The use and abuse of data: nomograms and talking to patients about clinical medicine

The Development of a Model to Predict Cognitive Decline Within 12 Months in Home Care Clients

AIM

To develop and validate a model to predict cognitive decline within 12 months for home care clients without a diagnosis of dementia.

DESIGN

We included all adults aged ≥ 18 years who had at least two interRAI Home Care assessments within 12 months, no diagnosis of dementia and a baseline Cognitive Performance Scale score ≤ 1. The sample was randomly split into a derivation cohort (75%) and a validation cohort (25%). Significant cognitive decline was defined as an increase (deterioration) in Cognitive Performance Scale scores from '0' or '1' at baseline to a score of ≥ 2 at the follow-up assessment.

METHODS

Using the derivation cohort, a multivariable logistic regression model was used to predict cognitive decline within 12 months. Covariates included demographics, disease diagnoses, sensory and communication impairments, health conditions, physical and social functioning, service utilisation, informal caregiver status and eight interRAI-derived health index scales. The predicted probability of cognitive decline was calculated for each person in the validation cohort. The c-statistic was used to assess the model's discriminative ability. This study followed the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) reporting guidelines.

RESULTS

A total of 6796 individuals (median age: 82; female: 60.4%) were split into a derivation cohort (n = 5098) and a validation cohort (n = 1698). Logistic regression models using the derivation cohort resulted in a c-statistic of 0.70 (95% CI 0.70, 0.73). The final regression model (including 21 main effects and 8 significant interaction terms) was applied to the validation cohort, resulting in a c-statistic of 0.69 (95% CI 0.66, 0.72).

CONCLUSION

interRAI data can be used to develop a model for identifying individuals at risk of cognitive decline. Identifying this group enables proactive clinical interventions and care planning, potentially improving their outcomes. While these results are promising, the model's moderate discriminative ability highlights opportunities for improvement.

A novel nomogram model to predict the overall survival of patients with retroperitoneal leiomyosarcoma: a large cohort retrospective study

Retroperitoneal leiomyosarcomas (RLS) are the second most common type of retroperitoneal sarcoma and one of the most aggressive tumours. The lack of early warning signs and delay in regular checkups lead to a poor prognosis. This study aims to create a nomogram to predict RLS patients' overall survival (OS). Patients diagnosed with RLS in the Surveillance, Epidemiology, and End Results (SEER) database between 2000 and 2018 were enrolled in this study. First, univariable and multivariable Cox regression analyses were used to identify independent prognostic factors, followed by constructing a nomogram to predict patients' OS at 1, 3, and 5 years. Secondly, the nomogram's distinguishability and prediction accuracy were assessed using receiver operating characteristic (ROC) and calibration curves. Finally, the decision curve analysis (DCA) investigated the nomogram's clinical utility. The study included 305 RLS patients, and they were divided into two groups at random: a training set (216) and a validation set (89). The training set's multivariable Cox regression analysis revealed that surgery, tumour size, tumour grade, and tumour stage were independent prognostic factors. ROC curves demonstrated that the nomogram had a high degree of distinguishability. In the training set, area under the curve (AUC) values for 1, 3, and 5 years were 0.800, 0.806, and 0.788, respectively, while in the validation set, AUC values for 1, 3, and 5 years were 0.738, 0.780, and 0.832, respectively. As evidenced by the calibration curve, the nomogram had high prediction accuracy. Moreover, DCA revealed that the nomogram had high clinical utility. Furthermore, the risk stratification system based on the nomogram could effectively categorise patients into three mortality risk subgroups. Therefore, the developed nomogram and risk stratification system may aid in optimising the treatment decisions of RLS patients to improve treatment prognosis and maximise their healthcare outcomes.

Predictive Nomogram of Subsequent Liver Metastasis After Mastectomy or Breast-Conserving Surgery in Patients With Nonmetastatic Breast Cancer

BACKGROUND

Metastasis accounts for the majority of deaths in patients with breast cancer. Liver metastasis is reported common for breast cancer patients. The purpose of this study was to construct a nomogram to predict the likelihood of subsequent liver metastasis in patients with nonmetastatic breast cancer, thus high-risk patient populations can be prevented and monitored.

METHODS

A total of 1840 patients with stage I-III breast cancer were retrospectively included and analyzed. A nomogram was constructed to predict liver metastasis based on multivariate logistic regression analysis. SEER database was used for external validation. C-index, calibration curve and decision curve analysis were used to evaluate the predictive performance of the model.

RESULTS

The nomogram included 3 variables related to liver metastasis: HER2 status (odds ratio (OR) 1.86, 95%CI 1.02 to 3.41; P = 0.045), tumor size (OR 3.62, 1.91 to 6.87; P < 0.001) and lymph node metastasis (OR 2.26, 1.18 to 4.34; P = 0.014). The C index of the training cohort, internal validation cohort and external validation cohort were 0.699, 0.814 and 0.791, respectively. The nomogram was well-calibrated, with no statistical difference between the predicted and the observed probabilities.

CONCLUSION

We have developed and validated a robust tool enabled to predict subsequent liver metastasis in patients with nonmetastatic breast cancer. Distinguishing a population of patients at high risk of liver metastasis will facilitate preventive treatment or monitoring of liver metastasis.

Predictive Nomogram for Midterm to Long-Term Prognosis in Patients with Papillary Renal Cell Carcinoma Based on Data from the Surveillance, Epidemiology, and End Results (SEER) Program

BACKGROUND This study aimed to develop a predictive nomogram for midterm to long-term prognosis in patients with papillary renal cell carcinoma (RCC) based on data from the US Surveillance, Epidemiology, and End Results (SEER) program. MATERIAL AND METHODS Clinical pathology data and follow-up information were obtained from the SEER database for patients with papillary RCC between 1997-2014. Univariate and multivariate Cox regression models evaluated the independent prognostic factors, and the nomogram was constructed to predict the 3-year, 5-year, and 10-year survival rates. Multiple parameters were estimated to evaluate the predictive values, including the concordance indices (C-indices), calibration plots, area under the receiver operator characteristics (ROC) curve, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision curve analysis (DCA). RESULTS The study included 13,926 patients with papillary RCC. Univariate and multivariate Cox regression analysis developed the nomogram that relied on the predictive variables of age, Fuhrman grade, TNM stage, surgery of the primary site, lymphadenectomy, and marital status. The C-indices of the novel model in the validation cohort were more satisfactory than those of the TNM classification. Accurate discrimination and calibration by the nomogram were identified in both cohorts. The NRI and IDI supported prediction improvements, and the DCA supported the nomogram's clinical significance. CONCLUSIONS A nomogram was developed to evaluate the prognosis of papillary RCC and to identify the patients who required specialized treatment. However, external validation of the predictive nomogram is required that also includes patients from other countries.

Predictive models in cancer management: A guide for clinicians

BACKGROUND

Predictive tools in cancer management are used to predict different outcomes including survival probability or risk of recurrence. The uptake of these tools by clinicians involved in cancer management has not been as common as other clinical tools, which may be due to the complexity of some of these tools or a lack of understanding of how they can aid decision-making in particular clinical situations.

AIMS

The aim of this article is to improve clinicians' knowledge and understanding of predictive tools used in cancer management, including how they are built, how they can be applied to medical practice, and what their limitations may be.

METHODS

Literature review was conducted to investigate the role of predictive tools in cancer management.

RESULTS

All predictive models share similar characteristics, but depending on the type of the tool its ability to predict an outcome will differ. Each type has its own pros and cons, and its generalisability will depend on the cohort used to build the tool. These factors will affect the clinician's decision whether to apply the model to their cohort or not.

CONCLUSIONS

Before a model is used in clinical practice, it is important to appreciate how the model is constructed, what its use may add over and above traditional decision-making tools, and what problems or limitations may be associated with it. Understanding all the above is an important step for any clinician who wants to decide whether or not use predictive tools in their practice.

Statistical consideration for clinical biomarker research in bladder cancer

OBJECTIVE

To critically review and illustrate current methodological and statistical considerations for bladder cancer biomarker discovery and evaluation.

METHODS

Original, review, and methodological articles, and editorials were reviewed and summarized.

RESULTS

Biomarkers may be useful at multiple stages of bladder cancer management: early detection, diagnosis, staging, prognosis, and treatment; however, few novel biomarkers are currently used in clinical practice. The reasons for this disjunction are many and reflect the long and difficult pathway from candidate biomarker discovery to clinical assay, and the lack of coherent and comprehensive processes (pipelines) for biomarker development. Conceptually, the development of new biomarkers should be a process that is similar to therapeutic drug evaluation-a highly regulated process with carefully regulated phases from discovery to human applications. In a further effort to address the pervasive problem of inadequacies in the design, analysis, and reporting of biomarker prognostic studies, a set of reporting recommendations are discussed. For example, biomarkers should provide unique information that adds to known clinical and pathologic information. Conventional multivariable analyses are not sufficient to demonstrate improved prediction of outcomes. Predictive models, including or excluding any new putative biomarker, need to show clinically significant improvement of performance in order to claim any real benefit. Towards this end, proper model building, avoidance of overfitting, and external validation are crucial. In addition, it is important to choose appropriate performance measures dependent on outcome and prediction type and to avoid the use of cutpoints. Biomarkers providing a continuous score provide potentially more useful information than cutpoints since risk fits a continuum model. Combination of complementary and independent biomarkers is likely to better capture the biological potential of a tumor than any single biomarker. Finally, methods that incorporate clinical consequences such as decision curve analysis are crucial to the evaluation of biomarkers.

CONCLUSIONS

Attention to sound design and statistical practice should be delivered as early as possible and will help maximize the promise of biomarkers for patient care. Studies should include a measure of predictive accuracy and clinical decision-analysis. External validation using data from an independent cohort provides the strongest evidence that a model is valid. There is a need for adequately assessed clinical biomarkers in bladder cancer.

Using administrative data to develop a nomogram for individualising risk of unplanned admission to intensive care

AIM

Although unplanned admissions to the intensive care unit (ICU) are associated with poorer prognoses, there is no published prognostic tool available for predicting this risk in an individual patient. We developed a nomogram for calculating the individualised absolute risk of unplanned ICU admission during a hospital stay.

METHOD

Hospital administrative data from a large district hospital of consecutive admissions from 1 January 2000 to 31 December 2006 of aged over 14 years was used. Patient data was extracted from 94,482 hospital admissions consisted of demographic and clinical variables, including diagnostic categories, types of admission and time and day of admission. Multivariate logistic regression coefficients were used to develop a predictive nomogram of individual risk to patients admitted to the study hospital of unplanned ICU admission.

RESULTS

A total of 672 incident unplanned ICU admissions were identified over this period. Independent predictors of unplanned ICU admissions included being male, older age, emergency department (ED) admissions, after-hour admissions, weekend admissions and six principal diagnosis groups: fractured femur, acute pancreatitis, liver disease, chronic airway disease, pneumonia and heart failure. The area under the receiver operating characteristic curve was 0.81.

CONCLUSION

The use of a nomogram to accurately identify at-risk patients using information that is readily available to clinicians has the potential to be a useful tool in reducing unplanned ICU admissions, which in turn may contribute to the reduction of adverse events of patients in the general wards.