Uploaded image for project: 'eCQM Issue Tracker'
  1. eCQM Issue Tracker
  2. CQM-1761

CMS-153 Denom Exclusion logic doesn't match intent of the exclusion

    • Icon: EC eCQMs - Eligible Clinicians EC eCQMs - Eligible Clinicians
    • Resolution: Answered
    • Icon: Minor Minor
    • Measure
    • None
    • Hide
      Thank you for your question. According to QDMv4.2, the Starts After End Of (SAE) timing operator asserts that the start date/time of an event (i.e., X-Ray order) on the left-hand side (LHS) is temporally after the end date/time of an event on the right-hand side (RHS) (Lab test order). The additional comparison operator (<= 7 day(s)) limits the maximum timeframe that the LHS event can start in relationship to the end of the RHS event. However, the LHS event may take place on the same day as the RHS event, but is must start after the end of the RHS event. In this scenario, since both events happen on the same day, the level of granularity need to determine if 'A' SAE 'B' would be time (hour and/or minute if necessary).
       
       
      We hope this addresses your question.
      Show
      Thank you for your question. According to QDMv4.2, the Starts After End Of (SAE) timing operator asserts that the start date/time of an event (i.e., X-Ray order) on the left-hand side (LHS) is temporally after the end date/time of an event on the right-hand side (RHS) (Lab test order). The additional comparison operator (<= 7 day(s)) limits the maximum timeframe that the LHS event can start in relationship to the end of the RHS event. However, the LHS event may take place on the same day as the RHS event, but is must start after the end of the RHS event. In this scenario, since both events happen on the same day, the level of granularity need to determine if 'A' SAE 'B' would be time (hour and/or minute if necessary).     We hope this addresses your question.
    • CMS153v4/NQF0033
    • Measure logic does not match intent

      The logic for the denominator exclusion is as follows:

      Denominator Exclusions =
      • OR:
      o AND: "Occurrence A of Laboratory Test, Order: Pregnancy Test" during "Measurement Period"
      o AND: Union of:
       "Medication, Order: Isotretinoin"
       "Diagnostic Study, Order: X-Ray Study (all inclusive)"
       <= 7 day(s) starts after end of "Occurrence A of Laboratory Test, Order: Pregnancy Test"

      Per Appendix C of the logic guidance, the granularity of the lookback indicates the gruanularoty to be used when computing that portion of the measure. (A lookback of <=7 days indicates we should compute to the granularity of days.) Since the logic for the Laboratory Test, Order data element is ‘start after end of’ and not concurrent, this means that a lab and a med/x-ray on the same day would not count.

      That does not seem in line with the intent of the measure, which likely envisions a workflow where the provider would order an result a pregnancy test, and then order a medication/x-ray upon a negative test result.

          [CQM-1761] CMS-153 Denom Exclusion logic doesn't match intent of the exclusion

          Rethinking this string of comments. Is the QDM description in Appendix B.2 really incorrect, or is the logic missing critical information. E.g., if the desire is that the order for medication happens at least 1 minute after the order for the pregnancy test, the logic should state that fact (consistent with Appendix B2 as written). Though more complex, it could be written to include another timing requirement:
          > 1 minute(s) starts after end of "Occurrence A of Laboratory Test, Order: Pregnancy Test".
          Please confirm with the measure developer, but this seems to be a more explicit set of logic rather than expecting an implementer to read through multiple examples in the QDM specification to interpret the intent correctly.

          Floyd Eisenberg added a comment - Rethinking this string of comments. Is the QDM description in Appendix B.2 really incorrect, or is the logic missing critical information. E.g., if the desire is that the order for medication happens at least 1 minute after the order for the pregnancy test, the logic should state that fact (consistent with Appendix B2 as written). Though more complex, it could be written to include another timing requirement: > 1 minute(s) starts after end of "Occurrence A of Laboratory Test, Order: Pregnancy Test". Please confirm with the measure developer, but this seems to be a more explicit set of logic rather than expecting an implementer to read through multiple examples in the QDM specification to interpret the intent correctly.

          I agree the interpretation from David Czulada is partially correct. In this example, the intent is to indicate seconds or minutes after the test is performed through the end of 7 days after the test (presumably to the end of the 7th day, not to 7 x 1440 minutes). The QDM section B.2. is explicit to indicate what to do if "no calculation unit is defined." In this case, the logic needs to be more explicit if there is to be a timing difference of 1 minute. I agree further clarity with examples in the QDM specification will help. However, I also think the measure developer should consider the intent. The reason for the exclusion is because a pregnancy test is performed for the purpose of determining if the pregnancy is performed solely for evaluating if isoretinon could be used or an Xray performed. While the QDM spec could be more expressive, the timing relationship is used rather than looking for the reason for the lab test is to determine a risk for treatment. Clinical workflow may not provide such information in the real world so the timing relationship is a surrogate for the information actually desired. I might ask if the result of the pregnancy test should be known before ordering the isoretinon or the Xray and, therefore, timing to one second or one minute may not be ideal. In the grand scheme of things, the "starts after or concurrent" is probably as effective as "starts after the end of". The actual difference is minute especially since this is a surrogate logic statement. One might question the cost of calculating such precision and consider how many extra exclusions would be captured if the logic were set to a day as described in the QDM.

          Floyd Eisenberg added a comment - I agree the interpretation from David Czulada is partially correct. In this example, the intent is to indicate seconds or minutes after the test is performed through the end of 7 days after the test (presumably to the end of the 7th day, not to 7 x 1440 minutes). The QDM section B.2. is explicit to indicate what to do if "no calculation unit is defined." In this case, the logic needs to be more explicit if there is to be a timing difference of 1 minute. I agree further clarity with examples in the QDM specification will help. However, I also think the measure developer should consider the intent. The reason for the exclusion is because a pregnancy test is performed for the purpose of determining if the pregnancy is performed solely for evaluating if isoretinon could be used or an Xray performed. While the QDM spec could be more expressive, the timing relationship is used rather than looking for the reason for the lab test is to determine a risk for treatment. Clinical workflow may not provide such information in the real world so the timing relationship is a surrogate for the information actually desired. I might ask if the result of the pregnancy test should be known before ordering the isoretinon or the Xray and, therefore, timing to one second or one minute may not be ideal. In the grand scheme of things, the "starts after or concurrent" is probably as effective as "starts after the end of". The actual difference is minute especially since this is a surrogate logic statement. One might question the cost of calculating such precision and consider how many extra exclusions would be captured if the logic were set to a day as described in the QDM.

          David Czulada added a comment -

          Unfortunately, the QDM documentation on this topic is as clear as mud.

          I believe the intent of QDM for this particular example would be:

          • Minutes is used first to determine that the relationship (starts after end) is true,
          • And then the delta is calculated in the given unit (days) to see if the comparison is true.

          If this is in-fact the intent of QDM, the spec needs to be updated to be more explicit. The spec does not make assertions, or show any examples, where an unstated time unit of measurement is used for the relationship comparison and the stated date unit is used to calcuate the duration comparison. However, the fourth example does specifically state that the unit specified is intended for the calculation of “duration" between the two date/time stamps.

          If the above is not the intent of QDM, the measure should be updated to "starts after or concurrent” in places where units less granular than minute(s) are used.

          QDM is consistent is specifying that all times are in fact times, and not solely dates. In every instance where time comparisons are used they are specified as "date/time”.

          -Dave Czulada

          David Czulada added a comment - Unfortunately, the QDM documentation on this topic is as clear as mud. I believe the intent of QDM for this particular example would be: Minutes is used first to determine that the relationship (starts after end) is true, And then the delta is calculated in the given unit (days) to see if the comparison is true. If this is in-fact the intent of QDM, the spec needs to be updated to be more explicit. The spec does not make assertions, or show any examples, where an unstated time unit of measurement is used for the relationship comparison and the stated date unit is used to calcuate the duration comparison. However, the fourth example does specifically state that the unit specified is intended for the calculation of “duration" between the two date/time stamps. If the above is not the intent of QDM, the measure should be updated to "starts after or concurrent” in places where units less granular than minute(s) are used. QDM is consistent is specifying that all times are in fact times, and not solely dates. In every instance where time comparisons are used they are specified as "date/time” . -Dave Czulada

          I don’t agree with your interpretation of Appendix B when it comes to handling temporal operators. B.2 specifically says that if no time unit is given, you use minutes. That implies that the temporal operator must also respect the given time units.
          See the examples from Appendix B.2:
          Example 1: Granularity to Seconds
          A: 2012-01-01 11:00:01
          B: 2012-01-01 11:00:02 A
          starts before B = TRUE

          Example 2: Granularity to Minutes
          A: 2012-01-01 11:00
          B: 2012-01-01 11:00
          A starts before B = FALSE

          This implies that the result starts before operator will change based on the time unit you are using. The examples you give seem to imply that the temporal operator should always be based on minutes, which conflicts with Appendix B.

          I believe the line in question in CMS-153 should be changed from this:
          <= 7 day(s) starts after end of "Occurrence A of Laboratory Test, Order: Pregnancy Test"
          To this:
          <= 7 day(s) starts after or concurrent with end of "Occurrence A of Laboratory Test, Order: Pregnancy Test"
          This would clear up any confusion with the measure logic.

          Anthony Galba (Inactive) added a comment - I don’t agree with your interpretation of Appendix B when it comes to handling temporal operators. B.2 specifically says that if no time unit is given, you use minutes. That implies that the temporal operator must also respect the given time units. See the examples from Appendix B.2: Example 1: Granularity to Seconds A: 2012-01-01 11:00:01 B: 2012-01-01 11:00:02 A starts before B = TRUE Example 2: Granularity to Minutes A: 2012-01-01 11:00 B: 2012-01-01 11:00 A starts before B = FALSE This implies that the result starts before operator will change based on the time unit you are using. The examples you give seem to imply that the temporal operator should always be based on minutes, which conflicts with Appendix B. I believe the line in question in CMS-153 should be changed from this: <= 7 day(s) starts after end of "Occurrence A of Laboratory Test, Order: Pregnancy Test" To this: <= 7 day(s) starts after or concurrent with end of "Occurrence A of Laboratory Test, Order: Pregnancy Test" This would clear up any confusion with the measure logic.

          Using the 'starts after end of' timing restricts the logic to ensure that the event on the left-hand side (LHS) starts temporally AFTER the end of the event on the right-hand side (RHS), and not concurrently. The further restriction of '<= 7 day(s)' then evaluates whether the LHS falls within this range. Per Appendix B.1, two events that occur on the same day will result in a duration of 0 days. Zero days falls within the '<= 7 day(s)' criteria and the logic is met. (Page 31 of the QDM explains these steps.)

          A <= 7 day(s) starts after end of B

          Example 1
          Event A starts at 02/09/2015 11:00
          Event B ends at 02/09/2015 10:00

          • Does A start after the end of B? Yes
          • Is A <= 7 day(s) after B? Yes - duration is 0 days
          • Criteria is met

          Example 2
          Event A starts at 02/09/2015 11:00
          Event B ends at 02/09/2015 12:00

          • Does A start after the end of B? No
          • Criteria is not met

          Example 3
          Event A starts at 02/09/2015 11:00
          Event B ends at 02/09/2015 11:00

          • Does A start after the end of B? No
          • Criteria is not met

          Mathematica EC eCQM Team added a comment - ​ Using the 'starts after end of' timing restricts the logic to ensure that the event on the left-hand side (LHS) starts temporally AFTER the end of the event on the right-hand side (RHS), and not concurrently. The further restriction of '<= 7 day(s)' then evaluates whether the LHS falls within this range. Per Appendix B.1, two events that occur on the same day will result in a duration of 0 days. Zero days falls within the '<= 7 day(s)' criteria and the logic is met. (Page 31 of the QDM explains these steps.) A <= 7 day(s) starts after end of B Example 1 Event A starts at 02/09/2015 11:00 Event B ends at 02/09/2015 10:00 Does A start after the end of B? Yes Is A <= 7 day(s) after B? Yes - duration is 0 days Criteria is met Example 2 Event A starts at 02/09/2015 11:00 Event B ends at 02/09/2015 12:00 Does A start after the end of B? No Criteria is not met Example 3 Event A starts at 02/09/2015 11:00 Event B ends at 02/09/2015 11:00 Does A start after the end of B? No Criteria is not met

          Your solution contradicts that last portion of Appendix B.2 of the QDM guide (second half of page 76), which says that if a unit is specified in the criterion, that unit should be used as the default unit for calculation. Since the unit of days is defined in ">= 7 days", we should be using a granularity of days.

          If these data elements are calculated to the granularity of a day, this means labs that occur on the same day as the med or xray would appear to happen concurrently. This does not meet the "starts after end" requirement.

          Link to guide: https://ecqi.healthit.gov/system/files/qdm/qdm_4_1_2.pdf

          Kelli Klippenstein (Inactive) added a comment - Your solution contradicts that last portion of Appendix B.2 of the QDM guide (second half of page 76), which says that if a unit is specified in the criterion, that unit should be used as the default unit for calculation. Since the unit of days is defined in ">= 7 days", we should be using a granularity of days. If these data elements are calculated to the granularity of a day, this means labs that occur on the same day as the med or xray would appear to happen concurrently. This does not meet the "starts after end" requirement. Link to guide: https://ecqi.healthit.gov/system/files/qdm/qdm_4_1_2.pdf

            edave Mathematica EC eCQM Team
            KelliKlippenstein Kelli Klippenstein (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            7 Start watching this issue

              Created:
              Updated:
              Resolved:
              Solution Posted On:
              Comment Posted On: