Cookies on this website
We use cookies to ensure that we give you the best experience on our website. If you click 'Continue' we'll assume that you are happy to receive all cookies and you won't see this message again. Click 'Find out more' for information on how to change your cookie settings.

Abstract Background: Clinical prediction models are increasingly made available on the Internet or as applications for smart phones. These models are available for use to both clinicians and the general public. However, the evidence of whether the model should be used is often unclear. Aim: The aim of this study was to review the availability of clinical prediction models as calculators on the Internet and identify the evidence base on the performance of the model. We also provide some guidance on principles to follow when preparing a prediction model calculator to be made available on the Internet. Methods: In March 2015, the Google search engine was used with a combination of fifteen search terms that described the concepts of prediction and calculator. For each search term, the first 50 hits (total 750 websites) were recorded. Websites that presented an online clinical prediction model that required manual entry of patient level information that produced a probability or risk of having an undiagnosed condition (diagnostic) or a probability or risk of developing a health condition in the future (prognostic), were eligible. Information such as the background of the model (e.g. country), intended patient population, intended user of the model (clinician or patient), information on how the prediction models were developed and any details about their validation model, and finally information on how the web-calculators are presented (graphical, text, lay terminology) and to be used were extracted using a pilot-tested extraction form. Results: A total of 116 models were included; only less than half of the websites cited references to the articles describing the development of model and only 8 websites cited references for the validation of the model on the website. Most of the prediction models are poorly documented on the Internet, with little information to help users actually use them. In many instances, it was unclear on who the prediction calculator was intended for (with only 44 mentioning the target group), less than 20% of websites provided help to use the model (including frequently asked questions). Only 25 models reported the description for each risk factors, nearly half of the models (n=56) presented no information or checks on the ranges of any continuous factors. Furthermore, many calculators (n=33) did not display warning messages when information is entered incorrectly. Conclusion: Prediction models are widely available on the Internet to support decision-making for clinicians and general public, yet the information presented alongside the models is inadequate, should be used with care.

Type

Poster