Predicting short-term suicidal thoughts in adolescents using machine learning: developing decision tools to identify daily level risk after hospitalization.

BACKGROUND: Mobile technology offers unique opportunities for monitoring short-term suicide risk in daily life. In this study of suicidal adolescent inpatients, theoretically informed risk factors were assessed daily following discharge to predict near-term suicidal ideation and inform decision algorithms for identifying elevations in daily level risk, with implications for real-time suicide-focused interventions. METHODS: Adolescents (N = 78; 67.9% female) completed brief surveys texted daily for 4 weeks after discharge (n = 1621 observations). Using multi-level classification and regression trees (CARTSs) with repeated 5-fold cross-validation, we tested (a) a simple prediction model incorporating previous-day scores for each of 10 risk factors, and (b) a more complex model incorporating, for each of these factors, a time-varying person-specific mean over prior days together with deviation from that mean. Models also incorporated missingness and contextual (study week, day of the week) indicators. The outcome was the presence/absence of next-day suicidal ideation. RESULTS: The best-performing model (cross-validated AUC = 0.86) was a complex model that included ideation duration, hopelessness, burdensomeness, and self-efficacy to refrain from suicidal action. An equivalent model that excluded ideation duration had acceptable overall performance (cross-validated AUC = 0.78). Models incorporating only previous-day scores, with and without ideation duration (cross-validated AUC of 0.82 and 0.75, respectively), showed relatively weaker performance. CONCLUSIONS: Results suggest that specific combinations of dynamic risk factors assessed in adolescents' daily life have promising utility in predicting next-day suicidal thoughts. Findings represent an important step in the development of decision tools identifying short-term risk as well as guiding timely interventions sensitive to proximal elevations in suicide risk in daily life.

Targeted care navigation to reduce hospital readmissions in 'at-risk' patients.

BACKGROUND: Care navigation is commonly used to reduce preventable hospitalisation. The use of Electronic Health Record-derived algorithms may enable better targeting of this intervention for greater impact. AIMS: To evaluate if community-based Targeted Care Navigation, supported by an Electronic Health Record-derived readmission risk algorithm, is associated with reduced rehospitalisation. METHODS: A propensity score matching cohort (5 comparison to 1 intervention cohort ratio) study was conducted in an 850-bed Victorian public metropolitan health service, Australia, from May to November 2017. Admitted acute care patients with a non-surgical condition, identified as at-risk of hospital readmission using an Electronic Health Record-derived readmission risk algorithm provide by the state health department, were eligible. Targeted Care Navigation involved telephone follow-up support provided for 30 days post-discharge by a registered nurse. The hazard ratio for hospital readmission was calculated at 30, 60 and 90 days post-discharge using multivariable Cox Proportional Hazards regression. RESULTS: Sixty-five recipients received care navigation and were matched to 262 people who did not receive care navigation. Excellent matching was achieved with standardised differences between groups being <0.1 for all 11 variables included in the propensity score, including the readmission risk score. The Targeted Care Navigation group had a significantly reduced hazard of readmission at 30 days (hazard ratio 0.34; 95% confidence interval: 0.12, 0.94) compared with the comparison group. The effect size was reduced at 60 and 90 days post-discharge. CONCLUSION: We provide preliminary evidence that Targeted Care Navigation supported by an Electronic Health Record-derived readmission risk algorithm may reduce 30-day hospital readmissions.

A risk algorithm that predicts alcohol use disorders among college students.

The first year of college may carry especially high risk for onset of alcohol use disorders. We assessed the one-year incidence of alcohol use disorders (AUD) among incoming first-year students, predictors of AUD-incidence, prediction accuracy and population impact. A prospective cohort study of first-year college students (baseline: N = 5843; response rate = 51.8%; 1-year follow-up: n = 1959; conditional response rate = 41.6%) at a large university in Belgium was conducted. AUD were evaluated with the AUDIT and baseline predictors with the Composite International Diagnostic Interview Screening Scales (CIDI-SC). The one-year incidence of AUD was 3.9% (SE = 0.4). The most important individual-level baseline predictors of AUD incidence were being male (OR = 1.53; 95% CI = 1.12-2.10), a break-up with a romantic partner (OR = 1.67; 95% CI = 1.08-2.59), hazardous drinking (OR = 3.36; 95% CI = 1.31-8.63), and alcohol use characteristics at baseline (ORs between 1.29 and 1.38). Multivariate cross-validated prediction (cross-validated AUC = 0.887) shows that 55.5% of incident AUD cases occurred among the 10% of students at highest predicted risk (20.1% predicted incidence in this highest-risk subgroup). Four out of five students with incident AUD would hypothetically be preventable if baseline hazardous drinking was to be eliminated along with a reduction of one standard deviation in alcohol use characteristics scores, and another 15.0% would potentially be preventable if all 12-month stressful events were eliminated. Screening at college entrance is a promising strategy to identify students at risk of transitioning to more problematic drinking and AUD, thus improving the development and deployment of targeted preventive interventions.

Using "Markers of Harm" to Track Risky Gambling in Two Cohorts of Online Sports Bettors.

Online gambling poses novel risks for problem gambling, but also unique opportunities to detect and intervene with at-risk users. A consortium of gambling companies recently committed to using nine behavioral "Markers of Harm'' that can be calculated with online user data to estimate risk for gambling-related harm. The current study evaluates these markers in two independent samples of sports bettors, collected ten years apart. We find over a two-year period that most users never had high enough overall risk scores to indicate that they would have received an intervention. This observation is partly due to characteristics of our samples that are associated with lower risk for gambling-related harm, but might also be due to overly high risk thresholds or flaws in the design of some markers. Users with higher average risk scores had more intraindividual variability in risk scores. Younger age and male gender were not associated with higher average risk scores. The most active users were more likely than other users to have ever exceeded risk thresholds. Several risk scores significantly predicted proxies of gambling-related harm (e.g., account closure). Overall, the current Markers of Harm system has some correctable limitations that future risk detection systems should consider adopting.

Developing and validating a risk algorithm to diagnose Neisseria gonorrhoeae and Chlamydia trachomatis in symptomatic Rwandan women.

BACKGROUND: Algorithms that bridge the gap between syndromic sexually transmitted infection (STI) management and treatment based in realistic diagnostic options and local epidemiology are urgently needed across Africa. Our objective was to develop and validate a risk algorithm for Neisseria gonorrhoeae (NG) and Chlamydia trachomatis (CT) diagnosis among symptomatic Rwandan women and to compare risk algorithm performance to the current Rwandan National Criteria for NG/CT diagnosis. METHODS: The risk algorithm was derived in a cohort (n = 468) comprised of symptomatic women in Kigali who sought free screening and treatment for sexually transmitted infections and vaginal dysbioses at our research site. We used logistic regression to derive a risk algorithm for prediction of NG/CT infection. Ten-fold cross-validation internally validated the risk algorithm. We applied the risk algorithm to an external validation cohort also comprised of symptomatic Rwandan women (n = 305). Measures of calibration, discrimination, and screening performance of our risk algorithm compared to the current Rwandan National Criteria are presented. RESULTS: The prevalence of NG/CT in the derivation cohort was 34.6%. The risk algorithm included: age < =25, having no/primary education, not having full-time employment, using condoms only sometimes, not reporting genital itching, testing negative for vaginal candida, and testing positive for bacterial vaginosis. The model was well calibrated (Hosmer-Lemeshow p = 0.831). Higher risk scores were significantly associated with increased prevalence of NG/CT infection (p < 0.001). Using a cut-point score of > = 5, the risk algorithm had a sensitivity of 81%, specificity of 54%, positive predictive value (PPV) of 48%, and negative predictive value (NPV) of 85%. Internal and external validation showed similar predictive ability of the risk algorithm, which outperformed the Rwandan National Criteria. Applying the Rwandan National Criteria cutoff of > = 2 (the current cutoff) to our derivation cohort had a sensitivity of 26%, specificity of 89%, PPV of 55%, and NPV of 69%. CONCLUSIONS: These data support use of a locally relevant, evidence-based risk algorithm to significantly reduce the number of untreated NG/CT cases in symptomatic Rwandan women. The risk algorithm could be a cost-effective way to target treatment to those at highest NG/CT risk. The algorithm could also aid in sexually transmitted infection risk and prevention communication between providers and clients.

Development and validation of a prediction algorithm for the onset of common mental disorders in a working population.

OBJECTIVE: Common mental disorders are the most common reason for long-term sickness absence in most developed countries. Prediction algorithms for the onset of common mental disorders may help target indicated work-based prevention interventions. We aimed to develop and validate a risk algorithm to predict the onset of common mental disorders at 12 months in a working population. METHODS: We conducted a secondary analysis of the Household, Income and Labour Dynamics in Australia Survey, a longitudinal, nationally representative household panel in Australia. Data from the 6189 working participants who did not meet the criteria for a common mental disorders at baseline were non-randomly split into training and validation databases, based on state of residence. Common mental disorders were assessed with the mental component score of 36-Item Short Form Health Survey questionnaire (score ⩽45). Risk algorithms were constructed following recommendations made by the Transparent Reporting of a multivariable prediction model for Prevention Or Diagnosis statement. RESULTS: Different risk factors were identified among women and men for the final risk algorithms. In the training data, the model for women had a C-index of 0.73 and effect size (Hedges' g) of 0.91. In men, the C-index was 0.76 and the effect size was 1.06. In the validation data, the C-index was 0.66 for women and 0.73 for men, with positive predictive values of 0.28 and 0.26, respectively Conclusion: It is possible to develop an algorithm with good discrimination for the onset identifying overall and modifiable risks of common mental disorders among working men. Such models have the potential to change the way that prevention of common mental disorders at the workplace is conducted, but different models may be required for women.

Comparison of cardiovascular risk algorithms in patients with vs without rheumatoid arthritis and the role of C-reactive protein in predicting cardiovascular outcomes in rheumatoid arthritis.

Objectives: The aims were to compare the performance of cardiovascular risk calculators, Framingham Risk Score (FRS) and QRISK2, in RA and matched non-RA patients and to evaluate whether their performance could be enhanced by the addition of CRP. Methods: We conducted a retrospective analysis, using a clinical practice data set linked to Hospital Episode Statistics (HES) data from the UK. Patients presenting with at least one RA diagnosis code and no prior cardiovascular events were matched to non-RA patients using disease risk scores. The overall performance of the FRS and QRISK2 was compared between cohorts, and assessed with and without CRP in the RA cohort using C-Index, Akaike Information Criterion (AIC) and the net reclassification index (NRI). Results: Four thousand seven hundred and eighty RA patients met the inclusion criteria and were followed for a mean of 3.8 years. The C-Index for the FRS in the non-RA and RA cohort was 0.783 and 0.754 (P < 0.001) and that of the QRISK2 was 0.770 and 0.744 (P < 0.001), respectively. Log[CRP] was positively associated with cardiovascular events, but improvements in the FRS and QRISK2 C-Indices as a result of inclusion of CRP were small, from 0.764 to 0.767 (P = 0.026) for FRS and from 0.764 to 0.765 (P = 0.250) for QRISK2. The NRI was 3.2% (95% CI: -2.8, 5.7%) for FRS and -2.0% (95% CI: -5.8, 4.5%) for QRISK2. Conclusion: The C-Index for the FRS and QRISK2 was significantly better in the non-RA compared with RA patients. The addition of CRP in both equations was not associated with a significant improvement in reclassification based on NRI.

Perceptions of Barriers and Facilitators to a Pilot Implementation of an Algorithm-Supported Care Navigation Model of Care: A Qualitative Study.

We aimed to explore managerial and project staff perceptions of the pilot implementation of an algorithm-supported care navigation model, targeting people at risk of hospital readmission. The pilot was implemented from May to November 2017 at a Victorian health service (Australia) and provided to sixty-five patients discharged from the hospital to the community. All managers and the single clinician involved participated in a semi-structured interview. Participants (n = 6) were asked about their perceptions of the service design and the enablers and barriers to implementation. Interviews were transcribed verbatim and analysed according to a framework approach, using inductive and deductive techniques. Constructed themes included the following: an algorithm alone is not enough, the health service culture, leadership, resources and the perceived patient experience. Participants felt that having an algorithm to target those considered most likely to benefit was helpful but not enough on its own without addressing other contextual factors, such as the health service's capacity to support a large-scale implementation. Deductively mapping themes to the integrated Promoting Action on Research Implementation in Health Services (i-PARIHS) framework highlighted that a formal facilitation would be essential for future sustainable implementations. The systematic identification of barriers and enablers elicited critical information for broader implementations of algorithm-supported models of care.

Comparing a Data Entry Tool to Provider Insights Alone for Assessment of COVID-19 Hospitalization Risk: Pilot Matched Cohort Comparison Study.

BACKGROUND: In March 2020, the World Health Organization declared COVID-19 a global pandemic, necessitating an understanding of factors influencing severe disease outcomes. High COVID-19 hospitalization rates underscore the need for robust risk prediction tools to determine estimated risk for future hospitalization for outpatients with COVID-19. We introduced the "COVID-19 Risk Tier Assessment Tool" (CRTAT), designed to enhance clinical decision-making for outpatients. OBJECTIVE: We investigated whether CRTAT offers more accurate risk tier assignments (RTAs) than medical provider insights alone. METHODS: We assessed COVID-19-positive patients enrolled at Emory Healthcare's Virtual Outpatient Management Clinic (VOMC)-a telemedicine monitoring program, from May 27 through August 24, 2020-who were not hospitalized at the time of enrollment. The primary analysis included patients from this program, who were later hospitalized due to COVID-19. We retroactively formed an age-, gender-, and risk factor-matched group of nonhospitalized patients for comparison. Data extracted from clinical notes were entered into CRTAT. We used descriptive statistics to compare RTAs reported by algorithm-trained health care providers and those produced by CRTAT. RESULTS: Our patients were primarily younger than 60 years (67% hospitalized and 71% nonhospitalized). Moderate risk factors were prevalent (hospitalized group: 1 among 11, 52% patients; 2 among 2, 10% patients; and ≥3 among 4, 19% patients; nonhospitalized group: 1 among 11, 52% patients, 2 among 5, 24% patients, and ≥3 among 4, 19% patients). High risk factors were prevalent in approximately 45% (n=19) of the sample (hospitalized group: 11, 52% patients; nonhospitalized: 8, 38% patients). Approximately 83% (n=35) of the sample reported nonspecific symptoms, and the symptoms were generally mild (hospitalized: 12, 57% patients; nonhospitalized: 14, 67% patients). Most patient visits were seen within the first 1-6 days of their illness (n=19, 45%) with symptoms reported as stable over this period (hospitalized: 7, 70% patients; nonhospitalized: 3, 33% patients). Of 42 matched patients (hospitalized: n=21; nonhospitalized: n=21), 26 had identical RTAs and 16 had discrepancies between VOMC providers and CRTAT. Elements that led to different RTAs were as follows: (1) the provider "missed" comorbidity (n=6), (2) the provider noted comorbidity but undercoded risk (n=10), and (3) the provider miscoded symptom severity and course (n=7). CONCLUSIONS: CRTAT, a point-of-care data entry tool, more accurately categorized patients into risk tiers (particularly those hospitalized), underscored by its ability to identify critical factors in patient history and clinical status. Clinical decision-making regarding patient management, resource allocation, and treatment plans could be enhanced by using similar risk assessment data entry tools for other disease states, such as influenza and community-acquired pneumonia. The COVID-19 pandemic has accelerated the adoption of telemedicine, enabling remote patient tools such as CRTAT. Future research should explore the long-term impact of outpatient clinical risk assessment tools and their contribution to better patient care.

Early Detection of Pneumonia with the Help of Dementia in Geriatric Hip Fracture Patients.

OBJECTIVE: To investigate the role of dementia in pneumonia among geriatric patients with hip fracture and further develop an algorithm for stratifying risk of developing postoperative pneumonia. METHODS: The algorithm was developed after retrospectively analyzing 1344 hip fracture patients in the National Clinical Research Center for Orthopedics, Sports Medicine, and Rehabilitation from 1992 to 2012. Twenty-eight variables were analyzed for evaluating the ability to predict postoperative pneumonia. The validation of the algorithm was performed in the MIMIC-III database after enrolling 235 patients. RESULTS: One thousand five hundred and seventy-nine patients were enrolled, 4.69% patients had postoperative pneumonia in our hospital, and 17.02% suffered pneumonia in the MIMIC-III database. Dementia patients had more postoperative pneumonia (12.68% vs 4.24%, P = 0.0075), as compared with patients without dementia. The algorithm included nine predictors: dementia, age, coronary heart disease, the American Society of Anesthesiologists score, surgical method, mechanical ventilation, anemia, hypoproteinemia, and high creatinine. Internal validation showed the algorithm with dementia could improve predictive performance, while external validation found the algorithm with or without dementia both had similar and good predictive ability. CONCLUSIONS: The algorithm has the potential to be a pragmatic risk prediction tool to calculate risk of pneumonia in clinical practice and it may also be applicable in critically ill hip fracture patients with dementia.

Combinatorial Analysis of Circulating Biomarkers and Maternal Characteristics for Preeclampsia Prediction in the First and Third Trimesters in Asia.

We aim to establish a prediction model for pregnancy outcomes through a combinatorial analysis of circulating biomarkers and maternal characteristics to effectively identify pregnant women with higher risks of preeclampsia in the first and third trimesters within the Asian population. A total of two hundred and twelve pregnant women were screened for preeclampsia through a multicenter study conducted in four recruiting centers in Taiwan from 2017 to 2020. In addition, serum levels of sFlt-1/PlGF ratio, miR-181a, miR-210 and miR-223 were measured and transformed into multiples of the median. We thus further developed statistically validated algorithmic models by designing combinations of different maternal characteristics and biomarker levels. Through the performance of the training cohort (0.848 AUC, 0.73−0.96 95% CI, 80% sensitivity, 85% specificity, p < 0.001) and the validation cohort (0.852 AUC, 0.74−0.98 95% CI, 75% sensitivity, 87% specificity, p < 0.001) from one hundred and fifty-two women with a combination of miR-210, miR-181a and BMI, we established a preeclampsia prediction model for the first trimester. We successfully identified pregnant women with higher risks of preeclampsia in the first and third trimesters in the Asian population using the established prediction models that utilized combinatorial analysis of circulating biomarkers and maternal characteristics.

Predicting Damage to Hop Cones by Tetranychus urticae (Acari: Tetranychidae).

Twospotted spider mite (Tetranychus urticae Koch) is a cosmopolitan pest of numerous plants, including hop (Humulus lupulus L.). The most costly damage from the pest on hop results from infestation of cones, which are the harvested product, which can render crops unsalable if cones become discolored. We analyzed 14 yr of historical data from 312 individual experimental plots in western Oregon to identify risk factors associated with visual damage to hop cones from T. urticae. Logistic regression models were fit to estimate the probability of cone damage. The most predictive model was based on T. urticae-days during mid-July to harvest, which correctly predicted occurrence and nonoccurrence of cone damage in 91 and 93% of data sets, respectively, based on Youden's index. A second model based on the ratio of T. urticae to predatory arthropods late in the season correctly predicted cone damage in 92% of data sets and nonoccurrence of damage in 77% of data sets. The model based on T. urticae abundance performed similarly when validated in 23 commercial hop yards, whereas the model based on the predator:prey ratio was relatively conservative and yielded false-positive predictions in 11 of the 23 yards. Antecedents of these risk factors were explored and quantified by structural equation modeling. A simple path diagram was constructed that conceptualizes T. urticae invasion of hop cones as dependent on prior density of the pest on leaves in early spring and summer, which in turn influences the development of predatory arthropods that mediate late-season density of the pest. In summary, the biological insights and models developed here provide guidance to pest managers on the likelihood of visual cone damage from T. urticae that can inform late-season management based on both abundance of the pest and its important predators. This is critically important because a formal economic threshold for T. urticae on hop does not exist and current management efforts may be mistimed to influence the pest when crop damage is most probable. More broadly, this research suggests that current management practices that target T. urticae early in the season may in fact predispose yards to later outbreaks of the pest.

Identifying UK travellers at increased risk of developing pneumococcal infection: a novel algorithm.

BACKGROUND: In 2016, the travel subcommittee of the UK Joint Committee on Vaccination and Immunisation (JCVI) recommended that 13-valent PCV (PCV13) could be offered to travellers aged over 65 years, visiting countries without infant PCV immunization programmes. This study aimed to identify, collate and review the available evidence to identify specific countries where UK travellers might be at an increased risk of developing pneumococcal infection. The data were then used to develop an algorithm, which could be used to facilitate implementation of the JCVI recommendation. METHODS: We conducted a systematic search of the published data available for pneumococcal disease, PCV vaccine implementation, coverage data and programme duration by country. The primary data sources used were World Health Organization databases and the International Vaccine Access Centre Vaccine Information and Epidemiology Window-hub database. Based on the algorithm, the countries were classified into 'high overall risk', 'intermediate overall risk' and 'low overall risk' from an adult traveller perspective. This could determine whether PCV13 should be recommended for UK adult travellers. RESULTS: A data search for a total of 228 countries was performed, with risk scores calculated for 188 countries. Overall, 45 countries were classified as 'high overall risk', 86 countries as 'intermediate overall risk', 57 countries as 'low overall risk' and 40 countries as 'unknown'. CONCLUSION: To our knowledge this is the first attempt to categorize the risk to UK adult travellers of contracting pneumococcal infection in each country, globally. These findings could be used by national travel advisory bodies and providers of travel vaccines to identify travellers at increased risk of pneumococcal infection, who could be offered PCV immunization.

The impact of communicating personal mental ill-health risk: A randomized controlled non-inferiority trial.

AIM: Risk algorithms predicting personal mental ill-health will form an important component of digital and personalized preventive interventions, yet it is unknown whether informing people of personal risk may cause unintended harm. This trial evaluated the comparative effect of communicating personal mental ill-health risk profiles on psychological distress. METHODS: Australian participants using a mood-monitoring app were randomly allocated to receiving their current personal mental ill-health risk profile (n = 119), their achievable personal risk profile (n = 118) or to a control group (n = 118) in which no risk information was communicated, in a non-inferiority trial design. The primary outcome was psychological distress at four-weeks as assessed on the Kessler Psychological Distress Scale. RESULTS: There was high attrition in the trial with 64% of data missing at follow up. Per-protocol (completer) analysis found that the lower bounds of the confidence intervals of the estimated mean change of the current risk (m = 0.19, 95% CI: -2.59- 2.98) and achievable risk (m = -0.09, 95% CI: -2.84 to 2.66) groups were within the non-inferiority margin of the control group's mean at follow up. Supplementary intention-to-treat analysis using Multivariate Imputation by Chained Equations (MICE) found that 98/100 imputed datasets of the current risk profile group, and all imputed datasets of the achievable risk profile group showed non-inferiority to the control group. CONCLUSIONS: This study provides preliminary support that providing personal mental health risk profiles does not lead to unacceptable worsening of distress compared to no risk feedback, although this needs to be replicated in a fully powered RCT.

Consensus Opinion for The Management of Soft Tissue Filler Induced Vision Loss.

There are multiple treatment strategies proposed for the management of vision loss related to the injection of soft tissue fillers. Currently, there is no internationally accepted consensus on the immediate management of soft tissue filler induced vision loss (STFIVL). A recent systematic review of the literature concluded that there is not enough evidence to support retrobulbar hyaluronidase, and alternative treatments require exploration. The available literature demonstrates the inconsistent and unproven success of retrobulbar and peribulbar hyaluronidase in reversal of soft filler induced vision loss. Various therapeutics have been used to aid the reversal of vision loss but with mixed outcomes. The current evidence base does not support the use of retrobulbar and peribulbar hyaluronidase. The use of retrobulbar hyaluronidase for reversing soft tissue filler induced vision loss is controversial. Its efficacy remains unproven and there is mixed evidence within the literature. The current evidence suggests that there may be an increased risk of introducing severe adverse events associated with retrobulbar hyaluronidase and may even exacerbate the problem for those clinicians who are not ophthalmology trained. Therefore, we recommend two alternative treatment pathways for ophthalmology and non-ophthalmology trained practitioners. The suggested goal of this publication is to understand the pathophysiology of STFIVL, recognize signs and symptoms, and to propose algorithms to manage vision loss for both non-ophthalmology and ophthalmology trained clinicians. Clinicians must act swiftly and arrange immediate transfer to an emergency department or ophthalmology specialist setting to give the patient the best chance of vision restoration. The focus of any intervention for non-ophthalmology trained clinicians should be based around the immediate use of non-invasive techniques.

Non-lab and semi-lab algorithms for screening undiagnosed diabetes: A cross-sectional study.

BACKGROUND: The terrifying undiagnosed rate and high prevalence of diabetes have become a public emergency. A high efficiency and cost-effective early recognition method is urgently needed. We aimed to generate innovative, user-friendly nomograms that can be applied for diabetes screening in different ethnic groups in China using the non-lab or noninvasive semi-lab data. METHODS: This multicenter, multi-ethnic, population-based, cross-sectional study was conducted in eight sites in China by enrolling subjects aged 20-70. Sociodemographic and anthropometric characteristics were collected. Blood and urine samples were obtained 2 h following a standard 75 g glucose solution. In the final analysis, 10,794 participants were included and randomized into model development (n = 8096) and model validation (n = 2698) group with a ratio of 3:1. Nomograms were developed by the stepwise binary logistic regression. The nomograms were validated internally by a bootstrap sampling method in the model development set and externally in the model validation set. The area under the receiver operating characteristic curve (AUC) was used to assess the screening performance of the nomograms. Decision curve analysis was applied to calculate the net benefit of the screening model. RESULTS: The overall prevalence of undiagnosed diabetes was 9.8% (1059/10794) according to ADA criteria. The non-lab model revealed that gender, age, body mass index, waist circumference, hypertension, ethnicities, vegetable daily consumption and family history of diabetes were independent risk factors for diabetes. By adding 2 h post meal glycosuria qualitative to the non-lab model, the semi-lab model showed an improved Akaike information criterion (AIC: 4506 to 3580). The AUC of the semi-lab model was statistically larger than the non-lab model (0.868 vs 0.763, P < 0.001). The optimal cutoff probability in semi-lab and non-lab nomograms were 0.088 and 0.098, respectively. The sensitivity and specificity were 76.3% and 81.6%, respectively in semi-lab nomogram, and 72.1% and 67.3% in non-lab nomogram at the optimal cut off point. The decision curve analysis also revealed a bigger decrease of avoidable OGTT test (52 per 100 subjects) in the semi-lab model compared to the non-lab model (36 per 100 subjects) and the existed New Chinese Diabetes Risk Score (NCDRS, 35 per 100 subjects). CONCLUSION: The non-lab and semi-lab nomograms appear to be reliable tools for diabetes screening, especially in developing countries. However, the semi-lab model outperformed the non-lab model and NCDRS prediction systems and might be worth being adopted as decision support in diabetes screening in China.

Simple non-laboratory- and laboratory-based risk assessment algorithms and nomogram for detecting undiagnosed diabetes mellitus.

BACKGROUND: The aim of the present study was to develop a simple nomogram that can be used to predict the risk of diabetes mellitus (DM) in the asymptomatic non-diabetic subjects based on non-laboratory- and laboratory-based risk algorithms. METHODS: Anthropometric data, plasma fasting glucose, full lipid profile, exercise habits, and family history of DM were collected from Chinese non-diabetic subjects aged 18-70 years. Logistic regression analysis was performed on a random sample of 2518 subjects to construct non-laboratory- and laboratory-based risk assessment algorithms for detection of undiagnosed DM; both algorithms were validated on data of the remaining sample (n = 839). The Hosmer-Lemeshow test and area under the receiver operating characteristic (ROC) curve (AUC) were used to assess the calibration and discrimination of the DM risk algorithms. RESULTS: Of 3357 subjects recruited, 271 (8.1%) had undiagnosed DM defined by fasting glucose ≥7.0 mmol/L or 2-h post-load plasma glucose ≥11.1 mmol/L after an oral glucose tolerance test. The non-laboratory-based risk algorithm, with scores ranging from 0 to 33, included age, body mass index, family history of DM, regular exercise, and uncontrolled blood pressure; the laboratory-based risk algorithm, with scores ranging from 0 to 37, added triglyceride level to the risk factors. Both algorithms demonstrated acceptable calibration (Hosmer-Lemeshow test: P = 0.229 and P = 0.483) and discrimination (AUC 0.709 and 0.711) for detection of undiagnosed DM. CONCLUSION: A simple-to-use nomogram for detecting undiagnosed DM has been developed using validated non-laboratory-based and laboratory-based risk algorithms.

Estimating overweight risk in childhood from predictors during infancy.

OBJECTIVE: The aim of this study was to develop and validate a risk score algorithm for childhood overweight based on a prediction model in infants. METHODS: Analysis was conducted by using the UK Millennium Cohort Study. The cohort was divided randomly by using 80% of the sample for derivation of the risk algorithm and 20% of the sample for validation. Stepwise logistic regression determined a prediction model for childhood overweight at 3 years defined by the International Obesity Task Force criteria. Predictive metrics R(2), area under the receiver operating curve (AUROC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated. RESULTS: Seven predictors were found to be significantly associated with overweight at 3 years in a mutually adjusted predictor model: gender, birth weight, weight gain, maternal prepregnancy BMI, paternal BMI, maternal smoking in pregnancy, and breastfeeding status. Risk scores ranged from 0 to 59 corresponding to a predicted risk from 4.1% to 73.8%. The model revealed moderately good predictive ability in both the derivation cohort (R(2) = 0.92, AUROC = 0.721, sensitivity = 0.699, specificity = 0.679, PPV = 38%, NPV = 87%) and validation cohort (R(2) = 0.84, AUROC = 0.755, sensitivity = 0.769, specificity = 0.665, PPV = 37%, NPV = 89%). CONCLUSIONS: Using a prediction algorithm to identify at-risk infants could reduce levels of child overweight and obesity by enabling health professionals to target prevention more effectively. Further research needs to evaluate the clinical validity, feasibility, and acceptability of communicating this risk.