Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “predictive modelling”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,441 records · Page 80Linked to original sources
PubMed2004

The utility of clinical and laboratory data for predicting intraabdominal injury among children.

BACKGROUND: The initial assessment of the child with blunt injury should lead ideally to a low rate of missed intraabdominal injury (IAI) while avoiding unnecessary imaging among children without IAI. The purpose of this study was to determine the utility of clinical and laboratory data for predicting the risk for IAI. METHODS: Among 351 children evaluated for possible blunt abdominal trauma, 23 variables potentially associated with IAI were determined retrospectively. Logistic regression and recursive partitioning were used to identify variables and develop predictive models. RESULTS: Logistic regression identified four positive predictors (abdominal tenderness, abrasion, ecchymoses, and alanine aminotransferase) and two negative predictors (injury caused by a motor vehicle crash and hematocrit) for IAI. The recursive partitioning model predicted the absence of IAI with a sensitivity of 100% (95% CI confidence interval, 86-100%) and a specificity of 87% (95% CI confidence interval, 81-91%) using abdominal examination and aspartate aminotransferase as discriminating variables. CONCLUSIONS: Physical examination combined with selected laboratory studies can be used to predict the risk of IAI accurately among children who sustain blunt trauma. Application of these findings may be useful in reducing costs and improving the accuracy of diagnosing IAI among children.

Abdominal Injuries↗
PubMed1999

Can postoperative nausea and vomiting be predicted?

BACKGROUND: Iletrospective studies fail to identify predictors of postoperative nausea and vomiting (PONV). The authors prospectively studied 17,638 consecutive outpatients who had surgery to identify predictors. METHODS: Data on medical conditions, anesthesia, surgery, and PONV were collected in the post-anesthesia care unit, in the ambulatory surgical unit, and in telephone interviews conducted 24 h after surgery. Multiple logistic regression with backward stepwise elimination was used to develop a predictive model An independent set of patients was used to validate the model RESULTS: Age (younger or older), sex (female or male), smoking status (nonsmokers or smokers), previous PONV, type of anesthesia (general or other), duration of anesthesia (longer or shorter), and type of surgery (plastic, orthopedic shoulder, or other) were independent predictors of PONV. A 10-yr increase in age decreased the likelihood of PONV by 13%. The risk for men was one third that for women. A 30-min increase in the duration of anesthesia increased the likelihood of PONV by 59%. General anesthesia increased the likelihood of PONV 11 times compared with other types of anesthesia. Patients with plastic and orthopedic shoulder surgery had a sixfold increase in the risk for PONV. The model predicted PONV accurately and yielded an area under the receiver operating characteristic curve of 0.785+/-0.011 using an independent validation set. CONCLUSIONS: A validated mathematical model is provided to calculate the risk of PONV in outpatients having surgery. Knowing the factors that predict PONV will help anesthesiologists determine which patients will need antiemetic therapy.

Adult↗
PubMed2005

The stress process perspective and adaptation of people with schizophrenia--an exploratory study.

BACKGROUND: Adaptation of people suffering from schizophrenia still raises numerous questions left unanswered by correlational studies and predictive models. A theoretical framework likely to bring new answers is the stress process perspective. Using the transactional model of coping, to which psychological and social variables have been added, this exploratory research studies the adaptation of 101 people with schizophrenia. METHOD: The research design is correlational with only one crosswise measure. RESULTS: The model predicts 60.7% of variance of adaptation and gives support that the transactional model of coping contributes to this variance. Five variables show a significant effect and account for 48% of the variance, and three interaction effects (two two-way and one three-way) add another 12.7% to the explained variance. CONCLUSION: Results give support to the significant impact of the variables education, age and negative symptoms on adaptation. They highlight the importance of accommodation, a cognitive strategy used by people with schizophrenia to increase their adaptation. They also challenge the broadly accepted assumption of a negative influence of life events on adaptation. Finally, the interaction effects allow us to better understand the mutual effects of variables on adaptation and confirm their relevance.

Adaptation, Psychological↗
PubMed1999

Psychiatric hospital service utilization of children and adolescents in state custody.

OBJECTIVE: To examine factors related to psychiatric hospitalization decision and length of stay of wards of the illinois Department of Child and Family Services. METHOD: A prospective design was implemented using the Childhood Severity of Psychiatric Illness (CSPI), a reliable, quantitative measure of psychiatric severity and its mediating factors. The CSPI was completed by hospital screeners upon conclusion of their crisis interviews. In addition to completing the CSPI, workers reported on demographic information, DSM-IV diagnoses, prescreening living arrangements, and length of hospital stay. RESULTS: CSPI variables could effectively predict decision to admit versus deflect. The overall accuracy of this statistically significant prediction model was 77.9%, which was replicated on a new sample. Factors associated with decision to hospitalize are clinical in nature; ratings of suicidality, dangerousness, and impulsivity contributed the most to the model. Predicting length of stay was only moderately successful. Despite achieving significance, the model accounted for just 15.1% of length of stay variance using a multiple regression. Factors associated with length of stay were largely nonclinical in nature: living arrangement stability, region of the hospitalization, and age. CONCLUSIONS: These results can be used to assess how decisions regarding level and duration of care are currently being made as a point of departure for quality improvement efforts.

Adolescent↗
PubMed2003

Assessing biological integrity using freshwater fish and decapod habitat selection functions.

Comparison between the number of taxa observed and the number expected in the absence of human impact is an easily understood and ecologically meaningful measure of biological integrity. This approach has been successfully applied to the assessment of the biological quality of flowing water sites using macroinvertebrates with the river invertebrate and classification system (RIVPACS) and its derivatives. In this paper, we develop a method similar to the RIVPACS predictive model approach to assess biological integrity at flowing-water sites using freshwater fish and decapod assemblages. We extend the RIVPACS approach by avoiding the biotic classification step and model each of the individual species separately. These assemblages were sampled at 118 least impacted (reference) sites in the Auckland region, New Zealand. Individual discriminant models based on the presence or absence of the 12 most common fish and decapod species were developed. Using the models, predictions were made using environmental measures at new sites to yield the probability of the capture of each of the 12 species, and these were combined to predict the assemblage expected at sites. The expected assemblage was compared to that observed using an observed over expected ratio (O/E). The models were evaluated using a number of internal tests including jackknifing, data partitioning, and the degree to which O/E values differed between reference sites and a set of sites perceived to be impaired by human impacts.

Animals↗
PubMed1997

Clinical applications of risk-assessment protocols in the management of individual patients.

BACKGROUND: Cardiac surgical databases are typically used to compare performance rather than to predict performance. Application of analytic techniques to representative clinical data allows for the creation of highly reliable prediction models for outcomes such as mortality after coronary artery bypass grafting. METHODS: The Northern New England Cardiovascular Disease Study Group collects clinical data from the six regional cardiac institutions. The four basic components of the one-page clinical data form are demographic information, comorbidity data, severity of illness data, and outcomes. From these, a mathematical model predicts the likelihood of inhospital mortality after coronary artery bypass grafting. RESULTS: The Northern New England Cardiovascular Disease Study Group mortality prediction rule was validated and found to predict and discriminate well. It is updated yearly based on the most current data. At Dartmouth-Hitchcock Medical Center, the rule has been combined with other decision-support tools based on American College of Cardiology/American Heart Association indications for coronary artery bypass grafting. These data are compared with data supplied by patients on their expectations for operation and their assessments of acceptable mortality risk. By expanding the database to include variables such as 5-year survival and additional outcomes, and additional interventions, an "electronic second opinion" is made available to clinicians. CONCLUSIONS: Although experience with decision-support tools has been positive, prediction rules have certain limitations. Their accuracy depends on data that are consistently tracked; thus, transfer to other patient populations must be approached with caution. The challenge for the future is to develop and validate similar techniques that apply to more difficult clinical situations.

Cardiovascular Diseases↗
PubMed1998

Seizures associated with fever: clinical data as predictors for normal biochemical blood levels.

UNLABELLED: We developed a predictive model to assess the probability of normal biochemical blood test results in children presenting with a seizure associated with fever. The models were based on various combinations of patient characteristics of the history and physical examination of 203 children. The characteristics included gender, age in years, previous history of febrile seizures, family history of febrile seizures, fever previous to the seizure, vomiting and diarrhoea previous to the seizure. Further, clinical characteristics of the seizure were considered: focal seizure signs, multiple seizure, seizure duration and rectal temperature at seizure. The outcome was defined as normal test results of serum levels of sodium (n=115, 68%), calcium (n=149, 89%) and glucose (n=173, 100%), according to the hospital reference values. The prevalence of abnormal test results was rather low and the abnormalities were outside the morbidity range. We used logistic regression to relate the outcome to the several clinical characteristics. The discriminative ability of the models was 0.63 (area under the receiver operating characteristic curve of the model predicting normal sodium), 0.66 (normal calcium) and 0.66 (both normal). The score chart we constructed is an additional tool to a carefully performed patient history and physical examination and it may help to decide if a biochemical test is indicated for the individual patient. CONCLUSION: In children with seizures associated with fever, abnormal biochemical blood test results are rare and outside the morbidity range. The biochemical tests are generally not required. In children with a low probability of a normal result as calculated by the score chart, the test may be indicated.

Blood Chemical Analysis↗
PubMed2001

Treatment of myelodysplastic syndromes.

The IPSS scoring system is useful to establish the appropriate treatment plan in MDS. Growth factors may alleviate both anemia and neutropenia in some MDS patients. Serum Epo levels and need for transfusion serve as good predictors of the erythroid response to the combination of Epo and G-CSF. Subgroups of MDS patients may respond favorably to immunosuppressive therapies such as CyA and ATG. Low-dose chemotherapy may also improve peripheral blood counts. Platelet counts, bone marrow cellularity, chromosome aberrations, and ringed sideroblasts combine to create a model predicting the response to low-dose ara-C. High-dose chemotherapy may lead to complete remission in about half of MDS patients, but the duration of remission is often short. The only proven curative therapy for MDS is allogeneic stem cell transplantation, resulting in an overall disease-free survival rate of about 40%. Only a minority of patients, however, can undergo allogeneic transplantation, both because of patient age and the availability of suitable donors. Autologous stem cell transplantation may be an option for selected patients who are unable to find allogeneic donors. Because the clinical features of patients with MDS are quite heterogeneous, the development of more accurate predictive models may be necessary to improve the efficacy of treatment.

Antineoplastic Combined Chemotherapy Protocols↗
PubMed1990

Which oral white lesions will become malignant? An image cytometric study.

We investigated the value of image cytometry in predicting the prognosis of oral epithelial lesions, whether or not they show dysplasia. Thirty-five oral epithelial lesions were studied retrospectively. Of these, 23 had later transformed to carcinoma and 12 had not. By means of the Leitz TAS image analyzer, 200 nuclei of epithelial cells and 20 nuclei of lymphocytes from each section were individually assessed for eight features related to shape and amount of stain and for six features related to chromatin pattern. The mean, standard deviation, and interquartile range of each feature were calculated, first for each section and then for each group. With the use of linear stepwise discriminant analysis we constructed a predictive model, which consisted of three variables related to chromatin pattern. The variables were mean margination, standard deviation of clumping, and standard deviation of condensation. In the jackknife classification, this model predicted the malignant potential of the lesions that later transformed to cancer with 86% predictive value and 83% sensitivity.

Adult↗
PubMed2006

Cardiovascular disease risk in subjects with type 2 diabetes mellitus in a population in southern Spain.

To calculate overall 10-year cardiovascular disease (CVD) risk for patients with type 2 diabetes we applied the UKPDS and SCORE prediction models to data derived from clinical notes of 1846 patients (mean age 65.5 years; 55.8% women) with type 2 diabetes attending eight Primary-Care Centres serving a catchment population of 200,000 citizens in Andalusia. The results showed obesity and high blood pressure present in >50%, established CVD in 24%, retinopathy in 30%, and nephropathy in 17%. Mean HbAlc level was 7.3%. Compliance with therapeutic goals was 54% for systolic blood pressure <130 mmHg, 39% for HbA1c <7% and 9% for LDL cholesterol <2.58 mmol/L. Approximately 33% were receiving treatment with metformin, statins, renin-angiotensin system inhibitors and anti-aggregation agents. UKPDS risk for coronary heart disease (CHD) was 23% and 16% for stroke. The SCORE 10-year mortality risk was 5%. Correlation coefficient between the two models predicting CVD risk was 0.68 (p<0.001). We conclude that, despite the European consensus that CVD is low in Mediterranean countries, the CVD risk factors in the type 2 diabetes sub-population in southern Spain is relatively high. Specific measures of health-care intervention are needed if CVD-associated morbido-mortality rates in these diabetic patients are to be reduced.

Adult↗
PubMed2003

Clay particle retention in small constructed wetlands.

Constructed wetlands (CWs) can be used to mitigate non-point source pollution from arable fields. Previous investigations have shown that the relative soil particle retention in small CWs increases when hydraulic load increases. This paper investigates why this phenomenon occurs, even though common retention models predict the opposite, by studying clay and silt particle retention in two Norwegian CWs. Retention was measured with water flow proportional sampling systems in the inlet and outlet of the wetlands, and the texture of the suspended solids was analyzed. The surface area of the CWs was small compared to the watershed area (approximately 0.07%), giving high average hydraulic loads (1.1 and 2.0 md(-1)). One of the watersheds included only old arable land, whereas the other included areas with disturbed topsoil after artificial land leveling. Clay particle retention was 57% for the CW in the first watershed, and 22% for the CW in the disturbed watershed. The different behavior of the wetlands could be due to differences in aggregate size and stability of the particles entering the wetlands. Results showed that increased hydraulic loads did affect CW retention negatively. However, as runoff increased, soil particles/aggregates with higher sedimentation velocities entered the CWs (e.g., the clay particles behaved as silt particles). Hence, clay particle settling velocity is not constant as assumed in many prediction models. The net result was increased retention.

Aluminum Silicates↗
PubMed2000

Alternative measures and models of hazardous consumption.

Hazardous alcohol consumption has been conceptualized (1) as all alcohol consumed on days when some threshold, usually approximately 60 g, is exceeded, and (2) as only that portion of intake that exceeds this threshold. The first measure is hypothesized to be a better predictor of acute alcohol-related outcomes, because of its greater capacity to discriminate between individuals who frequently exceed the hazardous threshold by a small amount and those who infrequently exceed the threshold by a large amount. To test this hypothesis, the two approaches were used to construct alternative estimates of a number of measures of hazardous consumption. Individually and in combination, these measures were compared in a series of multiplicative models predicting four alcohol-related outcomes: impaired driving, fighting, interpersonal problems and injuries. There was no consistent evidence for one approach being superior to the other as a predictor of these outcomes. In fact, the use of appropriate linear transformations of the consumption variables had a far greater effect on improving the proportion of variance explained. The most highly predictive models were those based on frequency of hazardous consumption, mean volume of hazardous intake consumed per hazardous drinking day, volume of nonhazardous intake and their interactions. Differences among models were small, and models using combinations of simple, easy-to-obtain measures performed nearly on a par with those utilizing far more complex measures.

Adolescent↗
PubMed2002

Sensitivity analysis for healthcare models fitted to data by statistical methods.

After fitting complex models to data using statistical methods, a sensitivity analysis can be carried out. This determines which parts of a model are causing the bulk of the uncertainty in the model predictions (model "output"), and is a decision-support tool for the modeller who contemplates refining a model further or collecting additional data. A simple methodology for carrying out a sensitivity analysis is described. It is envisaged that such a relatively quick insight-generating step would precede the use of a more formal decision-theoretic approach that would address specific questions. Its use is illustrated using a model for breast cancer screening previously published in this journal. A simpler 3-parameter screening model is used in a simulation study of the error of the method as a function of sample size.

Breast Neoplasms↗
PubMed1995

Breast reduction for symptomatic macromastia: can objective predictors for operative success be identified?

The purpose of this paper is to assess symptomatic macromastia, the relief of symptoms by operation, and predictors of symptom relief. The methods used have been retrospective chart review and a self-assessment patient questionnaire. One-hundred and thirty-three patients underwent an average 1660-gm reduction. Ninety-three percent reported a postoperative decrease in symptoms such as shoulder grooves and shoulder, neck, and back pain. Correlation between breast size and sign or symptom severity achieved significance only for the preoperative submammary rash (r = 0.33, p < 0.001). Patients lost an average of 8.9 lb postoperatively and were less overweight (49 versus 40 percent). Activity level increased postoperatively in 63 percent. Postoperative chest size correlated inversely with activity level (r = 0.35, p < 0.001). Thirty-nine percent of patients who took pain medications preoperatively were able to eliminate these postoperatively. The quantity of tissue removed did not correlate with outcome. A model predictive of symptom relief could not be developed (total R2 = 0.03). Reduction mammaplasty promoted relief of signs and symptoms of macromastia, but a predictive model of successful operation could not be developed.

Back Pain↗
PubMed1995

Hemodynamic predictors of atrial fibrillation or flutter after coronary artery bypass grafting.

The cumulated incidence of atrial fibrillation or flutter after coronary artery bypass grafting is 30%. The causes of these arrhythmias have not yet been sufficiently identified. We therefore undertook the present study to analyze the possible association of hemodynamic function during the various phases of coronary artery bypass grafting and the later development of atrial fibrillation/flutter. Hemodynamic function was measured with a pulmonary artery catheter in 120 consecutive patients undergoing elective coronary artery bypass surgery. Thirty-five (29%) of the patients developed atrial fibrillation/flutter. Logistic regression analysis identified independent predictors of atrial fibrillation/flutter. After induction of general anesthesia, the relative risk (95% confidence interval) of older age was 1.09/year (1.03-1.16), and the reduction in relative risk by an increase in left ventricular stroke work was 0.96/gm (0.93-0.99). After weaning from the extracorporeal circulation the independent significant predictors were age, relative risk 1.07/year (1.01-1.13), and increased central venous pressure, relative risk 1.12/mm Hg (1.00-1.26). At the time of admission to the intensive care unit, the relative risk of age was 1.10/year (1.03-1.18), and the relative risk of an increased central venous pressure was 1.26/mm Hg (1.06-1.49). However, the best prediction model (prediction after induction of general anesthesia) only provided a median predicted probability of atrial fibrillation/flutter of 0.37 for the patients who had atrial fibrillation/flutter, and a median predicted probability of atrial fibrillation/flutter of 0.20 for the patients without these arrhythmias. We identified possible hemodynamic predictors of atrial fibrillation/flutter after coronary bypass surgery, but the use of a risk stratification for development of atrial fibrillation/flutter based on hemodynamic function cannot be recommended.

Age Factors↗
PubMed2003

Asking neutral versus leading questions: implications for functional limitation measurement.

UNLABELLED: National surveys of older Americans routinely have included functional limitation items using either a leading approach ("how much difficulty do you have...") or a neutral approach ("do you have any difficulty..."). This article evaluates the performance of scales based on these two approaches. METHODS: Using responses from 595 randomly selected participants to the 1994 Health and Retirement Study, the authors compared prevalences and evaluated scales based on each approach with respect to the extent of missing data, face validity, reliability, predictive validity, convergent validity, and robustness of odds ratios in predictive models. RESULTS: The authors found that leading questions provided higher estimates of functional limitations than neutral questions, but both approaches yielded scales with similar validity and reliability. However, for both approaches, scales incorporating degree of difficulty had better validity and reliability than those based on counts of tasks. All four approaches yielded substantially similar coefficients in a model predicting disability onset. DISCUSSION: The authors conclude that, because they minimize survey time without compromising validity and reliability, items that explicitly capture degree of difficulty by asking "How much difficulty do you have..." may be the optimal approach for survey designers.

Activities of Daily Living↗
PubMed2001

Withdrawal of support in intracerebral hemorrhage may lead to self-fulfilling prophecies.

BACKGROUND: Withdrawal of support in patients with severe brain injury invariably leads to death. Preconceived notions about futility of care in patients with intracerebral hemorrhage (ICH) may prompt withdrawal of support, and modeling outcome in patient populations in whom withdrawal of support occurs may lead to self-fulfilling prophecies. METHODS: Subjects included consecutive patients with supratentorial ICH. Radiographic characteristics of the hemorrhage, clinical variables, and neurologic outcome were assessed. Attitudes about futility of care were examined among members of the departments of neurology and neurologic surgery through a written survey and case presentations. RESULTS: There were 87 patients with supratentorial ICH; overall mortality was 34.5% (30/87). Mortality was 66.7% (18/27) in patients with Glasgow Coma Score < or = 8 and ICH volume > 60 cm(3). Medical support was withdrawn in 76.7% (23/30) of patients who died. Inclusion of a variable to account for the withdrawal of support in a model predicting outcome negated the predictive value of all other variables. Patients undergoing surgical decompression were unlikely to have support withdrawn, and surgery was less likely to be performed in older patients (p < 0.01) and patients with left hemispheric hemorrhage (p = 0.04). Survey results suggested that practitioners tend to be overly pessimistic in prognosticating outcome based upon data available at the time of presentation. CONCLUSIONS: The most important prognostic variable in determining outcome after ICH is the level of medical support provided. Withdrawal of support in patients felt likely to have a "poor outcome" biases predictive models and leads to self-fulfilling prophecies. Our data show that individual patients in traditionally "poor outcome" categories can have a reasonable neurologic outcome when treated aggressively.

Aged↗
PubMed2005

Modeling the distribution of Schistosoma mansoni and host snails in Uganda using satellite sensor data and Geographical Information Systems.

The potential value of MODIS satellite sensor data on Normalized Difference Vegetation Index (NDVI) and land surface temperatures (LST) for describing the distribution of the Schistosoma mansoni-"Biomphalaria pfeifferi"/Biomphalaria sudanica parasite-snail system in inland Uganda, were tested by developing annual and seasonal composite models, and iteratively analysing for their relationship with parasite and snail distribution. The dry season composite model predicted an endemic area that produced the best fit with the distribution of schools with > or =5% prevalence. NDVI values of 151-174, day temperatures of 26-36 degrees C, and night temperatures of 15-20 degrees C were used as criteria for the prediction model. Using the same approach with host snail data indicated that most of Uganda is suitable "B. pfeifferi"/B. sudanica habitat, except for possibly the north-eastern region of the country. The parasite, however, appears to be restricted in its distribution in both the north-eastern and the south-western regions of Uganda. The absence of disease in the south-west can not be attributed to the absence of snail hosts. Results suggest a combination of satellite sensor data on temperature and standard climate data on precipitation, as the best ecological determinants of the S. mansoni-"B. pfeifferi"/B. sudanica system. Satellite composite models and logistic regression analysis, suggest low night time temperature as one of the significant factors inhibiting S. mansoni transmission in the south-western highland areas of Uganda. The developed models are, however, unique, representing species-specific ecologic preferences of the S. mansoni-"B. Pfeifferi"/B. sudanica system in inland Uganda. Further validation studies are needed to test the value of the model in other countries in East Africa.

Animals↗