Doxper

During disease outbreaks, rapid, anonymized and aggregated insights based on real world data is pivotal to direct the outbreak control efforts in the right direction. However, health systems that rely on paper based workflows that are not digitised at the point of care are unable to respond quickly enough without accurate, representative signals of changes happening on the ground.

Doxper digitises doctors’ case sheets without behaviour change, with an AI assisted, digital pen and paper solution. 

We have internally built capabilities on best in class medical handwriting recognition. Our proprietary method can model physician's handwriting by ingesting 100-300 prescriptions & case histories. For disease surveillance, our stand alone algorithmic accuracy today is good enough to be used without human intervention. This means, the solution can be scaled significantly without corresponding increase in human capital costs, especially in remote locations where surveillance and epidemiology data is either missing or significantly delayed.

Retrospective analysis shows that SARS-CoV-2 was circulating in a number of countries well before it was first recognized. Such problems occur in part because disease surveillance is often based on old-fashioned practices: frontline health workers noticing unusual patterns of symptoms and reporting them through analog channels.

Uneven, poor disease surveillance affects everyone, but especially the poor and vulnerable as governments are only able to respond after transmission is well established. Knee jerk reaction interventions and policies often lead to inefficient coordination and inadequate allocation of scarce resources.

One of the most powerful tools available for surveillance is to track the symptoms or provisional diagnosis from physician’s notes, which are often 90% correct. This data, if digitised, can preclude lab results and provide early, accurate signals of population level symptoms that are outside of normal thresholds for the location or the season.

Doctor’s notes for follow up visits, and longitudinal data must also supplement genomic surveillance of patients, but this is currently missing. Linking the two data sources will inform the rapid development and evolution of effective clinical protocols for a new virus/variant.

Doxper is an AI assisted and cloud hosted data capture solution that allows doctors to instantly digitise their case sheets using a digital pen and encoded paper. There is no major change in behaviour or workflow required.

Core Workflow:
1. Write with Doxper digital pen on regular, coded paper (codes are printed using a laser colour printer)
 
2. Handwritten strokes and data sync with Doxper app in real-time via Bluetooth. Option for offline memory storage.
 
3. Data is processed in the cloud with AI assisted handwriting recognition. 
 
Doxper serves as a data input bridge and can integrate with any existing EHR or disease surveillance system to plug gaps in data compliance, without compromising the variety and quality of data points that need to be captured. 
 
The template printed on the coded paper is fully customizable and can be produced in any format (A5, A4, A3, booklet etc), and with any combination of free text, structured fields or diagrams for annotation. Coded case sheets can also be printed on-site in hospitals or care settings with large patient volumes.

For the digitisation of paper based workflows, our customers and end users include physicians, nurses, receptionists, health workers and clinical researchers. 

For the analytics generated from the digitised records, our customers include Hospital administrators, Pharma companies, NGOs, Government and Clinical research organisations.
 
Doxper’s zero behaviour change digitisation technology combined with our handwriting recognition capabilities can significantly cover existing data gaps and lags in disease surveillance, especially in countries where there is poor or non-existent last-mile digitisation.
 
Today, we believe the most optimal input technology in the outpatient setting is pen and paper, especially in countries where there is high patient load and limited consultation time. While we continue on this path and expand our market leadership in India, we are also developing a suite of multi-modal input technologies to cover requirements across the user behaviour spectrum - from hybrid models (part of the data is captured on paper, and the rest, such as Rx advice on tablet/phone/app) to developing best in class AI, NLP engines for handwriting recognition. We are also innovating around unique paper surfaces that can be reused so that the consumable cost is significantly reduced, without compromising the user experience.
 
On the analytics front, we have started working with pharma companies on use cases such as prescription audits, brand launch tracking and patient journeys. This experience is helping us build capabilities for population health level surveillance and insights.

Doxper can strengthen disease surveillance, by picking up early signs of a potential pandemic, directly from authenticated physician’s notes, and in real-time. 

The solution can be easily integrated into a country’s existing disease surveillance system, to enable decision makers to act swiftly based on multiple sources of intelligence.
 
The system can also allow for effective response in a closed loop, contacting high risk patients with the latest guidance and interventions. Using physician’s notes as a source for disease surveillance also allows longitudinal data to be collated over multiple visits. Responders can create accurate patient cohorts for highly targeted interventions.

We are currently the market leader in the healthcare digitisation space in India, for outpatient settings and in hospitals, with a proven product market fit. Currently we are serving over 5000 active doctors, across independent clinics and 50 corporate hospitals. We have digitised records of more than 4 million patients, with longitudinal data across specialties. 

We believe we are well poised for further growth to serve public health use cases, in partnership with corporates, government and NGOs. Our research prototype that allows for elastic Boolean search of anonymised, aggregated data from physician’s handwritten case sheets has been well received by the pharmaceutical industry. The experience and learning from this can be applied to national level disease surveillance.

For the capture of data at the point of care, we have applied an existing technology/hardware for a new use case. It is the first clinical data capture system to use a digital pen and encoded paper, and has been the most successful in terms of scale and adoption in India. This means zero behaviour change for clinicians, nurses and health workers. Doxper is simple to train for and deploy. We have completed large, complex installations in hospitals within 45 days.

Existing input solutions such as typing are not acceptable as they increase physician burnout and affect patient experience.

Current data sources for surveillance systems, such as the manual, observational and anecdotal filling of specialised forms; even if these are online, there is often gross underreporting and delays in information being relayed back. Emerging real world data sources such as social media feeds, tweets or Google searches are not reliable, difficult to model on, or are only representative of urban centres.

Doxper's intuitive technology (>90% compliance among doctors) applied for disease surveillance would generate the most authentic, representative, and fastest source of real world data for early warning signals of a disease outbreak. 

As health systems of more than 124* countries are still mostly on pen and paper, a successful demonstration of Doxper for disease surveillance could be licenced and replicated globally in a short period of time.

*Internal estimate if a country has doctor-patient ratio of < 1:1000 as a proxy for digitisation gaps. Source W.H.O, 2015-2018.

Doxper is currently serving 5000 doctors, and over 4 million patients in the outpatient setting. Our goal in 1 year is to serve 15000 doctors and in 5 years over 100,000 doctors in India and around the world. Doxper is making a difference wherever pen touches paper in healthcare, and beyond geographical expansion, we see tremendous growth potential in new areas like public health, insurance and IPD/ICU settings.

1. Doctor Digitisation Compliance % :  Proportion of doctors actively digitising their case sheets and prescriptions using Doxper in the first 90 days and in the first 180 days.

2. Template Digitisation Compliance : For all the patients that doctors consult, what % of patients do they use Doxper for.

3. Data Transfer Frequency: What % of data points are coming in real-time vs those getting delayed due to offline use (data is automatically transferred when the Internet is available again or stable). For offline use, what is the average frequency of data transfer, in terms of days.

4. Handwritten Strokes to Text Accuracy : For the target data points and fields, what is our algorithm’s accuracy compared to human transcription? After establishing a baseline for 90 days, what is the % improvement in accuracy for the same set of doctors in the next 180 days?

5. Trending Diseases, Symptoms and Medicines: Aggregated, anonymised queries on our research elastic search prototype. Stress testing the system to check if data flows are authentic, and in line with known trends.

Full time employees: 98

Contract staff: 30

Randeep, the Chief Scientific Officer was involved in precision medicine and has several patents to his name while at Philips Research. He was leading the project in sequencing the first Indian Genome from the private sector. He has a deep healthcare domain understanding both from the usability side and also the data side. Pawan, the CTO, has 15 years experience in scalable systems architecture, design, security, UI & mobile app development. Prior to Doxper, he was at Flywheel, FirstRain, Snapstick & Visible Alpha. Shailesh, the CEO has more than a decade experience in managing large scale operations, sales and marketing teams at Schlumberger across the globe. All 3 Founders are graduates from India's premier Indian Institutes of Technology.

The company was founded in 2015, and the Founders along with the leadership team have successfully grown the business to its present size digitising over 5000 doctors and 4 million patients. In terms of hospital customers, Doxper is the largest and most successful digitisation solution in the outpatient setting.

Disease surveillance is tough. According to Klaucke et al in their 1988 publication “Guidelines for evaluating surveillance systems”, quality is measured in terms of seven dimensions: sensitivity, representativeness, timeliness, simplicity, flexibility, acceptability, and positive-negative predictive values.

In modern times, we should be blessed with more tools and data at our disposal to address the 7 factors above. However, the pervasive lack of digitisation in primary care excludes the most critical and accurate real world data: Insights from doctor’s prescriptions and case histories, and the speed at which these insights are relayed back.

Doxper has been successful at digitising outpatient records in the private sector. We are confident of our abilities to significantly contribute to disease surveillance. However, grant funding and support from larger organisations is needed to convince the State/Central governments to partner with a small, albeit fast growing startup for a pilot, and eventually a pan-India rollout. In short, pooling risk capital with the right expertise to energise the space.

We believe SOLVE can also give us a platform to make an impact beyond India in LMIC markets with similar challenges.

Financial: We are currently in our Series B round, and would like to expand scope to involve potential investors from outside India, in particular those looking to fund public health projects and business models.

Public Relations: We are currently mainly known in India, and not so much in the public health space. We are looking for 3rd party, unbiased reporting on our progress and potential, in the form of more case studies, white papers etc

Distribution: We have a proven product-market fit in India, in the private sector. We are looking to expand our client base into public health, and internationally. We recently signed a large order in Bangladesh, and are starting a pilot in Nigeria. 

1. Erstwhile Google Flu trends team: Experience/lessons in disease surveillance 

2. Google Research team in India: AI/ML expertise, to further enhance our handwriting recognition engine

3. Bill & Melinda Gates Foundation: Experience funding/executing with State and Central governments in India; Memorandum of Cooperation (MoC) with India's Ministry of Health and Family Welfare under which they provide technical, management, and program design support for key health initiatives; Strong data driven track record.

4. GSK: Disease surveillance is a form of real world data (RWD), and the quality of RWD is the highest when it comes directly from the doctors' prescriptions and case histories. GSK is one of the first pharma companies to fully embrace RWD, and their expertise could be invaluable.

5. Dr Evidence: They have worked with the top pharma companies globally in the RWD space, and have a specialised medical search engine that could be potentially integrated with Doxper.

6. Infosys: Deploying IT solutions at scale globally

Applying machine learning to improve handwriting recognition of physicians is central to our vision of a world where advancing human health is simpler and faster with pervasive digitisation. We have already made significant R&D strides in this area, and have the world’s largest training datasets of physician’s handwritten strokes. We recently partnered with a Fortune 50 company to supply them with our datasets, and are keen on more partnerships. 

Our handwriting recognition engine works as follows. Physician’s handwriting is modeled on the past historical data of the particular physician. A physician specific dictionary is curated that includes medicines, dosages, frequency, duration, diagnosis, labs & scans advised and other categories. AI is used to both reduce the data labeling effort by clustering similar looking scribbles as well as recognition of the characters and words occurring in the curated physician specific dictionary during live usage. Usually deep learning methods (CRNN) are data hungry, requiring millions of data points to converge and provide better outputs. Our proprietary method can model physician's handwriting by ingesting only 100-300 prescriptions & case histories. The initial accuracy is ~95% with a manual check by the physician herself or our internal medical transcription team.

Succeeding in this domain would substantially improve turnaround times from raw data to insights, reduce costs (less human transcription) and ease the workflow further for physicians - all of this without needing to change physician behaviour and ensuring >90% digitisation compliance.

Although TB is a notifiable disease, as per a recent report*, the number of cases notified last year in India dropped by 25% to 1.8 million, reversing tangible gains made in case detection since 2017.

Another industry estimate** pegs a loss of US$735 million every year to expired drugs because of the lack of digitisation in the Indian pharma supply chain.

 For both of the problem statements above, Doxper’s zero behaviour change technology can enable the capture of surveillance, supply chain and notifiable disease data, directly picking up crucial trendlines from the physician’s case sheet. The paucity of data is because of lack of last mile digitisation, and the primary care physicians’ notes contain the most comprehensive and valuable form of real world data to generate a variety of insights.

Doxper’s authentic case sheet can also solve for fraud, by ensuring that only genuine beneficiaries receive medicines and tests prescribed.

Sources:

* https://the-ken.com/story/the-funereal-pace-of-indias-anti-tb-drug-acquisition-efforts/#_=_

** https://the-ken.com/story/inside-indias-big-pharma-drug-distributor-team-up/