In considering the role of data analytics in the promotion of patient safety, it is possible to chart a path along which the role of data becomes increasingly integral and directive. This path can be diagrammed as follows: descriptive analytics ➡️ predictive analytics ➡️ prescriptive analytics.1,2,3
Effective descriptive analytics involves finding the right terms, measures, and categories to clearly characterize a subject. An example of outstanding descriptive analytics is Candello’s annual report, Medical Malpractice in America, which examined 10 years of malpractice data across all facets of medical professional liability—including case rates, financials, responsible clinical service, and contributing factors, among other areas.
The 2020 Benchmarking Report, The Power to Predict, represents predictive analytics, with a stride toward prescriptive analytics. Predictive analytics develops measures that show what aspects of malpractice cases make the outcome of interest—in this report, a case closing with an indemnity payment—more likely. This, in turn, enables patient safety and risk management teams to ascertain the common features and process breakdowns of such cases and so identify what actions need to be taken both to make care safer and reduce liability risk—in other words, prescriptive analytics.
Using logistic regression and random forest methods, the 2020 Candello Report identifies specific contributing factors that predict that a malpractice case will close with an indemnity payment. For example, a case with a contributing factor of a policy or protocol not being followed has an odds ratio of 2.45. That is, cases in which this contributing factor is present are 2.45 times more likely to close with an indemnity payment compared to cases in which this contributing factor is absent.
Knowing that a policy or protocol not being followed predicts an increased probability that a case will close with payment is actionable. Equipped with this knowledge, risk management and patient safety teams can examine cases that have this contributing factor to identify common elements that lend themselves to solutions that enhance patient safety.
Following this approach in the 2020 Candello Report, we found that many of the cases in which a policy or protocol was not followed involved patient falls. The next step suggested by these data are to review these types of cases and ascertain whether the issue is that a policy was not followed (for example, a fall risk assessment was not performed on admission, as required) or a new policy is needed (for example, there needs to be a policy that an identified fall risk in a patient must be communicated when that patient goes for diagnostic testing). This approach allows risks to be identified and mitigated, and it represents the transition from predictive analytics to prescriptive analytics.
PRESCRIPTIVE DATA IN THE CLINICAL REALM
In the clinical sphere, what makes clinical information valuable is its ability to influence your treatment decisions—in medical vernacular, it “changes your management” of the patient. A chest x-ray obtained on admission to the hospital that informs the decision whether to start antibiotics to treat a possible pneumonia is valuable information. In contrast, a chest x-ray obtained to provide reassurance that the pneumonia is getting smaller a few days after antibiotics were started may describe that that pneumonia is smaller. However, it provides minimal useful information, since the patient is going to have to complete the course of antibiotics regardless. As analytics progresses from descriptive to predictive to prescriptive, the results of the analysis become more likely to “change your management” in the patient safety sphere, and provide information that catalyzes improvements in patient safety.
Candello’s database (Candello)—with its hundreds of thousands of malpractice cases that are coded for contributing factors and other case attributes—provides the granular data needed to produce the type of predictive and prescriptive analytics found in the 2020 Candello Report, which drive improved patient safety and reduced liability risk.
 Soltanpoor R, Sellis T. Prescriptive Analytics for Big Data. In: Cheema MA, Zhang W, Chang L, eds. Databases Theory and Applications. Cham, Switzerland: Springer International Publishing; 2016:245-256.
 Krumeich J, Werth D, Loos P. Prescriptive control of business processes. Business & Information Systems Engineering. 2016;58(4):261-280.
 Lepenioti K, Bousdekis A, Apostolou D, Mentzas G. Prescriptive analytics: Literature review and research challenges. International Journal of Information Management. 2020;50:57-70.