Clinical Therapeutics/Volume 36, Number 11, 2014
Original Research
Patient-level Medication Regimen Complexity in Older Adults With Depression Sunny A. Linnebur, PharmD, FCCP, BCPS, CGP1; Joseph P. Vande Griend, PharmD, BCPS, CGP1; Kelli R. Metz, PharmD5; Patrick W. Hosokawa, MS2; Jan D. Hirsch, BS Pharm, PhD3,4; and Anne M. Libby, PhD1,5 1
Department of Clinical Pharmacy, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of Colorado, Aurora, Colorado; 2Colorado Health Outcomes Program, School of Medicine, University of Colorado, Aurora, Colorado; 3Department of Clinical Pharmacy, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of California San Diego, San Diego, California; 4Veterans Affairs of San Diego Healthcare System, San Diego, California; and 5Center for Pharmaceutical Outcomes Research, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of Colorado, Aurora, Colorado
ABSTRACT Purpose: Polypharmacy and medication adherence are well known challenges facing older adults. Medication regimen complexity increases the demands of self-care in the home. Some medication regimens may be more complex than others, especially when dosage form, frequency of dosing, and additional usage directions are included in complexity along with the number of medications In older adults with depression, it is unknown what features of their medications most influence their medication regimen complexity. Methods: A sample cohort of 100 adults ≥65 years old with a diagnosis of depression was randomly selected from electronic medical records (EMR) in ambulatory clinics at the University of Colorado (CU) and University of San Diego (SD). Demographic, medical history, and medication-related information was extracted from the EMR. Complexity was determined using the Medication Regimen Complexity Index (MRCI). IRB approval was obtained. Findings: The cohort mean age was 74.3 years (SD) and 79.7 years (CU). The mean unweighted Charlson comorbidity index for 1.0 (SD) and 1.8 (CU). The
Accepted for publication October 6, 2014. http://dx.doi.org/10.1016/j.clinthera.2014.10.004 0149-2918/$ - see front matter & 2014 Elsevier HS Journals, Inc. All rights reserved.
1538
mean number of medications was 7.1 and 8.0, with 1.1 and 1.2 depression meds, 5.4 and 4.3 nondepression prescription meds, and 0.6 and 2.4 OTC meds for the SD and CU cohorts, respectively. 66% of SD adults and 70% of CU adults took six or more meds. Individual MRCI scores were on average 17.62 (SD) and 19.36 (CU). Dosing frequency contributed to 57-58% of the MRCI score, with patients facing an average of 7–8 unique dosing frequencies in their regimen. In both cohorts, there was an average of 3 additional directions added to the regimens to clarify dosing. Implications: As expected, in our older adult cohorts with depression the majority of patients took multiple medications. Using a standardized instrument, we characterized the regimen complexity and found that it was increasingly complex due to numerous dosing forms, frequencies and additional directions for use. Patient-level medication regimen complexity should go beyond depression medication to encompass the patient’s entire regimen for opportunities to reduce complexity and improve ease of selfcare. (Clin Ther. 2014;36:1538–1546) & 2014 Elsevier HS Journals, Inc. All rights reserved.
Scan the QR Code with your phone to obtain FREE ACCESS to the articles featured in the Clinical Therapeutics topical updates or text GS2C65 to 64842. To scan QR Codes your phone must have a QR Code reader installed.
Volume 36 Number 11
S.A. Linnebur et al. Key words: aged, depression, geriatrics, medication adherence, medication regimen complexity, MRCI.
INTRODUCTION Older adults are often challenged by managing multiple chronic conditions and the many medications associated with those conditions. Medication use by older adults can be even further complicated by the complexity of the medication regimens. Medication regimen complexity has been found to be high in institutionalized elderly—a setting where older adults have assistance in managing their medications.1,2 In ambulatory settings as older adults manage medications on their own, increasing medication regimen complexity increases self-care demands. Complexity of medication regimens can vary by the number of medications, different dosage formulations, dosing frequencies, and additional directions for use. Data indicate that increasing medication regimen complexity and especially frequency of dosing are directly associated with medication nonadherence.2–7 Medication nonadherence can be associated with polypharmacy and self-care demands; these can precipitate adverse events, poor response to medications, poor control of medical conditions, disease state complications, and increased health care burden.8,9 A gap exists in the literature whereby studies of medication regimen complexity have measured complexity associated only with a disease-defined cohort, instead of for the entire patient. An additional gap in knowledge is whether regional variations in care have a significant effect on medication complexity for a given cohort. To characterize patient-level medication regimen complexity for older adults, we chose to evaluate a disease-defined cohort with depression diagnoses and antidepressant prescriptions. As a marker for potential adherence problems, we sought to assess patient-level medication regimen complexity for older adults with treated depression. Depression, a common comorbidity in the older adult, was used because it has been shown to adversely affect medication adherence for common chronic disease states in older adults.10–14 The patients used in this cohort represent patients whose adherence to a medication regimen would likely be affected by a complex medication regimen. We planned to measure medication regimen complexity of the patients’ antidepressant regimens and to
November 2014
contrast it with the medication regimen complexity attributable to all other medications the patients were taking. This would assist in understanding medication regimen complexity as a patient-level construct that may be a key piece of overall patient complexity. Additionally, the integral pieces comprising the patient’s medication regimen complexity could inform practitioners of targets to reduce or simplify regimens and possibly affect adherence. Validated tools for calculating medication regimen complexity exist in the literature. The Medication Regimen Complexity Index (MRCI) is a nondiseasespecific validated tool that measures medication regimen complexity using paper-coded medication regimens from medication lists.15 Scores are derived from weighted values of the regimen components (eg, dosage formulations, dosing frequencies, and specific administration instructions). The MRCI was originally validated in 1 disease-defined cohort (chronic obstructive pulmonary disease) and measured only the prescriptions for that disease state.15 An extension of that work, the parent study of the current study, broadened the validation using multiple disease cohorts, to a patient-level MRCI (pMRCI) measuring medication regimen complexity across all of a patient’s medications for the practical purpose of identifying patients with expected difficulty managing medication regimens and warranting interventions.16,17 The pMRCI is a sum of weighted subscores for 3 mutually exclusive medication categories: diseasespecific prescription medications (ie, antidepressant medications), all other prescription medications, and over-the-counter (OTC) medications recorded on a medication list. This study aimed to use the quantitative tool, the pMRCI, to measure medication regimen complexity and component parts for 2 cohorts of older adults with depression to better characterize the types of medication regimen complexities and possible associated self-care demands faced by older adults in the community. The objectives of this study were to evaluate the entire medication regimen and to determine potential targets to simplify the regimen and improve adherence. We hypothesized that cohorts from ambulatory care clinics at 2 different, geographically diverse, academic medical centers would be similar. We also hypothesized that our cohorts of patients with depression, known in the literature to be relatively nonadherent to medications, would
1539
Clinical Therapeutics have a high contribution to medication regimen complexity from nonantidepressant regimens. We expected to find certain features of the regimens that could offer areas for simplification that could affect adherence.
PATIENTS AND METHODS This study was a retrospective, cross-sectional study of 2 randomly selected cohorts of 100 adults 65 to 89 years of age with treated depression in active care from 2011 to 2012 at ambulatory clinics at the University of California San Diego (referred to later as CA) and the University of Colorado Anschutz Medical Campus (referred to later as CO). Two cohorts from different geographic regions were chosen to increase generalizability of the descriptive findings. Cohort size was determined based on a power calculation for the number needed to compare the distributional metrics of pMRCI values across sites and limited by person-time needed to code the pMRCI. The electronic health record (EHR) was the source of the medical, pharmacy, and patient data for the cohorts. Although using the EHR is a real-world representation of a patient’s active medication list, we sought to use additional EHR data as a conservative approach to improve the accuracy of active treatment. During the data extraction process, an additional step was added in which medications in the EHR were excluded if they had expired after 12 months of no renewal (eg, they were administratively discontinued). The 2 university sites were partners in a larger parent study of medication regimen complexity to examine regional variation in treatment patterns. A full description of the methodologies of the larger study is published elsewhere.16 This study was approved by both the University of California San Diego Human Research Protections Program and University of Colorado Multiple Institutional Review Boards.
Data Source and Definition of Cohorts Deidentified electronic data were used for cohort creation. Inclusion criteria were used to create an index visit from which the medication list was drawn. For each subject, the source of the data was a clinic EPIC-based EHR, drawn from a retrospective clinical data repository managed by a third party. The last visit recorded between July 1 and December 31, 2012 became designated as the index visit. Among the eligible older adults, we included only active patients
1540
defined as having at least 1 visit within the year before the index visit plus the index visit. Inclusion also required a qualifying medical diagnosis of depression (International Classification of Diseases, Ninth Revision 296.2, 296.3, 300.4, or 311) on the “current problem list” and at least 1 antidepressant (identified using National Drug Codes or drug name within the associated Generic Product Identifier [GPI] group 58 antidepressants [Medi-Span Electronic Drug File MED-File v2]) documented in the medication list. Additional medications were evaluated to exclude those medications that had been administratively discontinued. There were no other exclusion criteria. Among all eligible patient records, a random sample was chosen using a number generator to select the study cohorts (N ¼ 100 at both sites) based on power calculations needed to compare means across sites. Once the random sample of 100 records from each clinic was created, extracts were created that included anonymous data on current medication regimens, demographic factors, and active diagnoses. A clinical pharmacist (K.R.M.) coded the pMRCI using an electronic data capture tool that calculated 3 subscores for (1) antidepressant medications, (2) other prescription medications that were not antidepressants; and (3) OTC medications. Primary outcome measures were the overall pMRCI scores and subscores.
MRCI Coding The electronic data capture tool for the MRCI used for coding is publically available at http://www.ucden ver.edu/academics/colleges/pharmacy/Research/research areas/Pages/MRCTool.aspx.16 A screen shot is included as an Supplemental Appendix. The electronic data capture tool allowed the calculation of pMRCI as a simple sum of all 3 weighted subscores based on identical data capture forms for dose formulation, dosage frequency, and additional administration directions. Patient records were assigned a unique anonymous identifier, and their medication regimen was coded. Tool instructions developed by the research team were followed to determine the way in which medications were coded to the 3 medication categories (antidepressants, other prescription medications, and OTC medications) using some decision rules. For example, all antidepressants were coded as antidepressant medications even if they were possibly ordered for the treatment of another
Volume 36 Number 11
S.A. Linnebur et al. condition (eg, trazodone for insomnia). In addition, decision rules were put in place to consistently code for when a drug might be considered a prescription or OTC medication (eg, vitamin D).16
Data Analysis Coded patient-level data were exported from the electronic data capture tool for analysis with SAS version 9.3 software (SAS Institute, Inc, Cary, North Carolina). Age, Charlson comorbidity index, and pMRCI scores were compared between cohorts using an independent-sample t test. Sex was compared via χ2 test. Medication count was evaluated both as a continuous variable and as a categorical variable (1-5, 6-10, Z11) and was compared between cohorts both for medications as a whole and for each subcategory (depression, other prescriptions, OTC medications). Unique medications in the electronic medication list were counted, combining only where it appeared that the medicines were the same prescription but of variable dosages. Medication types were compared between groups by aggregating medications into categories via the first 2 digits of their GPI code. Correlation between continuous variables (medication count, MRCI score, and Charlson comorbidity index score) were evaluated using the Spearman correlation coefficient. The a priori level of significance was set at P o 0.05. The number of unique frequencies and additional directions was generated by counting each instance recorded in the ACCESS data capture tool.
Measures Dependent Variables MRCI Score/pMRCI The original validated MRCI tool was developed by George et al15 to assign complexity scores to a disease-defined medication regimen for individual patients based on the 3 components: (1) dosage forms, (2) dosing frequencies, and (3) additional directions (eg, take at a specified time, take in relation to food). The MRCI tool is a weighted sum with a base case weight of 1 given to tablet/capsule and weight of 1 given to frequency of once daily; thus, the lowest score from MRCI is 2. Other dosage formulations and dosing frequencies are assigned increasing weight related to the increasing difficulty in administration (eg, injectable agents, topical agents, every 4 hours as needed). Additional administration directions (eg,
November 2014
break or crush, take with food, take at a certain time of the day) associated with a medication and documented in the prescription also add to the score and are weighted according to difficulty in administration. pMRCI scores compile complexity scores for each of the 3 components for all of a patient’s medications.
RESULTS Table I presents population characteristics for the patient samples drawn from California (CA) and Colorado (CO). Characteristics were compared between the geographic groups to evaluate for similarities and differences between the groups. The mean age of the patients was in their 70s, and the majority were female. The patients in both cohorts were fairly healthy based on the Charlson comorbidity index, with the CO patients having slightly more comorbidities on average than the CA patients (P o 0.0001). The most common of the Charlson comorbidities was diabetes mellitus in CA and chronic obstructive pulmonary disease in CO, with the second most common in both cohorts being cancer. The average total count of medications was slightly, but not significantly, lower in CA (mean, 7.1) than in CO (mean, 8.0) (P ¼ 0.10). The difference appeared to be driven by the only significant difference between the cohorts: more OTC medications in the CO patients (mean, 0.6 [CA] compared with mean, 2.4 [CO], P o 0.0001). The distribution of polypharmacy (defined as medication count 45 medications) was also similar between groups with approximately two-thirds of all patients taking 45 medications. Medication counts ranged from 1 to 19 in the CA cohort and 2 to 18 in the CO cohort. Similarly, the pMRCI scores ranged from 2 to 47 and 4 to 48, respectively. Medication count (Figure 1) and Charlson comorbidity index (data not shown) were significantly associated with pMRCI score (r ¼ 0.92, P o 0.001 and r ¼ 0.22, P ¼ 0.0016, respectively). As expected, depression-related medications did not contribute as significantly to the medication regimen complexity compared with other prescription medications. In the CO cohort, OTC medications contributed more to the pMRCI than depression-related medications. Use of OTC agents in the Colorado cohort was common and, as such, was evaluated for potential indication (Figure 2). Conditions that were most
1541
Clinical Therapeutics
Table I. Population characteristics and medication use.
No. of patients Demographic characteristics Age, mean (SD) Sex (% female) Clinical characteristics Charlson comorbidity index, range, 0–17 Most common comorbidity group
California
Colorado*
100
100
74.3 (7.4) 76 0.9 (0–5) DM without complications, 18% Cancer, 17% 17.62 (10.05) 7.1 (1–19)
P
79.7 (6.1) 81
o0.0001 0.39
1.8 (0–6) COPD, 35%
o0.0001 N/A
Second most common comorbidity group Cancer, 30% N/A Mean pMRCI 19.36 (10.21) 0.21 Mean medication count, all (range) 8.0 (2–18) 0.07 By category Mean no. of depression medications (range) 1.1 (1–3) 1.2 (1–3) 0.02 Mean no. of other prescription medications, (range) 5.4 (0–15) 4.3 (0–14) 0.03 Mean no. of over-the-counter medications (range) 0.6 (0–6) 2.4 (0–11) o0.0001 By quantity 1–5 medications, % 34 30 0.78 6–10 medications, % 49 50 Z11 medications, % 17 20 Most common GPI group (% of total prescriptions) Antidepressants, 16.3% Antidepressants, 15.7% N/A Second most common GPI group Antihyperlipidemic, 7.0% Antihyperlipidemic, 7.2% N/A Third most common GPI group Antihypertensive, 6.1% Antihypertensive, 6.9% N/A COPD ¼ chronic obstructive pulmonary disorder; DM ¼ diabetes mellitus; GPI ¼ generic product identifier; N/A ¼ statistical testing not applicable, presented for descriptive purposes; pMRCI ¼ patient-level Medication Regimen Complexity Index. * Some cohort descriptive statistics were presented in Table 1 of Libby et al (2013)16 for the same depression cohort; additional descriptive information is presented in original form here.
frequently treated with OTC medications included heart health, constipation, pain, and indigestion. Figure 3 displays the average pMRCI scores by site, with colors to differentiate dosage formulation, dosing frequency, and administration directions; vertical bars also distinguish the pMRCI subscores for antidepressant medications, other prescription medications, and OTC medications. Table II provides the mean values for the total pMRCI for the CA and CO patients, along with the mean scores for each portion of the pMRCI calculation. The pMRCI for both patient groups comprised mostly points from the frequency of use and was due to the high number of medications recorded for each patient. Frequency of dosing was most often not limited to once-daily dosing. In the CA and CO patients, the mean number of dosing frequencies per patient, calculated for the 3 categories of
1542
medications, was 6.98 (SD 3.72) and 7.73 (SD 3.49), respectively. On average, approximately 3 “additional directions” for use (eg, break/crush tablet) were recorded per patient in both cohorts. The most common additional direction in both groups was to take the medication at a specified time of day. More than one-half of all patients in both groups had more than 1 dosage form; the most common additional formulation beyond tablet/capsule was a cream (data not shown). Six percent (CA) and 11% (CO) of patients had Z3 distinct dose formulations on their medication list (data not shown).
DISCUSSION Using medication lists and active problem codes in EHRs, pMRCI scores were calculated and determined to be similar across 2 academic ambulatory clinics
Volume 36 Number 11
S.A. Linnebur et al.
MRCI 60
50
40
30
20
10
0 0
5
10 15 Medication Count Facility
California
20
25
Colorado
Figure 1. Medication Regimen Complexity Index (MRCI) and medication count.
located in CA and CO, suggesting that our description of this population has external validity. The pMRCI expanded the assessment of medication regimen complexity to include all prescription and nonprescription
Other 24%
Heart Health 22%
Indigestion (mild to severe) 15%
Constipation 22%
Pain 17%
Figure 2. Conditions treated with over-the-counter (OTC) agents.
November 2014
medications recorded in the medical record. Similar to Mansur et al,2 the majority of older adult patients in our 2 samples had documented polypharmacy, and their pMRCI scores reflected this. As expected, depression treatment did not contribute greatly to polypharmacy or medication regimen complexity. Rather, the pMRCI collected all other prescription and OTC medications to give a complete representation of medication regimen complexity and potentially enable a pharmacist to target sections of the medication regimen in older adults with depression to reduce complexity. When pMRCI score components were evaluated, it was found that dosing frequency contributed quantitatively the most to the patients’ pMRCI scores, with many patients being prescribed multiple unique frequencies for dosing. In addition, many patients had additional instructions for taking their medications beyond frequency (eg, take at a specific time of day). Multiple formulations, several unique dosing frequencies, and additional instructions likely complicate any patient’s ability to maintain appropriate and consistent medication administration practices. In the case of older adults with depression, the demands of maintaining adherence may be further complicated by limitations in vision, dexterity, or memory.
1543
Clinical Therapeutics
Mean Score 20
15
10
5
0
Disease
Other Rx OTC California Component
Total Score
Dosage Forms
Disease
Frequency
Total Score
Other Rx OTC Colorado Additional Directions
Figure 3. Mean patient-level Medication Regimen Complexity Index at California and Colorado clinics. OTC ¼ over-the-counter; Rx ¼ prescription.
The use of the pMRCI scores to assess medication regimen complexity greatly enhances the understanding of what older adults with depression may face with medication administration and adherence, beyond a simple pill count or pill burden. The pMRCI may prove to have utility as a risk prediction tool for medication regimen complexity and nonadherence, perhaps similar to less tailored metrics like the number of medications and the number of comorbidities as used in some medication therapy management programs. More research is needed to investigate the association of
pMRCI scores with adherence and patient outcomes. This was not possible given our data source, but is the next important step in this work to link the pMRCI to patient outcomes. In older adults, other factors such as social, demographic, and concomitant conditions could also play a large role in clinical outcomes. In addition, to be used as a point of care pMRCI tool for providers or patients, it is likely that the current electronic scoring version in ACCESS would need to be simplified to make the tool feasible and acceptable at the point of care.
Table II. Patient-level Medication Regimen Complexity Index) at California and Colorado clinics. California (n ¼ 100) Mean (SD) Medication count Total pMRCI Dosage form Frequency Additional directions
7.10 17.62 4.42 10.19 3.01
(3.73) (10.05) (2.78) (6.39) (2.67)
Range (IQR) 1-19 2-47 1-15 1-30.5 0-15
(4.5–9.5) (10.5–23.25) (2–6) (6–13.25) (1–5)
Colorado (n ¼ 100) Mean (SD) 7.96 19.36 5.43 11.06 2.87
(3.63) (10.21) (3.05) (6.08) (2.44)
Range (IQR) 2-18 4-48 1-15 2-30 0-10
(5–10) (11.75–25.5) (3–7) (6.5–14.5) (1–4)
IQR ¼ interquartile range; pMRCI ¼ Patient-level Medication Regimen Complexity Index.
1544
Volume 36 Number 11
S.A. Linnebur et al. Based on the high level of medication regimen complexity in both of our patient groups, our results suggest a need for pharmacist review of the patient’s entire medication regimen, not just disease-specific or prescription-only medications. The goal of this medication review would be to assess and reduce complexity to a manageable level for the patient if possible. Pharmacist review could identify opportunities to reduce the number of unnecessary prescription and OTC medications and to use once-daily regimens and reduced administration instructions. In older adults, reducing the number of dosage formulations may not be feasible, given the potential benefit for reduced systemic adverse events with the use of certain nonoral dosage forms (eg, topical or transdermal forms). In fact, our results indicate that creams are already being used at a high rate in the older adults with depression. Pharmacist review could also investigate actual nonadherence issues that patients may be experiencing due to specific complexities in their medication regimen and provide more targeted patient education about proper administration and how to overcome adherence-limiting regimen issues. Only 1 significant difference in medication use between CA and CO was observed; OTC use was higher on average among the CO patients. One possible reason for this is that the CO providers documented the OTC use at a higher rate than the CA providers. Data collection in this study was limited by what was documented in the electronic medical record; thus, it is very likely that the use of OTC medications was higher than documented in both study groups. Another possible reason for the difference in OTC use between cohorts is that patients in CO may be more interested in using nonprescription treatments. This study has several imitations. First, there was no direct measure that the medication was consumed by the patients, only that it was recorded as prescribed or continued at the index visit. By clinic protocols, however, each visit at both sites included a medication reconciliation component. The requirement of each patient record to include at least 2 visits in the year should have provided an opportunity for medication reconciliation to identify any medication changes and allow for optimal accuracy in the medication record. The pMRCI was hand-coded by a pharmacist using an Access database. Automated coding at the point of care would be much more clinically relevant for practitioners. McDonald et al18 recently published
November 2014
their successful attempt at using an automated tool to calculate the pMRCI from 90,000 EHRs. Their findings were very similar in terms of the mean number of medications and pMCRI for the population that they studied, indicating our handcoding and their automated coding may be consistent. Use of an automated system like that of McDonald et al could also reduce individual coding error and quickly provide the practitioner with validated complexity information at the point of making medication changes. Using this information has the potential to improve medication administration, adherence, adverse events, and other patient outcomes if interventions and education are used along with the pMRCI. In addition, pharmacists could use an automated pMRCI when they perform medication therapy management visits and medication reconciliation in the ambulatory and community setting. Finally, the results of this study are likely generalizable to other ambulatory older adults with depression but may not apply to older adults without depression.
CONCLUSIONS In this study, EHRs of older adults with depression from 2 distinct ambulatory clinics were evaluated using the pMRCI tool to assess medication regimen complexity. Medication regimen complexity and its components were very similar, despite geographic differences. On average, the patients took 45 medications per day, including OTC medications, and their medication regimen complexity directly correlated with the number of medications and comorbidities. The most influential features in the patient-level MRCI scores were dosing frequencies and dosage forms, and these scores were derived from the nondepression prescription medications and OTC agents. Using this information may provide a mechanism to positively reduce the medication regimen complexities of older adults with depression. Future research with the pMRCI tool could focus on assessing medication regimen complexity with an automated pMRCI tool at the point of care, which portion of the pMRCI relates the most to adherence, and the clinical effects of reducing medication regimen complexity in older adults with depression. Opportunities may be available to reduce medication regimen complexity, improve ease of self-care, improve adherence, and reduce adverse drug events in older adults.
1545
Clinical Therapeutics
ACKNOWLEDGMENTS This research was funded by The ALSAM Foundation Skaggs Scholars Program at the University of Colorado Skaggs School of Pharmacy and Pharmaceutical Sciences. Drs. Linnebur, Vande Griend, and Metz were responsible for the literature search, study design, data collection/coding, data interpretation, and writing of the manuscript. Mr. Hosokawa was responsible for the data analysis, data interpretation, figure and table creation, and writing of the manuscript. Dr. Hirsch was responsible for the literature search, study design, data interpretation, and writing of the manuscript. Dr. Libby was responsible for the literature search, study design, data analysis, data interpretation, and writing of the manuscript.
7.
8.
9.
10.
11.
CONFLICTS OF INTEREST The authors have indicated that they have no conflicts of interest regarding the content of this article.
12.
SUPPLEMENTAL MATERIAL Supplemental materials accompanying this article can be found in the online version at http://dx.doi.org/10. 1016/j.clinthera.2013.12.011
13.
REFERENCES 1. Advinha AM, de Oliveira-Martins S, Mateus V, et al. Medication regimen complexity in institutionalized elderly people in an aging society. Int J Clin Pharm. 2014;36:750– 756. 2. Mansur N, Weiss A, Beloosesky Y. Looking beyond polypharmacy: quantification of medication regimen complexity in the elderly. Am J Geriatr Pharmacother. 2012;10:223–229. 3. Conn VS, Taylor SG, Kelley S. Medication regimen complexity and adherence among older adults. J Nurs Scholarship. 1991;23:231–235. 4. Ingersoll KS, Cohen J. The impact on medication regimen factors on adherence to chronic treatment: a review of literature. J Behav Med. June 2008;31:213–224. 5. Balkrishnan R. Predictors of medication adherence in the elderly. Clin Ther. 1998;20:764–771. 6. Iskedjian M, Einarson TR, MacKeigan LD, et al. Relationship between daily dose frequency and adherence to
14.
15.
16.
17.
18.
antihypertensive pharmacotherapy: Evidence from a meta-analysis. Clin Ther. 2002;24:302–316. Corsonello A, Pedone C, Lattanzio F, et al. Regimen complexity and medication nonadherence in elderly patients. Ther Clin Risk Manag. 2009;5:209–216. Salazar JA, Poon I, Nair Mridula. Clinical consequences of polypharmacy in elderly: expect the unexpected, think the unthinkable. Expert Opin Drug Saf. 2007;6:695–704. Vik SA, Hogan DB, Patten SB, et al. Medication Nonadherence and Subsequent Risk of Hospitalisation and Mortality among Older Adults. Drugs Aging. 2006;23:345– 356. Qian J, Simoni-Wastila L, Rattinger GB. Association between depression and maintenance medication adherence among Medicare beneficiaries with chronic obstructive pulmonary disease. Int J Geriatr Psychiatry. 2014;29:49– 57. Sjosten N, Nabi H, Westerlund H, et al. Effect of depression onset on adherence to medication among hypertensive patients: a longitudinal study. J Hypertens. 2013;31:1477–1484. Caughey GE, Preiss AK, Vitry AI, et al. Does antidepressant medication use affect persistence with diabetes medicines? Pharmacoepidemiol Drug Saf. 2013;22:615–622. Dempe C, Junger J, Hoppe S, et al. Association of anxious and depressive symptoms with medication nonadherence in patients with stable coronary artery disease. J Psychosom Res. 2013;74:122–127. Rose AJ, Miller DR, Ozonoff A, et al. Gaps in monitoring during oral anticoagulation: insights into care transitions, monitoring barriers, and medication nonadherence. Chest. 2013;143:751–757. George J, Phun YT, Bailey MJ, et al. Development and validation of the medication regimen complexity index. Ann Pharmacother. 2004;38:1369–1376. Libby AM, Fish DN, Hosokawa PW, et al. Patient-level medication regimen complexity across populations with chronic disease. Clin Ther. 2013;35:385–398. Hirsch JD, Metz KR, Hosokawa PW, et al. Validation of a patient-level medication regimen complexity index as a possible tool to identify patients for medication therapy management intervention. Pharmacotherapy. 2014;34:826– 835. McDonald MV, Peng TR, Sridharan S, et al. Automating the medication regimen complexity index. J Am Med Inform Assoc. 2013;20:499–505.
Address correspondence to: Sunny A. Linnebur, PharmD, FCCP, BCPS, CGP, University of Colorado Skaggs School of Pharmacy and Pharmaceutical Sciences, Mail Stop C238, 12850 East Montview Blvd., Aurora, CO 80045. E-mail: [email protected]
1546
Volume 36 Number 11
S.A. Linnebur et al.
SUPPLEMENTAL APPENDIX Screen shot of main screen of MRCI electronic data capture tool.
November 2014
1546.e1
Original Research
Patient-level Medication Regimen Complexity in Older Adults With Depression Sunny A. Linnebur, PharmD, FCCP, BCPS, CGP1; Joseph P. Vande Griend, PharmD, BCPS, CGP1; Kelli R. Metz, PharmD5; Patrick W. Hosokawa, MS2; Jan D. Hirsch, BS Pharm, PhD3,4; and Anne M. Libby, PhD1,5 1
Department of Clinical Pharmacy, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of Colorado, Aurora, Colorado; 2Colorado Health Outcomes Program, School of Medicine, University of Colorado, Aurora, Colorado; 3Department of Clinical Pharmacy, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of California San Diego, San Diego, California; 4Veterans Affairs of San Diego Healthcare System, San Diego, California; and 5Center for Pharmaceutical Outcomes Research, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of Colorado, Aurora, Colorado
ABSTRACT Purpose: Polypharmacy and medication adherence are well known challenges facing older adults. Medication regimen complexity increases the demands of self-care in the home. Some medication regimens may be more complex than others, especially when dosage form, frequency of dosing, and additional usage directions are included in complexity along with the number of medications In older adults with depression, it is unknown what features of their medications most influence their medication regimen complexity. Methods: A sample cohort of 100 adults ≥65 years old with a diagnosis of depression was randomly selected from electronic medical records (EMR) in ambulatory clinics at the University of Colorado (CU) and University of San Diego (SD). Demographic, medical history, and medication-related information was extracted from the EMR. Complexity was determined using the Medication Regimen Complexity Index (MRCI). IRB approval was obtained. Findings: The cohort mean age was 74.3 years (SD) and 79.7 years (CU). The mean unweighted Charlson comorbidity index for 1.0 (SD) and 1.8 (CU). The
Accepted for publication October 6, 2014. http://dx.doi.org/10.1016/j.clinthera.2014.10.004 0149-2918/$ - see front matter & 2014 Elsevier HS Journals, Inc. All rights reserved.
1538
mean number of medications was 7.1 and 8.0, with 1.1 and 1.2 depression meds, 5.4 and 4.3 nondepression prescription meds, and 0.6 and 2.4 OTC meds for the SD and CU cohorts, respectively. 66% of SD adults and 70% of CU adults took six or more meds. Individual MRCI scores were on average 17.62 (SD) and 19.36 (CU). Dosing frequency contributed to 57-58% of the MRCI score, with patients facing an average of 7–8 unique dosing frequencies in their regimen. In both cohorts, there was an average of 3 additional directions added to the regimens to clarify dosing. Implications: As expected, in our older adult cohorts with depression the majority of patients took multiple medications. Using a standardized instrument, we characterized the regimen complexity and found that it was increasingly complex due to numerous dosing forms, frequencies and additional directions for use. Patient-level medication regimen complexity should go beyond depression medication to encompass the patient’s entire regimen for opportunities to reduce complexity and improve ease of selfcare. (Clin Ther. 2014;36:1538–1546) & 2014 Elsevier HS Journals, Inc. All rights reserved.
Scan the QR Code with your phone to obtain FREE ACCESS to the articles featured in the Clinical Therapeutics topical updates or text GS2C65 to 64842. To scan QR Codes your phone must have a QR Code reader installed.
Volume 36 Number 11
S.A. Linnebur et al. Key words: aged, depression, geriatrics, medication adherence, medication regimen complexity, MRCI.
INTRODUCTION Older adults are often challenged by managing multiple chronic conditions and the many medications associated with those conditions. Medication use by older adults can be even further complicated by the complexity of the medication regimens. Medication regimen complexity has been found to be high in institutionalized elderly—a setting where older adults have assistance in managing their medications.1,2 In ambulatory settings as older adults manage medications on their own, increasing medication regimen complexity increases self-care demands. Complexity of medication regimens can vary by the number of medications, different dosage formulations, dosing frequencies, and additional directions for use. Data indicate that increasing medication regimen complexity and especially frequency of dosing are directly associated with medication nonadherence.2–7 Medication nonadherence can be associated with polypharmacy and self-care demands; these can precipitate adverse events, poor response to medications, poor control of medical conditions, disease state complications, and increased health care burden.8,9 A gap exists in the literature whereby studies of medication regimen complexity have measured complexity associated only with a disease-defined cohort, instead of for the entire patient. An additional gap in knowledge is whether regional variations in care have a significant effect on medication complexity for a given cohort. To characterize patient-level medication regimen complexity for older adults, we chose to evaluate a disease-defined cohort with depression diagnoses and antidepressant prescriptions. As a marker for potential adherence problems, we sought to assess patient-level medication regimen complexity for older adults with treated depression. Depression, a common comorbidity in the older adult, was used because it has been shown to adversely affect medication adherence for common chronic disease states in older adults.10–14 The patients used in this cohort represent patients whose adherence to a medication regimen would likely be affected by a complex medication regimen. We planned to measure medication regimen complexity of the patients’ antidepressant regimens and to
November 2014
contrast it with the medication regimen complexity attributable to all other medications the patients were taking. This would assist in understanding medication regimen complexity as a patient-level construct that may be a key piece of overall patient complexity. Additionally, the integral pieces comprising the patient’s medication regimen complexity could inform practitioners of targets to reduce or simplify regimens and possibly affect adherence. Validated tools for calculating medication regimen complexity exist in the literature. The Medication Regimen Complexity Index (MRCI) is a nondiseasespecific validated tool that measures medication regimen complexity using paper-coded medication regimens from medication lists.15 Scores are derived from weighted values of the regimen components (eg, dosage formulations, dosing frequencies, and specific administration instructions). The MRCI was originally validated in 1 disease-defined cohort (chronic obstructive pulmonary disease) and measured only the prescriptions for that disease state.15 An extension of that work, the parent study of the current study, broadened the validation using multiple disease cohorts, to a patient-level MRCI (pMRCI) measuring medication regimen complexity across all of a patient’s medications for the practical purpose of identifying patients with expected difficulty managing medication regimens and warranting interventions.16,17 The pMRCI is a sum of weighted subscores for 3 mutually exclusive medication categories: diseasespecific prescription medications (ie, antidepressant medications), all other prescription medications, and over-the-counter (OTC) medications recorded on a medication list. This study aimed to use the quantitative tool, the pMRCI, to measure medication regimen complexity and component parts for 2 cohorts of older adults with depression to better characterize the types of medication regimen complexities and possible associated self-care demands faced by older adults in the community. The objectives of this study were to evaluate the entire medication regimen and to determine potential targets to simplify the regimen and improve adherence. We hypothesized that cohorts from ambulatory care clinics at 2 different, geographically diverse, academic medical centers would be similar. We also hypothesized that our cohorts of patients with depression, known in the literature to be relatively nonadherent to medications, would
1539
Clinical Therapeutics have a high contribution to medication regimen complexity from nonantidepressant regimens. We expected to find certain features of the regimens that could offer areas for simplification that could affect adherence.
PATIENTS AND METHODS This study was a retrospective, cross-sectional study of 2 randomly selected cohorts of 100 adults 65 to 89 years of age with treated depression in active care from 2011 to 2012 at ambulatory clinics at the University of California San Diego (referred to later as CA) and the University of Colorado Anschutz Medical Campus (referred to later as CO). Two cohorts from different geographic regions were chosen to increase generalizability of the descriptive findings. Cohort size was determined based on a power calculation for the number needed to compare the distributional metrics of pMRCI values across sites and limited by person-time needed to code the pMRCI. The electronic health record (EHR) was the source of the medical, pharmacy, and patient data for the cohorts. Although using the EHR is a real-world representation of a patient’s active medication list, we sought to use additional EHR data as a conservative approach to improve the accuracy of active treatment. During the data extraction process, an additional step was added in which medications in the EHR were excluded if they had expired after 12 months of no renewal (eg, they were administratively discontinued). The 2 university sites were partners in a larger parent study of medication regimen complexity to examine regional variation in treatment patterns. A full description of the methodologies of the larger study is published elsewhere.16 This study was approved by both the University of California San Diego Human Research Protections Program and University of Colorado Multiple Institutional Review Boards.
Data Source and Definition of Cohorts Deidentified electronic data were used for cohort creation. Inclusion criteria were used to create an index visit from which the medication list was drawn. For each subject, the source of the data was a clinic EPIC-based EHR, drawn from a retrospective clinical data repository managed by a third party. The last visit recorded between July 1 and December 31, 2012 became designated as the index visit. Among the eligible older adults, we included only active patients
1540
defined as having at least 1 visit within the year before the index visit plus the index visit. Inclusion also required a qualifying medical diagnosis of depression (International Classification of Diseases, Ninth Revision 296.2, 296.3, 300.4, or 311) on the “current problem list” and at least 1 antidepressant (identified using National Drug Codes or drug name within the associated Generic Product Identifier [GPI] group 58 antidepressants [Medi-Span Electronic Drug File MED-File v2]) documented in the medication list. Additional medications were evaluated to exclude those medications that had been administratively discontinued. There were no other exclusion criteria. Among all eligible patient records, a random sample was chosen using a number generator to select the study cohorts (N ¼ 100 at both sites) based on power calculations needed to compare means across sites. Once the random sample of 100 records from each clinic was created, extracts were created that included anonymous data on current medication regimens, demographic factors, and active diagnoses. A clinical pharmacist (K.R.M.) coded the pMRCI using an electronic data capture tool that calculated 3 subscores for (1) antidepressant medications, (2) other prescription medications that were not antidepressants; and (3) OTC medications. Primary outcome measures were the overall pMRCI scores and subscores.
MRCI Coding The electronic data capture tool for the MRCI used for coding is publically available at http://www.ucden ver.edu/academics/colleges/pharmacy/Research/research areas/Pages/MRCTool.aspx.16 A screen shot is included as an Supplemental Appendix. The electronic data capture tool allowed the calculation of pMRCI as a simple sum of all 3 weighted subscores based on identical data capture forms for dose formulation, dosage frequency, and additional administration directions. Patient records were assigned a unique anonymous identifier, and their medication regimen was coded. Tool instructions developed by the research team were followed to determine the way in which medications were coded to the 3 medication categories (antidepressants, other prescription medications, and OTC medications) using some decision rules. For example, all antidepressants were coded as antidepressant medications even if they were possibly ordered for the treatment of another
Volume 36 Number 11
S.A. Linnebur et al. condition (eg, trazodone for insomnia). In addition, decision rules were put in place to consistently code for when a drug might be considered a prescription or OTC medication (eg, vitamin D).16
Data Analysis Coded patient-level data were exported from the electronic data capture tool for analysis with SAS version 9.3 software (SAS Institute, Inc, Cary, North Carolina). Age, Charlson comorbidity index, and pMRCI scores were compared between cohorts using an independent-sample t test. Sex was compared via χ2 test. Medication count was evaluated both as a continuous variable and as a categorical variable (1-5, 6-10, Z11) and was compared between cohorts both for medications as a whole and for each subcategory (depression, other prescriptions, OTC medications). Unique medications in the electronic medication list were counted, combining only where it appeared that the medicines were the same prescription but of variable dosages. Medication types were compared between groups by aggregating medications into categories via the first 2 digits of their GPI code. Correlation between continuous variables (medication count, MRCI score, and Charlson comorbidity index score) were evaluated using the Spearman correlation coefficient. The a priori level of significance was set at P o 0.05. The number of unique frequencies and additional directions was generated by counting each instance recorded in the ACCESS data capture tool.
Measures Dependent Variables MRCI Score/pMRCI The original validated MRCI tool was developed by George et al15 to assign complexity scores to a disease-defined medication regimen for individual patients based on the 3 components: (1) dosage forms, (2) dosing frequencies, and (3) additional directions (eg, take at a specified time, take in relation to food). The MRCI tool is a weighted sum with a base case weight of 1 given to tablet/capsule and weight of 1 given to frequency of once daily; thus, the lowest score from MRCI is 2. Other dosage formulations and dosing frequencies are assigned increasing weight related to the increasing difficulty in administration (eg, injectable agents, topical agents, every 4 hours as needed). Additional administration directions (eg,
November 2014
break or crush, take with food, take at a certain time of the day) associated with a medication and documented in the prescription also add to the score and are weighted according to difficulty in administration. pMRCI scores compile complexity scores for each of the 3 components for all of a patient’s medications.
RESULTS Table I presents population characteristics for the patient samples drawn from California (CA) and Colorado (CO). Characteristics were compared between the geographic groups to evaluate for similarities and differences between the groups. The mean age of the patients was in their 70s, and the majority were female. The patients in both cohorts were fairly healthy based on the Charlson comorbidity index, with the CO patients having slightly more comorbidities on average than the CA patients (P o 0.0001). The most common of the Charlson comorbidities was diabetes mellitus in CA and chronic obstructive pulmonary disease in CO, with the second most common in both cohorts being cancer. The average total count of medications was slightly, but not significantly, lower in CA (mean, 7.1) than in CO (mean, 8.0) (P ¼ 0.10). The difference appeared to be driven by the only significant difference between the cohorts: more OTC medications in the CO patients (mean, 0.6 [CA] compared with mean, 2.4 [CO], P o 0.0001). The distribution of polypharmacy (defined as medication count 45 medications) was also similar between groups with approximately two-thirds of all patients taking 45 medications. Medication counts ranged from 1 to 19 in the CA cohort and 2 to 18 in the CO cohort. Similarly, the pMRCI scores ranged from 2 to 47 and 4 to 48, respectively. Medication count (Figure 1) and Charlson comorbidity index (data not shown) were significantly associated with pMRCI score (r ¼ 0.92, P o 0.001 and r ¼ 0.22, P ¼ 0.0016, respectively). As expected, depression-related medications did not contribute as significantly to the medication regimen complexity compared with other prescription medications. In the CO cohort, OTC medications contributed more to the pMRCI than depression-related medications. Use of OTC agents in the Colorado cohort was common and, as such, was evaluated for potential indication (Figure 2). Conditions that were most
1541
Clinical Therapeutics
Table I. Population characteristics and medication use.
No. of patients Demographic characteristics Age, mean (SD) Sex (% female) Clinical characteristics Charlson comorbidity index, range, 0–17 Most common comorbidity group
California
Colorado*
100
100
74.3 (7.4) 76 0.9 (0–5) DM without complications, 18% Cancer, 17% 17.62 (10.05) 7.1 (1–19)
P
79.7 (6.1) 81
o0.0001 0.39
1.8 (0–6) COPD, 35%
o0.0001 N/A
Second most common comorbidity group Cancer, 30% N/A Mean pMRCI 19.36 (10.21) 0.21 Mean medication count, all (range) 8.0 (2–18) 0.07 By category Mean no. of depression medications (range) 1.1 (1–3) 1.2 (1–3) 0.02 Mean no. of other prescription medications, (range) 5.4 (0–15) 4.3 (0–14) 0.03 Mean no. of over-the-counter medications (range) 0.6 (0–6) 2.4 (0–11) o0.0001 By quantity 1–5 medications, % 34 30 0.78 6–10 medications, % 49 50 Z11 medications, % 17 20 Most common GPI group (% of total prescriptions) Antidepressants, 16.3% Antidepressants, 15.7% N/A Second most common GPI group Antihyperlipidemic, 7.0% Antihyperlipidemic, 7.2% N/A Third most common GPI group Antihypertensive, 6.1% Antihypertensive, 6.9% N/A COPD ¼ chronic obstructive pulmonary disorder; DM ¼ diabetes mellitus; GPI ¼ generic product identifier; N/A ¼ statistical testing not applicable, presented for descriptive purposes; pMRCI ¼ patient-level Medication Regimen Complexity Index. * Some cohort descriptive statistics were presented in Table 1 of Libby et al (2013)16 for the same depression cohort; additional descriptive information is presented in original form here.
frequently treated with OTC medications included heart health, constipation, pain, and indigestion. Figure 3 displays the average pMRCI scores by site, with colors to differentiate dosage formulation, dosing frequency, and administration directions; vertical bars also distinguish the pMRCI subscores for antidepressant medications, other prescription medications, and OTC medications. Table II provides the mean values for the total pMRCI for the CA and CO patients, along with the mean scores for each portion of the pMRCI calculation. The pMRCI for both patient groups comprised mostly points from the frequency of use and was due to the high number of medications recorded for each patient. Frequency of dosing was most often not limited to once-daily dosing. In the CA and CO patients, the mean number of dosing frequencies per patient, calculated for the 3 categories of
1542
medications, was 6.98 (SD 3.72) and 7.73 (SD 3.49), respectively. On average, approximately 3 “additional directions” for use (eg, break/crush tablet) were recorded per patient in both cohorts. The most common additional direction in both groups was to take the medication at a specified time of day. More than one-half of all patients in both groups had more than 1 dosage form; the most common additional formulation beyond tablet/capsule was a cream (data not shown). Six percent (CA) and 11% (CO) of patients had Z3 distinct dose formulations on their medication list (data not shown).
DISCUSSION Using medication lists and active problem codes in EHRs, pMRCI scores were calculated and determined to be similar across 2 academic ambulatory clinics
Volume 36 Number 11
S.A. Linnebur et al.
MRCI 60
50
40
30
20
10
0 0
5
10 15 Medication Count Facility
California
20
25
Colorado
Figure 1. Medication Regimen Complexity Index (MRCI) and medication count.
located in CA and CO, suggesting that our description of this population has external validity. The pMRCI expanded the assessment of medication regimen complexity to include all prescription and nonprescription
Other 24%
Heart Health 22%
Indigestion (mild to severe) 15%
Constipation 22%
Pain 17%
Figure 2. Conditions treated with over-the-counter (OTC) agents.
November 2014
medications recorded in the medical record. Similar to Mansur et al,2 the majority of older adult patients in our 2 samples had documented polypharmacy, and their pMRCI scores reflected this. As expected, depression treatment did not contribute greatly to polypharmacy or medication regimen complexity. Rather, the pMRCI collected all other prescription and OTC medications to give a complete representation of medication regimen complexity and potentially enable a pharmacist to target sections of the medication regimen in older adults with depression to reduce complexity. When pMRCI score components were evaluated, it was found that dosing frequency contributed quantitatively the most to the patients’ pMRCI scores, with many patients being prescribed multiple unique frequencies for dosing. In addition, many patients had additional instructions for taking their medications beyond frequency (eg, take at a specific time of day). Multiple formulations, several unique dosing frequencies, and additional instructions likely complicate any patient’s ability to maintain appropriate and consistent medication administration practices. In the case of older adults with depression, the demands of maintaining adherence may be further complicated by limitations in vision, dexterity, or memory.
1543
Clinical Therapeutics
Mean Score 20
15
10
5
0
Disease
Other Rx OTC California Component
Total Score
Dosage Forms
Disease
Frequency
Total Score
Other Rx OTC Colorado Additional Directions
Figure 3. Mean patient-level Medication Regimen Complexity Index at California and Colorado clinics. OTC ¼ over-the-counter; Rx ¼ prescription.
The use of the pMRCI scores to assess medication regimen complexity greatly enhances the understanding of what older adults with depression may face with medication administration and adherence, beyond a simple pill count or pill burden. The pMRCI may prove to have utility as a risk prediction tool for medication regimen complexity and nonadherence, perhaps similar to less tailored metrics like the number of medications and the number of comorbidities as used in some medication therapy management programs. More research is needed to investigate the association of
pMRCI scores with adherence and patient outcomes. This was not possible given our data source, but is the next important step in this work to link the pMRCI to patient outcomes. In older adults, other factors such as social, demographic, and concomitant conditions could also play a large role in clinical outcomes. In addition, to be used as a point of care pMRCI tool for providers or patients, it is likely that the current electronic scoring version in ACCESS would need to be simplified to make the tool feasible and acceptable at the point of care.
Table II. Patient-level Medication Regimen Complexity Index) at California and Colorado clinics. California (n ¼ 100) Mean (SD) Medication count Total pMRCI Dosage form Frequency Additional directions
7.10 17.62 4.42 10.19 3.01
(3.73) (10.05) (2.78) (6.39) (2.67)
Range (IQR) 1-19 2-47 1-15 1-30.5 0-15
(4.5–9.5) (10.5–23.25) (2–6) (6–13.25) (1–5)
Colorado (n ¼ 100) Mean (SD) 7.96 19.36 5.43 11.06 2.87
(3.63) (10.21) (3.05) (6.08) (2.44)
Range (IQR) 2-18 4-48 1-15 2-30 0-10
(5–10) (11.75–25.5) (3–7) (6.5–14.5) (1–4)
IQR ¼ interquartile range; pMRCI ¼ Patient-level Medication Regimen Complexity Index.
1544
Volume 36 Number 11
S.A. Linnebur et al. Based on the high level of medication regimen complexity in both of our patient groups, our results suggest a need for pharmacist review of the patient’s entire medication regimen, not just disease-specific or prescription-only medications. The goal of this medication review would be to assess and reduce complexity to a manageable level for the patient if possible. Pharmacist review could identify opportunities to reduce the number of unnecessary prescription and OTC medications and to use once-daily regimens and reduced administration instructions. In older adults, reducing the number of dosage formulations may not be feasible, given the potential benefit for reduced systemic adverse events with the use of certain nonoral dosage forms (eg, topical or transdermal forms). In fact, our results indicate that creams are already being used at a high rate in the older adults with depression. Pharmacist review could also investigate actual nonadherence issues that patients may be experiencing due to specific complexities in their medication regimen and provide more targeted patient education about proper administration and how to overcome adherence-limiting regimen issues. Only 1 significant difference in medication use between CA and CO was observed; OTC use was higher on average among the CO patients. One possible reason for this is that the CO providers documented the OTC use at a higher rate than the CA providers. Data collection in this study was limited by what was documented in the electronic medical record; thus, it is very likely that the use of OTC medications was higher than documented in both study groups. Another possible reason for the difference in OTC use between cohorts is that patients in CO may be more interested in using nonprescription treatments. This study has several imitations. First, there was no direct measure that the medication was consumed by the patients, only that it was recorded as prescribed or continued at the index visit. By clinic protocols, however, each visit at both sites included a medication reconciliation component. The requirement of each patient record to include at least 2 visits in the year should have provided an opportunity for medication reconciliation to identify any medication changes and allow for optimal accuracy in the medication record. The pMRCI was hand-coded by a pharmacist using an Access database. Automated coding at the point of care would be much more clinically relevant for practitioners. McDonald et al18 recently published
November 2014
their successful attempt at using an automated tool to calculate the pMRCI from 90,000 EHRs. Their findings were very similar in terms of the mean number of medications and pMCRI for the population that they studied, indicating our handcoding and their automated coding may be consistent. Use of an automated system like that of McDonald et al could also reduce individual coding error and quickly provide the practitioner with validated complexity information at the point of making medication changes. Using this information has the potential to improve medication administration, adherence, adverse events, and other patient outcomes if interventions and education are used along with the pMRCI. In addition, pharmacists could use an automated pMRCI when they perform medication therapy management visits and medication reconciliation in the ambulatory and community setting. Finally, the results of this study are likely generalizable to other ambulatory older adults with depression but may not apply to older adults without depression.
CONCLUSIONS In this study, EHRs of older adults with depression from 2 distinct ambulatory clinics were evaluated using the pMRCI tool to assess medication regimen complexity. Medication regimen complexity and its components were very similar, despite geographic differences. On average, the patients took 45 medications per day, including OTC medications, and their medication regimen complexity directly correlated with the number of medications and comorbidities. The most influential features in the patient-level MRCI scores were dosing frequencies and dosage forms, and these scores were derived from the nondepression prescription medications and OTC agents. Using this information may provide a mechanism to positively reduce the medication regimen complexities of older adults with depression. Future research with the pMRCI tool could focus on assessing medication regimen complexity with an automated pMRCI tool at the point of care, which portion of the pMRCI relates the most to adherence, and the clinical effects of reducing medication regimen complexity in older adults with depression. Opportunities may be available to reduce medication regimen complexity, improve ease of self-care, improve adherence, and reduce adverse drug events in older adults.
1545
Clinical Therapeutics
ACKNOWLEDGMENTS This research was funded by The ALSAM Foundation Skaggs Scholars Program at the University of Colorado Skaggs School of Pharmacy and Pharmaceutical Sciences. Drs. Linnebur, Vande Griend, and Metz were responsible for the literature search, study design, data collection/coding, data interpretation, and writing of the manuscript. Mr. Hosokawa was responsible for the data analysis, data interpretation, figure and table creation, and writing of the manuscript. Dr. Hirsch was responsible for the literature search, study design, data interpretation, and writing of the manuscript. Dr. Libby was responsible for the literature search, study design, data analysis, data interpretation, and writing of the manuscript.
7.
8.
9.
10.
11.
CONFLICTS OF INTEREST The authors have indicated that they have no conflicts of interest regarding the content of this article.
12.
SUPPLEMENTAL MATERIAL Supplemental materials accompanying this article can be found in the online version at http://dx.doi.org/10. 1016/j.clinthera.2013.12.011
13.
REFERENCES 1. Advinha AM, de Oliveira-Martins S, Mateus V, et al. Medication regimen complexity in institutionalized elderly people in an aging society. Int J Clin Pharm. 2014;36:750– 756. 2. Mansur N, Weiss A, Beloosesky Y. Looking beyond polypharmacy: quantification of medication regimen complexity in the elderly. Am J Geriatr Pharmacother. 2012;10:223–229. 3. Conn VS, Taylor SG, Kelley S. Medication regimen complexity and adherence among older adults. J Nurs Scholarship. 1991;23:231–235. 4. Ingersoll KS, Cohen J. The impact on medication regimen factors on adherence to chronic treatment: a review of literature. J Behav Med. June 2008;31:213–224. 5. Balkrishnan R. Predictors of medication adherence in the elderly. Clin Ther. 1998;20:764–771. 6. Iskedjian M, Einarson TR, MacKeigan LD, et al. Relationship between daily dose frequency and adherence to
14.
15.
16.
17.
18.
antihypertensive pharmacotherapy: Evidence from a meta-analysis. Clin Ther. 2002;24:302–316. Corsonello A, Pedone C, Lattanzio F, et al. Regimen complexity and medication nonadherence in elderly patients. Ther Clin Risk Manag. 2009;5:209–216. Salazar JA, Poon I, Nair Mridula. Clinical consequences of polypharmacy in elderly: expect the unexpected, think the unthinkable. Expert Opin Drug Saf. 2007;6:695–704. Vik SA, Hogan DB, Patten SB, et al. Medication Nonadherence and Subsequent Risk of Hospitalisation and Mortality among Older Adults. Drugs Aging. 2006;23:345– 356. Qian J, Simoni-Wastila L, Rattinger GB. Association between depression and maintenance medication adherence among Medicare beneficiaries with chronic obstructive pulmonary disease. Int J Geriatr Psychiatry. 2014;29:49– 57. Sjosten N, Nabi H, Westerlund H, et al. Effect of depression onset on adherence to medication among hypertensive patients: a longitudinal study. J Hypertens. 2013;31:1477–1484. Caughey GE, Preiss AK, Vitry AI, et al. Does antidepressant medication use affect persistence with diabetes medicines? Pharmacoepidemiol Drug Saf. 2013;22:615–622. Dempe C, Junger J, Hoppe S, et al. Association of anxious and depressive symptoms with medication nonadherence in patients with stable coronary artery disease. J Psychosom Res. 2013;74:122–127. Rose AJ, Miller DR, Ozonoff A, et al. Gaps in monitoring during oral anticoagulation: insights into care transitions, monitoring barriers, and medication nonadherence. Chest. 2013;143:751–757. George J, Phun YT, Bailey MJ, et al. Development and validation of the medication regimen complexity index. Ann Pharmacother. 2004;38:1369–1376. Libby AM, Fish DN, Hosokawa PW, et al. Patient-level medication regimen complexity across populations with chronic disease. Clin Ther. 2013;35:385–398. Hirsch JD, Metz KR, Hosokawa PW, et al. Validation of a patient-level medication regimen complexity index as a possible tool to identify patients for medication therapy management intervention. Pharmacotherapy. 2014;34:826– 835. McDonald MV, Peng TR, Sridharan S, et al. Automating the medication regimen complexity index. J Am Med Inform Assoc. 2013;20:499–505.
Address correspondence to: Sunny A. Linnebur, PharmD, FCCP, BCPS, CGP, University of Colorado Skaggs School of Pharmacy and Pharmaceutical Sciences, Mail Stop C238, 12850 East Montview Blvd., Aurora, CO 80045. E-mail: [email protected]
1546
Volume 36 Number 11
S.A. Linnebur et al.
SUPPLEMENTAL APPENDIX Screen shot of main screen of MRCI electronic data capture tool.
November 2014
1546.e1
