sep_epic Power BI documentation & findings

[staceypark]

discussion from 9/21 huddle with @jeffhoerle

Where will the clinic selection feature be located?

• So far, the only way for the clinic selection and the resulting visualizations can be nested within one location is to use the solution Ash already created in his demo
• Otherwise, Jeff recommends we have the clinic selection in a separate place.
• The issue is that the clinic selection cannot be "cached," in multiple ways:
  • You cannot "build up" a clinic selection list meaning if you want to filter by provider first, add the results to a list, then filter by clinic location, and add that result to list, there's no way for that initial "provider filter" list to be held.
  • We also cannot save a set of clinics to go back to in the way we have the "get previous clinic list" for running patient level data or running the team data table.

[staceypark]

We can get data by either importing the data ahead of time or implementing a direct connection. Importing the data is restricted to 10gb. Importing data ahead and working with it is much faster than doing a direct connection. VA already has documentation on how to setup the row level security permissioning for this method.

Due to the 10gb restriction, we may need to have 18 separate visn data models instead of one national data model. But, @jeffhoerle is hopeful so far that we can put it all together in one model.

Key value proposition: Important for our clinicians/end users to be able to be in the "driver seat" of creating their clinic selections/data sets and querying them aka transform the data themselves.

Need to update PTSD_OMHO "staging data base" then have PBI refresh its data based on PTSD_OMHO. PTSD_OMHO needs to be updated by using SQL script. In the past, this was the splashSourceIncrementalUpdate. We need an identical process in PBI. We update on the monthly schedule, because otherwise the data "drops" drastically since all the data for the month is not in it and we've learned not to do this with our user base.

PBI updates can be scheduled (this is the current propagation step), but we do not currently know of a way for PTSD_OMHO to be done automatically.

Replicating data & viz tabs, with modified queries that give the same results. Jeff is doing this one station at a time, for a smaller set of data to work with.

[lijenn]

Discussed 10/20 Support Workgroup:

@jeffhoerle Found 2 bugs (?):

1) CC Params - True Missed Appt %
Calculation indicates if the appointment was cancelled on the same day vs than when the appt was scheduled.
E.g. If the appt was cancelled 2 hours before the appointment time, it will show up as a missed appointment.

@lzim Can you confirm if cancellations happening between midnight and the time of the appointment occurring should not show up as missed appointments?

• NOT A BUG - True Missed Appt is an intentional business rule, keep it the same as it is.
• CC to potential "Bugs"
  DECISION "Not a Bug:" True missed appointment % - Yes, cancel on the same day of the appointment is "true" because the provider is unlikely to be able to use the slot for another Veteran.

This is a VHA definition (not unique to MTL).
3 Date Times associated with Appts:

• Made
• Occurred
• Cancelled

Double checked on the definition (from the (i) in the sim ui):

The proportion of care coordination appointments with this team that no-showed or were cancelled after the appointment was supposed to have happened. (pct)

Out of 4.8 mil no-shows, 900,000 were cancelled on the same day but before the appointment.

2) CC Params - Calculates # of visits by/per week. For schedules for a clinic, there were only 30 weeks that had data/visits.

@jeffhoerle Thinks that the median is using only using the weeks that contain data versus 52 weeks/a year.
If this is happening in CC, then we may see the same issue in the MM Params tabs.

@lzim Should we be using 52 weeks as the median rather than weeks that contain data?

• DECISION: "Plug in" zeros in Power BI, then calculate.
• We carefully considered the way zeros are handled in the param tabs now.
• Right now the median is being calculated only for the observations we have.
• So, if calculations are being applied on 30 weeks of observations, not 52 weeks of the year.
• But, if zeros are meaningful, then they make up of the distribution of observations of that service and should be included in the calculations.

---

@dkngenda Yes, we can we use the ClinicSelections a team makes to define their EBP reach numerators/denominators?

• We could export the Operations selections in Excel from PTSD_OMHO on a01 (operations server).
• David needs to merge those selections with the Research queries on ORD workgroups on b02 (research server).
• Will be possible using LocationSID from the Location dimension table.

[mnallajerla]

Closing this issue. Referencing it in PBI Documentation (#2310) & Tracking cards (#2399)