Development of nomograms to assess the risk of clinical outcome

Abstract

A methodology is presented for developing nomograms for assessing and stratifying the risk of a clinical outcome based on the created virtual data set using the R software environment. The virtual data set included input numerical and factor variables (variable types correspond to the R software documentation) and outcome. For quantitative variables, descriptive statistics were calculated at all levels of the outcome variable, and mosaic diagrams were constructed for factor variables. As a model that describes the association of input variables with the outcome, a logistic regression model was used. A bootstrap method was applied to validate and evaluate the model performance. The calculated validity indicators showed an acceptable discriminatory ability of the predictive model. The statistical calibration demonstrated the proximity of the model’s calibration curve to the ideal calibration curve. Based on the logistic regression coefficients, a nomogram was constructed using which the risk value of a specific outcome was calculated for each subject (patient). It is shown that with the help of the presented technique it is possible to stratify patients effectively by the risk of an adverse outcome, thus adequately altering the diagnosis and treatment tactics. The use of a nomogram greatly simplifies risk assessment and can be used in paper form as a supplement to the patient examination protocol. The article contains the codes of the R programming language with explanations.

Keywords

nomogram / virtual data / bootstrap method / prediction of clinical outcome / logistic regression / R language / model validation / risk stratification / risk assessment

References

[1] Корнеенков, А.А. Использование модифицированной процедуры последовательного распознавания Вальда для определения исхода оперативного лечения у пациентов с болезнью Меньера / А.А. Корнеенков [и др.] // Росс. оториноларингол. – 2018. – № 3 (94). – С. 54–59.

[2] Светуньков, И.С. Методы социально-экономического прогнозирования в 2 т. Т. 2 Модели и методы: учебник и практикум для академического бакалавриата / И.С. Светуньков, С.Г. Светуньков. – М.: Издательство Юрайт, 2018. – C. 17–25.

[3] Кабаков, Р.И. R в действии. Анализ и визуализация данных в программе R / И. Кабаков; пер. с англ. П.А. Волковой. – М.: ДМК Прес, 2016. – 588 с.

[4] Cleveland, W.S. Robust Locally Weighted Regression and Smoothing Scatterplots / W.S. Cleveland // American Statistical Association. – 1979. – Vol. 74, № 368. – P. 829–836.

[5] Efron, B. Bootstrap methods: Another look at the jackknife / B. Efron // Ann. Statist. – 1979. – № 7. – P. 1–26.

[6] Hartigan, J.A. Mosaics for Contingency Tables / J.A. Hartigan, B. Kleiner // Computer Science and Statistics: Proceedings of the 13th Symposium on the Interface. – New York: Springer, 1981. – P. 268–273.

[7] Harrell, F.E.Jr. Regression Modeling Strategies / E.F.Jr. Harrell. – Switzerland: Springer International Publishing, 2015. – 582 p.

RIGHTS & PERMISSIONS

Korneenkov A.A., Kuzmin S.G., Dergachev V.B., Borisov D.N.