Do changes in health reveal the possibility of undiagnosed pancreatic cancer? Development of a risk-prediction model based on healthcare claims data

Abstract

Background and objective Early detection methods for pancreatic cancer are lacking. We aimed to develop a prediction model for pancreatic cancer based on changes in health captured by healthcare claims data. Methods We conducted a case-control study on 29,646 Medicare-enrolled patients aged 68 years and above with pancreatic ductal adenocarcinoma (PDAC) reported to the Surveillance Epidemiology an End Results (SEER) tumor registries program in 2004–2011 and 88,938 age and sex-matched controls. We developed a prediction model using multivariable logistic regression on Medicare claims for 16 risk factors and pre-diagnostic symptoms of PDAC present within 15 months prior to PDAC diagnosis. Claims within 3 months of PDAC diagnosis were excluded in sensitivity analyses. We evaluated the discriminatory power of the model with the area under the receiver operating curve (AUC) and performed cross-validation by bootstrapping. Results The prediction model on all cases and controls reached AUC of 0.68. Excluding the final 3 months of claims lowered the AUC to 0.58. Among new-onset diabetes patients, the prediction model reached AUC of 0.73, which decreased to 0.63 when claims from the final 3 months were excluded. Performance measures of the prediction models was confirmed by internal validation using the bootstrap method. Conclusion Models based on healthcare claims for clinical risk factors, symptoms and signs of pancreatic cancer are limited in classifying those who go on to diagnosis of pancreatic cancer and those who do not, especially when excluding claims that immediately precede the diagnosis of PDAC.

Early detection of pancreatic cancer

“…Electronic health records (EHR) offer opportunities to utilize longitudinal and cumulative healthcare data to build prediction models. Parametric models that inform high-risk for PDAC among individuals with NOD or new-onset pre-diabetes have been developed, with varying predictive performances for PDAC [103][104][105][106] . A model of NOD patients residing in Minnesota, USA incorporated age at onset of DM, and changes in weight and blood glucose to identify persons at 4.5% 3-year predicted risk of PDAC with 78% sensitivity and 80% specificity (AUC = 0.87) 103 .…”

“…A Medicare claims-based model among individuals with NOD incorporating multiple health indicators including pancreatitis, dyspepsia, depression, abdominal pain, weight jaundice, nausea/vomiting has been applied to an elderly population with NOD. The model identifies persons at 3.5% predicted risk of PDAC over 1-year (AUC = 0.73) 106 . Insurance claims-based models with diagnoses, such as pancreatitis, dyspepsia, abdominal pain, weight loss, and jaundice, incorporating parametric 106 or machine-learning methods 107 have also been developed with limited diagnostic performance (AUC = 0.73).…”

“…Insurance claims-based models with diagnoses, such as pancreatitis, dyspepsia, abdominal pain, weight loss, and jaundice, incorporating parametric 106 or machine-learning methods 107 have also been developed with limited diagnostic performance (AUC = 0.73). 106,107 . Reliance on coded diagnoses without laboratory parameters limited the performance of these models.…”

Diabetes Mellitus in the Middle-Aged and Elderly Population (>45 Years) and Its Association With Pancreatic Cancer: An Updated Review

“…Among the included articles, four studies explained that LTDM is an independent risk factor for PC [5,7,8,9]. Seven studies showed that NOD precedes the onset of PC [9][10][11][12][18][19][20], and 13 studies proposed as to how NOD with certain clinical factors and screening tests can be used as a predictor for PC screening [1,7,12,14,16,18,19,[21][22][23][24][25][26][27]. Seven studies mentioned the relation between PC and DM after pancreatic surgery [28][29][30][31][32][33][34].…”

“…A prediction model of medicare enrollees provided some information about the risk factors that could be used to test for emergent diagnosis of PC. However, it was not applicable for population screening since they have excluded the data for three months before the development of PC [24]. All the above studies are to be validated further by conducting population-based prospective interventional studies to find out the best strategy.…”