PubMed Health⌕ Search

Biomedical subjects

S Lemeshow

Publications and source records attributed to S Lemeshow.

At least 37 records · Page 2Linked to original sources
PubMed1996

Confidence intervals for measures of interaction.

Interaction, defined as departure of disease rates from an additive model, can be measured by the relative excess risk due to interaction, or the attributable proportion due to interaction. Point estimates can be obtained using multiple logistic regression. Using simulated case-control data, we compare several confidence interval estimation techniques for these measures. These include a symmetrical interval based on the delta method estimate of the variance, and three types of bootstrap confidence intervals. One such bootstrap method has coverage closest to the nominal level and is the most evenly balanced with respect to the direction in which intervals miss the true value. The estimation methods are applied to data from an actual case-control study, and the results are interpreted in light of the simulation study.

Case-Control Studies↗
PubMed1996

Factors affecting the performance of the models in the Mortality Probability Model II system and strategies of customization: a simulation study.

OBJECTIVES: To examine the impact of hospital mortality and intensive care unit (ICU) size on the performance of the Mortality Probability Model II system for use in quality assessment, and to examine the ability of model customization to produce accurate estimates of hospital mortality to characterize patients by severity of illness for clinical trials. DESIGN: Prospective evaluation of model performance, using retrospective data. SETTING: Data for the simulation were assembled from six adult medical and surgical ICUs in Massachusetts and New York. PATIENTS: Consecutive admissions (n = 4,224) to the Massachusetts and New York ICUs were studied. The mortality rate in the database was 18.7%. INTERVENTIONS: A computer simulation of several different hospital mortality rates and ICU sample sizes, using a multicenter database of consecutive ICU admissions, was utilized. We simulated 20 different mortality rates by randomly changing the outcomes at hospital discharge from "survived" to "deceased" and from "deceased" to "survived". Four sample size simulations used 75%, 50%, 25%, and 10% of the database. Ten replications of each mortality rate and samples size were constructed, and model calibration and discrimination were assessed for each replication. Model coefficients were customized, using logistic regression. MEASUREMENTS AND MAIN RESULTS: Vital status at hospital discharge was the outcome measure among the ICU patient population. Model performance was assessed using the Hosmer-Lemeshow C statistic for calibration, and the area under the receiver operating characteristic curve for discrimination. Goodness-of-fit tests and receiver operating characteristic curve areas demonstrated that the models were sensitive to differences in hospital mortality, indicating that they are useful quality assurance tools. Goodness-of-fit tests were more sensitive than the receiver operating characteristic curve areas. The further the hospital mortality rate diverged from the original rate, the worse the performance of the model. Sample size had an impact on these results. The smaller the sample size, the less likely the model was to perform poorly. Model coefficients were successfully customized to demonstrate that improved model performance can be achieved when necessary for clinical trial stratification. CONCLUSION: Mortality Probability Model II models can be used to assess quality of care in ICUs, but the size of the sample should be considered when assessing calibration and discrimination.

Adult↗
PubMed1996

Effect of changing patient mix on the performance of an intensive care unit severity-of-illness model: how to distinguish a general from a specialty intensive care unit.

OBJECTIVE: To analyze the effects of patient mix diversity on performance of an intensive care unit (ICU) severity-of-illness model. DESIGN: Multiple patient populations were created using computer simulations. A customized version of the Mortality Probability Model (MPM) II admission model was used to ascertain probabilities of hospital mortality. Performance of the model was assessed using discrimination (area under the receiver operating characteristic curve) and calibration (goodness-of-fit testing). SETTING: Intensive care units. PATIENTS: Data were collected from 4,224 ICU patients from two Massachusetts hospitals (Baystate Medical Center, Springfield, MA; University of Massachusetts Medical Center, Worcester, MA) and two New York hospitals (Albany Medical Center, Albany, NY; Ellis Hospital, Schenectady, NY). INTERVENTIONS: Random samples were taken from a database. The percentage of patients with each model disease characteristic was varied by assigning weights (ranging from 0 to 10) to patients with a disease characteristic. Three simulations were run for each of 15 model variables at each of 16 weights, totaling 720 simulations. MEASUREMENTS AND MAIN RESULTS: The area under the receiver operating characteristic curve and model fit were assessed in each random sample. Removing patients with a given disease characteristic did not affect discrimination or calibration. Increasing frequency of patients with each disease characteristic above the original frequency caused discrimination and calibration to deteriorate. Model fit was more robust to increases in less frequently occurring patient conditions. From the goodness-of-fit test, a critical percentage for each admission model variable was determined for each disease characteristic, defined as the percentage at which the average p value for the test over the three replications decreased to < .10. CONCLUSIONS: The concept of critical percentages is potentially clinically important. It might provide an easy first step in checking applicability of a given severity-of-illness model and in defining a general medical-surgical ICU. If the critical percentages are exceeded, as might occur in a highly specialized ICU, the model would not be accurate. Alternative modeling approaches might be to customize the model coefficients to the population for more accurate probabilities or to develop specialized models. The MPM approach remained robust for a large variation in patient mix factors.

Aged↗
PubMed1996

Resource use implications of do not resuscitate orders for intensive care unit patients.

This study describes the use of do not resuscitate (DNR) orders for ICU patients in four northeastern U.S. teaching hospitals and investigates the relationship between DNR orders and length of stay. The use of detailed data from the mortality probability model (MPM II) study on 6,290 consecutive ICU admissions to general adult medical and surgical ICUs during 1989 through 1991 allows us to control for severity of illness and the time during the ICU stay at which the DNR order was entered. About 12.8% of patients were DNR during their ICU stay, including more than half of nonsurvivors. The percentage of patients with DNR was higher for older and more severely ill patients. Most DNR orders were issued after 72 h in the ICU, but many were issued during the first ICU day. Nonsurvivors with early (first 24 h) DNR had shorter mean and median ICU and hospital stays than the comparison group of non-DNR patients. The percentage of patients with very long ICU (> 30 d) and hospital (> 60 d) stays was smaller among DNR patients. The use of DNR orders, particularly early in the ICU stay, may be associated with significant resource use reduction for an identifiable group of patients.

Adult↗
PubMed1995

Confidence interval estimates of an index of quality performance based on logistic regression models.

This paper considers an index of hospital quality performance defined as the ratio of the observed number deaths to the number predicted by a fitted logistic regression model. We study tests and confidence intervals under two different scenarios depending on the availability of an estimate of the covariance matrix of the coefficients from the fitted logistic regression model. We propose parametric as well as bootstrap-based confidence intervals. We apply the methods to an analysis of the performance of 27 intensive care units.

Aged↗
PubMed1995

Customized probability models for early severe sepsis in adult intensive care patients. Intensive Care Unit Scoring Group.

OBJECTIVE: To develop customized versions of the Simplified Acute Physiology Score II (SAPS II) and the 24-hour Mortality Probability Model II (MPM II) to estimate the probability of mortality for intensive care unit patients with early severe sepsis. DESIGN AND SETTING: Logistic regression models developed for patients with severe sepsis in a database of adult medical and surgical intensive care units in 12 countries. PATIENTS: Of 11,458 patients in the intensive care unit for at least 24 hours, 1130 had severe sepsis based on criteria of the American College of Chest Physicians and the Society of Critical Care Medicine (systemic inflammatory response syndrome in response to infection, plus hypotension, hypoperfusion, or multiple organ dysfunction). RESULTS: In patients with severe sepsis, mortality was higher (48.0% vs 19.6% among other patients) and 28-day survival was lower. The customized SAPS II was well calibrated (P = .92 for the goodness-of-fit test) and discriminated well (area under the receiver operating characteristic [ROC] curve, 0.78). Performance in the validation sample was equally good (P = .85 for the goodness-of-fit test; area under the ROC curve, 0.79). The customized MPM II was well calibrated (P = .92 for the goodness-of-fit test) and discriminated well (area under the ROC curve, 0.79). Performance in the validation sample was equally good (P = .52 for the goodness-of-fit test; area under the ROC curve, 0.75). The models are independent of each other; either can be used alone to estimate the probability of mortality of patients with severe sepsis. CONCLUSIONS: Customization provides a simple technique to apply existing models to a subgroup of patients. Accurately assessing the probability of hospital mortality is a useful adjunct for clinical trials.

Adult↗
PubMed1995

Outcome prediction for individual intensive care patients: useful, misused, or abused?

Probabilities of hospital mortality provide meaningful information in many contexts, such as in discussions of patient prognosis by intensive care physicians, in patient stratification for analysis of clinical trial data by researchers, and in hospital reimbursement analysis by insurers. Use of probabilities as binary predictors based on a cut point can be misleading for making treatment decisions for individual patients, however, even when model performance is good overall. Alternative models for estimating severity of illness in intensive care unit (ICU) patients, while demonstrating good agreement for describing patients in the aggregate, are shown to differ considerably for individual patients. This suggests that identifying patients unlikely to benefit from ICU care by using models must be approached with considerable caution.

Bias↗
PubMed1995

A comparison of intensive care unit utilization in Alberta and western Massachusetts.

OBJECTIVE: To analyze differences in intensive care unit (ICU) utilization between a Canadian province and a U.S. area. DESIGN: Retrospective data analysis of hospital discharge data and existing data from an international study of severity of illness in ICU patients. SETTING: Administrative data for the province of Alberta and the four counties of western Massachusetts for the years 1990 to 1991 were used. Detailed data on consecutive ICU admissions from two Alberta hospitals, one western Massachusetts hospital, and 24 other U.S. hospitals for 3 months in 1991 were used. MEASUREMENTS AND MAIN RESULTS: ICU use and hospital mortality rates were compared for 50,030 hospital admissions divided into 11 patient groups. ICU days per million population were two to three times as great in western Massachusetts as in Alberta. The primary reason was higher ICU incidence (percent of hospitalized patients treated in the ICU) rather than a difference in hospital admission rate or length of ICU stay. ICU incidence in western Massachusetts was significantly higher in ten of 11 patient groups--for the coronary bypass surgery group, there was no difference. The hospital mortality rate in western Massachusetts was similar to, or higher than, the mortality rate in Alberta. In Alberta, a much higher proportion of ICU patients received mechanical ventilation. For elective surgery patients, the ICU severity of illness was lower in western Massachusetts and in other U.S. hospitals than in Alberta. CONCLUSIONS: Western Massachusetts hospitalized patients are more likely to be treated in an ICU than are similar patients in Alberta. There is no evidence that the greater ICU utilization in western Massachusetts led to a lower hospital mortality rate.

Aged↗
PubMed1994

Modeling the severity of illness of ICU patients. A systems update.

OBJECTIVE: To review recent revisions of systems for estimating the probability of hospital mortality of adult intensive care unit (ICU) patients. Emphasis on comparison of components of systems and potential uses. DATA SOURCES: Published articles in which the systems were presented. STUDY SELECTION: Acute Physiology and Chronic Health Evaluation (APACHE III), Simplified Acute Physiology Score (SAPS II), and Mortality Probability Models (MPM II) are the major severity systems for ICU patients. DATA EXTRACTION: Information on variables collected in the systems, characteristics of databases from which they were developed, and reported performance of models were evaluated from published articles. DATA SYNTHESIS: APACHE III and SAPS II produce a score and probability of hospital mortality based on worst values of several variables during the first 24 hours in ICU. The MPM II system has four models, one at ICU admission and one at 24, 48, and 72 hours into the ICU stay. The SAPS II and MPM II models can be implemented from published information. The APACHE III score can be calculated from published information; weights to convert score to probability are proprietary. All reported good areas under receiver operating characteristic curve. Goodness of fit was good for SAPS II and MPM II models and was not reported for APACHE III models. CONCLUSIONS: All models were based on rigorous research and reported performance is good. All can be used to assist in assessing prognosis, to compare ICU performance, and to stratify patients for clinical trials. Direct comparison on a common cohort is needed.

Adult↗
PubMed1994

A retrospective-cohort study of occupational exposure to hexavalent chromium.

A retrospective cohort study was conducted to examine the risk of mortality, cancer, and other adverse health outcomes, at the United States' largest chromate chemicals manufacturing facility in Castle Hayne, North Carolina. This facility, built in 1971, was designed to reduce the high levels of chromium exposure found at most older facilities. Exposure assessment was based on analysis of more than 5,000 personal breathing zone samples collected over a 15-year period. A questionnaire was used to collect relevant occupational, medical, smoking, and other information from current and former employees. Analysis of the cohort's mortality experience found no substantial departures from that expected based on external comparisons, although evidence of a healthy worker effect was observed. Internal cohort analyses were limited by relatively small numbers; however, a subgroup of employees who transferred from older facilities was found to have higher risks of mortality (odds ratio = 1.27 for each 3 years of previous exposure; 90% confidence interval (CI) = 1.07-1.51) and cancer (odds ratio = 1.22 for each 3 years of previous exposure; 90% CI = 1.03-1.45). While this subgroup represented only 11% of the individuals in this study, they accounted for 46% (6/13) of all observed cancers (excluding skin cancers) and 60% (3/5) of lung cancers. There was no increased risk of mortality or cancer among employees who worked only at the newer facility. As an etiologic research study, the results are limited by the relatively small number of subjects and short follow-up; nevertheless, the findings can be used to design and implement a prospective surveillance system for monitoring the health of chromate production workers.

Carcinogens, Environmental↗
PubMed1994

Mortality probability models for patients in the intensive care unit for 48 or 72 hours: a prospective, multicenter study.

OBJECTIVE: To develop models in the Mortality Probability Model (MPM II) system to estimate the probability of hospital mortality at 48 and 72 hrs in the intensive care unit (ICU), and to test whether the 24-hr Mortality Probability Model (MPM24), developed for use at 24 hrs in the ICU, can be used on a daily basis beyond 24 hrs. DESIGN: A prospective, multicenter study to develop and validate models, using a cohort of consecutive admissions. SETTING: Six adult medical and surgical ICUs in Massachusetts and New York adjusted to reflect 137 ICUs in 12 countries. PATIENTS: Consecutive admissions (n = 6,290) to the Massachusetts/New York ICUs were studied. Of these patients, 3,023 and 2,233 patients remained in the ICU and had complete data at 48 and 72 hrs, respectively. Patients < 18 yrs of age, burn patients, coronary care patients, and cardiac surgical patients were excluded. OUTCOME MEASURE: Vital status at the time of hospital discharge. RESULTS: The models consist of five variables measured at the time of ICU admission and eight variables ascertained at 24-hr intervals. The 24-hr model demonstrated poor calibration and discrimination at 48 and 72 hrs. The newly developed 48- and 72-hr models--MPM48 and MPM72--contain the same 13 variables and coefficients as the MPM24. The models differ only in their constant terms, which increase in a manner that reflects the increasing probability of mortality with increasing length of stay in the ICU. These constant terms were adjusted by a factor determined from the relationship between the data from the six Massachusetts and New York ICUs and a more extensive data set, from which the ICU admission Mortality Probability Model (MPM0) and MPM24 were developed. This latter data set was assembled from ICUs in 12 countries. The MPM48 and MPM72 calibrated and discriminated well, based on goodness-of-fit tests and area under the receiver operating characteristic curve. CONCLUSIONS: Models developed for use among ICU patients at one time period are not transferable without modification to other time periods. The MPM48 and MPM72 calibrated well to their respective time periods, and they are intended for use at specific points in time. The increasing constant terms and associated increase in the probability of hospital mortality exemplify a common clinical adage that if a patient's clinical profile stays the same, he or she is actually getting worse.

Adult↗
PubMed1994

A method for assessing the clinical performance and cost-effectiveness of intensive care units: a multicenter inception cohort study.

OBJECTIVES: To present an approach for assessing intensive care unit (ICU) performance which takes into account both economic and clinical performance while adjusting for severity of illness. To present a graphic display which permits comparisons among a group of hospitals. DESIGN: A multicenter, inception cohort study. SETTING: Twenty-five ICUs in U.S. hospitals that participated in the European and North American Study of Severity Systems for ICU Patients. PATIENTS: Consecutive patients (n = 3,397) admitted to ICUs in participating hospitals between September 30, 1991 and December 27, 1991. Excluded were coronary care patients, burn patients, cardiac surgery patients and patients aged < 18 yrs. MEASUREMENTS AND MAIN RESULTS: The clinical performance index is the difference between observed hospital survival rate and survival rate predicted by the Mortality Probability Model measuring severity of illness at ICU admission. The economic performance (resource use) measure is a length of stay index, Weighted Hospital Days, which weights ICU days more heavily than non-ICU days. The economic performance index is the difference between actual mean resource use and the resource use predicted by a regression including severity of illness and percent of surgical patients. Both the clinical and economic performance indices are standardized to show how far a particular hospital is from the overall mean and are graphed together. Most of the 25 hospitals lie within 1 SD of the mean on both clinical and economic performance scales. The graph makes it easy to identify those hospitals that are outside this range. There is no evidence of a trade-off between high clinical performance and high economic performance; i.e., it is possible to achieve both. CONCLUSIONS: Cross-indexing of clinical and economic ICU performance is easy to calculate. It has potential as a research and evaluation tool used by physicians, hospital administrators, payers, and others.

Adult↗
PubMed1994

Why severity models should be used with caution.

There are now two validated time points for predicting hospital mortality of ICU patients--at admission and at 24 hours. The best purposes include evaluation of high clinical performance ICUs and for patients being enrolled in clinical trials. For the latter purpose, the model must be calibrated in the individual hospital to ensure that the model is applicable. This can be estimated by using goodness-of-fit testing. There are fewer uses for physiology scores and increased emphasis on converting scores to probabilities. For individual patient application, the model should be demonstrated to have high discrimination, as measured by the area under the receiver operating characteristic curve, and high calibration, as defined by goodness-of-fit testing. Although models have improved substantially and are now based on much larger databases, there is considerable uncertainty in their application for insurance purposes, triage, regulatory applications, sanctions against individual physicians, and cost containment. Current models may not adequately describe important ICU conditions such as adult respiratory distress syndrome and multi-organ dysfunction occurring after 24 hours into ICU care. For family discussions regarding prognosis of individual patients, ICU severity models must be used cautiously at admission or after 24 hours, with the understanding of the strengths and weakness of estimating probabilities of hospital mortality. The mathematical link between physiology score and estimation of hospital mortality is established only for the time point of 24 hours after ICU admission. Calibration and discrimination of the admission and 24-hour models also must be performed within each hospital in which individual probabilities are presented to families. It may be possible to customize a probability model such as MPM to achieve a high level of calibration at the individual hospital level.

Bias↗
PubMed1993

Mortality Probability Models (MPM II) based on an international cohort of intensive care unit patients.

OBJECTIVE: To revise and update models in the Mortality Probability Model (MPM II) system to estimate the probability of hospital mortality among 19,124 intensive care unit (ICU) patients that can be used for quality assessment within and among ICUs. DESIGN AND SETTING: Models developed and validated on consecutive admissions to adult medical and surgical ICUs in 12 countries. PATIENTS: A total of 12,610 patients for model development, 6514 patients for model validation. Patients younger than 18 years and burn, coronary care, and cardiac surgery patients were excluded. OUTCOME MEASURE: Vital status at hospital discharge. RESULTS: The admission model, MPM0, contains 15 readily obtainable variables. In developmental and validation samples it calibrated well (goodness-of-fit tests: P = .623 and P = .327, respectively, where a high P value represents good fit between observed and expected values) and discriminated well (area under the receiver operating characteristic curve = 0.837 and 0.824, respectively). The 24-hour model, MPM24 (developed on 10,357 patients still in the ICU at 24 hours), contains five of the admission variables and eight additional variables easily ascertained at 24 hours. It also calibrated well (P = .764 and P = .231 in the developmental and validation samples, respectively) and discriminated well (area under the receiver operating characteristic curve = 0.844 and 0.836 in the developmental and validation samples, respectively). CONCLUSIONS: Among severity systems for intensive care patients, the MPM0 is the only model available for use at ICU admission. Both MPM0 and MPM24 are useful research tools and provide important clinical information when used alone or together.

Adult↗
PubMed1993

Using severity measures to describe high performance intensive care units.

This article describes the use of various scores and probabilities to clinically categorize patients in the adult intensive care unit. Some of the limitations of these severity measures are reviewed including variable definitions, timing of measurements, and whether models can be used for individual patients. Also, this article discusses how probability models may be used to compare similar types of intensive care units using standardized clinical and cost performance indices.

Hospital Mortality↗
PubMed1992

Confidence interval estimation of interaction.

Relative excess risk due to interaction, the proportion of disease among those with both exposures that is attributable to their interaction, and the synergy index have been proposed as measures of interaction in epidemiologic studies. This paper presents the methodology for obtaining confidence interval estimates of these indices utilizing routinely available output from multiple logistic regression software.

Confidence Intervals↗
PubMed1992

Resource utilization among intensive care patients. Managed care vs traditional insurance.

BACKGROUND: There is considerable evidence that members of managed care organizations use fewer hospital resources than patients covered by traditional health insurance. While intensive care might seem to be an unlikely setting for such differences to exist, the relationship between health coverage and use of intensive care has not been examined. METHODS: We conducted a cross-sectional analysis of consecutive intensive care unit admissions at a regional tertiary care teaching hospital. Patients in managed care plans (n = 159) and with traditional insurance (n = 389) were compared with respect to length of stay, hospital charges, charges for specific services, and use of mechanical ventilation. The analysis controlled for severity of illness, as measured by the Mortality Probability Model, case mix, and mortality. The whole sample as well as subsamples representing medical, emergency surgery, and elective surgery patients were examined. RESULTS: The managed care group, on average, had short stays (both hospital and intensive care unit), lower charges, and less use of mechanical ventilation than the traditionally insured group. Average differences of about 30% to 40% were observed. The finding held for the whole sample as well as the medical and emergency surgery subsamples. The differences were more pronounced in the patients with lowest severity of illness. CONCLUSION: Even in a setting where there would appear to be relatively little room for discretion in treatment decisions, incentives associated with type of health insurance seemed to affect resource use.

Cost Control↗