SQL DB table Relationship - sql

Question

I am trying to create a small DB based on flat files that I import from a non Database System. The Import is working, the DB is good but I added a new table that contains data from another system. I am trying to create a relationship between the tables but because one table has duplicate rows (flat file is the source) I am not able to set that relationship.
Example: Table 1 lists all procedures done for a patient by a physician.. the patient can have many of the same procedures on the same day by the same physician (hence the duplicate rows) ... Table 2 has a list of Physicians and their ID #s ... I want to set up a relationship between the two tables based on the physician's name but I am getting errors because of the non unique data.
Anyone has a tip?
Thanks

Answer 1

the patient can have many of the same procedures on the same day by the same physician (hence the duplicate rows)
Normally, you should be able to set a foreign key relationship from Table1 to Table2 even in the presence of duplicate rows. This kind of error usually means you're trying to set the foreign key in the wrong table.
-- Your "Table2"
create table physicians (
physician_id integer primary key,
physician_name varchar(35) not null -- names are not unique
);
insert into physicians values
(1, 'Doctor Who'), (2, 'Dr. Watson');
create table patients (
patient_id integer primary key,
patient_name varchar(35) not null -- names are not unique
);
insert into patients values
(100, 'Melville, Herman'), (101, 'Poe, Edgar Allen');
-- Your "Table1"
-- Allows multiple physicians per date.
create table patient_procedures(
patient_id integer not null references patients (patient_id),
physician_id integer not null references physicians(physician_id),
procedure_date date not null default current_date,
procedure_name varchar(15) not null,
primary key (patient_id, physician_id, procedure_date, procedure_name)
);
insert into patient_procedures values
(100, 1, '2012-01-02', 'CBC'),
(100, 1, '2012-01-02', 'Thyroid panel');

Answer 2

I'm not sure from your description where the duplicate-data problem is. You have:
Table 1: procedures. Could be lots of rows for same physician
Table 2: physicians. Should be 1 row per physician (but there may be duplicates)
The relationship that would make sense would be 1[table 2 physician row] -> many[table 1 procedures rows]. i.e. table 2 would be the primary key table in the relationship: each table 2 row relating to between 0 and "many" table 1 rows. If you try to create this kind of relationship, then multiple duplicate table 1 rows are not a problem.
If, however, you have multiple rows per physician in table 2, then you won't be able to create this kind of relationship, because table 2 rows are not unique and thus can't act as the primary key element in the relationship. The problem then is one of data-cleansing: figuring out which rows in table 2 are duplicates, updating the table 1 rows to point to just ONE physician out of the duplicates, and then deleting the duplicate rows from table 2.
You mention physician ID#s and physician names. Physician name would be a bad choice for a unique key; if a user tries to add a new physician called "John Smith" when there already is another physician of that name, either
You've set up a unique index on PhysicianName, their change gets rejected, and you have an irate user; or
You haven't, and all the existing physician's (let's call him John A. Smith) procedures will be associated with the other physician (let's call him John B. Smith) and vice versa.
The relationship should be set up using the physician ID. If Table 1 (Procedures) includes a physician ID column, you're in luck. If it only includes physician Name, then you may have a data-cleansing problem, if there are already duplicate physician Names in Table 2.