Intro
For Pharma companies looking to differentiate and strengthen their drugs with a beyond-the-molecule offering, a clinical decision support system (CDSS or a clinical decision support algorithm) could be an option.
CDSS are becoming increasingly common to augment physicians’ decision-making as the number of treatment options increases and sub-populations within a disease are more often diagnosed and classified causing treatment algorithms to get more complicated.
This article is aimed at brand/portfolio managers or those working in Pharma/MedTech innovation teams. It will give you an introduction to CDSS and cover concepts to help you understand if the development of one could support your goals.
We’ll cover
- The history and classification of CDSS
- The benefits and challenges with implementing them
- Two examples of CDSS
- Opportunities where they could be used by a pharma company
History and Definition
CDSS originated in the late 70s as clunky academic tools, poorly integrated and often unusable outside the leading centers where they were developed. The prototype systems included MYCIN, fpr antibiotic selection, and INTERNIST-1, which aided physicians in diagnosing complex diseases.
Throughout the 1980s and 90s, their use spread, with large hospitals designing CDSS based on digital flow charts and Boolean logic to support adverse drug event monitoring, antibiotic prescribing, ventilator management, and other needs internally.
As hospitals adopted more digital systems in the late 1990s and 00s, integrating these tools with Electronic Health Records (EHR) became essential. This enabled the automatic sharing of patient information with CDSS and reduced input errors.
The explosion in machine learning and artificial intelligence over the last 10 years has meant the development of increasingly complex assistants that can access larger datasets and give recommendations and guidance. However, this has brought its own challenges, which are discussed below.
The number of publications for new technology is always good for observing growth and foreshadowing adoption. With a detailed search on PubMed, you can see in the image below growth from <100 a year in the 90s and 00s to 100+ a year since 2011.
The literature provides multiple approaches to classifying systems based on multiple characteristics. When you start reading the classifications below, you can begin thinking about what might be suitable or wouldn’t work in your disease or therapy area.
- Underlying decision-making process: The original and simplest models were digital coding of flow charts, known as Knowledge-Based models. These are still a compelling option in many instances.
- Human-computer interaction: How do users set up, interact with, and control the system? Some may be voice-enabled or use a traditional keyboard/mouse through a PC.
- Active-passive: Defined by whether information is sent to HCPs automatically through alerts or whether the HCP/user needs to input information or control the system.
- System function: This is an older classification that is nearly obsolete, as most systems now support diagnosis and then recommend the next steps. However, in the literature, this is described as the system deciding “what is true,” aka diagnosis, and then “what to do.”
Benefits and challenges
We know that treating patients, including the multiple stakeholders involved in their care and administration, is complex, and introducing a new component in the journey could significantly impact healthcare delivery.
This is a source of multiple benefits but presents many challenges that must be addressed in the design, testing and implementation.
Before discussing the points, it is important to note that CDSS should not be thought of as replacing HCPs but as a partnership, with the physician’s knowledge and patient interaction complimented by the CDSS’s ability to harness data sets or observations and interpret them in ways humans can not.
Benefits
- Patient safety: Whether flagging potential drug-drug interactions, ensuring correct doses are assigned, or raising side effects that could exacerbate comorbidities, this is a huge advantage for using CDSS, especially in stretched healthcare systems.
- Adherence to clinical guidelines /educational tools: There is strong evidence that clinicians find it difficult to implement therapeutic guidelines in practice (31, 32). This is unsurprising given the complexity of many diseases and the time the average clinician needs to read and adjust their practice, especially when covering multiple diseases.
- Cost containment: health systems can ensure appropriate diagnostic tests and medicines are used (including generics or cheaper alternatives) and potentially reduce patient length of stay through better care.
- Scalability: With the integration of hospital systems, CDSS can often be deployed across an entire network so HCPs can benefit regardless of location or facility
- Interoperability and data sharing: If patients are interacting with different specialists, their data from different sources are captured and incorporated into the decision-making process
Challenges
- Patient and HCP trust: Several studies have shown that HCPs have concerns about CDSS quality, which has only increased with the number that rely on advice generated by “black box” AI systems and the fact they do not always incorporate patient preference
- Fragmented workflows: User studies in healthcare and other industries have shown if extra software windows are needed or the software does not smoothly integrate with existing workflows, they will face an uphill battle for adoption
- Alert fatigue: For passive systems, there are concerns about the amount of signal noise that could be generated, especially in a world where people are already overwhelmed with notifications
- Impact on skill and HCP collaboration: Historically, HCPs collaborated and checked each other’s works, sharing opinions on the best course of action, which some may skip/ completely replace with CDSS. There is also the potential that, over time users of CDSS come to rely on them completely, resulting in weaker skills themselves.
- Data quality / incorrect content: Ensuring the database a tool is based on is kept up to date, and for more intelligent CDSS, they access the correct sources and do not “hallucinate” information is an essential challenge to manage
Example Clinical Decision Support Systems
To better understand CDSS, I’ll briefly outline 2 tools developed and used in different settings that can help achieve outcomes
We’ll start with healthcare conglomerate McKesson’s Clear Value Plus, an oncology CDSS powered by National Comprehensive Cancer Network guidelines.
The tool provides guidance for physicians at the point of care, ensuring they use the most up-to-date best practices.It is designed to highlight evidence-based treatment options that enable personalized medicine and appropriate diagnostic tests.
A magazine interview with a KOL gave the example of a community oncologist who hadn’t seen a patient with advanced pancreatic cancer for four months and couldn’t recall what regimens would be needed or what biomarkers to test from memory. With Clear Value Plus, this information is displayed on the screen after entering the patient stage.
The tool also provides visibility into treatment costs, including the cost-effectiveness of different options. It accelerates prior authorization and reimbursement by confirming regimen selections are on guideline and on pathway and gives the option to participate in value-based care programs.
A second example, in a very different setting, is an algorithm to guide antibiotic prescription in Tanzania.
Bacterial antimicrobial resistance was responsible for 1.27 million deaths in 2019, with the highest burden in sub-Saharan Africa. Resource-constrained countries like Tanzania prescribe antibiotics in 80-90% of outpatient cases regardless of diagnosis, which is often an inappropriate action in many cases.
Ifakara Health Institute in Tanzania, collaborating with several Swiss universities, developed a CDSS that helped reduce antibiotic prescriptions to 23.2% at interventional facilities vs. 70.1% in the control facilities providing their usual care.
This large decrease in prescription of antibiotics did not impact quality of care, with a non-inferior coprimary outcome of clinical failure on day 7, and no difference in secondary safety outcomes, including death and non-referred secondary hospitalizations.
A tool like this shows the overwhelming benefit of ensuring patients receive the correct medication to address health needs and reduce inappropriate medication use.
Why might a pharma company be interested in developing a CDSS?
With the background and examples for CDSS, we’ll now look at how they could give a competitive edge in crowded markets for a Pharma brand to help differentiate itself and gain an edge.
These examples wouldn’t be suitable for all companies facing the situation described, and a feasibility study engaging your target audience with primary research would be essential, alongside analysis of the regulatory conditions, understanding stakeholder unmet needs who would interact with the CDSS, and assessing internal capabilities to develop and implement/commercialize.
Examples:
- A company has identified that the prescription volume for its drug does not match the expected epidemiology of the sub-population it is designed for, as it is not being correctly diagnosed or routed.
- A company’s drug has been newly included at an earlier stage in the treatment algorithm or guidelines based on long-term follow-up, real-world evidence, or other data.
- A company wants to build data on the patient journey, population characteristics, and HCP decisions for future treatments in its portfolio.
- As a way of engaging HCPs and patients to increase brand name awareness ahead of generic entry
- Pure revenue generation approach using internal digital capabilities and knowledge of a disease area to develop a subscription tool
- Performance of a charitable/altruistic act to develop a CDSS for an example like the antibiotics situation in Tanzania
Wrap Up
This is a brief introduction to CDSS. The links below will help you better understand the tools’ history, benefits, challenges, and many other aspects.
As a pharma leader, this will likely have sparked some ideas about how it could support your brand or portfolio goals, or you may have ruled it out as unsuitable due to complexities in your area. Either way, if you would like to discuss the topic in a consultation, get in touch.
Also, if anyone has encountered any interesting CDSS or has experience using them.
References
https://www.nature.com/articles/s41746-020-0221-y – Great detailed overview on decision support tools
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5171504/ – historical review of clinical decision support tools including early years
https://openheart.bmj.com/content/10/2/e002432 – Detailed review of benefits and challenges with using CDSS
https://www.ncbi.nlm.nih.gov/books/NBK543516/ – Book chapter from a data science point of view on the technical classification and aspects of CDSS
https://www.ontada.com/providers-solutions/clear-value-plus/ – Marketing website for Clear Value Plus
https://www.nature.com/articles/s41591-023-02633-9 – Article on Tanzanian CDSS