Homework1: Opioid Pills Analysis

In this homework, we will use a dataset from Washington Post for opioid pills analysis.

Read the story: How deeply did prescription opioid pills flood your county? See here. (using a private window to input an email to view or download the news PDF)

The database represents transaction of opioid pills sales by a buyer sold by a provider. We mainly analyze the transactions by the buyers. 

In this homework, we will build a database for California and Kentucky, and compare the differences. 

The California State dataset can be downloaded here. The zip file size is 433.8MB, and the unzipped dataset is  7.66GB.  

The Kentucky State dataset can be downloaded here. The zip file size is 106.6MB, and the unzipped dataset is 2.15GB. 

In addition, we have a US zip code database, which includes population data for each zip code. The file is available to download. 

Total points: 20.   Due: 11:59pm EST, March 2,2025. 

1. Create DEA table, load the data, and create indexes.

1a (1 points). Create a DEA database table with following information (column descriptions in schemas.xlsx) and put the SQL into a file createdeatable.sql:

CSE532.DEA(REPORTER_DEA_NO,REPORTER_BUS_ACT,REPORTER_NAME,REPORTER_ADDL_CO_INFO,REPORTER_ADDRESS1,

REPORTER_ADDRESS2,REPORTER_CITY,REPORTER_STATE,REPORTER_ZIP,REPORTER_COUNTY,BUYER_DEA_NO,

BUYER_BUS_ACT,BUYER_NAME,BUYER_ADDL_CO_INFO,BUYER_ADDRESS1,BUYER_ADDRESS2,BUYER_CITY,

BUYER_STATE,BUYER_ZIP,BUYER_COUNTY,TRANSACTION_CODE,DRUG_CODE,NDC_NO,DRUG_NAME,QUANTITY

,UNIT,ACTION_INDICATOR,ORDER_FORM_NO,CORRECTION_NO,STRENGTH,TRANSACTION_DATE,CALC_BASE_WT_IN_GM,

DOSAGE_UNIT,TRANSACTION_ID,Product_Name,Ingredient_Name,Measure,MME_Conversion_Factor,

Combined_Labeler_Name,Reporter_family,dos_str,MME).

Note: 

TRANSACTION_DATE needs to be a DATE format. Please make sure your datatypes can support the queries in Q3. 

To save space and improve performance, you can use the COMPRESS option for your table. e.g., create mytable(...)  COMPRESS YES.

1b (1 points). Load the csv file into the database by modifying the loading script. Run it as:

db2 -tf load.sql

Validate there were 3,595,806,356  prescription pain pills supplied to Kentucky, and 15,268,725,465  prescription pills supplied to California.   (This information was copied from WP website and not accurate).

(DOSAGE_UNIT is the total number of pills in a transaction/record.)

1c (1 point). Based on queries in 3, create proper indexes to make the queries more efficient (createdeaindexes.sql). Attributes used in predicates and used in GROUP BY are candidates for indexes. 

2. Create and load data for zip code population table (the population in each zip code).

2a (1 points). Create a table CSE532.ZIP  (zip,state,county,pop) to represent the population of zip codes, where "pop" refers to "population" column in the zip_code_database.csv file (createzip.sql). The documentation of the zip code database can be found at here.

Note that there might be duplicated zip codes in the table.  A zip code can cross counties. Some zip codes have zero populations as they may represent a special administration or military region.  

2b (1 point). Create a loading script to load the zip_code_database.csv to the database (zipload.sql). Note that you have to specify columns to be loaded. Refer to load documentation on how to specify columns to be loaded using "METHOD P". 

3. Write SQL queries. (Note that we are only querying pills by buyers.) 

3a (5 points). Return counts of pills for each state ( of buyer)  and for each county, by year and by month using a single OLAP query (q3a.sql).  DOSAGE_UNIT is the total number of pills in a transaction. Save result in q3a.txt. 

Note: month() and year() functions extract month and year from a DATE datatype respectively. 

3b (5 points). Return smoothed counts of pills with a moving window (preceding 2 months and following 2 months) for Perry County in Kentucky (q3b.sql).   DOSAGE_UNIT is the total number of pills in a transaction. Save result in q3b.txt. 

Draw the two curves with Excel or other drawing tools to show the difference between smoothed count and normal count trends without smoothing.

Note: 

1) month() and year() functions extract month and year from a DATE datatype respectively. 

2) You may use CONCAT or "||" to concatenate two values. To generate a uniform "yeardate" representation (e.g., 200801), you may also consider to use CASE function. (example case function)

 3) Consider to use common table expression. 

3c (5 points). Find the top 15 zip codes for each state (California and Kentucky), with most pills sold in terms of MME when normalized by the population in the zip codes, i.e., zip codes with most sold total MME of pills per person. Please use RANK function. (q3c.sql)

Note:

A zip code be be associated with multiple counties. You may want to check if that is true before writing the query. 

MME is a normalized amount which better represents the total strength of the pills sold in the transaction.

Homework Submission

Please zip your SQL scripts and results:

createdeatable.sql

load.sql

createdeaindexes.sql

createzip.sql

zipload.sql

q3a.sql (and result q3a.txt)

q3b.sql (and result q3b.txt)

q3c.sql (and result q3c.txt)

A readme.txt file (optional) explaining anything not included. 

Please go to BrightSpace, and submit the zip file under homework 1.