Search

Prediction models aim to use available data to predict a health state or outcome that has not yet been observed. Prediction is primarily relevant to clinical practice, but is also used in research, and administration. While prediction modeling involves estimating the relationship between patient factors and outcomes, it is distinct from casual inference. Prediction modeling thus requires unique considerations for development, validation, and updating. This document represents an effort from editors at 31 respiratory, sleep, and critical care medicine journals to consolidate contemporary best practices and recommendations related to prediction study design, conduct, and reporting. Herein, we address issues commonly encountered in submissions to our various journals. Key topics include considerations for selecting predictor variables, operationalizing variables, dealing with missing data, the importance of appropriate validation, model performance measures and their interpretation, and good reporting practices. Supplemental discussion covers emerging topics such as model fairness, competing risks, pitfalls of "modifiable risk factors", measurement error, and risk for bias. This guidance is not meant to be overly prescriptive; we acknowledge that every study is different, and no set of rules will fit all cases. Additional best practices can be found in the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines, to which we refer readers for further details.

Background & aimsA comprehensive analysis of changes in symptoms and functioning during and after direct-acting antiviral (DAA) therapy for chronic hepatitis C virus (HCV) infection has not been conducted for patients treated in real-world clinical settings. Therefore, we evaluated patient-reported outcomes (PROs) in a diverse cohort of patients with HCV treated with commonly prescribed DAAs.MethodsPROP UP is a US multicenter observational study of 1,601 patients with HCV treated with DAAs in 2016-2017. PRO data were collected at baseline (T1), early on-treatment (T2), late on-treatment (T3) and 3-months post-treatment (T4). PRO mean change scores were calculated from baseline and a minimally important change (MIC) threshold was set at 5%. Regression analyses investigated patient and treatment characteristics independently associated with PRO changes on-treatment and post-treatment.ResultsOf 1,564 patients, 55% were male, 39% non-white, 47% had cirrhosis. Sofosbuvir/ledipasvir was prescribed to 63%, sofosbuvir/velpatasvir to 21%, grazoprevir/elbasvir to 11%, and paritaprevir/ombitasvir/ritonavir + dasabuvir to 5%. During DAA therapy, mean PRO scores improved slightly in the overall cohort, but did not reach the 5% MIC threshold. Between 21-53% of patients experienced >5% improved PROs while 23-36% experienced >5% worse symptoms. Of 1,410 patients with evaluable sustained virologic response (SVR) data, 95% achieved SVR. Among those with SVR, all mean PRO scores improved, with the 5% MIC threshold met for fatigue, sleep disturbance, and functioning well-being. Regression analyses identified subgroups, defined by age 35-55, baseline mental health issues and a higher number of health comorbidities as predictors of PRO improvements.ConclusionsIn real-world clinical practices, we observed heterogeneous patient experiences during and after DAA treatment. Symptom improvements were more pronounced in younger patients, those with baseline mental health issues and mu

BackgroundSymptom burden, medical comorbidities, and functional well-being of patients with chronic hepatitis C virus (HCV) initiating direct acting antiviral (DAA) therapy in real-world clinical settings are not known. We characterized these patient-reported outcomes (PROs) among HCV-infected patients and explored associations with sociodemographic, liver disease, and psychiatric/substance abuse variables.Methods and findingsPROP UP is a large US multicenter observational study that enrolled 1,600 patients with chronic HCV in 2016-2017. Data collected prior to initiating DAA therapy assessed the following PROs: number of medical comorbidities; neuropsychiatric, somatic, gastrointestinal symptoms (PROMIS surveys); overall symptom burden (Memorial Symptom Assessment Scale); and functional well-being (HCV-PRO). Candidate predictors included liver disease markers and patient-reported sociodemographic, psychiatric, and alcohol/drug use features. Predictive models were explored using a random selection of 700 participants; models were then validated with data from the remaining 900 participants. The cohort was 55% male, 39% non-white, 48% had cirrhosis (12% with advanced cirrhosis); 52% were disabled or unemployed; 63% were on public health insurance or uninsured; and over 40% had markers of psychiatric illness. The median number of medical comorbidities was 4 (range: 0-15), with sleep disorders, chronic pain, diabetes, joint pain and muscle aches being present in 20-50%. Fatigue, sleep disturbance, pain and neuropsychiatric symptoms were present in over 60% and gastrointestinal symptoms in 40-50%. In multivariable validation models, the strongest and most frequent predictors of worse PROs were disability, unemployment, and use of psychiatric medications, while liver markers generally were not.ConclusionsThis large multi-center cohort study provides a comprehensive and contemporary assessment of the symptom burden and comorbid medical conditions in patients with HCV trea

BackgroundSymptom burden, medical comorbidities, and functional well-being of patients with chronic hepatitis C virus (HCV) initiating direct acting antiviral (DAA) therapy in real-world clinical settings are not known. We characterized these patient-reported outcomes (PROs) among HCV-infected patients and explored associations with sociodemographic, liver disease, and psychiatric/substance abuse variables.Methods and findingsPROP UP is a large US multicenter observational study that enrolled 1,600 patients with chronic HCV in 2016-2017. Data collected prior to initiating DAA therapy assessed the following PROs: number of medical comorbidities; neuropsychiatric, somatic, gastrointestinal symptoms (PROMIS surveys); overall symptom burden (Memorial Symptom Assessment Scale); and functional well-being (HCV-PRO). Candidate predictors included liver disease markers and patient-reported sociodemographic, psychiatric, and alcohol/drug use features. Predictive models were explored using a random selection of 700 participants; models were then validated with data from the remaining 900 participants. The cohort was 55% male, 39% non-white, 48% had cirrhosis (12% with advanced cirrhosis); 52% were disabled or unemployed; 63% were on public health insurance or uninsured; and over 40% had markers of psychiatric illness. The median number of medical comorbidities was 4 (range: 0-15), with sleep disorders, chronic pain, diabetes, joint pain and muscle aches being present in 20-50%. Fatigue, sleep disturbance, pain and neuropsychiatric symptoms were present in over 60% and gastrointestinal symptoms in 40-50%. In multivariable validation models, the strongest and most frequent predictors of worse PROs were disability, unemployment, and use of psychiatric medications, while liver markers generally were not.ConclusionsThis large multi-center cohort study provides a comprehensive and contemporary assessment of the symptom burden and comorbid medical conditions in patients with HCV trea

Background. We examined whether viral dynamics in the genital tract during the natural history of acute human immunodeficiency virus type 1 (HIV-1) infection could explain efficient heterosexual transmission of HIV. Methods. We measured HIV-1 concentration in blood and semen samples from patients with acute and long-term HIV-1 infection. We explored the effect of changes in viral dynamics in semen on the probability of transmission per coital act, using a probabilistic model published elsewhere. Results. Considered over time from infection, semen HIV-1 concentrations, in men with acute infection, increase and decrease in approximate parallel with changes occurring in blood. Modeling suggests that these acute dynamics alone are sufficient to increase probability of heterosexual transmission by 8-10-fold between peak (day 20 after infection, based on the model) and virologic set points (day 54 and later after infection). Depending on the frequency of coitus, men with average semen HIV-1 loads and without sexually transmitted diseases (STDs) would be expected to infect 7%-24% of susceptible female sex partners during the first 2 months of infection. The predicted infection rate would be much higher when either partner has an STD. Conclusions. Empirical biological data strongly support the hypothesis that sexual transmission by acutely infected individuals has a disproportionate effect on the spread of HIV-1 infection. Acute hyperinfectiousness may, in part, explain the current pandemic in heterosexual individuals

Impaired hepatic bile acid export may contribute to development of cholestatic drug-induced liver injury (DILI). The multidrug resistance-associated proteins (MRP) 3 and 4 are postulated to be compensatory hepatic basolateral bile acid efflux transporters when biliary excretion by the bile salt export pump (BSEP) is impaired. BSEP inhibition is a risk factor for cholestatic DILI. This study aimed to characterize the relationship between MRP3, MRP4, and BSEP inhibition and cholestatic potential of drugs. The inhibitory effect of 88 drugs (100 mu M) on MRP3- and MRP4-mediated substrate transport was measured in membrane vesicles. Drugs selected for investigation included 50 BSEP non-inhibitors (24 non-cholestatic; 26 cholestatic) and 38 BSEP inhibitors (16 non-cholestatic; 22 cholestatic). MRP4 inhibition was associated with an increased risk of cholestatic potential among BSEP non-inhibitors. In this group, for each 1% increase in MRP4 inhibition, the odds of the drug being cholestatic increased by 3.1%. Using an inhibition cutoff of 21%, which predicted a 50% chance of cholestasis, 62% of cholestatic drugs inhibited MRP4 (P < 0.05); in contrast, only 17% of non-cholestatic drugs were MRP4 inhibitors. Among BSEP inhibitors, MRP4 inhibition did not provide additional predictive value of cholestatic potential; almost all BSEP inhibitors were also MRP4 inhibitors. Inclusion of pharmacokinetic predictor variables (e. g., maximal unbound concentration in plasma) in addition to percent MRP4 inhibition in logistic regression models did not improve cholestasis prediction. Association of cholestasis with percent MRP3 inhibition was not statistically significant, regardless of BSEP-inhibition status. Inhibition of MRP4, in addition to BSEP, may be a risk factor for the development of cholestatic DILI.

Introduction: Triage of patients is critical to patient safety, yet no clear information exists as to the utility of initial vital signs in identifying critically ill older emergency department (ED) patients. The objective of this study is to evaluate a set of initial vital sign thresholds as predictors of severe illness and injury among older adults presenting to the ED.Methods: We reviewed all visits by patients aged 75 and older seen during 2007 at an academic ED serving a large community of older adults. Patients’ charts were abstracted for demographic and clinical information including vital signs, via automated electronic methods. We used bivariate analysis to investigate the relationship between vital sign abnormalities and severe illness or injury, defined as intensive care unit (ICU) admission or ED death. In addition, we calculated likelihood ratios for normal and abnormal vital signs in predicting severe illness or injury.Results: 4,873 visits by patients aged 75 and above were made to the ED during 2007, and of these 3,848 had a complete set of triage vital signs. For these elderly patients, the sensitivity and specificity of an abnormal vital sign taken at triage for predicting death or admission to an ICU were 73% (66,81) and 50% (48,52) respectively (positive likelihood ratio 1.47 (1.30,1.60); negative likelihood ratio 0.54 (0.30,0.60).Conclusion: Emergency provider assessment and triage scores that rely primarily on initial vital signs are likely to miss a substantial portion of critically ill older adults. [West J Emerg Med. 2013;14(5):453–460.]

Introduction: Triage of patients is critical to patient safety, yet no clear information exists as to the utility of initial vital signs in identifying critically ill older emergency department (ED) patients. The objective of this study is to evaluate a set of initial vital sign thresholds as predictors of severe illness and injury among older adults presenting to the ED.Methods: We reviewed all visits by patients aged 75 and older seen during 2007 at an academic ED serving a large community of older adults. Patients’ charts were abstracted for demographic and clinical information including vital signs, via automated electronic methods. We used bivariate analysis to investigate the relationship between vital sign abnormalities and severe illness or injury, defined as intensive care unit (ICU) admission or ED death. In addition, we calculated likelihood ratios for normal and abnormal vital signs in predicting severe illness or injury.Results: 4,873 visits by patients aged 75 and above were made to the ED during 2007, and of these 3,848 had a complete set of triage vital signs. For these elderly patients, the sensitivity and specificity of an abnormal vital sign taken at triage for predicting death or admission to an ICU were 73% (66,81) and 50% (48,52) respectively (positive likelihood ratio 1.47 (1.30,1.60); negative likelihood ratio 0.54 (0.30,0.60).Conclusion: Emergency provider assessment and triage scores that rely primarily on initial vital signs are likely to miss a substantial portion of critically ill older adults. [West J Emerg Med. 2013;14(5):453–460.]

PurposeCancer cachexia is a devastating and understudied illness in patients with head and neck squamous cell carcinoma (HNSCC). The primary objective was to identify clinical characteristics and serum levels of cytokines and cachexia-related factors in patients with HNSCC. The secondary objective was to detect the occurrence of cytokine and cachexia-related factor gene expression in HNSCC tumors.Experimental designFor the primary objective, cross-sectional data were obtained from prospectively recruited patients identified as cachexia cases and matching cachexia-free controls. For the secondary objective, a retrospective cohort design with matched controls was used.ResultsClinical characteristics associated with cancer cachexia in HNSCC were T(4) status (P = 0.01), increased C-reactive protein (P = 0.01), and decreased hemoglobin (P < 0.01). Exploratory multiplex analysis of serum cytokine levels found increased interleukin (IL)-6 (P = 0.04). A highly sensitive ELISA confirmed the multiplex result for increased IL-6 in cachectic patients (P = 0.02). Quality of life was substantially reduced in patients with cachexia compared with noncachectic patients (P < 0.01). All tumors of HNSCC patients both with and without cachexia expressed RNA for each cytokine tested and the cachexia factor lipid-mobilizing factor. There were no statistically significant differences between the cytokine and cachexia factor RNA expression of cachectic and noncachectic patients (each P > 0.05). No tumors expressed the cachexia factor proteolysis-inducing factor.ConclusionWe have identified clinical characteristics and pathophysiologic mechanisms associated with cancer cachexia in a carefully defined population of patients with HNSCC. The data suggest that the acute-phase response and elevated IL-6 are associated with this complex disease state. We therefore hypothesize that IL-6 may represent an important therapeutic target for HNSCC patients with cancer cachexia.

PurposeCancer cachexia is a devastating and understudied illness in patients with head and neck squamous cell carcinoma (HNSCC). The primary objective was to identify clinical characteristics and serum levels of cytokines and cachexia-related factors in patients with HNSCC. The secondary objective was to detect the occurrence of cytokine and cachexia-related factor gene expression in HNSCC tumors.Experimental designFor the primary objective, cross-sectional data were obtained from prospectively recruited patients identified as cachexia cases and matching cachexia-free controls. For the secondary objective, a retrospective cohort design with matched controls was used.ResultsClinical characteristics associated with cancer cachexia in HNSCC were T(4) status (P = 0.01), increased C-reactive protein (P = 0.01), and decreased hemoglobin (P < 0.01). Exploratory multiplex analysis of serum cytokine levels found increased interleukin (IL)-6 (P = 0.04). A highly sensitive ELISA confirmed the multiplex result for increased IL-6 in cachectic patients (P = 0.02). Quality of life was substantially reduced in patients with cachexia compared with noncachectic patients (P < 0.01). All tumors of HNSCC patients both with and without cachexia expressed RNA for each cytokine tested and the cachexia factor lipid-mobilizing factor. There were no statistically significant differences between the cytokine and cachexia factor RNA expression of cachectic and noncachectic patients (each P > 0.05). No tumors expressed the cachexia factor proteolysis-inducing factor.ConclusionWe have identified clinical characteristics and pathophysiologic mechanisms associated with cancer cachexia in a carefully defined population of patients with HNSCC. The data suggest that the acute-phase response and elevated IL-6 are associated with this complex disease state. We therefore hypothesize that IL-6 may represent an important therapeutic target for HNSCC patients with cancer cachexia.

Background. We examined whether viral dynamics in the genital tract during the natural history of acute human immunodeficiency virus type 1 (HIV-1) infection could explain efficient heterosexual transmission of HIV. Methods. We measured HIV-1 concentration in blood and semen samples from patients with acute and long-term HIV-1 infection. We explored the effect of changes in viral dynamics in semen on the probability of transmission per coital act, using a probabilistic model published elsewhere. Results. Considered over time from infection, semen HIV-1 concentrations, in men with acute infection, increase and decrease in approximate parallel with changes occurring in blood. Modeling suggests that these acute dynamics alone are sufficient to increase probability of heterosexual transmission by 8-10-fold between peak (day 20 after infection, based on the model) and virologic set points (day 54 and later after infection). Depending on the frequency of coitus, men with average semen HIV-1 loads and without sexually transmitted diseases (STDs) would be expected to infect 7%-24% of susceptible female sex partners during the first 2 months of infection. The predicted infection rate would be much higher when either partner has an STD. Conclusions. Empirical biological data strongly support the hypothesis that sexual transmission by acutely infected individuals has a disproportionate effect on the spread of HIV-1 infection. Acute hyperinfectiousness may, in part, explain the current pandemic in heterosexual individuals