Assessing markers from ambulatory laboratory tests for predicting high-risk patients (original) (raw)
Related papers
Measuring population health risks using inpatient diagnoses and outpatient pharmacy data
Health Services Research, 2001
OBJECTIVE: To examine and evaluate models that use inpatient encounter data and outpatient pharmacy claims data to predict future health care expenditures. DATA SOURCES/STUDY DESIGN: The study group was the privately insured under-65 population in the 1997 and 1998 MEDSTAT Market Scan (R) Research Database. Pharmacy and disease profiles, created from pharmacy claims and inpatient encounter data, respectively, were used separately and in combination to predict each individual's subsequent-year health care expenditures. PRINCIPAL FINDINGS: The inpatient-diagnosis model predicts well for the low-hospitalization under-65 populations, explaining 8.4 percent of future individual total cost variation. The pharmacy-based and in patient-diagnosis models perform comparably overall, with pharmacy data better able to split off a group of truly low-cost people and inpatient diagnoses better able to find a small group with extremely high future costs. The model th at uses both kinds of data performed significantly better than either model alone, with an R2 value of 11.8 percent . CONCLUSIONS: Comprehensive pharmacy and inpatient diagnosis classification systems are each helpful for discriminating among people according to their expected costs. Properly organized and in combination these data are promising predictors of future costs.
Prediction of Hospital Mortality Rates by Admission Laboratory Tests
Clinical Chemistry, 2005
Background: The aim of this study was to explore whether electronically retrieved laboratory data can predict mortality in internal medicine departments in a regional hospital. Methods: All 10 308 patients hospitalized in internal medicine departments over a 1-year period were included in the cohort. Nearly all patients had a complete blood count and basic clinical chemistries on admission. We used logistic regression analysis to predict the 573 deaths (5.6%), including all variables that added significantly to the model. Results: Eight laboratory variables and age significantly and independently contributed to a logistic regression model (area under the ROC curve, 88.7%). The odds ratio for the final model per quartile of risk was 6.44 (95% confidence interval, 5.42–7.64), whereas for age alone, the odds ratio per quartile was 2.01 (95% confidence interval, 1.84–2.19). Conclusions: A logistic regression model including only age and electronically retrieved laboratory data highly pr...
The American journal of managed care, 2009
To contrast the advantages and limitations of using medication, diagnostic, and cost data to prospectively identify candidates for care management programs. Risk scores from prior-cost information and a set of clinically based predictive models (PMs) derived from diagnostic and medication data sources, as well as from a combination of all 3 data sources, were assigned to a national sample of commercially insured, non-elderly adults (n = 2,259,584). Clinical relevance of risk groups and statistical performance using future costs as the outcome were contrasted across the PMs. Compared with prior cost, diagnostic and medication-based PMs identified high-risk groups with a higher burden of clinically actionable characteristics. Statistical performance was similar and in some cases better for the clinical PMs compared with prior cost. The best classification accuracy was obtained with a comprehensive model that united diagnostic, medication, and prior-cost risk factors. Clinically based ...
How well does diagnosis-based risk-adjustment work for comparing ambulatory clinical outcomes?
Health Care Management Science, 2009
This paper examines the empirical consistency of the Diagnosis Cost Groups/Hierarchical Condition Categories (DCG/HCC) risk-adjustment method for comparing 7-day mortality between hospital-based outpatient departments (HOPDs) and freestanding ambulatory surgery centers (ASCs). We used patient level data for the three most common outpatient procedures provided during the 1997-2004 period in Florida. We estimated base-line logistic regression models without any diagnosis-based risk adjustment and compared them to logistic regression models with the DCG/HCC risk-adjustment, and to conditional logit models with a matched cohort riskadjustment approach. We also evaluated models that adjusted for primary diagnoses only, and then for all available diagnoses, to assess how the frequently absent secondary diagnoses fields in ambulatory surgical data affect risk-adjustment. We found that risk-adjustment using both diagnosis-based methods resulted in similar 7-day mortality estimates for HOPD patients in comparison with ASC patients in two out of three procedures. We conclude that the DCG/HCC risk-adjustment method is relatively consistent and stable, and recommend this risk-adjustment method for health policy research and practice with ambulatory surgery data. We also recommend using riskadjustment with all available diagnoses.
PLoS ONE, 2013
Background: We explored the use of routine blood tests and national early warning scores (NEWS) reported within 624 hours of admission to predict in-hospital mortality in emergency admissions, using empirical decision Tree models because they are intuitive and may ultimately be used to support clinical decision making. Methodology: A retrospective analysis of adult emergency admissions to a large acute hospital during April 2009 to March 2010 in the West Midlands, England, with a full set of index blood tests results (albumin, creatinine, haemoglobin, potassium, sodium, urea, white cell count and an index NEWS undertaken within 624 hours of admission). We developed a Tree model by randomly splitting the admissions into a training (50%) and validation dataset (50%) and assessed its accuracy using the concordance (c-) statistic. Emergency admissions (about 30%) did not have a full set of index blood tests and/or NEWS and so were not included in our analysis. Results: There were 23248 emergency admissions with a full set of blood tests and NEWS with an in-hospital mortality of 5.69%. The Tree model identified age, NEWS, albumin, sodium, white cell count and urea as significant (p,0.001) predictors of death, which described 17 homogeneous subgroups of admissions with mortality ranging from 0.2% to 60%. The cstatistic for the training model was 0.864 (95%CI 0.852 to 0.87) and when applied to the testing data set this was 0.853 (95%CI 0.840 to 0.866). Conclusions: An easy to interpret validated risk adjustment Tree model using blood test and NEWS taken within 624 hours of admission provides good discrimination and offers a novel approach to risk adjustment which may potentially support clinical decision making. Given the nature of the clinical data, the results are likely to be generalisable but further research is required to investigate this promising approach.
Health Services and Outcomes Research Methodology, 2006
Comparing clinical outcomes in observational studies often requires adjustment for comorbid disease. The objective of this study was to compare the performance of risk adjustment measures derived from different data sources to predict the clinical outcomes of mortality and hospitalization. We compared the predictive ability of self-reported comorbidity measures to those derived from administrative diagnosis codes and pharmacy data to predict all-cause mortality and hospitalizations in a large sample of veterans receiving care in the Veterans Affairs outpatient clinic setting. In logistic regression models to predict mortality adjusting for age and gender, the Seattle Index of Comorbidity, SF-36, Charlson Index, Diagnosis Cost Groups, and RxRisk had similar discriminatory power ranging between 0.73 and 0.74. The Adjusted Clinical Groups and Chronic Illness and Disability Payment System were less accurate in prediction mortality. Although all measures performed less well in predicting hospitalizations, administrative measures performed better than self-reported measures. We conclude that self-reported morbidity measures had similar performance to administrative and pharmacy measures to predict mortality in a larger outpatient sample, but under-performed these measures in predicting hospitalization. While models using self-report measures can typically only be run on subsamples of patients for which models using The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.
QJM : monthly journal of the Association of Physicians, 2016
The Biochemistry and Haematology Outcome Model (BHOM) relies on the results from routine index blood tests to predict the patient risk of death. We aimed to externally validate the BHOM model. We considered all emergency adult medical patients who were discharged from Northern Lincolnshire and Goole (NLAG) hospital in 2014. We compared patient characteristics between NLAG (the validation sample) and the hospital where BHOM was developed. We evaluated the predictive performance, according to discriminative ability (with a concordance statistic, c), and calibration (agreement between observed and predicted risk). There were 29 834 emergency discharges of which 24 696 (83%) had complete data. In comparison with the development sample, the NLAG sample was similar in age, blood test results, but experienced a lower mortality (4.7 vs. 8.7%). When applied to NLAG, the BHOM model had good discrimination (c-statistic 0.83 [95% CI 0.823-0.842]). Calibration was good overall, although the BHOM...
PubMed, 2016
Objectives: With the advent of healthcare payment reform, identifying high-risk populations has become more important to providers. Existing risk-prediction models often focus on chronic conditions. This study sought to better understand other factors to improve identification of the highest risk population. Study design: A retrospective cohort study of a paneled primary care population utilizing 2010 data to calibrate a risk prediction model of hospital and emergency department (ED) use in 2011. Methods: Data were randomly split into development and validation data sets. We compared the enhanced model containing the additional risk predictors with the Minnesota medical tiering model. The study was conducted in the primary care practice of an integrated delivery system at an academic medical center in Rochester, Minnesota. The study focus was primary care medical home patients in 2010 and 2011 (n = 84,752), with the primary outcome of subsequent hospitalization or ED visit. A total of 42,384 individuals derived the enhanced risk-prediction model and 42,368 individuals validated the model. Predictors included Adjusted Clinical Groups-based Minnesota medical tiering, patient demographics, insurance status, and prior year healthcare utilization. Additional variables included specific mental and medical conditions, use of high-risk medications, and body mass index. Results: The area under the curve in the enhanced model was 0.705 (95% CI, 0.698-0.712) compared with 0.662 (95% CI, 0.656-0.669) in the Minnesota medical tiering-only model. New high-risk patients in the enhanced model were more likely to have lack of health insurance, presence of Medicaid, diagnosed depression, and prior ED utilization. Conclusions: An enhanced model including additional healthcare-related factors improved the prediction of risk of hospitalization or ED visit.
Evaluating multivariate risk scores for clinical decision making
Family medicine, 2008
The medical literature abounds with risk scores that can help clinicians predict the probability of disease. Risk scores are popular and attractive because they synthesize the effects of several different risk factors for disease in a way that is otherwise too complex for the human mind alone to analyze. However, to optimize clinical decision making, users of risk scores need to consider the factors and methods used to create the score and to recognize the potential limitations of risk scores. Clinicians should consider the patient populations in which the risk score has been developed and validated, the risk factors included in the score, how computing the score might fit into the flow of daily practice, and how risk scores can help estimate pretest probability. An awareness of the uses and potential limitations of using risk assessment tools will aid the clinician in daily clinical decision making. These tools may grow in importance and use with increasing utilization of electroni...
Medical Care, 2008
To develop a risk-adjustment methodology that maximizes the use of automated physiology and diagnosis data from the time period preceding hospitalization. Design: Retrospective cohort study using split-validation and logistic regression. Setting: Seventeen hospitals in a large integrated health care delivery system. Subjects: Patients (n ϭ 259,699) hospitalized between January 2002 and June 2005. Main Outcome Measures: Inpatient and 30-day mortality. Results: Inpatient mortality was 3.50%; 30-day mortality was 4.06%. We tested logistic regression models in a randomly chosen derivation dataset consisting of 50% of the records and applied their coefficients to the validation dataset. The final model included sex, age, admission type, admission diagnosis, a Laboratory-based Acute Physiology Score (LAPS), and a COmorbidity Point Score (COPS). The LAPS integrates information from 14 laboratory tests obtained in the 24 hours preceding hospitalization into a single continuous variable. Using Diagnostic Cost Groups software, we categorized patients as having up to 40 different comorbidities based on outpatient and inpatient data from the 12 months preceding hospitalization. The COPS integrates information regarding these 41 comorbidities into a single continuous variable. Our best model for inpatient mortality had a c statistic of 0.88 in the validation dataset, whereas the c statistic for 30-day mortality was 0.86; both models had excellent calibration. Physiologic data accounted for a substantial proportion of the model's predictive ability.