Designing a smart patient health info collection form for public health
DURATION
9 months
TYPE
User Research | Service Design | UX/UI
CLIENT
Wadhwani AI & Ministry of Health & Family Welfare, India
*This is a summary of the full case study due to NDA.
** Mobile view has limited information, use desktop.
Problem
Statement
Our team was responsible for building an AI differential diagnosis recommendation model. This model would recommend various diagnosis to the consulting general practitioner during Tele-consultation on eSanjeevani app. However, the dataset that we received for model training was low standard prompting us to explore how we can collect structured patient health information using the eSanjeevani tele consultation app.
The Beginning
The Assumption
Patient Data collection is a time taking process that effects the doctor patient interaction.
Research
eSanjeevani platform has 2 types of interaction
-
Patient can reach the Doctor directly using eSanjeevani app
-
Rural Patient can reach the doctor through Community Health officer
​We began the research with observing how the eSanjeevani tele consultation happens in different levels of public healthcare in Urban spaces.
Form Prototype
We started with creating a form prototype based on the patient health data we had from eSanjeevani's tele consultation database.
We tried multiple open source tools but eventually ended up building it in house due to factors like User experience, costing and data privacy.
Form prototype
1. Personification of Health worker: Sanjeevani
Branding to build trust in the form collecting personal data. Layman language instead of medical language for 123 symptoms captured from eSanjeevani's completed consultations.
We built a 44,000 word rule based form to collect patient health information alongside program managers to test the user experience and understand the user needs from user and doctor's point of view.
2. Modelling real world conversations into form
Replication of an offline conversation between patient and medical practitioner which begins with collecting all symptoms and then digging into specifics following Patient centric model
3. Multiple choice over long form text entry
We provided options on the basis of available data to avoid typing as we assumed it was not a preferred activity due to the time taken. Typing was kept as a secondary option.
Process 1.0
The primary research process evolved on the basis of access to the facilities with program support and permission from stakeholders and involved urban spaces, Tertiary(Big) hospitals & Online interviews.
We focussed on both types of eSanjeevani services: Direct Patient to doctor and Patient-CHO-Doctor interactions.
01
User feedback using
Google store reviews
Secondary research
02
Tele-consultation laws & research review
Secondary research
03
Observing eSanjeevani
in State run hospital
Contextual Interviews
Field research
04
Observing Tele-consultation in Tertiary hospital
Contextual Interviews
Field research
05
Online interviews of
Service providers on eSanjeevani
Primary research
06
Form prototype & testing
Prototype Development
Usability Testing
Process
Outcome 1.0
From the field visit, interviews & conversations with various stakeholders, we learned some key aspects about the public healthcare system in India.
1. Overview of ecosystem
The scale and complexity of the public health ecosystem make eSanjeevani adoption difficult. AI use cases must be sustainable and not burden health workers. The Ayushman Bharat policy's focus on primary health care offers a potential area for intervention for AI.
2. Imbalance in HR & infra creates burden
There is a shortage of specialists in India, most are in urban areas, employed by the government at the tertiary level or in the private sector. All patients often congregate at the tertiary level, highlighting the need to address issues at primary level to reduce load at tertiary level.
3. User types for eSanjeevani
There were 3 key user types identified: Patient, Providers and Admin. Our initial understanding focussed on patients who would use the tele consultation model directly seeking specialists that we learned were not key users.
4. Existing knowledge to identify all user actions
Since the entire eSanjeevani app was being developed by an external vendor, we conducted various meetings and a visioning exercise to align with the various information category that the form would take care of.
5. Longer than 5 mins made users drop off
We evaluated 6 users with varying digital literacy levels for the form's content, interaction, and flow. The major finding was the need to reduce the time spent collecting data, as the need for capturing detailed information effects the user experience.
Process 2.0
We established that eSanjeevani platform is not most active in Patient to Doctor consultations. We also established that Patient health data is critical in tele consultation, however, we needed to understand the the end to end platform as well as the Patient-CHO-Doctor interaction.
Field operation changes
01
Need for unbiased responses without fear
We aligned ourself with a partner for this field visit and did not use Ministry in our introductions. We found that the conversations were more fluent & open.
02
Preplanned selection of states and facilities
We identified complete hub & spoke models in two states which are opposite on public health performance spectrum in the states.
03
Trained team members to support the research
Team members received orientation, took notes, and discussed them in debriefing sessions to conclude learnings.
01
Foundational research in 2 states to understand end to end service
Contextual Interviews
Card sorting
02
Revision of the form based on Usability testing in UPHC
Prototype Development
03
Usability testing of the revised form with Health worker (CHO)
Usability Testing
Process
Outcome 2.0
1. Hub & spoke system for eSanjeevani
Key users in this ecosystem are Community Health Officers at Health & Wellness Centers and General Practitioners at PHC/DH, who face interaction challenges with the eSanjeevani product. Patients are the beneficiaries in the OPD model.
2. Defined end to end service of the form
By addressing each touchpoint of the eSanjeevani platform, we were able to identify what expectations the CHO & the doctor have from features. We created 3 separate modules: Registration, Patient Health collection and Doctor-Patient bridge.
3. Personification didn't work for Health workers
The CHO's preferred conversations to be to the point and wanted a straight forward tone reducing reading/response time. The form was treated like a tool by the CHO while for users interacting directly with the form, the tone mattered.
4. Reading manual text entry is a key aspect
We learned that users describe symptoms in various ways, making a comprehensive list impossible. Since users preferred typing symptoms, they would often type and add, we needed to train a model to learn from hybrid textual entries.
Final Product
Since the form was part of eSanjeevani app, we needed to redesign it to meet their UX/UI style and make it responsive as we found that all the community health workers used their mobile for tele consultation.
Key features
1. Multi select responses
The CHO is shown most common symptoms based on data from that particular geography. All of the questions come with pre given options to reduce manual entry.
2. Type & add in Indian languages
We provided a "+Add" option for the user to type and add symptom details. We found that users preferred that instead of searching through the options and selecting the available one.
3. Doctor style to the point tone
We modified the tone of the product for the Community Health worker to be straight forward.
We removed additional statements that were meant to build a sense of comfort (relevant for patient)
What we
have achieved
so far
1. Structured Patient Health data collection
Since, the form is deployed it has helped support 122 Million online consultations with a more robust data collection than before.
2. Mapping of eSanjeevani platform
We have been able to map the different stakeholders, actors of the platform as well as the potential and limitations. The learnings have helped identify the next direction for the product improvement.
Next Steps
Voice over text
Our form is currently based on selection of pre existing options and Type to search and add symptoms (to reduce the need for typing). However, as screen interaction poses a challenge in user experience with the patient, we are exploring Voice capabilities of AI to record and transcribe the conversations.