Merge pull request #4 from jcunninghame/jcunni/sprucing-up

Jcunni/sprucing up

main_page.py

@@ -4,89 +4,107 @@
 import snowflake.connector as sn
 from dotenv import load_dotenv
 import os
+import util
 
 load_dotenv()
 
-st.markdown("# High Level Summary")
-st.sidebar.markdown("# Claims Summary")
-
 # Connect and fetch data
-con = sn.connect(
- user=os.getenv('SNOWFLAKE_USER'),
- password=os.getenv('SNOWFLAKE_PASSWORD'),
- account=os.getenv('SNOWFLAKE_ACCOUNT'),
- warehouse=os.getenv('SNOWFLAKE_WH'),
- role=os.getenv('SNOWFLAKE_ROLE')
-)
-cs = con.cursor()
-cs.execute("""SELECT PT.*, PB.MEMBER_COUNT FROM TUVA_PROJECT_DEMO.PMPM.PMPM_TRENDS PT
+conn = util.connection()
+
+@st.cache_data
+def summary_data():
+ query = """SELECT PT.*, PB.MEMBER_COUNT, PHARMACY_SPEND FROM TUVA_PROJECT_DEMO.PMPM.PMPM_TRENDS PT
  LEFT JOIN (SELECT CONCAT(LEFT(YEAR_MONTH, 4), '-', RIGHT(YEAR_MONTH, 2)) AS YEAR_MONTH, 
- COUNT(*) AS MEMBER_COUNT
+ COUNT(*) AS MEMBER_COUNT,
+ SUM(PHARMACY_PAID) AS PHARMACY_SPEND
  FROM TUVA_PROJECT_DEMO.PMPM.PMPM_BUILDER
  GROUP BY YEAR_MONTH) AS PB
- ON PT.YEAR_MONTH = PB.YEAR_MONTH;""")
+ ON PT.YEAR_MONTH = PB.YEAR_MONTH;"""
+
+ data = util.safe_to_pandas(conn, query)
+ data['year_month'] = pd.to_datetime(data['year_month'], format='%Y-%m').dt.date
+ return data
+
+@st.cache_data
+def gender_data():
+ query = """SELECT GENDER, COUNT(*) AS COUNT FROM TUVA_PROJECT_DEMO.CORE.PATIENT GROUP BY 1;"""
+ data = util.safe_to_pandas(conn, query)
+ return data
+
+@st.cache_data
+def race_data():
+ query = """SELECT RACE, COUNT(*) AS COUNT FROM TUVA_PROJECT_DEMO.CORE.PATIENT GROUP BY 1;"""
+ data = util.safe_to_pandas(conn, query)
+ return data
 
-data = cs.fetch_pandas_all()
-data['YEAR_MONTH'] = pd.to_datetime(data['YEAR_MONTH'], format='%Y-%m').dt.date
+@st.cache_data
+def age_data():
+ query = """SELECT CASE 
+ WHEN div0(current_date() - BIRTH_DATE, 365) < 49 THEN '34-48'
+ WHEN div0(current_date() - BIRTH_DATE, 365) >= 49 AND div0(current_date() - BIRTH_DATE, 365) < 65 THEN '49-64'
+ WHEN div0(current_date() - BIRTH_DATE, 365) >= 65 AND div0(current_date() - BIRTH_DATE, 365) < 79 THEN '65-78'
+ WHEN div0(current_date() - BIRTH_DATE, 365) >= 79 AND div0(current_date() - BIRTH_DATE, 365) < 99 THEN '79-98'
+ WHEN div0(current_date() - BIRTH_DATE, 365) >= 99 THEN '99+' END
+ AS AGE_GROUP,
+ COUNT(*) AS COUNT
+ FROM TUVA_PROJECT_DEMO.CORE.PATIENT
+ GROUP BY 1
+ ORDER BY 1;"""
+ data = util.safe_to_pandas(conn, query)
+ return data
+data = summary_data()
+demo_gender = gender_data()
+demo_race = race_data()
+demo_age = age_data()
 
-st.markdown("### Summary of Claims")
+st.markdown("# Summary of Claims")
 start_date, end_date = st.select_slider("Select date range for claims summary",
- options=data['YEAR_MONTH'].sort_values(),
- value=(data['YEAR_MONTH'].min(), data['YEAR_MONTH'].max()))
-filtered_data = data.loc[(data['YEAR_MONTH'] >= start_date) & (data['YEAR_MONTH'] <= end_date), :]
+ options=data['year_month'].sort_values(),
+ value=(data['year_month'].min(), data['year_month'].max()))
+filtered_data = data.loc[(data['year_month'] >= start_date) & (data['year_month'] <= end_date), :]
+
+st.markdown("### High Level Summary")
+st.markdown("""At a glance, see the total medical spend and PMPM for the chosen time period. As well as a trend
+graph for other important financial metrics""")
+st.sidebar.markdown("# Claims Summary")
 
 # Summary Metrics
-total_spend = filtered_data['MEDICAL_SPEND'].sum()
-total_member_months = filtered_data['MEMBER_MONTH_COUNT'].sum()
+total_spend = filtered_data['medical_spend'].sum()
+total_member_months = filtered_data['member_month_count'].sum()
 avg_pmpm = total_spend/total_member_months
+total_pharm_spend = filtered_data['pharmacy_spend'].sum()
 
-col1, col2, col3 = st.columns([1.5,1,1])
-col1.metric("Total Medical Spend", '${:,.2f}'.format(total_spend))
-col2.metric("Total Member Months", total_member_months)
-col3.metric("Average PMPM", '${:,.2f}'.format(avg_pmpm))
+col1, col2, col3, col4 = st.columns([1,1,1,1])
+col1.metric("Total Medical Spend", '${}'.format(util.human_format(total_spend)))
+col2.metric("Total Member Months", util.human_format(total_member_months))
+col3.metric("Average PMPM", '${}'.format(util.human_format(avg_pmpm)))
+col4.metric('Total Pharmacy Spend', '${}'.format(util.human_format(total_pharm_spend)))
 
 st.divider()
-y_axis = st.selectbox('Select Metric for Trend Line', [x for x in data.columns if x != 'YEAR_MONTH'])
+y_axis = st.selectbox('Select Metric for Trend Line', [x for x in data.columns if x != 'year_month'])
 
 if y_axis:
- st.line_chart(filtered_data, x='YEAR_MONTH', y=y_axis)
+ st.line_chart(filtered_data, x='year_month', y=y_axis)
 
 # Patient Demographic Section
 st.divider()
-st.subheader('Patient Demographics')
-st.write('Patient data static, not filtered by claims date sliders currently')
-
-cs.execute("""SELECT GENDER, COUNT(*) AS COUNT FROM TUVA_PROJECT_DEMO.CORE.PATIENT GROUP BY 1;""")
-demo_gender = cs.fetch_pandas_all()
-cs.execute("""SELECT RACE, COUNT(*) AS COUNT FROM TUVA_PROJECT_DEMO.CORE.PATIENT GROUP BY 1;""")
-demo_race = cs.fetch_pandas_all()
-cs.execute("""SELECT CASE 
- WHEN div0(current_date() - BIRTH_DATE, 365) < 49 THEN '34-48'
- WHEN div0(current_date() - BIRTH_DATE, 365) >= 49 AND div0(current_date() - BIRTH_DATE, 365) < 65 THEN '49-64'
- WHEN div0(current_date() - BIRTH_DATE, 365) >= 65 AND div0(current_date() - BIRTH_DATE, 365) < 79 THEN '65-78'
- WHEN div0(current_date() - BIRTH_DATE, 365) >= 79 AND div0(current_date() - BIRTH_DATE, 365) < 99 THEN '79-98'
- WHEN div0(current_date() - BIRTH_DATE, 365) >= 99 THEN '99+' END
- AS AGE_GROUP,
- COUNT(*) AS COUNT
- FROM TUVA_PROJECT_DEMO.CORE.PATIENT
- GROUP BY 1
- ORDER BY 1;""")
-demo_age = cs.fetch_pandas_all()
+st.markdown('### Patient Demographics')
+st.markdown("""The patient population during this claims period was mostly `female`, `white` and largely 
+over the age of 65, with nearly half of patients falling into the `65-78` age group""")
+st.write(' Please note that patient data is static, and not filtered by claims date sliders currently')
 
 demo_col1, demo_col2 = st.columns([1, 2])
 with demo_col1:
- plost.donut_chart(demo_gender, theta='COUNT',
- color=dict(field='GENDER', scale=dict(range=['#F8B7CD', '#67A3D9'])), legend='left',
+ plost.donut_chart(demo_gender, theta='count',
+ color=dict(field='gender', scale=dict(range=['#F8B7CD', '#67A3D9'])), legend='left',
  title='Gender Breakdown')
 with demo_col2:
  plost.bar_chart(
- demo_age, bar='AGE_GROUP', value='COUNT', legend=None, use_container_width=True,
+ demo_age, bar='age_group', value='count', legend=None, use_container_width=True,
  title='Counts by Age Group'
  )
 
 plost.bar_chart(
- demo_race, bar='RACE', value='COUNT', color='RACE', legend='bottom', use_container_width=True,
- title='Counts by Race'
+ demo_race, bar='race', value='count', color='race', legend='bottom', use_container_width=True,
+ title='Counts by Race', height=400
 )
-
-con.close()

pages/drilldown.py

@@ -0,0 +1,82 @@
+import streamlit as st
+import plost
+import pandas as pd
+import snowflake.connector as sn
+from dotenv import load_dotenv
+import os
+import util
+
+conn = util.connection()
+
+@st.cache_data
+def pmpm_data():
+ query = """SELECT PT.*, PB.MEMBER_COUNT, PB.PHARMACY_SPEND FROM TUVA_PROJECT_DEMO.PMPM.PMPM_TRENDS PT
+ LEFT JOIN (SELECT CONCAT(LEFT(YEAR_MONTH, 4), '-', RIGHT(YEAR_MONTH, 2)) AS YEAR_MONTH, 
+ COUNT(*) AS MEMBER_COUNT,
+ SUM(PHARMACY_PAID) AS PHARMACY_SPEND 
+ FROM TUVA_PROJECT_DEMO.PMPM.PMPM_BUILDER
+ GROUP BY YEAR_MONTH) AS PB
+ ON PT.YEAR_MONTH = PB.YEAR_MONTH;"""
+ data = util.safe_to_pandas(conn, query)
+ data['year_month'] = pd.to_datetime(data['year_month'], format='%Y-%m').dt.date
+ data['pharmacy_spend'] = data['pharmacy_spend'].astype(float)
+ return data
+
+@st.cache_data
+def condition_data():
+ query = """SELECT
+ CONCAT(date_part(year, FIRST_DIAGNOSIS_DATE), '-', lpad(date_part(month, FIRST_DIAGNOSIS_DATE), 2, 0)) AS DIAGNOSIS_YEAR_MONTH,
+ CONDITION,
+ COUNT(*) AS CONDITION_CASES,
+ AVG(LAST_DIAGNOSIS_DATE + 1 - FIRST_DIAGNOSIS_DATE) AS DIAGNOSIS_DURATION
+ FROM TUVA_PROJECT_DEMO.CHRONIC_CONDITIONS.TUVA_CHRONIC_CONDITIONS_LONG
+ GROUP BY 1,2
+ ORDER BY 3 DESC;"""
+ data = util.safe_to_pandas(conn, query)
+ return data
+
+pmpm_data = pmpm_data()
+cond_data = condition_data()
+
+st.markdown("# A Further Look")
+st.markdown( """The page below offers insight into several key metrics from the data, including PMPM, 
+pharmacy spend and chronic conditions""")
+st.sidebar.markdown("# Drilldown")
+
+st.markdown("### PMPM Breakdown and Pharmacy Spend Trends")
+start_date, end_date = st.select_slider("Select date range for claims summary",
+ options=pmpm_data['year_month'].sort_values(),
+ value=(pmpm_data['year_month'].min(), pmpm_data['year_month'].max()))
+
+filtered_pmpm_data = pmpm_data.loc[(pmpm_data['year_month'] >= start_date) & (pmpm_data['year_month'] <= end_date), :]
+filtered_pmpm_data['Metric'] = 'Average PMPM'
+
+pmpm_cats = ['inpatient_pmpm', 'outpatient_pmpm', 'office_visit_pmpm', 'ancillary_pmpm', 'other_pmpm']
+grouped_pmpm = filtered_pmpm_data.groupby(by='Metric', as_index=False)[pmpm_cats].mean()
+
+st.divider()
+st.markdown("""Inpatient and Outpatient spend are the largest drivers of PMPM during this time period""")
+plost.bar_chart(data=grouped_pmpm, bar='Metric', value=pmpm_cats, stack='normalize',
+ direction='horizontal', legend='top', height=200)
+st.markdown("### Total Pharmacy Spend Over Claim Period")
+st.markdown("""Pharmacy Spend appears largely steady between 2016 and 2018, averaging between `$100k-$200k` a month.
+however we do see larger spikes in April 2017 and February 2018""")
+st.line_chart(data=filtered_pmpm_data, x='year_month', y='pharmacy_spend')
+
+st.divider()
+
+st.markdown("### Top 5 Condition Diagnoses Over Claim Period")
+st.markdown("""The chart below shows trends in new cases of the top five chronic conditions during the 
+claims period selected.""")
+msk = (cond_data['diagnosis_year_month'] >= str(start_date)) & (cond_data['diagnosis_year_month'] <= str(end_date))
+filtered_cond_data = cond_data.loc[msk, :]
+top5_conditions = filtered_cond_data.groupby('condition')['condition_cases'].sum().nlargest(5)
+msk = filtered_cond_data['condition'].isin(top5_conditions.index)
+top5_filtered_cond = filtered_cond_data.loc[msk, :]
+
+plost.line_chart(data=top5_filtered_cond,
+ x='diagnosis_year_month',
+ y='condition_cases',
+ color='condition',
+ pan_zoom=None,
+ height=400)