Abstract
For those with high blood pressure, diabetes, or high cholesterol, adherence to a home medication regimen is important for health. Reductions in the number of daily medication-taking events or daily pill burden improve adherence. A novel advice-giving computer application was developed using the SMART platform to generate advice on how to potentially simplify home medication regimens. MedMinify generated advice for 41.3% of 1,500 home medication regimens for adults age 60 years and older with chronic medical conditions. If the advice given by MedMinify were implemented, 320 regimen changes would have reduced daily medication-taking events while an additional 295 changes would have decreased the daily pill burden. The application identified four serious drug-drug interactions and so advised against taking two pairs of medications simultaneously. MedMinify can give advice to change home medication regimens that could result in simpler home medication-taking schedules.
Introduction
To minify is to reduce the amount of something1. We studied how to use information technology to help minify the number of times consumers have to take medications each day and the quantity of pills they have to take. It is important to simplify home medication regimens in these ways to facilitate medication adherence. While simpler home medication-taking schedules may diminish medication mishaps2, this research is justified on the basis that better adherence resulting from simplified medication regimens improves the health and well being of consumers.
For individuals diagnosed with high blood pressure, high cholesterol, or diabetes, adherence to home medication regimens is known to improve health3. Conversely, lack of adherence to home medication regimens for chronic conditions causes considerable harm4. The overall cost of non-adherence to medication regimens has been estimated to be $100 billion per year in the United States5. Several categories of strategies to increase medication adherence have proven modestly effective5. These include strategies to improve communication of what to take and why, strategies to support behavior change such as self-monitoring of adherence, strategies to decrease out-of-pocket costs of medications, and strategies to cue consumers to take their medications such as reminder systems and calendars5. The most effective approaches combine strategies from multiple categories5. However, in terms of effect size, the evidence indicates that greater adherence improvements result from decreasing the complexity of home medication regimens than from combining educational, behavioral, and cueing interventions5. One strategy to decrease the complexity of medication regimens is to standardize the times of daily medication-taking events and limit their number to a maximum of four6. Another strategy is for health care providers to conduct medication regimen reviews intended to simplify medication regimens7. To facilitate these latter two strategies, we have developed an advice giving system to assist in the process of simplifying home medication regimens.
This paper describes the architecture and early stage evaluation of MedMinify, a new application that offers advice on how a home medication schedule could be made simpler by reducing the number of daily medication-taking events and the daily pill burden while accounting for drug-drug interactions. Because problems with medication adherence are widespread, MedMinify was developed to interoperate with a broad range of electronic health record systems. Our intent is to avoid the need to recreate the functionality of MedMinify in every different electronic health record (EHR) system. For the purpose of simplifying regimens, MedMinify advises users as to the availability of sustained-release or fixed-dose combination drug products and indicates whether certain drug interactions pertain. The literature supports the potential effectiveness of regimen simplification to improve medication adherence7,8.
Background and Significance
In a systematic review of the association between the number of daily medication-taking events and adherence, Claxton, Cramer, and Pierce found mean adherence to a medication regimen declined as daily medication-taking events increased. Mean adherence for one, two, three, and four daily medication-taking events was 79% (SD 14), 69% (SD 15), 65% (SD 16), and 51% (SD 20) respectively8. The mean adherence differences between one daily medication-taking event and either three or four daily medication-taking events were statistically significant as was the mean difference between two daily medication-taking events and four daily medication-taking events8. In light of more recent concurring evidence, these findings indicate that decreasing daily medication-taking events results in improved adherence9,10.
Controlled studies of a standardized home medication schedule where patients take all of their medications either at breakfast, lunch, dinner or bedtime, have shown that in most cases consumers do not need to take medications for many common chronic conditions more than four times a day11–13. These studies suggest there is an opportunity to minify the number of daily medication-taking events12.
Adherence to home medication regimens is also influenced by the number of pills taken per day or daily pill burden. To minify the number of pills taken daily can also be seen as a necessary adherence intervention10,14. In a recent meta-analysis, higher pill burden was associated with lower medication adherence to medication regimens that treat human immunodeficiency virus infection10. In another study agreement with the statement, “I am already taking too many medications”, was a predictor of very low adherence amongst women being treated for osteoporosis15.
One difficulty in minifying daily medication-taking events and daily pill burden to simplify home medication schedules is that consumers may actually increase the number of scheduled daily medication-taking events on their own for a variety of reasons16. A partial explanation for this behavior could be concern over drug-drug interactions16. Consumers may assume that taking oral medications at different times of day mitigates adverse events from drug interactions. Actually, relatively few drug interactions result from simultaneous oral ingestion17. More often drug interactions result from systemic effects at the sites in the body where drug metabolism, pharmacologic action, or drug excretion take place17. To counteract their largely false beliefs, what consumers may need is explicit information regarding which drugs they should and should not ingest simultaneously.
Previous research has examined the use of information technology to assist in reducing medication regimen complexity and polypharmacy18. Others have reported on the automated generation of pill cards and printed home medication schedules13. We have used information technology differently to give advice on how to minify daily medication-taking events and daily pill burden while ensuring that the advice given is in concordance with what is known about the relatively few serious drug-drug interactions that result from the simultaneous ingestion of drugs.
This research is significant for several reasons. While the majority of patients take multiple medications for multiple conditions, many interventional trials have focused on adherence to single medication therapies or to treatments for one condition. We intentionally sought to analyze prescriptions for all scheduled tablets and capsules consumers must take to adhere to their home medication regimens. In addition, whereas pharmacist-led medication regimen review helps to simplify medication regimens, a lack of pharmacist time to conduct home medication regimen reviews limits this intervention7. An information resource that gives advice on how to minify daily medication-taking events and daily pill burden may decrease the amount of time required to review home medication regimens for simplification opportunities. Such advice may also apply during medication reconciliation tasks. Finally, to facilitate future integration with EHR systems we set a goal to build MedMinify so that it would interoperate with a variety of EHR systems from the outset.
Motivated by these research findings and the clinical significance of the problem of medication adherence, this paper reports whether an advice giving system capable of offering potentially useful advice to simplify medication regimens could be built using the Substitutable Medical Apps, Reusable Technology (SMART) platform, how frequently potentially useful advice to assist in simplifying medication regimens could be generated, and in what ways and by how much the advice generated, if heeded, could change actual home medication regimens.
Research Questions
The following three research questions were investigated.
Is the Substitutable Medical Apps, Reusable Technology (SMART) platform robust and extensible enough to support the architectural and functional requirements of a system capable of giving advice on how to potentially minify daily medication-taking events and daily pill burden?
What types of advice and how much of each type can be generated by analyzing home medication regimens with an information resource capable of identifying candidate regimens for minification of daily medication-taking events and daily pill burden while accounting for drug-drug interactions from simultaneous ingestion?
To what extent would the requirements of home medication regimens change if advice given by a computer application on how to minify daily medication-taking events and daily pill burden were implemented?
Methods
SMART Application Development
MedMinify was developed to meet the following requirements. The application had to be able to access and analyze prescription data organized as medication regimens for individuals. To provide drug product advice, MedMinify had to be able to identify the active ingredients in all current prescriptions for an individual and check for other substitutable drug products containing the same active ingredients. To ascertain whether drug-drug interactions pertain, MedMinify required a drug-drug interaction knowledge table and the routines to check whether the active ingredients in the current prescriptions for an individual could be found in the drug-drug interaction knowledge table. Finally, to facilitate regimen review by the investigators, MedMinify had to render key facts about each individual’s regimen within a browser window along with any advice that the application generated.
Software design and application development of MedMinify involved the SMART platform (smartplatforms.org), the Django Web framework (Version 1.5, djangoproject.com), and the RxNorm application programming interfaces (rxnav.nlm.nih.gov/RxNormAPIs.html#). The organization of the MedMinify application is depicted in Figure 1. The SMART Reference EMR was used for development, testing, and evaluation of MedMinify. SMART afforded data models for individuals and their corresponding prescription records with an application programming interface (API) for the SMART Reference EMR. The Django Web framework afforded a Python-based Web development environment integrated with SQLite database technology. The Django Web framework also included a host of ready to use tools that enabled the development, testing, and rapid iterative refinement of the application. The RxNorm APIs afforded automated lookup of drug products marketed in the United States and their ingredients. MedMinify used Javascript executed in a browser along with Hyper Text Markup Language and Cascading Style Sheets to call the SMART API, render and format application output, respectively (Figure 1).
Figure 1.
The MedMinify application architecture with a partial screen shot of its user interface. At the top of the image is a cutout of part of a user’s view in a browser with counts of the Medications, Pill Burden, and Different Pills for one regimen. A local instance of the SMART Reference EMR platform was loaded with 1,500 sampled multi-drug enteral medication regimens (Rx). MedMinify used the Django Web framework and a SQLite database. Drug-drug interaction knowledge was loaded into the MedMinify SQLite database.
Several field value conversions were undertaken to convert data from the University of Michigan’s EMR to the formats used by the SMART Reference EMR. These data conversions were made using rules in a spreadsheet. For example, the quantity of each dosage form to be administered was not a discrete field in the University of Michigan’s EMR prescription data but the SMART EMR accepted a discrete value for the quantity of pills to be administered. The quantity of pills to take at each medication-taking event was extracted from the instructions field in the original University EMR prescription data. A prescription instruction “Take 2 tablets daily” resulted in an extracted numeric quantity value of “2” for use by SMART. Similar data conversions were conducted for the frequency data from the University’s EMR to transform the frequencies provided into the format used by SMART.
To support the drug interaction identification function of MedMinify, knowledge from drug interaction monographs was used (First DataBank, Inc. (FDB), San Francisco, CA). FDB provided the investigators with select fields from a complete set of their highest severity (FDB severity level 1) drug interaction monographs. The fields provided were the drug interaction monograph title, drug interaction mechanism, and clinical effects fields. Keyword analysis of the FDB knowledge source was conducted to select only those drug interactions related to simultaneous ingestion of medications. The 34 keywords included “gut”, “intestine”, “bioavailability”, “absorption”, “gastric”, and “pH”. A list of 59 unique, severity level 1 drug-drug interactions that are mechanistically associated with simultaneous ingestion was developed. After determining the drug pairs associated with each of the selected 59 drug-drug interactions and removing duplicates, a list of 559 interacting, active drug ingredient pairs for interactions due to simultaneous ingestion was created. For each active drug ingredient involved, the RxNorm ingredient (IN) code was identified. A free software application was used to load this drug interaction ingredient pair knowledge into the MedMinify application database (SQLiteStudio 2.1.5 by Pawel Salawa). An example row from this drug-drug interaction knowledge table includes the antacid omeprazole (RxNorm IN = 7646) and the antiviral drug atazanavir (RxNorm IN = 343047). It is known that omeprazole lowers the acidity of the gastric juices decreasing the solubility of atazanavir, disrupting atazanavir absorption, and thereby diminishing the potential efficacy of atazanavir.
After application development and testing using the publicly available SMART Reference EMR online sandbox, a local copy of the SMART Reference EMR was installed and made operable on an Apple Mac Mini computer running Mac OS version 10.8.5. A fictional name generator (http://homepage.net/name_generator/) was used to generate 1,500 fake names for loading the 1,500 randomly sampled regimens into the local SMART instance for use with MedMinify. After the 1,500 fictitious patients and corresponding regimens were loaded into SMART, data validation occurred by examining the SMART database. MedMinify application testing and use was conducted using an Apple MacBook Pro laptop running Mac OS version 10.8.5 connected via a local area network to the local SMART instance. The Safari browser was used (version 6.1.1).
Data source and sample
This study evaluated the function of a computer application we developed called MedMinify. To conduct the evaluation we used deidentified prescription data from the electronic medical record (EMR) system at the University of Michigan Hospitals and Health Systems (EpicCare Ambulatory Electronic Medical Record, Epic, Verona, WI). The Institutional Review Board of the University of Michigan reviewed and approved this study. Up to date medication regimens, provided as records of prescriptions for 41,903 unique individuals, were selected via query using the following criteria, a–c:
Each individual whose medication regimen was selected was alive and of a chronological age greater than or equal to 60 years on the date of the query, January 15, 2014.
Each individual whose medication regimen was selected had at least one medication list update documented in the University of Michigan’s EMR during calendar year 2013 or the first two weeks of 2014 to ensure regimen currency.
Each individual’s University of Michigan EMR profile had an International Classification of Diseases (ICD version 9) coded diagnosis for hypertension (401.x, 405.x, 416.x, 459.3x) and/or diabetes (249.x, 250.x, 253.5) and/or hypercholesterolemia (272.x).
These three criteria were specifically chosen to provide real medication regimens for adults using multiple medications to treat common diseases. This query of the University of Michigan’s EMR resulted in data for 380,932 prescriptions. Besides patient identifiers such as age and gender, these data included prescription fields for ordering date, prescription start date, prescription end date, generic drug name, drug product strength, frequency, medication dosage form, route of administration, quantity to dispense, number of refills, medication therapeutic subclass, prescriber instructions, and RxNorm Semantic Clinical Drug code. Not all of these data were usable. A data exclusion diagram for the 380,932 prescriptions is given in Figure 2.
Figure 2.
Prescription data was excluded for expired prescriptions, prescriptions for non-drug items, non-enteral prescriptions, take-as-needed prescriptions, single event prescriptions, and prescriptions with indeterminate frequencies. Prescriptions missing frequencies, product strengths, or RxNorm Semantic Drug codes were also removed.
After exclusions were applied and removal of prescriptions missing data, 123,845 active prescriptions remained for 35,042 unique individuals for scheduled and recurring home medications with known product strengths administered via an enteral route (e.g., orally or sublingually). Next, 8,105 complete medication regimens consisting of only one medication were also excluded leaving 115,740 prescriptions comprising the home medication regimens of 26,937 individuals with more than one scheduled and recurring medication of known product strength administered via an enteral route. These data included enteral medications prescribed for use on a weekly (e.g., “three times a week”) or monthly (e.g., “once a month”) schedule. For this study, these 26,937 medication regimens represent the population of qualified multi-drug enteral medication regimens used to treat chronic hypertension, hypercholesterolemia and diabetes. The 26,937 regimens include regimens for 14,403 females (53.5%) and 12,534 males (46.5%). The 26,937 individuals included had an average age of 71 years (SD 8.3, Range 60 to 103).
Because each regimen required approximately 10 seconds of time to analyze with MedMinify and 30 seconds more to assess if advice was given, instead of analyzing all 26,937 regimens a random sample of 1,500 regimens consisting of 6,241 prescriptions was taken from the population of 26,937 qualified regimens using SPSS (version 21). Random sampling was preferred over a selective sample of the most complex regimens because we believed random sampling to be the most conservative test of MedMinify’s capabilities. The sample size of 1,500 regimens was chosen to give a desired margin of error of 0.025 for 95% confidence intervals for Chi-squared tests with one degree of freedom19. Sampling was done without replacement resulting in a hypergeometric distribution. However because the population of 26,937 is much larger than the sample size of 1,500 it was assumed that the binomial distribution provided a reasonably good approximation for calculating proportion confidence intervals. Therefore the Agresti-Coull adjusted Wald interval for large sample sizes was used to calculate the confidence intervals for sample proportions reported in this paper20.
Generation of advice to minify daily medication-taking events or daily pill burden using a computer application
All 1,500 randomly sampled regimens were analyzed using MedMinify. For each regimen, MedMinify calculated the number of prescriptions in the regimen, the daily pill burden, the number of unique pills in the regimen, and the maximum number of daily medication-taking events required by the regimen. Next, for every immediate-release pill taken more than once a day, MedMinify checked RxNorm to see whether a sustained-release product was available as a potential substitute. If so, MedMinify recommended that the sustained-release product or products be considered as substitutes. Subsequently, for every pair of active ingredients prescribed as two single ingredient products, MedMinify checked RxNorm to see whether a fixed-dose combination drug product was available that combined two active ingredients in one pill. If so, MedMinify recommended that the fixed-dose combination product or products be considered as substitutes. Next, MedMinify screened every pair-wise combination of active drug ingredients in each regimen against the table of known FDB severity level 1 interacting drug-drug pairs where the drug-drug interaction is associated with simultaneous ingestion. When a potential drug-drug interaction was identified, MedMinify gave advice not to take the two interacting drugs at the same time of day. Finally, MedMinify counted all instances when a single active ingredient in a regimen appeared in the drug-drug interaction knowledge table (i.e., one drug of an interacting pair) even where no drug interaction was found. Because this was a lab function study, there were no consumer end users and none of the advice generated was ever implemented.
Quantification of how much home medication regimen requirements would change if the computer-generated advice were implemented
All advisories given by MedMinify to change sampled medication regimens were documented and counted. Assuming MedMinify’s advice was accepted and implemented, the potential change or changes to each regimen where advice was generated was assessed in two ways. First, if a sustained-release product or products was recommended, the pharmacist investigator (AJF) checked to see whether the maximum number of medication-taking events per day could be reduced, and, if so, by how many daily events. (The opportunity to minify the maximum number of medication-taking events per day depends on whether the recommendation to substitute a sustained-release product applies to the prescription with the most daily medication-taking events.) Next, if either a sustained-release or a fixed-dose combination product or both were recommended by MedMinify, the pharmacist investigator checked to see whether the daily pill burden could be reduced, and if so by how much. When overlaps implementing this advice arose, for example when the same active ingredient could be taken via a sustained-release drug product or as a component of a fixed-dose combination drug product, preference was given to minifying the number of daily medication-taking events over minifying the daily pill burden.
Results
1. Is the Substitutable Medical Apps, Reusable Technology (SMART) platform robust and extensible enough to support the architectural and functional requirements of a system capable of giving advice on how to potentially minify daily medication-taking events and daily pill burden?
The MedMinify application is an example of a SMART Connect application (Figure 1). SMART Connect provides the SMART JavaScript client for the purpose of communicating from a SMART container via the SMART API when connecting to the SMART Reference EMR with a browser. To meet MedMinify’s requirement for access to prescription data organized as regimens, SMART permitted MedMinify access to prescription data stored in its database with one minor limitation. The SMART Reference EMR data model could not fully represent prescriptions with specific days and times in coded fields (e.g., “Take 1 tablet at 2:00 p.m. on Tuesdays and Thursdays.”). A few prescriptions had to be represented in less detail with respect to dose timing upon conversion to SMART. To meet the requirement for RxNorm integration, procedural code written in a high-level programming language, Python, was used to manage calls and responses via the RxNorm prescribable APIs. The Django Web framework played a central role in the architecture of MedMinify. To meet the requirement for integration of drug-drug interaction knowledge, the Django Web framework provided an integrated SQLite database. Finally, to meet the requirement for rendering output to the screen within the SMART container, the Django Web framework provided a web server to serve various files including index.html and analyzed.html (Figure 1). The combination of SMART and the Django Web framework afforded a stable, extensible platform upon which MedMinify was successfully built.
2. What types of advice and how much of each type can be generated by analyzing home medication regimens with an information resource capable of identifying candidate regimens for minification of daily medication-taking events and daily pill burden while accounting for drug-drug interactions from simultaneous ingestion?
Three sets of results address the second research question, quantification of the overall amount of advice generated, categorization of the generated advice by type, and an analysis of MedMinify’s drug-drug interaction checking function. Descriptive statistics and results quantifying the number of recommendations given as advice for minifying the 1,500 sampled regimens are given in Table 1 where the sampled regimens are categorized according to the maximum number of daily medication-taking events.
Table 1.
Sampled qualified multi-drug enteral medication regimens (n = 1,500) data for individuals of age greater than or equal to 60 years with diagnoses for hypertension or diabetes or hypercholesterolemia (alone or in combination) according to the maximum number of daily medication-taking events per regimen.
| Number of Daily Medication-Taking Events Category | Number of Sampled Regimens | Number of Sampled Regimens with Recommendations from MedMinify | Average Number of Prescriptions per Sampled Regimen | Average Daily Pill Burden per Sampled Regimen |
|---|---|---|---|---|
| Max. 8 events daily | 1 | 0 | 7.0 | 26.0 |
| Max. 6 events daily | 3 | 2 | 5.3 | 15.6 |
| Max. 5 events daily | 4 | 2 | 6.3 | 15.8 |
| Max. 4 events daily | 44 | 22 | 5.4 | 11.3 |
| Max. 3 events daily | 178 | 126 | 5.7 | 10.4 |
| Max. 2 events daily | 643 | 388 | 4.6 | 6.7 |
| Max. 1 event daily | 626 | 79 | 3.2 | 3.4 |
| Max. 0* events daily | 1 | 0 | 2.0 | 0.0 |
| TOTALS | 1,500 | 619 | ||
One regimen included only prescriptions for medications to be taken weekly and no medications to be taken daily.
MedMinify identified 619 (41.3%) of the 1,500 randomly sampled regimens, 95% CI [38.7, 43.7] as candidates for minifying daily medication-taking events or daily pill burden.
In Table 2 below, the 619 regimens where at least one recommendation to change the regimen was given are further categorized by the type of recommendation. Population estimates for the proportion of regimens subject to each type of advice are given as 95% confidence intervals in Table 2.
Table 2.
Counts and percentages of sampled regimens (n = 1,500) are provided by recommendation type with confidence intervals for population estimates of the percentages of regimens for which MedMinify would generate recommendations for the study population.
| Recommendations Generated per Sampled Regimen | Number of Sampled Regimens | % of Sampled Regimens (95% CI) |
|---|---|---|
| Any Recommendation | 619 | 41.3% (38.8, 43.8) |
| Sustained Release Product Recommendations Only | 407 | 27.1% (24.9, 29.4) |
| Fixed-Dose Combination Product Recommendations Only | 114 | 7.6% (6.3, 8.9) |
| Sustained Release and Fixed Dose Combination Products | 94 | 6.3% (5.0, 7.5) |
| Sustained Release Product and Drug-Drug Interaction | 3 | 0.2% (0.0, 0.4) |
| Drug-Drug Interaction Only | 1 | 0.1% (0.0, 0.3) |
For the 1,500 sampled regimens, MedMinify identified at least one drug on the FDB drug-drug interaction list in 1,209 (81%) of regimens, 95% CI [78.6, 82.6]. MedMinify recommended against taking two drugs at the same time in four cases due to drug-drug interactions related to simultaneous ingestion of drug products. Three of the four drug interactions involved the drugs cyclosporine and simvastatin. (There is a risk that competition for influx transporter proteins in the small intestine and liver could change the serum levels and efficacy of either drug21. It is not clear whether taking cyclosporine and simvastatin at different times of day mitigates the risk of harm from this interaction.) The fourth case of a drug-drug interaction detected by MedMinify involved the two drugs rosuvastatin and gemfibrozil. (There is a risk of muscle damage when using these drugs simultaneously. The exact mechanism of this drug interaction is unknown but there is a suggestion in the literature that simultaneous ingestion may play a role in the interaction22. It is not clear whether taking rosuvastatin and gemfibrozil at different times of day mitigates the risk of harm from this drug interaction.)
3. To what extent would the requirements of home medication regimens change if advice given by a computer application on how to minify daily medication-taking events and daily pill burden were implemented?
If MedMinify’s advice were accepted and implemented for all of the regimens for which recommendations were generated, 320 out of the 1,500 regimens (21.3%, 95% CI [19.3, 23.4]) would have at least one fewer daily medication-taking event. Three regimens would actually gain an additional daily medication-taking event to mitigate an identified drug-drug interaction by taking the interacting drugs at two different times of day. The daily pill burden for an additional 295 of the 1,500 regimens (19.7%, 95% CI [17.7, 21.8]) would be minified by an average of 1.4 pills per regimen. Table 3 includes more detail about these regimen changes.
Table 3.
Changes for sampled qualified multi-drug enteral medication regimens (n = 1,500) according to the maximum number of daily medication-taking events per regimen if all of MedMinify’s advice were implemented.
| Number of Daily Medication-Taking Events Category | Number of Sampled Regimens | Number of Sampled Regimens with Fewer Daily Medication-Taking Events | Number of Sampled Regimens with Decreased Daily Pill Burden Only | New Average Daily Pill Burden per Sampled Regimen if Advice Implemented (Difference) |
|---|---|---|---|---|
| Max. 8 events daily | 1 | 0 | 0 | 26.0 (0) |
| Max. 6 events daily | 3 | 1 | 1 | 13.3 (−2.3) |
| Max. 5 events daily | 4 | 1 | 1 | 14.8 (−1.0) |
| Max. 4 events daily | 44 | 10 | 12 | 10.2 (−1.1) |
| Max. 3 events daily | 178 | 80 | 45 | 9.3 (−1.1) |
| Max. 2 events daily | 643 | 228 | 157 | 5.8 (−0.9) |
| Max. 1 event daily | 626 | N/A† | 79 | 3.3 (−0.1) |
| Max. 0* events daily | 1 | N/A | N/A | 0.0 (0) |
| TOTALS | 1,500 | 320 | 295 |
One regimen included only prescriptions for medications to be taken weekly
One is a minimum daily medication-taking event. MedMinify does not provide advice to discontinue medication therapies.
Discussion
Taking medications every day at home to treat chronic diseases is difficult for most people. Whether and how well one adheres to a home medication schedule is a result of a complex set of influences including economic factors, social support, individual dedication to the task, and beliefs about the utility of prescribed medication treatments5,23. For individuals with means who have supportive relationships and believe that taking medications is beneficial, adherence to complex medication regimens is still difficult8. More attention on simplifying regimens is needed to make every home medication regimen as easy to execute as possible6.
Pharmacists have long been champions of simplifying medication regimens and reducing polypharmacy24. However, many consumers are unaware that their pharmacist can assist them to simplify their home medication regimens. Relatively few consumers are assured to have the simplest possible medication regimens. MedMinify could improve the efficiency of time-consuming home medication regimen reviews provided unsystematically to consumers today.
In this initial evaluation of the function of the MedMinify advice-giving system in the lab, advice to minify daily medication-taking event and daily pill burden was generated for 41.3%, 95% CI [38.7, 43.7] of actual home medication regimens for adults with one or more of three common chronic diseases. This result is bolstered by a previous analysis we did where we found that similar drug product substitution advice could be generated for 38.4% of 2,944 home medication regimens of retirees surveyed in 2007 and by results from Elliott that potential changes to reduce complexity were identified in 45.7% of reviewed medication regimens at hospital discharge7,25.
Based on a careful analysis of the 1,500 sampled regimens we believe the 41.3% figure may overestimate the actual percentage of candidate regimens subject to applicable simplification advice. However, if for reasons of consumer preference, clinical indication, drug product cost, or perceived low marginal utility from changing prescriptions only 1 out of 5 home medication regimens for which advice was given by MedMinify would actually be simplified in practice, the fraction of regimens changed would be approximately 8% in this population. Routine screening with an information resource to try and minify daily medication-taking events or daily pill burden may be justified if 8 out of every 100 regimens could be made simpler. This is a particularly important finding when one considers that 15.3%, 95% CI [13.5, 17.2] of home medication regimens in this population require individuals to take medications more than two times a day (Table 1). We found that 151 of these 230 complex regimens (65.6%) were candidates for simplification, and that the greater fraction of the advice given to simplify these regimens would have resulted in diminishing the number of daily medication-taking events (Table 3). Because the regimens studied were real and current regimens for actual individuals, it is reasonable to generalize these findings to similar consumer populations.
Specific recommendations to substitute sustained-release products for immediate-release ones were the most common recommendations made by MedMinify (Table 2). An example would be to use the antihypertensive beta-blocker carvedilol as a sustained-release product once a day instead of twice a day as an immediate-release product. MedMinify generated fixed-dose combination recommendations for a total of 208 out of the 1,500 regimens (Table 2). An example of a fixed-dose combination recommendation would be to substitute a pill that contains both the antihypertensive drugs hydrochlorothiazide and lisinopril instead of taking these two drugs as separate pills.
Consumers may separate their medication-taking tasks throughout the day due to confusion over prescription instructions or concerns about drug-drug interactions16. It was expected that very few drug-drug interactions would be identified in real regimens like those studied because of prior drug interaction screening. In this study, while most regimens included one drug in the drug interaction knowledge table, rarely were two interacting drugs found in the same regimen. While commonly used medications such as the “cholesterol-lowering statins” are involved in severity level 1 drug-drug interactions, to cause an interaction these drugs had to have been paired with other drugs that are rarely used in this population. By checking for the rare drug-drug interactions that can be managed by taking medications at different times of day, MedMinify can be used to indicate how most medications prescribed to treat high blood pressure, diabetes, and high cholesterol can safely be taken together at the same time of day.
This study has several limitations. With respect to the methods used, the clinical appropriateness of the advice offered by MedMinify was not assessed. Also prescriptions for non-enteral drug products and take-as-needed prescriptions were excluded making the regimens studied less complex than they actually are. With respect to limits on MedMinify’s functionality, the application cannot currently reconcile medication-taking events with individual preferences for meal times, work hours, or sleep schedules. MedMinify does not yet account for the differing cost of various medication products neither for prescription insurance coverage. MedMinify does not recognize the difference between temporary or trial prescriptions and other prescriptions. Also, by limiting the scope of drug-drug interaction assessment to only severity level 1 interactions as defined by FDB, we surely overlooked some less severe but important drug-drug interactions.
Future work will address some of the limitations. We also intend to study having MedMinify give other types of advice including advice about duplicative prescriptions, potentially inappropriate medications26, and polypharmacy. We also look forward to further developing and evaluating a consumer-oriented user interface for MedMinify that will gather details about an individual’s daily schedule and provide advice that is informed by it.
To achieve an ideal level of adherence to home medication regimens requires much more than decreasing the number of medication-taking events or the daily pill burden. However, adherence is improved when home medication-taking schedules become simpler. Minifying daily medication-taking events and daily pill burden is necessary, but not sufficient, to improve adherence to medication regimens used to treat chronic conditions.
Conclusion
A new advice-giving system developed for the SMART platform, and the advice it gave about drug products and drug-drug interactions, was studied. Advice to minify the number of medication-taking events per day and the number of pills taken daily was generated for 41.3% of home medication regimens for adults with chronic conditions. The percentage of regimens for which advice was given increased as the complexity of the home medication regimens increased. Even if only a fraction of the advice given resulted in regimen changes, a considerable number of home medication regimens would include fewer daily medication-taking events involving fewer pills. This study provides preliminary evidence that an information resource could be used routinely to help simplify home medication-taking schedules for adults with chronic medical conditions.
Acknowledgments
This research would not have been possible without support from the following individuals and their organizations: Dr. Joan Kapusnik-Uner of First Databank, Inc. who provided the drug interaction knowledge, Lalitha Natarajan and James Law of the University of Michigan who queried the medication regimen data, Dr. John Kilbourne, Head, Medical Subject Headings, National Library of Medicine and the RxNorm team, and Josh Mandel, Dr. Kenneth Mandl, Dr. Pascal Pfiffner, and Nikolai Schwertner of the Substitutable Medical Apps, Reusable Technologies (SMART) platform project, a Strategic Health IT Advanced Research Program (SHARP) project funded by the Office of the National Coordinator for Health IT of the U.S. Department of Health and Human Services.
References
- 1.“minify, v”, Oxford English Dictionary, 3rd Edition, OED Online. The Oxford University Press; 2014. [Google Scholar]
- 2.Willson MN, Greer CL, Weeks DL. Medication regimen complexity and hospital readmission for an adverse drug event. Ann Pharmacother. 2014 Jan;48(1):26–32. doi: 10.1177/1060028013510898. [DOI] [PubMed] [Google Scholar]
- 3.Deaton A. The great escape: health, wealth, and the origins of inequality. Princeton University Press; 2013. [Google Scholar]
- 4.Mennini FS, Marcellusi A, von der Schulenburg JM, Gray A, Levy P, Sciattella P, Soro M, Staffiero G, Zeidler J, Maggioni A, Schmieder RE. Cost of poor adherence to anti-hypertensive therapy in five European countries. Eur J Health Econ. 2014 Jan 5; doi: 10.1007/s10198-013-0554-4. PMID: 24390212. [DOI] [PubMed] [Google Scholar]
- 5.Bosworth HB, Granger BB, Mendys P, Brindis R, Burkholder R, Czajkowski SM, Daniel JG, Ekman I, Ho M, Johnson M, Kimmel SE, Liu LZ, Musaus J, Shrank WH, Whalley Buono E, Weiss K, Granger CB. Medication adherence: a call for action. Am Heart J. 2011 Sep;162(3):412–24. doi: 10.1016/j.ahj.2011.06.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Isham G., chair . Standardizing medication labels: confusing patients less, workshop summary. 2008. Roundtable on health literacy, Institute of Medicine. [Google Scholar]
- 7.Elliott RA. Reducing medication regimen complexity for older patients prior to discharge from hospital: feasibility and barriers. J Clin Pharm Ther. 2012;(37):637–642. doi: 10.1111/j.1365-2710.2012.01356.x. [DOI] [PubMed] [Google Scholar]
- 8.Claxton AJ, Cramer J, Pierce A. A Systematic Review of the Associations between dose regimens and medication compliance. Clinical Therapeutics. 2001;23(8):1296–1310. doi: 10.1016/s0149-2918(01)80109-0. [DOI] [PubMed] [Google Scholar]
- 9.Cohen C, Elion RA, Frank I, Kloser P, Sherer R, Squires KE, Corklin S, Tebas P. Once-daily antiretroviral therapies for HIV infection: consensus of an advisory committee of the international association of physicians in AIDS care. JIAPAC. 2002;(1):141–145. doi: 10.1177/154510970200100406. [DOI] [PubMed] [Google Scholar]
- 10.Nachega JB, Parienti JJ, Uthman OA, Gross R, Dowdy DW, Sax PE, Gallant JE, Mugavero MJ, Mills EJ, Giordano TP. Lower pill burden and once-daily antiretroviral treatment regimens for HIV infection: a meta-analysis of randomized controlled trials. Clin Infect Dis. 2014 Mar 5; doi: 10.1093/cid/ciu046. PMID 24457345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kripalani S, Jacobson TA. Illustrated medication schedules improve medication adherence in at-risk patients with coronary heart disease [abstract] J Gen Intern Med. 2010;25(S3):S301. [Google Scholar]
- 12.Gazmararian J, Jacobson KL, Pan Y, Schmotzer B, Kripalani S. Effect of a pharmacy-based health literacy intervention and patient characteristics on medication refill adherence in an urban health system. Ann Pharmacother. 2010;44(1):80–87. doi: 10.1345/aph.1M328. [DOI] [PubMed] [Google Scholar]
- 13.Schnipper JL, Roumie CL, Cawthon C, Businger A, Dalal AK, Mugalla I, Eden S, Jacobson TA, Rask KJ, Vaccarino V, Gandhi TK, Bates DW, Johnson DC, Labonville S, Gregory D, Kripalani S, PILL-CVD Study Group Rationale and design of the pharmacist intervention for low literacy in cardiovascular disease (PILL-CVD) study. Circ Cardiovasc Qual Outcomes. 2010;3(2):212–219. doi: 10.1161/CIRCOUTCOMES.109.921833. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Stange D, Kriston L, von-Wolff A, Baehr M, Dartsch DC. Reducing cardiovascular medication complexity in a German university hospital: effects of a structured pharmaceutical management intervention on adherence. J Manag Care Pharm. 2013 Jun;19(5):396–407. doi: 10.18553/jmcp.2013.19.5.396. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Solomon DH, Brookhart MA, Tsao P, Dundaresan D, Andrade SE, Mazor K, Yood R. Predictors of very low adherence with medications for osteoporosis. Osteoporos Int. 2011;(22):1737–1743. doi: 10.1007/s00198-010-1381-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wolf MS, Curtis LM, Waite K, Bailey SC, Hedlund LA, Davis TC, Shrank WH, Parker RM, Wood AJ. Helping patients simplify and safely use complex prescription regimens. Arch Intern Med. 2011 Feb 28;171(4):300–5. doi: 10.1001/archinternmed.2011.39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hansten PD, Horn JR. The top 100 drug interactions. H&H Publishing; 2013. [Google Scholar]
- 18.Farrish S, Grando A. Ontological approach to reduce complexity in polypharmacy. AMIA Annu Symp Proc. 2013 2013 Nov 16;:398–407. [PMC free article] [PubMed] [Google Scholar]
- 19.Krejcie RV, Morgan DW. Determining sample size for research activities. Ed Psych Meas. 1970;30:607–10. [Google Scholar]
- 20.Brown LD, Cai TT, DasGupta A. Interval Estimation for a proportion. Statistical Science. 2001;16:101–133. [Google Scholar]
- 21.Kalliokoski A, Niemi M. Impact of OATP transporters on pharmacokinetics. Br J Pharmacol. 2009 Oct;158(3):693–705. doi: 10.1111/j.1476-5381.2009.00430.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Bergman E1, Matsson EM, Hedeland M, Bondesson U, Knutson L, Lennernäs H. Effect of a single gemfibrozil dose on the pharmacokinetics of rosuvastatin in bile and plasma in healthy volunteers. J Clin Pharmacol. 2010 Sep;50(9):1039–49. doi: 10.1177/0091270009357432. [DOI] [PubMed] [Google Scholar]
- 23.Norell SE. Improving medication compliance: a randomised clinical trial. Br Med J. 1979;2(6197):1031–3. doi: 10.1136/bmj.2.6197.1031. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Lenaghan E1, Holland R, Brooks A. Home-based medication review in a high risk elderly population in primary care-the POLYMED randomised controlled trial. Age Ageing. 2007 May;36(3):292–7. doi: 10.1093/ageing/afm036. [DOI] [PubMed] [Google Scholar]
- 25.Flynn AJ, Klasnja P, Friedman CP. Taking it easy – a needs analysis for computer-generated advice to simplify home medication regimens; AMIA Annu Symp Proc; 2013 Nov 16. [Google Scholar]
- 26.Shade MY, Berger AM, Chaperon C. Potentially Inappropriate Medications in Community-Dwelling Older Adults. Res Gerontol Nurs. 2014 Feb 19;:1–15. doi: 10.3928/19404921-20140210-01. [DOI] [PubMed] [Google Scholar]