Contact Us

Contact Us

AI in Healthcare: ML-based Integration Options for Healthcare Software

The Importance of AI and ML Integration in Healthcare Software

Artificial Intelligence and Machine Learning are no longer the future of healthcare – they are the drivers of the modern-day healthcare industry. AI-integrated healthcare solutions are driving improvements in patient care and operational efficiency. As AI-integrated healthcare software becomes more widely accepted and used, it’s reshaping how healthcare services are delivered. These solutions enable the analysis of diverse data types, ranging from electronic health records to real-time patient monitoring data, leading to significant improvements in healthcare delivery.

Did you know? 23% of US healthcare executives acknowledge the improvement in clinical results through AI and ML technologies.

Integration Options for ML-based Clinical Decision Support Systems (CDSS)

Many Healthcare companies are utilizing AI to develop solutions for the healthcare industry. These systems run on data and thus require access to patient data. In addition, healthcare systems need to incorporate these breakthrough solutions into their workflow. This article delves into various strategies for seamlessly integrating these Al-based integrated healthcare solutions into Electronic Health Records (EHRs) and Healthcare Systems. Specifically, we focus on integrating the back-end systems of Clinical Decision Support Systems (CDSS). These CDSS systems predict the risk status of various medical conditions.

Typically, such systems must receive patient data via HL7 and other interfaces. They also monitor patients’ clinical data in the electronic health record (EHR) and notify healthcare providers (HCPs) in case of increased medical condition risk.

We further explore five integration options for ML-based Clinical Decision Support Systems (CDSS) into Electronic Health Records (EHRs) and healthcare systems. Each option is explained, and their pros and cons are discussed, providing a comprehensive analysis for healthcare organizations seeking AI and ML-based integrated solutions.

Option 1: HL7 2.x (V2) backend integration

HL7 V2, As a longstanding integrated healthcare option

Since the late eighties, this option has emerged as a trusted integrated healthcare services solution. HL7.org reports that 95% of US healthcare companies use HL7 V2 integrations, and it is used in over 35 countries. Hl7 V2 can still be used for data ingestion for healthcare data exchange mechanisms. 

Data exchange mechanisms and event-driven messaging in HL7 V2

Operating on socket-level communication, HL7 V2 typically exchanges data with a socket-level communication and has a pipe-delimited data format. Most HL7 data transfers use unsolicited messages in a point-to-point TCP/IP connection. These unsolicited messages are sent to a receiving system for processing. Usually, unsolicited messages are event-driven. For example, updating a patient’s demographics data triggers an ADT: A08 event, which prompts the system to send an unsolicited message to the receiving system to update patient demographics.

HL7 V2 offers versatile applications and serves both ingestion and predictive analytics needs. An HL7 V2 system can provide SCH: S12 events reporting all scheduled appointments to predict a patient’s visit chances to a scheduling-based system. In an inpatient setting, the ML system actively seeks out and consumes ADT A01 and Results R01-based messages to learn about new patient arrivals. Any artificial intelligence application for orders can consume ORM: O01. It is possible to use various events similarly in outbound messages. For instance, an ORU: R01 message can be sent actively in a risk assessment, such as “We will send an ORU: R01 message regarding your medical condition.”

Although it is widely adopted, implementing HL7 on a large scale presents challenges. Each interface implementation requires many weeks of collaboration between hospital integration staff, EHR vendors, and receiving systems integration staff. One usually needs an interface engine to convert data into HL7 V2 format and bridge the gap between health systems.

HL7 V2 operates behind the scenes discretely to the user interface. It relies entirely on the EHR’s display to communicate with healthcare providers. Despite challenges and complexities, HL7 V2 is integral to Al-based integrated healthcare solutions. 

Pros and cons of HL7 V2 integration

Pros: 
  • Worldwide adoption
  • Widely used in healthcare software development
  • EHR support available
  • No unsolicited messages
  • Once the interface is live, minimal ongoing cost
Cons:
  • Time taking process to go Live in each hospital
  • Socket-level point-to-point connections
  • Many dependencies and work required by hospital IT
  • No UI
  • There is no control over how/if the alert will show proactively
  • Not centrally located dashboard option
  • Limited number of events shown by the health industry

Option 2: Consolidator APIs integration

Challenges with HL7 V2 implementation and consolidator APIs as an alternative integrated healthcare solution

Although HL7 V2 interface integration is widely adopted worldwide, there are some difficulties in implementing the HL7 v2 interface. Providers delivering these services are limited to hospitals and entities where this has already been implemented. They charge heavily for these services, reducing the time to bring the interface up at the hospital.

Vendors offering consolidator API services

Some of these vendors in this category are listed below:

  • Redox (redoxengine.com)
  • Commure (commure.com)
  • 1upHealth (1uphealth.com) 

These providers and vendors collect the data using various integration options. When API calls are made to collect data, they are charged. Vendors like Redox can send risk alerts about medical conditions to an EHR and it’s up to the EHR to show the result. The other drawback of this option is its user interface does not allow users to interact. In simple wording, it is a backend—integration option.

Redox Network Graphical presentation: Ai based integrated healthcare systems
Source: Redox (redoxengine.com)

Pros and cons of consolidator API integration

Pros:
  • Most EHR support available
  • Fast onboarding at each hospital
  • Very little work is required by hospital IT staff
  • No dependencies
Cons:
  • No EHR embedded User interface(UI) option. The user interface can be built by utilizing this option. This process will require separate authentication and payment for each data access through API.
  • There is no control over results or how they will shown.
  • Not centrally located sepsis risk dashboard
  • High ongoing cost for live interface

Option 3: Smart on FHIR (Fast Healthcare Interoperability Resources)

FHIR, A modern healthcare interoperability standard

SMART on FHIR is an open platform built on the HL7 FHIR standard. It allows developers to construct applications that use the SMART (Substitutable Medical Apps, Reusable Technology) platform for app publishing and integration. They may use FHIR to access data using this platform. On FHIR, SMART is a user-invoked capability. 

A widget hosted within an electronic health record (EHR) or installed in a mobile or online application can use single sign-on (SSO) to authenticate with an EHR. 

FHIR’s role in facilitating data exchange between healthcare systems

HL7 FHIR V1 was released in 2014. Since then, it has become the second most popular way to integrate into healthcare settings. Most hospitals plan to switch from HL7 V2 to some FHIR for their connections. One of its greatest strengths is its user-friendliness for software developers.

FHIR provides an on-demand option instead of unsolicited messages. Users can request FHIR “Resources” from an FHIR server by calling a Restful API. FHIR resources are healthcare resources, like patients, orders, providers, interactions, etc. 

When FHIR resources are transmitted, they are in JSON or XML formats, which are straightforward for any programmer to comprehend. FHIR uses Auth and HTTPS  protocols to ensure security.

Benefits of FHIR integration, including improved interoperability and flexibility

SMART on FHIR is optimal for integrating and displaying a machine-learning solution within an existing EHR. Implementing and accessing data using FHIR is effortless and prompt, with no ongoing expenses.

The SMART on FHIR standard enables the integration of a single sign-on mechanism between an Electronic Health Record (EHR) system and a SMART application. Once users log into an electronic health record (EHR), they will not need to re-enter their username and password to access the application that shows hazards associated with medical conditions. This app can access patient data in the Host EHR using FHIR.

Challenges of FHIR implementation

SMART on FHIR may not be suitable for data ingestion because it relies on a user interface and its primary purpose for a single patient’s data set.

Option 4: CDS Hooks (Clinical Decision Support)

Clinical Decision Support, or CDS, Hooks is a FHIR-based technology created by SMART. This technology, powered by AI, assists clinicians in carrying out their daily tasks.

CDS Hooks, technology, and benefits

It permits third-party CDS systems to connect with an EHR through a “hook” mechanism integration into the workflow. Implementing CDS connections would enable physicians to utilize external clinical decision support systems to determine the significance of patient-related data.

Implementation process

The third-party Clinical Decision Support (CDS) system can provide data as “cards” to the EHR. These devices enable support systems or end-users to view EHR information, facilitating real-time patient data access. At multiple workflow stages within an EHR, these connections can be triggered.

For instance, the following scenarios can trigger a hook: – Accessing a patient record – Prescribing a drug – Placing an order.

Card types and their functions 

In response to the hook, the clinical decision support system may offer a range of response cards. The available cards can be classified into three distinct types:

Information Card: This card offers textual information to the clinician or user. 

Suggestion Card: This card offers a distinct and precise recommendation for including a clickable button that allows the user to consent. The electronic health record (EHR) is automatically filled with suggested information by clicking.

Link Card for the application: This card connects to a SMART system.

Pros and cons of CDS hooks

Pros:
  • Send an alert every time someone opens a patient record
  • A new rule from CMS says that all EHR vendors must offer help soon. Link it to the SMART on the FHIR app to view patient information
Cons:
  • Cerner and most other EHRs on the market do not provide this feature as of this writing 
  • By default, there is no way to send a warning
  • There needs to be a way for users to see a dashboard with multiple patients and their related risk factors

Option 5: Clinical collaboration & communication Platforms review

Besides mainstream Al-based integrated healthcare options, alternative ways to communicate and collaborate exist, such as CC&C platforms. These systems can be used to communicate predictions made by AI. These solutions are rarely used to ingest data in AI applications. From an integration point of view, the following are the three vendors that we assessed.

  • Vocera
  • Tiger Connect
  • Spok

Each of the above-listed platforms provides a wide range of features and options for patient communication. The primary purpose and focus was to find the integration options available to alert providers if a patient is at high risk of developing medical conditions.

Vocera

The Vocera platform has phone and web apps for providers and customers. Different sources can send new information about a patient. This information can be set up to be shown to the right HCP based on the type of information sent and set up. Vocera can handle many HL7 V2 events for data transmission. For example, when a patient leaves a hospital, an HL7 ADT A03 event is triggered and sent to Vocera. Once configured, the system will send this information as a warning to the assigned provider’s Vocera Mobile app. 

Similarly, we can trigger an ORU^R40 event (Unsolicited Alert Observation) to report a medical situation to Vocera. Vocera can then show the risk prediction to the right HCP.

Vocera platform Working Screenshot, Ai based healthcare integration
Source: vocera.com

You can find out more about how to connect to Vocera using HL7 V2 to share ML predictions of patients at the following URLs: 

  1. https://pubs.vocera.com/solution-concepts/hl7-solution/production/help/hl7- solution_hl7_help/index.html
  2. https://pubs.vocera.com/solution-concepts/hl7-solution/production/help/hl7- solution_hl7_help/index.html#solution-concepts/hl7- solution/production/topics/ hl7_solution_integrations_about.html

Spok

Spok is another vendor in this category that offers a similar set of features. Spok provides three ways to collaborate: HL7 V2, a custom API, and HL7 FHIR. The custom API option is currently on their development list, but the other options are already accessible.

The following table shows more information about the possible merging scenarios and choices:

Spok Comparison Table for Ai based Integrated healthcare option
Spok Comparison Table for Ai based Integrated healthcare option

Tiger Connect

Tiger Connect offers an advanced method for providers and customers to communicate. They offer a web and mobile app that includes features for various communication needs. The Tiger Connect mobile app functions as a customizable smart card, showing essential information to the healthcare provider (HCP) and the patient after an event has been recorded.

Tiger Connect Integration Div , A ML Based integrated healthcare solution
Source: Tiger Connect
Smart card functionality

One can set up a response button on each smart card. The individual can view the card or use the smart card to perform a pre-configured action.

Tiger Connect features an Open API for interaction. Additionally, it provides a comprehensive SDK for JavaScript (JS), Android, and iOS. The Restful API contains a wealth of information, empowering developers.

Utilization of server-sent events (SSE)

The platform uses server-sent events (SSE) to inform interested clients about real-time changes. Messages, status changes, and other events can be events. An “Event Stream” gets all the events. The Event Stream notifies all users about events as the primary method. “Present” indicates a person’s connection to the stream. Developers can add events to the events stream, make groups, add people, and send notes to groups using the SDK. 

Medical condition risk prediction in an Al-based integrated healthcare 

The Event Stream can send any ML-based medical condition risk prediction or patient warning. Groups can put users interested in risk notifications.

Tiger Connect Platform Diagram: Healthcare integration system
Source: Tiger Connect

You can find further details on the following URLs:

https://developer.tigertext.com/reference#api-endpoints

https://developer.tigertext.com/docs/tigerconnect-overview

Conclusion

In the rapidly evolving landscape of AI-based integrated healthcare, five key integration options stand out: HL7 V2.x, Consolidator APIs, SMART on FHIR, CDS Hooks, and Clinical Collaboration and Communication Solutions. Each offers unique advantages and potential drawbacks, tailored to specific system goals and design requirements.

Key Takeaways

  • AI and ML are driving improvements in patient care and operational efficiency in the healthcare industry.
  • Integration options for ML-based CDSS vary in terms of complexity, cost, and user interface.
  • HL7 V2 is a widely adopted standard, but its implementation can be time-consuming and costly.
  • Consolidator APIs offer fast onboarding but come with high ongoing costs.
  • Smart on FHIR provides an open platform for app development and integration.
  • CDS Hooks enable real-time patient data access and triggering of hooks.
  • Clinical Collaboration & Communication Platforms offer alternative ways to communicate and collaborate.

With Artificial Intelligence now central to healthcare, Technosoft excels in integrating AI solutions with existing systems, enhancing efficiency and patient care. Our expertise ensures that we effectively navigate the complexities of each integration option to deliver seamless, innovative, and impactful solutions. By leveraging our deep industry knowledge and cutting-edge technology, we empower healthcare providers to make data-driven decisions, streamline workflows, and improve patient outcomes. At Technosoft, we are not just integrating systems; we are pioneering the future of healthcare, one intelligent solution at a time.

Starting any AI Healthcare Integration Project? Get your questions answered in a Free 30-minute consultancy!

 

Comments