Changes in the agreement between the Minimum Data Set and hospital Medicare claims measures of dementia

Introduction

Previous research shows that Medicare claim measures of dementia based on International Classification of Disease (ICD) coding lack sensitivity, especially for men, younger adults, and racial/ethnic minorities.^1–6 One study found that the validity of the dementia measure based on Medicare claims data increased between 2006 to 2012 and that racial/ethnic differences narrowed over time.¹ No research to our knowledge has examined trends in the validity of hospital Medicare documentation of dementia since 2012. The validity may have changed due to increased recognition of dementia, changes from ICD-9 to ICD-10 diagnosis codes, and changes in payment related to disease severity in Medicare Advantage plans.⁷ The objective of this study was to examine the agreement between a clinical Minimum Data Set measure of dementia and a diagnosis of dementia documented on a hospital claim across three points in time. A second objective was to examine the extent to which the agreement varied by age, sex, and race/ethnicity.

Methods

This study linked national Minimum Data Set (MDS) assessments with the Medicare Provider Analysis and Review (MedPAR) file from 2012–2017. The MDS is a federally mandated patient assessment conducted upon admission to a nursing home and then at least quarterly.⁸ In order to be classified as having an MDS-based diagnosis of dementia, individuals had documentation of dementia and a score of 2 or higher (indicating at least mild cognitive impairment) on the Cognitive Function Scale (CFS).⁸

All acute care hospitalizations among individuals with an MDS assessment completed within 120 days prior to the index hospitalization between 2012–2017 were identified (n=10,249,356). We limited our sample to hospitalizations from individuals with an MDS-based diagnosis of dementia, who were not comatose, and who were white, Black, Asian, or Hispanic given the small sample sizes of other racial/ethnic groups.

Using the MDS documentation of dementia as a reference standard, we reported the percent of hospitalizations with a dementia diagnosis noted on the corresponding claim based on ICD-9 or ICD-10 diagnosis codes between years 2012–2017 by age, sex, and race/ethnicity. Additional information about the data and methods used for these analyses can be found in the Brown Digital Repository.

Results

A total of 1,273,348 individuals with dementia as defined in the MDS experienced 2,251,749 hospitalizations between the years 2012 and 2017. When the MDS noted dementia in 2012, 76.9% of hospitalizations had dementia documented on the Medicare claim (Table 1). The agreement increased to 86.3% in 2017. The agreement between the two measures of dementia differed by age: in 2017, 79.2% of hospital claims for patients aged 66–74 had a diagnosis of dementia compared to older age categories that had dementia noted in over 85% of records. Women were slightly more likely to have a diagnosis of dementia documented in the Medicare hospital claims data. Compared to white patients, Asian patients were less likely to have a diagnosis of dementia noted. Approximately 84.2% of hospital records from Asian patients had a diagnosis of dementia compared to 86.6% of records from white patients.

Table 1.

Agreement between a dementia diagnosis on the Minimum Data Set and hospital Medicare claim among hospitalized individuals with dementia, over time

	Year 2012		Year 2014		Year 2017
	Number of hospitalizations among individuals with an MDS diagnosis of dementia	Diagnosis present on hospital claim (%)	Number of hospitalizations among individuals with an MDS diagnosis of dementia	Diagnosis present on hospital claim (%)	Number of hospitalizations among individuals with an MDS diagnosis of dementia	Diagnosis present on hospital claim (%)
Total population	437,921	76.9	369,193	83.1	343,958	86.3
Age group
66–75	64,281	69.0	57,473	74.8	58,751	79.2
76–85	175,782	77.3	145,416	83.5	136,745	86.6
86–95	197,858	79.0	166,304	85.6	152,730	86.3
Sex
Male	161,442	75.5	140,232	81.7	137,336	85.1
Female	276,479	77.6	228,961	83.9	213,003	86.9
Race
White	344,996	77.5	288,351	83.5	267,010	86.6
Black	72,666	75.0	62,967	81.9	59,176	85.3
Asian	7,378	71.9	6,693	80.7	7,279	84.2
Hispanic	12,881	74.8	11,182	81.2	10,493	85.0

Notes: Data came from the 2012, 2014, and 2017 inpatient MedPAR file and the MDS. Hospitalizations occurred within 120 days of an MDS assessment. In the MDS, dementia was defined based on a diagnosis of dementia and a Cognitive Function Scale score of at least 2. The hospital diagnosis of dementia was defined based on an ICD-9 or ICD-10 code.

Discussion

The overall increase in the accuracy of hospital claims may indicate that clinicians are recognizing dementia at higher rates in this setting. The remaining lack of documentation may be caused by physicians attributing cognitive problems to “normal aging” or prioritizing the documentation of physical ailments.⁹ Overall, there are encouraging signs that the gaps across age and race groups have narrowed over time; they do persist and warrant further attention. Hospital claims of patients aged 66–75 were less likely to be accurate than other age groups, which is consistent with previous studies^1,3 and suggests physicians do not always look for signs of dementia in younger adults. Our study also found that Asian patients were less likely to have a diagnosis of dementia documented during the hospitalization, which could be related to language barriers between patients and clinicians.

Our sample consisted entirely of hospital claims that had a preceding nursing home stay, which limits generalizability to other populations. We also used the MDS assessments as the reference standard. Although the MDS includes a validated measure of cognitive function, there could be remaining measurement error.

Older adults with dementia have medical needs that differ from individuals without dementia and require targeted care.¹⁰ Improved screening efforts may improve documentation of dementia diagnosis during a hospital stay and potentially improve medical decision making. Future research should also consider how the bias in using the Medicare diagnosis of dementia in hospital claims may affect findings regarding the medical care and health care costs of individuals with dementia.