AWS for Industries

Streamline Medical Contact Center Interactions with AI

Across the board, healthcare providers are struggling to serve patients within the funds available to them. Remote healthcare, along with AI powered automation of doctors’ non-value adding activities (such as repetitive conversations), will allow doctors to have more time for patients and enable healthcare providers to reduce their costs.

Why automate Medical Contact Centers?

According to How automation of patient data workflows is transforming primary care, healthcare across the world is overburdened with patients, even before the impact of the resent pandemic. As populations grow, and get older, healthcare systems are put under increasing strain.

COVID19 created human-to-human contact issues. Everything ‘contactless’ was in huge demand, from contactless payments to remote healthcare.

For this reason, health organizations such as the National Health Services (NHS) Medical Contact Center, are offering Medical Contact Centers for patients to talk to doctors over a phone or other virtual channels. However, this has also created its own kind of burden on medical professionals.

When it comes to how medical professionals spend their working hours it needs to be where it adds the most value–caring for patients. Now, instead, medical professionals spend a significant portion of their time in non-value adding activities, filling out forms, wrestling with data and spreadsheets, and collecting initial information about patients. These are continually repetitive activities such as asking about the patient’s conditions, allergies or underlying diseases.

Anything that can make these processes more efficient will reduce the burden on medically qualified staff and generate significant savings.

Patient data workflows can be automated in a number of ways, such as online forms replacing paper ones, alongside a wide range of automatic and data-driven workflow processes. In practice, during automated calls, AI-powered, natural language processing, self-service diagnostic voice/chat bots could ensure patients are channeled towards the right kind of care straight away.

Not everyone is going to need to see a doctor or nurse, whether virtually or in-person. A self-diagnostic system, especially if this was connected to a wearable device, would make it far easier and quicker for patients to get the treatment they need more quickly.

What is Medical Contact Centers automation?

Automation enhancements of the Medical Contact Center can streamline the patient/doctor interaction and offer improved efficiencies.

The voice/chat bot (medical bot) integrated into the Medical Contact Center, can automate the commonly asked questions of the patient/doctor conversation. The medical bot asks questions and is able to collect the initial info about patients (including patient’s conditions, allergies or underlying diseases) before the patient even begins interaction with an actual doctor. This automation of a common and repetitive patient/doctor conversation frees doctors’ time so they can prioritize their medical resources better.

The logic of the conversation, the hierarchy and the content of the medical bot questions are editable and can be changed subject to the customer or country requirements. Biometric identification of the patients allows the medical bot to greet the patient personally and reduce time for identification of the patient.

AI powered analysis of the medical bot/patient conversation can discover hidden information about patients, their emotions and main concerns. Through a query of the medical bot’s conversation, a basis for diagnosis can be provided to the doctor to follow-up on. Thus, doctors can potentially resolve the patients’ query faster, thereby reducing the handling time and improving the resolution. This creates more quality time with the doctor as well as more patients being seen by the doctor.

In addition, the patient/medical bot and the patient/doctor conversations are automatically captured removing the need for non-value-added activities (such as filling out forms and recording the patients’ information). Instead, this automation allows doctors to spend more time in caring for their patients.

AI empowered near real-time monitoring of conversations offers the doctor an immediate patients’ sentiment analysis and their key issues. AI would also help to redact the Personally Identifiable Information (PII) when storing the data for analytical purposes.

This newly de-identified captured data can be integrated with other relevant data, in a data lake, with AI providing predictions and trends for a patients’ treatments. An analytics layer on top of the data lake would allow for research of pandemics and other population health phenomena, as well as feedback to improve future medical bot conversations.

How does it work?

Let’s walk through the potential case of Jane Doe, a 62-year-old female presenting with a fever, chills and a cough. She decides to call her doctor, who has implemented a new automated Medical Contact Center.

Automated Medical Bot Conversation
When Jane Doe calls her doctor, she first encounters a medical bot. By using Amazon Connect Voice ID (a feature of Amazon Connect) which provides near real-time voice authentication, Jane  is able to identify herself and connect the medical bot to her stored records. The medical bot greets Jane by her name and proceeds to ask her about her symptoms from a set of predetermined questions. After recording Jane’s current condition, the medical bot asks Jane about her medical history, chronic diseases, allergies and medications, if this was not already in her record.

All of Jane’s responses/information is then collated within the electronic health record (EHR) system for the doctor to review before starting a conversation with Jane. The information is also captured by Amazon Kinesis Data Firehose for further processing down the line.

Consultation with the Doctor
When the doctor starts the conversation with Jane Doe, he/she is aware of Jane’s current condition including fever and sore throat (possibly indicative of an infection) as well as her medical history (in particular diabetes as well as high blood pressure). During her one-on-one online consultation with the doctor, Jane expresses her current concerns. The recorded consultation is analyzed by Contact Lens for Amazon Connect pointing out Jane’s main concerns of shortness of breath and accelerated heart-beat.

From the medical bot conversation, the doctor is already aware of her increasing difficulties in breathing during the last few days and that it is causing her to panic from not being able to breath. The doctor is also aware from the medical bot conversation on medical history that Jane has been hospitalized for severe pneumonia two years ago. The doctor reverify this information during their conversation.

In light of this information, the doctor initiates a home visit to check Jane’s potential need for respiratory assistance due to a possible infection as well as further diagnostic and therapeutic measures.

The initial information supplied to the doctor by the medical bot, allowed for an efficient and expedient one-on-one consultation with Jane as well as the fast and effective resolution of her enquiry. Jane was also very relieved to be able to acquire fast medical assistance in her time of need.

The interaction between the medical bot and Jane, as well as her consultation with the doctor, is able to be recorded into the EHR system and by Amazon Kinesis Data Firehose. A summary report of the consultation can then be forwarded to the consequent treating medical practitioner. This means the doctor doesn’t need to waste his/her time in filling out forms and compiling reports.

AI Powered Data Analysis
Jane’s anonymized (without any PII) stored data information is added, along with other callers anonymized information, into the data lake. There it can be integrated with other complementary data (such as demographics) and provide a basis for analytics models to research common diseases phenomena. By training AI in these historical data patterns, scientists can discover novel ways of treating diseases.

The next step is to incorporate Amazon HealthLake (HealthLake) into this solution. HealthLake is a HIPAA-eligible service offering healthcare and life sciences companies a complete view of an individual or patient population health data for query and analytics at scale. By bringing together the patient data from various resources (including the doctors’ notes, lab test results and more) it can provide the trends of the patient’s health. This comprehensive information is provided in a straightforward and understandable way to the doctor, at the time when the patient is calling, to improve the quality and efficiency of the doctor’s diagnose.

In Jane’s case, HealthLake will store not just the recorded conversations. It will also store the clinical notes, lab reports, X-Rays, CTs, MRIs, EKG, medications and even her telemedicine information from any wearable device (such as a continuous heart monitor or blood pressure device).

HealthLake uses specialized machine learning models (like natural language processing) to tag, index, and structure health data. Integrated medical natural language processing (NLP) has been trained to understand and extract meaningful information from unstructured healthcare data. HealthLake can identify trends and make predictions. As per Figure 1, HealthLake presents a chronological order of medical events so doctors can look at trends (like disease progression over time) giving doctors visibility to improve care and make it more efficient.

Figure 1 HealthLake - Extracted meaning from health data

Figure 1 HealthLake – Extracted meaning from health data

When Jane has the next consultation with the doctor, HealthLake will provide trends and predictions of Jane’s lung and heart disease progression, medications that Jane was taking along with her test results. The doctor will have the full visibility of Jane’s circumstances helping him/her to make fast and informed next steps.

Technical Architecture

Figure 2 Sample architecture of Medical Contact Center with AI powered automation

Figure 2 Sample architecture of Medical Contact Center with AI powered automation

An essential component of any automated Medical Contact Center is Amazon Connect, a cloud contact center providing integration with AWS services and other applications. It does call routing subject to predefined conditions.

Amazon Connect Voice ID provides machine learning powered, near real-time caller authentication that makes voice interactions in Medical Contact Center more secure and efficient. It replaces painful and time-consuming manual authentication process. Based on the patient identification, the medical bot already knows the patient’s personal information (such as age, address and others) so the patient is personally greeted by the medical bot.

The flow of conversation between the Patient and the Medical Contact Center is designed using Amazon Connect Flows. The medical bot is powered by Amazon Lex. Amazon Lex is an AWS conversational interface offering automation of chat and voice conversations. Amazon Lex provides the advanced deep learning functionalities of automatic speech-to-text (ASR) for converting speech-to-text and, natural language understanding (NLU) to recognize the intent from the text.

As Amazon Connect is natively integrated with Amazon Lex, all the attributes from the medical bot that live in Amazon Lex, are available in the Amazon Connect Flows. The medical bot utilizes a set of predefined utterances and intents to lead the conversation and is fulfilled by AWS Lambda functions.

Amazon Polly, is a Text-to-Speech (TTS) service, which is also natively integrated with Amazon Connect, and is used to ask the questions to the patients when they call.

Amazon DynamoDB stores the structure of the conversations and the question logic inputs. The logic can be altered, without changing the underlying code, allowing for adjustment for specific customer or country requirement. The doctor can fine tune the medical bot’s questions to elicit better answers from patients.

Contact Lens for Amazon Connect uses AI to analyze the conversation from the medical bot, and the consultation with the doctor, to extract the sentiment of the patient and important key words relevant to diseases. At the same time, it redacts the patient sensitive PII information, prior to sending the analyzed conversation to the analytics system.

The automated medical bot conversation and the doctor’s consultation are streamed to an Amazon Simple Storage Services (Amazon S3) bucket by using Amazon Kinesis Data Streams. Amazon S3 provides the data lake capabilities allowing for conversations to be stored and integrated with other relevant data such as research or demographics, to enhance the data set for further analysis. This is different from HealthLake’s AI capabilities to provide the comprehensive patient information in a straightforward and understandable way to the doctor.

AWS Glue is a fully featured extract, transform and load (ETL) service which crawls Amazon S3 data, discovers the metadata, and creates a data catalogue.

Amazon Athena is already integrated with AWS Glue, and by using SQL queries can query data in Amazon S3.

The analytics layer uses Amazon QuickSight as its analysis dashboard. Amazon QuickSight utilizes the Amazon Glue Data Catalogue and provides a number of predefined analytics models that are applied against the conversations and other complementary data (such as demographics) to research common diseases phenomena. By training AI in these historical data patterns, scientists can discover different ways of treating diseases. Figure 3 shows an example of a QuickSight report created from the medical bot conversation.

Figure 3 Sample report from the automated bot conversations

Figure 3 Sample report from the automated bot conversations


  1. Integration with electronic health record (EHR) systems was not addressed in this blog. There can be number of different vendors of EHR system but standard interfaces, such as Fast Healthcare Interoperability Resources (FHIR), can facilitate integration. HealthLake supports interoperable standards, like the FHIR standard format, for integration with healthcare systems including EHR. We are planning to address this in a follow-up blog.
  2. At this stage the data lake of the Medical Contact Center is kept separate from the HealthLake. The Medical Contact Center data lake is focused on storing and analyzing conversations enriched with other relevant data. However, in the next blog, we are planning to consolidate the two and bring the conversations/consultations into the HealthLake as one of the data sources. This will empower this solution to use all existing AI capabilities of HealthLake for the Medical Contact Center.


In times when, across the board, healthcare providers are struggling to serve ever growing numbers of patients, within the funds available to them, we can use advancements in AI to automate non-value-added doctors’ activities. This can improve efficiency and reduce the cost of doctors’ consultations. Automation of the repeatable questions improves the quality of care―allowing the doctors to focus on patients instead of capturing administrative and other repeatable information. Collected information, along with other integrated data sources, create the basis for diseases research.

In addition, Amazon HealthLake employs specialized machine learning models to bring together the patient data from various resources to present a chronological order of medical events. Doctors can look at trends and predictions like disease progression over time. This gives doctors visibility to improve care and make it more efficient.

Amazon is well positioned to deliver end-to-end health care solutions. In addition to Contact Centers and a wide range of AI/machine learning (ML) capabilities, Amazon offers a wide range of IoT Services for telemedicine as well as specific Telehealth solutions.

To find out more about how AWS or our partners can help you please contact your Representative.

For Further Reading

Nada Reinprecht

Nada Reinprecht

Nada Reinprecht is an AWS Senior Partner Solution Architect passionate about innovative, and advancing technology that serves people. Before AWS she worked with Accenture, IBM and others within Industry designing and delivering solutions for customers in Australia, US, Europe and UK. Her rich experience in technology now extends into Healthcare, Pharma, Power and Utilities, Manufacturing and Sustainability. Nada is the Accenture AWS Business Group EMEA technical lead for Connected Customer Experience, using Amazon Connect. Nada loves bush walking, yoga and running.