Data-Driven Decision Making
The Key to Making Better Decisions
The Basic Idea
Many of the decisions we make every day are grounded in models, or rather, representations of how something works. Sometimes these models are mental, like when we look into an empty room and envision how we may decorate it. Models may also be visual, like referencing Google Maps when traveling to a new destination. We frequently use these models to help us make better decisions and to simplify tasks.
The models we use to make decisions can also be mathematical, such as abstract models that employ mathematical language to describe the behavior of a system.7 This particular type of model falls under the umbrella of decision science: the data-driven, interdisciplinary application of behavioral sciences, business, computer science and technology.6 In essence, decision science is about using data to optimize the decision-making process.
Although most people do not use decision science explicitly in their day-to-day lives, it is commonly used by businesses and economists to undertake risk analysis, cost-benefit and cost-effectiveness analysis, simulation modeling and more.7
Theory, meet practice
TDL is an applied research consultancy. In our work, we leverage the insights of diverse fields—from psychology and economics to machine learning and behavioral data science—to sculpt targeted solutions to nuanced problems.
Big data: Large amounts of information that companies and statisticians accumulate, often from consumers. This often comes from data mining.
Data mining: A process companies use to transform raw data into usable and informative material. These data are often then used to build machine learning programs.
Cost benefit analysis: A systematic method of estimating the strengths and weaknesses of an approach. It provides businesses with a better understanding of which approach would result in maximum benefit and savings.3
Data science: An interdisciplinary field that uses scientific algorithms, methods and techniques to extract insights from structured and unstructured data. Data sciences view data as a tool for business optimization and refinement; it relies on big data and machine learning.2
Decision science: An application of complex quantitative techniques to facilitate and inform the decision-making process. Decision sciences view data as a tool for hypothesis development and decision making using logic and probability.2
Machine learning: A branch of artificial intelligence that involves building applications and computer programs who learn from data and improve their accuracy over time. Machines are trained to find patterns in data.
Risk analysis: A process that estimates the probability of an adverse event occurring, within corporate, governmental, or environmental industries.4 Risk analysis ensures that there is a balance between taking risks and reducing potential negative consequences.
Decision sciences stem from multiple approaches including risk analysis, cost-benefit analysis, and data science. It is important to remember that decision sciences are rooted in the underlying processes of how we make decisions; this is why they can relate to other fields but are fundamentally different from the field of data science, for example.
Risk analysis broadly refers to the process of assessing the probability of an adverse event occurring within business, government, or even the environmental sector. An application within the realm of decision science is to protect public health by communicating information about health risks within a framework that is useful for decision making.4 It’s especially important now, as the public is becoming increasingly aware of how their own health is impacted by environmental hazards and pollution. Therefore, governments, health organizations and private organizations need to make informed decisions to ensure they are able to respond to public wants and needs accordingly. A limitation, however, is that risk is based in probability, meaning it is virtually impossible to be precise when calculating risk. Moreover, risk analysis cannot account for the impact of extreme events.4,7
Cost-benefit analysis was conceptualized in 1848 by engineer Jules Dupuit but wasn’t implemented into public policy until 1958. Cost benefit analysis was then expanded to address topics relating to substance abuse and chemical waste. Now, it is most frequently used in economics, specifically welfare economics, as well as public policy. Its characteristic feature is measuring costs and benefits in the same unit, meaning that benefits are often measured in financial terms.7 A limitation of however, is that for long-term projects, a cost-benefit analysis may not be able to account for financial issues such as inflation or interest rates.3
Data science was officially termed in 2001 by computer scientist William S. Cleveland. Cleveland argued that statistics should extend beyond theory or data mining and instead be used for innovation. Now, data science is extremely powerful, as it informs virtually every aspect of our lives; it is the back-end source of information for all things technology. Data science has far-reaching effects into our everyday decision making.
Decision science is therefore an umbrella term for many quantitative techniques such as risk analysis and cost-benefit analysis that are used to improve decision making. It analyzes decisions as units of analysis, meaning it can and has informed decisions in law, education, environmental regulation, public health, and public policy. Decision science differs from other approaches, such as data science, in that it is concerned with making choices based on available data, rather than producing new data.7
Making decisions can be incredibly difficult and stress-inducing for individuals. Consider, however, how difficult it is to make a decision for the future of an entire company or government. Despite the massive amount of data companies and governments receive, and the expertise of their task forces, there is always the potential of interference by cognitive biases and subjective opinions.
Decision science mitigates some biases and provides companies and governments with a clear and remarkably accurate prediction of a potential decision outcome. Further, by calculating the risks and rewards of a decision, companies or policy makers can develop a better understanding of what they are getting into before making a final choice.
Prenatal genetic screening in Ontario
Non-invasive prenatal testing (NIPT) became available in Ontario in 2013. The Ministry of Health and Long-Term Care (MOHLTC) publicly funded many prenatal genetic screening tests since 1993; however, the NIPT was only available to those willing to pay for it when it first became available. The NIPT touted its ability to improve accuracy and safety in genetic testing, which heightened public interest. In response, the Ministry appointed a Prenatal Genetic Screening Group (PGSG) in 2014 to advise on ongoing screening practices and make recommendations for improved prenatal genetic screening in Ontario. Specifically, the Ministry requested an economic evaluation which included the costs and performance outcomes associated with NIPT.1
The purpose of prenatal genetic screening is to identify fetal chromosomal abnormalities, such as Down syndrome (Trisomy 21), in women of advanced maternal age (those older than 35). Early screening methods were referred to as amniocentesis, which involved a risk between 0.01 to 0.5% of fetal loss. More modern analyses were minimally invasive and became a routine part of check-ups for those older than 35. In Ontario, four screening tests were available at the time. By conducting an examination of the existing screening measures, the number of women who utilized them, and their effectiveness, the committee recommended that NIPT should be offered to women, despite the risks. Moreover, it was recommended that introducing NIPT into the public system could lead to more detection of Down syndrome cases, as well as fewer invasive tests and fewer related pregnancy losses. The committee notes, however, that NIPT should be used primarily as a contingency test, as making it mandatory would involve four to five times the cost. In this situation, the committee conducted risk-analysis and cost-benefit analysis to examine whether the NIPT would be a reasonable test to incorporate into the health system.
Informing public health guidelines
The US Preventive Services Task Force was created in 1984 with the aim of improving the health of Americans by making evidence-based recommendations about clinical services. These services include screenings, counselling services, and preventative medications. They base their recommendations on those of other agencies, empirical evidence from studies, and population data. An example of a recommendation of theirs is to improve screening for lung cancer.
In their statement, the task force reveals that lung cancer is the second most common form of cancer and leading cause of death in the US. They emphasize that the most important risk factor for lung cancer is smoking, as well as older age. Based on the existing data from a scoping review, and existing treatments and interventions, the task force argues that adults aged 50-80 years who have a 20 pack-per-year smoking history should be screened for lung cancer every year using low-dose computed tomography. This should be continued until a person has not smoked for 15 years.5
The process underlying the recommendations proposed by the US Preventive Services Task Force illustrates the usefulness of employing decision sciences, including cost-benefit analysis, and risk analysis. By considering the risks and costs associated with integrating screening and scans in preventing lung cancer, the authors can provide helpful recommendations.
Related TDL Resources
If you’re interested in learning more about decision sciences and why they matter, this article differentiates between data science and decision science and dives deeper into the applications of the latter.
If you’re interested in the related field of data science, this article outlines its history and how it permeates our everyday lives.
- Beck, D., Toole, J., John-Baptiste, A., (2015). Deciding Value for Money: Improving Prenatal Genetic Screening in Ontario in: Speechley, M., & Terry, A.L. [eds] Western Public Health Casebook 2015. London, ON: Public Health Casebook Publishing.
- Decision scientist vs. data scientist. (n.d.). Data Science Central. https://www.datasciencecentral.com/profiles/blogs/decision-scientist-vs-data-scientist
- How cost-benefit analysis process is performed. (n.d.). Investopedia. https://www.investopedia.com/terms/c/cost-benefitanalysis.asp
- How risk analysis works. (n.d.). Investopedia. https://www.investopedia.com/terms/r/risk-analysis.asp
- Recommendation: Lung cancer: Screening | United States preventive services Taskforce. (n.d.). United States Preventive Services Taskforce. https://uspreventiveservicestaskforce.org/uspstf/recommendation/lung-cancer-screening
- What is decision science? (2017, June 5). NDMU Online. https://online.ndm.edu/news/analytics/what-is-decision-science/
- What is decision science? (2020, October 20). Center for Health Decision Science. https://chds.hsph.harvard.edu/approaches/what-is-decision-science/