Effects of a comprehensive brain computed tomography deep learning model on radiologist detection accuracy.

OBJECTIVES: Non-contrast computed tomography of the brain (NCCTB) is commonly used to detect intracranial pathology but is subject to interpretation errors. Machine learning can augment clinical decision-making and improve NCCTB scan interpretation. This retrospective detection accuracy study assessed the performance of radiologists assisted by a deep learning model and compared the standalone performance of the model with that of unassisted radiologists. METHODS: A deep learning model was trained on 212,484 NCCTB scans drawn from a private radiology group in Australia. Scans from inpatient, outpatient, and emergency settings were included. Scan inclusion criteria were age ≥ 18 years and series slice thickness ≤ 1.5 mm. Thirty-two radiologists reviewed 2848 scans with and without the assistance of the deep learning system and rated their confidence in the presence of each finding using a 7-point scale. Differences in AUC and Matthews correlation coefficient (MCC) were calculated using a ground-truth gold standard. RESULTS: The model demonstrated an average area under the receiver operating characteristic curve (AUC) of 0.93 across 144 NCCTB findings and significantly improved radiologist interpretation performance. Assisted and unassisted radiologists demonstrated an average AUC of 0.79 and 0.73 across 22 grouped parent findings and 0.72 and 0.68 across 189 child findings, respectively. When assisted by the model, radiologist AUC was significantly improved for 91 findings (158 findings were non-inferior), and reading time was significantly reduced. CONCLUSIONS: The assistance of a comprehensive deep learning model significantly improved radiologist detection accuracy across a wide range of clinical findings and demonstrated the potential to improve NCCTB interpretation. CLINICAL RELEVANCE STATEMENT: This study evaluated a comprehensive CT brain deep learning model, which performed strongly, improved the performance of radiologists, and reduced interpretation time. The model may reduce errors, improve efficiency, facilitate triage, and better enable the delivery of timely patient care. KEY POINTS: • This study demonstrated that the use of a comprehensive deep learning system assisted radiologists in the detection of a wide range of abnormalities on non-contrast brain computed tomography scans. • The deep learning model demonstrated an average area under the receiver operating characteristic curve of 0.93 across 144 findings and significantly improved radiologist interpretation performance. • The assistance of the comprehensive deep learning model significantly reduced the time required for radiologists to interpret computed tomography scans of the brain.

Multimorbidity, mortality, and HbA1c in type 2 diabetes: A cohort study with UK and Taiwanese cohorts.

BACKGROUND: There is emerging interest in multimorbidity in type 2 diabetes (T2D), which can be either concordant (T2D related) or discordant (unrelated), as a way of understanding the burden of disease in T2D. Current diabetes guidelines acknowledge the complex nature of multimorbidity, the management of which should be based on the patient's individual clinical needs and comorbidities. However, although associations between multimorbidity, glycated haemoglobin (HbA1c), and mortality in people with T2D have been studied to some extent, significant gaps remain, particularly regarding different patterns of multimorbidity, including concordant and discordant conditions. This study explores associations between multimorbidity (total condition counts/concordant/discordant/different combinations of conditions), baseline HbA1c, and all-cause mortality in T2D. METHODS AND FINDINGS: We studied two longitudinal cohorts of people with T2D using the UK Biobank (n = 20,569) and the Taiwan National Diabetes Care Management Program (NDCMP) (n = 59,657). The number of conditions in addition to T2D was used to quantify total multimorbidity, concordant, and discordant counts, and the effects of different combinations of conditions were also studied. Outcomes of interest were baseline HbA1c and all-cause mortality. For the UK Biobank and Taiwan NDCMP, mean (SD) ages were 60.2 (6.8) years and 60.8 (11.3) years; 7,579 (36.8%) and 31,339 (52.5%) were female; body mass index (BMI) medians (IQR) were 30.8 (27.7, 34.8) kg/m2 and 25.6 (23.5, 28.7) kg/m2; and 2,197 (10.8%) and 9,423 (15.8) were current smokers, respectively. Increasing total and discordant multimorbidity counts were associated with lower HbA1c and increased mortality in both datasets. In Taiwan NDCMP, for those with four or more additional conditions compared with T2D only, the mean difference (95% CI) in HbA1c was -0.82% (-0.88, -0.76) p < 0.001. In UK Biobank, hazard ratios (HRs) (95% CI) for all-cause mortality in people with T2D and one, two, three, and four or more additional conditions compared with those without comorbidity were 1.20 (0.91-1.56) p < 0.001, 1.75 (1.35-2.27) p < 0.001, 2.17 (1.67-2.81) p < 0.001, and 3.14 (2.43-4.03) p < 0.001, respectively. Both concordant/discordant conditions were significantly associated with mortality; however, HRs were largest for concordant conditions. Those with four or more concordant conditions had >5 times the mortality (5.83 [4.28-7.93] p <0.001). HRs for NDCMP were similar to those from UK Biobank for all multimorbidity counts. For those with two conditions in addition to T2D, cardiovascular diseases featured in 18 of the top 20 combinations most highly associated with mortality in UK Biobank and 12 of the top combinations in the Taiwan NDCMP. In UK Biobank, a combination of coronary heart disease and heart failure in addition to T2D had the largest effect size on mortality, with a HR (95% CI) of 4.37 (3.59-5.32) p < 0.001, whereas in the Taiwan NDCMP, a combination of painful conditions and alcohol problems had the largest effect size on mortality, with an HR (95% CI) of 4.02 (3.08-5.23) p < 0.001. One limitation to note is that we were unable to model for changes in multimorbidity during our study period. CONCLUSIONS: Multimorbidity patterns associated with the highest mortality differed between UK Biobank (a population predominantly comprising people of European descent) and the Taiwan NDCMP, a predominantly ethnic Chinese population. Future research should explore the mechanisms underpinning the observed relationship between increasing multimorbidity count and reduced HbA1c alongside increased mortality in people with T2D and further examine the implications of different patterns of multimorbidity across different ethnic groups. Better understanding of these issues, especially effects of condition type, will enable more effective personalisation of care.

Global Prevalence of Colorectal Neoplasia: A Systematic Review and Meta-Analysis.

BACKGROUND & AIMS: Most colorectal cancers (CRC) arise from colorectal adenomas, yet there is not enough information on global prevalence to inform health care policy. We examined the prevalence of any type of adenomas, advanced adenomas (AADs), and CRC according to age, sex, ethnicity, geographic regions, and anatomic location (proximal vs distal). METHODS: MEDLINE and Embase were searched from their inception through May 1, 2018, to identify population-based, observational studies that reported the prevalence of colorectal neoplasia. Studies on participants 15 years or older, with a sample size of 500 persons or more, were included. Metaprop (College Station, TX) was used to model within-study variability by binomial distribution and Freeman-Tukey Double Arcsine Transformation to stabilize the variances. The prevalence figures were presented by proportions and their 95% CIs using random-effects models. RESULTS: Our meta-analysis included 70 studies involving 637,414 individuals. The overall prevalence rates of adenoma (23.9%; 95% CI, 22.2%-25.8%), AAD (4.6%; 95% CI, 3.8%-5.5%), and CRC (0.4%, 95% CI, 0.3%-0.5%) were calculated. Subgroup analysis indicated that prevalence values (adenomas, AADs, and CRCs) were higher among men (29.7%, 6.5%, and 0.8%, respectively) than women (19.3%, 3.8% and 0.4%, respectively), among older adults (25.9%, 5.2%, and 0.6%, respectively) than younger adults (14.6%, 1.6%, and 0.1%, respectively), among Caucasians (23.7%, 6.6%, and 0.5%, respectively) than other ethnicities, in European countries (25.9%, 8.4%, and 0.8%, respectively) than other countries, and among patients with proximal (25.9%, 5.3%, and 0.1%, respectively) vs distal neoplasia. CONCLUSIONS: In a systematic review and meta-analysis, we found a high prevalence of colorectal neoplasia among some populations. This indicates a need to expand CRC screening programs for these groups. The pooled prevalence estimates can be used as quality indicators for established CRC screening programs.

Associations between multimorbidity and glycaemia (HbA1c) in people with type 2 diabetes: cross-sectional study in Australian general practice.

OBJECTIVES: To explore the prevalence of multimorbidity as well as individual and combinations of long-term conditions (LTCs) in people with type 2 diabetes (T2D) attending Australian general practice, using electronic health record (EHR) data. We also examine the association between multimorbidity condition count (total/concordant(T2D related)/discordant(unrelated)) and glycaemia (glycated haemoglobin, HbA1c). DESIGN: Cross-sectional study. SETTING: Australian general practice. PARTICIPANTS: 69 718 people with T2D with a general practice encounter between 2013 and 2015 captured in the MedicineInsight database (EHR Data from 557 general practices and >3.8 million Australian patients). PRIMARY AND SECONDARY OUTCOME MEASURES: Prevalence of multimorbidity, individual and combinations of LTCs. Multivariable linear regression models used to examine associations between multimorbidity counts and HbA1c (%). RESULTS: Mean (SD) age 66.42 (12.70) years, 46.1% female and mean (SD) HbA1c 7.1 (1.4)%. More than 90% of participants with T2D were living with multimorbidity. Discordant conditions were more prevalent (83.4%) than concordant conditions (69.9 %). The three most prevalent discordant conditions were: painful conditions (55.4%), dyspepsia (31.6%) and depression (22.8%). The three most prevalent concordant conditions were hypertension (61.4%), coronary heart disease (17.1%) and chronic kidney disease (8.5%). The three most common combinations of conditions were: painful conditions and hypertension (38.8%), painful conditions and dyspepsia (23.1%) and hypertension and dyspepsia (22.7%). We found no associations between any multimorbidity counts (total, concordant and discordant) or combinations and HbA1c. CONCLUSIONS: Multimorbidity was common in our cohort of people with T2D attending Australian general practice, but was not associated with glycaemia. Although we did not explore mortality in this study, our results suggest that the increased mortality in those with multimorbidity and T2D observed in other studies may not be linked to glycaemia. Interestingly, discordant conditions were more prevalent than concordant conditions with painful conditions being the second most common comorbidity. Better understanding of the implications of different patterns of multimorbidity in people with T2D will allow more effective tailored care.

Multimorbidity, glycaemic variability and time in target range in people with type 2 diabetes: A baseline analysis of the GP-OSMOTIC trial.

AIMS: To explore associations between multimorbidity condition counts (total; concordant (diabetes-related); discordant (unrelated to diabetes)) and glycaemia (HbA1c; glycaemic variability (GV); time in range (TIR)) using data from a randomised controlled trial examining effectiveness of continuous glucose monitoring (CGM) in people with type 2 diabetes (T2D). METHODS: Cross-sectional study: 279 people with T2D using baseline data from the General Practice Optimising Structured MOnitoring To Improve Clinical outcomes (GP-OSMOTIC) trial from 25 general practices in Australia. Number of long-term conditions (LTCs) in addition to T2D used to quantify total/concordant/discordant multimorbidity counts. GV (measured by coefficient of variation (CV)) and TIR derived from CGM data. Multivariable linear regression models used to examine associations between multimorbidity counts, HbA1c (%), GV and TIR. RESULTS: Mean (SD) age of participants 60.4 (9.9) years; 40.9% female. Multimorbidity was present in 89.2% of participants. Most prevalent comorbid LTCs: hypertension (57.4%), painful conditions (29.8%), coronary heart disease (22.6%) and depression (19.0%). No evidence of associations between multimorbidity counts, HbA1c, GV and TIR. CONCLUSIONS: While multimorbidity was common in this T2D cohort, it was not associated with HbA1c, CV or TIR. Future studies should explore factors other than glycaemia that contribute to the increased mortality observed in those with multimorbidity and T2D.

Obesity paradox in associations between body mass index and diabetes-related hospitalization and mortality in patients with type 2 diabetes: Retrospective cohort studies.

AIM: Diabetes is a primary cause of hospitalization and in-hospital mortality. However, studies exploring the relationships between body mass index (BMI) and hospitalization-related and mortality-related outcomes in patients with type 2 diabetes are lacking. METHODS: Our data were obtained from two independent retrospective cohort studies, namely, the Taiwan Diabetes Study (Taiwan DS), providing hospitalization outcome measures, and the Taichung Diabetes Study (Taichung DS) that can be linked with the National Death Registry dataset. BMI and hospitalization, in-hospital mortality, and all-cause and cause-specific death events were analyzed by Cox proportional hazard regression model. RESULTS: A total of 3,541, 38,779, and 10,399 patients died during hospitalization, hospitalized for all-cause and diabetes-related events, respectively, in the Taiwan DS cohort. Moreover, 685 deaths were identified in the Taichung DS cohort. Compared with patients with increasing-but-acceptable-risk obesity, multivariable-adjusted hazard ratios (HRs) of in-hospital mortality, all-cause hospitalization, hospitalization due to diabetes, hypoglycaemia, and renal failure for patients who were underweight were 2.09 (95% confidence interval 1.73, 2.51), 1.39 (1.28, 1.50), 1.69 (1.49, 1.90), 1.87 (1.34, 2.61), and 1.55 (1.26, 1.91). Adjusted HRs of all-cause mortality and non-expanded CVD-related mortality in patients with underweight were 2.02 (1.28, 3.21), and 2.27 (1.28, 4.03). CONCLUSIONS: The BMI associated with the best survival and less hospitalization was the higher-high-risk obesity (≥ 27.5 kg/m2) category. We observed obesity paradox for mortality outcomes, which should be addressed by further research, particularly on whether randomized controlled trials of adopting a healthy lifestyle for patients with obesity can improve type 2 diabetes patients' survival.

Associations between multimorbidity, all-cause mortality and glycaemia in people with type 2 diabetes: A systematic review.

INTRODUCTION: Type 2 diabetes (T2D) is a major health priority worldwide and the majority of people with diabetes live with multimorbidity (MM) (the co-occurrence of ≥2 chronic conditions). The aim of this systematic review was to explore the association between MM and all-cause mortality and glycaemic outcomes in people with T2D. METHODS: The search strategy centred on: T2D, MM, comorbidity, mortality and glycaemia. Databases searched: MEDLINE, EMBASE, CINAHL Complete, The Cochrane Library, and SCOPUS. Restrictions included: English language, quantitative empirical studies. Two reviewers independently carried out: abstract and full text screening, data extraction, and quality appraisal. Disagreements adjudicated by a third reviewer. RESULTS: Of the 4882 papers identified; 41 met inclusion criteria. The outcome was all-cause mortality in 16 studies, glycaemia in 24 studies and both outcomes in one study. There were 28 longitudinal cohort studies and 13 cross-sectional studies, with the number of participants ranging from 96-892,223. Included studies were conducted in high or upper-middle-income countries. Fifteen of 17 studies showed a statistically significant association between increasing MM and higher mortality. Ten of 14 studies showed no significant associations between MM and HbA1c. Four of 14 studies found higher levels of MM associated with higher HbA1c. Increasing MM was significantly associated with hypoglycaemia in 9/10 studies. There was no significant association between MM and fasting glucose (one study). No studies explored effects on glycaemic variability. CONCLUSIONS: This review demonstrates that MM in T2D is associated with higher mortality and hypoglycaemia, whilst evidence regarding the association with other measures of glycaemic control is mixed. The current single disease focused approach to management of T2D seems inappropriate. Our findings highlight the need for clinical guidelines to support a holistic approach to the complex care needs of those with T2D and MM, accounting for the various conditions that people with T2D may be living with. SYSTEMATIC REVIEW REGISTRATION: International Prospective Register of Systematic Reviews CRD42017079500.

Impact of multimorbidity count on all-cause mortality and glycaemic outcomes in people with type 2 diabetes: a systematic review protocol.

INTRODUCTION: Type 2 diabetes (T2D) is a leading health priority worldwide. Multimorbidity (MM) is a term describing the co-occurrence of two or more chronic diseases or conditions. The majority of people living with T2D have MM. The relationship between MM and mortality and glycaemia in people with T2D is not clear. METHODS AND ANALYSIS: Medline, Embase, Cumulative Index of Nursing and Allied Health Complete, The Cochrane Library, and SCOPUS will be searched with a prespecified search strategy. The searches will be limited to quantitative empirical studies in English with no restriction on publication date. One reviewer will perform title screening and two review authors will independently screen the abstract and full texts using Covidence software, with disagreements adjudicated by a third reviewer. Data will be extracted using a using a Population, Exposure, Comparator and Outcomes framework. Two reviewers will independently extract data and undertake the risk of bias (quality) assessment. Disagreements will be resolved by consensus. A narrative synthesis of the results will be conducted and meta-analysis considered if appropriate. Quality appraisal will be undertaken using the Newcastle-Ottawa quality assessment scale and the quality of the cumulative evidence of the included studies will be assessed using the Grading of Recommendations, Assessment, Development and Evaluation approach. This protocol was prepared in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols guidelines to ensure the quality of our review. ETHICS AND DISSEMINATION: This review will synthesise the existing evidence about the impact of MM on mortality and glycaemic outcomes in people living with T2D and increase our understanding of this subject and will inform future practice and policy. Findings will be disseminated via conference presentations, social media and peer-reviewed publication. PROSPERO REGISTRATION NUMBER: CRD42017079500.