Fast Way To Get Values From Previous Record

Hello,

I know that I've seen this question asked on here before, but I can't
find an answer that gives me the performance that I need.

I have a table that stores events for users:

CREATE TABLE Lead_Action_History (
lead_action_seq INT IDENTITY NOT NULL,
lead_action_date DATETIME NOT NULL,
lead_seq INT NULL,
operator_id VARCHAR(20) NOT NULL,
call_time INT NOT NULL,
CONSTRAINT PK_Lead_Action_History PRIMARY KEY (lead_action_seq) )
GO

The table has a foreign key to another table through the lead_seq
column:

CREATE TABLE Lead_Master (
lead_seq INT IDENTITY NOT NULL,
state CHAR(2) NOT NULL,
CONSTRAINT PK_Lead_Master PRIMARY KEY (lead_seq) )
GO

I need to write a query that will give me a sum of call_time broken
down by a column that is in the table joined through the lead_seq.
However, if the lead_seq for a row is NULL then I need to use the
lead_seq for the previous row (based on lead_action_date) for the same
operator.

This is what I came up with:

SELECT LM.state, SUM(call_time)
FROM Lead_Action_History LAH
INNER JOIN Lead_Master LM ON (LM.lead_seq = LAH.lead_seq)
OR (LAH.lead_seq IS NULL
AND LM.lead_seq = (SELECT TOP 1
LAH2.lead_seq
FROM
Lead_Action_History LAH2
WHERE
LAH2.operator_id = LAH.operator_id
AND LAH2.lead_seq
IS NOT NULL
ORDER BY
LAH2.lead_action_date DESC))
GROUP BY LM.state


The problem is that Lead_Action_History has millions of records and
any solution that I've found involves one or more subqueries on it
which kills performance. I am going to look at using a covering index
with the solution above, but I thought that someone here might have
another way of doing this.

I can't really change the structure, but I can play with the indexing.
I would still be curious though how other people model this type of
temporal data in a way that makes it easy to work with.

Thanks!
-Tom.