SQL Server + Entity CF won't use table index without hint - sql-server

Question

I have an Entity class which generates a fairly simple table: (obfuscated)
public class TableName
{
public int TableNameId { get; set; }
public int OtherTableId { get; set; }
public int UserId { get; set; }
public bool IsActive { get; set; }
// removed other columns for brevity
public virtual OtherTable OtherTable { get; set; }
public virtual User User { get; set; }
}
Entity automatically creates a primary key on TableNameId and foreign keys with indexes on OtherTableId and UserId.
The table will only ever have one record where IsActive is true for each combination of UserId and OtherTableId, but records are deactivated and replaced frequently.
Records are located with statements similar to this:
var record =
context.TableName
.FirstOrDefault(x => x.UserId == userId
&& x.OtherTableId == otherTableId
&& x.IsActive);
We found that with no other indexes, performance while pulling a record was poor since it has to scan through the UserId and OtherTableId indexes to find the records where IsActive = 1.
To address this, we added this index:
create nonclustered index ix_TableName_UserId_OtherTableId_IsActive
on TableName (UserId, OtherTableId, IsActive)
However, SQL does not use it. Instead, with this index in place it does a scan on the table's primary key!
Since Entity is pulling the entire row, I understand that the above index is not covering. However, in every other case we've had we're able to add an index similar to this one and it results in an index seek combined with a key lookup and gives satisfactory performance. In this case it will not use the index we're creating unless we use an index hint in the select statement. We'd like to avoid this since it requires custom SQL which kind of defeats the purpose of Code First Entity.
After much trial and error, the only index that SQL will use on its own is one that includes every column and is thus fully covering. However, I'd rather avoid this since it means essentially storing the table twice and 99% of the rows will be useless since IsActive = 0.
--- Edits below ---
For the code above, I found that Entity generates the following SQL. I don't know why it uses a subselect, but the SQL optimizer seems to filter it out (running the full SQL or the inner select results in the same plan).
SELECT TOP (1)
[Project1].[TableNameId] AS [TableNameId],
[Project1].[OtherTableId] AS [OtherTableId],
[Project1].[UserId] AS [UserId],
[Project1].[IsActive] AS [IsActive]
FROM ( SELECT
[Extent1].[TableNameId] AS [TableNameId],
[Extent1].[OtherTableId] AS [OtherTableId],
[Extent1].[UserId] AS [UserId],
[Extent1].[IsActive] AS [IsActive]
FROM [dbo].[TableName] AS [Extent1]
WHERE ([Extent1].[UserId] = <userId>)
AND ([Extent1].[OtherTableId] = <otherTableId>)
AND ([Extent1].[IsActive] = 1)
) AS [Project1]