PaperEMR

Over 10m health interactions every day in Africa are documented - on paper. Without digital data from these interactions it’s impossible to improve healthcare. For example, 80% of the over 2m deaths annually due to medical errors are preventable (https://www.who.int/news-room/...); data on health interactions would allow this.

Most health data is captured on paper, which is available everywhere, familiar to users, and costs very little. Generating digital data from paper records is however very difficult and expensive, involving painstaking, error-prone manual transcription. The most common alternative - entering data directly into smartphones, tablets or computers - is however also very expensive and difficult to implement, requiring huge investments in infrastructure and digital change management. More importantly, direct digital data entry systems in health (eg. electronic medical records systems) have a huge, negative impact on doctor-patient interactions (https://www.newyorker.com/maga...).

Our current understanding of global health challenges and solutions come from focused research projects that generate data around specific hypotheses. There is very little routine data generated in global health. Routine data on the 10m+ daily interactions can solve diverse healthcare problems, from inappropriate antibiotic use to last-mile supply chains for lifesaving medical products.

We are solving the challenge of sustainably generating routine health with currently available infrastructure and human resources. Importantly, we are generating these data with a clear focus on upstream challenges that can be addressed (eg. access to primary healthcare or essential medicines).

PaperEMR combines two simple steps that can be performed by anyone:

a) Document data on paper, and

b) Digitise the data by taking a picture

Both steps have been simplified to allow documentation and data capture in any setting, without requiring bespoke devices or training.

Documentation on paper: Health worker (community health worker, pharmacist, nurse or doctor) or patient documents health interactions (eg. consultation, dispensing) on paper forms with structured data entry fields. Structured forms can be pre-printed, or printed on demand using rubber stamps. Forms are developed with clients and users, involving human centred design approaches, generating a shared understanding of what data is important to capture, in which format, and with what level of error tolerance. Pictorial aides are often used to support health workers with low literacy.

Taking a picture of the paper record: Once the interaction is documented, the form is digitised by taking a picture using any device with a camera and browser (including feature phones). Typically health workers use their personal phones to digitise the health record. The application is packaged into a web-page, which is accessed by the health worker (does not involve downloading an app). The "progressive web app" works offline after the first time it is accessed. Data entered on paper is extracted on the device using computer vision algorithms, even when offline. 

Reviewing captured data: Data extracted from the paper form is displayed to the user within 1-2 seconds (even when offline). The interface allows the user to easily verify captured data and correct errors if any. Data are captured from both "bubble sheet" fields (computer vision-based optical mark recognition algorithms are >99% accurate), and handwriting fields (machine learning-based handwriting recognition algorithms are ~90% accurate). Extraction algorithms have all been developed and perfected in-house. Error correction is simple and intuitive, allowing for near-100% accuracy of digitisation.

Sharing data to the cloud: Since the data is extracted on the device, the size of data transferred to the cloud is very small. Typical forms generate ~50bits of data. These data can be synced to cloud servers either using mobile internet or SMS (if offline).

Data visualisation: Data are updated on cloud-based dashboards within minutes of taking a picture, and are available to users (eg. health workers), beneficiaries (eg. patients) and health system managers.

These steps are all seen in this short video where PaperEMR supports community health workers in remote, marginalised nomadic populations: 

PaperEMR is a solution for high-quality data capture at low cost. There are diverse uses and users of this solution currently, impacting lives in different ways:

a) Primary healthcare for non-communicable diseases (NCDs): NCDs kill 41 million people each year, equivalent to 71% of all deaths globally. We have published how PaperEMR improves documentation of NCDs in primary care: https://www.ncbi.nlm.nih.gov/p...

b) Antibiotic use in primary healthcare: Almost 5m deaths each year are associated with antibiotic resistance. We have published how PaperEMR contributes to measuring and improving antibiotic use among primary healthcare providers: https://www.ncbi.nlm.nih.gov/p...

c) Last-mile access to life-saving health products: Millions of people in rural parts of LMICs lack access to basic medicines like oral rehydration salts or anti-malarials. PaperEMR is used to support last-mile supply chains of essential medicines using community health workers: https://insupplyhealth.com/ins...

d) Postpartum haemorrhage (PPH) is the most common cause of maternal mortality, responsible for up to 35% of maternal deaths. Lack of availability of safe blood for transfusion is a major root cause. PaperEMR is currently used in an NIH-funded study across three counties in Kenya to understand demand for blood.

e) Every facility-based delivery in Migori County in Kenya is documented using PaperEMR. We are currently preparing manuscripts based on the insights generated from these data on >6,000 births. Importantly, the problem of poor data on maternity services was articulated by the county government, who directly procured the solution from Health-E-Net.

Across these different projects, the lack of data is a common challenge. Access to routine data at low cost can solve multiple problems at different levels of health systems.

Once data is captured, PaperEMR can be (and is) integrated with other tools (eg. telemedicine technology or imaging tools) to deliver an integrated solution. Health-E-Net provides some of these solutions.

Health-E-Net was founded by Dr. Pratap Kumar, who is a medical doctor with a PhD in neuroinformatics. He is also a senior lecturer at the Strathmore University Business School in Nairobi where he teaches innovation in health, and has extensive collaborations with healthcare organisations across East Africa. He spent his early career as a medical doctor in India, and volunteered with community-based organisations involved in physical and mental rehabilitation. This experience was the basis for founding Health-E-Net, to be able to better link patients in underserved communities with a global skill pool of health workers. His work in global market access with big pharma provides a deep understanding of the health system challenges involved.

Dr. Meghan Kumar, the co-founder, did her PhD on the economics of community health systems and is an Asst. Prof. in Health Economics at the London School of Hygiene and Topical Medicine. She has worked extensively with community health workers, coordinating implementation research activities across Africa and Asia. She is spearheading the use of participatory methods in health economics. Together the founders have extensive clinical, implementation and academic experience in global health.

The team includes experienced IT developers and managers who have created novel, robust and scalable technologies with a completely Nairobi-based team.

PaperEMR is trying to change how patients, providers and managers view digital health. We recently published a paper titled "Digital ≠ Paperless" to highlight that paper can be effectively integrated into digital systems, and how a hybrid paper-digital system has various advantages over digital-only systems: https://www.ncbi.nlm.nih.gov/p...

This level of change requires support on multiple fronts. However the primary reason to apply is for the visibility that the Challenge can provide, and the access to numerous potential clients that can unlock the use of PaperEMR at scale.

Paper is integral to global health systems, and replacing paper by digital-only systems is hugely challenging and potentially harmful.

So far, capturing data from paper has mainly focused on handwriting recognition, which is not accurate enough for practical use in healthcare settings. People often use the stereotypical "doctors' handwriting" on prescriptions to highlight the difficulty of the problem.

PaperEMR is a novel combination of computer vision and machine learning with knowledge of health data capture that makes data capture from paper a practical reality.

We have demonstrated that PaperEMR can be used to capture data on diverse health interactions, and be used (and loved) by nomadic community health workers and doctors alike.

With few improvements to reduce the cost and effort of client acquisition, we expect that PaperEMR can be the default solution for documentation and data capture in every situation where direct digital data entry is not practical.

We aim to directly impact 5m people every year within 5 years - the beneficiaries of health transactions where data is captured using PaperEMR. 

This will be achieved mainly through commercial growth strategy, focusing on organisations working in partnership with pharmacies and agrovets (the first point of contact for human and animal health problems). These include pharmaceutical manufacturers and distributors, who lack the insights into sales of their products at the >500,000 small pharmacies and agrovets across Africa. PaperEMR is ideally placed to deliver data on demand for healthcare products (and services), and to enable clients to engage deeply with a distributed, predominantly rural population, and unlock services like remote prescriptions and consultations and rural health insurance.

The steps we intend to take over the next year include

1. Focussed business development with pharmaceutical manufacturers and distributors

2. Technology development to reduce client acquisition costs (eg. automated generation of scannable forms through a web interface)

3. Technology development to create APIs to integrate PaperEMR with commonly used sales and CRM software

As a business-to-business organisation, our impact will be measured through the impact of our clients.

Our main metrics are:

a) number of clients

b) number of users

c) number of beneficiaries

Improving healthcare requires sustainable ways to measure health and healthcare service delivery.

Current approaches to measure health and healthcare service delivery are predominantly focussed tools that "primarily inform funders rather than practical performance improvements."

The problem can be distilled down to the impracticality of direct digital data entry at scale (i.e. a techonology adoption problem).

Paper is a highly versatile, low-cost, and familiar interface for documenting almost any interaction. However capturing data from paper requires solving on two fronts:

a) the design of paper documentation tools to support reliable, high-quality data entry with minimal training and minimal additional effort

b) the design of simple data capture tools that can be used on personal devices, with minimal training and minimal effort

If these two aspects are solved, it is conceivable that existing documentation workflows can be modified in small ways to incorporate a "digitisable form" which can be digitised by currently available human resources, with devices currently available to them, and using workflows (i.e. taking a picture on a mobile device) that are familiar to them.

The additional benefit is that users will be integrated into the digital system by instantly visualising the data they capture, and the insights that those data reveal.

The log-frame version:

a) Activities: 

1. Designing and implementing scannable versions of paper forms currently used for documentation 

2. Modifying the PaperEMR application to digitise data from the scannable forms, and visualise data on web-based dashboards

3. Providing users with instructions to fill and scan forms and visualise data

4. Use of scannable forms to document health services

5. Regular alerts and reports on the data

b) Outputs

1. Digital data on services delivered (eg. medicines dispensed, diagnoses, referrals, etc.)

2. Data dashboards providing analyses and visualisations

c) Outcomes

1. Improvements on pre-defined success criteria (eg. antibiotic prescribing, frequency of stockouts, etc.)

2. A digitally aware and involved workforce

1. A progressive web app built using javascript

2. Bespoke computer vision and machine learning algorithms

3. Cloud servers, SMS

4. Paper

The primary healthcare providers themselves collect the data on their services. They do this as part of routine documentation of their services (a current aspect of their jobs). PaperEMR makes this documentation simpler and more meaningful to healthcare providers.

Occasionally frontline workers have been incentivised to document and collect data using PaperEMR. These incentives can often be part of business practices (eg. sales commissions) or acceptable cost of outcomes (eg. cost of identifying a new case of TB and linking them to care).

Users are mostly excited to be part of the digital ecosystem with minimal effort.

Health-E-Net's staff are all based in Nairobi, Kenya, and we recruit locally. The team is 50% female (including the developer team), and we have members from diverse faiths and socio-cultural backgrounds. Our lead developer works actively with Kenya's deaf community, supporting them with digital tools (eg. videos in Kenyan sign language).