Reporting of artificial intelligence prediction models.

By Gary S. Collins, Karel G. M. Moons. The Lancet, 2019

558 citations recorded by the source corpus.

www.thelancet.com Vol 393 April 20, 2019 1577 shortcut, investing in magic bullets such as vaccines, antiretrovirals, and bednets that could be distributed without a well functioning healthcare system, and even in war zones. These programmes saved millions of lives, but failed to launch the anticipated virtuous cycle of development, poverty reduction, and improved health systems. Many African countries remain deeply impoverished, and although child mortality has plummeted, weak health systems remain ill equipped to address soaring numbers of deaths and disability from heart disease, cancer, and other non communicable diseases. Bollyky and colleagues’ findings suggest a possible reason for this dispiriting reality: many of the countries with the worst health systems are governed by corrupt autocrats who retain power by force and can ignore the welfare of their people without repercussions. To date, the global development community has skirted the complications of politics, instead emphasising medical programmes and the empower ment of women, the LGBT+ (lesbian, gay, bisexual, transgender, and all other identities) community, and people with disabilities—ie, people who do not directly challenge government power. This emphasis is sometimes justified by the longstanding assumption that in poor countries, especially, dictators are better at getting things done because they can ignore the demands of petty, competing constituencies. However, Bollyky’s findings support the theory that dictators might themselves be a cause of poverty and illness, and that democrats, however befuddled and disorganised, better serve their people. Unfortunately, the rights of prodemocracy activists are routinely violated with impunity, especially in those African countries most beset by poverty and ill health. Recent elections in Kenya, Uganda, Ethiopia, Rwanda, the Democratic Republic of the Congo, Gabon, Cameroon, and Zimbabwe were all marred by credible rigging allegations, which donors, in most cases, dismissed. When two opposition members of parliament in donor darling Uganda were beaten and crippled by security forces inside parliament in September, 2017, not one international donor or human rights organisation spoke out. Global health advocacy groups need to do more than clamour for more funding and occasionally bemoan corruption. They need to call on Washington (USA), Brussels (Belgium), London (UK), and other donors to impose sanctions on dictators, including those who cooperate with western military aims. As Rudolf Vichow, one of the pioneers of modern public health, wrote after witnessing the ravages of typhus on the oppressed peasants of Silesia, “politics is nothing but medicine at a larger scale”.

Subfields: Cardiology and Cardiovascular Medicine, Epidemiology

View Paper: https://www.semanticscholar.org/paper/e300dd84a6546c40aa19fc8cd92cf4e871ac2cc2

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cardiology and Cardiovascular Medicine

In Epidemiology

Citations in the corpus

Improving Reproducibility in Machine Learning Research (A Report from the NeurIPS 2019 Reproducibility Program) (2020, background): “Other guidelines are under development outside of the ML community, namely for the application of AI tools in clinical trials (Liu et al., 2019) and health-care (Collins and Moons, 2019).”
Key challenges for delivering clinical impact with artificial intelligence (2019, unknown): “…version of the TRIPOD statement that is specific to machine learning prediction algorithms (TRIPOD-ML) is in development and will focus on the introduction of machine learning prediction algorithms, establishing methodological and reporting standards for machine learning studies in healthcare [64].”
Deep learning in histopathology: the path to the clinic (2021, unknown): “Ideally, such trials would use clinical outcomes as end points to demonstrate long-term effects 115 and apply standardized reporting methods such as TRIPOD-AI 134 , which is currently under development.”
Risk of bias in studies on prediction models developed using supervised machine learning techniques: systematic review (2021, methodology): “Currently, extensions of TRIPOD and PROBAST for prediction models developed using machine learning are under development (TRIPOD-AI, 38 39 As sample size contributed largely to the overall high risk of bias, future methodological research could focus on determining the appropriate sample sizes for…”
Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension (2020, unknown): “32,62 Other potentially relevant guidelines, which are agnostic to study design, are registered with the EQUATOR Network.”
Supervised machine learning: A brief primer (2020, methodology): “…also should follow standard reporting guidelines such as the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement (Collins et al., 2015), which are currently being adapted for machine learning applications (Collins & Moons, 2019).”
External validation of prognostic models: what, why, how, when and where? (2020, background): “However, the inherent complexity of these models complicates risk calculation and external validation of such ‘black-box’ models is still highly infrequent [39].”
Guidelines for clinical trial protocols for interventions involving artificial intelligence: the SPIRIT-AI Extension (2020, unknown): “…of AI models should refer to TRIPOD-ML (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis–Machine Learning) 24 and STARD-AI (Standards For Reporting Diagnostic Accuracy Studies–Artificial Intelligence) 51 , both of which are currently under development.”
Metrics Reloaded: Recommendations for image analysis validation (2022, unknown): “Make uncertainties in rankings (if any) explicit [93] Depending on the specific application, apply existing reporting guidelines, such as BIAS [59] (for challenges), TRIPOD [67] (for validation of prediction models, including recommendations that are applicable to the validation of medical image ana”