_names.FPL_200_SERIES]
         )
-        self.df[field_names.LIFE_EXPECTANCY_LOW_INCOME_FIELD] = (
+        self.df[field_names.LOW_LIFE_EXPECTANCY_LOW_INCOME_FIELD] = (
-            life_expectancy_threshold & self.df[field_names.FPL_200_SERIES]
+            low_life_expectancy_threshold & self.df[field_names.FPL_200_SERIES]
-        )
-        self.df[field_names.HEALTHY_FOOD_LOW_INCOME_FIELD] = (
-            healthy_food_threshold & self.df[field_names.FPL_200_SERIES]
         )
 
         self._increment_total_eligibility_exceeded(health_eligibility_columns)
@@ -514,23 +502,25 @@ class ScoreL(Score):
         return self.df[health_eligibility_columns].any(axis="columns")
 
     def _workforce_factor(self) -> bool:
-        # Where unemployment is above X%
+        # Where unemployment is above Xth percentile
         # or
-        # Where median income is less than Y% of the area median income
+        # Where median income as a percent of area median income is above Xth percentile
         # or
-        # Where the percent of households at or below 100% of the federal poverty level is greater than Z%
+        # Where the percent of households at or below 100% of the federal poverty level
+        # is above Xth percentile
         # or
-        # Where linguistic isolation is greater than Y%
+        # Where linguistic isolation is above Xth percentile
         # AND
-        # Where the high school degree achievement rates for adults 25 years and older is less than 95%
+        # Where the high school degree achievement rates for adults 25 years and older
-        # (necessary to screen out university block groups)
+        # is less than Y%
+        # (necessary to screen out university tracts)
 
         # Workforce criteria for states fields.
         workforce_eligibility_columns = [
             field_names.UNEMPLOYMENT_LOW_HS_EDUCATION_FIELD,
             field_names.POVERTY_LOW_HS_EDUCATION_FIELD,
             field_names.LINGUISTIC_ISOLATION_LOW_HS_EDUCATION_FIELD,
-            field_names.MEDIAN_INCOME_LOW_HS_EDUCATION_FIELD,
+            field_names.LOW_MEDIAN_INCOME_LOW_HS_EDUCATION_FIELD,
         ]
 
         high_scool_achievement_rate_threshold = (
@@ -546,14 +536,12 @@ class ScoreL(Score):
             >= self.ENVIRONMENTAL_BURDEN_THRESHOLD
         )
 
-        median_income_threshold = (
+        low_median_income_threshold = (
             self.df[
-                field_names.MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD
+                field_names.LOW_MEDIAN_INCOME_AS_PERCENT_OF_AMI_FIELD
                 + field_names.PERCENTILE_FIELD_SUFFIX
             ]
-            # Note: a high median income as a % of AMI is good, so take 1 minus the threshold to invert it.
+            >= self.ENVIRONMENTAL_BURDEN_THRESHOLD
-            # and then look for median income lower than that (not greater than).
-            <= 1 - self.ENVIRONMENTAL_BURDEN_THRESHOLD
         )
 
         linguistic_isolation_threshold = (
@@ -581,8 +569,8 @@ class ScoreL(Score):
             poverty_threshold & high_scool_achievement_rate_threshold
         )
 
-        self.df[field_names.MEDIAN_INCOME_LOW_HS_EDUCATION_FIELD] = (
+        self.df[field_names.LOW_MEDIAN_INCOME_LOW_HS_EDUCATION_FIELD] = (
-            median_income_threshold & high_scool_achievement_rate_threshold
+            low_median_income_threshold & high_scool_achievement_rate_threshold
         )
 
         self.df[field_names.UNEMPLOYMENT_LOW_HS_EDUCATION_FIELD] = (
@@ -624,23 +612,31 @@ class ScoreL(Score):
             threshold_cutoff_for_island_areas=self.ENVIRONMENTAL_BURDEN_THRESHOLD,
         )
 
+        # Also check whether low area median income is 90th percentile or higher
+        # within the islands.
+        low_median_income_as_a_percent_of_ami_island_areas_criteria_field_name = (
+            f"{field_names.LOW_CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009} exceeds "
+            f"{field_names.PERCENTILE}th percentile"
+        )
+        self.df[
+            low_median_income_as_a_percent_of_ami_island_areas_criteria_field_name
+        ] = (
+            self.df[
+                field_names.LOW_CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009
+                + field_names.PERCENTILE_FIELD_SUFFIX
+            ]
+            >= self.ENVIRONMENTAL_BURDEN_THRESHOLD
+        )
+
         workforce_combined_criteria_for_island_areas = (
             self.df[unemployment_island_areas_criteria_field_name]
             | self.df[poverty_island_areas_criteria_field_name]
-            # Also check whether area median income is 10th percentile or lower
+            | self.df[
-            # within the islands.
+                low_median_income_as_a_percent_of_ami_island_areas_criteria_field_name
-            | (
-                self.df[
-                    field_names.CENSUS_DECENNIAL_AREA_MEDIAN_INCOME_PERCENT_FIELD_2009
-                    + field_names.PERCENTILE_FIELD_SUFFIX
             ]
-                # Note: a high median income as a % of AMI is good, so take 1 minus the threshold to invert it.
-                # and then look for median income lower than that (not greater than).
-                < 1 - self.ENVIRONMENTAL_BURDEN_THRESHOLD
-            )
         ) & (
             self.df[field_names.CENSUS_DECENNIAL_HIGH_SCHOOL_ED_FIELD_2009]
-            > self.LACK_OF_HIGH_SCHOOL_MINIMUM_THRESHOLD
+            >= self.LACK_OF_HIGH_SCHOOL_MINIMUM_THRESHOLD
         )
 
         percent_of_island_tracts_highlighted = (