An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Kevin Fiscella, MD, MPH, Jonathan Tobin, PhD, Subrina Farah, MS, Wendi Cross, PhD, Jennifer Carroll, MD, MPH, and Amneris Luque.
Author Information and AffiliationsPatient engagement is the foundation for improving health care, particularly for people living with HIV (PLWH). The Patient Activation Measure (PAM) is a widely used and validated metric for assessing patients' engagement and self-management empowerment.
Patient activation is associated with improved adherence, better health outcomes, and lower costs. There is a dearth of scientific data on how to improve patient activation—particularly among individuals with the lowest activation, who are often disproportionately minority, lower-income, older, and less educated than those with higher levels of activation. The Research team used a community-based participatory research approach involving PLWH, HIV clinicians, community-based organizations, and researchers to develop a patient activation program for PLWH. Get Ready and Empowered About Treatment (GREAT) uses a smart, web-enabled device (Apple iPod) with an electronic Personal Health Record (ePHR).
Our primary aim was to assess the impact of the GREAT program on patient activation (using PAM) among PLWH. Secondary outcomes included improvements in eHealth literacy (eHEALS), decision-making self-efficacy (DSES), patient involvement in care (Perceived Involvement in Care Scale [PICS]), patient-reported adherence to antiretroviral therapy (ART) and viral load suppression, health status/quality of life (SF-12), receipt of evidence-based care, and reductions in patient activation disparities.
We recruited 360 PLWH from 4 practices in greater Rochester, New York, and from 4 federally qualified health centers in the New York City metro area. Participants were randomized 1:1 to the GREAT intervention or control arm. Patients in the control arm received usual care from their HIV clinician during the intervention period and usual care HIV case management assistance based on patient needs and resources available in the practice.
Patients in the GREAT training program received six 90-minute group-based training sessions on how to use the iPod and ePHR and how to search the web for health information, and they had a coaching session before their HIV visit. All clinicians received one 60-minute training session focused on supporting patient empowerment. All participants received their own iPod, although participants in the control arm did not receive the iPod until 12 months postrandomization. The primary outcome was changes in the PAM. Secondary outcomes included changes in eHEALS, DSES, PICS, ART adherence and viral suppression, SF-12, and receipt of recommended care (ie, preventive screening, testing, and immunizations received during the 12 months following randomization). Participants were assessed at baseline (T0), 6 to 8 weeks postrandomization (T1), and at the end of the study (T2)—ie, 12 months postrandomization—using data abstracted from charts. We examined bivariate relationships and used mixed models that controlled for site and cohort effects.
Participants who were randomized to the intervention had statistically significant improvements in PAM (P < .05). Effects were greatest among those in the lowest quartile for PAM scores at baseline. Significant improvements were also observed for eHealth literacy (P < .001) and involvement in care (P < .05). No statistically significant effects were observed for DSES, ART adherence, confidence in adherence, HIV viral suppression, SF-12, or receipt of evidence-based care. Intervention effects were similar by race/ethnicity and education level except for eHealth literacy, where effects were stronger for minority participants.
A multicomponent intervention improved patient activation and other empowerment-related constructs. Effects on patient activation were largest among those with lowest baseline patient activation; however, the intervention did not improve ART adherence, viral load, or receipt of evidence-based care.
Findings were limited by lack of full blinding of participants and research staff. Effects were largest among participants with lowest baseline PAM scores.
Patient activation is defined as “understanding one's role in the care process and having the knowledge, skills, and confidence to manage one's health and health care.”1 Patient activation is critical to improving patient self-management and reducing avoidable emergency department (ED) visits and hospitalizations.2 Lower levels of patient activation contribute to disparities in health care3; HIV care is a prime example. Poor and minority patients, including persons living with HIV (PLWH), report lower partnership with their clinicians,4 miss more office visits,5 ask fewer questions during their visits,6-8 report less confidence in self-management,9 and more frequently miss doses10 or stop taking their antiretroviral therapy (ART).11 Patients with low levels of education and/or low income tend to know less about their medications,12 report fewer benefits from cancer screening,12 and report less involvement in cancer screening decisions.13 In other words, social disadvantage likely undermines elements of patient activation—eg, knowledge, skills, and confidence—and is associated with less access to resources to support self-management.
Prior studies have found that low patient activation among PLWH contributes to disparities in adherence14,15 and viral suppression,16,17 and ultimately to disparities in HIV treatment outcomes,18-20 including hospitalizations and mortality.21,22 Death from HIV-related causes is among the top 10 leading causes of death among Black and Latino people aged 20 to 54 years.23 Among Black people aged 35 to 44 years, HIV ranks in the top 5 causes of death.23 We hypothesize that addressing disparities in patient activation provides one means for mitigating the impact of these health outcomes on PLWH.
For many people, patient activation depends on access to relevant web-based information—ie, disease-specific information and data on one's personal health information, which are typically available through vetted health information websites and electronic portals that have become part of most electronic health records (EHRs) systems, respectively.
Evidence is weak that availability of patient portals to EHRs alone, or even use of handheld electronic personal health records (ePHRs) alone, improves self-management.24,25 Access to ePHRs must be coupled with targeted interventions that address patient activation, yet relatively little is known about how to improve patient activation, including use of PHRs among PLWH.26,27 We were unable to identify any randomized controlled trials that tested interventions designed explicitly to improve patient activation among PLWH.
Our conceptual model for improving patient activation is based on the Capability, Opportunity, Motivation and Behavior model for behavior change.28 Patient activation requires motivation to assume greater involvement in one's own health care. It requires capability, including requisite knowledge and skills as well as confidence that one can take greater ownership. And it requires a supportive opportunity or context, including access to relevant online information and digital tools.29,30 Motivation toward a goal is based on perceived value and expectation for success in achieving it. Many socially disadvantaged patients have not had role models for patient activation, much less experience with how activation might be viewed as beneficial. Many socially disadvantaged patients lack access to reliable sources of health information beyond brief medical encounters. The experience of being disadvantaged and socially disempowered undermines self-efficacy, particularly during medical encounters where social distance is large (eg, between a person living in poverty and a physician).31 Last, the digital revolution in health care could leave older, lower-income, and less-educated people behind, creating worse disparities.32 Age, education, and income are each associated with access to a broadband connection,33,34 going online for health information,35,36 knowledge of computers and smart devices, and use of online patient portals.36,37 Addressing disparities in patient activation requires improving internet access (eg, through devices that can access the web via free WiFi or low-cost data plans) and providing disadvantaged groups with the knowledge, confidence, and skills necessary to go online when access is available.
The Get Ready and Empowered About Treatment (GREAT) project directly targeted these drivers of patient activation. It addressed motivation through a unique blend of external incentives (smart devices) and internal incentives; the latter were based on group work that promotes patient perceived autonomy (ability to enact one's own goals and preferences), competence (learning new skills), and relatedness (co-learning and assisting other group members). The project imparted knowledge through group discussion and promotion of highly reliable sources of information. Training addressed confidence through small achievable steps in proficiency, and scaffolding supported small next steps in the learning process.
A smart device was provided, and, notably, participants were taught to use an ePHR. The project leveraged the growing availability of portals by training patients to access high-quality health information and to learn how to transfer their information to their own ePHR for personal use. National data show that patients, particularly those with low income, low education, and chronic conditions, report greater benefits from having access to their ePHR.38 The study aims are shown in Table 1.
Study Aims and Hypotheses.
Initially begun in August 2009, our project was grounded in the principles of community-based participatory research (CBPR). From its inception, patients and stakeholders have been active partners in each phase of the project.39,40 In Rochester, New York, this partnership has included 4 PLWH, 4 HIV medical directors, and representatives from 2 HIV service organizations, in addition to 2 outreach workers and clinicians and researchers from a practice-based research network (PBRN). In New York City (NYC), Clinical Directors Network (CDN, www.CDNetwork.org), a primary care PBRN and AHRQ-designated Center of Excellence (P30) for primary care practice-based research and learning, has worked closely with the clinical and administrative staff, including the medical and nursing directors and HIV and peer educators at the 2 federally qualified health centers (FQHCs) that participated in the NYC pilot, and the current proposal is based on feedback from their active engagement as well as from other clinicians who participated in the CDN-PBRN.
These partnerships involved cross-training between members in CBPR approaches as well as research and methods of collaboration, including data sharing, decision-making, and revenue sharing. The meeting facilitator ensured that every member had the opportunity to speak and contribute. The group agreed that all aggregated data would be distributed among members and that key decisions, including next steps in the research project, would be decided by the group. Each participating organization was compensated for use of space, promotion of projects, and staff time; individual patient stakeholders were compensated based on time contributed to the project. Meetings were held quarterly at 7:00 AM at a convenient location and utilized both in-person (Rochester) and online (CDN, NYC) participation.
The original idea for the proposal came from a PLWH who pointed out that lack of information and limited access to their test results data, specifically their HIV viral load, represented barriers. Our patient partners further noted that many patients declined offers for printouts of test results due to fear of a breach of confidentiality (ie, a friend, family member, or acquaintance may inadvertently see printed personal health information). Our patient partners initially suggested use of a secure, password-protected electronic device to store patient information. The partnership considered various options, including use of standard cellphones (most lack the ability to run suitable applications), use of smartphones (prohibitive initial cost and need for wireless telephone contracts that many patients would not qualify for nor afford), and standalone devices (need for separate device beyond standard cellphone). After weighing these considerations, the partnership decided that the iPod Touch would be the best option because it has many of the features of smartphones minus the expensive service contract, is web-accessible at free Wi-Fi hotspots, and could access and utilize the full range of apps and websites as a smartphone. Notably, this device was chosen by our CBPR group years before the study began and before widespread availability of other electronic web-enabled devices.
Patients and stakeholders have been involved throughout the project and are now engaged in developing a proposal to PCORI for dissemination and implementation among community-based organizations serving PLWH. Patients wanted training to enhance their ability to self-manage their condition and communicate more effectively with their clinicians. We adopted an iterative, user-centered approach to designing the training and designing the ePHR application. Specifically, we collected both quantitative and qualitative data from training participants during each step of the process and used those data to inform next steps. Notably, we presented findings from each phase to our patient partners, clinicians, and other stakeholders for further input. This approach guided the type and frequency of training, design of the app, research questions, and outcome measures.
Our patients and stakeholders were interested in the following primary research question: Will our intervention improve empowerment and engagement in patients relative to their current, usual treatment? Thus, our overarching aim was to compare the effectiveness of our multimodal intervention with usual care. This decision was also based on the absence of existing patient activation interventions in routine practice.
To guide us in assessing a range of study designs on their relative strengths and weaknesses, we used PCORI Methodology Standards.41 Our primary goal was to generate findings with real-world implications—ie, how to empower PLWH while minimizing bias from unmeasured confounding. Thus, we selected a pragmatic randomized controlled trial as our design. It allowed us to compare preassessments and postassessments of participants randomized to the intervention (experimental treatment) with preassessments and postassessments from participants randomized to usual care (comparator = usual care). Given the nature of our intervention—including its reliance on training and specific tools—the risk for contamination between participants was deemed small. We considered other designs, including randomizing by practice site (cluster randomized trial), stepped wedge randomized trial, and staggered enrollment trials, but did not choose these designs due too few clusters, data collection burden, and a relatively short time frame (3-year project). To maximize generalizability, we opted for a pragmatic trial that employs broad “real-world” inclusion criteria for participants. We adopted inclusive recruitment methods and tracking enrollment by subgroup—eg, Black, Latino, White non-Latino, and highest educational level—to minimize enrollment bias in our sample. Without resorting to oversampling, we enrolled participants who mirrored the patient population in the participating practices and NYS more broadly.
Participants in both arms received the smart device (iPod Touch); however, only participants in the experimental arm received the device and were trained on its use during the trial. Usual care/control arm participants received the device and instructions on its use after the T2 follow-up measures were obtained (2-4 weeks following the participants' subsequent HIV visit). The study was approved by the University of Rochester Human Subjects Review Board and CDN IRB, and it was registered at ClinicalTrials.gov (NCT02165735).
Our eligibility criteria were broadly inclusive to reach a broad sample of typical patients living with HIV. We included patients with confirmed HIV diagnoses who were at least 18 years of age and who received care at a participating site. We excluded those who were unable to provide informed consent or had insufficient English proficiency to participate. We opted for direct recruitment of participants through onsite, face-to-face discussions at the time of patient visits and/or practice outreach to minimize any bias from differences in reading skills, social networks, or referrals. We adopted slightly different approaches, depending on the practice context. In Rochester, the clinician/designee introduced the study and referred interested patients to the research assistant (RA). The RA then met privately with potential participants to review the study protocol and obtain consent from those interested. In NYC, we used phone calls, based on patients identified from medical records, and intercepted patients who presented for visits with their HIV clinician during dedicated sessions for HIV care.
Rochester sites included 3 dedicated HIV practices and 1 FQHC with a large HIV population. The HIV practices included 2 hospital-based HIV practices and a practice largely devoted to the needs of HIV-positive and sexual minority individuals. The sites in the greater NYC area were 4 CDN-member FQHCs with large HIV-positive populations.
We designed an intervention to promote patient activation among PLWH by targeting key drivers (motivation, knowledge, skills, confidence, and technology). The intervention consisted of (1) a customized ePHR for PLWH using an iPod Touch to access the internet, (2) six 90-minute group-based training sessions on use of the iPod and the ePHR and on searching the web for health information, (3) a previsit coaching session, and (4) clinician training in supporting patient empowerment. Sixty-five participants who required extra help (or missed a session) met individually with trainers to catch up, for a total of 72 extra sessions.
We selected the Apple iPod Touch® based on its widespread availability and adoption, versatility (number of applications, web-enabled), intuitive interface/usability, relatively low cost ($199), no requirement for phone or internet service provider contract, and overall desirability and “coolness” factor.
We loaded each device with our customized ePHR, developed for PLWH, named UrHealth. We had previously created this app from a user-centered design with one-on-one interviews and then formally piloted it before undertaking this study; we did not conduct further usability testing during the current study. Key features include (1) drop-down menus of common HIV medications with accompanying pill photographs; (2) common lab tests with brief, understandable explanations; (3) a setting to create reminders for appointments and taking and refilling medications; (4) a personalized “prompt list” of potential questions for patients to ask their clinicians. These questions are generated from data the patients enter into their devices, such as date of birth. For example, if the patient is less than 26 years old, he or she will be prompted to ask about the HPV vaccine (unless he or she has already entered dates of its receipt). The patient can personalize these prompts by prioritizing them and, most important, to augment them with their own formulated questions. Participants provided written consent for us to track use of the UrHealth App.
The iPod represented a potential incentive or “hook” to encourage participants to enroll. We explicitly targeted key intrinsic motivational factors (autonomy, competence, and relatedness) as ways to engage patients during the groups. Each of the six 90-minute sessions focused on development of a basic eHealth competency by encouraging patient practice, mutual support and assistance, and celebration of successes. Participants themselves entered the data; trainers provided guidance and encouraged participants to bring their device to their appointments and confirm data they entered with their clinicians, although no direct digital data exchange took place between the practice EHR and the ePHR. Thus, clinicians had access only when participants specifically showed them the device. Key tasks are summarized in Table 2.
Key Tasks for Group Training.
After participants in the intervention arm completed their last session, a health coach met with each patient before his or her next HIV office visit. This one-time individual coaching session was designed to reinforce skills learned during the group training and to prepare patients to be engaged in their office visit with their clinician. A training manual and fidelity checklist addressed this one-on-one session (see Appendix). Key tasks for the previsit coaching session are summarized in Table 3.
Key Tasks for Previsit Coaching.
We trained all clinicians to engage patients in the visits through one 60-minute, educational “lunch and learn”—ie, a CME-accredited session that addressed the purpose of the project and basic clinician behaviors to facilitate empowerment. We used the Invite-Listen-Summarize model42 of communication and incorporated evidence-based communication skills.43-45 We employed training approaches used in previous studies.46-48 These approaches are supported by systematic reviews of randomized controlled trials of clinician training.49,50 We demonstrated the behaviors via brief videos with standardized patients (Table 4).51 We believe these approaches can help foster greater uniformity in clinician receptivity to activated patients, even though clinicians may be exposed to patients with different levels of activation.
Key Tasks for Clinician Training.
We developed a detailed training manual (see Appendix) for training professional staff and peers to conduct the training sessions. We conducted train-the-trainer sessions followed by remote videoconferencing training. Sessions emphasized enthusiasm, active engagement of the group through questions and sharing, staying on task and time, and ensuring that all members can rehearse a particular skill during the session. We assessed fidelity using an observation checklist based on key process and content elements for each session. Trained observers checked for fidelity criteria during in-person observations. We retrained trainers when fidelity drift occurred.
Consistent with features for a pragmatic trial, we did not adopt an attention control; instead, patients randomized to the control arm (usual care) received care according to the practice's guidelines. Control patients received the same smart device as did those in the intervention; however, they were not trained in using it, nor did they receive the ePHR app during the study. Most practices had case managers who assisted patients with addressing access barriers and promoted adherence. Case managers were available to all patients regardless of study arm, although their availability differed by site.
Data were collected at baseline T0 (immediately before randomization) and at T1 (8 weeks postrandomization and shortly after the activation training), T2 (2-4 weeks following the participants' subsequent HIV visit), and T3 (12 months postrandomization). RAs administered survey questions to participants at T0, T1, and T2. RAs abstracted medical record data at T3. Neither participants nor RAs were fully blinded.
Our primary outcome, the Patient Activation Measure (PAM), was chosen by our patients and stakeholders. We used the short-form PAM, a 13-item scale that assesses patient knowledge, skill, and confidence for self-care.52 Raw scores are transformed from 0 to 100. PAM is reliable (Cronbach's α = .91 with a Rasch person statistic = 0.81)52 and has previously been used to assess patient activation among PLWH.26,53,54 Notably, the higher PAM levels are prospectively associated with meaningful outcomes such as improved adherence and lower ED use by PLWH.2 We aimed to capture related dimensions including eHealth literacy (eHealth Literacy Scale [eHEALS], Cronbach's α = .94),55,56 patient involvement in care (Patient Involvement in Care Scale [PICS], Cronbach's α = .73),57 and patient self-confidence in health care decision-making (Decision-making Self-Efficacy Scale [DSES], Cronbach's α = .92).58
Using a structured case report form of explicit criteria, trained RAs performed chart abstraction for viral loads and evidence-based preventive care (see Appendix). When a screening, immunization, or test was absent from the record, it was coded as not performed. The timing of data collection is summarized in Table 5.
Study Measures, Data Source, and Timing.
The possible range of the key scales is shown in Table 6.
Measures and Range of Scores.
Our study statistician generated sequential identification (ID) numbers using computer-generated random numbers stratified by site (Rochester or NYC). Before participant enrollment, 360 sequentially numbered, opaque, sealed envelopes were created. Each envelope contained the IDs on the outside of the envelope and experimental arm assignment (intervention or usual care) on the inside. The RA opened the envelopes sequentially, only after the participant had completed the baseline (T0) assessment, so that the arm assignment was concealed from the RA before the T0 assessment.
We took steps to minimize missing data by ensuring that participants responded to all items. We obtained multiple contacts' telephone numbers from participants and made repeated efforts to track them down. Our analysis strategy relied on mixed-effects models of all available observations from participants. This approach optimizes use of nonmissing data under the same ignorable missing data assumption (“missing at random” [MAR]) that underlies the validity of standard multiple imputation procedures.65 We used multiple imputation based on all respondents when fewer than half of items were missing from a particular scale. In those rare (<1%) instances where half or more of items were missing, the scale was treated as missing. We tested for the MAR assumption using Little's test.66 We also created dummy variables for whether the variable was missing and assessed associations with variables. We conducted sensitivity analyses in which we excluded participants who responded with consistent extreme values to scales.65 For the main effects analysis, we used an intent-to-treat approach such that all participants were analyzed based on the arm to which they were initially assigned, regardless of their level of participation in the intervention. All analyses were conducted using SAS version 9.4.
For H1.1 to H3.1, we estimated detectable difference using the method described by Donner and Klar,67 a significance level of .05, and a minimum power of 80%. The intracluster correlation coefficient (0.0207) was estimated from our pilot data and is consistent with other studies.68 Based on planned enrollment of 360 participants and an estimated 15% dropout rate, we calculated power to detect a standardized effect size of 0.51 (difference in mean/SD for the full sample) for each of the measures in Aims 1 to 3. An effect size of 0.35 for PAM is associated with statistically significant although modestly clinically higher ART adherence and viral load suppression.53
For H4.1 and H4.2, we ran Monte Carlo simulations to determine treatment interaction effects. Based on the simulated data, we ran generalized linear model procedures and observed 80% power to detect a difference of 3.7 in PAM between (theoretical range, 1-100) low- and high-activation groups based on a sample size of 300.69,70
Our general approach was to compare the 2 study arms in terms of changes in the dependent variable from baseline, using separate models for patient activation, eHealth literacy, and decisional self-efficacy outcomes using GEE models that controlled for clustering with site (Rochester and NYC) and within training cohort. In our mixed models, we included study arms; cohort; timeline; race; ethnicity; gender; age; computer use; education; familiarity with an iPod, iPad, or iPhone; income; internet use; and marital status as covariates. We followed a forward stepwise selection of significant factors in the final model from the full model. All models were adjusted for any potential confounders, including arm; cohort; time; race; ethnicity; gender; age; prior general computer use; education; familiarity with an iPod, iPad or iPhone; income, internet use; and marital status. To assess the intervention effect, we entered experimental arm as an interaction term with time (arm × time). We imputed missing values (<1%) using multiple imputation.65,71 To assess the potential for extreme values to affect the results, we conducted sensitivity analyses in which we excluded participants who reported consistently extreme values for a particular scale. In secondary analyses, we used the same statistical models using a “per protocol” analyses among participants who used the UrHealth app at least once in the 6 months following their group training.
To address Aim 4, which represents our heterogeneity of treatment effects (HTE) analysis, we entered interaction terms into the models with treatment arm. Given our focus on addressing disparities in social advantage—eg, low health literacy, minority race/ethnicity, and low education—we included prespecified interaction terms based on lowest quartile cutoffs for activation and eHealth literacy in addition to minority race/ethnicity and low education. When the interaction term was significant (P < .05), we conducted stratified analyses to assess the impact of the intervention on these subgroups. We also assessed whether clinician behavior affects outcomes, by entering an interaction term for clinician behavior using the Instrument on Doctor-Patient Communication Skills,64 measured at baseline (T0) using a median cutoff.
A total of 360 participants were enrolled between August 2014 and March 2016, including 240 from the greater Rochester, New York, area and 120 from the greater New York City area. A total of 694 participants were screened for eligibility, 471 met eligibility criteria, and 360 were randomized (see consort diagram, Figure 1). All 180 patients randomized to the intervention arm were considered for the final analyses. From the control arm, 179 participants were considered for the final analyses, and 1 person was excluded after they were later found to be ineligible for the study due to previous participation in the GREAT pilot. The characteristics of participants in each arm are summarized in Table 7. There were no statistically significant differences between the 2 arms.
Participant Flow Through the Study.
Baseline Characteristics of Study Participants.
Participants' baseline demographics were well matched between the intervention and control arms (Table 7), and 84% of intervention participants attended at least 1 group training session. The mean number of sessions attended (not including individual makeup sessions) among all participants in the intervention was 3.6 out of 6; 3 in 4 participants (76%) attended a previsit coaching session.
We assessed use of the UrHealth app by identifying those using the app during the training sessions, among those assigned to the intervention arm and by assessing frequency of login and screen use. We were able to identify 133 participants (74% of those in the intervention) who used the app once ever and 109 participants (61% of intervention participants) who used the app at least once in the 6 months following their training (control participants did not have access to the app during the study). The 109 participants used the app post-training for a mean of 18 times (a mean usage time of 131 minutes during the 6-month period following their group training). In order of frequency, the most commonly opened screens were tests, events, and the to-do list (which included the question prompts).
Our primary study outcome was patient activation as measured by PAM. Patient activation at baseline was similar for intervention (72.1 ± 1.34) and control (70.8 ± 1.36) arms (P > .05). In mixed models that controlled for patient characteristics and clustering of patients by cohort and by site (all outcomes discussed are presented in Table 8), patient activation increased more in the intervention arm than in the control (73.35 ± 1.13 vs 70.53 ± 1.14; P < .027). In sensitivity analyses, results remained significant when extreme values were omitted. In the final mixed model for changes in PAM, cohort group, education, and familiarity with an iPod/iPad/iPhone were statistically significant predictors of improvement in PAM.
Effect of the GREAT Intervention on Study Outcomes.
Results were also statistically significant for mixed models that included potential confounders for eHealth literacy, patient perceptions of involvement in care, and patient perceptions of clinician communication (ie, Instrument on Doctor-Patient Communication Skills [IDPCS]). Each, except for the IDPCS, remained statistically significant after further adjustment for multiple comparisons. Again, exclusion of extreme values did not alter the results. Effects were not significant for adherence self-efficacy, patient-reported adherence, or improvements in viral load. There was no significant change in mental or physical health or receipt of evidence-based care.
We observed marginally statistically significant improvement in patient perceptions of clinicians' communication skills (P = .049; Table 8), although this was no longer statistically significant after adjusting for multiple comparisons.
Results were not statistically significant for patient confidence in adhering to cART, patient self-reported adherence to cART, or undetectable viral load (which improved in both arms). We conducted secondary analyses among participants with a detectable viral load at baseline and among participants who reported missing at least 1 dose of their ART in the past week. Neither of these subgroup analyses showed significant effects. We examined changes in viral load among participants who used the app at least once and found no effects for the intervention compared with control. In secondary analyses, we explored absence of effects by examining the relationships between PAM and adherence and adherence and viral load. We found statistically significant but weak correlations between PAM and self-reported adherence (R = 0.21, P < .0001) and between self-reported adherence and undetectable viral load (R = 0.15, P < .01). These findings suggest that the path between PAM and viral load, while present, is too weak to show detectable effects in our analysis (see Table 8 for all outcomes).
There was no statistically significant improvement in overall receipt of evidence-based preventive services (Table 8); a sensitivity analysis that restricted the intervention arm to participants who used the app at least once also failed to show statistically significant effects.
There was no statistically significant improvement in mental or physical health status (Table 8).
Participants with PAM scores in the lowest quartile showed statistically significant improvements when receiving the intervention (Table 8). The relative magnitude of these improvements is shown in Figure 2. In a mixed model of the subgroup of patients in the lowest quartile of PAM scores (n = 90), the intervention resulted in an increase in PAM of 3.9. In contrast, effects were smaller (3.3) for those in the upper 3 quartiles and results did not reach statistical significance. The lowest quartile approximates persons who scored at level 1 and 2 for PAM. While regression to the mean likely contributes to greater within-group improvements—ie, high-quartile and low-quartile groups—it is unlikely to explain the smaller differences between groups—ie, between the intervention and control arms. Last, we examined a 1-stage improvement in PAM among participants who scored at PAM stage 1, where the impact on outcomes is likely to be greatest. The results showed that the intervention was associated with an improvement to higher stage (odds ratio: 2.10; 95% CI, 1.08-4.07).
PAM by Quartile Groups.
We repeated all the analyses among participants who used the app at least once post-training. In mixed models, we observed larger effects (often double) among app users than among those observed in the full analysis (see Appendix). However, statistically significant effects were the same as in the intent-to-treat analysis—ie, statistically significant effects only for patient activation, eHealth Literacy, and involvement in care.
We conducted a similar analysis for eHealth literacy and observed similar although relatively larger effects. Participants in the lowest quartile (n = 99) showed the largest improvement from the intervention.
We observed no statistically significant interactions by race/ethnicity or education for the interventions effects on tPAM, suggesting that similar effects were observed across these groups (Table 8); however, we did observe a statistically significant interaction in eHealth literacy for minority participants. Stratified analysis showed that minority participants showed greater improvements in eHealth literacy from the intervention than did nonminority participants (Figure 3). In a mixed model that adjusted for other characteristics, the minority × intervention interaction term was statistically significant (P = .0275).
eHealth Literacy Changes for Minority vs Nonminority Group by Intervention.
There were no adverse events, serious or otherwise, that had any plausible relationship to the intervention.
In this randomized trial, we found that a multimodal patient self-management program for PLWH improved patient activation, eHealth literacy, and involvement in care.
Clinicians and practices serving PLWH are confronted with the question: What can be done to improve patient activation and eHealth literacy among low-income PLHW? Our findings suggest that this multimodal intervention, which involves group training cofacilitated by patient peers, can accomplish this aim.
To our knowledge, this is the first randomized controlled trial of an intervention that was explicitly designed to improve patient activation among PLWH. McInnes et al conducted a pre-post pilot (no control group) of a similar intervention among 14 low-income veterans living with HIV or HCV. Patient activation improved, but results were not statistically significant.54
Our effects are potentially important. An observational study showed that a 5-point difference in PAM was associated with an 18% and 8% higher odds for ART adherence and viral suppression, respectively.53 Among patients with PAM scores in the lowest quartile in our study, our intervention resulted in a nearly a 4-point adjusted increase in PAM compared with the control arm. Among participants with the lowest PAM scores (level 1), the intervention resulted in doubling of the odds ratio of moving up to the next level of activation. This increase from level 1 to 2 has been associated with a 30% reduction in the odds of having an ED visit.72 Similarly, patients with PAM level 1 are significantly more likely to develop a new a chronic disease in the 3 years of observation, with odds ratios ranging from 1.21 to 1.31.73
Other trials have assessed interventions for patients with other conditions—eg, diabetes, heart failure, hypertension, and in disease prevention.74-83 Our effects on PAM are larger than those from a randomized controlled trial of a Wellness Portal.77 Results are similar to those observed from a randomized controlled trial involving low socioeconomic status patients being discharged from the hospital who were randomly assigned to trained community health workers or usual care,84 but smaller than those from a heart failure self-management trial.74 Our intervention appears to be unique in being explicitly designed by and for PLWH who are disproportionately poorer, less educated, and of minority status than the general population. It was designed to provide PLWH not only with knowledge and confidence in managing their health and health care but also with skills and tools (ie, a smart device) that enables self-management.
We did not observe statistically significant changes in HIV viral suppression or improvements in processes of care (ie, screening, immunization, and other preventive care). However, our study was not designed nor powered to examine viral load suppression, which is strongly linked to ART adherence. Nearly, three-quarters of participants had undetectable viral loads at baseline. We did not specifically focus on adherence skills and habits, which are strongly linked to viral load suppression.85 Based on a modest relationship between PAM and adherence,53 the improvements in PAM in this study were likely too small to show detectable effects on adherence or viral suppression.
Our intervention offers potential for implementation and scalability. The primary challenge is staff time to conduct training, and support for patients who cannot afford devices. However, many HIV practices, particularly those receiving Ryan-White HIV/AIDS funding, offer case management and group peer support programs.86 Another option would be use of reimbursable group medical visits to conduct these trainings.87
Our training manual provides a step-by-step, session-by-session approach for implementation. The key resources needed include staff and peer facilitators for groups, meeting space for groups, access to Wi-Fi (to minimize data use), and means to support patients who don't currently own a smart device acquiring one. Despite persistent disparities by age and education in use of smart devices, national data show steady increases in such use, primarily for smartphones. Our experience is that a minority of patients lack access to these devices. In addition, we have developed a version of our ePHR that runs on Android devices, including Android phones. This new Android version will improve implementation by no longer requiring use of an iOS device (eg, iPod Touch or iPhone). We plan to use this device and offer a Spanish option if our PCORI dissemination and implementation funding is awarded.
Our patient inclusion criteria were very broad. Inclusion of community health centers probably enriched the participation of lower SES patients. We excluded very few participants. We excluded patients who participated in pilot versions of the program and a couple who lacked basic English language proficiency. We have addressed this limitation by developing a Spanish version of the intervention. Our practice criteria were also broad. All HIV practices in greater Rochester participated in addition to 4 large FQHCs in New York and New Jersey.
Consistent with hypothesis 4.1, we observed that patients in the lowest quartile of activation showed the greatest improvement in activation. Some of this change might represent regression to mean. Notably, we observed comparable improvements by race, ethnicity, and education, with the exception of eHealth literacy, where minority participants improved more with the intervention than did nonminority participants.
It is possible that patient activation training coupled with 1-time coaching before a single visit is insufficient to improve process of care. Our results in this respect contrast with those from a randomized controlled trial of an ePHR for patients with serious mental illness.88 Differences may reflect the intervention design—eg, in that study patients were instructed to personally hand the clinician a printout of their PHR. Differences from that study also likely reflect the large increase in the number of outpatient visits that the Druss et al intervention had compared with the usual care arm (14.9 vs 0.5).88
Ceiling effects hindered our ability to detect differences in suppressed viral load. More than 3 in 4 participants had viral suppression at baseline. Thus, we were significantly underpowered to detect changes in this secondary outcome.
Our design was pragmatic. We compared the intervention with usual care rather than with an attention control group; thus, we cannot exclude the possibility that groups per se improved patient activation. However, we are not aware that patient activation can be improved without addressing the core attitudes and skills that undergird this construct.
For disadvantaged patients, a 15-minute office visit is often insufficient time to address their concerns, provide evidence-based preventive care, engage in shared decision-making, deliver relevant chronic care including support for patient self-management,89 and address social determinants of health. About 5 minutes of each primary care visit are devoted to the patient's chief concern and a minute or less each to the remaining issues.90 Most primary care visits address at least 3 concerns,91 but disadvantaged patients present with a greater number of concerns,92 many of which are related to the social determinants of health. Our clinician training emphasized helping patients to prioritize their concerns for the visit. For example, if the patient prioritized discussing back pain, a new rash, and life stress, then many of the questions prompted from the ePHR might not have been addressed. Thus, a single visit may not have been sufficient to address the combination of prompted issues and the patient's own primary concerns. Further, clinicians might not have universally ordered tests, immunizations, or cancer screening requested by the patient.
Further, it is conceivable that clinician training and cross-patient contamination sensitized clinicians to care processes among both treatment arms, thus biasing results toward the null. Last, while we documented continued use of the app by participants, it is possible that use was not sufficiently intensive to generate changes on more distal health outcomes.
Lack of blinding of participants and staff represents a potential threat to internal validity. Although participants were not aware of the specific study aims, participants receiving the intervention were aware of its focus—ie, training patients to be more activated. This knowledge might have biased their responses to the various scales. Similarly, while we asked RAs not to ask participants which group they were assigned, we cannot be certain that participants did not provide this information, directly or indirectly. Thus, we cannot exclude the potential for lack of blinding of participants or staff to have biased the results.
There is a cost (roughly $200) associated with purchase of an iPod Touch™, and people must carry this device along with their cellphone. We have recently addressed this challenge by developing a version of the app for Android phones. Suitable Android phones can now be purchased for $60, to avoid use of 2 separate devices.
Use of smart devices can be a challenge for older persons with low vision. One participant mentioned that use of a stylus made key pressing easier; as result, we provided styluses to participants.
The GREAT program was designed to promote patient activation and eHealth literacy to mitigate the impact of the digital divide on patient activation. Our trial results show that while we succeeded in enhancing patient activation, the primary outcome, those improvements were not associated with increases in medication adherence, reductions in HIV viral load, or increases in the delivery of preventive care. This project provides a foundation for next steps, including training modules that are more targeted toward ART adherence. Based on participants' feedback, there is strong patient desire for additional modules that address lifestyle concerns, including nutrition and physical activity, and use of apps that support them.
Nonetheless, there are significant barriers to using ePHRs to address health care disparities. Currently, only a handful of patients access these web-based ePHRs.38 Moreover, while the disparities for minorities in online access continue to decline overall,93 disparities in access to and use of web-based ePHRs persist. Specifically, Black people, Latino people, and those with low SES or those who are older use web-based ePHRs less frequently than do their counterparts.38,94-100 These disparities reflect the combination of lower health literacy combined with the so-called digital divide. Data from the US Census Bureau attest that poor, minority, and older patients use the web less often.101 National surveys (and our own preliminary data) suggest that 3 factors largely contribute to the digital divide:33 cost of web access, lack of interest, and lack of skills. In the absence of targeted mitigation strategies,102 access to web-based ePHRs will likely worsen rather than ameliorate health care disparities. One potential mitigating factor is that Black and Latino people are more likely to use smartphones to access the web than are White people.34 We created an Android version of the app to take advantage of this growing use of smartphones by minorities.
We confirmed our primary hypothesis—that this intervention would improve patient activation—but we did not observe improvements in adherence or recommended care. Findings from this study on activation add to the evidence on successful approaches that improve patient activation among patients with chronic conditions and address a notable gap in evidence regarding interventions for activating PLWH. The intervention effects seem to be greatest among those with lowest activation, suggesting potential targeting of this group.
The internal validity of these findings is strengthened by use of a randomized design, use of validated measures, intervention fidelity checks, good rates of participant retention, and very low rates of missing data; the primary threat to internal validity was inability to blind participants or research assistants.
External validity is strengthened by broad participant eligibility criteria, high rates of trial participation among potentially eligible patients, and use of trained peer facilitators. External validity is further strengthened by diversity among the practices that participated, including all 4 major HIV practices in Rochester and a diverse range of FQHCs in the New York City/northern New Jersey region.
The intervention was effective in improving patient activation, eHealth literacy, and involvement in care. These constructs are associated with important health outcomes, such as ED visits, avoidable hospitalizations, adherence, improved decision-making, experience with care, and reduction in disparities.2,103-105 Nonetheless, the GREAT intervention did not show statistically significant improvements in self-reported ART adherence; viral load suppression; or uptake of evidence-based testing, immunizations, or preventive care. Future studies should consider focusing activation interventions on nonadherent participants with detectable HIV virus.
Given these considerations, the GREAT intervention is likely to be of interest to practices and health plans concerned with improving patient activation. However, the absence of effects on adherence and update of evidence-care suggests that this intervention alone is not sufficient to appreciably improve these outcomes. Research should address targeted, possibly supplementary, approaches to adherence and uptake of evidence-based care for PLWH.
Research reported in this report was funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (AD1306-03104). Further information available at: https://www.pcori.org/research-results/2013/helping-people-living-hiv-learn-skills-manage-their-care
G.R.E.A.T. Training Manual 2014 (PDF, 7.6M)
Medical Abstraction Form (PDF, 535K)
Fiscella K, Tobin J, Farah S, Cross W, Carroll J, Luque A. (2018). Helping People Living with HIV Learn Skills to Manage Their Care. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/12.2018.AD.130603104
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
