ACTION

I interviewed:

● Intake Coordinators who receive forms and work with OCR

● Case Managers who communicate with doctors and patients, to ensure all documents and information is complete and accurate

● Reimbursement Specialists who communicate with insurance companies and pharmacies

● Patient Navigators who communicate with patients to keep them informed throughout the process, answer questions, and basically hold patient's hand

● Call Center agents who service both patients and employees; they answer patient questions and guide patients through filling out forms online, and answer employee questions and/or triage employee calls

Employee Interviews revealed:

Patients call the call center for:

● 50% of calls cover standard, simple responses, and triaging calls to respective areas.
● 50% calls are more complex, request human high touch, and last for up to 45 minutes.

● Agents handle 80-90 cases per day.

Employees call the call center for:
To be connected to another department, to ask for someone’s contact info, or to ask how some internal system work.

● 70% of doctors’ offices submit forms via fax, resulting in forms being submitted incomplete
- Doctor’s forms can have wrong diagnosis code or missing dates, etc.
- Doctors’ offices also submit majority of patient forms, that patients fill out while in doctor’s office. Patients often miss signature. Legally, process cannot start without patient’s signature, or verbal consent.

● Issue with forms' content and structure, be it paper or online, was repetition of the same question multiple times on one form, which prompted some patients to abandon the form.

● Patient & Prescriber forms, for the same patient and same medical issue, sometimes arrive to our client's Co-Pay assistance division, at different times. Reconciliation is automated albeit not 100% successful. So, some manual matching takes place, leaving room for error and slowing the process down.

● Patient Support Staff is overwhelmed and does not have sufficient time to dedicate to explain patients their options.

● Lack of transparency in timeline and status of the investigation, to the patient. They often don’t know where their case is in the process, and what will happen when.

RESULT

My strategy was to identify areas where conversational AI, meaning IVR, Bot and Virtual Assistant can help:
 

1. Verbal consent - Help patient provide verbal consent, that their information can be used, to help the them pay less for their treatment.

2. Status update - Patient can retrieve basic status update on their case.

3. Triaging calls - Employee calls can be triaged; to be connected to another department, to ask for someone’s contact info.

4. Request for materials - Sharps containers, marketing materials, and travel kits calls, can be fully automated.

5. FAQ - Clinical questions (medication side effects, how to use your inhaler, etc).

6. Help patients fill out forms - Virtual assistant can do that, in the comfort of patient's home.

Benefits (of CUI, conversational UI) to the business:

1. Bot can help avoid incomplete forms and missing signature, and help avoid inaccurate spelling.

2. If humans have more time to help patients fill out their forms, and address impediments to enrollment, so the end result is increased enrollment.

3. If Bot can handle mundane and easy tasks, humans can handle more complex cases. And more of them, in shorter amount of time. So patients spend less time waiting in cue = less abandonment = increased enrollment.

4. Bot can answer some of the inbound calls, agents can spend more time spend on outbound calls = more patients reached = Outbound calls are usually following up with patients.

Accessible, inclusive and multi-modal

1. COVID reduced physical interaction between HCPs and patients, the amount, length and frequency of it. Patients are not in doctor's offices, clinics or hospitals (unless they have to be). As a result, interest and demand for DAs exploded in the healthcare space.
    

2. Sapient research showed that not all patients have a computer. Such as:
   • elderly
   • those who do not need it, who work with their hands (construction, cook, home healthcare aid, plumber, carpenter, etc.)
   • those who do not know how to use a mouse or trackpad (elderly, mostly).
   • those who cannot afford a computer
   
   But everyone has a phone.
    

3. Not everyone has an email, or knows how to use it (elderly). So you need to find your patients where they are at. Just like you did with Secure Messaging. You skipped computers, websites, and emails and went straight to text.
    

4. Patients do not like interacting with GUIs (graphic user interfaces)
   • One patient said that although she owns a laptop, she simply does not use it as much, cause it’s a “hassle”.
   • Another patient said that she does not like using websites on her phone.
   
   Lessons learned: find people where they are at: on their devices, using text and voice.

User stories & sample dialogs

Here are mockups for 3 potential Bot scenarios:
● Verbal consent
● Status update
● Triaging calls

Verbal consent:
Frequent scenario: Case is in progress. The only thing that’s missing is Patient Consent. The client currently obtains consent by:

● Patient signs the form online.

● Patient signs paper form, takes picture and submits the picture.

● Via phone, with a live agent – but that’s only temporary measure, during COVID.

Client would like to enable Verbal Consent, and record it. Here’s how that can be achieved:

● Patient receives a text message, along the lines:

Logins are a problem In regards to the text message above, common obstacle to login-based websites, is that user does not recall their username and/or password. The route with the Bot circumvents that obstacle. As a reference point, as a means of identification, financial services and banks’ Bots today, create a voice imprint / identity, for each user, while the user is talking, for the first time. That’s the most up-to-date identification type in voice. For the purposes of the interaction shown below, we are going to use a more modest, traditional method, such as name, DOB and zip code.

Here's how Verbal Consent can be provided via chatbot:

Status update:
Frequent scenario: Case is in progress. Patient receives regular Status Update text messages, via the Secure Messaging system that’s already in place.

Here are 2 common status updates from everyday life:

Amazon

Amazon updates that user receives via email or text. Please note the Amazon is being used as an example of concept, NOT content.

Status update #1: We received your request, and your package is being prepared.
Visit our website for more information.
User then opens the timeline in the app, if they have it installed on their device, or they go to Amazon’s mobile website, and use their login info, to view the complete timeline.

Status update #2: Your package is en route. Visit our website for more information.

Status update #3: Your package has been delivered.

You can go online, or in your app, to check on details. In most cases, the brief text update would suffice.

Pharmacy

Either you receive automated text, or you call your pharmacy, to check on the status of your prescription.

You get 3 statuses:

Status update #1: You are out of refills. Would you like us to call your doctor?

Status update #2: Your prescription is being prepared. We’ll call you when it’s ready

Status update #3: Your prescription is ready for pickup.

If you still want to talk a pharmacist, you have that option. In most cases, the brief automated update would suffice.

Your patient can receive the same type of brief status update via text:

Or, if they call, Status Update can be provided via chatbot:

Triaging calls:

This is how triaging calls could look like:

Introduction to our persona

Upper left: section names and issues with Account.

Upper middle: issues with Payments.

General Troubleshooting scenarios.

Main part: my ideas for improved Customer Service, and channels of how I think it should be delivered.

My additional ideas, outside of what client asked for. But what I thought would be of value. Set #1.

My additional ideas, set #2.

How would it look like implemented: