Interactive tools for inpatient medication tracking: a multi-phase study with cardiothoracic surgery patients

Abstract

Objective Prior studies of computing applications that support patients’ medication knowledge and self-management offer valuable insights into effective application design, but do not address inpatient settings. This study is the first to explore the design and usefulness of patient-facing tools supporting inpatient medication management and tracking.

Materials and Methods We designed myNYP Inpatient, a custom personal health record application, through an iterative, user-centered approach. Medication-tracking tools in myNYP Inpatient include interactive views of home and hospital medication data and features for commenting on these data. In a two-phase pilot study, patients used the tools during cardiothoracic postoperative care at Columbia University Medical Center. In Phase One, we provided 20 patients with the application for 24–48 h and conducted a closing interview after this period. In Phase Two, we conducted semi-structured interviews with 12 patients and 5 clinical pharmacists who evaluated refinements to the tools based on the feedback received during Phase One.

Results Patients reported that the medication-tracking tools were useful. During Phase One, 14 of the 20 participants used the tools actively, to review medication lists and log comments and questions about their medications. Patients’ interview responses and audit logs revealed that they made frequent use of the hospital medications feature and found electronic reporting of questions and comments useful. We also uncovered important considerations for subsequent design of such tools. In Phase Two, the patients and pharmacists participating in the study confirmed the usability and usefulness of the refined tools.

Conclusions Inpatient medication-tracking tools, when designed to meet patients’ needs, can play an important role in fostering patient participation in their own care and patient-provider communication during a hospital stay.

BACKGROUND AND SIGNIFICANCE

Studies focused on the effects of providing patients with access to their medical records have highlighted patients’ positive experiences with having such access, including improved communication, self-care, and participation in planning their own care. ¹ However, hospitalized patients and their family members have largely missed opportunities to benefit from the systems used to manage inpatient care, including the ability to resolve discrepancies in patient data, supplement that data, and use it to inform timely, collaborative decision making about patients’ care. ^2–5

To date, most research on patient access to clinical data has focused on personal health record systems or online patient portals. ^6,7 These technologies provide select portions of the patient’s medical record, often as found in a healthcare system’s electronic health record (EHR) system, and frequently include secure messaging with providers, prescription refills, and appointment scheduling. ^8–10 In recent studies, patients’ ability to view portions of their EHR positively impacted their communication with their caregivers, enhanced knowledge of their own health, improved their self-management skills, and allowed for increased follow-up on abnormal test results. ^1,11 Noting that communication between physicians and patients is one of the few factors that are strong enough to be predictive of patient adherence to therapies, ^12–17 recent work has examined the potential for EHRs to enhance physician-patient communication during the clinical encounter. ^18,19

Patients’ Medication Information Needs

In parallel, research on patients’ medication information needs has thus far addressed medication management in outpatient or home settings, ^20–23 long-term medication management, ^24,25 and medication management for specific care transitions. ²⁶ In contrast, this study examines how computing tools can be designed to meet specific medication information needs that arise in the inpatient setting.

Applications for medication monitoring and management have particular relevance for patients who have just undergone cardiothoracic surgery, given their relatively short lengths of stay and their need to self-manage medications postoperatively, in the home setting. The self-management needs of patients with chronic heart disease post-hospitalization have been well-documented in studies on HeartCareII, which included the creation of patient-tailored web resources, within an inpatient portal, that were accessible at home. ⁶

However, a lack of knowledge about medications and poor communication with clinicians – originating within the inpatient setting – create critical barriers to patients’ post-treatment medication adherence, shared decision making between patients and physicians, and therapeutic outcomes. ⁵ Patients’ knowledge of medications administered in-hospital has been shown to be poor. In a recent study of hospitalized patients’ knowledge about prescribed warfarin therapy, Shuaib et al. ²⁷ found that 58% of patients were unaware of any adverse effects of this therapy, while 27% had experienced adverse effects, and 12% required further hospitalization as a result of such effects. Cumbler et al. ²⁸ found that 44% of hospital patients believed they were receiving a medication that they were not, and 96% were unable to recall the name of at least one medication that they had been prescribed during their hospitalization. Access to medication lists during care could aid patients’ knowledge of the medications they are prescribed during their hospital stay.

In addition to improving patients’ knowledge of their medication therapies, providing mechanisms for patient feedback on medication lists improves medical record accuracy, in turn promoting patient safety. ²⁹ Dullabh et al., ⁴ in their study of 1500 users of the MyGeisinger personal health record system, found that 30% of these users submitted feedback on their recorded medication lists, with 89% of these correspondences including requests to change information in the shared medication list in order to improve its accuracy.

These studies support the need for technology that assists patients’ medication knowledge and management throughout all stages of their care. However, the inpatient environment is an important but underexplored setting: although emerging studies show a promising connection between hospital patients’ ongoing access to their care information and their post-treatment outcomes, the information needs of hospital patients often go unmet. ^30–32 According to O’Leary et al., ³³ a substantial portion of hospitalized patients do not understand their own care plan, including their primary diagnosis and medication changes, which impacts these patients’ ability to assume their own care after discharge. In a recent study of 150 hospitalized (general medical) patients that assessed patients’ current use of information technology and their interest in receiving electronic information, nearly half of the study participants reported an interest in receiving electronic information during their hospitalization, with a particularly high interest in receiving medication lists. ³⁴ Medication lists with Infobuttons (ie, tailored medication information) were also identified as desired features in a patient-centered tool kit designed to engage hospitalized patients and their care partners in their plan of care. ⁵

Motivated by these findings, our study investigates how tools can support patients’ medication information needs during inpatient care. Prey and colleagues ³⁵ have synthesized the literature on patient engagement in inpatient settings to date, concluding that inpatient engagement technologies have been limited and calling for further research exploring the design of technology for this setting. With their design of an electronic bedside communication center that presents patients in general medical units with tailored information, Dykes et al. ³⁶ took important steps toward enabling patient engagement during inpatient care. The participants in this study anticipated that the communication center designed by the authors would be useful.

Complementary research by Caligtan et al. ²⁹ identified patients’ high-level information requirements for promoting both healthcare management needs as well as understanding of the care environment and specified medication information as one of five core needs. Subsequent work by Dykes et al. ⁵ described a participatory design process in intensive care and oncology units, to identify workflow and system usability constraints. In contrast, we aim to explore inpatient medication information needs in detail and validate the design of medication tracking tools developed to meet those needs.

We began our iterative design process with a preliminary needs assessment consisting of in situ interviews with patients and nurses. ³² The results of these interviews informed the preliminary design of the means of viewing inpatient medications and the development and pilot study of an inpatient application infrastructure, ⁴⁴ described further below. We now present the results of the following two phases of our iterative design for the first study, of which we are aware, focused on responses to a set of inpatient medication-tracking tools designed to meet the specific needs of patients in the inpatient setting.

Designing for Patient Engagement

To design medication-tracking tools for the inpatient setting, we looked to conceptualizations of “activation” and “participation” – each originating in the Chronic Care Model ³⁷ and articulated in successive frameworks and instruments. In particular, we drew from three relevant tenets underlying Hibbard et al.’s Patient Activation Measure. ^38,39 First, we aim to enable patients’ awareness of their therapies to support their engagement in activities that maintain functioning. Second, by providing this awareness, coupled with mechanisms for reporting questions and comments, we aim to promote patients’ involvement in treatment choices. Third, through awareness and involvement in treatment choices, medication-tracking tools can support patients’ collaboration with providers. Patient activation, “the skills and confidence that equip patients to become actively engaged in their health care,” can be increased through interventions that foster patients’ skills and confidence, leading to improved health outcomes in addition to lowering costs. ⁴⁰ In 2012, the Agency for Healthcare Research and Quality recognized the importance of such activation in their Patient and Family Engagement Framework, which encourages interventions that promote the participation of patients and family members in patient care, in part through keeping patients informed during their hospitalization. ⁴¹

Additionally, we drew on the patient work framework ⁴² by proposing advancements in informatics tools that emphasize patients’ perspectives on the usefulness of informatics solutions. To determine the usefulness of the tools we developed, we adopted concepts from the Technology Acceptance Model (TAM) ⁴³ – specifically, perceived ease of use, perceived usefulness, and behavioral intent to use. The goal of the tools we have developed is to enhance patient engagement by promoting patient activation and facilitating patients’ participation in their own care. To achieve this goal, a critical prerequisite is that these tools are perceived to be easy to use and useful for meeting patients’ information needs, influencing behavioral intent to use and, ultimately, actual use (acceptability).

OBJECTIVE

In this paper, we first present the myNYP Inpatient application. We describe two phases of a larger iterative, user-centered process of aligning the design of interactive medication-tracking tools with the information needs of cardiothoracic surgery patients and ensuring that the interaction and presentation techniques of the tools match the abilities and preferences of the target population. Our research questions relate to the usability and the actual use of the tools as well as their usefulness for meeting information needs that particularly arise in the inpatient setting. They include:

1. Will participants make use of the medication-tracking tools in myNYP Inpatient? (Phase One)

2. Will participants find the medication-tracking tools in myNYP Inpatient usable? (Phases One and Two)

3. Do the medication-tracking tools in myNYP Inpatient meet patients’ information needs? (Phases One and Two)

Designing medication-tracking tools for the inpatient setting required novel approaches to structuring and presenting large numbers of medication orders. To create patient-friendly means of viewing these data, we added interactive groupings of medication orders in Phase Two of the study, using a hybrid approach to combine medication lists with graphical schedules. We further augmented the presentation of inpatient medications to include the data types prioritized by patients, in Phase One of the study, and by clinical pharmacists, in Phase Two.

After introducing the interactive medication-tracking tools in myNYP Inpatient, we describe a two-phase, field-based evaluation of these tools in a postoperative cardiothoracic surgery setting ( Table 1 ). In the first phase of the study, we recruited 20 patients and two of their family members over a 16-week period. In the second phase of the study, five clinical pharmacists, along with 12 patients, evaluated tool refinements and participated in short interviews, conducted over a 3-week period, in which we examined their preferences for these tools.

MATERIALS AND METHODS

Study Context

Study Setting

This two-phase field study took place in a 30-bed postoperative cardiothoracic surgery step-down unit at Columbia University Medical Center at New York-Presbyterian Hospital (NYP). This medical center is a Level I trauma center and a major teaching hospital with a busy cardiothoracic surgery service. The medical center’s EHR (Sunrise Clinical Manager, Allscripts Corp., Chicago, IL, USA) has been in use since 2004. Approval for the study was obtained from the Columbia University Medical Center Institutional Review Board.

myNYP.org

In 2009, NYP launched its personal health record portal, www.myNYP.org . The portal uses Microsoft HealthVault (Microsoft Corp., Redmond, WA, USA) to store demographic and clinical data and is available to all adult NYP patients. It offers educational content, including a set of materials focused on cardiovascular conditions, tests, and procedures, through a Health Library module.

myNYP Inpatient Architecture

Formatted for tablet computers, myNYP Inpatient is a custom prototype application for inpatient use ( Figures 1–3 ). The application supports a touch-based user interface. A diagram of the system architecture is shown in Figure 3 . A Data Access Module, using a “message broker” model, ⁴⁴ queries the inpatient EHR using structured query language (SQL), for patient-specific information, such as demographics, care providers, medication orders, and detailed medication administration records. An audit module logs user actions for research and security purposes. In both phases of the study, medication-tracking tools in myNYP Inpatient ran in the Safari web browser on first-generation Apple iPad devices.

Application content reflects nearly real-time data. For example, when a physician modifies a medication order or a nurse records the administration of a medication in the EHR, the tablet application displays the new information almost instantaneously. We assessed the feasibility of the deployment and usage of the application through a small pilot study. ⁴⁵ This previous study included a short supervised usage session, with views of inpatient medication data that included only a limited amount of information. Through the pilot study, we learned that patients desired easier ways to track new medication orders and verify the medications they took at home. Several participants also requested the ability to record information they wanted to communicate to their care team. Thus, our next iteration of these tools, presented herein, builds on our prior work by adding these capabilities. In addition, studying the use of medication-tracking tools in myNYP Inpatient using a larger number of patients over a longer time period allowed us to glean further insights about the tools’ usability and usefulness as well as how they can evolve to best meet inpatients’ information needs.

PHASE ONE MATERIALS AND METHODS

myNYP Inpatient: Medication-Tracking Tools

Herein, we introduce three tools comprising the medication-tracking capabilities of myNYP Inpatient: Hospital Medications, Home Medications, and Patient Notepad. Figures 1 and 2 show views of the user interface, as they appeared in Phase One of the study, and Figures 4 and 5 show views of the user interface as they appeared in Phase Two of the study.

Hospital Medications

Touching the “Hospital Medications” link on the application home screen launches interactive views of the medications the patient was administered in-hospital, as recorded in the patient’s medical record ( Figures 1 , 4 , and 5 ). The patient’s current, pending, and discontinued medication orders are displayed, along with information on the medication dose, its frequency of administration, and the administration time. Touching the name of a medication launches a JavaScript procedure that sends a web service request to MedlinePlus ⁴⁶ to display context-specific medication information maintained by the National Library of Medicine and targeted to consumers. ⁴⁷ The Hospital Medications tool, shown in Figure 1 , includes two medication lists, ordered alphabetically by medication name, in a tabular format. Touching the last administration time for a medication opens a window that shows the entire administration history of the medication for the duration of the patient’s hospitalization.

Home Medications

Interactive views of home medications include text descriptions of the medication name and usage instructions, including dose ( Figure 1 ). Because the information in this tool reflects the patient’s home medication list as recorded at their intake, the tool’s structure is limited to the fields available in the EHR to record this information. In accordance with hospital procedure, patients were directed to notify their care team if they had concerns about the accuracy of either the home medications or allergies listed.

Patient Notepad

Touching the “Record Questions and Comments” link on the application home screen launches a free-text entry field and a view of the patient’s previously-recorded text. The user can enter the text as a question or a comment ( Figure 2 ) – each posts to a custom research pane in the EHR, associated with the patient’s medical record (the recorded text need not pertain only to medications). This study includes an analysis of patients’ use of the Patient Notepad tool for medication-specific questions and comments.

Phase One Study Sample

Twenty-one postoperative cardiothoracic surgery step-down patients were approached for the study. Exclusion criteria included patients with severe mental illness, a history of substance abuse, language barriers, or severe physical discomfort. Screening criteria required that the patient’s attending physician identify them as being in good enough health to participate in the study (eg, capable of ambulating with assistance or on their own and off of step-down monitors). We registered 20 patients with the myNYP Inpatient application after obtaining their informed consent, provided each participant with a temporary user ID and password generated for the study, and connected each of the tablets running the application to the study server via a limited-access, virtual private network.

Phase One Procedures

After a 15-min training and observation session to introduce and explain the features of myNYP Inpatient and observe a demonstration of its use, the iPad was left with the patient to use in an unsupervised manner over a 24–48-h period. Patients were told that information typed into the Patient Notepad could be viewed in the EHR, but comments and questions should be addressed through face-to-face communication with the healthcare team. The researchers returned only to answer patients’ questions about the study or to conduct their debriefing interview. Thus, all the patients participating in this phase of the study used the application under supervision for training and usability assessment, while unsupervised use was captured to assess actual user behavior.

We noted our observation data related to the usability of the application on paper (eg, errors and whether and how they were reversed, trouble navigating to information, and patients’ questions and difficulties related to vocabulary and interactions). ^48–50 We conducted an audio-recorded interview at the conclusion of the unsupervised use period to assess the patient’s subjective satisfaction with and perceptions of the usefulness of the tools. Semi-structured interview questions were based on the primary concepts from the TAM ⁴³ – specifically, perceived ease of use, perceived usefulness, and behavioral intent to use (sample questions are shown in Box 1 ). We compared the patients’ response data related to these themes with audit log data (ie, records of patients’ actual use of the tools) and performed a descriptive analysis. ⁵¹

Phase One Data Analysis

To differentiate between supervised and unsupervised use of the application, we manually cross-referenced patient responses to the interview questions with their audit logs: we matched event timestamps in log data with the initiation and conclusion of the supervised observations and interview. Auditing methods recorded patient selections of application features and state changes in the application, (eg, “select ‘Hospital Medications, Dose History’”).

All interview data were transcribed verbatim. Using a subset of field notes along with interview transcriptions segmented at speaker turns, four members of the research team (J.P., A.S., L.W., J.W.) together reviewed these data iteratively, reaching consensus on the themes and sub-themes informed by TAM, for qualitative coding. A senior member of the research team (S.B.) with expertise in qualitative analysis and technology acceptance models reviewed the coding and made suggestions for refinements. The resulting codes were applied to the remaining qualitative data in a deductive fashion.

PHASE ONE RESULTS

As shown in Table 2 , 20 patients (8 female) ages 26–81 (mean = 58.35 ) participated in the study. Additionally, two of the patients’ spouses participated in the interviews. Of the 20 patients enrolled in our study, 14 (70%) used the application while unsupervised (referred to as “Active Users” in Table 2 ). Patient interviews lasted 20–30 min (interviews with active users ran longer, in order to capture their experiences).

Interview Data

Our deductive thematic coding of the patient interview data comprises the primary themes related to the medication-tracking tools: use, usability, and usefulness, along with several subthemes (eg, data presentation, navigation, and interaction as subthemes of usability; and information needs, communication needs, and resolving data as subthemes of usefulness). In Table 3 , we provide illustrative quotations from the patient interviews corresponding to each of these themes and subthemes. To organize the quotations across study phases, we labelled this group of patients as AP1–AP20.

Eighteen of the twenty patients enrolled in Phase One of the study (including the 14 “active users” who used the tools during the unsupervised use period) indicated that they found the tools in the application to be useful. Although six participants in Phase One of the study did not use the tools during the unsupervised use period, four of these six participants appreciated the availability of the tools’. Below, we elaborate on these findings as well as others related to the application’s use and usability.

Audit Log Data

Of the 14 participants using the application tools actively during the unsupervised use period, the average number of selections made by a participant was 13. All 14 participants accessed the Hospital Medications tool at least once (up to 25 times). Seven of the fourteen participants accessed the Home Medications tool at least once (up to 12 times), and seven accessed the Patient Notepad at least once (up to 4 times). Among the 14 participants, 10 used the Hospital Medications tool more than any other tool; 1 participant each used the Patient Notepad and the Home Medications tool most frequently. The remaining two participants accessed the Patient Notepad and the Hospital Medications tool the most, with equal frequency. From the Hospital Medications view, 3 of the 14 active users navigated to MedlinePlus resources during the unsupervised use period (logging 3, 6, and 17 selections, respectively), while only 1 of the 14 active users accessed their medication administration history.

Comparing Interview and Audit Log Data

Use and Usefulness of Medication-Tracking Tools

Hospital Medications

Although the frequency with which participants accessed the Hospital Medications tool varied, most found it useful to access the information contained therein, even if they only did it once, during their application use session. Reasons for this tool’s usefulness included the ability to share the information with family (AP8, AP15), to remain informed about prior medication administration (AP6, AP10), and to search for further information about their medications (AP5). Several patients suggested additional data that could better meet their needs, including explanations of the medication and its class in lay terms (AP1, AP7), the expected schedule for the medication’s administration (AP5, AP10, AP11), and the medication’s method of administration (AP7, AP20).

Most patients appreciated the availability of their hospital medications data, though some patients found the size of the inpatient medication lists to be overwhelming and felt discomfort reading the side effects and warnings associated with certain medications (AP4, AP6; see Table 3 ). As AP4 said of her experience, “it’s good – it’s informative,” adding that it was, “a little scary when I saw all the medications I was given in a 5-day period.” During the observation period, AP4 navigated to MedlinePlus resources about amiodarone and experienced distress while reading the warnings for and possible side effects of the medication, indicating, “this is only going to stress me out.” In the following section, we discuss how our design was refined to address patients’ concerns related to the size of their inpatient medication lists.

Home Medications

To our surprise, we also found that although the Home Medications tool contained “static” data (ie, updates to the home medication list were not expected during the patient’s hospital stay), some patients accessed the tool frequently. For AP13, Home Medications accounted for 11 selections of 28, and AP5 selected the Home Medications tool three times in 24 h. We learned via the patient interviews that patients shared their home medication data with their family members (AP6, AP8) and accessed it in order to resolve discrepancies in the data (AP8, AP10; see Table 3 ).

Patient Notepad

We were unsure whether patients receiving postsurgical care would find the Patient Notepad feature useful, given that typing can prove physically burdensome. Because no prior study has inquired into patients’ preferences for communicating electronic comments or questions to their care team members during inpatient care, we were uncertain about patients’ preferences for handling such communication. In fact, all but one of the active users indicated that they found the Patient Notepad useful. Most (9 of 12) also mentioned that they preferred sharing data in a “two-way” communication style (ie, posting questions and comments to the EHR and receiving an electronic, versus an in-person, reply).

Electronic Inpatient Communication: Perceived Benefits and Concerns

Although patients and participating spouses volunteered several benefits of two-way electronic inpatient communication for medication-related questions and comments, they also raised concerns about this method of communication. The noted benefits included the ability to log side effects (AP1), to refer back to details of communication later that they might otherwise forget (AP5, AP6, AP11), and to support question “triage” to help clinicians prioritize communications (AP6, AP10) with the ability to reply quickly (AP4, AP6). One spouse who participated in the patient interviews, the wife of AP8, suggested that electronic communication could reduce “sonic noise” during her spouse’s hospital stay, and participant AP5 mentioned that electronic communication could ease the physical discomfort of speaking. Finally, patients remarked that the tool can allow family members to note their own questions about medications (AP5, AP6, AP8).

Several patients raised concerns related to two-way electronic communication during inpatient care. Some wished to direct questions and comments to an individual member of the care team rather than the entire team and raised concerns about accountability and maintaining relationships with specific caregivers (AP2, AP11, AP12). Two participants mentioned wanting to track the status of a question (eg, unread, read, replied-to) (AP1, AP11). Still other patients were interested in electronic communication but wanted examples of questions and comments to get them started (AP1, AP17). These concerns point to the need for further research to systematically investigate issues related to both the feasibility and safety of this communication model.

Usability of the Medication-Tracking Tools

Reviewing field observation notes and usability-related interview responses revealed two primary concerns. First, some of the interactive features of the Hospital Medications tool, such as selecting the last administration time to see the history of the medication’s administration, were not readily apparent to users (AP1, AP9, AP15, AP16). Second, the organization of the inpatient medication lists in a scrollable, tabular format both limited discovery of discontinued medications (AP16) and made for an overwhelming experience for users upon loading (AP4, AP6). In addition, some patients complained about the need to zoom in frequently, due to the small font size in the application (AP4, AP11).

Acceptability

The fact that 70% of the Phase One participants were able to, and did, use the application unsupervised is encouraging. One limitation of this phase of the study was the time frame available to elicit patient usage of the application – we do not know if we could expect such use over the course of many days of care, nor whether such use could influence patients’ longer-term adoption of patient portal technology. Still, patients’ perceptions that the tools were usable and useful, taken together with their demonstrated usage, point toward the positive potential of the acceptability of these tools during postoperative cardiothoracic surgery care.

SUMMARY OF REQUIRED REFINEMENTS TO TOOLS

From our Phase One data analysis, we distilled required refinements to the application tools based on both the primary usability concerns expressed by the study participants and the information needs we discovered through the patient interviews. We scoped the following research phase to address the Hospital Medications tool, noting concerns regarding the presentation of large medication lists (usability), making the interactive content clearer (usability), and enabling better patient awareness of their overall therapeutic progress, medication class identification, and tracking of timing and route of medication administration (usefulness: information needs).

PHASE TWO MATERIALS AND METHODS

Refinements to Hospital Medications Tool

Presenting Large Numbers of Hospital Medications

To better understand the expected volume of hospital medication orders for individual patients, we performed a retrospective review of the EHR data of 12 cardiothoracic surgery patients with common profiles. The average number of unique medications ordered for these patients (when differentiating administration forms, such as liquid vs pill) was 63.5. This includes “as needed” medications and represents multiple stages of care. Still, for patients taking only a few medications at home (if any), this number can be alarming. We used these data to design alternative techniques for presenting medication lists and reviewed them with clinical pharmacists prior to Phase Two of the study.

Supporting Awareness of Therapeutic Progress and Expected Course of Administration

With two participating clinical pharmacists (C1 and C2), we first designed and evaluated chronological “timeline” views of patient’s medication data. Spatial arrangement of the medication data in a timeline format required that predetermined grouping metrics be applied to the data in order to render chronological views. However, groups still cluttered these views while limiting patient exploration of the data – details of the drugs were not readily apparent and browsing medication orders by name, class, or administration method would not be possible. Below, we describe how we approached these challenges in the resulting design.

Resulting Design

We created a hybrid graphical and list-based view of the Hospital Medications tool ( Figures 4 and 5 ) with three groups: Active Medication Orders, Medications Ordered Today, and Completed or Discontinued Medications, organized based on medication order data. Graphical summaries of the daily administration schedule for the medication are rendered for each medication entry. The frequency field is generated based on parsing the instructions of the medication order with keyword matching (eg, “9 a.m. daily” and “twice daily”) and an analysis of the medication administration history (eg, if given twice daily and historically given at 10 a.m. and 8 a.m., dots are drawn at places corresponding roughly to these times in the abstracted schedule).

The data groups are collapsible, to allow the user to focus on a single group, and each contains six sortable fields, describing the class, name, dose, frequency of administration, route of administration, and uses of the medication. Uses of the medications were extracted from HON-certified online sources (the “Health on the Net” [HON] Foundation certifies the credibility of health-related websites through physician review) and validated by the clinical pharmacist who participated in the study. Active Medication Orders is the only group expanded by default.

Phase Two Study Samples and Procedures

Clinical Pharmacist Sample and Procedures

Five clinical pharmacists, including the two that participated in the design process of the Phase Two tool, evaluated subsequent iterations of the Hospital Medications view for further refinement and to prepare for the patients’ evaluation. The clinical pharmacists performed a variation on a cognitive walkthrough, ⁵² in which they used the user interface populated with historical patient data to find information about heparin (an active order), simvastatin (a discontinued order), and oxycodone with acetaminophen (a new order). The pharmacists were asked to select the interactive fields for each medication entry and to comment on the information quality and appropriateness found therein.

The user interface also contained sample medications representing classes that are common to cardiothoracic surgical care (ie, angiotensin-converting enzyme [ACE] inhibitors, analgesics, antiaggregants, antiarrhythmics, anticoagulants, β-blockers, diuretics, and statins). To assess preliminary usability, the pharmacists were asked to use a think-aloud protocol when interacting with the user interface (written notes captured the pharmacists’ think-aloud data). Walkthroughs were performed using teleconferencing software ( n = 3) with screen-sharing enabled or in-person ( n = 2); all of the walkthroughs were audio-recorded and transcribed.

Patient Sample and Procedures

Fifteen postoperative cardiothoracic surgery step-down patients were approached for participation in this user study. Inclusion and exclusion criteria as well as the recruitment and informed consent processes matched those of Phase One of the study. After obtaining informed consent from the patients, we asked them to interact with the redesigned view of their medication data, using a think-aloud protocol. ⁵² The redesigned Hospital Medications view was populated with medication order data using a JavaScript import function – the patients participating in Phase Two of the study were not registered with the full myNYP Inpatient application, used only the Hospital Medications tool, and did so only under supervision. We concluded with patient interviews, conducted using a subset of the questions asked in the Phase One patient interviews that pertained to the usability and usefulness of the Hospital Medications tool.

Phase Two Data Analysis

Using the TAM-informed coding scheme constructed in Phase One of the study, one researcher (L.W.) applied deductive thematic coding to both the patient and pharmacist interview response data. Think-aloud data were cross-referenced with the participants’ interview responses.

PHASE TWO RESULTS

Clinical Pharmacists Evaluation

Cognitive walkthroughs revealed that sorting functionality helped the clinical pharmacists locate and interact with patients’ medication data. The ability to sort medications by class, administration method, and administration frequency allowed pharmacists to explore the data according to the questions they expected from patients about the medications. Adding the Medications Ordered Today group as a single view allowed the pharmacists to better track changes to patients’ medication therapies. The abstracted daily schedules ( Figures 4 and 5 , see the “Frequency” column) gave approximate times for medication administration events while yielding clearer views of the medications. In addition, we found that using this means of representation could mitigate expectations that a medication must be given to a patient at a specific time, when, in fact, a time range for administration is appropriate.

The pharmacists also recommended adding alternative ways to identify a medication (through images and pronunciation guides), safety information about the medication, and additional presentation and interaction techniques for managing large medication lists. Table 4 highlights themes and corresponding quotations from the clinician and patient interviews. Clinician pharmacist participants in this phase are referred to as C1–C5.

Patient Evaluation

Twelve patients (four female) participated, of the fifteen approached. The participants’ ages were in the range of 48–88 (mean = 66). Two participants’ family members participated along with the patient in the interviews. The participants’ subjective satisfaction with the interactive Hospital Medications tool was high – all of the participants indicated that they appreciated the availability of the information in the Home Medications tool at the bedside (in one case, only for their family members) and thought that the design of the tool met their primary information needs. The patient participants in this phase are referred to as BP1–BP12.

The patients participating in Phase Two of the study indicated that the class, name, dose, frequency of administration, route of administration, and uses of the medication were essential data to include in a primary view of inpatient medication lists, with touch access to dosage, administration history, and educational content for each medication. Several patients brought up additional information needs and functionality requirements, such as better support for tracking changes in related medication orders (BP5, BP7, BP11) and more obvious links from the generic version of a drug to a familiar brand name (BP2, BP4). In some cases, patients also desired reconciliation of the inpatient medication list with what they could expect at discharge (BP6, BP7; see Table 4 ).

The patients in this phase of the study were uniformly positive about the tool’s sorting features and the incorporation of individual medication schedules into list-based views. When explaining why they desired these features, we learned that medication classes, doses, and administration methods (eg, intravenous [IV] liquids, injections, liquids, powders, pills) changed frequently, depending on the severity of the patients’ side effects, current problems, and individual responses to drug therapies. Thus, patients depended on clinical caregivers to administer drugs, while finding it challenging to maintain awareness of the schedules of medication administration and their own progress. Our approach to combining medication lists with abstract graphical schedules allowed for the incorporation of sorting functionality, which enabled flexible exploration of medication orders according to meaningful data types while providing patients with the means to track large volumes of medication orders. Still, patients also suggested that sorting could be automated, to more clearly indicate which medications could be expected to be ordered next (BP12; see Table 4 ).

Comparing Pharmacists’ and Patients’ Perspectives

Although both patients and pharmacists found the features of the tool to be useful, the patients emphasized features that helped them understand what to expect during the course of their hospital stay, including automatic updating of active medication views to highlight the next expected medication and functionality to track therapy replacements (ie, one drug replacing another). These findings point toward novel opportunities for health information systems to better support the identification of relationships between multiple inpatient medication orders as well as between inpatient and home medications, and to present these relationships to patients in ways that are understandable to them.

The pharmacists emphasized features related to clinical workflow and medication safety, including medication tracking through abstracted schedules, to mitigate patient expectations; the inclusion of safety information related to switching narrow-therapeutic-index drugs; and alternative methods of identifying medications. Both pharmacists and patients mentioned the importance of including familiar brand names of generic drugs in the immediate view of the medication lists. To realize these features, computational methods are needed to identify and display associations between medications with which the patient is familiar (ie, brand-name drugs taken at home) and generic equivalents prescribed in the hospital. Similarly, although images of inpatient medications (eg, pills) could aid patients’ identification of them, techniques are needed to choose the correct image (ie, the correct color, shape, and imprint) for the particular generic product used to fulfill the order.

DISCUSSION

Both the cardiothoracic surgery patients and clinical pharmacists in our study found that inpatient medication-tracking tools played an important role in fostering patient participation and patient-provider communication during a patient’s hospital stay. Our multi-phase study resulted in: 1) interactive, inpatient medication-tracking tools used by patients in a two-phase inpatient pilot study, with a high level of subjective satisfaction (RQ1, RQ2); 2) an assessment of patients’ medication information needs in the inpatient setting that incorporated perspectives from both patients and clinical pharmacists (RQ3); and 3) design insights on prioritizing and structuring inpatient medication data for interactive use in the inpatient setting (RQ 1–3). These contributions can guide future development of patient-facing clinical information systems that are designed for inpatient use.

Relationship to Prior Research

Our findings expand upon prior knowledge in three ways.

First, we distilled specific tool design requirements resulting from our analysis of both bedside interviews and patients’ actual use of a set of interactive tools during their inpatient care. Through these means, we were able to determine patients’ initial in situ responses to the tools ^29,36,49 while also capturing patients’ experiences over a period of in-hospital care. By conducting patient interviews after 1–2 days of unsupervised use, we captured participants’ responses to updating data as well as other situational factors. This revealed unexpected usage (eg, patients accessing “static” data, such as the Home Medications list, frequently, to resolve the correctness of the data, and many individual visits to MedlinePlus by a single user) and provided situated responses (eg, electronic communication preferences, based on needs arising during care). These insights answer questions in informatics research related to the acceptability of this class of technology for use during postsurgical inpatient care and can inform the design of future tools to better accommodate expected use cases.

Second, although we see our tools as only one of many needed innovations to support patients’ participation in their own care, ^29,36,45 we focused on medication-related tools independently, in order to uncover details of patients’ needs pertaining to this class of data. General medication management tools ^21–26 are typically designed with the goal of promoting patient adherence and compliance for a smaller number of medications, with fewer changes expected to medication regimes and a longer time frame for decision making related to a medication’s side effects and benefits. To manage the typically large numbers of medication orders in the postoperative cardiothoracic surgery setting, we arrived at techniques to summarize order data to allow for quick medication identification based on dosage, schedule, and class while providing summarized explanations for many drugs simultaneously. Our techniques also structure and organize these data to support patient awareness of medication changes and new medication orders.

Prior studies by Dykes et al. ³⁶ identified medication schedules (eg, “daily routine schedules”) as being important for patients during cooperative care planning. The patients who participated in our study also prioritized daily schedules for each of their medications, desiring these in the primary view of the Hospital Medications tool. However, we learned that patients not only emphasized the value of these schedules for care planning and communication, as found in the intensive care and oncology populations, ³⁶ but that they also tracked medication therapies in order to maintain a sense of their health status and therapeutic progress, to support their own monitoring of side effects and responses to replacement therapies, as well as to play a role in maximizing medication safety. They achieved these goals through the use of the Patient Notepad and Home Medications tool together with the Hospital Medications tool.

Our findings related to medication safety concur with those of Dullabh et al., ⁴ who reported that patients used the MyGeisinger system to resolve errors or incompleteness in their own medication data, proposing corrections that often resulted in revised patient records and increased the accuracy of medication lists. In this study, patients’ acceptance of the system influenced patient safety: patients who initiated medication feedback had used MyGeisinger features, including messaging, more often than the average user. By grounding our study efforts in TAM-based concepts to confirm the ease of use and the usefulness of our tools, we aimed to promote such acceptance of our application in the inpatient setting.

Our findings related to the importance of monitoring also blur, to an extent, the boundaries of patient work ⁴² toward self-management of care – often conceptualized as relating to the patient’s everyday life – and the inpatient setting. Brennan and colleagues ⁶ assessed technology-enhanced clinical practice, which combined professional nursing with patients’ and nurses’ use of computer systems, finding discernable benefits in early post-discharge stages of care for chronic cardiac disease patients. To facilitate patients’ long-term self-management, the authors of this study included education, self-monitoring, and communication features in the design of their computing systems. ⁶ We found that patients used and appreciated the same types of features in the inpatient setting; although they relied on caregivers to prescribe and administer medications, patients consulted short explanations and MedlinePlus resources about their therapies and strove to maintain awareness of their own therapeutic progress while logging questions and concerns for their care team. An important avenue for future work includes closer examinations of whether and how inpatient tools can effectively promote skills and attitudes that, in turn, impact patients’ longer-term self-management of care.

Finally, we explored patients’ attitudes toward electronic communication about inpatient medication data and use cases, preferences, and concerns related to the handling of inpatient electronic questions and comments, which we expect to be different from electronic messaging through traditional patient portals. ⁷ Sun et al.’s ¹¹ content analysis of a large sample of electronic patient-provider messaging in the MyPortal system (open to all the adult patients of a large Midwestern medical center) found that patient-provider messaging was generally related to maintaining health over time, ie, used for prescription-related requests and appointment scheduling (primarily) as well as to discuss diagnostic test results, requests for updates on their general physical status, psychosocial support, and advice.

Although future work is needed to compare inpatient messaging behavior, on a large scale, to traditional patient portal messaging practices, we were able to glean patients’ preferences for communicating electronic messages and receiving electronic replies through the EHR. Patients’ desires (ie, quicker replies, the potential to “triage” questions by importance, aids for memory) illustrate a shorter decision-making and education timeframe and the prioritization of recovery over the general health maintenance needs found in traditional patient portal messaging. ^7,11 However, many of the concerns raised by the patients who participated in our study echo the findings of Sun et al. (namely, patients’ need to reconcile the different media used to communicate that could result in “unattached portal messages” – ie, messages that could have been answered in face-to-face meetings or phone calls, with no electronic record – as well as their desire to direct messages to specific clinical caregivers in order to maintain relationships with them ¹¹ ). Future studies comparing content analyses of patient-caregiver messaging across different care contexts are needed to confirm our field observations. In addition, as noted by Dullabh et al., ⁴ further studies are needed to explore how patients can effectively provide feedback and manage electronic questions and replies related to a variety of clinical data.

Study Limitations

Our study has several limitations. Like many qualitative studies, we enlisted a relatively small number of patients and clinical pharmacists to participate. The study was conducted at a single academic medical center in an urban location, and, therefore, our results may not be generalizable to other settings. Patients who did not speak English or who were physically unable to use a tablet computer were excluded from the study.

We encountered some challenges regarding the fidelity of the data in the system audit logs, specifically that we could not differentiate between patient and visitor usage of the application. To overcome this challenge, we attempted to deduce from the patients’ interview responses, in combination with the audit data, who had used the application during the unsupervised use sessions. There were also limited types of interaction data available in the audit logs: individual “swipe” gestures, keystrokes, and browser-initiated state changes (eg, use of the “Back” button) were not captured. In future studies, adding usability software to the tablet computer may enable fine-grained recognition of these behaviors.

Finally, although we captured rich feedback from patients, some spouses and clinical pharmacists, we did not investigate the perspectives of other clinicians, patients’ other family members, or other “care partners.” ⁵³ Further research should investigate design decisions that support family members’ use of patient engagement software tools, including privacy and confidentiality considerations.

An important design challenge concerns how data reporting tools can incorporate expert communication strategies that consider both the technical and emotional complexity of the reported data. When patient-facing technology is widely deployed, it will reach people in fluctuating emotional states and might deliver information that has life-altering implications. Research focused on the development and testing of automated methods for presenting medication information to patients, particularly considering emotional sensitivity, is a critical next step in designing effective patient-centered systems and applications.

CONCLUSIONS

In this article, we presented the myNYP Inpatient application and described a user-centered process to align the design of interactive medication-tracking tools with the information needs of postoperative cardiothoracic surgery patients. This is the first study, of which we are aware, that examines how such tools can meet the specific medication information needs of patients that arise in the inpatient setting. Through a two-phase pilot study with postoperative cardiothoracic surgery patients and clinical pharmacists, we found that the interaction and presentation techniques of the tools matched the abilities and preferences of the target population. We described how such tools structured inpatient medication lists and facilitated inpatient communication in ways that postoperative cardiothoracic surgery patients and clinical pharmacists found valuable. Important design considerations for future iterations of these tools concern careful formatting, presentation, and timing of the delivery of detailed inpatient medication data to both individuals and family members that considers the technical and emotional complexity of data, data privacy, and electronically mediated patient-provider communication in the inpatient setting.

CONTRIBUTORS

Drs L.W., S.B., S.F., and D.K.V. conceptualized the study design and interpreted the study data. Drs L.W. and D.K.V. contributed user interface design and software development and analyzed the audit data. Dr L.W. managed the Institutional Review Board processes and protocols and wrote the manuscript. Drs L.W. and A.S. developed the usability questionnaires. Ms J.W. and Ms J.P. collected informed consent from patients, conducted inpatient training sessions and interviews, and produced transcriptions of the audio recordings. Drs L.W., S.B., and A.S., along with J.W. and Ms J.P., contributed the thematic analysis. Dr S.R. coordinated study efforts, served as a scientific advisor, and cared for study patients.

FUNDING

This work was supported by Agency for Healthcare Research and Quality grant numbers R36HS021393 and R01HS21816, National Library of Medicine grant number T15LM00707, and National Institute of Nursing Research grant number T32NR007969.

COMPETING INTERESTS

None.

ACKNOWLEDGEMENTS

The authors would like to thank Tasnim Raza, MD for his support in implementing the study, Jun Yang for his efforts in implementing the myNYP Inpatient application, and Danielle Cauthen for her assistance with the design of the user interface.

REFERENCES