Myelodysplastic syndromes (MDSs) constitute a heterogenous group of hematopoietic stem cell disorders [1] in which the bone marrow produces immature white blood cells, red blood cells, platelets, or a combination of all three. MDS can be rapidly progressive and is associated with a risk of evolution into acute myeloid leukemia (AML) [2]. AML, an orphan disease, is an aggressive cancer of the blood, consisting of a group of relatively well-defined hematopoietic neoplasms [1]. The disease begins as abnormal blood cells that are produced in the bone marrow and can spread throughout the circulatory system and beyond if not diagnosed and treated quickly [3]. Both AML and MDS are hematologic diseases that generally affect older adults; they are uncommon before the ages of 45 and 50 years and are most commonly diagnosed in patients who are in their late 60s or 70s [4,5].

The annual incidence of AML in the United States is approximately 3.75 cases per 100,000 people, and the incidence rate of MDS is approximately 3.30 cases per 100,000 persons [1]. White men have the highest incidence of MDS (6.9 cases per 100,000 people) [6]. In 2018, there were an estimated 19,520 new cases of AML, and there are approximately 14,275 new cases of MDS annually according to data from 2010 to 2014 in the United States [6]. The number of new cases diagnosed each year is likely to increase as the size of the elderly population increases.

Prognosis is poor for patients with AML and MDS, and effective treatment options are limited. Prognosis for AML generally decreases rapidly with age—5-year survival in adults is estimated to be around 24%; however, this drops to less than 5% in individuals aged 65 years and older [7]. The treatment approach for MDS depends on several factors, including the type and risk group of MDS and the patients’ age and overall health. Most diagnosed patients are older or in poor health and, thus, are not suitable candidates for a stem cell transplant, which is potentially the only cure for MDS. Alternative treatments for MDS, such as supportive care (transfusions or blood cell growth factors) along with chemotherapy, serve only to relieve symptoms and avoid complications and side effects. The standard treatment for patients with AML is intensive chemotherapy, typically involving cytarabine-based regimens, with the goal being to achieve complete remission. Intensive chemotherapy commonly requires prolonged hospitalization as it is associated with severe myelosuppression and an increased risk of treatment-related mortality. As a consequence, elderly patients with AML who present with poor performance status and/or comorbidities are generally not considered candidates for intensive chemotherapy because the risks are thought to outweigh the benefits. Until recently, treatment options for these patients were limited to clinical trials, nonintensive agents (hypomethylating agents or low-dose cytarabine), or best supportive care [8]. Reported treatment patterns in older AML patients vary; however, estimates using data from 2000 to 2014 suggest that only 40% to 60% of these patients receive any therapy for their disease [9], in part, because of a lack of options, but other reasons are not well understood [10,11].

Owing to poor prognosis and limited treatment options for patients with AML and MDS ineligible for intensive chemotherapy, there are corresponding unmet medical needs. These conditions also impair patients’ health-related quality of life (HRQoL) and are associated with high symptom burden [12]. Patients’ needs can be better addressed by gaining a deeper understanding of their perspectives on the disease, treatment options, and the impact of both on their lives and those of their caregivers. This information is valuable in the decision-making process—including end-of-life treatment decisions—but is challenging to capture using traditional methods or data sources (eg, electronic medical records and patient questionnaires) because of the rapid progression of the disease and patients’ age at diagnosis. Recent studies have highlighted the need to evaluate patient-reported outcomes for symptoms of AML or MDS to outline areas of focus for clinicians and researchers [13-15]. However, studies assessing experiences, preferences, and perspectives regarding treatment decisions for patients with AML or MDS are scarce, especially for those who are ineligible for intensive chemotherapy [16,17]. Those studies are crucial to better inform drug discovery and regulatory decisions and imperative for improving the patient experience.

Disease-specific platforms (eg, patient forums and discussion boards) on social media represent a virtual community where patients and their caregivers spontaneously share experiences and perspectives related to their disease and treatments. This information can provide an instant snapshot of the current humanistic and economic burden of illness in patients with AML or MDS and provide insight into the rationale for patients’ treatment decisions. The Food and Drug Administration reinforced this idea in June 2018, when it made recommendations encouraging the use of social media to shed light on the patient’s perspectives regarding symptoms and impacts of their diseases, stressing the opportunity to inform medical product development and enhance regulatory decision making [18].

The objective of this study was to use social media on disease-specific platforms to identify themes related to disease and treatment experiences of patients with AML or MDS who are ineligible for intensive chemotherapy and their caregivers. It aimed to capture the most important factors that drive patients’ treatment decisions by identifying patients’ priorities and their reasons for pursuing certain types of treatments over others, including preferences in end-of-life situations.

An initial search of disease-specific social media was conducted, focusing on the feasibility of addressing the study objectives (finalized on April 23, 2018). The search strategy focused on identifying AML and MDS–specific social media forums and discussion boards, using key search terms such as AML patient forum and MDS patient discussion. An exhaustive list of search terms used can be found in Multimedia Appendix 1.

Generic social media sites such as Facebook, Google+, and Twitter were not considered because of the added complication of filtering out irrelevant material and because access to data is governed by complex terms and conditions and data access fees. Searches were conducted using US and UK Google search engines, the first 5 pages of results were screened by title, and relevant forums were summarized. The searches yielded 41 sites for consideration for use in the study, which were assessed based on the following criteria:

Of the 41 sites identified during the initial search, 19 met the selection criteria. The following 3 forum sites with the highest number of posts were selected for this study:

Data extraction algorithms specific to each site were written in the R computational programming language (version 3.5.1 [23]) to extract posts programmatically from each site. Data extracted included username, post content, thread title, date of the post, and URL. A total of 44,456 posts (AML=21,654 and MDS=22,802) were extracted from 7811 users (AML=6373 and MDS=1438).

All data were programmatically deidentified at the time of extraction, and postal codes and addresses, email addresses, Web links, and telephone numbers were all removed. Following extraction, the data were processed and cleaned. Duplicate posts and non-Unicode Transformation Format-8 characters were removed. Misspellings were corrected using a dictionary of known misspellings in disease-related social media data compiled by the authors. The R hunspell package [24] was then used to identify additional misspellings. Despite these efforts, it is possible that some misspellings remained in the final data. Data were then assembled into individual user posting history files in .txt file format. These files contained all post content from each user chronologically. Extracted posts with the username deleted_user could not be assigned to user posting histories. Therefore, these were retained as individual posts and assumed to be individual users.

This study focused on patients with AML or MDS who were ineligible for intensive chemotherapy and their caregivers. The inclusion and exclusion criteria are provided in Textboxes 1 and 2, respectively. Intensive and nonintensive chemotherapies are listed in Textbox 3.

Following the manual review of poster history files, qualitative data analysis (QDA) was undertaken for each of the objectives. Posts were qualitatively reviewed in the R Qualitative Data Analysis Package. Thematic analysis was used, taking a top-down approach to data, using hypothesis coding (themes and codes determined a priori) to develop a coding dictionary based on the key objectives [25]. Text segments relevant to the objectives were highlighted, and open coding was used to add codes based on reviewed data. The occurrence of themes was summarized using descriptive statistics.

The posting histories of 892 users (AML=608 and MDS=284) including 20,265 posts (AML=15,435 and MDS=4830) were manually reviewed against the inclusion and exclusion criteria of the study, and, ultimately, 347 users (AML=220 and MDS=127) and 4176 posts (AML=1443 and MDS=2733) were included in the study. For the qualitative review, the 85 users included in the study contributed 1152 posts (AML=1132 and MDS=20). Of the 85 users, 73 (AML=68 and MDS=5) had posts relevant to the study. For the targeted qualitative analysis, 79 users (AML=40 and MDS=39) had posts containing the search terms, and on manual review of posts, 45 users (AML=27 and MDS=18) had posts relevant to the study objectives. For further details, see Figure 1.

The following 5 overarching themes were identified from the results of the QDA: humanistic burden, treatment decisions, unmet needs, life milestones, and economic burden. The number and proportion of users mentioning each theme are shown in Table 2.

Patients’ perspective is a vital component to understanding treatment decisions patients make and better informing future decisions. The treatment landscape for patients with AML or MDS particularly lacks this frame of reference. Previous research has examined the definition of patients with AML or MDS who are ineligible for intensive chemotherapy, available treatment options, end-of-life care, and health care resource utilization [26-29]. However, no published research, to the best of our knowledge, has investigated patient preferences for treatment of this population or has explored the use of social media via disease-specific forums to research the experiences of these patients and their caregivers. This study was a novel design that combined the use of unstructured social media text and natural language processing with an established text analysis technique, QDA, to attempt to better characterize and understand the perceptions and priorities of patients with AML or MDS who are not eligible for intensive chemotherapy.

Although the focus of this study was patients with AML who are not eligible for intensive chemotherapy, some findings agreed with those of previously published studies that surveyed the more general AML population. The burden of physical impacts of the disease and/or treatment, including fatigue and weakness, infections, and fever, was frequently reported in this study and in published works that examined similar impacts [30,31]. This study also showed how important psychological and emotional impacts (eg, anxiety, depression, fear, and confusion) were to patients. These impacts were often caused by a lack of knowledge of the condition and its treatment or anticipated test results; these aspects were also reflected in previous studies [31].

Social media also provide caregivers and family members with a wider platform to discuss how they are impacted. This study demonstrated the impact on these individuals who were burdened by additional responsibilities to manage the patient’s affairs on top of their usual roles and duties. In turn, these burdens sometimes had a negative effect on the emotional and psychological well-being of the caregivers.

The study indicated that patients felt there was a lack of treatment options and did not feel that age alone should be a reason to not receive a particular treatment. There was also a clear theme around a general lack of information and knowledge regarding the condition and treatment options, resulting in the feeling that the patients and caregivers were unable to make fully informed decisions.

This study provided a unique perspective on the experiences and views of patients dealing with end-of-life situations, a situation not easily explored with a survey-based approach. Patients discussed goals that were broader and more complex than simply extending survival. Patients and caregivers highly valued time at home and with family or desired to reach certain life milestones or family events. These priorities might be considered less by clinicians and researchers, but their impact on mental well-being, survival, and HRQoL can be equally as important as qualitative markers and should be considered during the treatment decision-making process.

A recent survey of patients with AML concluded that treatment options that slow disease progression and do not compromise HRQoL are important in this population [15]. In keeping with this, our targeted analysis found that 44% of patients discussing treatment choices mentioned impact on HRQoL as a key influence in opting for one therapy over another. Surprisingly, aspects related to home and family (42%) were equally influential in choosing a treatment. Patients continued to emphasize the importance of being at home, reaching specific life events, and spending time with family. Interestingly, the negative effect of treatments on HRQoL was most cited in reasons for refusing certain treatments, with patients often stating a preference for higher HRQoL over quantity of life for this condition. Influences under the home and family theme were cited as the reasons for making or accepting certain treatment choices.

The use of social media offers a unique and complementary approach to capture the perspectives of patients and their caregivers. Published studies that aimed to characterize AML patients’ perspectives using a patient survey approach varied in sample size from 32 to 82 patients [30-32]. However, these studies looked at a range of AML patient groups and ages and covered signs, symptoms, impacts, positive and negative concepts, receiving information, and making treatment decisions. Buckley et al [30] and LeBlanc et al [32] recruited patients from a single hospital setting, which could have resulted in a lack of more general representation.

This social media study is the first, to our knowledge, to look specifically at the experience of AML and MDS patients ineligible for intensive chemotherapy, and the approach was uniquely adaptable to include more patients as needed. A pool of 347 patients qualified for this study—saturation was reached at 85 for the QDA of objectives other than treatment decisions, and 45 patients were used for a targeted analysis of treatment decision making. The sample sizes in this study were comparable with the largest of the published studies for AML and far exceeded guidelines for sample sizes discussed in the literature [31,33].

The use of social media has the added benefit of providing a more geographically diverse sample and reducing timelines for collecting information. Just as important, if not more so, this approach has the potential to collect unprompted responses by patients and caregivers. Many patients may be too weak or frail to participate in patient surveys and in health-related social media forums, whereas caregivers often participate on the patient’s behalf. The inclusion of the experiences of caregivers provides a unique and potentially more comprehensive perspective to the approach of using social media.

A recent publication compared the use of disease-related social media, concept elicitation interviews, and group concept mapping to understand the perspectives of patients with ankylosing spondylitis. The study found that disease-related forums on social media uncovered the most concepts/perceptions, used the least amount of resources, and allowed access to a larger amount of material [34]. Our study also found the use of social media to be an efficient and cost-effective way to collect data of this type. In addition, this study presented a novel insight into the experiences and opinions of a patient group that had not been previously studied. Considering the rarity of the condition and the difficulties in recruiting patients with AML who are not eligible for intensive chemotherapy, this study drew from a relatively large sample size for a more representative look at this patient group. Another strength of this study was that the themes and concepts were driven by the discussions between the patients and caregivers themselves—they were not guided by any moderator, medical professional, or researcher and are therefore less likely to be impacted by information bias than traditional questionnaire or interview studies. Instead, they offer a more accurate and novel reflection of issues that are most important to patients and caregivers.

Social media, however, are a fairly unregulated source of information and should be used with caution. Compared with traditional patient interview studies, the biases present in disease-related social media data are currently not well understood; therefore, it is difficult to characterize the types of bias that could impact the results of this study. However, it should be noted that no study is free of all bias.

Social media provide a window into patients’ and caregivers’ perspectives using a previously untapped approach. The ability to get a more representative and current depiction of what matters to patients with AML or MDS and their caregivers can have an important influence on the treatment landscape and the resulting decision-making process.

This work highlighted the themes that are important to patients with AML or MDS and their caregivers, which have previously been given little consideration but, nonetheless, have a clear influence on their experiences and treatment decisions. The following themes were considered as very important: improved QoL over extension of life, being at home, spending time with family, and living to reach certain milestones. This study also confirmed what had been previously reported in the more general AML population—physical impacts on QoL such as AEs are important to patients and play a role in treatment decisions. Our approach, however, offered the unique ability to examine a wider range of unsolicited topics and provided a more in-depth look at these discussions.

Physicians and other medical staff should consider these softer preferences and goals and use this knowledge to engage in more comprehensive, informative discussions with patients and caregivers, especially given the lack of treatment options for this patient population. This knowledge also emphasizes the need to close the information gap for patients and caregivers regarding treatment options and can help clinicians make better-informed recommendations for their patients. Furthermore, the findings of this study can be used to inform further research in this population and allow for patients’ preferences to be taken into account during drug development and for regulatory decisions.