We described the implementation of a digital health intervention based on mobile text messaging to promote access to information about warfarin therapy. The text messages were validated with experts and patients and reached a CVC above 0.80 in the first round of assessment. We found that this intervention was well-accepted by primary care patients, most of them reported that it helped them to change their lifestyle and eating habits and affirmed that their knowledge on anticoagulation improved after receiving the SMS.
It is well-known that warfarin therapy has a narrow therapeutic index, wide variability, dose-response, and potential drug-drug and drug-food interactions. Therefore, the management of this treatment is challenging in the clinical practice. Periodic monitoring with INR tests is required to guide dose adjustments and reduce the risk of adverse events (5). Educating patients regarding anticoagulation therapy is an essential step to success, to support their engagement to the treatment and improving self-care (7, 8). In this intervention, the fact that 86% of the patients reported greater knowledge of warfarin therapy following the messages, as well as the increasing proportion of patients who achieved ≥ 75% correct answers in the OAK test, demonstrate the impact of such strategies.
One important principle in this intervention was the elaboration of the text messages considering the translation of scientific evidence to a simpler and accessible language using only a limited number of characters. We chose the Delphi method to validate the messages which were scrutinized by patients on warfarin therapy and experts in the field (21–23). This method allowed a multidisciplinary and collective construction of our bank of messages, which was very well-received by the evaluators and the CVC reached in the first round of evaluation did not require another round. In the early stages of the study development, we opted to include not only experts, but also the patients' perception about the content of the messages. Patients' views are fundamental in all stages of the process, as they are the target of this program. The development of the content is often underreported text-messaging interventions (32). However, as Ybarra et. al. noted, behavior changes depend mostly on what is given (content) rather than how it is given (delivery format) (33). Therefore, involving patients in the process to act as evaluators was valuable to better tailor it to the intended audience, as well as to understand how to communicate more effectively with them.
After the first round of the intervention (pilot), two patients lost access to their mobile phone numbers, and one gave up because he could not read the messages. In Brazil, it is common to change phone numbers, because of the affordability of prepaid phone plans. However, in the second phase, the proportion of patients who received the messages and completed the intervention was significantly lower. It is important to notice that the second phase took place during the first wave of the COVID-19 pandemic, and we had a higher proportion of patients who died or stopped to use warfarin compared to the first phase. Periodic surveillance to ensure that patients are receiving text messages is recommended by the World Health Organization, and has shown in the present study to be essential for the success of this type of intervention (30).
After the pilot, in the post-intervention satisfaction evaluation, lifestyle changes were reported by around 86% of the patients. When analyzing these changes in-depth, almost 80% of patients reported changes in their food consumption habits. In the open-ended question inquiring them about these changes, most patients reported changes in the consumption of green foods, such as green leaves and other vegetables. Historically, healthcare professionals used to advise patients taking warfarin to completely remove green foods from their diet. This was a recommendation because these foods are sources of vitamin K, which can enhance the effect of warfarin, potentially leading to a higher incidence of hemorrhagic complications. However, the current body of evidence demonstrates that maintaining a consistent intake of vitamin K is more important than the quantity (34–37). Even though this is the current recommendation, many patients and healthcare professionals are still unaware of this information. We believe that this is one of the reasons why most patients cited changes in their green foods’ consumption.
Regarding their overall knowledge about warfarin therapy, 86.2% of patients reported greater knowledge of their therapy in the satisfaction questionnaire. In the OAK test, there was a trend of increase in the proportion of patients who achieved ≥ 75% of correct answers (p = 0.0703). Previous studies have shown that an improvement in knowledge is associated with increased time within the therapeutic range and lower incidence of complications (7, 8, 38). A very high proportion of the patients were already adherent to the treatment, when assessed by the MAT test. Therefore, it was not possible to assess the impact on adherence.
On the questionnaire's open question some patients reported difficulties reading the messages for two different reasons: (1) Reduced visual acuity - which affected their readability on the mobile phone screen; (2) Difficulties to comprehend the messages. To overcome these difficulties, they needed the help of others (such as caregivers or family members) who read the SMS content for them. In Brazil, although illiteracy has reduced over the last decades, it still is a social issue. A national survey conducted in 2016 showed that 11.8 million people could not read or write, and illiteracy increased with age and lower family income (39). On the top of that, there are the individuals who are not illiterate, but still have low health literacy (40). In our intervention, most participants were more than 60 years-old, lived with less than four minimum wages, and studied less than 8 years. This social context can explain why patients reported reading difficulties, and this is an important aspect to take it into consideration when planning further interventions for this target audience.
As part of the study protocol, focus groups would have been the final stage of the first phase of this project (41). However, as this approach requires face-to-face interaction between researchers and participants, it was suspended due to the COVID-19 pandemic. Safety for both our researchers and participants (most of which were at-risk groups for severe COVID-19 disease) could not be guaranteed in this situation. Furthermore, the patients had limited access to videoconferencing tools that would enable us to adapt these interviews to a digital approach.
The scale-up of the SMS project supplied the demands of the large number of patients who said they would like to keep receiving messages. We also included the ones who could not receive the SMS before because they were recruited after the pilot had started. This round was designed to test how the project would be in “real-life”, without constant monitoring by a research team. Due to the COVID-19 pandemic, we chose to incorporate new messages using the same know-how to give patients information about the novel disease. It shows that this type of intervention can be easily adjusted to incorporate new information using the same low-cost technology. In our experience, the lack of constant monitoring caused two pitfalls: (1) The automated system presented persistent failures to some of the patients, so they never received the messages; (2) Social problems were much more prevalent than they were in the first round, with reports of nine deaths and five patients who stopped to take warfarin. Although our intention was to test a real-life implementation, these issues could be easily identified if we had implemented a surveillance system as we did in the first round. This is an important lesson for the future, and is in accordance with the recommendation by the guideline of World Health Organization for digital interventions (30).
We also assessed the long-term impact the SMS had in the patients’ perception. Although it is known that the perceived positive effect of the intervention decreases with time, there is a lack of evidence analyzing the duration of these benefits (42, 43). In our analysis, 9 (31%) of the interviewees could not recall any of the message's content after seven months of the end of the intervention. This assessment points out that, to be sustainable and scalable real-life interventions should have in mind the need for periodic rounds of messages, especially when targeting long-term chronic illness. Also in this phase, another interesting insight is that some patients referred to a feeling of “being taken care of”. This reflects that this type of intervention can have an emotional impact on patients that cannot be easily assessed by tests. Still in the qualitative analysis, one patient gave feedback that he would prefer the content to be presented to him by voice, because of his difficulty in reading. This suggestion is very relevant for the next steps of the project to make our intervention more accessible to patients who have limited literacy or visual impairments. Although voice-based interventions have not been extensively explored so far, some studies showed this strategy could improve health outcomes (44, 45). Indeed, more studies testing this form of delivering information needs to be done.
This study of a digital intervention using SMS to primary care patients taking warfarin was very well accepted by patients and had an impact on their knowledge of warfarin therapy. To the best of our knowledge, this is the first study to describe a SMS intervention for this target audience. We also showed the lack of monitoring may compromise an intervention, and a surveillance strategy should be ensured even in scale-up projects. Finally, we believe that a strategy using voice-based messages would be an interesting way to overcome the barriers in the implementation of such strategies for elderly patients, who may have difficulty in reading small font-size SMS, and for those who have limited literacy, which is a very common situation in LMICs.
This study has limitations. In this intervention, our sample size was small, and further studies are needed to generalize our findings, as well as to assess the impact of the intervention on clinical outcomes. The satisfaction questionnaire is a self-perception, which can lead to bias. In addition, the variables we measured for assessing impact are not clinical, and we cannot claim that a better understanding of warfarin therapy according to the OAK test would lead to a better clinical management.