INFO5306 Enterprise Healthcare Information Systems
Assignment 2 (Individual) – Health Data Analysis
20 Marks in Total
Weight: 10% of overall UoS
Due Date: Report (Week 13 - Wed 6th November 5pm)
Objective
This assessment is designed to let you demonstrate your understanding of the data analysis software you
used during the lab sessions, such as Microsoft Excel Pivot. You are required to prepare a report that
answers several questions specified in this document.
Data Sets and Software Availability:
• All the software you need to do this assignment are available on the Lab PCs. You can do the
assignment using your own PC but it is your responsibility to get the same answer as the Lab PC.
• For Q1, the recommended tool is Microsoft Excel Pivot Table. The required dataset is:
o Medicare_Provider_Charge_Outpatient_APC30_CY2013.csv and has 44,356 rows.
• For Q2, the recommended tool is Microsoft Excel Pivot Table. The required datasets are:
o Medicare_Provider_Charge_Outpatient_APC30_CY2013.csv
o Medicare_Provider_Charge_Outpatient_APC30_CY2012.csv
o Medicare_Provider_Charge_Outpatient_APC30_CY2011.csv
• For Q3, the recommended tool is MATLAB. The required dataset is:
o Diagnostic_Breast_Cancer.xlsx and has 461 rows.
• For Q4, the recommended tool is MATLAB. The required dataset is “SkinLesionDataset”. It has a
train folder (5,400 images) and a val folder (12 images). More details given in Q4.
Warning: The data files are strictly used for the INFO5306 assignment only. You are not allowed to
copy, distribute, adapt, display or include the data in any situations outside this class. The data maybe
inaccurate or misleading when used in real life.
2
Semester 2, 2019
Submission and Notes:
• You must use the provided answer sheet “info5306-assign2-answer-sheet.docx”.
• You need to submit an electronic copy using Canvas.
• For the electronic copy, you must submit your answers using your Canvas account. You may
submit as many times as you like until the due date. You need to submit two files to Canvas:
o One answer sheet file that includes all your answers.
▪ You need to rename it to “info5306-ass2-STUDENTID”, where STUDENTID is
your student ID.
▪ Save your submission as a PDF.
o One zip file containing all your workspace files for question 1 to 4, for example Q1.xlsx
files for Question 1.
▪ Your workspace files will be used to check how you a
ived at your answer.
▪ Matlab workspace can be saved using “save(filename)” command. More details
can be found here: https:
au.mathworks.com/help/matla
ef/save.html
▪ Name the zip file “info5306-assign2- STUDENTID.zip”
• You can access MATLAB on their personally owned PC while they are enrolled at the
University. For more information see http:
sydney.edu.au/students/student-it/apps.html.
• Each question indicates the tool that is recommended that you use to develop your answer.
However, you may use other tools. You must check with the lecturers first and also indicate the
tool used in your assignment answer.
• The tutorial materials covered the approach to the assignment questions. However, some of the
questions may require additional study.
TIP: Read the questions carefully.
https:
au.mathworks.com/help/matla
ef/save.html
https:
au.mathworks.com/help/matla
ef/save.html
http:
sydney.edu.au/students/student-it/apps.html
http:
sydney.edu.au/students/student-it/apps.html
3
Semester 2, 2019
Q1. [5 marks in total] (Recommended Tool: Excel)
The dataset includes estimated hospital-specific charges for 30 Ambulatory Payment Classification
(APC) Groups paid under the Medicare Outpatient Prospective Payment System (OPPS) for Calendar
Years (CY XXXXXXXXXXFor these APCs, the estimated average charges and the average Medicare payments are
provided at the individual hospital level.
APC Code and description identifying the Ambulatory Payment Classification.
APCs are a classification system where individual services (Healthcare
Common Procedure Coding System [HCPCS] codes) are assigned based on
similar clinical characteristics and similar costs.
Provider ID The provider identifier assigned to the Medicare certified hospital facility.
Provider Name The name of the provider.
Provider Street Address The provider’s street address.
Provider City The city where the provider is located.
Provider State The state where the provider is located.
Provider Zip Code The provider’s zip code.
Hospital Refe
al
Region (HRR)
Description
The Hospital Refe
al Region (HRR) where the provider is located.
Outpatient Services The number of services billed by the provider for outpatient hospital
services.
Average Estimated
Submitted Charges
The provider's average estimated submitted charge for services covered by
Medicare for the APC. These will vary from hospital to hospital because of
differences in hospital charge structures.
Average Total Payments The average of total payments to the provider for the APC including the
Medicare APC amount. Also included in Total Payments are co-payment
and deductible amounts that the patient is responsible for.
You may give your answers to two decimal places. Note that in the template tables provided, the exact
wordings of the headings may be different.
Q1.1. [1 mark] Calculate the top five providers that received the highest average payments for their
services. Sort the values from high to low. Your table should look like the one below.
Provider Payment
1
2
3
4
5
Grand Total
4
Semester 2, 2019
Q1.2. [1 mark] Which two types of APC had the lowest average in the Average Estimated Submitted
Charges in states CA, NY, and TX? What were the charges. Your table should look like the one below.
APC … Estimated Submitted Charges
1
2
Grand Total
Q1.3. [1 mark] You will need to add a new data field Average Ratio to answer this question. The
Average Ratio is defined as (Average Total Payments/Average Estimated Submitted Charges). Find out
which APCs have the highest maximum and lowest minimum Average Ratio. Your table should look like
the one below.
APC Highest Maximum
1
Grand Total
Q1.4 [2 marks] Use PivotTable to create one histogram table (bin size of 5000, minimum = 0 and
maximum = 40,000) to demonstrate the number of Outpatient Services in Average Estimated Charge for
each Region. The answer should look like the following
Midwest Region
Bin size = 5000 Number of Outpatient Services
0 – 4999
5000 – 9999
…
…
…
Northeast Region
Bin size = 5000 Number of Outpatient Services
0 – 4999
5000 – 9999
…
…
…
…
…
5
Semester 2, 2019
Q2. [5 marks total] (Recommended Tool: Excel)
You are working as an analyst and you are requested to generate a short report to analyse changing
patterns in the data of the US Medicare Outpatient Prospective Payment System. You are asked to look at
the data from a spatial and temporal perspective, and to analyse changes from Year 2011 to 2013.
Your task:
• Analyse the data, and identify a pattern.
• Create report that:
o Summarises the changes in the selected pattern over time. [1 mark]
o Quantifies whether the pattern changes differently at different locations. [1 mark]
o Includes at least one informative chart or graph. [1 mark]
o Makes a recommendation about future strategy based on this data. [1 mark]
o Is written in a clear style and format. [1 mark]
o MAXIMUM of two pages long.
HINTS:
▪ You should analyse the data and use it to justify your findings
▪ You may need to create new fields
▪ You may need to learn new techniques or functions by yourself
6
Semester 2, 2019
Q3. [5 marks total] (Recommended Tool: MATLAB)
You are given a
east cancer dataset. It contains the co
ect diagnosis made by physicians (ground truth)
as well as by four different algorithms. The attributes of the data are as follows:
• Diagnosis:
o 1 = malignant
o 0 = benign
• Algorithm_A, Algorithm_B, Algorithm_C and Algorithm_D:
o 1 = predicted malignant
o 0 = predicted benign
Q3.1. [1 mark] Calculate the number of true positives (TP), false positives (FP), true negatives (TN) and
false negatives (FN) for each algorithm.
Algorithm A Algorithm B Algorithm C Algorithm D
TP
FP
TN
FN
Q3.2. [1 mark] Calculate the Sensitivity, Specificity, Precision, Recall, F-score and Accuracy for these
four algorithms.
Algorithm A Algorithm B Algorithm C Algorithm D
Sensitivity
Specificity
Precision
Recall
F-score
Accuracy
Q3.3. [1 mark] Create a Receiver Operating Characteristics curve for each of the four algorithms.
Q3.4. [2 marks] You have been asked to make a recommendation about the algorithm that should be
used in an automated decision support system. Which of the four algorithms would you recommend?
Justify your answer.
7
Semester 2, 2019
Q4. [5 marks total] (Recommended Tool: MATLAB)
Skin cancer is the most common cancer in Australia, with melanoma being the most deadly form.
Dermoscopy is a skin imaging modality that has demonstrated improvement for diagnosis of skin cancer
compared to unaided visual inspection. However, even for experienced dermatologists, diagnosis by
human vision can be subjective, inaccurate and non-reproducible. This is attributed to the complexity of
lesion due to variations in size and shape, fuzzy lesion boundaries and different skin colour types.
Motivated by this difficulty, you are required to develop a computer-aided diagnosis (CAD) systems that
can assist the dermatologists’ clinical diagnosis.
Dataset:
You are given a SkinLesionDataset. It has a train folder (5,400 images) and a val folder (12 images). Data
structure is like the followings:
Your Task:
The goal for this CAD is to classify dermoscopic images among 8 different diagnostic categories (AK,
BCC, BKL, DF, MEL, NV, SCC and VASC). You are required to build a simple convolutional neural
networks (CNN) based on the train folder. The trained CNN model will be used to predict the clinical
diagnosis of the images in the val folder. The answer