Policy for Aggregate Result Minimum Cell Size
Closed this issue · 2 comments
Abel Kho, leader of the Data Privacy and Security group has gained consensus that the cell size should be 11 or greater for alignment with CMS. Goal is to have PCORnet policy in place to support initial analyses and queries using the DRN for the obesity surveys.
Policy passed and revised guidance discussed with coordinating center.
Had a minor clarification regarding query cell size and the interpretation of CMS via RESDAC. Copying that thread on to this ticket as a reference.
Russ Waitman (Client) Posted On: 27 February 2017 03:45 PM
Dear RESDAC,
Thank you for all your support for our RESDAC driven data requests for CMS data to support PCORnet.
We’ve had a related question come up for PCORnet which may also arise as we report out from our CMS cohort characterization work and as we’re both using CMS data and helping direct the PCORnet Data Committee, we were hopeful you might help us align PCORnet to be supportive of CMS requirements.
Imagine we have a analysis program that is exploring injury in relation to a chronic disease diagnosis.
Q1. “How many patients are over 65 years old at one healthsystem in a year?”
- 100000 might be the answer with 500000 excluded who were under 65 years old.
Q2. Then the analysis code asks “How many of those patients have Alzheimer’s disease at one healthsystem in a year”
- the answer reported in tabular form is that 1000 meet this inclusion criteria and 99000 are excluded.
Q3. Then the analysis asks “How many of those with Alzheimer’s from Q2 had a wrist fracture following their Alzheimer’s diagnosis?”=
- the answer reported in tabular form is NULL patients meet this inclusion criteria because the cell size is less than 11 but it also reports that 995 patients were excluded.
- So, strictly speaking the cell size of 5 wasn’t reported but through subtraction it’s quite obvious that there were only 5 patients that met criteria.
If I was conducting this analysis using CMS data would that table be in violation of CMS’ cell size policy or be ok?
Sincerely,
Russ Waitman, PhD
Response from RESDAC:
From: Data.pcornet [mailto:mailman-bounces@listserv.kumc.edu] On Behalf Of Matt McFalls
Sent: Tuesday, February 28, 2017 7:46 AM
To: Russ Waitman
Cc: data.pcornet@listserv.kumc.edu; Maren Wennberg; data.pcornet-owner@listserv.kumc.edu
Subject: [#BMK-965-97006]: minimum cell size question
-New staff reply-
Hi Russ,
From your example, since the denominator of 1000 is reported, the 995 excluded would imply a cell size of 5 remaining, and we would interpret this as violating the cell size policy. This example would need to display the data in an alternate way that does not reveal a count < 11.
I hope this helps, let us know if you have any questions.
Matt
Matt McFalls, MPH
Technical Advisor
Research Data Assistance Center (ResDAC)
888-973-7322 (Toll Free)
www.resdac.org
Response by Lesley Curtis with the PCORnet coordinating center:
From: Lesley Curtis [mailto:lesley.curtis@duke.edu]
Sent: Tuesday, February 28, 2017 9:29 AM
To: Russ Waitman; 'resdac@umn.edu'; 'Sturtevant, Jessica'
Cc: data.pcornet@listserv.kumc.edu; Maren Wennberg; data.pcornet-owner@listserv.kumc.edu; Carton, Thomas W (tcarton@tulane.edu); 'Jasmin Phua'; Kho, Abel (Abel.Kho@nm.org)
Subject: Re: [#BMK-965-97006]: minimum cell size question
Thanks for bringing this to our attention again, Russ. We understand the issue and propose that PMP v2.0 (scheduled for a March 2017 release) include a hard-coded minimum cell size of <11 for the attrition table. The program will maintain user-specified low cell count threshold (default <11) for all other masking. We’ll update the low cell count guidance to reflect this decision.
Note that there will be a few queries distributed before PMP v2.0 is available.
Thanks,
Lesley