Predicting hypertension control using machine learning.

Hypertension is a widely prevalent disease and uncontrolled hypertension predisposes affected individuals to severe adverse effects. Though the importance of controlling hypertension is clear, the multitude of therapeutic regimens and patient factors that affect the success of blood pressure control makes it difficult to predict the likelihood to predict whether a patient's blood pressure will be controlled. This project endeavors to investigate whether machine learning can accurately predict the control of a patient's hypertension within 12 months of a clinical encounter. To build the machine learning model, a retrospective review of the electronic medical records of 350,008 patients 18 years of age and older between January 1, 2015 and June 1, 2022 was performed to form model training and testing cohorts. The data included in the model included medication combinations, patient laboratory values, vital sign measurements, comorbidities, healthcare encounters, and demographic information. The mean age of the patient population was 65.6 years with 161,283 (46.1%) men and 275,001 (78.6%) white. A sliding time window of data was used to both prohibit data leakage from training sets to test sets and to maximize model performance. This sliding window resulted in using the study data to create 287 predictive models each using 2 years of training data and one week of testing data for a total study duration of five and a half years. Model performance was combined across all models. The primary outcome, prediction of blood pressure control within 12 months demonstrated an area under the curve of 0.76 (95% confidence interval; 0.75-0.76), sensitivity of 61.52% (61.0-62.03%), specificity of 75.69% (75.25-76.13%), positive predictive value of 67.75% (67.51-67.99%), and negative predictive value of 70.49% (70.32-70.66%). An AUC of 0.756 is considered to be moderately good for machine learning models. While the accuracy of this model is promising, it is impossible to state with certainty the clinical relevancy of any clinical support ML model without deploying it in a clinical setting and studying its impact on health outcomes. By also incorporating uncertainty analysis for every prediction, the authors believe that this approach offers the best-known solution to predicting hypertension control and that machine learning may be able to improve the accuracy of hypertension control predictions using patient information already available in the electronic health record. This method can serve as a foundation with further research to strengthen the model accuracy and to help determine clinical relevance.

Identification of acute aortic syndromes based on cross-sectional variability of Hounsfield units.

BACKGROUND: A characteristic feature of communicating aortic dissections (CD) is the dissection flap between the true and false lumen. However, in intramural hematomas (IMH) a flap is not visible. We aimed to determine if cross-sectional HU variability allow reliable identification of aortic dissections including IMH. METHODS: We included 362 patients presenting with acute chest pain (CP) or respiratory distress (RD) and underwent contrast-enhanced CTA with or without ECG-gating. In the derivation group we included 72 CP patients with and 74 without AAS. In the validation group we included 108 CP or RD patients with and 108 without AAS. The adventitial border of the aorta was visually identified and measurements were performed at 6 locations along the ascending and descending aorta. At each cross-section 5 circular ROI measurements of HU were made and the maximum HU difference calculated. RESULTS: In the derivation and validation group the maximum difference in HUs at any one location was significantly higher for AAS subjects than controls (validation group: median = 128.5 vs. 34.0, p-value Wilcoxon two-sample test <0.001). In the validation group, the estimated AUC was 0.939 with 95% CIs of [0.906, 0.972], indicating that the maximum difference in HUs is a strong predictor of AAS (p < 0.001). CONCLUSION: Our data provide evidence that cross-sectional variability of Hounsfield Unit reliably identifies aortic dissection including IMH in dedicated ECG-gated aorta scans but also non-gated chest CTs with limited aortic contrast enhancement. These results suggest that this approach could be feasible for an automated algorithm for identification of AAS.

All things considered, my risk for diabetes is medium: A risk personalization process of familial risk for type 2 diabetes.

BACKGROUND: A positive family history of type 2 diabetes (T2D) has been associated with risk awareness and risk-reducing behaviours among the unaffected relatives. Yet, little is known about how people with a positive family history for diabetes develop and manage their personal sense of risk. OBJECTIVE: To characterize two key concepts, salience and vulnerability, within the familial risk perception (FRP) model among unaffected individuals, at increased familial risk for T2D. DESIGN: We conducted a mixed method study. Descriptions of salience and vulnerability were collected through semi-structured interviews. Participant's perception of self-reported risk factors (family history, age, race/ethnicity, medical history, weight and exercise) was measured using the Perceived Risk Factors for T2D Tool and was compared to a clinical evaluation of the same risk factors. RESULTS: We identified two components of salience: (a) concern for developing T2D and (b) risk awareness triggers, and two features of vulnerability: (a) statement of risk and (b) risk assessment devices. Although few participants (26%) were concordant between their perceived and clinical overall T2D risk, concordance for individual risk factors was higher, ranging from 42% (medical history) to 90% (family history). DISCUSSION AND CONCLUSION: Both familial and non-familial events lead people to contemplate their T2D risk, even among people who have a positive family history. Participants often downplayed their overall risk and underestimated their overall risk compared to a clinical risk assessment of the same self-reported risk factors. Clinicians could leverage key components of the FRP process as way to engage patients in risk reduction strategies earlier.

Cleveland Clinic's Center for personalized healthcare: setting the stage for value-based care.

Cleveland Clinic (OH, USA) launched the Center for Personalized Healthcare in 2011 to establish an evidence-based system for individualizing care by incorporating unique patient characteristics, including but not limited to genetic and family health history information, into the standard medical decision-making process. Using MyFamily, a web-based tool integrated into our electronic health record, a patient's family health history is used as a surrogate for genetic, environmental and behavioral risks to identify those with an elevated probability of developing disease. Complementing MyFamily, the Personalized Medication Program was created for the purpose of identifying gene-drug pairs for integration into clinical practice and developing the implementation tools needed to incorporate pharmacogenomics into the clinical workflow. We have successfully implemented the gene-drug pairs HLA-B*57:01-abacavir and TPMT-thiopurines into patient care. Our efforts to establish personalized medical care at Cleveland Clinic may serve as a model for large-scale integration of personalized healthcare.