Realizing the Potential of Clinical Risk Prediction Models

Abstract

A t the turn of the last century, physicians were largely guided by lessons passed down in training, their own personal experience, and the experiences of their colleagues. Although this approach produced thoughtful clinicians, a key limitation remained-even the busiest, most experienced providers could see only so many patients, experience only a limited number of outcomes, and often struggled to ascertain the accuracy of diagnoses or the effectiveness of treatment. These challenges to delivery of safe and effective patient care were subsequently addressed by a growing focus on progressively larger and better designed cohort studies and randomized clinical trials and later by distilling these insights into clinical practice guidelines and appropriate use criteria to help summarize the rapidly evolving medical literature. Article see p 368Despite this exponential growth in well-conducted clinical research, a barrier in applying these studies into clinical care is that an individual patient may not obtain the average benefit observed in a clinical trial. It is well recognized that the heterogeneity of treatment effect across a population can be obscured by focusing only on the mean treatment effect in a population.1-3 Depending on a patient's age, sex, comorbidities, and other characteristics, that patient may benefit greatly from the same treatment that poses a significant risk for another. 4 Providers, after all, are concerned with delivering the safest and most effective treatment for a particular individual, rather than a population of patients. Accordingly, a growing focus has been placed on developing and implementing tools to identify which patients are likely to benefit from a particular treatment or strategy, those who may be harmed, and those for whom balanced risks and benefits exist that should prompt discussions between patients and providers about that patient's goals and preferences for care. To address this need, there has been a proliferation of published clinical prediction models to help tailor treatment to risk.In this issue of Circulation: Cardiovascular Quality and Outcomes, Wessler et al 5 report findings from a rigorous review of clinical prediction models in cardiovascular disease. Examining over 20 years of contemporary literature, they found ≈800 models focusing on conditions across the entire spectrum of cardiovascular disease. In fact, they found that the number of new clinical prediction models has doubled each decade. The growth of prediction models is encouraging, as the entire profession seeks to better understand the outcomes of their patients and how best to optimize these outcomes. Importantly, however, they highlight many important challenges, including the design, development, and testing of prediction models.With respect to the design of clinical prediction models, it is critically important that the results of the model would alter a clinical decision. If a physician will not treat a low-or high-risk patient differently, then what value is there in using a model? With respe...

Personalized evidence based medicine: predictive approaches to heterogeneous treatment effects

“…Nonetheless, substantial barriers still need to be to overcome 148. We list a few of the outstanding research questions related to the problems covered in this article in the Questions for future research box.…”

(Cost-)effectiveness of an individualised risk prediction tool (PERSARC) on patient’s knowledge and decisional conflict among soft-tissue sarcomas patients: protocol for a parallel cluster randomised trial (the VALUE-PERSARC study)

“…However, there are also important challenges including the design, development and testing of these risk prediction models. Until now, studies have mostly focused on developing risk prediction models 49. For example, there are many risk prediction models for several cancer types, but these models have not been validated and are, therefore, frequently not appropriate for use in clinical practice 50.…”

“…Until now, studies have mostly focused on developing risk prediction models. 49 For example, there are many risk prediction models for several cancer types, but these models have not been validated and are, therefore, frequently not appropriate for use in clinical practice. 50 The use of risk prediction models in clinical practice for patients with cancer is another important challenge.…”

Development and validation of a real‐world model to predict 1‐year Level 3 (severe) hypoglycaemia risk in adults with diabetes (theiNPHORMstudy, United States)

“…age, gender, medical history) to predict an outcome occurring in future 17 . The resulting model links research to practice by producing an evidence‐based, patient‐specific risk estimate that clinicians can use to guide safer management 18,19 . In a 2020 ADA report, Gabbay et al endorsed the increased uptake of such decision‐support tools as a strategy to enrich personalized care 20 …”

“…17 The resulting model links research to practice by producing an evidence-based, patient-specific risk estimate that clinicians can use to guide safer management. 18,19 In a 2020 ADA report, Gabbay et al endorsed the increased uptake of such decision-support tools as a strategy to enrich personalized care. 20 While in the United States, several SH prognostic models existevincing strong interest in the field-none predict the full spectrum of Level 3 event risk.…”