inferential_statistics_project3

July 9, 2017

1 Hospital Readmissions Data Analysis and Recommendations for Re-

duction
1.0.1 Background
In October 2012, the US governments Center for Medicare and Medicaid Services (CMS) began re-
ducing Medicare payments for Inpatient Prospective Payment System hospitals with excess read-
missions. Excess readmissions are measured by a ratio, by dividing a hospitals number of pre-
dicted 30-day readmissions for heart attack, heart failure, and pneumonia by the number that
would be expected, based on an average hospital with similar patients. A ratio greater than 1
indicates excess readmissions.

1.0.2 Exercise Directions

In this exercise, you will: + critique a preliminary analysis of readmissions data and recommenda-
tions (provided below) for reducing the readmissions rate + construct a statistically sound analysis
and make recommendations of your own
More instructions provided below. Include your work in this notebook and submit to your
Github account.

1.0.3 Resources
Data source: https://data.medicare.gov/Hospital-Compare/Hospital-Readmission-
Reduction/9n3s-kdb3
More information: http://www.cms.gov/Medicare/medicare-fee-for-service-
payment/acuteinpatientPPS/readmissions-reduction-program.html
Markdown syntax: http://nestacms.com/docs/creating-content/markdown-cheat-sheet
****

In [123]: %matplotlib inline

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import bokeh.plotting as bkp
from scipy import stats
from mpl_toolkits.axes_grid1 import make_axes_locatable
import statsmodels.stats.api as sms

1
In [5]: # read in readmissions data provided
hospital_read_df = pd.read_csv('data/cms_hospital_readmissions.csv')

In [12]: len(hospital_read_df["Hospital Name"])

Out[12]: 16860

In [6]: hospital_read_df.head()

Out[6]: Hospital Name Provider Number State \

0 FROEDTERT MEMORIAL LUTHERAN HOSPITAL 520177 WI
1 PROVIDENCE HOSPITAL 90006 DC
2 BEAUFORT COUNTY MEMORIAL HOSPITAL 420067 SC
3 ADVOCATE CHRIST HOSPITAL & MEDICAL CENTER 140208 IL
4 BRAZOSPORT REGIONAL HEALTH SYSTEM 450072 TX

Measure Name Number of Discharges Footnote \

0 READM-30-HIP-KNEE-HRRP 242 NaN
1 READM-30-HIP-KNEE-HRRP 247 NaN
2 READM-30-HIP-KNEE-HRRP 586 NaN
3 READM-30-HIP-KNEE-HRRP 965 NaN
4 READM-30-HIP-KNEE-HRRP 149 NaN

Excess Readmission Ratio Predicted Readmission Rate \

0 1.9095 10.8
1 1.7521 9.2
2 1.5836 7.6
3 1.5760 9.0
4 1.5308 8.2

Expected Readmission Rate Number of Readmissions Start Date End Dat

0 5.6 38.0 07/01/2010 06/30/201
1 5.3 33.0 07/01/2010 06/30/201
2 4.8 53.0 07/01/2010 06/30/201
3 5.7 95.0 07/01/2010 06/30/201
4 5.4 20.0 07/01/2010 06/30/201

1.1 Preliminary Analysis

In [36]: # deal with missing and inconvenient portions of data
clean_hospital_read_df = hospital_read_df[hospital_read_df['Number of Disc
clean_hospital_read_df.loc[:, 'Number of Discharges'] = clean_hospital_rea
clean_hospital_read_df = clean_hospital_read_df.sort_values('Number of Dis

/Users/howellyu/anaconda/lib/python3.5/site-packages/pandas/core/indexing.py:465: S
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

2
See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/i
self.obj[item] = s

In [37]: # generate a scatterplot for number of discharges vs. excess rate of readm
# lists work better with matplotlib scatterplot function
x = [a for a in clean_hospital_read_df['Number of Discharges'][81:-3]]
y = list(clean_hospital_read_df['Excess Readmission Ratio'][81:-3])

fig, ax = plt.subplots(figsize=(8,5))
ax.scatter(x, y,alpha=0.2)

ax.fill_between([0,350], 1.15, 2, facecolor='red', alpha = .15, interpolat

ax.fill_between([800,2500], .5, .95, facecolor='green', alpha = .15, inter

ax.set_xlim([0, max(x)])
ax.set_xlabel('Number of discharges', fontsize=12)
ax.set_ylabel('Excess rate of readmissions', fontsize=12)
ax.set_title('Scatterplot of number of discharges vs. excess rate of readm

ax.grid(True)
fig.tight_layout()

3
1.2 Preliminary Report
Read the following results/report. While you are reading it, think about if the conclusions are cor-
rect, incorrect, misleading or unfounded. Think about what you would change or what additional
analyses you would perform.
A. Initial observations based on the plot above + Overall, rate of readmissions is trending
down with increasing number of discharges + With lower number of discharges, there is a greater
incidence of excess rate of readmissions (area shaded red) + With higher number of discharges,
there is a greater incidence of lower rates of readmissions (area shaded green)
B. Statistics + In hospitals/facilities with number of discharges < 100, mean excess readmis-
sion rate is 1.023 and 63% have excess readmission rate greater than 1 + In hospitals/facilities
with number of discharges > 1000, mean excess readmission rate is 0.978 and 44% have excess
readmission rate greater than 1
C. Conclusions + There is a significant correlation between hospital capacity (number of dis-
charges) and readmission rates. + Smaller hospitals/facilities may be lacking necessary resources
to ensure quality care and prevent complications that lead to readmissions.
D. Regulatory policy recommendations + Hospitals/facilties with small capacity (< 300)
should be required to demonstrate upgraded resource allocation for quality care to continue oper-
ation. + Directives and incentives should be provided for consolidation of hospitals and facilities
to have a smaller number of them with higher capacity and number of discharges.

1.2.1 Exercise
Include your work on the following in this notebook and submit to your Github account.
A. Do you agree with the above analysis and recommendations? Why or why not?
B. Provide support for your arguments and your own recommendations with a statistically
sound analysis:

1. Setup an appropriate hypothesis test.

2. Compute and report the observed significance value (or p-value).
3. Report statistical significance for = .01.
4. Discuss statistical significance and practical significance. Do they differ here? How does this
change your recommendation to the client?
5. Look at the scatterplot above.
What are the advantages and disadvantages of using this plot to convey information?
Construct another plot that conveys the same information in a more direct manner.

You can compose in notebook cells using Markdown: + In the control panel at the top,
choose Cell > Cell Type > Markdown + Markdown syntax: http://nestacms.com/docs/creating-
content/markdown-cheat-sheet

1.2.2 A.
In A. Initial Observation, we does not know whether the readmission rate is trending down as the
hospital capasity decreases since there is no calculation support, and visually we can hardly tell
whether readmission rate goes down.

4
 In B. We cannot tell whether the mean of readmission rate is different for hospital with capacity
> 1000 and < 100. It is necessary to run a 2 sample t test before we make any conclusion and
therefore the conclusion made in C cannot be presumed correct.

1.2.3 B-1
H0 : There is NO difference of the means excess rate between hospital with capacity > 1000 and <
100.

In [111]: data_100 = clean_hospital_read_df[(clean_hospital_read_df['Number of Disc

(clean_hospital_read_df['Number of D
data_1000 = clean_hospital_read_df[clean_hospital_read_df['Number of Disc
mean_data_100 = data_100['Excess Readmission Ratio'].mean()
mean_data_1000 = data_1000['Excess Readmission Ratio'].mean()
var_data_100 = data_100['Excess Readmission Ratio'].var()
var_data_1000 = data_1000['Excess Readmission Ratio'].var()

print("The mean readmission rate of hospital with capacity < 100 is %.3f,
% round(mean_data_100, 3), "variance is %.5f,"% round(var_data_100,
"%.2f have excess readmission rate > 1." %
round(len(data_100[data_100["Excess Readmission Ratio"]>1])/len(data
print("The mean readmission rate of hospital with capacity > 1000 is %.3f
% round(mean_data_1000, 3), "variance is %.5f," % round(var_data_10
"%.2f have excess readmission rate > 1." %
round(len(data_1000[data_1000["Excess Readmission Ratio"]>1])/len(da

The mean readmission rate of hospital with capacity < 100 is 1.023, variance is 0.0
The mean readmission rate of hospital with capacity > 1000 is 0.978, variance is 0.
Their difference in mean is 0.044.

The calculation above is consistent with the number provided in Preliminary Report.

1.2.4 B - 2 run 2 sample t test to compare the means

In [61]: pd.DataFrame.hist(clean_hospital_read_df, column='Excess Readmission Ratio

Out[61]: array([[<matplotlib.axes._subplots.AxesSubplot object at 0x119759f98>]], d

5
In [83]: pd.DataFrame.hist(data_100, column='Excess Readmission Ratio') # normal

Out[83]: array([[<matplotlib.axes._subplots.AxesSubplot object at 0x119962898>]], d

6
In [84]: pd.DataFrame.hist(data_1000, column='Excess Readmission Ratio') # normal,

Out[84]: array([[<matplotlib.axes._subplots.AxesSubplot object at 0x119981940>]], d

In [146]: diff = mean_data_100 - mean_data_1000

t, p_value = stats.ttest_ind(data_100["Excess Readmission Ratio"], data_1
print("The t score is %.3f" % round(t, 3), "and the p-value is almost %.1

cm = sms.CompareMeans(sms.DescrStatsW(data_100["Excess Readmission Ratio"

sms.DescrStatsW(data_1000["Excess Readmission Ratio
CI = cm.tconfint_diff(alpha = 0.01, usevar='unequal')
CI

The t score is 10.061 and the p-value is almost 0.0

Out[146]: (0.029225356358290495, 0.059340501645796843)

1.2.5 B-3.
Since p-value < 0.01, we reject the null hypothesis and conclude that there is a difference between
2 samples.

7
1.2.6 B-4. Statistical Significant Vs. Practical Significance
Althrough the difference between 2 samples has been proved to be statistically significant, which
means the will be correlation between the 2 variables, if we take a look at the confidence, we are
99% sure that the difference between the means of 2 samples will be caught between 0.029 and
0.059. Therefore, the difference is not enough to catch a true correlation in real life.
1. The statement There is a significant correlation between hospital capacity (number of
discharges) and readmission rates. is misleading since it does not consider the practical signifi-
cance.
2. The statement Smaller hospitals/facilities may be lacking necessary resources to ensure
quality care and prevent complications that lead to readmissions. may not be true since lower
discharge number does not necessarily imply that the hospital is small, even though in many cases
we might find this true.

1.2.7 B-5. Scatter Plot

The scatter plot provides us with direct visually aid about the data. Thus, we can have a better
understanding comaring to simply look and the data frame and it is easier for us to make real-life
judgement instead of getting misled by small statistical significant difference (just like in this case).
However, in many other cases, a scatter plot fails to provide us a very deep understanding of the
data, and without support of numbers it is very likely to make bad judgement calls.

In [172]: bins = np.linspace(clean_hospital_read_df['Number of Discharges'].min(),

clean_hospital_read_df['Number of Discharges'].max(),
groups = clean_hospital_read_df.groupby(np.digitize(clean_hospital_read_d

#digitize() Return the indices of the bins to which each value in input a
plt.plot(groups.aggregate(lambda x: np.mean(x[x > 0.5]))['Number of Disch
list(groups.aggregate(lambda x: np.mean(x[x > 0.5]))['Excess Rea
color='b', lw=2, marker='s')
plt.xlabel('Number of discharges', fontsize=12)
plt.ylabel('Mean excess rate \nof readmissions', fontsize=12)
plt.title('Comparing binned discharges to \nmean excess rate of readmissi
plt.ylim(0.65,1.15)
plt.xlim(0, 7000)
plt.show()

8
9