Medication Regimen Complexity among Community-Dwelling Older Adults with Incident Mild Cognitive Impairment or Dementia

Introduction

Beyond medication count, complex medication regimens may be especially risky and burdensome for people with dementia or mild cognitive impairment (MCI) and their caregivers. The Medication Regimen Complexity Index (MRCI), which incorporates dosage form, frequency, and additional directions,^1,2 may be a useful tool to identify people with dementia or MCI who would benefit from deprescribing. This study sought to automate MRCI calculation in a large, real-world database of people with MCI or dementia and to examine contributions of specific MRCI components to overall complexity.

Methods

This was a cross-sectional study using existing medical record data from 7 Minnesota counties in the Rochester Epidemiology Project (REP) medical records-linkage sytem,³ which captures information from healthcare provided to 90% of the residents of the region.⁴ We searched REP electronic indexes to identify residents aged ≥65 with incident MCI or dementia from January 1, 2015 through December 31, 2017 (Supplementary Table S3). We searched the REP for outpatient medication prescriptions and self-reported medications for individuals in the 30 days before and after their dementia diagnosis.

The MRCI is a sum of 3 weighted subscores: form/route (Part A), frequency (Part B), and additional instructions (Part C).¹ Information on form, route and frequency were obtained from electronic prescription information. We examined the free-text and frequency fields for text patterns corresponding to Part C (e.g., “crush,” “meal,” and “bedtime”).⁵ Decision rules were created for cases not clearly addressed in the MRCI instructions (Supplement). Two geriatricians (ARG, SN) refined the algorithm by searching for text patterns that had not been accounted for in earlier steps and to adjudicate discrepancies. We calculated MRCI scores first as a sum of the weighted scores for Part A and Part B only, and second as a sum of the weighted scores for all three parts.⁶ Higher scores indicated greater complexity.

Patient characteristics were summarized and tested using chi-square tests. MRCI scores were summarized with median (interquartile range, IQR); differences were tested using the Kruskal-Wallis test. Correlations were summarized with Spearman correlation coefficients.

Results

The cohort consisted of 3,976 people and 29,059 linked medication records (Supplementary Tables S1 and S2). Among people with ≥1 medication, two central nervous system-active medications – opioids and antidepressants – together comprised 8% of medication prescriptions (Supplementary Table S2). Median MRCI scores across demographic/clinical characteristics are in Table 1. The median MRCI score was 12 (IQR: 5–25) calculated using Parts A and B and 14 (IQR: 6–29) calculated using all three parts. The biggest contributor to MRCI score was dosing frequency (Part B). Medication count was associated with MRCI score (Spearman r=0.91, P<0.01; Supplementary Figure S2). There was wide variation in MRCI scores among patients with the same number of medications (Figure 1).

Table 1.

Medication regimen complexity index (MRCI) by patient characteristics

		Parts A (form/route) and B (frequency)		Parts A (form/route), B (frequency), and C (additional directions)
Characteristic	N (%)	Median (IQR)a	P-valueb	Median (IQR)	P-valueb
Total	3,976	12 (5–24)		14 (6–29)
Sex
Men	1,742 (43.8)	12 (5–25)	0.51	14 (7–30)	0.23
Women	2,234 (56.2)	12 (5–24)		13 (6–28)
Age group (years)
65–70	540 (13.6)	12 (5–24)	0.01	14 (7–29)	0.03
>70–75	590 (14.8)	13 (6–25)		15 (7–29)
>75–80	757 (19.0)	11 (5–24)		13 (6–29)
>80–85	783 (19.7)	11 (5–21)		13 (6–26)
>85	1,306 (32.8)	12 (6–25)		15 (7–29)
Race
White	3,788 (95.3)	12 (5–24)	0.08	14 (6–28)	0.18
Black	34 (0.9)	9 (6–19)		11 (9–25)
Asian	51 (1.3)	15 (7–25)		16 (7–29)
2 or more/Other/missing	103 (2.6)	16 (7–31)		18 (8–38)
Ethnicity
Hispanic	66 (1.7)	15 (7–28)	<0.01	18 (9–34)	<0.01
Non-Hispanic	3,590 (90.3)	12 (6–25)		14 (7–30)
Unknown	320 (8.0)	7 (4–14)		9 (5–17)
Charlson index
0	328 (8.2)	6 (3–12)	<0.01	7 (4–14)	<0.01
1–2	1,178 (29.6)	9 (4–17)		10 (5–20)
≥3	2,470 (62.1)	15 (7–28)		17 (8–34)
MRCI	-
Part A score (form/route)	-	3 (1–6)		3 (1–6)
Part B score (frequency)	-	9 (4–19)		9 (4–19)
Part C score (additional directions)	-	-		2 (0–4)

^aIQR: interquartile range

^bKruskal-Wallis test

Figure 1.

Variation in MRCI scores across regimens with the same number of medications

The bottom box shows the regimen of a patient with an MRCI score of 9 (33% lower than average for patients with 5 medications) and the top box of a patient with an MRCI score of 17 (33% higher than average)

Discussion

Frequency was the biggest contributor to medication regimen complexity in this cohort of patients with MCI or dementia, similar to studies in non-dementia populations.⁵ MRCI scores varied widely among patients with the same number of medications and may more accurately capture patients’ and caregivers’ lived experience than medication count.

Complex medication regimens may increase the risk of poor health outcomes.⁷ The steps we undertook to calculate the MRCI could be used to identify people with MCI and dementia who may be most likely to benefit from deprescribing. Opioids and antidepressants were among the top 10 most common medication classes in our cohort. Reducing use of central nervous system-active medications may be important to reducing complexity for this population, as use of such medications is common among people living with dementia and associated with numerous adverse health outcomes.⁸

To implement the MRCI for pragmatic deprescribing trials, it would need to be automated for use within electronic medical records in real time. Part C was difficult to automate because of the wide variety of ways in which special administration instructions can be expressed, requiring coder discretion. We found that calculating the MRCI score using only Parts A and B was comparable to incorporating the Part C subscore in terms of identifying patients with high versus low complexity.⁶

The MRCI may prove to be more useful than number of medications for identifying patients with MCI or dementia who are most likely to benefit from deprescribing interventions. Addressing high medication regimen complexity – for example, by eliminating medications that are taken multiple times per day or have complicated administration instructions - could reduce self-care demands, prevent institutionalization and lessen caregiver strain.⁹ A limitation of this research is that misclassification of medication use is possible.

In conclusion, this study characterized medication regimen complexity among people with MCI or dementia. Future studies should assess the impact of reducing MRCI scores on clinical outcomes, including adverse events and patient- or caregiver-reported measures of treatment burden.