[show abstract][hide abstract] ABSTRACT: The objective of this study is to investigate the relationship between a physician's subjective mortality prediction and the level of confidence with which that mortality prediction is made.
The study is a prospective cohort of patients less than 18 years of age admitted to a tertiary Paediatric Intensive Care Unit (ICU) at a University Children's Hospital with a minimum length of ICU stay of 10 h. Paediatric ICU attending physicians and fellows provided mortality risk predictions and the level of confidence associated with these predictions on consecutive patients at the time of multidisciplinary rounds within 24 hours of admission to the paediatric ICU. Median confidence levels were compared across different ranges of mortality risk predictions.
Data were collected on 642 of 713 eligible patients (36 deaths, 5.6%). Mortality predictions greater than 5% and less than 95% were made with significantly less confidence than those predictions <5% and >95%. Experience was associated with greater confidence in prognostication.
We conclude that a physician's subjective mortality prediction may be dependent on the level of confidence in the prognosis; that is, a physician less confident in his or her prognosis is more likely to state an intermediate survival prediction. Measuring the level of confidence associated with mortality risk predictions (or any prognostic assessment) may therefore be important because different levels of confidence may translate into differences in a physician's therapeutic plans and their assessment of the patient's future.
Journal of Medical Ethics 06/2004; 30(3):304-7. · 1.42 Impact Factor
[show abstract][hide abstract] ABSTRACT: continuous set of non-singular 3-D affine transforms, which are the most general linear transformations available, and where by continuity we imply sub-pixel accuracy; our dissimilarity measure is Euclidean, which is maximum-likelihood assuming additive white Gaussian noise; finally, our search strategy is multi-scale, for fast convergence, and iterative, based on a variation of the MarquardtLevenberg (ML) algorithm for non-linear least-square optimizations .
[show abstract][hide abstract] ABSTRACT: Length of stay in the pediatric intensive care unit (PICU) is a reflection of patient severity of illness and health status, as well as PICU quality and performance. We determined the clinical profiles and relative resource use of long-stay patients (LSPs) and developed a prediction model to identify LSPs for early quality and cost saving interventions.
Nonconcurrent cohort study.
A total of 16 randomly selected PICUs and 16 volunteer PICUs.
A total of 11,165 consecutive admissions to the 32 PICUs.
LSPs were defined as patients having a length of stay greater than the 95th percentile (>12 days). Logistic regression analysis was used to determine which clinical characteristics, available within the first 24 hrs after admission, were associated with LSPs and to create a predictive algorithm. Overall, LSPs were 4.7% of the population but represented 36.1% of the days of care. Multivariate analysis indicated that the following factors are predictive of long stays: age <12 months, previous ICU admission, emergency admission, no CPR before admission, admission from another ICU or intermediate care unit, chronic care requirements (total parenteral nutrition and tracheostomy), specific diagnoses including acquired cardiac disease, pneumonia, and other respiratory disorders, having never been discharged from the hospital, need for ventilatory support or an intracranial catheter, and a Pediatric Risk of Mortality III score between 10 and 33. The performance of the prediction algorithm in both the training and validation samples for identifying LSPs was good for both discrimination (area under the receiver operating characteristics curve of 0.83 and 0.85, respectively), and calibration (goodness of fit, p = .33 and p = .16, respectively). LSPs comprised from 2.1% to 8.1% of individual ICU patients and occupied from 15.2% to 57.8% of individual ICU bed days.
LSPs have less favorable outcomes and use more resources than non-LSPs. The clinical profile of LSPs includes those who are younger and those that require chronic care devices. A predictive algorithm could help identify patients at high risk of prolonged stays appropriate for specific interventions.
Critical Care Medicine 04/2001; 29(3):652-7. · 6.12 Impact Factor
[show abstract][hide abstract] ABSTRACT: The use of the wavelet transform is explored for the detection of differences between brain functional magnetic resonance images (fMRI's) acquired under two different experimental conditions. The method benefits from the fact that a smooth and spatially localized signal can be represented by a small set of localized wavelet coefficients, while the power of white noise is uniformly spread throughout the wavelet space. Hence, a statistical procedure is developed that uses the imposed decomposition orthogonality to locate wavelet-space partitions with large signal-to-noise ratio (SNR), and subsequently restricts the testing for significant wavelet coefficients to these partitions. This results in a higher SNR and a smaller number of statistical tests, yielding a lower detection threshold compared to spatial-domain testing and, thus, a higher detection sensitivity without increasing type I errors. The multiresolution approach of the wavelet method is particularly suited to applications whe...
[show abstract][hide abstract] ABSTRACT: Investigation of associations of the diagnostic diversity and volumes with efficiency and quality of care.
Prospective observational study.
Thirty-two pediatric intensive care units (PICUs), 16 selected by random cluster sampling, and 16 volunteering.
Consecutive admissions of 11,165 patients.
The main outcome measures were length of PICU stay (LOS) and mortality rate, adjusted by generalized linear regression and multivariate logistic regression, respectively. Each diagnosis was categorized into 21 predefined, mutually exclusive categories. Diagnostic diversity of each PICU was characterized by an information-theoretical measure (entropy). For a patient-level analysis, the associations of this measure and PICU patient volume with outcomes were using regression models. For an institution-level analysis, the outcome measures of each PICU were adjusted using ratios of observed/predicted (by the regression models) values, and the associations of these ratios with diagnostic diversity and patient volume were investigated using linear bivariate regressions. Diagnostic diversity ranged in the PICUs from 0.823 to 0.928, when standardized to the uniform distribution with entropy of 1. Congenital heart diseases (12.6%) head traumas (11.5%), other central nervous system conditions (9.7%), and pneumonias (8.7%) constituted the largest diagnostic categories. Patient-level analysis indicated that longer adjusted LOS was associated with larger diagnostic diversity (p <.0001) and lower admission volumes (p <.0001). However, for a given increase in diagnostic diversity, a large LOS increase was associated with low-volume, but not high-volume units. Severity-adjusted mortality rates were inversely related (p =.036) only with admission volumes, but not diagnostic mix. Institution-level standardized LOS ratios correlated with diagnostic diversity (r2 = 0.145; p =.031). Institution-level standardized mortality ratios were inversely related (r2 = 0.123; p =.049) with admission volumes.
Patient volumes encountered in a PICU are important for maintaining quality and efficiency of care. In low-volume units, fewer diagnoses and higher volumes were both associated with higher efficiencies. In high volume units, diagnosis-specific volumes were generally large enough for achieving diagnosis-independent efficiency. Diagnostic mix was not associated with PICU mortality ratios, but higher PICU volumes were associated with lower mortality rates.
Pediatric Critical Care Medicine 10/2000; 1(2):133-9. · 2.35 Impact Factor
[show abstract][hide abstract] ABSTRACT: None of the currently available physiology-based mortality risk prediction models incorporate subjective judgements of healthcare professionals, a source of additional information that could improve predictor performance and make such systems more acceptable to healthcare professionals. This study compared the performance of subjective mortality estimates by physicians and nurses with a physiology-based method, the Pediatric Risk of Mortality (PRISM) III. Then, healthcare provider estimates were combined with PRISM III estimates using Bayesian statistics. The performance of the Bayesian model was then compared with the original two predictions.
Concurrent cohort study.
A tertiary pediatric intensive care unit at a university affiliated children's hospital.
Consecutive admissions to the pediatric intensive care unit.
For each of the 642 consecutive eligible patients, an exact mortality estimate and the degree of certainty (continuous scale from 1 to 5) associated with the estimate was collected from the attending, fellow, resident, and nurse responsible for the patient's care. Bayesian statistics were used to combine the PRISM III and certainty weighted subjective predictions to create a third Bayesian estimate of mortality. PRISM III discriminated survivors from nonsurvivors very well (area under curve [AUC], 0.924) as did the physicians and nurses (AUCs attendings, 0.953; fellows, 0.870; residents, 0.923; nurses, 0.935). Although the AUCs of the healthcare providers were not significantly different from the AUCs of PRISM III, the Bayesian AUCs were higher than both the healthcare providers' AUCs (p < or = .09 for all) and PRISM III AUCs. Similarly, the calibration statistics for the Bayesian estimates were superior to the calibration statistics for both the healthcare providers and PRISM III models.
The results of this study demonstrated that healthcare providers' subjective mortality predictions and PRISM III mortality predictions perform equally well. The Bayesian model that combined provider and PRISM III mortality predictions was more accurate than either provider or PRISM III alone and may be more acceptable to physicians. A methodology using subjective outcome predictions could be more relevant to individual patient decision support.
Critical Care Medicine 08/2000; 28(8):2984-90. · 6.12 Impact Factor
[show abstract][hide abstract] ABSTRACT: To evaluate the relative resource use of pediatric intensive care unit (PICU) patients who had been born prematurely.
Nonconcurrent cohort study.
Consecutive admissions to 16 voluntary PICUs.
A total of 431 formerly premature patients (FPP) and 5,319 nonpremature patients.
Patients with a history of prematurity and a prematurity-related complication or an anatomical deformity were compared for demographic and resource requirements to a group of non-premature patients by a bivariable logistic regression analysis that controlled for age as a co-morbid factor.
Compared with other patients, FPP were younger (34.9 +/- 2.2 months vs. 72.4 +/- 1.0 months; p < .001), readmitted to the PICU more often during the same hospitalization (11.1% vs. 5.5%; p < .001), used more chronic technologies (ventilators, gastrostomy tubes, tracheostomy tubes, and parenteral nutrition; 30.3% vs. 5.6%; p < .001), and had longer lengths of stay (5.98 +/-0.59 days vs. 3.56 +/- 0.12 days; p = .004). FPP had significantly higher use of ventilators (45.5% vs. 35.0%; p < .007) and lower use of arterial catheters (27.8% vs. 35.9%, p = .006) and central venous catheters (16.9% vs. 20.9%, p = .026) than nonprematures. The need for other PICU resources, including vasopressors, were similar.
FPP used more chronic and acute care resources than patients who were not prematurely born. Continued improvements in neonatal care will influence change in many aspects of the health care system. This will also affect the delivery of care to the current patient base of the PICU.
Critical Care Medicine 03/2000; 28(3):848-53. · 6.12 Impact Factor
[show abstract][hide abstract] ABSTRACT: Prognostication is central to developing treatment plans and relaying information to patients, family members, and other health care providers. The degree of confidence or certainty that a health care provider has in his or her mortality risk assessment is also important, because a provider may deliver care differently depending on their assuredness in the assessment. We assessed the performance of nurse and physician mortality risk estimates with and without weighting the estimates with their respective degrees of certainty.
Subjective mortality risk estimates from critical care attendings (n = 5), critical care fellows (n = 9), pediatric residents (n = 34), and nurses (n = 52) were prospectively collected on at least 94% of 642 eligible, consecutive admissions to a tertiary pediatric intensive care unit (PICU). A measure of certainty (continuous scale from 0 to 5) accompanied each mortality estimate. Estimates were evaluated with 2 x 2 outcome probabilities, the kappa statistic, the area under the receiver operating characteristics curve, and the Hosmer and Lemeshow goodness-of-fit chi(2) statistic. The estimates were then reevaluated after weighting predictions by their respective degree of certainty.
Overall, there was a significant difference in the predictive accuracy between groups. The mean mortality predictions from the attendings (6.09%) more closely approximated the true mortality rate (36 deaths, 5.61%) whereas fellows (7.87%), residents (10.00%), and nurses (16.29%) overestimated the mean overall PICU mortality. Attendings were more certain of their predictions (4.27) than the fellows (4.01), nurses (3.79), and residents (3.75). All groups discriminated well (area under receiver operating characteristics curve range, 0.86-0.93). Only PICU attendings and fellows did not significantly differ from ideal calibration (chi(2)). When mortality predictions were weighted with their respective certainties, their performance improved.
The level of medical training correlated with the provider's ability to predict mortality risk. The higher the level of certainty associated with the mortality prediction, the more accurate the prediction; however, high levels of certainty did not guarantee accurate predictions. Measures of certainty should be considered when assessing the performance of mortality risk estimates or other subjective outcome predictions.
[show abstract][hide abstract] ABSTRACT: Low frequency drift (0.0-0.015 Hz) has often been reported in time series fMRI data. This drift has often been attributed to physiological noise or subject motion, but no studies have been done to test this assumption. Time series T*2-weighted volumes were acquired on two clinical 1.5 T MRI systems using spiral and EPI readout gradients from cadavers, a normal volunteer, and nonhomogeneous and homogeneous phantoms. The data were tested for significant differences (P = 0.001) from Gaussian noise in the frequency range 0.0-0.015 Hz. The percentage of voxels that were significant in data from the cadaver, normal volunteer, nonhomogeneous and homogeneous phantoms were 13.7-49.0%, 22.1-61.9%, 46.4-68.0%, and 1.10%, respectively. Low frequency drift was more pronounced in regions with high spatial intensity gradients. Significant drifting was present in data acquired from cadavers and nonhomogeneous phantoms and all pulse sequences tested, implying that scanner instabilities and not motion or physiological noise may be the major cause of the drift.
[show abstract][hide abstract] ABSTRACT: As physiology based assessments of mortality risk become more accurate, their potential utility in clinical decision support and resource rationing decisions increases. Before these prediction models can be used, however, their performance must be statistically evaluated and interpreted in a clinical context. We examine the issues of confidence intervals (as estimates of survival ranges) and confidence levels (as estimates of clinical certainty) by applying Pediatric Risk of Mortality III (PRISM III) in two scenarios: (1) survival prediction for individual patients and (2) resource rationing.
A non-concurrent cohort study.
32 pediatric intensive care units (PICUs).
10608 consecutive patients (571 deaths).
For the individual patient application, we investigated the observed survival rates for patients with low survival predictions and the confidence intervals associated with these predictions. For the resource rationing application, we investigated the maximum error rate of a policy which would limit therapy for patients with scores exceeding a very high threshold. For both applications, we also investigated how the confidence intervals change as the confidence levels change. The observed survival in the PRISM III groups >28, >35, and >42 were 6.3, 5.3, and 0%, with 95% upper confidence interval bounds of 10.5, 13.0, and 13.3%, respectively. Changing the confidence level altered the survival range by more than 300% in the highest risk group, indicating the importance of clinical certainty provisions in prognostic estimates. The maximum error rates for resource allocation decisions were low (e. g., 29 per 100000 at a 95% certainty level), equivalent to many of the risks of daily living. Changes in confidence level had relatively little effect on this result.
Predictions for an individual patient's risk of death with a high PRISM score are statistically not precise by virtue of the small number of patients in these groups and the resulting wide confidence intervals. Clinical certainty (confidence level) issues substantially influence outcome ranges for individual patients, directly affecting the utility of scores for individual patient use. However, sample sizes are sufficient for rationing decisions for many groups with higher certainty levels. Before there can be widespread acceptance of this type of decision support, physicians and families must confront what they believe is adequate certainty.
Intensive Care Medicine 12/1998; 24(12):1299-304. · 5.26 Impact Factor
[show abstract][hide abstract] ABSTRACT: The development and validation of a pediatric emergency department severity of illness assessment method, using hospital admission as the primary outcome.
A random sample of 25% of ED charts from 4 consecutive months in a university-affiliated pediatric hospital was reviewed, after exclusion of children with minor injuries and children triaged to the nonurgent clinic. Sampled data included components of the medical history, physical findings, physiologic variables, diagnoses, and ED therapies. Univariate and multivariate logistic regression analyses, with bootstrapping validation, were performed to develop a bias-corrected model estimating the probability of hospital admission.
Of the 2,683 ED patients whose records were reviewed, 643 (24%) were admitted to the hospital. The final model, which yielded a Pediatric Risk of Admission (PRISA) score, included the following: 3 components of the medical history, 3 chronic disease factors, 9 physiologic variables, 2 therapies, and 4 interaction terms. Overall, the number of hospital admissions was well predicted in both the 80% development and 20% validation samples. In the former, 514 admissions were predicted and 514 were observed; in the latter, 126.9 admissions were predicted and 129 were observed. The Hosmer-Lemeshow goodness-of-fit test demonstrated good agreement between observed and expected admissions in consecutive deciles of admission probability; total chi2 was 10.49 (P=.233) for the development sample and 11.85 (P=.222) for the validation sample. The areas under the receiver operating characteristic curves (+/-SE) were .86+/-.011 and .825+/-.024, respectively. As the risk of hospital admission increased, the proportions of patients using unique hospital-based resources and using ICU resources increased, and the proportion of patients dying increased.
The probability of admission to the hospital can reliably be estimated from data available during the pediatric ED stay. Applications for this method include studies of quality and efficiency of care and measurements of severity of illness.
Annals of Emergency Medicine 09/1998; 32(2):161-9. · 4.29 Impact Factor
[show abstract][hide abstract] ABSTRACT: Steady-state arterial spin-tagging MRI approaches were used to quantitate regional cerebral blood flow increases in prefrontal cortex during a working memory (“two-back”) task in six normal subjects. Statistically significant increases in cerebral blood flow in prefrontal cortex were observed in all six subjects: the average increase in cerebral blood flow in activated prefrontal cortex regions was 22 ± 5 cc/100 g/min (23 ± 7%). The results demonstrate that spin-tagging approaches can be used to follow focal activation in prefrontal cortex during cognitive tasks.
[show abstract][hide abstract] ABSTRACT: Assessment of pediatric intensive care unit (PICU) efficiency with a length of stay prediction model and validation of this assessment by an efficiency measure based on daily use of intensive care unit-specific therapies.
Inception cohort study of data acquired between 1989 and 1994.
Thirty-two PICUs, 16 selected randomly and 16 volunteering.
Consecutive admissions of 10,658 patients (466 deaths) who stayed at least 2 hours and up to 12 days in the PICU.
Length of stay and its prediction from a model with admission day data (PRISM III-24, diagnostic factors, mechanical ventilation). For validation 11 PICUs recorded each patient's "efficient" days, that is, days when at least one PICU-specific therapy was given. PICU efficiency was computed as either the ratio of the observed efficient days or the days accounted for by the predictor variables to the total care days, and the agreement was assessed by Spearman's rank correlation analysis.
The total care days provided by each PICU (n = 32) were well predicted by the length of stay model (r = 0.946). The agreement in 11 validation PICUs between therapy-based efficiency (range 0.30 to 0.67) and predictor-based efficiency (range 0.31 to 0.63) was excellent (rank correlation r = 0.936, p < 0.0001).
PICU efficiency comparisons with either method are nearly equivalent. Predictor-based efficiency has the advantage that it can be computed from admission day data only.
Journal of Pediatrics 07/1998; 133(1):79-85. · 4.04 Impact Factor
[show abstract][hide abstract] ABSTRACT: The use of the wavelet transform is explored for the detection of differences between brain functional magnetic resonance images (fMRI's) acquired under two different experimental conditions. The method benefits from the fact that a smooth and spatially localized signal can be represented by a small set of localized wavelet coefficients, while the power of white noise is uniformly spread throughout the wavelet space. Hence, a statistical procedure is developed that uses the imposed decomposition orthogonality to locate wavelet-space partitions with large signal-to-noise ratio (SNR), and subsequently restricts the testing for significant wavelet coefficients to these partitions. This results in a higher SNR and a smaller number of statistical tests, yielding a lower detection threshold compared to spatial-domain testing and, thus, a higher detection sensitivity without increasing type I errors. The multiresolution approach of the wavelet method is particularly suited to applications where the signal bandwidth and/or the characteristics of an imaging modality cannot be well specified. The proposed method was applied to compare two different fMRI acquisition modalities. Differences of the respective useful signal bandwidths could be clearly demonstrated; the estimated signal, due to the smoothness of the wavelet representation, yielded more compact regions of neuroactivity than standard spatial-domain testing.
IEEE Transactions on Medical Imaging 05/1998; 17(2):142-54. · 4.03 Impact Factor
[show abstract][hide abstract] ABSTRACT: To determine if decomplexification of heart rate dynamics occurs in critically ill and injured pediatric patients. We hypothesized that heart rate power spectra, a measure of heart rate dynamics, would inversely correlate with measures of severity of illness and outcome.
A prospective clinical study.
A 12-bed pediatric intensive care unit (ICU) in a tertiary care children's hospital.
One hundred thirty-five consecutive pediatric ICU admissions.
We compared heart rate power spectra with the Pediatric Risk of Mortality (PRISM) score, the Pediatric Cerebral Performance Category (PCPC), and the Pediatric Overall Performance Category (POPC). We found significant negative correlations between minimum low-frequency and high-frequency heart rate power spectral values recorded during ICU stay and the maximum PRISM score (log low-frequency heart rate power vs. PRISM, r2 = .293, p < .001; and log high-frequency heart rate power vs. PRISM, r2 = .243, p < .001) and outcome at ICU discharge (log low-frequency heart rate power vs. POPC or PCPC, r2 = .429, p < .001; and log high-frequency heart rate power vs. POPC or PCPC, r2 = .271, p < .001).
Our data support the hypothesis that measures of heart rate power spectra are inversely related and negatively correlated to severity of illness and outcome in critically ill and injured children. The phenomenon of decomplexification of physiologic dynamics may have important clinical implications in critical illness and injury.
Critical Care Medicine 03/1998; 26(2):352-7. · 6.12 Impact Factor
[show abstract][hide abstract] ABSTRACT: We present an automatic subpixel registration algorithm that minimizes the mean square intensity difference between a reference and a test data set, which can be either images (two-dimensional) or volumes (three-dimensional). It uses an explicit spline representation of the images in conjunction with spline processing, and is based on a coarse-to-fine iterative strategy (pyramid approach). The minimization is performed according to a new variation (ML*) of the Marquardt-Levenberg algorithm for nonlinear least-square optimization. The geometric deformation model is a global three-dimensional (3-D) affine transformation that can be optionally restricted to rigid-body motion (rotation and translation), combined with isometric scaling. It also includes an optional adjustment of image contrast differences. We obtain excellent results for the registration of intramodality positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) data. We conclude that the multiresolution refinement strategy is more robust than a comparable single-stage method, being less likely to be trapped into a false local optimum. In addition, our improved version of the Marquardt-Levenberg algorithm is faster.
[show abstract][hide abstract] ABSTRACT: To determine the effectiveness of cardiopulmonary resuscitation (CPR) in the pediatric intensive care unit (ICU).
A nonconcurrent cohort study of consecutive admissions.
Thirty-two pediatric ICUs.
Consecutive admissions to 32 pediatric ICUs.
Pediatric ICU patients were followed for the occurrence of a cardiopulmonary arrest (external cardiac massage for at least 2 mins). Patients who were in a state of continuous cardiopulmonary arrest on admission, or who never achieved stable vital signs, were excluded from the study. A total of 205 patients, from a sample of 11,165 (1.8%) pediatric admissions, experienced a cardiopulmonary arrest. Overall, 28 (13.7%) patients survived to hospital discharge. Neither mean ages nor age distribution affected survival. Only two diagnostic categories, traumatic illness, and other etiologies, were associated with survival. None of the patients fitting this category survived (p = .0028). The durations of CPR for survivors and nonsurvivors were 22.5 +/- 10.1 and 24.8 +/- 1.9 mins, respectively (p = .015). For CPR durations of <15 mins, 15 to 30 mins, and >30 mins, the survival rates were 18.6%, 12.2%, and 5.6%, respectively (linear trend p = .022). Thirty-five (17.1%) patients had a cardiopulmonary arrest before pediatric ICU admission and another arrest in the pediatric ICU. Only two (5.7%) of these 35 patients survived to discharge. Pediatric ICU survival decreased as the number of pediatric ICU arrests increased. Patients with one arrest (n = 155), two arrests (n = 29), and more than three arrests (n = 21) experienced survival rates of 14%, 14%, and 9.5%, respectively. Severity of illness, as measured by the Pediatric Risk of Mortality III score, was a significant predictor of survival (p < .001).
Pediatric ICU cardiac arrest is an uncommon event. When it does occur, prehospital CPR, duration of resuscitation, traumatic etiology, and severity of illness are important factors associated with survival.
Critical Care Medicine 12/1997; 25(12):1951-5. · 6.12 Impact Factor