Missing Data on Pain Items in MDS 3.0 Nursing Home Assessment

Missing Data on Pain Items in MDS 3.0 Nursing Home Assessment

Current Version Date: 
08/03/2017
Purpose: 

This article has three goals: (1) to describe missing patterns on pain variables; (2) to describe the difference between real missing and skip patterns; (3) to describe which assessments should be used for calculating pain measures. This information is most relevant for researchers who work on either creating their own pain measures or constructing CMS quality measures.

The new MDS 3.0 requires nursing home staff to interview residents regarding health conditions, such as pain, mood and cognitive function through direct interview, and assumes that residents’ voice is important. Thus, missing due to non-response is an important issue to consider.

(1)  Description of Pain Items

MDS 3.0 measures the following pain variables: For more detailed description, please refer to the section J of the MDS 3.0 RAI Manual (https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/NursinghomeQualityInits/MDS30RAIManual.html ).

Pain Items:

  • J0100: Pain Management (5-Day Look Back)
  • J0200: Should Pain Assessment Interview Be Conducted?
  • J0300: Pain Presence
  • J0400: Pain Frequency
  • J0500: Pain Effect on Function
  • J0600: Pain Intensity (A: Numeric Rating Scale, B: Verbal Descriptor Scale)
  • J0700: Should the Staff Assessment for Pain Be Conducted?
  • J0800: Indicators of Pain (Staff-Report)        
  • J0850: Frequency of Indicator of Pain or Possible Pain (5-Day Look Back)
  • Other Items Related to Skip Pattern:
  • J1100: Shortness of Breath (dyspnea)

While calculating the pain measure, there are three types of missing that may lead to exclusions, blank “ ”, dash “-” and “9” or “99”. Missing due to a skip patterns occurs when scoring on one resident’s item governs how scoring is completed for one or more additional items. While a skip pattern is encountered, the following item would be left blank, “ ”. When the staff are unable to determine the response to an item, this item will be coded as dash, “-”. When a resident is unable to answer a question, this item will be coded as “9” or “99”.

 

Table 1. Missing Patterns on Pain Variables

Summary of Missing by Pain Item (N=67,266)

Pain Item

Real Missing:

Blank for All J0200-J1100 “ ”

Blank Due to Skip “ ”

Dash “-”

Unable to Answer “9”

Unable to Answer “99”

J0100A

 

 

1.40%

 

 

J0100B

 

 

1.44%

 

 

J0100C

 

 

2.07%

 

 

J0200

0.11%

 

12.72%

 

 

J0300

0.11%

10.07%

16.19%

2.03%

 

J0400

0.11%

(10.07+33.95)%

16.30%

0.25%

 

J0500A

0.11%

(10.07+33.95)%

16.38%

0.27%

 

J0500B

0.11%

(10.07+33.95)%

16.39%

0.32%

 

J0600A

0.11%

(10.07+33.95+7.80)%

16.78%

 

0.50%

J0600B

0.11%

(10.07+33.95+23.65)%

17.72%

0.21%

 

J0700

0.11%

(10.07+33.95)%

14.39%

 

 

J0800A

0.11%

71.25%

6.49%

 

 

J0800B

0.11%

71.25%

6.42%

 

 

J0800C

0.11%

71.25%

6.48%

 

 

J0800D

0.11%

71.25%

6.48%

 

 

J0800Z

0.11%

71.25%

6.45%

 

 

J0850

0.11%

(71.25+12.69)%

6.61%

 

 

 

(2) The Difference between Real Missing and Skip Patterns

It is important to distinguish reasons for missing, which we group into “real missing” and skip patterns below, especially if researchers are calculating their own pain measures vs. calculating CMS Quality Measures, where you may have different criteria in terms of selecting assessments.

Real Missing

We summarized the coding rules for missing stated in the MDS 3.0 QM User’s Manual (https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/NursingHomeQualityInits/NHQIQualityMeasures.html) in the section (1). 

Some reports treat all blanks as correct skip patterns based on the coding rules. However, based on our analyses of the January-June 2011 data, it is worth noting that some individuals (albeit small) have all pain items left blank (0.11%) except J0100 which does not include a skip pattern. Thus, we believe it is important to distinguish valid skips from real missing. For those 0.11% residents, pain items are “real” missing, because they cannot provide any information regarding pain condition. Thus, real missing includes those with “-“, 9, 99, and the 0.11% with all pain items blank.

Skip Patterns Represented by Blank Cell

Figure 1 shows a decision tree starting from J0200, “Should Pain Assessment Interview Be Conducted?” (see Figure 1 graphic in Additional Resources)

By answering “yes”, the staff will continue to J0300, “Pain Presence.” By answering “no” in J0200 for a resident, the staff skips to and completes J0800, “Indicators of Pain or Possible Pain”, and J0300-J0700 will be left blank for this resident (10.07%).

  • At J0300, for 33.95% of residents, the staff skipped to J0800 or J1100, and J0400-J0700 were left blank for these residents.
  • For J0600, “Pain Intensity”, there are two different scales, “J0600A: Numeric Rating Scale” and “J0600B: Verbal Descriptor Scale”. These two scales are exclusive. The 7.80% residents whose J0600A were left blank have values in J0600B. The 23.65% residents whose J0600B were left blank have values in J0600A. If the staff didn’t skip to J0800, J0800 and J0850 will be left blank (71.25%). Since J0850 is based on J0800, and the staff skipped from J0800Z to J1100 for 12.69% of residents, J0850 will be left blank for these 12.69% residents.

Various Meanings of the Dash

Some dashes might be a skip pattern. Among those residents who have J0200 coded as a dash, some of them still have a dash for J0800, and some will have a meaningful value in J0800, which means that a skip pattern from J0200 to J0800 is encountered for some of these residents.

For J0300, the staff could either skip to J0800/J1100 or continue to J0400, but some residents with a dash for J0300 have a meaningful value in J0700 and J0800. They have a value in J0800 due to the skip pattern. They have a value of“1” in J0700, because those have dashes in J0300 will still have dashes in J0400, and J0700 is based on J0400 (J0700 = “1”/“Yes” if J0400 = “9”/ “Unable to Answer”), which means that dashes in J0400 might be “Unable to Answer.” In sum, a dash could represent “staff is unable to determine the response”, “resident is unable to answer” or a skip pattern.

 

(3) Which Assessment Should Be Used 

To calculate the MDS 3.0 pain measure suggested by CMS—the percent of residents who self-report moderate to severe pain, the QM Manual suggests only using target assessments, because target assessments represent the residents’ status at the end of the latest episode within the chosen target period. Target assessments are the latest assessments of residents that meet three criteria stated in section 3 and 4 in the Chapter 1 of the QM Manual. You may modify the CMS rules to suit your research needs. Note: for CMS quality measures other than pain, we might use initial assessments representing the resident’s status as soon as possible after the admission that marks the beginning of the episode or conduct look-back scan to determine whether such events or conditions occurred during the look-back period. The criteria for selecting target assessments are stated in the Chapter 1 of the MDS 3.0 QM User’s Manual (see above for link). While calculating the CMS pain measure, target assessments with J0200, J0300, J0400 and J0600 missing will be dropped if one is meeting the exclusion criteria (as specified in the manual). Exclusion criteria are stated in the Chapter 2 of the MDS 3.0 QM User’s Manual.

 

Conclusions

Due to the nature of dash category described above, researchers should be cautious about treating it as missing while conducting analyses other than calculating the CMS pain measure. Also, researchers should be cautious about treating the blank resulting from a skip pattern as missing. We recommend coding and verifying skip and missing patterns while selecting your study sample, such as self-report pain sample vs. staff-report pain sample, where blank cells for self-report pain items may mean that residents who had such blank cells were in staff-report sample.

Additional Resources: 
AttachmentSize
PDF icon Figure 1.pdf1 MB
Article Number: 
31
Author(s): 
This work was performed under CMS Contract Number HHSM-500-2015-00558G
Related Data File Families: 

Disclaimer

The process and materials mentioned as part of this KnowledgeBase article are current, as of the publication date on the article, to the best of our knowledge. The examples provided are correct in the aggregate but may not apply to every subgroup or circumstance that a researcher may wish to study. It is up to the researcher to conduct analysis and confirm that the patterns described in this KnowledgeBase article apply to his/her particular study. If your research findings appear to contradict the advice provided, please contact ResDAC at resdac@umn.edu.