Inspiration
Every year, 795,000 people in the U.S. suffer from strokes [1]. Although rehabilitation begins in the hospital, 1-2 days after the stroke, patients are discharged to independent speech, physical, and occupational rehabilitation facilities within 5-7 days [2]. Depending on the severity of the stroke, the recovery process can take weeks, months, or even years as patients struggle with a variety of cognitive, physical, emotional, and sleep-related symptoms. Recovery also tends to slow down after 6 months, and some patients may experience other setbacks, such as pneumonia, a heart attack or a second stroke [3].
Additionally, doctors are often overbooked and not spending enough time with patients. An NIH study found that the median length of visits was 15.7 minutes [4]. The University of Chicago determined that a primary care physician would need 26.7 hours per day to see an average number of patients for the recommended period of time [5].
This is a serious problem for individuals recovering from strokes, since a strong doctor-patient relationship, oversight of the rehabilitation process, and a social support network is crucial to the long-term well-being of the patient. Thus, our goal is to develop an easily accessible application to support patient adherence to recovery and rehabilitation. The application also aims to connect healthcare providers and patients outside of a medical setting, allowing doctors to deliver better quality of care during the precious visit time with patients.
[1] https://www.strokeinfo.org/stroke-facts-statistics
[2] https://www.hopkinsmedicine.org/health/conditions-and-diseases/stroke/stroke-recovery-timeline
[3] https://www.cdc.gov/stroke/treatments.htm
[4] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2254573
[5] https://news.uchicago.edu/story/primary-care-doctors-would-need-more-24-hours-day-provide-recommended-care
What it does
🎯With recoverWrite, we hope to encourage patients to monitor their health long-term, and continue to foster that relationship between the patient and their healthcare providers. We accomplish this with three goals in mind:
- Record: collect information from patients about their journey that can be shared with healthcare providers
- Reflect: give patients the ability to look back at their journey while providing analytics about their progress
- Recover: grow the bond between patients and healthcare providers as patients recover
Our key features include:
- 😷 Patient Dashboard: Provides an overview of the rehabilitation exercises for the week along with a counter to track their efforts. Also includes a a list view of their logs and a graph of their sentiment analysis over time. There is also a journaling section separated into 2 parts: 1) Patients have an easy form to fill out including checkboxes and toggles to log their progress and 2) an optional section to provide detailed feedback, which can be optionally shared with healthcare providers.
- 🩺 Healthcare Provider Dashboard: Since healthcare providers do not always have the time or resources to look through every single journal entry, the sentiment analysis graph provides a quick and easy way to view the status of the patient. Entries will be flagged based on a sentiment analysis threshold, so they are appropriately notified.
- 😀 Sentiment Analysis: To selectively the most important journal entries that demands manual assessment by a healthcare provider, we perform a simple sentiment analysis on the patient's journal entry. This evaluates how positive, neutral, or negative the sentence is through a quantitative metric.
How we built it
🦋 We built the core iOS app using Flutter for functionality and implementation and used GitHub to keep track of our edits.
🧠 For our sentiment analysis, we used Flask to implement our HuggingFace models.
Challenges we ran into
**1. Stroke rehabilitation is heavily reliant on social support
- Healthcare provider availability is limited
- Filtering and flagging journal entries for healthcare providers**
We found that stroke rehabilitation involves a social component that technology cannot replace. Thus, for 1) we pivoted our work to assist development of a strong social support network and encourage patients to embrace mindfulness. Another significant challenge we discovered when researching the problem was the limited access to healthcare providers. To overcome 2), we decided to focus our work on maintaining a strong patient-doctor connection outside of a medical setting, helping healthcare providers gather a more substantial patient history. Finally, to mitigate the necessity for doctors to sift through large qualities of patient journals (3), we found success in applying Natural Language Processing techniques to analyze sentiment and assess the patient's general emotional state in their journal entries.
Accomplishments that we're proud of
🏆 We were proud of how much we could integrate in terms of the sentiment analysis. We fine tuned our model over time after much discourse and feedback from all team members to consider what would be most helpful to patients and healthcare providers. For instance, we pivoted from analyzing the entire entry to specific sentences.
What we learned
We learned how to tackle challenges as a team and work towards a common goal under time pressure. We also learned how develop code collaboratively in parallel through Github version control. When designing the product, we also adopted a "stacked" development, where features build on top of one another, instead of building a single all-or-nothing feature.
Additionally, we also sharpened our technical skills in Flutter app development and deep learning in the domain of Natural Language Processing. Finally, we enjoyed the various workshops throughout HopHacks and networking with the sponsor company representatives!
What's next for recoverWrite
Future ideas for recoverWrite include (but are certainly not limited to):
- 💬 Speech to Text: given their complications after their stroke, a patient may not have the fine motor skills to actually type anything on their keyboard to log their progress. This would make our app much more accessible to patients who have different impairments.
- 🗃️ Backend Database: since each healthcare provider may see thousands of patients each year, it would be crucial to have an efficient way to store all of the summarized logs for each patient.
- 🔒 Security: on that note, it would then be important for data to be heavily protected using the highest standards since this patient data is confidential.
- 🏵️ Milestones: we wanted to further motivate patients to continue to use our app to log their progress. Using information from analyzing the logs, we want to provide periodic encouragement specific to the patient (ex. "you are feeling 24% more pain-free over the last month than the previous month").
Thank You!
✨ Thanks for reading this far, we hope you enjoyed reading about our project. If you have any questions, please feel free to contact us!
recoverWrite: the write road to recovery
Log in or sign up for Devpost to join the conversation.