HIT Midterm Paper 1

stemswedishΤεχνίτη Νοημοσύνη και Ρομποτική

15 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

99 εμφανίσεις

HIT Midterm Paper
1













Georgetown University



Sentiment

Analysis Paper



HESY 566



The Graduate School of Arts and Sciences

Health Systems Administration Program


By


Charlotte L. Kreger

07
-
05
-
201
0


HIT Midterm Paper
2


Introduction:


P
atient satisfactions scores have shown as
opportunity for growth and
improvement both in an acute care setting and in a
n

outpatient

setting. Continued
advancements
communication technology while mostly positive have

sometimes

had
unforeseen
and sometimes unsettling
outcomes.

The sometimes overly vociferous
rating
of clinicians

on such s
ites such as Yelp,
Angie’s List,

and physicianratings.com

(to name but a few)
have opened new realms

where
uncensored and unsolicited
comments can be posted, reviewed, and perhaps commented on
a
t

nauseum by

passing

cyber reader
s
.

This rise in social networking has fueled the market
ing

of
personal opinion.

Though the cost of conducting individual patient surveys is of concern,

many researchers and analys
ts have crossed over to focus on

patient com
plaints
1,2
.
This documented information can be valuable in analyzing and comparing clinicians
throughout different regions

for the benefit of impro
ving patient experiences and
satisfaction
.

This can be
better
accomplished by applying sentiment analysis.


Background:



Sentiment analysis

can broadly be defined as the exploration and analysis of
data to determine a speaker or writer’s attitude or feeling
s

in order to

discover
meaningful patterns
2
.
Sentiment analysis
, also referred

to as
opinion mining or
data
mining, was first developed in the field of statistics
. The mining process was initially

associated with a negative connotation and analysts used it blindly

to

s
earch within data
without

a particular

purpose
3
.

L
ater

the term was

adopted for commercial

applications of
artificial intelligence to develop algorithms t
hat could be
applied to machine learning
3
.
The earliest reviews of

the mining process

from the 1990s indicate it was

a
HIT Midterm Paper
3


disappointment in the corporate world because the technology

at that time

w
as
immature
4
.

The current

use of the

technology is being used for analyzing, modeling, and
predicating across a variety of fields

to give a competitive edge

to various organizational
models
. These fields include

the corporate scene,

multiple science dis
ciplines,
engineering, p
harmaceuticals, and medicine
3
.
Opinions in
any

market place can make
or break a product, but
valuable lessons can be learned from the already available data.
This information
is important in changing behavior, experiences, and ultim
ately the
bottom line
5
.


Traditional
data
analysis is hypothetico
-
deductive, meaning the analyst creates a
hypothesis and then draws a conclusion based on the analytical tests performed. Data
mining tools can be used for hypothesis testing

but the nature of this method allows for
an inductive learning approach
;

meaning an analyst tests

first,

then hypothesizes
4
.

Data
mining examines a collection of data and then asks, “What are all of the hypotheses that
this data supports?”
7(p.31)
.

A data
driven approach is attractive because it allows analysts
to look at the relationships and patterns within the data that may have otherwise been
missed
3
.

Prather
8
found that real
-
time clinical data from actual patients can be significantly
rich in clinical
detail. Dates can usually be correlated with a specific data point. Thus,
analysts can derive important insight to clinical knowledge and patient specific
experiences.


Harrison
9
determined
that mining medical and laboratory data is currently very
challen
ging. Also there may be political and legal challenges. However, he is optimistic
HIT Midterm Paper
4


that substantial potential exists.
As the data volume begins to grow, standard data
representations will become more prevalent and the issues around clinical data mining
will

decrease.



Rodiques

and collegues
10
, with Health Organizations,
believes

that establishing a
boarder perspective on evidence
-
based practice (EBP) cannot occur without information
systems and advanced technology. They studied the importance of nursing researchers
in their ability to turn repositories

of data

into useful knowled
ge available to guide
interdisciplinary nursing practice into outcomes. T
heir impact on

nursing applications
including outcomes
in research, decision support, i
nternet based resear
ch, and
(EBP)
.

Alemi and Hurd
1
1

noted

that
patient surveys are important t
o document whether
care processes have provoked an improvement to an organization’s or a clinician’s
performance. Unsolicited patient complaints are valuable
in providing insight to the
patient’s voice. Measuring satisfaction rates helps to identify defects in the care;
complaints measure the defects in the care process. By calculating and documenting
both the satisfaction and the complaint rates, an
significan
t

amount of knowledge can
be obtained.



Method
s:



Days to Next Complaint


The data

for this project

was made available by Dr. Farrokh Alemi, Ph.D. who is
a professor at Georgetown University in the Health Systems Administration Department.
The official data collector was Vikas Arya.

Arya

complied data for this analysis by
accessing

“RateMDs.com” (
http://ratemds.com/
). From the opening page, the link “Find
HIT Midterm Paper
5


Doctors By Specialty” was chosen. From the next screen she selected the following
combinations of states and specialty:

1.

DC and Pediatrician

2.

VA and Pediatrician

3.

MD and

Pediatrician



Once the data was compiled into one large data set
,

the analysis was conducted
.
The data included ratings on the

office staff,
pe
diatrician punctuality, pediatrician
helpful
ness, and finally pediatrician knowledge
. To begin, pediatricians
with less than
four comments listed by parents and or patients were automati
cally eliminated from

the
analysis. Next, physician comments were converted to numerical ratings.
Then we
determined if a
n

office interaction was negative giving the visit a “Yes” or if the office
interaction was positive giving the visit a “No”. A visit was deemed

as “Yes” if a 3.5 or
less was scored in the
pediatrician
punctual, helpful, and knowledgeable categories. A
“No”

was assigned to a visit if a pediatrician scored more than a 3.5 in
any

of these
three categories.
Staff ratings
were

not included in this
analysis because these ratings
were not a direc
t reflection of the individual pediatricians
.

Then
,
we focused on th
e

days/analysis until the next pediatrician complaint

would occur
.

We chose to focus on the days/visits until the next complaint versus the
analysis of rate of complaint because less data is required

for this type of analysis
11
.

If D
designates the daily
probability of a complaint, then

the value can be found from the
following formula:




























From here we calculated the time to the
next complaint by using the following

HIT Midterm Paper
6



formula:
















Sentiment Analysis

Once the above steps where completed, each individual comment gathered from
“RateMDs.com” (
http://ratemds.com/
) was explored. Polarization was established by
reviewing each comment to classify adjective word and/or word combinations as
positive or negative. The adjectives were sorted and then each word frequency

was
counted. “Zeros” were replaced with “ones” in the event that a word did not appeared in
both the positive and negative adjective lists. The final adjective ratio was calculated
and the results displayed as a number with praises over complaints.

Result
s:


Days to Next Complaint

Below
in Table 1
is a summary e
xample

showing

six pediatricians

who are

practicing in the state of Virginia.
It is easy to see that with the minimal amount of
information obtained on each pediatrician, an analyst can then calculate and predict the
information displayed in the
table. For example, Dr. Kaftarian
’s first

complaint was on
August 5, 2008
; his

last comp
laint w
as noted on December 15, 2009;
he had one
complaint

associated with his

name at the time of data collection; the daily probability of
a complaint occurring

is 0.002
;
and it is predicted that his

next complaint w
ill be in four
hundred
ninety
-
six

days from now
respectfully
.


Data on 117

pediatricians were was
collected and

fully analyzed
: (DC
--
2 pediatricians; VA
--
60 pediatricians; and MD
--
55
HIT Midterm Paper
7


pediatricians)
. Appendix 1

contains all pediatricians in DC, VA, and MD used in this
s
tudy
.






Pediatrician Summary Table

Sample

Pediatricians

Date of First

Complaint

Date of Last

Complaint

Number of

Complaints

D of
Complaints

Average
Days to NC

Kaftarian, DC

8/5/08

12/15/09

1

0.0020

496.00

Lang, DC

6/26/06

5/28/09

4

0.0037

265.75

Zollicoffer,
MD

7/17/08

7/1/09

2

0.0057

173.50

Mengers, MD

6/6/07

7/17/09

2

0.0026

385.00

Hopkins, MD

11/12/07

11/9/09

3

0.0041

241.67

Brinkmann, MD

5/2/06

11/14/08

1

0.0011

926.00

Table
1



Sentiment Analysi
s


A

total
of 200

comments were reviewed
in detail
during

th
is portion of the data
analysis:

66

comments
where negative (meaning the patient the pediatrician scored a
3.5

or less
i
n the punctual, helpful and knowledgeable categories) and 144 comments
were positive (meaning the pediatrician scored
3.5

or greater
i
n the punctual, helpful
and knowledgeable categories)
.

Table 2 displays the f
our

most commonly found words
from

the
positive comment
s analyzed
.
A more detailed list of

commonly occurring

positive and negative
word or word combinations

can be found in
Appendix 1
.
Table 2
displays t
he four adjectives that
appeared

most frequentl
y. The word
knowledgeable

occurred thirty
-
three more times

with a

positive
connotation than a negative one
.
Loves

and
patient

were
adjectives found to be
thirteen more times

likel
y

noted

in a positive
comment

and the word
always

was twelve

times

more positive than negative.



HIT Midterm Paper
8


Sentiment Analysis

Word

Praise/Complaint Ratio

knowledgeable

33

loves

13

patient

13

always

12

Table 2



Discussion:



Days to Next Complaint

and
Sentiment Analysis
tie together to show the
relationship between a pediatrician’s complaint time and negative comments associated
with him/her as a pediatric health care provider. In other words,

if a

pediatrician
has a
short time period between patient/pa
rent complaints, then the likeliness of
having
negative

word/word combination

adjective
(
s
)

associated with a com
ment
is greater.


Limitations must be recognized in this study as having the potential to shift the
analytical results. First, the sample size
was small. Second, the human subjectivity
when classifying word/word combinations as either positive or negative must be
realized. The third limitation occurred when calculating the word ratios
.
During this
portion of the data analysis,
a

zero


was substi
tuted for a

one

.

For example, if the
word
knowledgeable

occurred 33 times in the positive comments and 0 times in the
negative comments



cannot be calculated
.
Therefore, we calculated



= 33.



As evident by the data, it can be understo
od that the
interplay

between

words

and meaning can be

power
ful
.

Thousands of physician/patient interactions occur every
day
and each individual involved will interpret and possibly derive different meaning
from the

same

interaction. This is why w
hen conducting analys
is
,

in attempt to interpret
how
one specific

individual may have been feeling at a specific point in time, it is
important to remove the temptation of subjectivity. I
f an analytical approach is taken

to
HIT Midterm Paper
9


conduct sentiment analysis
,

then subjectivity will be

removed. Further algorithm

develop
ment is suggested in order to continue advancing the
interpretations elicited
from sentiment analysis. As technology and sentiment analysis advance, health care
providers will have access to a new and plentiful
database
.
Physicians will have the
abili
ty to
get feedback on their daily patient interactions in real time and to make
adjustment where necessary to maximize growth potential
.




































HIT Midterm Paper
10




References


1.

Allen L.W., Creer E., Leggitt M.: Developing a patient complaint tracking system
to improve performance.
Jt Comm J Qual Improv

26:217
-
226, Apr. 2000.

2.

Bendall
-
Lyon D., Powers T.L.: The role of complaint management in the service
recovery process.
Jt Comm J

Qual Improv

27:278
-
286, May 2001.

3.

Berry MJA, Linhoff G.
Data Mining Techniques for Marketing, Sales and
Customer Support
. New York: Wiley; 1997.

4.

Berger Anne M., Berger Charles R. Data Mining as a Tool for Research and
Knowledge Development in Nursing.
C
omputers, Informatics, Nursing

22:123
-
131, May/June 2004.

5.

Colier K, Carey B, Grusy E, Marjanieme C, Sautter D. A perspective on data
mining. Paper presented at: Center for Data Insight at Northern Arizona
University; July 1998; Flagstaff, Ariz. Available
at:
http://insight.cse.nau.edu/downloads/DM%20Perspecive%20v2.pdf
. Accessed
June 2010.

6.

Wright, Alex. Mining the Web for Feelings, Not Facts.
The New York Times
,
August 24, 2009.

7.

Pyle D. Data Preparation of Data Mining. San Francisco, Calif: Morgan
Kaufmann; 1999.

8.

Prather J.
Exploratory Data Analysis to Detect Preterm Risk Factors
. Durham,
NC: Department of Biomedical Engineering, Duke University; 2000.
Dissert
ation
Abstr

Int

62:952.

9.

Harrison James H. Introduction to the Mining of Clinical Data.
Clin Lab Med
.
28:1
-
7, 2008.

10.

Rodriques RJ. Information systems: the key to evidence
-
based health practice.
Bull World Health Organ
. 2000;

78(11): 1344
-
1351.

11.

Alemi, Farrokh, Hurd, Patrick. Rethinking Satisfaction Surveys: Time to Next
Complaint.
Jt Comm on Acc of Healthcare Org

35:155
-
16, March 2009.
















HIT Midterm Paper
11




Appendix 1


Pediatrician Summary Table


Pediatrician

Name/State


Date of
First

Complaint


Date of Last

Complaint


Number of

Complaints


D of
Complaints


Average
Days to
NC

Dr. Hayk Kaftarian, DC

8/5/08

12/15/09

1

0.0020

496.00

Dr. Nicole Lang, DC

6/26/06

5/28/09

4

0.0037

265.75

Dr. Michael Zollicoffer, MD

7/17/08

7/1/09

2

0.0057

173.50

Dr.
Smita Mengers, MD

6/6/07

7/17/09

2

0.0026

385.00

Dr. Katherine L Hopkins, MD

11/12/07

11/9/09

3

0.0041

241.67

Dr. Kirsten Brinkmann, MD

5/2/06

11/14/08

1

0.0011

926.00

Dr. Jay Bernstein, MD

9/19/07

10/18/09

3

0.0039

252.33

Dr. Najla Abdur
-
Rahman, MD

5/21/07

8/17/09

1

0.0012

818.00

Dr. Ziad Idriss, MD

11/30/05

4/30/09

2

0.0016

622.50

Dr. Seth Eaton, MD

8/18/07

11/29/09

8

0.0096

103.25

Dr. Robert Miller, MD

9/26/08

12/7/09

6

0.0137

71.83

Dr. Maegan Chaney, MD

6/10/08

1/27/09

1

0.0043

230.00

Dr.
Jackie Gilliard

6/12/07

10/22/09

4

0.0046

214.75

Dr. Michael Lasser, MD

3/6/07

8/27/09

2

0.0022

451.50

Dr. Kevin Kasych, MD

8/13/08

3/5/10

1

0.0018

568.00

Dr. Corinne Coyner, MD

1/5/08

12/15/09

1

0.0014

709.00

Dr. Aziza Alam, MD

3/30/08

10/28/09

1

0.0017

576.00

Dr. Ogunrinde, MD

1/23/06

3/8/09

2

0.0018

569.00

Dr. Sandra Takai, MD

1/7/07

9/27/09

2

0.0020

496.00

Dr. Giorgio Kulp, MD

9/19/07

10/24/09

2

0.0026

382.00

Dr. Aruna Khurana

4/12/07

3/31/10

6

0.0055

179.67

Dr. Paul Ambush, MD

11/13/07

2/4/09

0

0.0000

-

Dr. Daniel Hunsinger, MD

5/11/07

10/28/09

2

0.0022

449.50

Dr. Dianna E. Abney, MD

6/22/07

1/21/10

4

0.0042

235.00

Dr. Clifford Galanis, MD

10/21/08

8/20/09

0

0.0000

-

Dr. Elias Gouel, MD

3/8/09

8/31/09

0

0.0000

-

Dr. Samuel Libber,
MD

6/19/06

9/2/09

3

0.0026

389.33

Dr. Samuel Williams, MD

3/3/04

8/25/09

0

0.0000

-

Dr. Amina Watson, MD

11/16/08

3/30/10

1

0.0020

498.00

Dr. Donald Schneider, MD

3/13/06

3/4/09

3

0.0028

361.33

Dr. Farhad Aliabadi, MD

12/29/05

2/7/09

1

0.0009

1135.00

Dr. Stephen Joseph Hittman,
MD

10/27/05

3/27/10

0

0.0000

-

Dr. Deborah Bittar, MD

2/3/06

4/6/10

6

0.0039

252.83

Dr. Peter Ferra, MD

3/27/08

4/3/09

1

0.0027

371.00

Dr. Travis Ganunis, MD

6/7/07

10/26/09

0

0.0000

-

HIT Midterm Paper
12


Dr. Arif Mannan, MD

1/7/08

1/5/10

1

0.0014

728.00

Dr. Peter Rowe, MD

8/6/05

2/28/09

1

0.0008

1301.00

Dr. Laura Lieberman, MD

1/1/06

3/8/09

4

0.0034

289.50

Dr. Arnold Brenner, MD

10/7/07

3/16/10

3

0.0034

296.00

Dr. Marianne Fridberg, MD

8/4/05

5/15/08

0

0.0000

-

Dr. Oliver Galita, MD

1/10/06

4/6/10

0

0.0000

-

Dr. Jeffries Bucci, MD

3/29/07

3/29/10

2

0.0018

547.00

Dr. Victor Abdow, MD

5/27/05

4/1/10

13

0.0073

135.15

Dr. Eugene Sussman, MD

12/18/06

11/18/09

2

0.0019

532.00

Dr. Jeffrey Bernstein, MD

12/18/04

9/14/09

0

0.0000

-

Dr.
Jason Goldstein, MD

3/28/08

2/26/10

1

0.0014

699.00

Dr. Tahir Sait, MD

4/7/06

2/6/10

0

0.0000

-

Dr. Arthur Guarinello, MD

8/22/06

9/1/09

0

0.0000

-

Dr. Giovanni Impeduglia, MD

1/3/06

9/13/07

0

0.0000

-

Dr. Michael J. Scobie, MD

1/12/06

7/7/09

2

0.0016

635.00

Dr. Ken Schuberth, MD

3/9/04

10/7/09

3

0.0015

678.33

Dr. Stephen Cooper, MD

11/10/08

11/19/09

1

0.0027

373.00

Dr. Steven Alcuri, MD

1/22/07

2/20/10

3

0.0027

374.00

Dr. Frank Israel, MD

9/14/05

1/29/10

1

0.0006

1597.00

Dr. Diane Landrum, MD

7/20/07

3/9/09

1

0.0017

597.00

Dr. Leigh Naughton, MD

3/13/07

7/11/08

1

0.0021

485.00

Dr. Sunitha Venugopal, MD

2/24/04

12/11/08

0

0.0000

-

Dr. Claudia E. Sussdorf, VA

4/19/06

7/29/09

8

0.0067

148.63

Dr. Barbara A. Stevens, VA

6/18/08

9/19/09

3

0.0066

151.67

Dr. Crestwood, VA

1/12/08

2/23/10

4

0.0052

192.25

Dr. Katherine Abbott
Crestwood, VA

2/27/07

6/28/09

5

0.0059

169.40

Dr. Maura Carroll, VA

9/19/07

8/17/09

2

0.0029

348.00

Dr. Kathleen Lundgren, VA

6/24/08

7/21/09

2

0.0051

195.00

Dr. Diane
Dubinsky, VA

2/13/07

11/16/09

1

0.0010

1006.00

Dr. Bernadette Wells, VA

8/25/06

1/11/10

8

0.0065

153.38

Dr. Maria Juanpere, VA

2/18/05

2/15/09

2

0.0014

728.00

Dr. Leonard Touchette, VA

5/8/07

11/29/09

4

0.0043

233.00

Dr. Gayle Schrier Smith, VA

5/21/07

2/18/10

1

0.0010

1003.00

Dr. Jacqueline Hoang, VA

1/21/09

12/22/09

2

0.0060

166.50

Dr. Linda Purcell, VA

2/28/09

7/28/09

1

0.0067

149.00

Dr. Ghazala Y. Khan, VA

3/12/07

3/15/08

1

0.0027

368.00

Dr. Wendy C. Ault, VA

2/24/06

3/14/09

1

0.0009

1113.00

Dr. Stacey Staatz, VA

4/27/07

11/22/08

0

0.0000

0.00

Dr. Michele Reilly, VA

8/19/07

9/2/08

0

0.0000

0.00

Dr. Ruth Vogel, VA

3/22/06

11/6/09

4

0.0030

330.25

Dr. Marijana Ducic, VA

11/30/05

9/4/09

2

0.0015

686.00

Dr. Maura Eriksson, VA

7/15/07

5/19/09

5

0.0074

133.80

Dr. Paula Fergusson, VA

11/11/04

2/4/09

1

0.0006

1545.00

Dr. Lynne Myers, VA

7/12/04

6/26/08

1

0.0007

1444.00

HIT Midterm Paper
13


Dr. Anuradha Takanti, VA

5/13/08

9/9/09

1

0.0021

483.00

Dr. Anne Bradshaw, VA

7/6/06

1/21/10

1

0.0008

1294.00

Dr. Candace
Fugate,VA

11/7/05

2/18/09

0

0.0000

0.00

Dr. Diane Halpin, VA

7/12/04

3/17/09

4

0.0023

426.25

Dr. D. Gregory Bott, VA

6/16/08

7/31/09

8

0.0195

50.25

Dr. Allen E. Aaronson,VA

12/19/05

3/31/09

1

0.0008

1197.00

Dr. Bassam Atiyeh, VA

9/20/06

12/9/09

4

0.0034

293.00

Dr. Charles Amory, VA

9/18/08

10/5/09

0

0.0000

-

Dr. Luther Beasley, VA

7/10/06

9/7/09

0

0.0000

-

Dr. Mark Holman, V A

12/5/05

6/18/09

5

0.0039

257.20

Dr. Louis Puppo, VA

11/13/07

3/1/10

2

0.0024

418.50

Dr. Michael Caplan, VA

7/2/07

8/27/08

0

0.0000

-

Dr. Rui Rodrigues, VA

7/31/08

5/19/09

0

0.0000

-

Dr. Jeremy Fishelberg, VA

1/22/08

1/9/09

0

0.0000

-

Dr. Grover Robinson III
, VA

10/30/04

2/27/07

3

0.0035

282.33

Dr. Evan Karp, VA

5/29/08

7/10/09

0

0.0000

-

Dr. David Katz, VA

7/25/08

3/4/10

1

0.0017

586.00

Dr. R. Skyler McCurley, VA

8/2/05

6/14/09

0

0.0000

-

Dr. Thomas Rowe, VA

8/20/08

10/7/09

1

0.0024

412.00

Dr. Timothy Walker, VA

11/15/04

2/11/09

1

0.0006

1548.00

Dr. Christopher Wrubel, VA

3/20/07

9/14/07

0

0.0000

-

Dr.
Richard Schwartz, VA

8/20/06

3/9/10

1

0.0008

1296.00

Dr. Ely Mouchahoir, VA

4/10/08

10/21/09

1

0.0018

558.00

Dr. Christopher Rossbach, VA

1/11/07

8/9/09

0

0.0000

-

Dr. Martin Forman, VA

6/29/06

8/2/08

2

0.0026

381.50

Dr. Arthur G. Gower III
, VA

11/3/05

3/22/10

4

0.0025

399.00

Dr. Vito Giannuzzi, VA

6/22/05

4/8/09

1

0.0007

1385.00

Dr. Walter Kilby, VA

2/9/06

9/21/09

5

0.0038

263.00

Dr. Michael Amster, VA

7/15/07

8/14/09

2

0.0026

379.50

Dr. William Rees, VA

6/13/06

4/11/07

3

0.0099

99.67

Dr. Gary
Bergman, VA

8/29/07

6/23/09

4

0.0060

165.00

Dr. William Carr, VA

9/28/06

8/19/09

8

0.0076

131.00

Dr. William Feldman, VA

3/12/07

10/4/08

4

0.0070

142.00

Dr. Brendan Sullivan, VA

11/12/08

3/17/10

3

0.0061

162.33

Dr. Brian Butcher, VA

1/13/09

2/8/10

2

0.0051

194.50

Dr. James R. Baugh, VA

6/21/07

7/21/09

3

0.0039

252.67

Dr. John D. Farrell, Jr., VA

11/30/05

8/16/09

5

0.0037

270.00









HIT Midterm Paper
14




Appendix 2


Adjective Word/Word Combination Summary Table

Word

Word Frequency

K
nowledgeable

33

L
oves

13

P
atient

13

A
lways

12

E
xtremely
, rushed, excellent

8

R
ushed

8

N
ice

7

E
fficient
, highly, kind

6

E
xperienced

5

Q
uickly
, trust

5

A
mazing
, attentive, cares, compassionate, listens,
respects, warm

4

A
ccessible
, awesome, calm, comfortable,
competent, concerned dedicated, fantastic,
genuine, incredibly, reassuring, smart, terrific

3

A
fraid
, birth, count on, family, friend, fun,
immediately, important, laid back, lucky, marvelous,
miss, never, pleased, positive,

quick, rare, really,
responsive, satisfied, straightforward, switch, travel,
understanding

2

C
ancelled
, misdiagnosed, mistake, negative, not,
pushy, screwed, thanks, uncaring, uncomfortable,
waiting

0.5

P
roblems
, unapproachable, worse, wrong

0.333333333

U
nprofessional

0.25

H
orrible

0.2