Category: Statistics

J Med Internet Res. 2022 Sep 28;24(9):e40547. doi: 10.2196/40547.

ABSTRACT

Scholars from the health and medical sciences have recently proposed the term social informatics (SI) as a new scientific subfield of health informatics (HI). However, SI is not a new academic concept; in fact, it has been continuously used in the social sciences and informatics since the 1970s. Although the dominant understanding of SI was established in the 1990s in the United States, a rich international perspective on SI has existed since the 1970s in other regions of the world. When that perspective is considered, the fields of understanding can be structured into 7 SI schools of thought. Against that conceptual background, this paper contributes to the discussion on the relationship between SI and HI, outlining possible perspectives of SI that are associated with health, medical, and clinical aspects. This paper argues against the multiplication and inconsistent appearance of the term SI when newly used in health and medical sciences. A more explicit name for the area that uses health and social data to advance individual and population health might be helpful to overcome this issue; giving an identity to this new field would help it to be understood more precisely and bring greater separation. This labeling could be fruitful for further segmentation of HI, which is rapidly expanding.

PMID:36169995 | DOI:10.2196/40547

JMIR Form Res. 2022 Sep 28;6(9):e36118. doi: 10.2196/36118.

ABSTRACT

BACKGROUND: There are a myriad of language cues that indicate depression in written texts, and natural language processing (NLP) researchers have proven the ability of machine learning and deep learning approaches to detect these cues. However, to date, these approaches bridging NLP and the domain of mental health for Bengali literature are not comprehensive. The Bengali-speaking population can express emotions in their native language in greater detail.

OBJECTIVE: Our goal is to detect the severity of depression using Bengali texts by generating a novel Bengali corpus of depressive posts. We collaborated with mental health experts to generate a clinically sound labeling scheme and an annotated corpus to train machine learning and deep learning models.

METHODS: We conducted a study using Bengali text-based data from blogs and open source platforms. We constructed a procedure for annotated corpus generation and extraction of textual information from Bengali literature for predictive analysis. We developed our own structured data set and designed a clinically sound labeling scheme with the help of mental health professionals, adhering to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) during the process. We used 5 machine learning models for detecting the severity of depression: kernel support vector machine (SVM), random forest, logistic regression K-nearest neighbor (KNN), and complement naive Bayes (NB). For the deep learning approach, we used long short-term memory (LSTM) units and gated recurrent units (GRUs) coupled with convolutional blocks or self-attention layers. Finally, we aimed for enhanced outcomes by using state-of-the-art pretrained language models.

RESULTS: The independent recurrent neural network (RNN) models yielded the highest accuracies and weighted F1 scores. GRUs, in particular, produced 81% accuracy. The hybrid architectures could not surpass the RNNs in terms of performance. Kernel SVM with term frequency-inverse document frequency (TF-IDF) embeddings generated 78% accuracy on test data. We used validation and training loss curves to observe and report the performance of our architectures. Overall, the number of available data remained the limitation of our experiment.

CONCLUSIONS: The findings from our experimental setup indicate that machine learning and deep learning models are fairly capable of assessing the severity of mental health issues from texts. For the future, we suggest more research endeavors to increase the volume of Bengali text data, in particular, so that modern architectures reach improved generalization capability.

PMID:36169989 | DOI:10.2196/36118

JAMA Psychiatry. 2022 Sep 28. doi: 10.1001/jamapsychiatry.2022.2983. Online ahead of print.

ABSTRACT

IMPORTANCE: Self-reported trauma exposure has consistently been found to be a risk factor for major depressive disorder (MDD), and several studies have reported interactions with genetic liability. To date, most studies have examined gene-environment interactions with trauma exposure using genome-wide variants (single-nucleotide variations [SNVs]) or polygenic scores, both typically capturing less than 3% of phenotypic risk variance.

OBJECTIVE: To reexamine genome-by-trauma interaction associations using genetic measures using all available genotyped data and thus, maximizing accounted variance.

DESIGN, SETTING, AND PARTICIPANTS: The UK Biobank study was conducted from April 2007 to May 1, 2016 (follow-up mental health questionnaire). The current study used available cross-sectional genomic and trauma exposure data from UK Biobank. Participants who completed the mental health questionnaire and had available genetic, trauma experience, depressive symptoms, and/or neuroticism information were included. Data were analyzed from April 1 to August 30, 2021.

EXPOSURES: Trauma and genome-by-trauma exposure interactions.

MAIN OUTCOMES AND MEASURES: Measures of self-reported depression, neuroticism, and trauma exposure with whole-genome SNV data are available from the UK Biobank study. Here, a mixed-model statistical approach using genetic, trauma exposure, and genome-by-trauma exposure interaction similarity matrices was used to explore sources of variation in depression and neuroticism.

RESULTS: Analyses were conducted on 148 129 participants (mean [SD] age, 56 [7] years) of which 76 995 were female (52.0%). The study approach estimated the heritability (SE) of MDD to be approximately 0.160 (0.016). Subtypes of self-reported trauma exposure (catastrophic, adult, childhood, and full trauma) accounted for a significant proportion of the variance of MDD, with heritability (SE) ranging from 0.056 (0.013) to 0.176 (0.025). The proportion of MDD risk variance accounted for by significant genome-by-trauma interaction revealed estimates (SD) ranging from 0.074 (0.006) to 0.201 (0.009). Results from sex-specific analyses found genome-by-trauma interaction variance estimates approximately 5-fold greater for MDD in male participants (0.441 [0.018]) than in female participants (0.086 [0.009]).

CONCLUSIONS AND RELEVANCE: This cross-sectional study used an approach combining all genome-wide SNV data when exploring genome-by-trauma interactions in individuals with MDD; findings suggest that such interactions were associated with depression manifestation. Genome-by-trauma interaction accounts for greater trait variance in male individuals, which points to potential differences in depression etiology between the sexes. The methodology used in this study can be extrapolated to other environmental factors to identify modifiable risk environments and at-risk groups to target with interventions.

PMID:36169986 | DOI:10.1001/jamapsychiatry.2022.2983

Melanoma Res. 2022 Sep 28. doi: 10.1097/CMR.0000000000000855. Online ahead of print.

ABSTRACT

This study was designed to screen for preliminary evidence of predictive markers of melanoma response to PD-1 blockade. We hypothesized that the following immune markers would be positive predictors of response: increased densities of CD103+CD8+ T cells or Th1 lineage T-bet+ T cells, high expression of CXCL9-11 and presence of tertiary lymphoid structures. Conversely, we hypothesized that the high expression of barrier molecules would be a negative predictor of response. Patients with advanced melanoma treated with pembrolizumab were identified, and clinical response as well as overall survival data were collected. Tumor samples were evaluated by multiplex immunofluorescence histology. All statistical analyses were performed in R Studio and Microsoft Excel using the Mann-Whitney U test, chi-square test, Spearman’s rank correlation and Kaplan-Meier survival curves. Sixty-five advanced melanoma patients were identified, of whom 46 met inclusion criteria and were included in this study. Increased densities (P = 0.04) and proportions (P = 0.02) of CD8+ T cells expressing CD103+ were associated with complete response (CR) to pembrolizumab. Improved survival was associated with increased proportions of CD8+ cells expressing CD103 (P = 0.0085) as well as decreased density of periplakin+ cells (P = 0.012) and periplakin+SOX10+ cells (P = 0.0012). The density and proportion of CD8+ T cells expressing CD103+ positively correlated with PD-L1 expression, though PD-L1 expression was not significantly correlated with outcomes. This screening study found that increased density and proportion of CD8+ T cells expressing CD103 and decreased density of periplakin were associated with positive outcomes in patients with melanoma metastases treated with pembrolizumab and may warrant further study.

PMID:36169985 | DOI:10.1097/CMR.0000000000000855

JAMA Psychiatry. 2022 Sep 28. doi: 10.1001/jamapsychiatry.2022.2990. Online ahead of print.

ABSTRACT

IMPORTANCE: Suicide is a leading cause of death in the United States, having increased more than 30% from 2000 to 2018. An inexpensive, safe, widely available treatment for preventing suicidal behavior could reverse this trend.

OBJECTIVE: To confirm a previous signal for decreased risk of suicide attempt following prescription fills for folic acid in a national pharmacoepidemiologic study of patients treated with folic acid.

DESIGN, SETTING, AND PARTICIPANTS: A within-person exposure-only cohort design was used to study the dynamic association between folic acid (vitamin B9) prescription fills over a 24-month period and suicide attempts and intentional self-harm. Data were collected from a pharmacoepidemiologic database of US medical claims (MarketScan) for patients with private health insurance who filled a folic acid prescription between 2012 and 2017. The same analysis was repeated with a control supplement (cyanocobalamin, vitamin B12). Data were analyzed from August 2021 to June 2022.

EXPOSURE: Folic acid prescription fills.

MAIN OUTCOME AND MEASURE: Suicide attempt or intentional self-harm resulting in an outpatient visit or inpatient admission as identified by codes from the International Statistical Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification.

RESULTS: Data on 866 586 patients were collected; 704 514 (81.30%) were female, and 90 296 (10.42%) were 60 years and older. Overall, there were 261 suicidal events during months covered by a folic acid prescription (5 521 597 person-months) for a rate of 4.73 per 100 000 person-months, compared with 895 suicidal events during months without folic acid (8 432 340) for a rate of 10.61 per 100 000 person-months. Adjusting for age and sex, diagnoses related to suicidal behavior, diagnoses related to folic acid deficiency, folate-reducing medications, history of folate-reducing medications, and history of suicidal events, the hazard ratio (HR) for folic acid for suicide events was 0.56 (95% CI, 0.48-0.65), with similar results for the modal dosage of 1 mg of folic acid per day (HR, 0.57; 95% CI, 0.48-0.69) and women of childbearing age (HR, 0.60; 95% CI, 0.50-0.73). A duration-response analysis (1-mg dosage) revealed a 5% decrease in suicidal events per month of additional treatment (HR, 0.95; 95% CI, 0.93-0.97). The same analysis for the negative control, cyanocobalamin, found no association with suicide attempt (HR, 1.01; 95% CI, 0.80-1.27).

CONCLUSIONS AND RELEVANCE: This large-scale pharmacoepidemiologic study of folic acid found a beneficial association in terms of lower rates of suicide attempts. The results warrant the conduct of a randomized clinical trial with suicidal ideation and behavior as outcomes of interest. If confirmed, folic acid may be a safe, inexpensive, and widely available treatment for suicidal ideation and behavior.

PMID:36169979 | DOI:10.1001/jamapsychiatry.2022.2990

Environ Health Perspect. 2022 Sep;130(9):97008. doi: 10.1289/EHP10889. Epub 2022 Sep 28.

ABSTRACT

BACKGROUND: Based on human and animal experimental studies, exposure to ambient carbon monoxide (CO) may be associated with cardiovascular disease outcomes, but epidemiological evidence of this link is limited. The number and distribution of ground-level regulatory agency monitors are insufficient to characterize fine-scale variations in CO concentrations.

OBJECTIVES: To develop a daily, high-resolution ambient CO exposure prediction model at the city scale.

METHODS: We developed a CO prediction model in Baltimore, Maryland, based on a spatiotemporal statistical algorithm with regulatory agency monitoring data and measurements from calibrated low-cost gas monitors. We also evaluated the contribution of three novel parameters to model performance: high-resolution meteorological data, satellite remote sensing data, and copollutant (${\mathrm{PM}}_{2.5}$, ${\mathrm{NO}}_2$, and ${\mathrm{NO}}_x$) concentrations.

RESULTS: The CO model had spatial cross-validation (CV) $R^2$ and root-mean-square error (RMSE) of $0.70\text{and}0.02\text{parts per million}$ (ppm), respectively; the model had temporal CV $R^2$ and RMSE of 0.61 and $0.04\text{ppm}$, respectively. The predictions revealed spatially resolved CO hot spots associated with population, traffic, and other nonroad emission sources (e.g., railroads and airport), as well as sharp concentration decreases within short distances from primary roads.

DISCUSSION: The three novel parameters did not substantially improve model performance, suggesting that, on its own, our spatiotemporal modeling framework based on geographic features was reliable and robust. As low-cost air monitors become increasingly available, this approach to CO concentration modeling can be generalized to resource-restricted environments to facilitate comprehensive epidemiological research. https://doi.org/10.1289/EHP10889.

PMID:36169978 | DOI:10.1289/EHP10889

J Magn Reson Imaging. 2022 Sep 28. doi: 10.1002/jmri.28435. Online ahead of print.

ABSTRACT

BACKGROUND: Metastatic lymph nodal number (LNN) is associated with the survival of nasopharyngeal carcinoma (NPC); however, counting multiple nodes is cumbersome.

PURPOSE: To explore LNN threshold and evaluate its use in risk stratification and induction chemotherapy (IC) indication.

STUDY TYPE: Retrospective.

POPULATION: A total of 792 radiotherapy-treated NPC patients (N classification: N0 182, N1 438, N2 113, N3 59; training group: 396, validation group: 396; receiving IC: 390).

FIELD STRENGTH/SEQUENCE: T1-, T2- and postcontrast T1-weighted fast spin echo MRI at 1.5 or 3.0 T.

ASSESSMENT: Nomogram with (model B) or without (model A) LNN was constructed to evaluate the 5-year overall (OS), distant metastasis-free (DMFS), and progression-free survival (PFS) for the group as a whole and N1 stage subgroup. High- and low-risk groups were divided (above vs below LNN- or model B-threshold); their response to IC was evaluated among advanced patients in stage III/IV.

STATISTICAL TESTS: Maximally selected rank, univariate and multivariable Cox analysis identified the optimal LNN threshold and other variables. Harrell’s concordance index (C-index) and 2-fold cross-validation evaluated discriminative ability of models. Matched-pair analysis compared survival outcomes of adding IC or not. A P value < 0.05 was considered statistically significant.

RESULTS: Median follow-up duration was 62.1 months. LNN ≥ 4 was independently associated with decreased 5-year DMFS, OS, and PFS in entire patients or N1 subgroup. Compared to model A, model B (adding LNN, LNN ≥ 4 vs <4) presented superior C-indexes in the training (0.755 vs 0.727) and validation groups (0.676 vs 0.642) for discriminating DMFS. High-risk patients benefited from IC with improved post-IC response and OS, but low-risk patients did not (P = 0.785 and 0.690, respectively).

CONCLUSIONS: LNN ≥ 4 is an independent risk stratification factor of worse survival in entire or N1 staging NPC patients. LNN ≥ 4 or the associated nomogram has potential to identify high-risk patients requiring IC.

EVIDENCE LEVEL: 4 TECHNICAL EFFICACY: 4.

PMID:36169976 | DOI:10.1002/jmri.28435

JAMA Netw Open. 2022 Sep 1;5(9):e2233649. doi: 10.1001/jamanetworkopen.2022.33649.

ABSTRACT

IMPORTANCE: Value in health care is quality per unit cost (V = Q/C), and an emergency department-based intensive care unit (ED-ICU) model has been associated with improved quality. To assess the value of this care delivery model, it is essential to determine the incremental direct cost of care.

OBJECTIVE: To determine the association of an ED-ICU with inflation-adjusted change in mean direct cost of care, net revenue, and direct margin per ED patient encounter.

DESIGN, SETTING, AND PARTICIPANTS: This retrospective economic analysis evaluated the cost of care delivery to patients in the ED before and after deployment of the Joyce and Don Massey Family Foundation Emergency Critical Care Center, an ED-ICU, on February 16, 2015, at a large academic medical center in the US with approximately 75 000 adult ED visits per year. The pre-ED-ICU cohort was defined as all documented ED visits by patients 18 years or older with a complete financial record from September 8, 2012, through June 30, 2014 (660 days); the post-ED-ICU cohort, all visits from July 1, 2015, through April 21, 2017 (660 days). Fiscal year 2015 was excluded from analysis to phase in the new care model. Statistical analysis was performed March 1 through December 30, 2021.

EXPOSURES: Implementation of an ED-ICU.

MAIN OUTCOMES AND MEASURES: Inflation-adjusted direct cost of care, net revenue, and direct margin per patient encounter in the ED.

RESULTS: A total of 234 884 ED visits during the study period were analyzed, with 115 052 patients (54.7% women) in the pre-ED-ICU cohort and 119 832 patients (54.5% women) in the post-ED-ICU cohort. The post-ED-ICU cohort was older (mean [SD] age, 49.1 [19.9] vs 47.8 [19.6] years; P < .001), required more intensive respiratory support (2.2% vs 1.1%; P < .001) and more vasopressor use (0.5% vs 0.2%; P < .001), and had a higher overall case mix index (mean [SD], 1.7 [2.0] vs 1.5 [1.7]; P < .001). Implementation of the ED-ICU was associated with similar inflation-adjusted total direct cost per ED encounter (pre-ED-ICU, mean [SD], $4875 [$15 175]; post-ED-ICU, $4877 [$17 400]; P = .98). Inflation-adjusted net revenue per encounter increased by 7.0% (95% CI, 3.4%-10.6%; P < .001), and inflation-adjusted direct margin per encounter increased by 46.6% (95% CI, 32.1%-61.2%; P < .001).

CONCLUSIONS AND RELEVANCE: Implementation of an ED-ICU was associated with no significant change in inflation-adjusted total direct cost per ED encounter. Holding delivery costs constant while improving quality demonstrates improved value via the ED-ICU model of care.

PMID:36169958 | DOI:10.1001/jamanetworkopen.2022.33649

JAMA Netw Open. 2022 Sep 1;5(9):e2233667. doi: 10.1001/jamanetworkopen.2022.33667.

ABSTRACT

IMPORTANCE: Whether interprofessional collaboration is effective and safe in decreasing hospital length of stay remains controversial.

OBJECTIVE: To evaluate the outcomes and safety associated with an electronic interprofessional-led discharge planning tool vs standard discharge planning to safely reduce length of stay among medical inpatients with multimorbidity.

DESIGN, SETTING, AND PARTICIPANTS: This multicenter prospective nonrandomized controlled trial used interrupted time series analysis to examine medical acute hospitalizations at 82 hospitals in Switzerland. It was conducted from February 2017 through January 2019. Data analysis was conducted from March 2021 to July 2022.

INTERVENTION: After a 12-month preintervention phase (February 2017 through January 2018), an electronic interprofessional-led discharge planning tool was implemented in February 2018 in 7 intervention hospitals in addition to standard discharge planning.

MAIN OUTCOMES AND MEASURES: Mixed-effects segmented regression analyses were used to compare monthly changes in trends of length of stay, hospital readmission, in-hospital mortality, and facility discharge after the implementation of the tool with changes in trends among control hospitals.

RESULTS: There were 54 695 hospitalizations at intervention hospitals, with 27 219 in the preintervention period (median [IQR] age, 72 [59-82] years; 14 400 [52.9%] men) and 27 476 in the intervention phase (median [IQR] age, 72 [59-82] years; 14 448 [52.6%] men) and 438 791 at control hospitals, with 216 261 in the preintervention period (median [IQR] age, 74 [60-83] years; 109 770 [50.8%] men) and 222 530 in the intervention phase (median [IQR] age, 74 [60-83] years; 113 053 [50.8%] men). The mean (SD) length of stay in the preintervention phase was 7.6 (7.1) days for intervention hospitals and 7.5 (7.4) days for control hospitals. During the preintervention phase, population-averaged length of stay decreased by -0.344 hr/mo (95% CI, -0.599 to -0.090 hr/mo) in control hospitals; however, no change in trend was observed among intervention hospitals (-0.034 hr/mo; 95% CI, -0.646 to 0.714 hr/mo; difference in slopes, P = .09). Over the intervention phase (February 2018 through January 2019), length of stay remained unchanged in control hospitals (slope, -0.011 hr/mo; 95% CI, -0.281 to 0.260 hr/mo; change in slope, P = .03), but decreased steadily among intervention hospitals by -0.879 hr/mo (95% CI, -1.607 to -0.150 hr/mo; change in slope, P = .04, difference in slopes, P = .03). Safety analyses showed no change in trends of hospital readmission, in-hospital mortality, or facility discharge over the whole study time.

CONCLUSIONS AND RELEVANCE: In this nonrandomized controlled trial, the implementation of an electronic interprofessional-led discharge planning tool was associated with a decline in length of stay without an increase in hospital readmission, in-hospital mortality, or facility discharge.

TRIAL REGISTRATION: isrctn.org Identifier: ISRCTN83274049.

PMID:36169957 | DOI:10.1001/jamanetworkopen.2022.33667

Eur Thyroid J. 2022 Sep 1:ETJ-22-0142. doi: 10.1530/ETJ-22-0142. Online ahead of print.

ABSTRACT

OBJECTIVE: Thyroid disease in women of reproductive age is mainly of autoimmune origin, and thyroid peroxidase antibodies (TPO-Ab) as well as thyroglobulin antibodies (Tg-Ab) are key markers. Adding to this, much focus in pregnancy is on euthyroid women who are thyroid antibody positive. Evidence to substantiate the cut-offs for definition of thyroid autoantibody positivity in early pregnant women is warranted.

METHODS: Stored serum samples from 14,030 Danish pregnant women were used for measurement of TPO-Ab, Tg-Ab, TSH, and free thyroxine (ADVIA Centaur XPT, Siemens Healthineers, Erlangen, Germany). Among all women, a reference cohort of 10,905 individuals was identified for establishment of antibody cut-offs. Percentile cut-offs for TPO-Ab and Tg-Ab determined using Regression on Order Statistics (the reference cohort). The established cut-offs were then applied (the full cohort), and frequencies of early pregnancy as well as later diagnosis of hypothyroidism were evaluated.

RESULTS: The highest established cut-offs (95th, 97.5th, and 99th percentile) were 59, 68, and 81 U/mL for TPO-Ab, and 33, 41, and 52 U/mL for Tg-Ab. When the cut-offs were applied in the full cohort, 11.0%, 10.2%, and 9.7% were TPO-Ab positive, whereas 13.3%, 12.3%, and 11.2% were Tg-Ab positive. Antibody positive women (TPO-Ab and/or Tg-Ab) had higher median TSH and were more like to have hypothyroidism in early pregnancy and to be diagnosed with hypothyroidism during follow-up.

CONCLUSIONS: This large study established and evaluated pregnancy-specific cut-offs for TPO-Ab and Tg-Ab. The findings are important regarding classification of exposure in pregnancy and assessment of thyroid autoimmunity per se.

PMID:36169923 | DOI:10.1530/ETJ-22-0142