Effectiveness of a personalized health profile on specificity of self-management goals among people living with HIV in Canada: a protocol for a blinded pragmatic randomized controlled trial

Goal setting – a unique aspect of self-management

Goal setting occupies a pivotal place in self-management of chronic conditions. Self-management interventions that include a goal setting component have been shown to be effective in increasing patients’ progress towards better health outcomes in the face of chronic diseases^15–20. Patient-centered goals are defined as goals set with patients actively engaged, through discussing goals, setting plans, and evaluating their condition afterwards²¹.

Until now, studies on goal setting in chronic disease management have used goal setting as a means to improve health outcomes with goals set collaboratively. Self-defined goals, without clinicians’ input, have not received much attention. It could be argued that at the population level, where there is no direct input from the healthcare professionals, setting good quality goals and making plans for action are difficult and can threaten achievement of targeted health outcomes.

One possible strategy to empower patients to self-manage their condition is providing feedback through sharing specific information about their health profile to stimulate the setting of specific goals²². Previous studies show that personalized feedback is more effective for improvements in lifestyle behaviors, specifically in terms of health risk indicators, such as body mass index, diet, and physical activity^23–25. Personalized rather than general feedback is more accepted by patients for management of health and well-being as they can relate it to their own health targets and make informed choices about risk assessment and lifestyle modification²⁶. This study contributes to the understanding of the quality of patient-formulated self-management goals (hereafter self-defined goals) by testing the effectiveness of providing feedback using a personalized health profile.

Objectives

This study aims to estimate, among people living with HIV, to what extent providing feedback on their health outcomes, compared to no feedback, will affect number and specificity of self-defined goals.

Conceptual model for the present study

Individuals set their health-related goals according to their personal health reference point “the level of mental, emotional, and physical health people believe possible or necessary to make the progress they seek”²⁷. A health reference point is formed by the individuals’ health status and circumstances and the trade-offs they are willing to make. Health feedback profiles can provide an understanding of the health reference point that can enable individuals to set goals that ultimately produce the desired effect on health outcomes. Because deterioration in the health outcomes of people with a chronic condition occur over a long time period, goal setting can be used as an intermediate step to raise awareness of areas for improvement and set in place action plans. These are the building blocks of self-management.

Cognitive ability is one factor that could modify the effect of health feedback on self-defined goals. Cognitive deficits such as problems with memory, attention, comprehension, executive functioning, or speed of information processing are likely to interfere with writing high quality goals²⁸. As self-management approaches have mostly been implemented as part of clinical programs, the recommendations have been tailored to the individual’s cognitive capacity²⁹. Figure 1 depicts the relationships explained above.

Figure 1. Conceptual model.

Context for the study: Brain Health Now (BHN)

This project is a sub-study of the BHN study (http://brainhealthnow.mcgill.ca), which is a multi-site Cohort Multiple Randomized Controlled Trial (cmRCT) across five sites in Canada³⁰. This design is based on using a fully characterized cohort as the basis for recruiting people into multiple trials. The characterization of the cohort was based on the Wilson-Cleary model, a biopsychosocial model integrating multiple components of health-related quality of life³¹. The particular focus is brain health, recognized as a multi-dimensional construct reflecting the brain’s role in cognition, mood, emotional stability, motivation and energy³². The outcome measurement strategy for the BHN comprised 20 different self-report, patient-reported outcome measures, one performance outcome measure, as well as questions about age, sex, education, drug and alcohol consumption, smoking, and nutrition. Cognitive ability of all participants has been measured with the Brief Cognitive Ability Measure (B-CAM).

The first phase of the study – Positive Brain Health Now (+BHN) – was conducted over the course of 5 years from 2013 to 2018, recruiting and characterizing more than 800 HIV+ men and women with four assessments, scheduled 9 months apart. Details of this study have been published elsewhere^30,33,34. This cohort has been extended (Action for Brain Health – ABHN) to cover the period until 2023 with annual follow-up.

This study has been approved by institutional research ethics (McGill University Health Centre Research Ethics Board, ABHN_Goals 2020-5728) and was registered as a trial on 25 November 2019 (Clinicaltrials.gov NCT04175795).

Study design

A blinded, stratified, pragmatic randomized controlled trial design will be used to assess the impact of the personalized health feedback profile on the number and specificity of self-defined goals. Participants will be randomized (1:1 allocation ratio) into one of two groups, the intervention or control group, stratified based on gender. The randomization is computer generated by the study statistician using randomization.com. For this study, it will not be possible to blind the participants as the intervention is their personalized health profile. To avoid bias in assessing outcomes, those collecting and analyzing the data will be kept blind to group assignment.

Study population

The study population for this trial, ABHN_Goals, is people who were enrolled in either +BHN study or its extension (ABHN) who agreed to be contacted for further sub-studies and who have access to the internet. Briefly, cohort members were people HIV+ for at least 1 year. Excluded were people with dementia, co-morbidity affecting cognition, substance abuse, or life-threatening illnesses³⁰.

Study procedure

All +BHN participants who consented for sub-studies will be contacted by the intermediary staff (i.e., designated research assistants) either by phone or e-mail to be recruited for this study. Participants will be asked to provide and/or confirm their e-mail addresses so that an e-mail containing the information regarding the study, their unique access code, and the link to access to a specialized web-based goal setting platform (available in French and English) could be sent to them. The intervention group will receive their Dashboard via e-mail along with instructions on goal setting and tips to improve brain health. The control group will receive only the goal setting instructions and tips. Figure 2 shows the study flow diagram.

Figure 2. Flow diagram for goal setting trial in HIV cohort.

Data collection will be through a web-based goal setting platform (LimeSurvey hosted on a McGill server). All participants will receive a token to access to the platform. After electronically signing the consent form (Extended data³⁵), participants will be directed to the main page to write their goals and fill out questions that follow each goal. Participants in both groups will be asked to think of the top 3 to 5 actions they would like to take to improve their health condition in the assigned boxes. For each goal, participants will be also asked questions about their self-regulatory plans (i.e., barriers and potential solutions).

A maximum of 2 weeks will be considered so that participants can think and reflect on their goals (a reminder e-mail will be sent after the first week). Participants who do not answer the survey within the two weeks will receive a second reminder e-mail to fill out the online survey. Participants who do not answer to the survey after the second reminder will not be re-contacted. At the end of the study (2 weeks + 2 weeks extension after the second reminder e-mail = 4 weeks in total for each participant), participants in the control group will receive their Dashboard via e-mail. Only research assistants will be in communication (phone and/or email) with participant. No in-person visit is required for this project. Step by step instruction as to how fill out the survey has been also made as a PDF file in English and French for participants who might need some help filling out the survey.

Intervention

The intervention tested in this study is inspired by the knowledge-to-action framework³⁶. It involves providing feedback [about modifiable contributors of health and quality of life] by sending participants their personalized health profiles. As part of the knowledge translation plan, a personalized profile of specific actionable health outcomes with evidence of life impact^34,37 termed as “My Personal Brain Health Dashboard” has been created for each BHN participant (Figure 3). The Dashboard has 15 different actionable items (selected from pre-existing validated measures) and covers information on brain health outcomes, health and quality of life ratings, and lifestyle factors. All items have sufficient track record of validity and are known as a reasonable substitute for multi-item questionnaires for obtaining and monitoring patient relevant outcomes “on aspects of health that only patients can report (e.g., stress, pain, fatigue, etc.)”³⁸. Cognitive debriefing interviews were conducted with members of the HIV community in Montreal and Vancouver to pre-test the Dashboard for its interpretability and usefulness for goal setting. Interpretability of the Dashboard was improved by minor modifications to item wording, adjustment of colors to differentiate outcome categories, and adding extra explanations.

Figure 3. My Personal Brain Health Profile.

A requirement for setting achievable goals is that people have contemplated and prepared themselves to set goals and take actions. However, if patients do not realize that there are aspects of their life that are suboptimal and amenable to intervention, they cannot set goals for targeting these areas. Our hypothesis is that the Dashboard helps patients to formulate achievable goals by helping them to understand their health status. The Dashboard is designed to 1) provide a summary of patients’ health outcomes in terms of their cognitive and physical functionality, emotions, perception of health and quality of life along with changes in these health outcomes over time between their visits and 2) help patients compare their results with the optimal range and see how far or close they are to reach the target. This way, patients would be prepared to move towards identification of topics for setting goals. The Dashboard will act as a prepared list of items where patients can find the areas that they need to improve, select, prioritize, and set goals for. As mentioned earlier, cognitive ability of all BHN participants has been measured with the B-CAM. This study is thus ideally placed to take into account the effect of cognitive ability on goal setting.

Main outcome

This study will test the hypothesis that providing feedback to people living with a chronic condition such as HIV, using personalized health profile, results in a person’s higher ability to formulate specific goals. The two main outcomes in this study are the number and specificity of self-management goals. For goal specificity, text mining techniques will be used. This outcome will be reported as a discrete variable – number of specific words (matched to a developed lexicon) per person-goals. The expectation is that the intervention group will set more goals and have more words matching the lexicon than the control group. The numerator for this estimator is total number of specific words per group; the denominator is the number of cumulated number of goals set per person (e.g., if 10 people set 5 goals each, this represents 50 person-goals).

Patient goals are mostly formulated collaboratively with the client and a clinician according to SMART¹ criteria, which facilitates setting good quality goals. There are no standardized and validated measuring criteria for goal formulation. Furthermore, the literature on goal setting shows that goals are most often evaluated according to the person’s progress towards their goal or the outcome (i.e., measure of change)³⁹.

To measure the quality of self-defined goals, which most often lack the structure and format of clinically SMART goals, techniques of text mining would appear to be well suited. Text mining helps us to extract and understand specific elements of the goals from the textual data. In this study, text mining techniques and algorithms will be used to 1) distinguish components of a patient goal, 2) evaluate specificity of components in a person goal according to our previously collected vocabularies (lexicon), and 3) expand our lexicon based on mass evaluation of around 400 to maximum 2000 self-defined goals.

Overview of the text mining procedure

We have created an initial lexicon (i.e., start list of words; Extended data³⁵). Our lexicon contains two separated lists of vocabularies. One list contains words that convey vague information and another list contains words that make goals more specific. This lexicon was made using two sources of data. The first was a set of goals set collaboratively (with a healthcare professional) during a project on health outcomes post-hospitalization. Words that represented each component of SMART criteria were manually extracted. The second source arose from cognitive interviews conducted with members of HIV community piloting a goal setting exercise for this future trial. Those words that corresponded to each component of SMART criteria were also manually extracted. This preliminary list of words was presented to a multidisciplinary group of researchers and clinicians familiar with goal setting in the field of physiotherapy, occupational therapy, psychology, epidemiology, and nursing. Through this process, initial goal setting lexicon was shaped.

Next, lemmatization function will be used to reach the root (i.e., word’s lemma) of the words in the lexicon. This process, lemmatization, will be done in order to enrich the lexicon with words that have similar meanings to the words in our lexicon, synonyms. A set of guidelines has been defined for evaluating each component of the goals. The guideline and further details of the text mining algorithms are explained in the Extended data³⁵.

Self-defined goals (written in English or French) in text format will be classified using text mining algorithms to find words that represent components of SMART (available in the lexicon). Total number of words per person-goals that match the lexicon will be reported according to the components of SMART.

Analytical approach

The main outcome of this study is a rate, based on the cumulated count of the number of specific words per person-goal. Thus, the analysis aims to compare rates of the response variable (i.e., word count/person-goal) across groups. The total number of specific words per person-goals will be calculated for each group and Poisson regression will be used to estimate the rate ratio and 95% confidence intervals and compare rate ratios between men and women using an interaction term:

Log (E(Rate of Goal Specificity)) = (groups) + (Gender) + (B-CAM scores) + (groups*B-CAM scores)

Characteristics of participants (e.g., age, sex, education, employment status, etc.) in both groups will be analyzed using descriptive statistics. SAS 9.4 software (SAS Institute Inc) will be used for statistical analysis.

Confidentiality

All personal and demographic information of participants have been collected during the BHN study. All direct identifiers such as names and email addresses have been removed and is only available to the research assistants who are responsible for recruitment. Participants are all identified by a code number (username assigned to them). All information that is collected for the study will be kept in a secure and password protected file and stored for 7 years at the Centre for Outcome Research and Evaluation, McGill University Health Centre research Institute.

Sample size

The sample size calculation for this “two-arm randomized controlled trial” is based on the requirements for the hypothesis tests for a rate ratio of formulating specific goals with a two-sided level of significance of 5% and a power of 80%. In order to find the value for the null hypothesis, a pilot study with 20 people (10 per group) has been done. Samples of goals have been collected to compare the difference in the number of specific words per person-goals (i.e., rates) between self-defined goals (i.e., e^β0) and collaboratively set goals^{2 40} (i.e., e^β1). For self-defined group, the average number of goals was 3 (30 person-goals) with the mean number of specific words per goal of 4, for a rate of 120/30 or 4. For collaboratively set goals, the average number of goals was 5 (50 person-goas) with the mean number of specific words per goal of 6 for a rate of 300/50 or 6. Using these data, a rate ratio (RR) of 1.5 (6:4) and 95% confidence interval of (1.12, 1.99) was calculated.

Considering equal group allocation, an event rate ratio of 1.5 has been used as the point estimate. Considering a projected survey response of 20%⁴¹, an estimation of confidence interval for comparison of two proportion showed that a minimum sample size of 210 per group would provide a 95% confidence interval that excludes 1 (i.e., 1.23, 1.83).

Limitations and potential solutions

This study will be the first trial using text mining in the context of goal setting and is not exempt from limitations. A main limitation for this study is the potential for low participation (i.e., low survey response), particularly for the control group as they might be less motivated because they will not be receiving their Dashboards until after survey submission. The measurement and analytic approach of this study requires a relatively high participation rate. Although this has been factored into the sample size calculation, to maximize recruitment, the recruitment and reminder e-mails will remind participants in the control group that their Dashboard will be sent to them shortly after filling out the survey.

Another pitfall facing this study is the potential of ambiguous results. This limitation arises from the complexity of natural language (i.e., words that can have more than one meaning and could be interpreted in different ways according to the context). Several steps will be taken to minimize this problem. First, the preliminary version of the lexicon has been developed using both well formulated goals or supervised goals (i.e., set in collaboration with a clinician following SMART criteria) and self-defined goals (i.e., with a written goal setting instruction). Second, the domain knowledge (i.e., words related to chronic disease and HIV self-management) has been integrated into our lexicon using relevant literature. So far, no text mining tool has included the domain knowledge. Third, different tagging algorithm will be tried to find the most accurate and robust result. Moreover, 20% of the goals will be randomly selected to be evaluated by a panel of 5 experts in the field of rehabilitation goal setting and these results will be compared with the ones obtained through text-mining.

Dissemination

The results of this trial will be publicly available and published in peer reviewed journals and disseminated at relevant conferences and meetings at the completion of the study. This study is novel in testing the usability of text mining to measure goal quality. The findings of this work will be rigorous and could be used as an example with other chronic health conditions. We aim to share the text mining methodology used in this research and its advantages and limitations in measuring goal quality. Also, we offer a comprehensive goal-related lexicon specific to people living with HIV in Canada. While our sample here is people with HIV, the lexicon could be used as a start point in similar research with other chronic conditions in Canada.

Underlying data

No underlying data is associated with this article.

Extended data

Open Science Framework: Goal Setting in HIV, https://doi.org/10.17605/OSF.IO/FW4SX³⁵.

This project contains the following extended data:

Reporting guidelines

Open Science Framework: SPIRIT checklist for ‘Effectiveness of a personalized health profile on specificity of self-management goals among people living with HIV in Canada: a protocol for a blinded pragmatic randomized controlled trial’, https://doi.org/10.17605/OSF.IO/FW4SX³⁵.

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).