find max value for specific column while still seeing other columns - postgresql

Question

For tables patient and labh
patient
id lastname
19 patientone
20 patienttwo
patientid lastname loinc datetime numerical
19 patientone 4548-4 2014-05-15 00:00:00 6.5
19 patientone 4548-4 2015-05-15 00:00:00 7.5
19 patientone 4548-4 2016-05-15 00:00:00 3.5
19 patientone 4548-4 2017-05-15 00:00:00 5.5
19 patientone 5000-3 2018-05-15 00:00:00 123
20 patienttwo 4548-4 2013-05-15 00:00:00 2.5
20 patienttwo 4548-4 2012-05-15 00:00:00 1.5
20 patienttwo 4548-4 2011-05-15 00:00:00 9.5
20 patienttwo 4548-4 2010-05-15 00:00:00 3.5
Desired output:
patientid lastname datetime numerical
19 patientone 2017-05-15 00:00:00 5.5
20 patienttwo 2013-05-15 00:00:00 2.5
The labh table hold lab values(numerical), the type of lab (loinc) and when they were done (datetime). I'd like to query for the most recent value of loinc=4548-4 , and i'd like the output to show both the date and the value.
i've tried this below and it shows the most recent dates, but I can't see the values (numerical) at the same time. when I add the numerical column, the it shows all the values, not just the most recent.
Select Distinct patient.id, patient.lastname, Max(Date_Trunc('day', labh.datetime)) As "Date" From patient Inner Join labh On patient.id = labh.patientid Where labh.loinc = '4548-4' Group By patient.id, patient.lastname, patient.firstname Order By patient.id

Answer 1

you haven't selected the numerical column in your query. You can use CTE to store the data temporarily through ranking on pratition over patient id and ordering each partition on the basis of date.
So, according to this, you can try:
WITH summary AS (
SELECT p.id as "Patient ID",
p.lastname as "Patient Name",
l.datetime As "Date",
l.numerical as "Numerical",
ROW_NUMBER() OVER (PARTITION BY p.id
ORDER BY l.datetime DESC) AS rank
FROM patient p
Inner Join labh l
On p.id = l.patientid)
SELECT "Patient ID",
"Patient Name",
"Date",
"Numerical"
FROM summary
WHERE rank = 1;
And this will give you:
Patient ID
Patient Name
Date
Numerical
19
patientone
2017-05-15T00:00:00.000Z
5.5
20
patienttwo
2013-05-15T00:00:00.000Z
2.5
UPDATE
As you've updated the question and changed the expectation, the modified query will be nothing but adding a where condition inside cte construction:
WITH summary AS (
SELECT p.id as "Patient ID",
p.lastname as "Patient Name",
l.datetime As "Date",
l.numerical as "Numerical",
ROW_NUMBER() OVER (PARTITION BY p.id
ORDER BY l.datetime DESC) AS rank
FROM patient p
Inner Join labh l
On p.id = l.patientid
where l.loinc = '4548-4') -- Added this line
SELECT "Patient ID",
"Patient Name",
"Date",
"Numerical"
FROM summary
WHERE rank = 1;
This will give you the same result:
Patient ID
Patient Name
Date
Numerical
19
patientone
2017-05-15T00:00:00.000Z
5.5
20
patienttwo
2013-05-15T00:00:00.000Z
2.5

Answer 2

In order to achieve what you're looking for in Postgres (and other SQL RDBMSes), you need to essentially identify the max value and its corresponding primary key, then join it with the rest of the data set you are looking to retrieve:
SELECT patient.*, labh.*
FROM patient
JOIN labh
ON patient.id = labh.patientid
JOIN (SELECT patientid, max(datetime)
FROM labh
GROUP BY patientid) maxvals
ON maxvals.patientid = labh.patientid AND
maxvals.datetime = labh.datetime