Dataset Information

The effects of data sources, cohort selection, and outcome definition on a predictive model of risk of thirty-day hospital readmissions.

ABSTRACT: Hospital readmission risk prediction remains a motivated area of investigation and operations in light of the hospital readmissions reduction program through CMS. Multiple models of risk have been reported with variable discriminatory performances, and it remains unclear how design factors affect performance.To study the effects of varying three factors of model development in the prediction of risk based on health record data: (1) reason for readmission (primary readmission diagnosis); (2) available data and data types (e.g. visit history, laboratory results, etc); (3) cohort selection.Regularized regression (LASSO) to generate predictions of readmissions risk using prevalence sampling. Support Vector Machine (SVM) used for comparison in cohort selection testing. Calibration by model refitting to outcome prevalence.Predicting readmission risk across multiple reasons for readmission resulted in ROC areas ranging from 0.92 for readmission for congestive heart failure to 0.71 for syncope and 0.68 for all-cause readmission. Visit history and laboratory tests contributed the most predictive value; contributions varied by readmission diagnosis. Cohort definition affected performance for both parametric and nonparametric algorithms. Compared to all patients, limiting the cohort to patients whose index admission and readmission diagnoses matched resulted in a decrease in average ROC from 0.78 to 0.55 (difference in ROC 0.23, p value 0.01). Calibration plots demonstrate good calibration with low mean squared error.Targeting reason for readmission in risk prediction impacted discriminatory performance. In general, laboratory data and visit history data contributed the most to prediction; data source contributions varied by reason for readmission. Cohort selection had a large impact on model performance, and these results demonstrate the difficulty of comparing results across different studies of predictive risk modeling.

SUBMITTER: Walsh C

PROVIDER: S-EPMC4261028 | biostudies-literature | 2014 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

The effects of data sources, cohort selection, and outcome definition on a predictive model of risk of thirty-day hospital readmissions.

Walsh Colin C Hripcsak George G

Journal of biomedical informatics 20140823

<h4>Background</h4>Hospital readmission risk prediction remains a motivated area of investigation and operations in light of the hospital readmissions reduction program through CMS. Multiple models of risk have been reported with variable discriminatory performances, and it remains unclear how design factors affect performance.<h4>Objectives</h4>To study the effects of varying three factors of model development in the prediction of risk based on health record data: (1) reason for readmission (pr ...[more]

PMID: 25182868

Similar Datasets

Project description:ImportanceThe Medicare Hospital Readmissions Reduction Program (HRRP) has disproportionately penalized hospitals that treat many black patients, which could worsen health outcomes in this population.ObjectiveTo determine whether short-term mortality rates increased among black and white adults 65 years and older after initiation of the HRRP and whether trends differed by race.Design, setting, and participantsIn a cohort study using an interrupted time-series analysis conducted from March 15, 2018, to January 23, 2019, in 3263 eligible acute care hospitals nationally, risk-adjusted mortality rates observed after Medicare started to impose penalties (October 1, 2012, to November 30, 2014) were compared with projections based on pre-HRRP trends (January 1, 2007, to March 31, 2010) among adults 65 years and older with acute myocardial infarction (AMI), heart failure (HF), or pneumonia. Observed-to-projected differences were then compared between racial groups.ExposuresHospital discharge during pre-HRRP and HRRP penalty periods.Main outcomes and measuresThirty-day postdischarge all-cause mortality.ResultsIn the 3263 acute care hospitals included in the analysis, black patients (627?373 index discharges) were more likely than white patients (5?845?130 index discharges) to be younger (mean [SD] age, 77.8 [8.3] vs 80.5 [8.2] years; P?<?.001), women (60.5% vs 53.7%; P?<?.001), dually covered by Medicare and Medicaid (45.7% vs 17.2%; P?<?.001), and treated at a penalized hospital (AMI, 82.8%; HF, 83.8%; and pneumonia, 82.6% vs 69.6%; 73.3%; and 71.7%; all P?<?.001). Pre-HRRP mortality rates for black vs white patients were 7.04% (95% CI, 6.75% to 7.33%) vs 7.47% (95% CI, 7.37% to 7.57%) for AMI, 6.69% (95% CI, 6.56% to 6.82%) vs 8.56% (95% CI, 8.48% to 8.64%) for HF, and 8.08% (95% CI, 7.88% to 8.27%) vs 8.27% (95% CI, 8.19% to 8.35%) for pneumonia. By the HRRP penalty period, observed mortality for AMI decreased more, relative to projections, among black than white patients (difference-in-differences, -1.65 percentage points; 95% CI, -3.19 to -0.10; P?=?.04). For HF, mortality increased relative to projections among white patients but not among black patients; however; mortality trends did not differ by race (difference-in-differences, -0.37 percentage points; 95% CI, -1.08 to 0.34; P?=?.31). For pneumonia, observed mortality was similar to projections in both racial groups, and trends did not differ by race (difference-in-differences, -0.54 percentage points; 95% CI, -1.66 to 0.59; P?=?.35). At both penalized and nonpenalized hospitals, mortality trends were similar or decreased more among black patients than white patients.Conclusions and relevanceIn this study of patients 65 years and older, short-term postdischarge mortality did not appear to increase for black patients under the HRRP, suggesting that certain value-based payment policies can be implemented without harming black populations. However, mortality seemed to increase for white patients with HF and this situation warrants investigation.

Dataset Information

The effects of data sources, cohort selection, and outcome definition on a predictive model of risk of thirty-day hospital readmissions.

Publications

The effects of data sources, cohort selection, and outcome definition on a predictive model of risk of thirty-day hospital readmissions.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets