MySQL: count consecutive rows with the same value - mysql

Question

I have read a few similar questions on counting consecutive rows, but none of them gave me a clear answer. I hope someone could give me some help with my problem. I have the following table as an example.
create table medical
(PatientID int,
Date Date,
TookTest int
);
insert into medical(PatientID, Date, TookTest)
values
(1, '2014-01-01', 1),
(1, '2014-01-05', 1),
(1, '2014-01-10', 1),
(2, '2014-01-01', 1),
(2, '2014-01-10', 0),
(2, '2014-01-20', 1),
(3, '2014-01-01', 1),
(3, '2014-01-07', 1),
(3, '2014-01-12', 1),
(3, '2014-01-21', 1),
(4, '2014-01-03', 1),
(4, '2014-01-05', 1),
(4, '2014-01-22', 0),
(4, '2014-01-27', 1)
This table is used to find out which patient took a medical test on certain dates. The PatientID and date columns are pretty self-explanatory. The last column, TookTest is a binary indicator column where 1 indicates that a patient took a test and 0 otherwise. The patientID and date are sorted at the time of this table's creation. I would like to count the number of patients who took tests at least 3 times consecutively. In our example, PatientID 1 and 3 took 3 or more tests. So the answer is 2. Could anyone show me how to write a query in MySQL? Thanks for your help in advance!

Answer 1

SELECT
m_id
FROM(
SELECT
m.PatientID AS m_id,
m.Date AS m_date,
m.TookTest,
IF(m.TookTest = 1 AND #b = m.PatientID, #a := #a +1, #a := 0) AS new_count,
#b := m.PatientID
FROM medical m
JOIN (
SELECT
#a := 0,
#b := 0
) AS t
) AS TEMP
WHERE new_count >= 2
GROUP BY m_id
this does the calculation for you.. only thing is it looks a little weird because the count starts at 0 instead of 1 so if its 3 consecutive the count will be 2. this does what you requested..... see the fiddle if you have questions http://sqlfiddle.com/#!2/22ba28/12

Answer 2

This code also works, too.
set #test = 0, #id=0, #count=0;
select m.id, max(count)
from (
select
#count := if(TookTest = 1 and PatientID = #id, #count+1, 0) as count,
#test := Tooktest,
#id := PatientID as id
from medical) as m
group by m.id
having max(count) >=2;
This code counts the historical max consecutive rows of TookTest as opposed to the most recent consecutive row count (This distinction is not relevant here because the example data is too small to make any difference between most recent consecutive count and historical max consecutive count.)
My coding background is R, Python and Java. Possibly because of my personal particular coding experience, it is hard for me to grasp using join twice in this context. My code above is a way to get around it. I hope this answer helps others in a similar situation like mine.