Cost minimization using an artificial neural network sleep apnea prediction tool for sleep studies

RATIONALE

More than a million polysomnograms (PSGs) are performed annually in the United States to diagnose obstructive sleep apnea (OSA). Third-party payers now advocate a home sleep test (HST), rather than an in-laboratory PSG, as the diagnostic study for OSA regardless of clinical probability, but the economic benefit of this approach is not known.

OBJECTIVES

We determined the diagnostic performance of OSA prediction tools including the newly developed OSUNet, based on an artificial neural network, and performed a cost-minimization analysis when the prediction tools are used to identify patients who should undergo HST.

METHODS

The OSUNet was trained to predict the presence of OSA in a derivation group of patients who underwent an in-laboratory PSG (n = 383). Validation group 1 consisted of in-laboratory PSG patients (n = 149). The network was trained further in 33 patients who underwent HST and then was validated in a separate group of 100 HST patients (validation group 2). Likelihood ratios (LRs) were compared with two previously published prediction tools. The total costs from the use of the three prediction tools and the third-party approach within a clinical algorithm were compared.

MEASUREMENTS AND MAIN RESULTS

The OSUNet had a higher +LR in all groups compared with the STOP-BANG and the modified neck circumference (MNC) prediction tools. The +LRs for STOP-BANG, MNC, and OSUNet in validation group 1 were 1.1 (1.0-1.2), 1.3 (1.1-1.5), and 2.1 (1.4-3.1); and in validation group 2 they were 1.4 (1.1-1.7), 1.7 (1.3-2.2), and 3.4 (1.8-6.1), respectively. With an OSA prevalence less than 52%, the use of all three clinical prediction tools resulted in cost savings compared with the third-party approach.

CONCLUSIONS

The routine requirement of an HST to diagnose OSA regardless of clinical probability is more costly compared with the use of OSA clinical prediction tools that identify patients who should undergo this procedure when OSA is expected to be present in less than half of the population. With OSA prevalence less than 40%, the OSUNet offers the greatest savings, which are substantial when the number of sleep studies done annually is considered.

Qualitative point-of-care D-dimer testing compared with quantitative D-dimer testing in excluding pulmonary embolism in primary care

BACKGROUND

General practitioners can safely exclude pulmonary embolism (PE) by using the Wells PE rule combined with D-dimer testing.

OBJECTIVE

To compare the accuracy of a strategy using the Wells rule combined with either a qualitative point-of-care (POC) D-dimer test performed in primary care or a quantitative laboratory-based D-dimer test.

METHODS

We used data from a prospective cohort study including 598 adults suspected of PE in primary care in the Netherlands. General practitioners scored the Wells rule and carried out a qualitative POC test. All patients were referred to hospital for reference testing. We obtained quantitative D-dimer test results as performed in hospital laboratories. The primary outcome was the prevalence of venous thromboembolism in low-risk patients.

RESULTS

Prevalence of PE was 12.2%. POC D-dimer test results were available in 582 patients (97%). Quantitative test results were available in 401 patients (67%). We imputed results in 197 patients. The quantitative test and POC test missed one (0.4%) and four patients (1.5%), respectively, with a negative strategy (Wells ≤ 4 points and D-dimer test negative) (P = 0.20). The POC test could exclude 23 more patients (4%) (P = 0.05). The sensitivity and specificity of the Wells rule combined with a POC test were 94.5% and 51.0% and, combined with a quantitative test, 98.6% and 47.2%, respectively.

CONCLUSIONS

Combined with the Wells PE rule, both tests are safe to use in excluding PE. The quantitative test seemed to be safer than the POC test, albeit not statistically significant. The specificity of the POC test was higher, resulting in more patients in whom PE could be excluded.

Pulmonary-Specific Intermountain Risk Score Predicts All-Cause Mortality via Spirometry, the Red Cell Distribution Width, and Other Laboratory Parameters

BACKGROUND

Pulmonary function testing parameters predict cardiovascular and mortality outcomes. Previously, risk scores were created using the basic metabolic profile and complete blood count, including the Intermountain Risk Score (IMRS). This study sought to develop similar pulmonary-specific risk scores for mortality prediction.

METHODS

Subjects evaluated by spirometry at 5 Intermountain Healthcare hospitals (females: n = 2,943; males: n = 2,495) were randomly assigned to risk score derivation (70% of subjects) or an independent validation set (the remaining 30%). Sex-specific scores used spirometry, age, and metabolic and blood count laboratory data. Cox regression β-coefficients formed the basis of risk score weightings.

RESULTS

Among females, pulmonary IMRS was strongly associated with 5-y mortality in the validation set (hazard ratio = 1.24 per +1 risk score, CI 1.16-1.33, P trend < .001), with C-statistics of C = 0.835 and C = 0.757 for derivation and validation, respectively. Among males, validation results were similarly significant (hazard ratio = 1.20 per +1 risk score value, CI 1.11-1.28, P trend < .001), with C = 0.755 and C = 0.699 in derivation and validation sets, respectively. Results were stronger for pulmonary basic metabolic profile risk score, with females having C = 0.815 (derivation) and C = 0.806 (validation), whereas males had C = 0.734 and C = 0.731.

CONCLUSIONS

Pulmonary-specific IMRS and pulmonary-specific basic metabolic profile risk score provided excellent discrimination of mortality among pulmonary subjects. These risk stratification tools combine familiar, relatively inexpensive, commonly-measured, standardized laboratory parameters with spirometry data. They may be electronically calculated and delivered at the point of care, providing meaningful risk information to assist clinicians in patient evaluations.

A noninvasive method for coronary artery diseases diagnosis using a clinically-interpretable fuzzy rule-based system

BACKGROUND

Coronary heart diseases/coronary artery diseases (CHDs/CAD), the most common form of cardiovascular disease (CVD), are a major cause for death and disability in developing/developed countries. CAD risk factors could be detected by physicians to prevent the CAD occurrence in the near future. Invasive coronary angiography, a current diagnosis method, is costly and associated with morbidity and mortality in CAD patients. The aim of this study was to design a computer-based noninvasive CAD diagnosis system with clinically interpretable rules.

MATERIALS AND METHODS

In this study, the Cleveland CAD dataset from the University of California UCI (Irvine) was used. The interval-scale variables were discretized, with cut points taken from the literature. A fuzzy rule-based system was then formulated based on a neuro-fuzzy classifier (NFC) whose learning procedure was speeded up by the scaled conjugate gradient algorithm. Two feature selection (FS) methods, multiple logistic regression (MLR) and sequential FS, were used to reduce the required attributes. The performance of the NFC (without/with FS) was then assessed in a hold-out validation framework. Further cross-validation was performed on the best classifier.

RESULTS

In this dataset, 16 complete attributes along with the binary CHD diagnosis (gold standard) for 272 subjects (68% male) were analyzed. MLR + NFC showed the best performance. Its overall sensitivity, specificity, accuracy, type I error (α) and statistical power were 79%, 89%, 84%, 0.1 and 79%, respectively. The selected features were "age and ST/heart rate slope categories," "exercise-induced angina status," fluoroscopy, and thallium-201 stress scintigraphy results.

CONCLUSION

The proposed method showed "substantial agreement" with the gold standard. This algorithm is thus, a promising tool for screening CAD patients.

Incident disability in older adults: prediction models based on two British prospective cohort studies

OBJECTIVE

To develop and validate a prediction model for incident locomotor disability after 7 years in older adults.

SETTING

Prospective British cohort studies: British Women's Heart and Health Study (BWHHS) for development and the English Longitudinal Study of Ageing (ELSA) for validation.

SUBJECTS

Community-dwelling older adults.

METHODS

Multivariable logistic regression models after selection of predictors with backward elimination. Model performance was assessed using metrics of discrimination and calibration. Models were internally and externally validated.

RESULTS

Locomotor disability was reported in BWHHS by 861 of 1,786 (48%) women after 7 years. Age, a history of arthritis and low physical activity levels were the most important predictors of locomotor disability. Models using routine measures as predictors had satisfactory calibration and discrimination (c-index 0.73). Addition of 31 blood markers did not increase the predictive performance. External validation in ELSA showed reduced discrimination (c-index 0.65) and an underestimation of disability risks. A web-based calculator for locomotor disability is available (http://www.sealedenvelope.com/trials/bwhhsmodel/).

CONCLUSIONS

We developed and externally validated a prediction model for incident locomotor disability in older adults based on routine measures available to general practitioners, patients and public health workers, and showed an adequate discrimination. Addition of blood markers from major biological pathways did not improve the performance of the model. Further replication in additional data sets may lead to further enhancement of the current model.

Advanced chronic kidney disease in non-valvular atrial fibrillation: extending the utility of R2CHADS2 to patients with advanced renal failure

BACKGROUND

The R2CHADS2 is a new prediction rule for stroke risk in atrial fibrillation (AF) patients wherein R stands for renal risk. However, it was created from a cohort that excluded patients with advanced renal failure (defined as glomerular filtration rate of <30 mL/min). Our study extends the use of R2CHADS2 to patients with advanced renal failure and aims to compare its predictive power against the currently used CHADS and CHA2DS2VaSc.

METHODS

This retrospective cohort study analyzed the 1-year risk for stroke of the 524 patients with AF at Metropolitan Hospital Center. AUC and C statistics were calculated using three groups: (i) the entire cohort including patients with advanced renal failure, (ii) a cohort excluding patients with advanced renal failure and (iii) all patients with GFR < 30 mL/min only.

RESULTS

R2CHADS2, as a predictor for stroke risk, consistently performs better than CHADS2 and CHA2DS2VsC in groups 1 and 2. The C-statistic was highest in R2CHADS compared with CHADS or CHADSVASC in group 1 (0.718 versus 0.605 versus 0.602) and in group 2 (0.724 versus 0.584 versus 0.579). However, there was no statistically significant difference in group 3 (0.631 versus 0.629 versus 0.623).

CONCLUSION

Our study supports the utility of R2CHADS2 as a clinical prediction rule for stroke risk in patients with advanced renal failure.

Clinical decision rule to prevent unnecessary chest X-ray in patients with blunt multiple traumas

BACKGROUND

Since the diagnostic yield of chest X-ray (CXR) is not high enough, when it is ordered for all the multiple trauma patients, this study was aimed to evaluate the relationship between clinical and CXR findings in order to formulate a clinical decision rule to prevent unnecessary CXR in these patients.

METHODS

Stable multiple blunt trauma patients referring to the ED were included. The clinical and radiographic findings of all the patients were collected and the relationships between these variables analysed. Finally, based on the regression coefficients (β) of the variables, the Thoracic Injury Rule-out Criteria (TIRC) were designed.

RESULTS

A total of 2607 patients were included (males: 78.9%, mean age: 34.1 ± 15.0 years). Age over 60 (β = 0.8; 95% CI: 0.27-1.34; P = 0.003), crepitation (β = 4.33; 95% CI: 1.65-7.0; P < 0.001), loss of consciousness (β = 3.16; 95% CI: 2.44-3.88; P < 0.001), decrease in pulmonary sounds (β = 2.67; 95% CI: 1.73-3.6; P < 0.001), chest wall pain (β = 2.12; 95% CI: 1.63-2.61; P < 0.001) and tenderness (β = 1.78; 95% CI: 1.26-2.27; P < 0.001), dyspnea (β = 1.3; 95% CI: 0.41-2.18; P = 0.004) and abrasion (β = 0.5; 95% CI: 0.22-0.83; P = 0.03) were independent factors predicting thoracic injury. CXR in stable conscious multiple blunt trauma patients under 60 years, without chest wall pain and tenderness, decrease in pulmonary sounds, crepitation, skin abrasion, and dyspnea did not provide any additional findings.

CONCLUSIONS

Based on TIRC, it seems that CXR in stable multiple blunt trauma patients who are conscious and under 60 and have no decrease in pulmonary sounds, no dyspnea, no thoracic skin abrasion, and no crepitation can be ignored.

A clinical prediction rule for diagnosing human infections with avian influenza A(H7N9) in a hospital emergency department setting

BACKGROUND

Human infections with avian influenza A(H7N9) virus are associated with severe illness and high mortality. To better inform triage decisions of hospitalization and management, we developed a clinical prediction rule for diagnosing patients with A(H7N9) and determined its predictive performance.

METHODS

Clinical details on presentation of adult patients hospitalized with either A(H7N9)(n = 121) in China from March to May 2013 or other causes of acute respiratory infections (n = 2,603) in Jingzhou City, China from January 2010 through September 2012 were analyzed. A clinical prediction rule was developed using a two-step coefficient-based multivariable logistic regression scoring method and evaluated with internal validation by bootstrapping.

RESULTS

In step 1, predictors for A(H7N9) included male sex, poultry exposure history, and fever, haemoptysis, or shortness of breath on history and physical examination. In step 2, haziness or pneumonic consolidation on chest radiographs and leukopenia were also associated with a higher probability of A(H7N9). The observed risk of A(H7N9) was 0.3% for those assigned to the low-risk group and 2.5%, 4.3%, and 44.0% for tertiles 1 through 3, respectively, in the high-risk group. This prediction rule achieved good model performance, with an optimism-corrected sensitivity of 0.93, a specificity of 0.80, and an area under the receiver-operating characteristic curve of 0.96.

CONCLUSIONS

A simple decision rule based on data readily obtainable in the setting of patients' first clinical presentations from the first wave of the A/H7N9 epidemic in China has been developed. This prediction rule has achieved good model performance in predicting their risk of A(H7N9) infection and should be useful in guiding important clinical and public health decisions in a timely and objective manner. Data to be gathered with its use in the current evolving second wave of the A/H7N9 epidemic in China will help to inform its performance in the field and contribute to its further refinement.

Predicting postoperative mortality after colorectal surgery: a novel clinical model

AIM

The aim of this study was to develop and externally validate a clinically, practical and discriminative prediction model designed to estimate in-hospital mortality of patients undergoing colorectal surgery.

METHOD

All consecutive patients who underwent elective or emergency colorectal surgery from 1990 to 2005, at the Zaandam Medical Centre, The Netherlands, were included in this study. Multivariate logistic regression analysis was performed to estimate odds ratios (ORs) and 95% confidence intervals (CIs) linking the explanatory variables to the outcome variable in-hospital mortality, and a simplified Identification of Risk in Colorectal Surgery (IRCS) score was constructed. The model was validated in a population of patients who underwent colorectal surgery from 2005 to 2011 in Barcelona, Spain. Predictive performance was estimated by calculating the area under the receiver operating characteristic curve.

RESULTS

The strongest predictors of in-hospital mortality were emergency surgery (OR = 6.7, 95% CI 4.7-9.5), tumour stage (OR = 3.2, 95% CI 2.8-4.6), age (OR = 13.1, 95% CI 6.6-26.0), pulmonary failure (OR = 4.9, 95% CI 3.3-7.1) and cardiac failure (OR = 3.7, 95% CI 2.6-5.3). These parameters were included in the prediction model and simplified scoring system. The IRCS model predicted in-hospital mortality and demonstrated a predictive performance of 0.83 (95% CI 0.79-0.87) in the validation population. In this population the predictive performance of the CR-POSSUM score was 0.76 (95% CI 0.71-0.81).

CONCLUSIONS

The results of this study have shown that the IRCS score is a good predictor of in-hospital mortality after colorectal surgery despite the relatively low number of model parameters.

Prediction of asymptomatic carotid artery stenosis in the general population: identification of high-risk groups

BACKGROUND AND PURPOSE

Because of a low prevalence of severe carotid stenosis in the general population, screening for presence of asymptomatic carotid artery stenosis (ACAS) is not warranted. Possibly, for certain subgroups, screening is worthwhile. The present study aims to develop prediction rules for the presence of ACAS (>50% and >70%).

METHODS

Individual participant data from 4 population-based cohort studies (Malmö Diet and Cancer Study, Tromsø Study, Carotid Atherosclerosis Progression Study, and Cardiovascular Health Study; totaling 23 706 participants) were pooled. Multivariable logistic regression was performed to determine which variables predict presence of ACAS (>50% and >70%). Calibration and discrimination of the models were assessed, and bootstrapping was used to correct for overfitting.

RESULTS

Age, sex, history of vascular disease, systolic and diastolic blood pressure, total cholesterol/high-density lipoprotein ratio, diabetes mellitus, and current smoking were predictors of stenosis (>50% and >70%). The calibration of the model was good confirmed by a nonsignificant Hosmer and Lemeshow test for moderate (P=0.59) and severe stenosis (P=0.07). The models discriminated well between participants with and without stenosis, with an area under the receiver operating characteristic curve corrected for over optimism of 0.82 (95% confidence interval, 0.80-0.84) for moderate stenosis and of 0.87 (95% confidence interval, 0.85-0.90) for severe stenosis. The regression coefficients of the predictors were converted into a score chart to facilitate practical application.

CONCLUSIONS

A clinical prediction rule was developed that allows identification of subgroups with high prevalence of moderate (>50%) and severe (>70%) ACAS. When confirmed in comparable cohorts, application of the prediction rule may lead to a reduction in the number needed to screen for ACAS.

Developing an international register of clinical prediction rules for use in primary care: a descriptive analysis

PURPOSE

We describe the methodology used to create a register of clinical prediction rules relevant to primary care. We also summarize the rules included in the register according to various characteristics.

METHODS

To identify relevant articles, we searched the MEDLINE database (PubMed) for the years 1980 to 2009 and supplemented the results with searches of secondary sources (books on clinical prediction rules) and personal resources (eg, experts in the field). The rules described in relevant articles were classified according to their clinical domain, the stage of development, and the clinical setting in which they were studied.

RESULTS

Our search identified clinical prediction rules reported between 1965 and 2009. The largest share of rules (37.2%) were retrieved from PubMed. The number of published rules increased substantially over the study decades. We included 745 articles in the register; many contained more than 1 clinical prediction rule study (eg, both a derivation study and a validation study), resulting in 989 individual studies. In all, 434 unique rules had gone through derivation; however, only 54.8% had been validated and merely 2.8% had undergone analysis of their impact on either the process or outcome of clinical care. The rules most commonly pertained to cardiovascular disease, respiratory, and musculoskeletal conditions. They had most often been studied in the primary care or emergency department settings.

CONCLUSIONS

Many clinical prediction rules have been derived, but only about half have been validated and few have been assessed for clinical impact. This lack of thorough evaluation for many rules makes it difficult to retrieve and identify those that are ready for use at the point of patient care. We plan to develop an international web-based register of clinical prediction rules and computer-based clinical decision support systems.

Indices of initial hepatitis C virus RNA reduction rate to predict efficacy of interferon-beta followed by peginterferon plus ribavirin for genotype 1b high viral load

AIM

Initial hepatitis C virus (HCV) RNA reduction was investigated as a potential index for sustained virological response (SVR) in the treatment of interferon (IFN)-β followed by peginterferon plus ribavirin (PEG IFN/RBV).

METHODS

The treatment course was retrospectively analyzed in 64 genotype 1b patients with a HCV RNA level of 5.0 logIU/mL or higher. IFN-β was administrated twice a day for 2 weeks followed by 24 or 48 weeks of PEG IFN/RBV. The serum HCV RNA level was measured by real-time polymerase chain reaction before administration and at 1, 2 and 4 weeks of therapy.

RESULTS

By the duration of PEG IFN administration, the SVR rates were 11% (2/18, <19 weeks), 64% (23/36, 20-24 weeks) and 40% (4/10, 25-72 weeks) (P = 0.0011, χ(2) -test). The SVR rate was high in patients in whom the HCV RNA level had decreased by 2.5 logIU/mL or greater at 1 week of IFN-β (29/55 [53%] vs 0/9 [0%], P = 0.0029, χ(2) -test). Among these patients, the SVR rate was even higher in those with continuous reduction in the first 2 weeks after the switch to PEG IFN/RBV (27/45 [60%] vs 2/10 [20%], P = 0.0048). Age below 65 years, no previous IFN course and good initial HCV RNA reduction were significantly associated with SVR on multivariate analysis, and the SVR rate was 95% (18/19) among these patients.

CONCLUSION

The 2.5 logIU/mL reduction in HCV RNA at 1 week of IFN-β and the continuous reduction just after the switch to PEG IFN/RBV are important SVR-predictive indices.

A clinical score can predict associated deep infiltrating endometriosis before surgery for an endometrioma

STUDY QUESTION

Is it possible to detect associated deep infiltrating endometriosis (DIE) before surgery for patients operated on for endometriomas using a preoperative clinical symptoms questionnaire?

SUMMARY ANSWER

A diagnostic score of DIE associated with endometriomas using four clinical symptoms defined a high-risk group where the probability of DIE was 88% and a low-risk group with a 10% probability of DIE.

WHAT IS KNOWN ALREADY

Many clinical symptoms are already known to be associated with DIE but they have not yet been used to build a clinical prediction model.

STUDY DESIGN, SIZE, DURATION

We built a diagnostic score of DIE based on a case control study of 326 consecutive patients operated on for an endometrioma between January 2005 and October 2011: 164 had associated DIE (DIE+) and 162 had no DIE (DIE-). We derived the score on a training sample obtained from a random selection of 2/3 of the population (211 patients, 101 DIE+, 110 DIE-), and validated the results on the remaining third (115 patients, 63 DIE+, 52 DIE-). The gold standard for the diagnosis of DIE was based on surgical exploration and histological diagnosis.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Participants were consecutive patients aged 18-42 years who underwent surgery for an endometrioma with histological confirmation and complete treatment of their endometriotic lesions: data for these women were extracted from a prospective database including a standardized preoperative questionnaire. On the training dataset, variables associated with DIE in a univariate analysis were introduced in a multiple logistic regression and selected by a backward stepwise procedure and a Jackknife procedure. A diagnostic score of DIE was built with the scaled/rounded coefficients of the multiple regression. Two cut-off values delimitated a high and a low risk group, and their diagnostic accuracy was tested on the validation dataset.

MAIN RESULTS AND THE ROLE OF CHANCE

Four variables were independently associated with DIE: visual analogue scale of gastro-intestinal symptoms ≥5 or of deep dyspareunia >5 (adjusted diagnostic odds ratio (aDOR) = 6.0, 95% confidence interval (CI) [2.9-12.1]), duration of pain greater than 24 months (aDOR = 3.8, 95% CI [1.9-7.7]), severe dysmenorrhoea (defined as the prescription of the oral contraceptive pill for the treatment of a primary dysmenorrhoea or the worsening of a secondary dysmenorrhoea) (aDOR = 3.8, 95% CI [1.9-7.6]) and primary or secondary infertility (aDOR = 2.5, 95% CI [1.2-4.9]). The sum of these variables weighted by their rounded/scaled coefficients constituted the score ranging from 0 to 53. A score <13 defined a low-risk group where the probability of DIE was 10% (95% CI [7-15] with a sensitivity of 95% (95% CI [89-98]) and a negative likelihood ratio of 0.1 (95% CI [0.0-0.3]). A score ≥35 defined a high-risk group where the probability of DIE was 88% (95% CI [83-92%]), with a specificity of 94% (95% CI [87-97]), and a positive likelihood ratio of 8.1 (95% CI [3.9-17.0]). The performance of the score was confirmed on the validation dataset with 11% of DIE+ patients having a score <13 (sensibility: 95%) and 90% of DIE+ patients having a score ≥35 (specificity: 94%).

LIMITATION, REASONS FOR CAUTION

This study was performed in a department specialized in DIE management. Score accuracy could be different in less specialized centres.

WIDER IMPLICATIONS OF THE FINDINGS

This score could have a major clinical impact on the time of diagnosis, the management of DIE and could reduce the cost of investigations by helping to identify high-risk patients, while preserving the quality of care.

STUDY FUNDING/COMPETING INTERESTS

The authors have no competing interests to declare. No grant supported the study.

Validation of dengue infection severity score

Objective

To validate a simple scoring system to classify dengue viral infection severity to patients in different settings.

Methods

The developed scoring system derived from 777 patients from three tertiary-care hospitals was applied to 400 patients in the validation data obtained from another three tertiary-care hospitals. Percentage of correct classification, underestimation, and overestimation was compared. The score discriminative performance in the two datasets was compared by analysis of areas under the receiver operating characteristic curves.

Results

Patients in the validation data were different from those in the development data in some aspects. In the validation data, classifying patients into three severity levels (dengue fever, dengue hemorrhagic fever, and dengue shock syndrome) yielded 50.8% correct prediction (versus 60.7% in the development data), with clinically acceptable underestimation (18.6% versus 25.7%) and overestimation (30.8% versus 13.5%). Despite the difference in predictive performances between the validation and the development data, the overall prediction of the scoring system is considered high.

Conclusion

The developed severity score may be applied to classify patients with dengue viral infection into three severity levels with clinically acceptable under- or overestimation. Its impact when used in routine clinical practice should be a topic for further study.

Validation of a clinical risk-scoring algorithm for severe scrub typhus

OBJECTIVE

The aim of the study reported here was to validate the risk-scoring algorithm for prognostication of scrub typhus severity.

METHODS

The risk-scoring algorithm for prognostication of scrub typhus severity developed earlier from two general hospitals in Thailand was validated using an independent dataset of scrub typhus patients in one of the hospitals from a few years later. The predictive performances of the two datasets were compared by analysis of the area under the receiver-operating characteristic curve (AuROC). Classification of patients into non-severe, severe, and fatal cases was also compared.

RESULTS

The proportions of non-severe, severe, and fatal patients by operational definition were similar between the development and validation datasets. Patient, clinical, and laboratory profiles were also similar. Scores were similar in both datasets, both in terms of discriminating non-severe from severe and fatal patients (AuROC =88.74% versus 91.48%, P=0.324), and in discriminating fatal from severe and non-severe patients (AuROC =88.66% versus 91.22%, P=0.407). Over- and under-estimations were similar and were clinically acceptable.

CONCLUSION

The previously developed risk-scoring algorithm for prognostication of scrub typhus severity performed similarly with the validation data and the first dataset. The scoring algorithm may help in the prognostication of patients according to their severity in routine clinical practice. Clinicians may use this scoring system to help make decisions about more intensive investigations and appropriate treatments.

Clinical risk-scoring algorithm to forecast scrub typhus severity

Background

The study explored clinical risk characteristics that may be used to forecast scrub typhus severity under routine clinical practices.

Methods

Retrospective data were collected from patients registered at two university-affiliated tertiary care hospitals in the north of Thailand, from 2004 to 2010. Key information was retrieved from in-patient records, out patient cards, laboratory reports and registers. Patients were classified into three severity groups: nonsevere, severe (those with at least one organ involvement), and deceased. Prognostic characteristics for scrub typhus severity were analyzed by a multivariable ordinal continuation ratio regression.

Results

A total of 526 patients were classified into nonsevere (n = 357), severe (n = 100), and deceased (n = 69). The significant multivariable prognostic characteristics for scrub typhus severity were increased body temperature (odds ratio [OR] = 0.58, 95% confidence interval [CI] = 0.45–0.74, P < 0.001), increased pulse rate (OR = 1.03, 95% CI = 1.01–1.05, P < 0.001), presence of crepitation (OR = 3.25, 95% CI = 1.52–6.96, P = 0.001), increased percentage of lymphocytes (OR = 0.97, 95% CI = 0.95–0.98, P = 0.001), increased aspartate aminotransferase (every 10 IU/L) (OR = 1.04, 95% CI = 1.02–1.06, P < 0.001), increased serum albumin (OR = 0.47, 95% CI = 0.27–0.80, P = 0.001), increased serum creatinine (OR = 1.83, 95% CI = 1.50–2.24, P < 0.001), and increased levels of positive urine albumin (OR = 1.43, 95% CI = 1.17–1.75, P < 0.001).

Conclusion

Patients suspicious of scrub typhus with low body temperature, rapid pulse rate, presence of crepitation, low percentage of lymphocyte, low serum albumin, elevated aspartate aminotransferase, elevated serum creatinine, and positive urine albumin should be monitored closely for severity progression.

Clinical risk-scoring algorithm to forecast scrub typhus severity

Purpose

To develop a simple risk-scoring system to forecast scrub typhus severity.

Patients and methods

Seven years’ retrospective data of patients diagnosed with scrub typhus from two university-affiliated hospitals in the north of Thailand were analyzed. Patients were categorized into three severity groups: nonsevere, severe, and dead. Predictors for severity were analyzed under multivariable ordinal continuation ratio logistic regression. Significant coefficients were transformed into item score and summed to total scores.

Results

Predictors of scrub typhus severity were age >15 years, (odds ratio [OR] =4.09), pulse rate >100/minute (OR 3.19), crepitation (OR 2.97), serum aspartate aminotransferase >160 IU/L (OR 2.89), serum albumin ≤3.0 g/dL (OR 4.69), and serum creatinine >1.4 mg/dL (OR 8.19). The scores which ranged from 0 to 16, classified patients into three risk levels: non-severe (score ≤5, n=278, 52.8%), severe (score 6–9, n=143, 27.2%), and fatal (score ≥10, n=105, 20.0%). Exact severity classification was obtained in 68.3% of cases. Underestimations of 5.9% and overestimations of 25.8% were clinically acceptable.

Conclusion

The derived scrub typhus severity score classified patients into their severity levels with high levels of prediction, with clinically acceptable under- and overestimations. This classification may assist clinicians in patient prognostication, investigation, and management. The scoring algorithm should be validated by independent data before adoption into routine clinical practice.

Pulmonary embolism severity index accurately predicts long-term mortality rate in patients hospitalized for acute pulmonary embolism

BACKGROUND

The Pulmonary Embolism (PE) Severity Index (PESI) is a clinical prognostic rule that accurately classifies PE patients into five risk classes with increasing mortality. PESI score has been validated in studies with a relatively short-term follow-up and its accuracy in predicting long-term prognosis has never been established.

METHODS

Consecutive patients admitted to the tertiary care hospital of Varese (Italy) with an objectively diagnosed PE between January 2005 and December 2009 were retrospectively included. Information on clinical presentation, diagnostic work-up, risk factors, treatment and mortality during a 1-year follow-up was collected.

RESULTS

Five hundred and thirty-eight patients were enrolled in this study. The mean age was 70.6 years (± SD 15.2), 44.4% of patients were male, and 27.9% had known cancer. One-year follow-up was available for 96.1% of patients. The overall mortality rate was 23.2% at 3 months, 30.2% at 6 months and 37.1% at 12 months. The discriminatory power of the PESI score to predict long-term mortality, expressed as the area under the ROC curve, was 0.77 (95%CI, 0.72-0.81) at 3 months, 0.77 (95%CI, 0.73-0.81) at 6 months and 0.79 (95%CI, 0.75-0.82) at 12 months. The PESI score confirmed its accurate prediction in patients without cancer. Simplified PESI had a similar overall accuracy to the original PESI at 3 and 6 months, but this was significantly lower at 1 year.

CONCLUSIONS

The results of this study suggest that PESI score may also be an accurate tool to define the 6-month and 1-year mortality rates in PE patients.

Nurse and physician agreement in the assessment of minor blunt head trauma

OBJECTIVE

The Pediatric Emergency Care Applied Research Network (PECARN) traumatic brain injury (TBI) clinical prediction rules identify children with minor blunt head trauma who are at low risk for clinically important traumatic brain injuries. We measured the agreement between the registered nurse (RN) and physician (MD) assessments.

METHODS

We performed a cross-sectional study of all children <18 years of age with minor blunt head trauma who presented to a single emergency department. RNs and MDs independently assessed each child and recorded age-based PECARN predictors. As symptoms can change over time, we included cases only when both evaluations were completed within 60 minutes. We used the κ statistic to measure RN-MD agreement, with the main analysis focusing on the overall PECARN rule agreement.

RESULTS

Of the 1624 eligible children, 1191 (73%) had evaluations completed by both RN and ED providers, of which 437 (37%) were in children <2 years of age. The median time between completions of the provider forms was 12 minutes (interquartile range 4-25 minutes). The overall agreement between the RN and MD was higher for the older children (κ 0.55, 95% confidence interval 0.49-0.61 for children 2-18 years versus κ 0.32, 95% confidence interval 0.23-0.41 for children <2 years).

CONCLUSIONS

The overall agreement between RN and MD for the PECARN TBI prediction rules was moderate for older children and fair for younger children. Initial RN assessments should be verified by the MD before clinical application, especially for the youngest children.

Diagnostic and prognostic prediction models

Risk prediction models can be used to estimate the probability of either having (diagnostic model) or developing a particular disease or outcome (prognostic model). In clinical practice, these models are used to inform patients and guide therapeutic management. Examples from the field of venous thrombo-embolism (VTE) include the Wells rule for patients suspected of deep venous thrombosis and pulmonary embolism, and more recently prediction rules to estimate the risk of recurrence after a first episode of unprovoked VTE. In this paper, the three phases that are recommended before a prediction model may be used in daily practice are described: development, validation, and impact assessment. In the development phase, the focus is on model development commonly using a multivariable logistic (diagnostic) or survival (prognostic) regression analysis. The performance of the developed model is expressed by discrimination, calibration and (re-) classification. In the validation phase, the developed model is tested in a new set of patients using these same performance measures. This is important, as model performance is commonly poorer in a new set of patients, e.g. due to case-mix or domain differences. Finally, in the impact phase the ability of a prediction model to actually guide patient management is evaluated. Whereas in the development and validation phase single cohort designs are preferred, this last phase asks for comparative designs, ideally randomized designs; therapeutic management and outcomes after using the prediction model is compared to a control group not using the model (e.g. usual care).

Prognostic value of the Geneva prediction rule in patients with pulmonary embolism

BACKGROUND

Assessment of pre-test probability of pulmonary embolism (PE) and prognostic stratification are two widely recommended steps in the management of patients with suspected PE. Some items of the Geneva prediction rule may have a prognostic value. We analyzed whether the initial probability assessed by the Geneva rule was associated with the outcome of patients with PE.

METHODS

In a post-hoc analysis of a multicenter trial including 1,693 patients with suspected PE, the all-cause death or readmission rates during the 3-month follow-up of patients with confirmed PE were analyzed. PE probability group was prospectively assessed by the revised Geneva score (RGS). Similar analyses were made with the a posteriori-calculated simplified Geneva score (SGS).

RESULTS

PE was confirmed in 357 patients and 21 (5.9%) died during the 3-month follow-up. The mortality rate differed significantly with the initial RGS group, as with the SGS group. For the RGS, the mortality increased from 0% (95% Confidence Interval: [0-5.4%]) in the low-probability group to 14.3% (95% CI: [6.3-28.2%]) in the high-probability group, and for the SGS, from 0% (95% CI: [0-5.4%] to 17.9% (95% CI: [7.4-36%]). Readmission occurred in 58 out of the 352 patients with complete information on readmission (16.5%). No significant change of readmission rate was found among the RGS or SGS groups.

CONCLUSIONS

Returning to the initial PE probability evaluation may help clinicians predict 3-month mortality in patients with confirmed PE. (ClinicalTrials.gov: NCT00117169).

The impact of maternal characteristics on the moderately premature infant: an antenatal maternal transport clinical prediction rule

OBJECTIVE

Moderately premature infants, defined here as those born between 30⁰/₇ and 34⁶/₇ weeks gestation, comprise 3.9% of all births in the United States and 32% of all preterm births. Although long-term outcomes for these infants are better than for less mature infants, morbidity and mortality are still substantially increased in comparison with infants born at term. There is an added survival benefit resulting from birth at a tertiary neonatal care center, and although many of these infants require tertiary level care, delivery at lower level hospitals and subsequent neonatal transfer are still common. Our primary aim was to determine the impact of maternal characteristics and antenatal medical management on the early neonatal course of the moderately premature infant. The secondary aim was to create a clinical prediction rule to determine which infants require intubation and mechanical ventilation in the first 24 h of life. Such a prediction rule could inform the decision to transfer maternal-fetal patients before delivery to a facility with a Level III neonatal intensive care unit (NICU), where optimal care could be provided without the requirement for a neonatal transfer.

STUDY DESIGN

Data for this analysis came from the cohort of infants in the Moderately Premature Infant Project (MPIP) database, a multicenter cohort study of 850 infants born at gestational age 30⁰/₇ and 34⁶/₇ weeks, with birth weight between 591 to 3540 g. [corrected], who were discharged to home alive. We built a logistic regression model to identify maternal characteristics associated with need for tertiary care, as measured by administration of surfactant. Using statistically significant covariates from this model, we then created a numerical decision rule to predict need for tertiary care.

RESULT

In multivariate modeling, four factors were associated with reduction in the need for tertiary care, including non-White race (odds ratio (OR)=0.5, (0.3, 0.7)), older gestational age, female gender (OR=0.6 (0.4, 0.8)) and use of antenatal corticosteroids (OR=0.5, (0.3, 0.8)). The clinical prediction rule to discriminate between infants who received surfactant, versus those who did not, had an area under the curve of 0.77 (0.73, 0.8).

CONCLUSION

Four antenatal risk factors are associated with a requirement for Level III NICU care as defined by the need for surfactant administration. Future analyses will examine a broader spectrum of antenatal characteristics and revalidate the prediction rule in an independent cohort.

Derivation and validation of a clinical prediction rule for delirium in patients admitted to a medical ward: an observational study

OBJECTIVES

To develop and validate a simple clinical prediction rule, based on variables easily measurable at admission, to identify patients at high risk of developing delirium during their hospital stay on an internal medicine ward.

DESIGN

Prospective study of two cohorts of patients admitted between 1 May and 30 June 2008 (derivation cohort), and between 1 May and 30 June 2009 (validation cohort).

SETTING

A tertiary hospital in Donostia-Gipuzkoa (Spain).

PARTICIPANTS

In total 397 patients participated in the study. The mean age and incidence of delirium were 75.9 years and 13%, respectively, in the derivation cohort, and 75.8 years and 25% in the validation cohort.

MAIN OUTCOME MEASURES

The predictive variables analysed and finally included in the rule were: being aged 85 years old or older, being dependent in five or more activities of daily living, and taking two or more psychotropic drugs (antipsychotics, benzodiazepines, antidepressants, anticonvulsant and/or antidementia drugs). The variable of interest was delirium as defined by the short Confusion Assessment Method, which assesses four characteristics: acute onset and fluctuating course, inattention, disorganised thinking and altered level of consciousness.

RESULTS

We developed a rule in which the individual risk of delirium is obtained by adding one point for each criterion met (age≥85, high level of dependence, and being on psychotropic medication). The result is considered positive if the score is ≥1. The rule accuracy was: sensitivity=93.4% (95% CI 85.5% to 97.2%), specificity=60.6% (95% CI 54.1% to 66.8%), positive predictive value=44.4% (95% CI 36.9% to 52.1%) and negative predictive value=96.5% (95% CI: 92% to 98.5%). The area under the receiver operator characteristic (ROC) curve was 0.85 for the validation cohort.

CONCLUSIONS

The presence or absence of any of the three predictive factors (age≥85, high level of dependence and psychotropic medication) allowed us to classify patients on internal medicine wards according to the risk of developing delirium. The simplicity of the variables in our clinical prediction rule means that the data collection required is feasible in busy medicine units.

Predictors of very low adherence with medications for osteoporosis: towards development of a clinical prediction rule

We developed a clinical prediction rule score to predict medication non-adherence for women prescribed osteoporosis treatment. When combined into a summative score, 62% with seven or more points on the score demonstrated very low adherence. This compares with 17% subjects with fewer than seven points (c-statistic = 0.74).

INTRODUCTION

Medication non-adherence is extremely common for osteoporosis; however, no clear methods exist for identifying patients at risk of this behavior. We developed a clinical prediction rule to predict medication non-adherence for women prescribed osteoporosis treatment.

METHODS

Women undergoing bone mineral density testing and fulfilling WHO criteria for osteoporosis were invited to complete a questionnaire and then followed for 1 year. Adjusted logistic regression models were examined to identify variables associated with very low adherence (medication possession ratio <20%). The weighted variables, based on the logistic regression, were summed, and the score was compared with the proportion of subjects with very low adherence.

RESULTS

One hundred forty two women participated in the questionnaire and were prescribed an osteoporosis medication. After 1 year, 36% (n = 50) had very low adherence. Variables associated with very low adherence included prior non-adherence with chronic medications, agreement that side effects are concerning, agreement that she is taking too many medications, lack of agreement that osteoporosis is a worry, lack of agreement that a fracture will cause disability, lack of agreement that medications help her stay active, and frequent use of alcohol. When combined into a summative score, 36 of the 58 subjects (62%) with seven or more points on the score demonstrated very low adherence. This compares with 14 of the 84 (17%) subjects with fewer than seven points (c-statistic = 0.74).

CONCLUSION

We developed a brief clinical prediction rule that was able to discriminate between women likely (and unlikely) to experience very low adherence with osteoporosis medications.

Validation of a clinical rule to predict complications of acute cough in preschool children: a prospective study in primary care

BACKGROUND

Few clinical rules have been derived let alone validated in primary care. A rule was derived to predict complications of acute cough in preschool children presenting to primary care. The clinical rule used the presence/absence of fever and/or chest signs to distinguish children at low, medium, and high risk of complications.

AIM

To validate a clinical rule for predicting complications of acute cough in preschool children in primary care.

DESIGN OF STUDY

Prospective cohort study.

SETTING

Thirteen general practices in Bristol and Tayside, UK.

METHOD

Preschool children with cough up to 28 days and without asthma were recruited. The same sociodemographic, clinical history, examination, and complications data as for the derivation study were collected. First, univariable logistic regression was used to explore the associations with complications, and then predictors with stronger relationships (P<0.2) were modelled using multivariable logistic regression. These predictors were compared with derivation predictors with respect to their strength of association with complications. The derivation predictors were used in the validation dataset to allow comparison of the post-test probabilities of complications between derivation and validation studies.

RESULTS

The presence of fever and chest signs in the validation study tended to be protective for complications, with univariable odds ratios (ORs) of 0.37 and 0.81 respectively, compared with ORs of 4.86 and 2.72 in the derivation study. However, 95% confidence limits were wide and evidence for two other possible reasons for these results were found: spectrum bias and confounding by indication.

CONCLUSION

No evidence was found to validate the clinical rule for predicting complications of acute cough, possibly as a result of spectrum bias, confounding by indication, and/or chance. As paediatric infectious illness is costly and associated with high rates of antibiotic use, further research is needed to derive and validate prediction rules.

Development and validation of a clinical prediction rule for angiotensin-converting enzyme inhibitor-induced cough

BACKGROUND

Angiotensin-converting enzyme inhibitors are effective for many cardiovascular diseases and are widely prescribed, but cough sometimes necessitates their withdrawal.

OBJECTIVE

To develop and validate a model that predicts, by using information available at first prescription, whether a patient will develop cough within 6 months.

DESIGN

Retrospective cohort study with derivation and validation sets.

SETTING

Outpatient clinics affiliated with an urban tertiary care hospital.

PATIENTS

Clinical data were collected from electronic charts. The derivation set included 1125 patients and the validation set included 567 patients.

INTERVENTIONS

None.

MEASUREMENTS

Angiotensin-converting enzyme inhibitor-induced cough assessed by predetermined criteria.

RESULTS

In the total cohort, 12% of patients developed angiotensin-converting enzyme inhibitor-induced cough. Independent multivariate predictors of cough were older age, female gender, non-African American (with East Asian having highest risk), no history of previous angiotensin-converting enzyme inhibitor use, and history of cough due to another angiotensin-converting enzyme inhibitor. Patients with a history of angiotensin-converting enzyme inhibitor-induced cough were 29 times more likely to develop a cough than those without this history. These factors were used to develop a model stratifying patients into 4 risk groups. In the derivation set, low-risk, average-risk, intermediate-risk, and high-risk groups had a 6%, 9%, 22%, and 55% probability of cough, respectively. In the validation set, 4%, 14%, 20%, and 60% of patients in these 4 groups developed cough, respectively.

CONCLUSIONS

This model may help clinicians predict the likelihood of a particular patient developing cough from an angiotensin-converting enzyme inhibitor at the time of prescribing, and may also assist with subsequent clinical decisions.