Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
Download

Jupyter notebook 2-Digging into participation, utilization, cost, and quality.ipynb

51 views
Tweet
Share
Share
Kernel: Python 2 (SageMath)

Notebook #2: Digging into participation, utilization, cost, and quality

This notebook...

  • looks at which CMS programs the cardiologist participates in

  • brings in data from CMS Aggregate Utilization and Cost data for CY2014

  • starts to compare the 8 given cardiologists to the population of 5,480 CA cardiologists we have Medicare data on

import pandas as pd
import numpy as np

pd.set_option('display.max_columns', None)

# Let's load the cardiologist file again
given_cardiologists = pd.read_csv("data/given_cardiologists_plus_national_downloadable_file.csv")

cms_program_cols = ['Participating in eRx',
                    'Participating in PQRS',
                    'Participating in EHR',
                    'Received PQRS Maintenance of Certification Program Incentive',
                    'Participated in Million Hearts']

given_cardiologists.replace(np.nan, "NULL", inplace=True)
given_cardiologists.groupby(cms_program_cols).size()
Participating in eRx Participating in PQRS Participating in EHR Received PQRS Maintenance of Certification Program Incentive Participated in Million Hearts NULL NULL NULL NULL NULL 2 Y Y NULL NULL 6 dtype: int64

Analysis of CMS program participation

  • We can see that none of the cardiologists are participating in eRX, Million Hearts, or have received the PQRS incentive.

  • 6 of the 8 physicians are participating in both PQRS and EHR incentive program -- we may want to prioritize these six

# Let's now look at the aggregate provider utilizationd data from CMS (source #3 in README.md)
all_provider_agg_util = pd.read_csv("data/data.cms.gov_4a3h-46r6_CY2015-Aggregate-Utilization.csv", low_memory=False)

# and join in the all_provider_agg_util data with given_cardiologist data into a new dataframe
given_cardiologists_plus_agg_util = pd.merge(given_cardiologists, all_provider_agg_util, how='left', left_on='npi', right_on='National Provider Identifier', suffixes=('_ndf','_agg'))
# write this merge to the file system for safe keeping
given_cardiologists_plus_agg_util.to_csv("data/given_cardiologists_plus_agg_util.csv")


# we may want to look at all cardiologists' aggregate information so let's bring in the original source #1 dataset and join it with this one as well
all_providers = pd.read_csv("data/data.medicare.gov_s63f-csi6_National-Downloadable_File.csv", low_memory=False)
all_cardiologists = all_providers[all_providers['Primary specialty'] == 'CARDIOVASCULAR DISEASE (CARDIOLOGY)']
all_cardiologists_in_ca = all_cardiologists[all_cardiologists['State'] == 'CA']
all_cardiologists_in_ca_plus_agg_util = pd.merge(all_cardiologists_in_ca, all_provider_agg_util, how='left', left_on='npi', right_on='National Provider Identifier', suffixes=('_ndf','_agg'))

# So what data do we have to work with now?
given_cardiologists_plus_agg_util

Analysis of given cardiologists and all CA cardiologists

# How many CA cardiologists can we compare these 8 with?
print "There are {0} cardiologists in CA that we will compare these {1} with".format(len(all_cardiologists_in_ca_plus_agg_util), len(given_cardiologists_plus_agg_util))
There are 5480 cardiologists in CA that we will compare these 8 with 
# We will use plotly for interactive charting and visual analytics
import plotly.plotly as py
import plotly.graph_objs as go
from plotly import tools

# helper method for creating hover over tooltips/labels
def list_of_provider_labels(df):
    labels = []
    for n, provider in df.iterrows():
        labels.append("NPI: {0}<br />Name: {1}, {2}".format(provider['npi'], provider['Last Name'], provider['First Name']))
    return labels

# given one df, this produces a two-column layout: [boxplot | scatterplot]
## Example usage:
## fig = plotly_box_and_scatter_plot(given_cardiologists_plus_agg_util, 'Graduation year')
## py.iplot(fig)
def plotly_box_and_scatter_plot(df, column_to_graph):
    data = [
        go.Box(
            name = column_to_graph,
            y = df[column_to_graph],
            boxmean = True
        ),
        go.Scatter(
            x = np.full(len(df), 0),
            y = df[column_to_graph].tolist(),
            mode = 'markers',
            text = list_of_provider_labels(df) #df['npi'].tolist(),
        )
    ]
    layout = go.Layout(
        showlegend=False,
        hovermode='closest',
        title="{0}".format(column_to_graph)
    )
    fig = go.Figure(data=data, layout=layout)
    return fig

# given two dfs, this produces a two-column layout: [given boxplot | given scatterplot | all boxplot | all scatterplot]
## Example usage:
## fig = plotly_compare_distribution_of_column(
##     column_to_graph = 'Percent (%) of Beneficiaries Identified With Ischemic Heart Disease ',
##     given_name = 'Nearby Cardios', given_df = given_cardiologists_plus_agg_util,
##     all_name = 'All CA Cardios', all_df = all_cardiologists_in_ca_plus_agg_util
##     y_min = 0, y_min = 100 # these are optional!
##     )
## py.iplot(fig)
def plotly_compare_distribution_of_column(column_to_graph, given_df, given_name, all_df, all_name, y_min=None, y_max=None):
    given_box = go.Box(
        name = given_name,
        y = given_df[column_to_graph],
        boxmean = True
    )
    given_scatter = go.Scatter(
        name = given_name,
        x = np.full(len(given_df), 0),
        y = given_df[column_to_graph].tolist(),
        mode = 'markers',
        text = list_of_provider_labels(given_df)
    )

    all_box = go.Box(
        name = all_name,
        y = all_df[column_to_graph],
        boxmean = True
    )
    all_scatter = go.Scatter(
        name = all_name,
        x = np.full(len(all_df), 0),
        y = all_df[column_to_graph].tolist(),
        mode = 'markers',
        text = list_of_provider_labels(all_df)
    )

    fig = tools.make_subplots(rows=1, cols=4)

    fig.append_trace(given_box, 1, 1)
    fig.append_trace(given_scatter, 1, 2)
    fig.append_trace(all_box, 1, 3)
    fig.append_trace(all_scatter, 1, 4)

    fig['layout'].update(title=column_to_graph, showlegend=True, hovermode='closest')
    fig['layout']['xaxis2'].update(title=given_name, type='category',  ticks='', showticklabels=False)
    fig['layout']['xaxis4'].update(title=all_name, type='category', ticks='', showticklabels=False)

    # if y_min and y_max are non-None, set all axes to the specified min and max
    if y_min != None and y_max != None:
        y_axis_layout = dict(range = [y_min, y_max], autorange = False)
        fig['layout']['yaxis1'].update(y_axis_layout)
        fig['layout']['yaxis2'].update(y_axis_layout)
        fig['layout']['yaxis3'].update(y_axis_layout)
        fig['layout']['yaxis4'].update(y_axis_layout)
    
    return fig

fig = plotly_box_and_scatter_plot(given_cardiologists_plus_agg_util, 'Percent (%) of Beneficiaries Identified With Heart Failure ')
py.iplot(fig)

fig = plotly_compare_distribution_of_column(
    column_to_graph = 'Percent (%) of Beneficiaries Identified With Heart Failure ',
    given_name = 'Nearby Cardios', given_df = given_cardiologists_plus_agg_util,
    all_name = 'All CA Cardios', all_df = all_cardiologists_in_ca_plus_agg_util)
py.iplot(fig)
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ] 
fig1 = plotly_compare_distribution_of_column(
    column_to_graph = 'Percent (%) of Beneficiaries Identified With Heart Failure ',
    given_name = 'Nearby Cardios', given_df = given_cardiologists_plus_agg_util,
    all_name = 'All CA Cardios', all_df = all_cardiologists_in_ca_plus_agg_util,
    y_min = 0, y_max = 100)

fig2 = plotly_compare_distribution_of_column(
    column_to_graph = 'Percent (%) of Beneficiaries Identified With Ischemic Heart Disease ',
    given_name = 'Nearby Cardios', given_df = given_cardiologists_plus_agg_util,
    all_name = 'All CA Cardios', all_df = all_cardiologists_in_ca_plus_agg_util,
    y_min = 0, y_max = 100)

display(py.iplot(fig1))
display(py.iplot(fig2))
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ] This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ] 
from IPython.display import display

# let's chart a bunch more interesting columns
columns_to_chart = ['Number of HCPCS', 
                    'Number of Services',
                    'Number of Medicare Beneficiaries',
                    'Number of Medical Services',
                    'Average Age of Beneficiaries',
                    'Average HCC Risk Score of Beneficiaries ',
                    'Number of Beneficiaries With Medicare & Medicaid Entitlement']

for column in columns_to_chart:
    fig = plotly_compare_distribution_of_column(
        column_to_graph = column,
        given_name = 'Nearby Cardios', given_df = given_cardiologists_plus_agg_util,
        all_name = 'All CA Cardios', all_df = all_cardiologists_in_ca_plus_agg_util)
    display(py.iplot(fig))
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ]
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ]
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ]
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ]
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ]
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ]
This is the format of your plot grid: [ (1,1) x1,y1 ] [ (1,2) x2,y2 ] [ (1,3) x3,y3 ] [ (1,4) x4,y4 ]