Article Text
PDF
Abstract
Hydroxyurea (HU) is an effective but underused disease-modifying therapy for patients with sickle cell anaemia (SCA). EMBRACE SCD, a sickle cell disease treatment demonstration project, aimed to improve access to HU by increasing prescription (Rx) rates by at least 10% from baseline in children with SCA.
The Model for Improvement was used as the quality improvement framework. HU Rx was assessed from clinical databases in three paediatric haematology centres. Children aged 9 months–18 years with SCA not on chronic transfusions were eligible for HU treatment. The health belief model was the conceptual framework to discuss with patients and promote HU acceptance. A visual aid showing erythrocytes under the effect of HU and the American Society of Hematology HU brochure were used as educational tools. At least 6 months after offering HU, a Barrier Assessment Questionnaire was given to assess reasons for HU acceptance and refusals. If HU was declined, the providers discussed with family again. We conducted chart audits to find missed opportunities to prescribe HU as one plan–do–study–act cycle.
At initial measurement, 50.2% of 524 eligible patients had HU prescribed. During the testing and initial implementation phase, the mean performance after 10 data points was 53%. After 2 years, the mean performance was 59%, achieving an 11% increase in mean performance and a 29% increase from initial to the last measurement (64.8% HU Rx). During a 15-month period, 32.1% (N=168) of the eligible patients who were offered HU completed the barrier questionnaire with 19% (N=32) refusing HU, mostly based on not perceiving enough severity of their children’s SCA or fearing side effects.
Reviewing patient charts for missed opportunity of offering HU with feedback and evaluating the reasons of declining HU via a questionnaire were key components in increasing HU Rx in our population.
- Chronic disease management
- Healthcare quality improvement
- Paediatrics
Data availability statement
Data are available upon reasonable request. The data to make the charts are available upon reasonable request.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Hydroxyurea is the best drug available to decrease morbidity in sickle cell anaemia but remains underused.
WHAT THIS STUDY ADDS
Quality improvement activities of provider education, assessing patient barriers and preferences, and chart audits with practice feedback contributed significantly to improve hydroxyurea prescription rates and patient acceptance.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
We demonstrated that a provider quality improvement network changed provider practice in prescribing hydroxyurea and increased patient medication acceptance.
Introduction
The Education and Mentoring to BRing Access to CarE for SCD (EMBRACE SCD) Network was a Health Resources and Services Administration (HRSA)-sponsored regional collaborative of eight southeastern states (Alabama, Florida, Georgia, Kentucky, Mississippi, North Carolina, South Carolina and Tennessee). Its purpose was to improve access to high-quality care for individuals living with sickle cell disease (SCD) and to increase the number of providers adhering to National Heart, Lung and Blood Institute (NHLBI) guidelines1 by engaging in different quality improvement (QI) projects.
Florida paediatric haematology centres chose as their primary QI project to increase the number of hydroxyurea (HU) prescription (Rx) rates in individuals with sickle cell anaemia (SCA), the most severe form of SCD. HU is the oldest and best studied disease-modifying therapy to treat SCA, with a long track record of safety, efficacy2 3 and cost-effectiveness.4 HU works by increasing fetal haemoglobin levels, which interferes with haemoglobin S polymerisation.5 The 2014 NHLBI Expert Panel guidelines1 strongly recommend HU for adults with SCA who have three or more severe vaso-occlusive crises during any 12-month period, chronic anaemia interfering with daily activities, or have severe or recurrent episodes of acute chest syndrome. The same recommendations are applied to children based on clinician experience. A recommendation of moderate strength suggested offering treatment with HU without regard to the presence of symptoms to infants aged 9 months and older based on the double-blind placebo-controlled BABY HUG Study,3 which showed benefit of HU when compared with placebo to prevent vaso-occlusive complications.
Unfortunately, HU remains underused in children and adults with SCA. In 2010, a survey of the members of the American Society of Pediatric Hematology/Oncology who cared for children with SCA revealed that only 9% of the responding members had 50%–90% of paediatric patients on HU, while 10% respondents had <10% patients on HU.6 An analysis of the Truven Health Analytics-IBM Watson Health MarketScan Medicaid database from 2009 to 2015 showed there was an increase in HU Rx rates from 14.3% in 2009 to 28.2% in 2015, which is still low, for a multistate population of children 1–19 years of age.7 Providers’ decision against prescribing HU included concerns about patient adherence with daily HU and laboratory monitoring and thinking post-pubertal females would not adhere to contraception while on HU.6 All participating sites across the EMBRACE SCD Network were charged with implementing QI projects focused on increasing prescription rates of HU. Administration in adults is outside of the scope of our report as the involved centres in this project treat children and adolescents with SCD. Therefore, we will concentrate on the populations followed by three paediatric programmes in Florida.
We present here details of our paediatric HU prescription QI project. Our aim statement and expected study outcome was to increase HU Rx rates by at least 10% from baseline (June 2019) by December 2021 in our combined population of over 500 eligible patients. We assessed that an increase of at least 10% was feasible and clinically meaningful.
Methods
Definition of SCD and SCA
SCD encompasses all genotypes including haemoglobin SS, S-ß0 thalassaemia, S-ß+ thalassaemia, haemoglobin SC and other more uncommon heterozygote combinations where one of the abnormal haemoglobins is S. SCA refers to the most serious genotypes, the first two enumerated in the list above. In this project, as we are talking about HU Rx, we are specifically targeting patients with SCA although all genotypes are followed in the clinical practices.
EMBRACE SCD Florida team
The three medical centres participating in the EMBRACE SCD Florida QI project were (University of Miami (UM); Salah Foundation at Broward Health (SFBH); Johns Hopkins All Children's Hospital (JHACH)) with (UM) serving as the lead site. In collaboration with the EMBRACE SCD Network Data Coordinating Center (DCC) at Atrium Health, Charlotte, North Carolina, USA, the EMBRACE SCD Florida team developed their HU QI aim statement, developed a key driver diagram and determined a comprehensive change package to drive iterative plan–do–study–act (PDSA) cycles geared at increasing the acceptance of HU among paediatric patients receiving care at the participating centres. Two of the centres designated as site 1 and site 3 for the purpose of this report had a single paediatric haematologist provider exclusively for patients with SCD assisted by one or two nurse practitioners; site 2 had three paediatric haematologist-oncologists taking care of patients with SCD. The nurse practitioners and the other physicians were educated about the QI project and reminded about the safety and benefit of HU in the attempt to increase prescription rates and establish uniformity in practice standards. Two national QI consultants (CC and LN) from the DCC had regular QI coaching calls with the sites and provided improvement science guidance to the entire network during the study. Members of three SCD community-based organisations (Sickle Cell Disease Association of America (SCDAA) Miami Chapter, SCDAA St. Petersburg, Sickle Cell Foundation, Inc., Tallahassee, all in Florida), which also participated in EMBRACE SCD Florida, contributed with ideas and feedback regarding the design and implementation of this project. In addition, all team members contributed during monthly virtual meetings to data-driven change using PDSA cycles.
QI methodology
We used the Model for Improvement8 to achieve our goals, starting with the aim statement of increasing HU Rx in the eligible population by 10% from baseline within a 30-month period (June 2019–December 2021). The outcome measures were the per cent of eligible patients with SCA who were prescribed HU from the total of eligible children and parent self-reported barriers to accept HU. We performed rapid PDSA cycles9 to test interventions, and prepared run charts10 and control charts11 to make data-driven decisions.
Patient identification and eligibility to receive HU
The first activity was to determine who were eligible patients for this QI project. To this end, we developed or expanded centre databases to verify all active patients, defined as those patients followed within the 2 previous years, which constituted our baseline patient pool. Patients were also classified as eligible for HU if they had haemoglobin SS or Sß0 thalassaemia, were not on chronic transfusion and were 9 months of age and older. We noted on the databases whether they were on HU or whether they have previously declined HU, per documentation in the electronic medical record (EMR).
Health belief model
As part of the QI coaching, all participating centre investigators who were the HU prescribers were educated about the health belief model (HBM) (figure 1)12 in July 2019 during a 2-hour in-person site initiation visit at each participating site and thereafter virtually as part of monthly meetings. HBM has been used successfully in public health in promoting preventive health measures, such as the acceptance of influenza and COVID-19 immunisations among patients and providers,13–15 in cancer screening16 17 and increasing the adherence to treatment18 among many other settings, and served in this QI project as the starting point to conduct conversations with patients and families regarding HU. The HBM has six constructs: perceived susceptibility, perceived severity, perceived barriers, perceived benefits, cues to action and self-efficacy. Both susceptibility and severity lead to a perceived threat. Both the barriers and the benefits determine a response. A decision to decline treatment with HU may result from the patient or parent perceiving no benefits or perceiving high risk from the therapy and low disease severity. Our implementation plan involved providing consistent education about HU benefits and risks to all eligible patients and caregivers. As part of HBM, clinicians educated clients about the maximum benefits and low chances of side effects for HU treatment to emphasise perceived benefits and decreased perceived barriers to action. Parents/patients were reminded about the risks involved with SCD (threat of illness). Once a decision was made, cues to actions that would motivate a person to change a behaviour could include reminders, information and guidance. The goal of these cues to action is to keep the person motivated to succeed. For self-efficacy or the confidence of patient’s ability to take action or decide to start and adhere to treatment, we provided encouragement and guidance, which are important to keep patients engaged and adherent to treatment.
The health belief model.
Further provider education
Another counselling technique, which was introduced to providers, was motivational interviewing.19 Participating providers were required to watch videos on the subject as one of the early preparatory activities to begin the QI project during a site visit from the lead institution team. The motivational interviewing approach helps to focus discussions with the patient by allowing patients to provide their thoughts on how to best improve on their health challenges, such as choosing the best way of adhering to treatment.
Definition of Rx rate
A person was considered taking HU if the Rx was given in the last 6 months and the patient reported taking HU at the next clinic visit. Rx with refills were repeated as clinically indicated. HU Rx was a combined decision between the provider and the patient or the patient’s parent depending on age.
Clinical monitoring during HU treatment
Clinical monitoring was not altered during the QI project, ranging from every month to every 3 months based on usual care, except that during the early COVID-19 pandemic (spring 2020), telemedicine substituted in-person monitoring. In general, monitoring consisted of clinical evaluations including history and physical examination, and laboratory assessments (complete blood counts, comprehensive chemistry panel and haemoglobinopathy panel to measure haemoglobin F level, which increases with HU treatment).
Data collection and analysis
Each site created Excel databases with all patients with SCD who were served, with separate tabs for the patients who were eligible and those who were ineligible for HU treatment. Patients who were on HU were identified and the date of treatment initiation was written down. The number of patients <18 years who were eligible to take HU and who completed clinic appointments at each centre on the previous month was reported as the denominator monthly. The number of patients <18 years who were eligible to take HU and who completed a clinic appointment at each centre with a current HU Rx was also collected as the numerator monthly. If a patient declined or discontinued HU, providers documented the reason for declining or discontinuation in the EMR. Data were entered into the central EMBRACE Network REDCap database managed by the National Institute for Children’s Health Quality. No personal health identifiers were reported. Mean frequencies or percentages were calculated. Run charts were created by the University of Miami staff (SE) based on the QI data, correlating with successive interventions. Trends in the data were discussed at our virtual QI meetings each month to provide incentive to continue with QI activities. For this report, we prepared control charts using QI Macros software.
Intervention with change ideas
We used several interventions or ‘change ideas’ to achieve the QI goals of this project, which were introduced individually as PDSA cycles to determine each one’s utility. The key driver diagram on figure 2 shows the interventions. The specific provider-focused change ideas included (a) provider education about when to prescribe HU to all patients 9 months and older who were not on chronic transfusion and had SCA, (b) EMR alert to serve as a provider reminder that HU could be prescribed to a patient once we opened the patient’s EPIC EMR implemented at site 1 only, and (c) random chart audit with practice feedback. Each site investigator audited 10 charts from patients who have attended the site’s clinic in the previous month and who met criteria for HU to assess for missed opportunities for HU Rx. The audited charts were chosen at random using Google random number generator from all eligible charts. The per cent of charts showing patients on HU on the previous month was submitted to the QI consultants by email. The QI consultants provided feedback to all the practice leads in monthly virtual Zoom meetings. In addition, when performing their own audits, the site investigators became aware of the missed opportunities regarding who were the patients who were not offered HU. In the case of the site with several providers (site 2) but not in sites 1 and 3, the physician practice lead shared this information in weekly meetings attended by all three practice paediatric haematology providers to make the others aware of who were the eligible patients who were not on HU and review reasons why they were not. Interventions a and c were performed at the three sites simultaneously. Missed opportunity for HU prescription was defined as the failure of the provider to approach an eligible child and/or parent to discuss HU treatment.
Key driver diagram. ASH, American Society of Hematology; EMR, electronic medical record; HU, hydroxyurea; SCD, sickle cell disease; PA, prior authorization required for medication approval; CBO, community based organizaton.
The specific patient-focused change ideas included the following: (a) showing the patient/caregiver a visual aid that simply outlines the effect of HU on the red blood cells demonstrating less sickling over time. This tool was created by the EMBRACE SCD Network and reviewed for health literacy level from a figure adapted from the article ‘How I use hydroxyurea to treat young patients with sickle cell anemia’20; (b) sharing the American Society of Hematology (ASH) HU brochure with patients when providers discussed HU to assist in educating patients and parents; and (c) a brief patient or parent/caregiver HU Barrier Assessment Questionnaire was deployed to evaluate the reasons for accepting or declining HU. During a 15-month period (November 2019–January 2021), the medical provider or nurse practitioner gave the paper questionnaire during the clinical encounter to all patients who met the criteria (at least 6 months have passed since starting or declining HU) (online supplemental file 1). We used the barrier assessment responses to better understand what the patients’ reasons were for accepting or declining HU Rx. This information was used to re-educate the families about HU immediately following the questionnaire completion. The number of families who changed their mind and accepted HU was collected. In the case of the HU Barrier Assessment Questionnaire, site 1 began in November 2019, whereas sites 2 and 3 began in June 2020.
Supplemental material
Results
At the beginning of the QI project in June 2019, there were 125, 208, and 191 eligible patients at sites 1, 2 and 3, respectively (table 1). The offer and refusal rates at the different institutions were 90.4% and 28% (site 1), 33.6% and 0.5% (site 2), and 61.8% and 1% (site 3). The initial HU Rx rate was 50.2% of 524 eligible patients among the three clinical site databases with a range of 33.2%–62.4% among the sites, with 62.4%, 33.2%, and 60.7% at sites 1, 2 and 3, respectively. The PDSA cycle with the EMR alert, which was implemented in EPIC at the site 1 only, took several months to implement and did not help to identify more patients as it was deployed each time we opened the chart as it could not technically exclude those patients already on HU, creating extra burden for providers.
Individual site and overall results for hydroxyurea (HU) prescription (Rx) rates in the EMBRACE collaborative
Monthly chart reviews were conducted successfully by the three centres, assisting in identifying eligible patients to offer HU. This was carried out with a one-page data collection sheet (online supplemental file 2). We made changes over time in the data collection sheets to capture the eligible patients (only SS and Sß0 thalassaemia), instead of including milder sickle cell genotypes who are not eligible for HU in the random sample, and collected the reasons why the providers had not prescribed HU.
Supplemental material
The patient-targeted interventions in successive PDSA cycles (visual aid, ASH HU brochure and Barrier Assessment Questionnaire) at the three participating sites were conducted successfully by our provider staff. The first two were adopted without changes and were used to inform patients visually and in written format. The Barrier Assessment Questionnaire was shared with few patients or parents before implementation and changes were made according to feedback.
Figure 3 shows the aggregate SPC P-chart. A baseline was not established prior to starting the QI collaborative; therefore, we used the first 10 points as our baseline mean. The mean baseline HU Rx was 53% among all three clinical sites. The annotated aggregate control chart reveals two signals of special cause variation, which signal improvement: the first eight data points above the original baseline are a shift, resulting in a recalculation of the mean in an upward direction of 59%, which is an 11% increase over the initial calculated mean of 53%, and the second signal is a trend of six increasing points, which is a trend upward with the last data point being 65%. The EMR alert, red blood cell pictures and visual aid, ASH HU brochure, a barrier questionnaire, and monthly chart audits with feedback are change ideas and are visually represented in the figure as annotations. In April 2020, during the peak of the COVID-19 pandemic, there was a transient decrease in the HU Rx rate because many appointments were cancelled unless patients required urgent evaluations. Very soon, we started with telehealth audiovisual appointments, but patients were eventually brought back to in-person encounters within 2 months. The last change idea, which consisted of an audit and review of 10 random charts every month by each site investigator prompting group feedback from the QI consultants, increased provider awareness about missed opportunities and gave the most overall improvement for a single change idea. Once the investigators realised that eligible patients were not offered HU, the site team was prepared to offer it at the next scheduled encounter. The second signal of improvement, as demonstrated by the upward trend of six points, likely reflects the higher reliability of all sites implementing the change ideas. Each site SPC P-chart is shown as online supplemental file 3. By June 2021 (last point measured), the HU Rx rates per centre were 85%, 47.6% and 67.8%, corresponding to sites 1, 2 and 3, respectively, representing an increase of 36.2%, 43.4% and 11.7%, and a combined increase of 29.1%. Table 1 summarises the baseline and last HU Rx rates for the three sites.
Supplemental material
P-chart of aggregate hydroxyurea (HU) prescriptions (Rx) in the EMBRACE sickle cell disease demonstration project. ASH, American Society of Hematology; CL, central line; EMR, electronic medical record; LCL, lower control limit; RBC, red blood cell; UCL, upper control limit.
The results of the barrier questionnaire are presented in table 2. The rate of completion of the barrier questionnaire was 100% of the patients sampled although the administration rate was variable according to the centre with 71.8%, (N=94 of 131), 7.1% (n=14 of 196), and 30.9% (N=60 of 194) of eligible patients at sites 1, 2, and 3, respectively, by the time we concluded this change idea (January 2021). In total, 32.1% (N=168) of the eligible patients by January 2021 had been assessed with this questionnaire, most of them (56.0%) at site 1. In our combined sample of patients/parents who completed the barrier questionnaire, 81% (N=136) were taking it and only 19% (N=32) declined. From the patients who accepted HU, 67% (N=91) did not encounter problems and 33% (N=45) did. Twenty-four per cent (N=11) of those who encountered problems forced themselves to take HU. The most common reason for declining was not stated by 31% of the respondents who refused HU. However, fear of side effects and fear of taking chemotherapy impacted significantly in the decision to refuse HU. In site 1, from 12 patients/parents who have declined and were further educated, 3 of them changed their minds and agreed to receive HU.
Acceptance and refusals of 168 patients assessed with the hydroxyurea (HU) barrier questionnaire
Looking at the effect of the PDSA cycles on clinic flow, we did not find that they impacted on clinic flow, according to nurse practitioner/provider self-report. On the contrary, it increased our focus on the intervention and allowed for better patient education.
Discussion
Recognising the importance of HU for patients with SCA and the reported underutilisation, our QI project was designed to increase Rx rates by at least 10% from baseline in a population of 524 children. The three centres in Florida started with a mean prescription rate of 50% (range 33%–61%). By using a uniform approach across all three sites, we increased the HU Rx rate by 29% from entry among patients with SCA to a mean of almost 65% with a range of 48%–85% at the last point measured. We considered all children who were not on chronic transfusion and who were 9 months of age and older and had SS or Sß0 thalassaemia to be eligible for HU. Therefore, expanding HU Rx to all eligible patients, rather than concentrating on the very few severe patients, allowed for extending Rx to many patients with SCA. However, we still used clinical judgement and were less likely to offer if a child had a high baseline haemoglobin F level or a child did not have any sickle cell complications.
There was heterogeneity in the outcome percentage achieved at the different centres, depending on the baseline per cent, with the practice without an identified sickle cell provider (several paediatric haematologists/oncologists taking care of patients with SCD) having the lowest prescription rates. Of note, this practice, site 2, had the highest number of eligible patients at baseline and the highest increase in their HU Rx, which demonstrates their significant improvement effort. Familiarising with the NHLBI HU guidelines in the case of site 2 for these providers who were haematologists but did not specialise in SCD, and expanding HU Rx to all eligible patients in the three sites rather than concentrating in the most severe patients were very important contributors to our success as prescribers. The QI consultants provided extremely valuable guidance to our efforts in pursuing different interventions by giving QI education and feedback and helping us interpret charts, contributing to the success of our QI project. We demonstrated that provider and patient education helps to increase Rx rates over time and should be an ongoing effort nationally. However, education is best paired with other interventions or change ideas for sustainable improvement.
In studying the different change ideas, the monthly chart review audits with provider feedback and the addition of the HU barrier questionnaire in sites 2 and 3 were the changes associated with an upward trend in Rx (figure 3). The monthly chart review audit with provider feedback made providers more likely to prescribe HU when interacting with eligible patients at the next clinical encounter as they realised their previous missed opportunity. The benefit of such intervention has been previously reported in the QI literature.21 The barrier questionnaire facilitated discussion about fears and reasons for declining with patients and caregivers. Showing the effect of HU on red cells or educating families with written information or just having a EMR alert did not seem to produce an effect in achieving higher Rx rates among our practices. We postulate that patients or caregivers either did not read the information or preferred verbal discussion rather than written or visual images. The EMR alert that was only launched at site 1 was repetitive every time the provider opened the chart and did not add any further information to what the provider already knew. It is possible that the alert could have brought added benefit in a practice with multiple providers who could be less knowledgeable about criteria to offer HU.
The HBM has been previously used as the framework to survey caregivers of children with SCD in either determining reasons for missing clinical appointments22 or promptly seeking care for fever.23 We did not find any report using this model to promote HU treatment in SCD at the provider–patient discussion level. Learning about the HBM helped us to frame the information provided to the patients and to understand patients’ decisions better. The barrier questionnaire revealed that many patients who agreed to take HU did not face side effects or barriers, although one-third of the patients who were taking HU did.
This project differs from another published HRSA collaborative24 in several aspects. Our QI project used the HBM as the framework to approach and understand parents/patients when offering HU. We also used visual aids and the ASH brochure as educational tools, and the Barrier Assessment Questionnaire to assess the reasons for declining and barriers found while taking HU. In contrast to the goals of that previous collaborative, which was to document HU counselling, we focused on HU Rx, which in our opinion, approximates better to the NHLBI guidelines of making HU accessible to the highest possible number of individuals with SCA.
Our QI projects have several limitations. We do not have data on the per cent of patients on HU at the participating sites prior to the start of our QI project in June 2019 because we wanted to change clinical practice quickly and did not have the resources to accurately assess prior monthly data. Therefore, we are unable to determine whether our earliest interventions such as provider education and patient education materials impacted the rate of HU prescribing. However, we did demonstrate a significant improvement beginning May 2020 as illustrated on our SPC chart after several interventions were deployed. Another limitation was the number of barrier questionnaires completed compared with the total number of patients who were started on HU at the three sites combined related to low survey administration rates at two participating sites. Furthermore, we do not know the reason for refusal of 10 (31%) parents/patients who declined HU. When reasons were stated, the parent or patient did not deem that SCD was severe enough to warrant taking medication (perceived severity) or feared side effects (perceived barrier), considering HU to be a chemotherapeutic agent. Only a minority changed their mind on further education.
We overcame the challenge of incorporating a QI project into busy clinical practices. Time is required to gather and enter data in databases. This type of exercise made our practices more organised and built the groundwork for future QI projects. We recommend the incorporation of QI projects into clinical practice. A pre-analysis and post-analysis of clinical and haematological benefits and safety might be considered to verify effect in the future, such as the evaluation of pain episodes before and after the adoption of new HU prescriptions. However, we do not expect to find differences from what has been widely published. In our clinical practices, we have continued to update databases regarding disease-modifying therapies we prescribe with the purpose of providing the best available treatments to our patients.
Data availability statement
Data are available upon reasonable request. The data to make the charts are available upon reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by University of Miami IRB (eprost number 20180445). Each institution obtained IRB approval for the HU QI project. The IRB waived informed consent.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors OAA, JK, JJS and IO planned the hydroxyurea prescription quality improvement project. OAA, SE, HR-C and ELJC executed the work. CC, LN and TB-M conducted data analysis and critical consultation on the quality improvement project. OAA wrote the manuscript with significant contribution and revisions from all coauthors and also accepts full responsibility for the work as guarantor and conduct of the project, had access to the data, and controlled the decision to publish. All authors gave final approval to the version to be published and agreed to be accountable for all aspects of the work.
Funding This work has been funded by Health Resources and Services Administration (HRSA) under grant U1EMC31108.
Competing interests ELJC and IO report employment by Forma Therapeutics, now part of Novo Nordisk. IO also reports funding by Patient-Centered Outcomes Research Institute (PCORI) for another study.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.