DHARA: Digital @Home, AI- enabled, Real time, Appropriate

Neglected Tropical Diseases (NTDs) are a group of 20 diseases that affect one sixth of the world’s population, an estimated 1.7 billion rural and urban poor in low-and-middle-income countries with limited access to basic water, sanitation and hygiene (WASH) and health care resources. Chronic NTDs like leprosy and lymphatic filariasis (LF) cause significant lifelong morbidity and disability. National programs however, focus on new case detection and treatment while people with chronic NTDs require long-term care for which government primary health facilities are their only recourse. Meanwhile, government health systems are perennially resource constrained, and hence consistently overlook NTD management even in highly endemic areas. Part of the reason is because data depicting the magnitude of the problem, which could influence service provision, simply does not exist. Additionally, grassroots healthcare workers lack the capacity to manage NTD care. The World Health Organization (WHO) recognizes that people needing NTD care largely remain out of reach of national health care delivery systems. Since 2005, WHO has moved away from a disease-centered approach to an integrated, intervention-based approach to counter these diseases, providing marginalized communities with better access to health care.(Source:https://www.who.int/news-room/commentaries/detail/neglected-tropical-diseases-a-proxy-for-equitable-development-and-shared-prosperity). Therefore, in high NTD endemic areas, measuring improvement in primary health care performance for these diseases serves is a proxy measure for universal coverage, equitable access, and an effective health care delivery system at the primary level.

DHARA is a  user-friendly grassroots-level digital platform that collects reliable data and builds the skills of frontline healthcare workers to provide home-based care. It improves primary health care performance and health outcomes through data-guided precise interventions, optimum resource allocation, and effective preventive and promotive home-based care. DHARA, which is Sanskrit for “continuous flow”, stands for Digital @ Home, Artificial Intelligence enabled, Real time, Appropriate interventions. It is a modular, android smartphone-based, artificial intelligence (AI) enabled platform developed to empower trained local grassroots level healthcare workers to screen, document, understand, grade, and provide customized interventions in real time at the patients’ homes for NTDs and WASH (a key determinant of many NTDs). DHARA delivers all these with Icons, Images, Intelligence, and Interventions in an iterative modular model.

1. The user interface of DHARA software has no language or text. It has “icons” that are understood universally and developed iteratively with multiple stakeholder/ user inputs.
2. Screening is done by trained health care workers asking questions based on existing validated content, guided by Icons and taking “images” that are automatically tagged  with date, time, and location for reliable documentation. 
3. With input data and images DHARA has a built-in logic or “intelligence” to understand, grade the condition, and recommend appropriate “interventions” that are also icons in motion or animations.

DHARA’s icon based software is created to be usable universally in all NTD endemic countries. Currently, trained local village women are using  smartphones with the DHARA application for service across two countries (India and Nepal) to screen half a million people over three years. Analysis shows that universal screening of every household is possible. This was confirmed by overlaying the metadata collected on DHARA on Google maps of habitations. Training through DHARA to empower local grassroots level health care workers is five times faster than paper-based survey training; it was possible for people with limited literacy to use the application and even non smartphone users were able to learn and use the application. ‘Visit to value generation / creation’, which is the time taken from the screening to the resultant intervention that meets the needs of the household / person affected, is done in real time and is 10 times faster through icons and images, compared to the analog paper-based process. Data with metadata is passively recorded at every step throughout the service leading to highly reliable service at the ‘point of need’, that is continuously measured and can be iteratively improved.  

DHARA addresses many barriers leading to poor primary health care performance: lack of local skills, lack of reliable documentation, and lack of continuous access. It allows a local person living in an affected rural remote community to be empowered and upskilled to reliably screen, document, understand the need, provide possible continuous home-based integrated interventions for both WASH and NTDs, facilitate hospital referral when necessary, and provide follow up home care. Our vision is to (1) extend the service through DHARA universally by empowering local grassroots level health care workers; (2) enhance the imaging, intelligence and implementation through Artificial Intelligence / Machine Learning (AI/ML), granular continuous analytics and interactive alerts and; (3) expand and optimize the integration with the State health system’s hospital based service.

Primary health care through government primary healthcare facilities are the only recourse for people living in remote and resource constrained rural areas. Primary health care therefore has several stakeholders. At a high level they are: the patients in their communities, the local grassroots level trained health care workers, the paramedical supervisors / first level medical service providers, primary care medical doctors, and the system administrators. DHARA has a positive impact across ALL stakeholders. Currently we are working in rural Nepal and India. 

Icon, Image, Intelligence- led, and 10-times faster DHARA helps the local grassroots level trained health care workers to interact more with patients and provide service in real time, leading to empowerment. Patients benefit by integrated management at home through DHARA, for not just the disease (NTDs) but also the determinants of disease (WASH), and can avoid the multi-dimensional access issues and disjointed and disintegrated current model of management. Chronic NTDs like leprosy and lymphatic filariasis (LF) cause significant lifelong morbidity and disability. Their management is done continuously, based on need, by a local member of their community. The data with metadata collected continuously and analyzed at granular level and layered, helps to create highly reliable dashboards that dynamically visualize  population coverage and service performance. DHARA’s interactive map view dashboards can help system administrators to improve and optimize primary health care. Paramedics and doctors can benefit from optimizing their time for appropriate specialized medical care rather than basic primary care activities and checking data for consistency and reliability.

This project is a collaboration between American Leprosy Missions (ALM),  the oldest and largest non-profit organization in the United States of America working to cure and care for people with leprosy and related diseases globally,  and the Hi Rapid Lab (HRL), the first and only startup company created by India’s largest public health organization, the Public Health Foundation of India, to fast-track the design and deployment of novel easy-to-use language-neutral icon, image, intelligence based primary health care digital services at home. The focus of HRL is research-led design, development and commercial service deployment of intelligent, innovative primary health,  mobility and social care services including areas of socio-economic improvement and livelihood generation.  Both organizations along with  project partners LEPRA Society in India, and Nepal Leprosy Trust and Nepal Leprosy Fellowship in Nepal, are working collaboratively on Neglected Tropical Diseases (NTDs) and Water, Sanitation and Hygiene (WASH) management since 2018, managing large population-based projects in remote rural areas in India and Nepal. DHARA was developed based on the learning experience and felt needs across stakeholders. Feedback from the trained female grassroots workers using the application and project field staff in both countries has been critical in iterative design and for improving the workflow from the field perspective. The core team members of HRL startup company are the faculty and students of India’s first MSc and PhD program in Health Informatics. Project New DHARA-WN (WASH & NTDs) with AI enabling  will help the team and the students of the MSc and PhD program develop expertise in Artificial Intelligence methods and integrate the learning in the program curriculum for wider and longer knowledge dissemination.

DHARA needs peer critical review at a global level along with further addition of best practices that need to be included in the workflow and intervention content. Through MIT Solve’s global network, we want to learn from and meet with potential partners globally for learning and implementation. DHARA with its icon user interface (UI) is universally implementable in any population. Through MIT Solve we hope to accelerate the adoption of our concept exponentially to hopefully 10-100 times more than what we can do alone, to reach the billion people affected by NTDs in the shortest possible time.

Also, given our resourcing requirements to meet our scaling goals, MIT Solve's connections will be invaluable to ensuring access to a broader pool of peer-to-peer networking, technology mentorship, impact measurement validation and support, media visibility and exposure and grant support.

The solution to improve primary health care performance for NTDs cannot be a single innovative service for a single disease; it has to address a multiplicity of determinants and it has to address the challenges of the social ecosystem in which it is delivered. DHARA has multiple innovations to address the multidimensional needs and challenges. The elegance of DHARA is the model in which seemingly simple innovations have been brought together for a transformative impact.

DHARA has multiple innovations with Icons, Images, Intelligence, Interventions in an iterative modular model to: (1) simplify screening; (2) amplify data quality and (3) exemplify service management. Simplification of screening is innovatively achieved by: icon based, language neutral, user interface and automated (in subsequent versions) image based signs and symptoms screening. Amplification of data collection is achieved by GPS tagging and passive metadata tracking. Exemplified service is achieved by intelligent algorithms that analyze data and provide real time household risk profile, disease grade and customized animation led intervention list. These innovations collectively enable DHARA to deliver highly reliable service at the point of need that is continuously measured and can be iteratively improved. In comparison most proposed solutions are unidimensional; while these may serve well for generating numbers as a pilot or a proof of concept, they fail to sustain in an ecosystem plagued by multi-dimensional issues.

DHARA is being implemented across several locations in India and Nepal to manage over half a million people:

• In Nirmal District, Telangana state, India (with due permissions from the state government),11 government frontline health workers are screening for socioeconomics, demographics, quality of household water, sanitation, and hygiene, nutrition, disability, and NTDs and providing at home customized interventions, and facilitating hospital referrals  to about 25,000 population using the DHARA application on their smartphones. 

• In Samastipur District, Bihar state, India,  130 local village women identified from 130 villages have been trained in DHARA software to provide similar services for a much larger population of 350,000. 

• In two Provinces of Nepal, 31 local village women have been trained to screen around 150,000 people and provide home based interventions using the DHARA application.

Data Collection: the  screening through the DHARA application  is sourcing about 100-150 MB of data with 40-60 images per household, totaling about 10-15 terabytes of data with 5 million images across 5 verticals per 3 month screening cycle. 10-15 service cycles are planned in the next 3 years. Total images collected will be over 50 million. Standard operating protocols (SoPs) have been developed so images taken by the DHARA application users  are captured in a prescribed method to limit inconsistencies.  The table below shows the type and approximate total number of images collected in 15 service cycles and the significance of automation.

Vertical

Image Description

Specific Significance

Number of Images

Socio Demographics

Front of house images capturing the type of house: five grades – ranging from well-built, solid construction, brick houses  to houses made of mud, bamboo, straw, thatch.

Grading  the social and economic household condition automatically with images

700,000

WASH

Water stagnation in front of house, water storage & treatment;  toilet, bath room, soap, animal presence

Identifying poor WASH living conditions automatically and providing real time health customized health promotion information 

28,000,000

Nutrition

Person – front and profile

Identifying approximate height and weight automatically and calculating BMI (body mass index) for nutritional status and nutrition advice

7,000,000

Disease (NTDs) – Leprosy and Lymphatic Filariasis

Images of limbs with different grades of filarial lymphedema;  leprosy eye, hand and foot disabilities 

Automated identifying of the signs and symptoms from images leading to automated diagnosis and disease/disability grading.

15,000,000

Our vision for the next 12 months is to:

(1) Extend the service by empowering about 500 local grassroots level health care workers to screen a million people; (2) fully buildout our current prototype software by “enhancing” the imaging, intelligence, implementation through original AI/ML algorithms based on over 50 million original images being collected at present, (3) reduce the patient service time by 75%, and the patient interaction time by 50%, (4) enhance the dashboard interactive feature, automated analytics and alerts by over 500%, and  (5) expand and optimize the integration with the larger health system’s hospital based service.

Our vision for the next 60 months or 5 years is to:

(1) extend the service by empowering about 5,000 local grassroots level health care workers across Asia and Africa to screen over 10 million people; (2) continuously enhance the imaging, intelligence and implementation through dynamic self-learning AI/ML, that provides granular continuous analytics and interactive alerts and predicts necessary responses and; (3) expand and optimize the integration with health systems across several countries.

Broadly, the activities are grouped under three work packages (WP):

“EXTEND” WP1: Improving DHARA service usability for field health care workers

Activity 1. Training local frontline workers

Activity 2. App based screening, documentation and disease management

Activity 3. Feedback from frontline workers for improving the workflow from the field perspective

“ENHANCE” WP2: Improving AI, patient / disease management outcomes at home.

Activity 4. Automated feature extraction from images

Activity 5. Appropriateness of workflow logic and suggested interventions

Activity 6. Patient engagement processes

Activity 7. Measuring patient adherence to app based guidance

Activity 8. Measuring improvement in disease condition and ADL (Activities of Daily Living)

“EXPAND” WP3: Improving the responsiveness and optimization of the health care system for local needs at the primary level.

Activity 9. Measuring the benefits to the health system

Activity 10. Development of a feedback-loop system for continuous improvement of services based on local needs

Indicators will be measured through an experimental approach, where data collected from present analog or paper-based service will be considered as the gold standard. A comparison study will be done between analog model and digital DHARA model for several parameters like:

1. Time taken for training process of healthcare workers

2. Time taken for the workflow – screening, documentation and disease management

3. Health care worker requirements for improving the workflow from the field perspective

4. The patient engagement process

5. Patient adherence to app based guidance

6. improvement in disease condition and activities of daily living (ADL)

7. Benefits to the health system

   1. Improvement in service coverage

   2. Comprehensiveness of data collection, aggregation, identification of clusters

   3. Improving continuity of data collection – creating longitudinal health profile

   4. Optimizing the medicine procurement for managing NTDs and allocating personnel for facility-based management

   5. Integration with Ayushman Bharat Digital Health Mission electronic healthcare database system (in India)

8. Development of a feedback loop system for continuous improvement of  services based on local needs

As these processes are multidimensional, the broad research focus will be disaggregated into smaller research questions to facilitate in-depth granular evaluation. Broadly, observational and experimental methodological approaches are considered with mixed methods (quantitative and qualitative) tools for the evaluation. For most research questions more than one tool is used, as triangulation is considered a more robust approach for measuring multidimensional indicators. Quantitative research methods like questionnaire evaluation, workflow videography timestamps, metadata metric evaluation and qualitative research methods like focus group discussion, in-depth interviews, secondary literature-based narrative reviews, content analysis, ethnographic research, case study research and a workflow journal will be used appropriately for the study.

THE PROBLEM: NTDs (disease) affect the poorest populations with least access to WASH, living in close contact with infectious vectors, domestic animals and livestock (determinants), in locations where health services are below par, thus continuing the cycle of disease and poverty. NTDs should be managed continuously at the point of need along with their causal determinants for achievable eradication.

STRATEGIES:

1. Explore – List the issues of NTDs and WASH, prioritize, and identify promising solutions for effective screening and efficient at-home interventions that address all the issues in the given ecosystem.

2. Empower – Create an integrated mobile digital platform with solutions, as a prototype that can be made available at the point of need as a pilot, and train local people to use and address the issues head on.

3. Extend – Learn from the pilot, iteratively modify, fully build the prototype to product, and train more people to create a critical mass of opinion and evidence that can lead the way for a new system

4. Enhance – Add more features and increase the value proposition of the product with more customized location-specific intelligence and implementation through dynamic self-learning AI/ML

5. Expand – Optimize and monetize the service and integrate with the larger health systems across several countries

KEY OUTPUTS:

Strategy 1. Explore 

Output 1.1. Documented microlevel workflow and issues through deep observational studies and focus group discussions (FGDs) with primary health care stakeholders.

Output 1.2. Solutions to address the identified  needs, and technology readiness levels (TRL) for implementation/further development through design thinking-led technology exploration .

Output 1.3. Multidimensional fit-for-purpose platform in the user ecosystem through iterative FGDs with mock prototypes and experimental user experience.

Strategy 2. Empower 

Output 2.1.  Prioritized fit-for-purpose solutions incorporated in the digital mobile platform.

Output 2.2. Village level local grassroots workers identified and trained, empowered as health care with digital solutions  to screen, document, understand, grade, and provide customized interventions in real time at home, universally for WASH and NTDs.

Output 2.3. Piloted solution’s usability by trained workers and comprehensiveness of the solution at the scale of large critical mass level population.

Strategy 3. Extend 

Output 3.1. Pilot study and data analysis quantitatively and qualitatively measures the progress and need for improvement.

Output 3.2. Fully built prototype with automated feature AI/ML, is used by trained local grassroots level health care workers, and continuous study of the service and data.

Output 3.3. Service monetization models identified through regular stakeholder meetings to expand the service and build consensus among stakeholders.

Strategy 4. Enhance 

Output 4.1. Addition of more analytical features that help in understanding the dynamics between disease and determinants.

Output 4.2. Predictive models that can prevent the transmission of disease and eradicate disease outbreaks.

Output 4.3. The digital platform aligned with health care data standards and disease management to add value and quality to patient life at home through AI.

Strategy 5. Expand 

Output 5.1. Mapped service ecosystem and documented benefits of workflow optimization.

Output 5.2. Monetization models identified through costing, cost benefit analysis at different levels .

ASSUMPTIONS

• The general progress of digitization continues to improve the quality of services, and regulations are supportive

• The public and involved stakeholder opinion on digitization led services are favorable, and at mass scale the unit cost benefit of digital services outweigh analog services.

RISKS AND MITIGATION

• Technical Risks – Any innovative and transformative product has a certain technical risks; DHARA with multiple innovations has inherent engineering challenges. Mitigation: Iterative development, early customer engagement and field trials for real-world pilot service at large scale.

• Market Risks (Service acceptance) – DHARA and its methodology is unique and far ahead of what is in service today; immediate acceptance would be a challenge. Mitigation: Pilot service roll out in multiple locations, multiple ecosystems and continuous follow up by design team and user engagement. The integration of opinions of stakeholders at multiple levels is key to DHARA’s ongoing field acceptance; and iterative changes will make DHARA even more usable and acceptable.

• Funding Risks – Obtaining required funding from experienced long-term supporters/ investors. Mitigation: Continue incremental development and add features in a modular prioritized model.

OUTCOMES:

The intended outcome are  to (1) extend the service through DHARA universally by empowering local grassroots level health care workers to manage the needs of over a billion people affected by or at-risk of NTDs, including WASH; (2) comprehensive coverage and automation through enhanced imaging, intelligence and implementation through AI/ML, with granular continuous analytics and interactive alerts and; (3) expand and optimize the integration with the larger health system’s hospital based service.

IMPACT: Potential elimination of NTDs and poor WASH globally including the remotest and poorest communities by empowering local grassroots level health care workers, or the next best alternative, to facilitate reasonable increased quality of life for people with NTDs and improve WASH.

NB: Strategies 1-Explore & 2-Empower are already underway successfully. MIT SOLVE’s support is required for Strategy 3-Extend, Strategy 4- Enhance, and Strategy 5-Expand

DHARA has multiple innovations with Icons, Images, Intelligence, Interventions in an iterative modular model. Low health literacy is a global issue. More than 800 studies have documented that most available health information is at levels that greatly exceed users’ abilities of comprehension. Very long-ago Confucius said ‘A picture is worth a thousand words’. Several researchers in health care have found that pictures can help non literate, low literate and even highly literate people comprehend health care information better. To address the issue of health communication, we identified and listed the key words representing disease signs and symptoms in the validated disease management questionnaires, and developed graphic digital illustrations, also known as ICONS for: asking, looking, listening and feeling; and another set of ICONS for recording positive, negative and other responses. The icons were iteratively revised and refined extensively with several rounds of inputs from a variety of health care stakeholders across all levels. At the Public Health Foundation of India, we have unprecedented access to health system stakeholders. The ICONS were then sequenced as a workflow and converted as a software application (APPS) for a smartphone. DHARA has a very unique ICON based User Interface (UI) which can now be used globally as the software is language-neutral and does not have a specific user language. After disease screening,  the first step of disease prevention or disease management is health promotion, which is usually educating people on what to do and what not to do. On the same model of creating icons, we also created motion icons or ANIMATIONS in a similar modular template.  The animations have a standardized format: (1) Step 1 - showing the setting; (2) Step 2 – showing the products and activity in animation; (3) Step 3 – supporting icon overlay for frequency of activity.

Automated intelligent imaging is done by Artificial Narrow Intelligence (ANI). It is defined as a specific type of artificial intelligence which usually outperforms humans in some very narrowly defined tasks. ANI needs a large amount of high quality data to yield accurate results. This data is currently being collected. ANI deep learning methods will be used for three objectives: (1) automated classification, grading of known types of images; and (2) identifying the outline of images to quantify the size of the feature when appropriate. Methods like convoluted neural networks (CNN)  with gradient boosting will be used to classify and grade images of known types. Initial work has been done to develop models to identify the type of house automatically. Similar method automation models will be developed for auto identification or grading features in images. Transfer Learning or Network Fine Tuning will then be applied to pretrained models to customize for the required number of classes.

In the villages where DHARA is implemented, women who are interested to learn and conduct digital mobile based services are nominated by their village leaders and women’s groups. A basic literacy and digital literacy assessment is done to understand the baseline literacy status, although in reality this is not a criteria for selection. The selected women are trained to be Community Resource Persons (CRPs) in NTDs and WASH  for their village. They are trained in a  stepwise manner:

1. First they are taught the meaning of the icons, tested for understanding and in the next step they are given a new smartphone loaded with a DHARA application. 

2. They are taught how to use the smartphone, the features – especially the camera feature – and then how to open and use the DHARA application. 

3. In the third stage, they are given hands-on training through case scenarios. The CRPs are guided in small groups by the project field staff in-person and remotely by the Hi Rapid Lab team. 

Each CRP is assigned about 500 households and does a daily screening of about 5-10 households. The CRPs are currently paid by the pilot project; in the future, the grassroots worker using DHARA could be paid by the service provider. As an empowered community member delivering service to their own people in need, the CRPs are sufficiently motivated. The data is shared with the health program staff and used for advocacy with the concerned departments.

At the forefront of ALM and HRL’s approach to serving marginalized people affected by leprosy, and other neglected tropical diseases, is inclusion. Over 20% of ALM’s  board are persons affected by leprosy. They are deeply committed to our mission, attend events, join staff weekly for meetings, and travel to project locations as possible. “Beneficiaries” are included in the design of tools and technology, as demonstrated by this project as well as other adaptive technology we are working on. ALM includes and is in regular contact with leprosy affected advocacy groups, like MORHAN in Brazil. 

ALM’s staff is diverse and is dispersed across the US and staff are located in Ghana, Libera, India and the UK hailing with experience every continent except Antarctica. Many have immigrant stories to share and celebrate dual citizenship. Our CEO was raised in the Democratic Republic of Congo and most staff have lived and worked cross-culturally. While we are required by the IRS to have a “headquarters” we are truly dispersed organization that accommodates “work from home” and strives to include everyone regardless of their location and work situation. 

Our programs empower women in traditional societies and help them to advocate for government healthcare coverage for which they have a right to access. 

Hi Rapid Lab (HRL) Private Limited is a startup company formed by a team of PHFI faculty members and students. The focus of HRL is research-led design, development and commercial service deployment of intelligent, innovative primary health mobility and social care services including areas of socio-economic improvement and livelihood generation for the marginalized. In association with PHFI the expanded commitment of HRL is to build the entire ecosystem and operation models that can help scale up, sustain primary health, mobility, social care services and turn them to an everyday, and possibly everywhere and every need, service. HRL proposes to shift the primary health care model from “people coming to a facility for treatment” to “@ home primary health care service” model. Although several revisions have been made over the period of years, the concept of “people coming to a facility for treatment” has largely remained unchanged and primary health care services remain synonymous with “inaccessibility, inadequacy and inefficiency”. HRL believes a local person from the community trained and empowered with our AI led, smart mobility supported digital platform can deliver universal primary health care service "@ home" and make primary health care "accessible, adequate and efficient". Three out of the five directors of Hi Rapid Lab are women. All the local grassroots level health care workers of DHARA are women, selected from the NTD- endemic villages where DHARA is being piloted, and trained. Almost a quarter of them are first time smartphone users, and through training become adept at data collection, screening, home care management, and behavior change communication – a vital healthcare and health information resource in their villages and communities that link with the primary health care system. The aim of DHARA is to empower local women with global best practices and deliver them precisely where and when it is needed, and this is happening.

DHARA has a very unique ICON based User Interface (UI) which can be used universally across the globe as the software is language neutral and does not have a specific user language by trained local grass root level health care workers to manage NTDs and WASH locally. Service providers including government health systems in this space look for validated solutions with a robust digital system and are willing to engage them to provide service as a partnership. The service providers do not pay for innovation or development.

ALM has supported DHARA for prototype development and has also supported the pilot field testing on the ground. Learning from the pilot is being analyzed and we are planning redevelopment to make DHARA robust for service with partners. Several partners have expressed interest to use DHARA on a fee or license based model. We hope to generate donor support for development and revenue from service providers who will be using DHARA. As a start-up company HRL is actively reaching out to potential investors to fund development and expand services.

DEVELOPMENT: ALM has invested USD 97,000  to support the development and field pilot of DHARA. HRL has also received financial support from Dr.Reddy's Laboratories, an India multinational pharmaceutical company which granted the Public Health Foundation of India (PHFI) USD 400,000 to develop a home based primary health care platform. Honda Motorcycle & Scooter India Ltd granted PHFI & HRL USD 70,000 to support the development of a two wheeler platform that will complement DHARA in carrying the medical products required for home based management.

SERVICE PROVIDERS: We are in discussion with companies like:

https://www.starhealth.in/

https://www.callhealth.com/

https://www.medicallyhome.com/

They are keen to expand health care services at home and are planning to consider DHARA’s unique user interface and home services. Creating a robust DHARA platform with field validation would greatly help DHARA to survive, sustain and scale up.