Get certified in Microsoft Fabric—for free! For a limited time, the Microsoft Fabric Community team will be offering free DP-600 exam vouchers. Prepare now
Hi. I have rows in a data table with patient id, admit datetime and discharge datetime. My goal is to count patients and create other measures in four-hour intervals so I need to explode these rows out to have a row for each four hour period. Think of dividing the day into four hour shifts, like this:
Shift | Period |
1 | 12AM-4AM |
2 | 4AM-8AM |
3 | 8AM-12PM |
4 | 12PM-4PM |
5 | 4PM-8PM |
6 | 8PM-12AM |
For example, say I have the following row for a patient encounter:
PatientID | DRGWeight | AdmitDTS | DchDTS |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM |
B | 1.78 | 2/7/24 11:37 AM | 2/12/24 1:49 PM |
What I need is something like the following:
PatientID | DRGWeight | AdmitDTS | DchDTS | CensusDTS | Shift |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/18/24 11:59 AM | 3 |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/18/24 3:59 PM | 4 |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/18/24 7:59 PM | 5 |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/18/24 11:59 PM | 6 |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/19/24 3:59 AM | 1 |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/19/24 7:59 AM | 2 |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/19/24 11:59 AM | 3 |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/19/24 3:59 PM | 4 |
A | 1.81 | 1/18/24 10:17 AM | 1/19/24 6:15 PM | 1/19/24 7:59 PM | 5 |
And of course the same for PatientID B, C, D, etc. Any ideas?
Solved! Go to Solution.
Sh
let
Source = {0..200},
#"Converted to Table" = Table.FromList(Source, Splitter.SplitByNothing(), null, null, ExtraValues.Error),
#"Added Custom" = Table.AddColumn(#"Converted to Table", "CensusDTS", each #datetime(2024,1,18,3,59,0) + #duration(0,[Column1]*4,0,0),type datetime),
#"Added Custom1" = Table.AddColumn(#"Added Custom", "Shift", each Number.Mod([Column1],6)+1,Int64.Type)
in
#"Added Custom1"
Pat
let
Source = Table.FromRows(Json.Document(Binary.Decompress(Binary.FromText("LcsxDoAgDEbhq5DOxPpXFOgGO4l7w/2voVSnl7zkM6NGkbAVrDAKSwrYFTm08a261qU4wz1oRqPuIpc3wtkB9PiBMMSXpupgPg==", BinaryEncoding.Base64), Compression.Deflate)), let _t = ((type nullable text) meta [Serialized.Text = true]) in type table [PatientID = _t, DRGWeight = _t, AdmitDTS = _t, DchDTS = _t]),
#"Changed Type" = Table.TransformColumnTypes(Source,{{"PatientID", type text}, {"DRGWeight", type number}, {"AdmitDTS", type datetime}, {"DchDTS", type datetime}}),
#"Added Custom" = Table.AddColumn(#"Changed Type", "Custom", (k)=> Table.SelectRows(Sh, each k[AdmitDTS]<=[CensusDTS] and [CensusDTS]<=k[DchDTS])),
#"Expanded Custom" = Table.ExpandTableColumn(#"Added Custom", "Custom", {"CensusDTS", "Shift"}, {"CensusDTS", "Shift"})
in
#"Expanded Custom"
This worked for me so I'm going to mark it as a solution but I actually ended up taking a different route. Here's what I did:
1. Create a discharge shift time column
= Table.AddColumn(#"Reordered Columns", "DchShiftTime", each if [DischargeTime] <= #time(3, 59, 0) then #time(3, 59, 0) else if [DischargeTime] <= #time(7, 59, 0) then #time(7, 59, 0) else if [DischargeTime] <= #time(11, 59, 0) then #time(11, 59, 0) else if [DischargeTime] <= #time(15, 59, 0) then #time(15, 59, 0) else if [DischargeTime] <= #time(19, 59, 0) then #time(19, 59, 0) else if [DischargeTime] <= #time(23, 59, 0) then #time(23, 59, 0) else null)
2. Merge that new column with discharge date to create a discharge datetime column. For each encounter this gives me a datetime for the last minute of the four-hour shift.
= Table.AddColumn(#"Added Conditional Column", "DchShiftDTS", each Text.Combine({Text.From([DischargeDT], "en-US"), Text.From([DchShiftTime], "en-US")}, " "), type datetime)
3. Determine the number of minutes between the discharge shift datetime and the admit datetime and make that an integer
= Table.AddColumn(#"Changed Type3", "IPStayMinutes", each (Duration.TotalMinutes( [DchShiftDTS]-[InpatientAdmitDTS])))
= Table.TransformColumnTypes(#"Added Custom",{{"IPStayMinutes", Int64.Type}})
4. Here is the nifty part! List all the datetimes for the admission datetime plus the number of minutes calculated in the previous step (+1 to get that last minute)
= Table.AddColumn(#"Changed Type1", "Minutes", each List.DateTimes( [InpatientAdmitDTS], [IPStayMinutes]+1, #duration(0,0,1,0)))
5. Expand that list to new rows and change the resulting column to type datetime
= Table.ExpandListColumn(#"Added Custom3", "Minutes")
= Table.TransformColumnTypes(#"Expanded Minutes",{{"Minutes", type datetime}})
6. Add a new time column based on the previous so I have just the time for each row
= Table.AddColumn(#"Changed Type2", "CensusTime", each DateTime.Time([Minutes]), type time)
7. Finally, the payoff! Filter the "CensusTime" column to just the six end of shift times, leaving me with one row per census 'moment' for an encounter
= Table.SelectRows(#"Inserted Time3", each [CensusTime] = #time(3, 59, 0) or [CensusTime] = #time(7, 59, 0) or [CensusTime] = #time(11, 59, 0) or [CensusTime] = #time(15, 59, 0) or [CensusTime] = #time(19, 59, 0) or [CensusTime] = #time(23, 59, 0))
The big problem is that it took a long time to load the data. With six years of data I started with 196K rows and ended up with 5.15 million rows. Five million rows isn't that significant but in between, with the calculation and expansion of the minute-by-minute rows, there would have been hundreds of millions of rows to process. If I modify this or do something similar in the future I might calculate by hour rather than minute in increase efficiency at the cost of some accuracy and precision.
Shift = INT(TIMEVALUE(CensusDTS)*6)+1
Thanks. However, that answer assumes I already have the CensusDTS column. How do I get the CensusDTS from the Admit & Discharge DTS columns?
FYI, I noticed this earlier but forgot to mention. In your DAX approach here you aren't returning the correct shift for the last row for patient A. Not sure if that means everything is off. If the discharge datetime is 1/19/2024 6:15 PM, the Census DTS should be 1/19/2024 7:59 PM, shift 5.
Awesome! That looks promising. Do you know of a way to do this in the Query Editor?
Sh
let
Source = {0..200},
#"Converted to Table" = Table.FromList(Source, Splitter.SplitByNothing(), null, null, ExtraValues.Error),
#"Added Custom" = Table.AddColumn(#"Converted to Table", "CensusDTS", each #datetime(2024,1,18,3,59,0) + #duration(0,[Column1]*4,0,0),type datetime),
#"Added Custom1" = Table.AddColumn(#"Added Custom", "Shift", each Number.Mod([Column1],6)+1,Int64.Type)
in
#"Added Custom1"
Pat
let
Source = Table.FromRows(Json.Document(Binary.Decompress(Binary.FromText("LcsxDoAgDEbhq5DOxPpXFOgGO4l7w/2voVSnl7zkM6NGkbAVrDAKSwrYFTm08a261qU4wz1oRqPuIpc3wtkB9PiBMMSXpupgPg==", BinaryEncoding.Base64), Compression.Deflate)), let _t = ((type nullable text) meta [Serialized.Text = true]) in type table [PatientID = _t, DRGWeight = _t, AdmitDTS = _t, DchDTS = _t]),
#"Changed Type" = Table.TransformColumnTypes(Source,{{"PatientID", type text}, {"DRGWeight", type number}, {"AdmitDTS", type datetime}, {"DchDTS", type datetime}}),
#"Added Custom" = Table.AddColumn(#"Changed Type", "Custom", (k)=> Table.SelectRows(Sh, each k[AdmitDTS]<=[CensusDTS] and [CensusDTS]<=k[DchDTS])),
#"Expanded Custom" = Table.ExpandTableColumn(#"Added Custom", "Custom", {"CensusDTS", "Shift"}, {"CensusDTS", "Shift"})
in
#"Expanded Custom"
This worked for me so I'm going to mark it as a solution but I actually ended up taking a different route. Here's what I did:
1. Create a discharge shift time column
= Table.AddColumn(#"Reordered Columns", "DchShiftTime", each if [DischargeTime] <= #time(3, 59, 0) then #time(3, 59, 0) else if [DischargeTime] <= #time(7, 59, 0) then #time(7, 59, 0) else if [DischargeTime] <= #time(11, 59, 0) then #time(11, 59, 0) else if [DischargeTime] <= #time(15, 59, 0) then #time(15, 59, 0) else if [DischargeTime] <= #time(19, 59, 0) then #time(19, 59, 0) else if [DischargeTime] <= #time(23, 59, 0) then #time(23, 59, 0) else null)
2. Merge that new column with discharge date to create a discharge datetime column. For each encounter this gives me a datetime for the last minute of the four-hour shift.
= Table.AddColumn(#"Added Conditional Column", "DchShiftDTS", each Text.Combine({Text.From([DischargeDT], "en-US"), Text.From([DchShiftTime], "en-US")}, " "), type datetime)
3. Determine the number of minutes between the discharge shift datetime and the admit datetime and make that an integer
= Table.AddColumn(#"Changed Type3", "IPStayMinutes", each (Duration.TotalMinutes( [DchShiftDTS]-[InpatientAdmitDTS])))
= Table.TransformColumnTypes(#"Added Custom",{{"IPStayMinutes", Int64.Type}})
4. Here is the nifty part! List all the datetimes for the admission datetime plus the number of minutes calculated in the previous step (+1 to get that last minute)
= Table.AddColumn(#"Changed Type1", "Minutes", each List.DateTimes( [InpatientAdmitDTS], [IPStayMinutes]+1, #duration(0,0,1,0)))
5. Expand that list to new rows and change the resulting column to type datetime
= Table.ExpandListColumn(#"Added Custom3", "Minutes")
= Table.TransformColumnTypes(#"Expanded Minutes",{{"Minutes", type datetime}})
6. Add a new time column based on the previous so I have just the time for each row
= Table.AddColumn(#"Changed Type2", "CensusTime", each DateTime.Time([Minutes]), type time)
7. Finally, the payoff! Filter the "CensusTime" column to just the six end of shift times, leaving me with one row per census 'moment' for an encounter
= Table.SelectRows(#"Inserted Time3", each [CensusTime] = #time(3, 59, 0) or [CensusTime] = #time(7, 59, 0) or [CensusTime] = #time(11, 59, 0) or [CensusTime] = #time(15, 59, 0) or [CensusTime] = #time(19, 59, 0) or [CensusTime] = #time(23, 59, 0))
The big problem is that it took a long time to load the data. With six years of data I started with 196K rows and ended up with 5.15 million rows. Five million rows isn't that significant but in between, with the calculation and expansion of the minute-by-minute rows, there would have been hundreds of millions of rows to process. If I modify this or do something similar in the future I might calculate by hour rather than minute in increase efficiency at the cost of some accuracy and precision.
Read about Table.Buffer and its siblings. Or do this all in DAX.
Thanks so much for all your help!
This is helpful, thanks much. I've come up with an alternative solution but yours is much more elegant. However, I'm not entirely sure what is going on in that M code and would like to understand so that I can incoporate it into my query.
When I try to incorporate this into my query that custom column on the patient encounters table returns as a function
and that function asks for a parameter
If you have the time could you explain what this is doing and how I might fix it?
1. "(k)" - that's a long story. When I started to learn about context in Power Query I saw an example that used k, and then I used it too, and have used it ever since. It means "current context" and you can use whatever string you like.
2. yeah, whatever range you need.
Note that your formula contains an extra "each" - you can either have the "each" (but nesting them gets nasty very quickly) or the explicit reference via (k)=> , but not both. Remove the first "each" and you get a table instead of a function.
If you want to learn Power Query, read Ben Gribaudo's primer.
Ach! Stupid cut and paste error!
Done for the day but I will try that out tomorrow. Thanks so much!
Could you share a straightforward solution? All I could find is a way to expand date ranges to create rows for each day between the admit & discharge dates. I haven't seen anything on how to break out ranges by hour or grouped hours.
That seems to be straightforward . What have you tried and where are you stuck?
Check out the October 2024 Power BI update to learn about new features.
Learn from experts, get hands-on experience, and win awesome prizes.
User | Count |
---|---|
113 | |
107 | |
106 | |
91 | |
67 |
User | Count |
---|---|
162 | |
133 | |
132 | |
93 | |
91 |