Cypher

Cypher is Neo4j's query language. It has some similarities to SQL, but getting comfortable with it requires some effort on your part. You can start off with the Get started with Cypher page on the Neo4j documentation. Some basic ideas, relevant to healthbase, are covered in this article.

1. Basic example

How do we find all practitioners who work at a specific hospital?

MATCH p=(n:hospital {healthbase_id: 'H:100151'})<-[r:WORKS_AT]-(m) RETURN p;

Let's break this down. The MATCH part of the query specifies what it is that we are looking for. This is a pattern, usually specified like this:

(node1)<-[r]-(node2)

This indicates that we want a relationship r between node1 and node2. Note that the direction of the arrow matters; it must match the direction of the relationship we want.

We can modify this pattern with filters. Nodes have labels (such as hospital or healthsystem) as well as properties (such as healthbase_id or name). Relationships too have labels (such as WORKS_AT) and attributes to them(such as score1). In this case we are looking for a node of type hospital with healthbase_id 'H:100151' which has a WORKS_AT relationship directed towards it. We haven't set a filter on m in this case.

2. Filter by score

Modifying the above query, let's restrict it to give us only strongly affiliated physicians.

MATCH p=(n:hospital {healthbase_id: 'H:100151'})<-[r]-(m:physician) WHERE r.score1 > 90 RETURN p LIMIT 300;

The WHERE clause allows us to specify that we need only relationships that have a score1 greater than 90. We also set a LIMIT on the number of relationships to return.

3. Check if a relationship property is present

You can further restrict the query to show you only relationships that fall into a certain therapeutic area by making sure they have the diagnosis code you want. You can use the exists function to do this.

MATCH p=(n:hospital {healthbase_id: 'H:100151'})<-[r]-(m:physician) WHERE r.score1 > 90 and exists(r.dx_R13) RETURN p LIMIT 10;

4. Path to any depth

In all of the above examples, we were only looking for direct connections. Neo4j also allows us to run queries that span across relationships. Take this for instance -

MATCH p=(n:healthsystem {healthbase_id: 'HS:E416011702'})<-[:OWNED_BY*..]-(m) RETURN p LIMIT 300;

In this case, we are trying to get all OWNED_BY relationships to healthsystem ID 'HS:E416011702', no matter what depth. This will give you many results, so we set a LIMIT on it to be safe.

5. Limits on depth

We can set limits on the depth too. This is recommended so as to not crash the server by looking for nodes that are too far away in the graph. You do this by specifying an asterix and two dots, like so [*..4].

MATCH p=(n:healthsystem {healthbase_id: 'HS:E416011702'})<-[*..4]-(m) WHERE m:hospital OR m:healthsystem RETURN p LIMIT 300;

6. Shortest path

Given two entities, how do you find the most direct relationship between them? You can do this by using the shortestPath function.

MATCH path=shortestPath((cs:healthsystem { healthbase_id: 'HS:E222529464' })-[*..10]-(ms:asc { healthbase_id: 'ASC:07C0001051' }))
RETURN path;

Neo4j comes with many such builtin functions.