Dataset Information

Using highly detailed administrative data to predict pneumonia mortality.

ABSTRACT:

Background

Mortality prediction models generally require clinical data or are derived from information coded at discharge, limiting adjustment for presenting severity of illness in observational studies using administrative data.

Objectives

To develop and validate a mortality prediction model using administrative data available in the first 2 hospital days.

Research design

After dividing the dataset into derivation and validation sets, we created a hierarchical generalized linear mortality model that included patient demographics, comorbidities, medications, therapies, and diagnostic tests administered in the first 2 hospital days. We then applied the model to the validation set.

Subjects

Patients aged ? 18 years admitted with pneumonia between July 2007 and June 2010 to 347 hospitals in Premier, Inc.'s Perspective database.

Measures

In hospital mortality.

Results

The derivation cohort included 200,870 patients and the validation cohort had 50,037. Mortality was 7.2%. In the multivariable model, 3 demographic factors, 25 comorbidities, 41 medications, 7 diagnostic tests, and 9 treatments were associated with mortality. Factors that were most strongly associated with mortality included receipt of vasopressors, non-invasive ventilation, and bicarbonate. The model had a c-statistic of 0.85 in both cohorts. In the validation cohort, deciles of predicted risk ranged from 0.3% to 34.3% with observed risk over the same deciles from 0.1% to 33.7%.

Conclusions

A mortality model based on detailed administrative data available in the first 2 hospital days had good discrimination and calibration. The model compares favorably to clinically based prediction models and may be useful in observational studies when clinical data are not available.

SUBMITTER: Rothberg MB

PROVIDER: S-EPMC3909106 | biostudies-literature | 2014

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Using highly detailed administrative data to predict pneumonia mortality.

Rothberg Michael B MB Pekow Penelope S PS Priya Aruna A Zilberberg Marya D MD Belforti Raquel R Skiest Daniel D Lagu Tara T Higgins Thomas L TL Lindenauer Peter K PK

PloS one 20140131 1

<h4>Background</h4>Mortality prediction models generally require clinical data or are derived from information coded at discharge, limiting adjustment for presenting severity of illness in observational studies using administrative data.<h4>Objectives</h4>To develop and validate a mortality prediction model using administrative data available in the first 2 hospital days.<h4>Research design</h4>After dividing the dataset into derivation and validation sets, we created a hierarchical generalized ...[more]

PMID: 24498090

Similar Datasets

Project description:BackgroundPneumonia has traditionally been classified into two subtypes: community-acquired pneumonia (CAP) and nosocomial pneumonia (NP). Recently, a new entity has been defined, called healthcare-associated pneumonia (HCAP). Few studies have investigated the potential of population-based, electronic, healthcare databases to identify the incidences of these three subtypes of pneumonia. The aim of this study was to estimate the burden of the three subtypes of pneumonia in elderly patients (aged 65+ years) in a large region of central Italy.MethodsA retrospective cohort study was performed using linked regional Hospital Information System and Mortality Register. All episodes of pneumonia in elderly patients, who were discharged from the hospital in 2006-2008, were selected for the study. Following a validated ICD-9-coding algorithm, incidents of pneumonia events were classified into three groups (HCAP; probable nosocomial pneumonia, PNP; and CAP). Hospitalisation rates were calculated by age group (65-79, 80+), gender, and year, using the population from the Institute of Statistics (ISTAT) census estimates as denominators.ResultsA total of 26,239 pneumonia events occurred in 24,338 patients residing in the Lazio region, aged 65+ years: 2257 HCAP, 6775 PNP, and 17,107 CAP. For all subtypes, the proportion of males was greater than females. Comorbidity status was more severe in HCAP than in the other categories. In-hospital mortality, 30-day mortality, and length of hospital stay were twice higher in HCAP than in CAP episodes. The annual incidence rates were 0.7, 2.1, and 5.4 episodes per 1000 residents for HCAP, PNP, and CAP, respectively. From 2006 to 2008, incidence rates slightly increased for all three subtypes.ConclusionHealth care databases can be used to give a timely and inexpensive picture of the epidemiology of pneumonia. HCAP represents a distinct category of pneumonia, with the longest stay, highest mortality, and the greatest comorbidity.

Project description:BackgroundOutcomes for coronary artery bypass surgery are of broadening interest, but the impact of data type on quality reporting has not been fully examined. We compared the performance of administrative and clinical data-based risk adjustment models at a tertiary-quaternary care hospital.MethodsWe used a prospective study design to test two risk adjustment models, one from administrative (Canadian Institute for Health Information [CIHI] Cardiac Care Quality Indicator) and one from clinical data (Society of Thoracic Surgeons), on cardiac surgical procedures performed between 2013 and 2016 (n = 1635). Our primary outcome was in-hospital mortality within 30 days of surgery. Model performance was established by comparing predicted and observed mortality, model calibration and handling of critical covariates.ResultsObserved mortality was 1.96%, which was the same as that predicted by the Society of Thoracic Surgeons model (1.96%), but significantly higher than that predicted by the CIHI model (1.03%). Despite both models having similar C statistics (0.756 CIHI; 0.758 Society of Thoracic Surgeons), the CIHI model showed significant underestimation of mortality among patients at higher risk. There was significant miscalibration of risk associated with 7 covariates: New York Heart Association class IV, congestive heart failure, ejection fraction less than 20%, atrial fibrillation, acute coronary insufficiency, cardiac compromise (shock, myocardial infarction < 24 h, intra-aortic balloon pump, cardiac resuscitation or preprocedure circulatory support) and creatinine concentration of 100 mg/dL or more. Together, these factors accounted for 84% of the difference in predicted mortality between the administrative and clinical models.InterpretationRisk prediction using administrative data underestimated risk of death, potentially inflating observed-to-predicted mortality ratios at hospitals with patients who are more ill. Caution is warranted when hospital reports of cardiac surgery outcomes are based on administrative data alone.

Dataset Information

Using highly detailed administrative data to predict pneumonia mortality.

Background

Objectives

Research design

Subjects

Measures

Results

Conclusions

Publications

Using highly detailed administrative data to predict pneumonia mortality.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets