Peer Audit

From CCMDB Wiki
Projects
Active?: legacy
Program: CC and Med
Requestor: internal



Legacy only, see below for details





The CCMDB Peer Audit was a real-time audit to quantify the variability in our data collection.

Purpose of an Audit

An audit is a scientific approach to obtaining a quantitative measure of the quality of the data data we collect. By quantity we mean "accuracy" and ease of "reproducibility" (precision) of the numerous elements we collect. Once we have a measure, the next phase is to collaborate as a team and identify any factors that are affecting reproducibility, and work together and make a plan to improve data quality........one step at a time. Examples of factors that may be affecting quality:

  • the collection process
  • source of information
  • documentation
  • guidelines
  • equipment
  • human factors: staffing, vacation, sick time etc.
  • data structure
  • etc......

GO TO: List of Factor affecting data quality

What the Peer Audit is and is not

The Peer Audit is not meant identify "wrong" data or to single out a specific collector who is doing something bad. We are trying to quantify the precision rather than the accuracy (reproducibility) of our individual data elements. A lot of us have hunches about where there are problems, this audit is to give us objective indicators.

  • In the analysis of the audit information, a site/unit that shows a low percentage in reproducibility of data elements, is not an indication that the collector at the site/unit is collecting poorly, nor does it indicate that the peer auditor for the site/unit is collecting poorly either. The audit analysis doesn't distinguish between who is better or worse, it only shows us is that reproducibility is not easy achieved. It provide us with direction as to where we need to focus most to find factors and make plans to continue to raise the quality of the data we collect.--TOstryzniuk 22:30, 29 September 2010 (CDT)

Goals and follow-ups to the peer audit

Julie will do comparative analysis between the audit data and the database data by element. The proportion of dissimilarity of values will be calculated over time and presented in a statistical control chart. Values found outside the prescribed or predetermined control limits be investigated.JMojica 10:16, 2 December 2009 (CST)

Once we have identified the elements which showed large discrepancies or variability, we will identify the reasons why, suggest changes to reduce the variation in the data, implement the changes and re-assess again to see if the change results in improvement in the quality of the data. This will largely happen ad-hoc using the wiki. We may also come back to you personally to find out why there are discrepancies, but this is to find the reasons and fix the underlying problem, not to criticize individuals.

We will also use our findings to correct data, but this is for a very small subset of our database and just coincidental.

Start Date

  • pilot by collectors - START TEST: Nov 17.09.TOstryzniuk 13:00, 17 November 2009 (CST)
    • sites to start Nov 17.09:
      • HSC SICU & MICU (Joyce and Lois)
      • HSC Med all wards (Gail, Con, Pat, Marie)
      • VIC Med (Tara and Shirley start Nov 18)
      • GRA Med all wards (Steph and Sheila)
      • STB CICU - Laura K
    • Start Week of Nov 23.09 on Thursday Nov 26.09
      • STB -all ICUs- MICU CICU & CCU (Kym and Darlene & Laura)
      • STB Med all wards (Deb, Elaine, Galye)
      • VIC Med - all wards (Wendy, Tara, Shirley)

Stop Date

  • June 18.10 - stopped. Will resume later in the year.
  • Please complete an audit for this week and also continue to follow and complete any audits to discharge that you have already started on your laptop/PDA.
  • Thank you everyone for the good work with the Peer audit!
  • The program is currently in the process of analyzing the information and this is now in the prelimary stages. We have 283 files to date and a few more that will still come in. The information will be shared with the Collection Team once the analysis is complete and a report is written. --TOstryzniuk 17:56, 17 June 2010 (CDT)

Restart Date

Processes and Procedures

Data Collection

  • Do not compare notes on the patients you are auditing as this would prevent us from getting an accurate idea how consistent our data is.
  • Remember, we want the audited profile to be unbiased. Don't audit a patient you have previously collected data on.
  • When sending in your audit profiles you must attached the initials of the person who actually did the audit, not the person who sent the audit profile.

Vacation/Sick - "covering for" or "going on" any type of leave

  • If you are COVERING for vacation/sick time on a ward that you have been assigned to audit on, for the week that you are covering:
    • A. do not do an audit.
    • B. check if the previous week's audit profile is already completed by the collector who has gone on vacation/sick time.
      • if YES, then keep your previous week's audit profile.
      • if NO, then drop your previous week's audit profile.
  • If you are the person that is GOING on vacation/sick leave and:
    • A. your audit profile has not been completed:
      • you will complete the audit when you return.
      • the person covering you while you are away, will not continue your audit profile.
    • B. your audit profile has been completed then:
      • the person covering for you must sent the file using your initials in the csv sent file.
Discussion
    • Basically-if you are doing vacation relief, you do not do audits for the person you are covering. When the person who was on vacation returns, they are responsible for finishing up their audit patients ( including pulling the chart from med records if the patient was discharged).--CMarks 18:37, 29 January 2010 (CST)
    • If the relief required is longer than 2 weeks (such as for sick leaves) clarify with Trish/Julie as to what should be done.--CMarks 18:42, 29 January 2010 (CST)

What is INCLUDED for collection

All data elements for patients

What is EXCLUDED for collection

ICU

  • TISS
  • pharmacy and lab tests for now. NOTE: In Jan or Feb 2010 when reduced lab and pharm list is implemented then it will be included in Peer audit.--TOstryzniuk 18:23, 2 December 2009 (CST)

ICU & MED

How it is collected -Peer Audit Partners

Every data collector (except community ICU) has an audit ward assigned in Peer Audit Partners.

One patient profile per week must be audited.

Starting on Thursday morning, the first patient admitted or transferred to the audit ward whom you have no prior information and whose chart you have seen the first time will be an audit patient and will be followed as if he or she were a patient admitted to the regular ward of that collector.

The serial numbers to be used for audit patients will be 111 to 140 (if you need higher, you will not be able to send your audit data, contact Trish or Tina). If a patient is not discharged by next Thursday, use the next number, e.g. 112. Re-use earlier numbers once they become available, i.e. once patient 111 is sent and deleted, use the number for the next audit patient.

How to Send

On the next send day a separate batch is sent for peer audit patients discharged during the previous week. To do this, make sure you either first delete your regular sent patients, or that you uncheck their Final Check checkbox.

Peer Audit output batch labelling

  • The records are sent as a separate batch with the following parameters
  • Batch label manual type in the letter "a"
  • Type in the initials of the person who actually did the audit (if default setting is not your intitals).
  • As of Dec 17.09 - DO NOT manually enter a DATE. Access (CCMDB) Version 1.9852 will now automatically put a date and time when you send peer audit file.
    • An EXAMPLE of audit batch label that is sent to the Regional Server: M_a_HSC_H4_2009_12_17_15-42-10_TB.cvs

  • Every audit batch is sent and starts with the labelled "a". What distinguishes one "a" batch from another is the date and initials included on the sent file.
  • Pagasa will move the file off regional server each week.


Data Sending

If the batch is labelled with an "a" (i.e. for audit patients) then CCMDB.mdb will send task and temp information to the following alternative audit locations:

Central Office Data Processing

Central Office- Data Analysis

  • Statistician will retrieve file from X:\\Med_CCMED\CCMDB|Peer Audit\backup location
  • match every field one on one, give count of good vs bad and degree of difference
  • do a pair-analysis for dxs, Admit 1 specific, and others regardless of diagnosis number

Follow-Up

  • Post accuracy scores to this article
  • further investigate causes for differences

QUESTIONS from Collectors

Long Stay Patients

  • Also, what happens if the patients we are auditing have prolonged lengths of stay (ie 6 mos to a year)?
    • You will continue to follow patients until discharged or moved from your audit ward. Julie can decide what she wants to do. Lets see how many end up being in this group. She will be able to see by the pending reports and will advise further.

EMIP/OVER pts

  • Also, Wendy and I now do S3 as well and all the EMIP/OVER patients and split the workload between our laptops. Do you want us to follow any of the S3 patients as most of these patients have been transferred from other wards and have been medically stable and are usually waiting placement?

Audit Transfers as well?

  • Do you want us to follow patients that are transferred between wards or just new admits?
    • Both. If the first patient that arrives on your audit ward on Thursday is a patient that was transferred over from your own collection ward do not audit that patient. Select the "next" patient that arrived on your audit ward on Thursday. The main idea of a peer audit is to repeat data collection for a patient by two totally different data collectors.--TOstryzniuk 13:29, 19 November 2009 (CST)

Data Integrity Checks in Access

  • query "check_peer_audit_audit_serial_reg_send" to ensure no peer audit serial numbers (111-140) get sent as regular data (i.e. batch <> "a")
  • query "check_peer_audit_invalid_serial_audited" to ensure that no non peer audit serial numbers get sent as peer audit files.(i.e. batch = "a")
  • query "Pending_Julie" and "Pending_Pagasa" to exclude peer audit serial numbers
  • query "Send_Tasks_Master" and "Send_Tasks" to exclude peer audit serial numbers
  • query "Send_Tmp_audit" to only send peer audit serial numbers
  • query "Send_Tmp" to exclude peer audit serial numbers
  • queries "Send_Tasks_audit" and "Send_Tasks_Master_audit" to deal with audit sending
  • table "peer_audit_partners" and query "check_peer_audit_location_bad" to limit which patients can be sent as audits
    • Checks above are in Version 1.9852. Rolled out Dec 16.09--TOstryzniuk 21:25, 16 December 2009 (CST)