Detecting diabetic retinopathy through machine learning on electronic health record data from an urban, safety net healthcare system.

OBJECTIVE: Clinical guidelines recommend annual eye examinations to detect diabetic retinopathy (DR) in patients with diabetes. However, timely DR detection remains a problem in medically underserved and under-resourced settings in the United States. Machine learning that identifies patients with latent/undiagnosed DR could help to address this problem. MATERIALS AND METHODS: Using electronic health record data from 40 631 unique diabetic patients seen at Los Angeles County Department of Health Services healthcare facilities between January 1, 2015 and December 31, 2017, we compared ten machine learning environments, including five classifier models, for assessing the presence or absence of DR. We also used data from a distinct set of 9300 diabetic patients seen between January 1, 2018 and December 31, 2018 as an external validation set. RESULTS: Following feature subset selection, the classifier with the best AUC on the external validation set was a deep neural network using majority class undersampling, with an AUC of 0.8, the sensitivity of 72.17%, and specificity of 74.2%. DISCUSSION: A deep neural network produced the best AUCs and sensitivity results on the test set and external validation set. Models are intended to be used to screen guideline noncompliant diabetic patients in an urban safety-net setting. CONCLUSION: Machine learning on diabetic patients' routinely collected clinical data could help clinicians in safety-net settings to identify and target unscreened diabetic patients who potentially have undiagnosed DR.

Schizophrenia hospitalization in the US 2005-2014: Examination of trends in demographics, length of stay, and cost.

ABSTRACT: Primarily we aimed to examine the crude and standardized schizophrenia hospitalization trend from 2005 to 2014. We hypothesized that there will be a statistically significant linear trend in hospitalization rates for schizophrenia from 2005 to 2014. Secondarily we also examined trends in hospitalization by race/ethnicity, age, gender, as well as trends in hospitalization Length of Stay (LOS) and inflation adjusted cost.In this observational study, we used Nationwide Inpatient Sample data and International Classification of Diseases, Eleventh Revisions codes for Schizophrenia, which revealed 6,122,284 cases for this study. Outcomes included crude and standardized hospitalization rates, race/ethnicity, age, cost, and LOS. The analysis included descriptive statistics, indirect standardization, Rao-Scott Chi-Square test, t-test, and adjusted linear regression trend.Hospitalizations were most prevalent for individuals ages 45-64 (38.8%), African Americans were overrepresented (25.8% of hospitalizations), and the gender distribution was nearly equivalent. Mean LOS was 9.08 days (95% confidence interval 8.71-9.45). Medicare was the primary payer for most hospitalizations (55.4%), with most of the costs ranging from $10,000-$49,999 (57.1%). The crude hospitalization rates ranged from 790-1142/100,000 admissions, while the US 2010 census standardized rates were 380-552/100,000 from 2005-2014. Linear regression trend analysis showed no significant difference in trend for race/ethnicity, age, nor gender (P > .001). The hospitalizations' overall rates increased while LOS significantly decreased, while hospitalization costs and Charlson's co-morbidity index increased (P < .001).From 2005-2014, the overall US hospitalization rates significantly increased. Over this period, observed disparities in hospitalizations for middle-aged and African Americans were unchanged, and LOS has gone down while costs have gone up. Further studies addressing the important disparities in race/ethnicity and age and reducing costs of acute hospitalization are needed.

How do motorcyclists manage mental tensions of risky riding?

BACKGROUND: Road traffic injuries, especially those involving motorcycles, are a particular concern in Iran. We aimed to identify the specific cognitive dissonances and consonances associated with risky riding among Iranian motorcyclists. METHODS: This was a grounded theory qualitative study of male motorcyclists who were ≥ 18 and were living in one of the three cities of Tehran, Isfahan and Ahwaz. Thirty four (n = 34) motorcyclists participated in 19 in-depth interviews and 5 focus-groups between January 2007 and February 2008. RESULTS: We identified four categories of motorcycle riders each endorsing a unique risk bias they employed to justify their risky ridings. The categories included: (1) Risk Managers who justified risky riding by doubting that it would result in negative outcomes if they are competent riders. (2) Risk Utilizers who justified risky riding as functional and practical that would enable them to handle daily chores and responsibilities more efficiently. (3) Risk Calculators who justified risky riding by believing that it will help them to avoid road crashes. (4) Risk Takers who justified risky riding by arguing that risky riding is thrilling and brings them peer recognition. CONCLUSION: Our findings reveal different groups of motorcyclists according to their different rationalizations for risky riding. Road safety advocates can benefit from our findings by matching relevant and appropriate interventions and incentives to these specific groups.

CEDRIC: a computerized chronic disease management system for urban, safety net clinics.

To meet the challenge of improving health care quality in urban, medically underserved areas of the US that have a predominance of chronic diseases such as diabetes, we have developed a new information system called CEDRIC for managing chronic diseases. CEDRIC was developed in collaboration with clinicians at an urban safety net clinic, using a community-participatory partnered research approach, with a view to addressing the particular needs of urban clinics with a high physician turnover and large uninsured/underinsured patient population. The pilot implementation focuses on diabetes management. In this paper, we describe the system's architecture and features.

Using the information-motivation behavioral model to predict sexual behavior among underserved minority youth.

BACKGROUND: Testing, refining, and tailoring theoretical approaches that are hypothesized to reduce sexual risk behaviors among adolescent subpopulations is an important task. Relatively little is known about the relationship between components of the information-motivation-behavior (IMB) model and sexual behaviors among underage minority youth. Using the IMB model, this study examines predictors of risky sexual behavior among underserved Hispanic and African-American youth. METHODS: This cross-sectional study was conducted with a sample of 380 youths aged 11-17 years recruited in Los Angeles, California, and utilized latent variable models to examine interrelationships and predictive relations among IMB model variables associated with risky sexual behavior. RESULTS: Sixty percent of the participants aged 15-17 and 1 out of 10 participants aged 11-12 reported prior sexual intercourse. Of the sexually active, more than half reported having unprotected sex and 11% had sexual intercourse with 4 or more partners. Results of the structural equation model indicated that older age and attitudes against sexual activities had significant, direct impacts on risky sexual behaviors. Behavioral refusal skills, positioned as an intervening variable, also significantly predicted less risky sex. Knowledge, attitudes against sexual activities, and perceived peer pressure against sexual behavior predicted sexual refusal skills. Additionally, there were significant indirect effects on risky sexual behavior mediated through behavioral refusal skills. CONCLUSION: A large number of disadvantaged minority urban youth are engaged in risky sexual behaviors. Intervention programs, particularly those targeting preadolescents, should focus on building long-lasting behavioral skills that emphasize the reduction of peer pressure and normative influences on risky sexual behaviors. Components of the IMB model clearly have a role in the design of efficacious interventions.

Comorbid chronic illness and the diagnosis and treatment of depression in safety net primary care settings.

OBJECTIVE: To estimate the impact of chronic medical conditions on depression diagnosis, treatment, and follow-up care in primary care settings. DESIGN: This was a cross-sectional study that used interviewer-administered surveys and medical record reviews. Three hundred fifteen participants were recruited from 3 public primary care clinics. Depression diagnosis, guideline-concordant treatment, and follow-up care were the primary outcomes examined in individuals with depression alone compared with individuals with depression and chronic medical conditions measured using the Charlson Comorbidity Index (CCI). RESULTS: Physician diagnosis of depression (32.6%), guideline-concordant depression treatment (32.7%), and guideline-concordant follow-up care (16.3%) were all low. Logistic regression analysis showed no significant difference in the likelihood of depression diagnosis, guideline-concordant treatment, or follow-up care in individuals with depression alone compared with those with both depression and chronic medical conditions. Participants with severe depression were, however, twice as likely to receive a diagnosis of depression as participants with moderate depression. In addition, participants with moderately severe and severe depression received much less appropriate follow-up care than participants with moderate depression. Among participants receiving a depression diagnosis, 74% received guideline-concordant treatment. CONCLUSION: Physician depression care in primary care settings is not influenced by competing demands for care for other comorbid medical conditions.

Correlates of self-diagnosis of chronic medical and mental health conditions in under-served African American and Latino populations.

OBJECTIVE: This study examines the correlates of self-diagnosis of chronic medical and mental health conditions in under-served minority populations. The Behavioral Model for Vulnerable Populations was employed to compare the predisposing and enabling characteristics of two groups: the first group consisted of individuals who self-reported their medical conditions without a presumptive or definitive physician diagnosis, while the second group consisted of individuals who self-reported their medical conditions with a presumptive or definitive physician diagnosis of their condition. STUDY SETTING: The sample consisted of 287 African American and Latino heads of household. This sample was obtained from a geographically defined random sample of 418 households from three urban public housing communities in Los Angeles County, California. STUDY DESIGN: This study was a cross-sectional, face-to-face, semistructured interview survey. RESULTS: Using logistic regression techniques and controlling for demographic characteristics, the results indicate that accessibility, affordability, continuity of medical care, and financial strains were the core concepts that explain the gap between self vs physician diagnosis of medical conditions. CONCLUSION: This study identifies unique characteristics of minority persons who claimed that their medical conditions had not been presented to or diagnosed by a medical provider in comparison to those who are formally diagnosed by medical providers. The study provides an entry point for further examination of correlates and sequels of self-diagnosis and its resultant effects on professional treatment-seeking in minority populations with certain medically important chronic conditions.

Correlates of complementary and alternative medicine utilization in depressed, underserved african american and Hispanic patients in primary care settings.

OBJECTIVES: This study seeks to examine the correlates of complementary and alternative medicine (CAM) use in depressed underserved minority populations receiving medical care in primary care settings. METHODS: A prospective study using interviewer-administered surveys and medical record reviews was conducted at 2 large outpatient primary care clinics providing care primarily to underserved African American and Hispanic individuals located in Los Angeles, California. A total of 2321 patients were screened for depression. Of these, 315 met the Patient Health Questionnaire-9 criteria for mild to severe depression. RESULTS: Over 57% of the sample reported using CAM sometimes or often (24%) and frequently (33%) for treatment of their depressive symptoms. Controlling for demographic characteristics, lack of health care coverage remained one of the strongest predictors of CAM use. Additionally, being moderately depressed, using psychotherapeutic prescription medications, and poorer self-reported health status were all associated with increased frequency of CAM utilization for treating depression. CONCLUSIONS: The underserved African American and Hispanic individuals meeting the diagnostic criteria for depression or subsyndromal depression use CAM extensively for symptoms of depression. CAM is used as a substitute for conventional care when access to care is not available or limited. Since CAM is used so extensively for depression, understanding domains, types, and correlates of such use is imperative. This knowledge could be used to design interventions aimed at improving care for depression.

Depression symptomatology and diagnosis: discordance between patients and physicians in primary care settings.

BACKGROUND: To examine the agreement between depression symptoms using an assessment tool (PHQ-9), and physician documentation of the same symptoms during a clinic visit, and then to examine how the presence of these symptoms affects depression diagnosis in primary care settings. METHODS: Interviewer administered surveys and medical record reviews. A total of 304 participants were recruited from 2321 participants screened for depression at two large urban primary care community settings. RESULTS: Of the 2321 participants screened for depression 304 were positive for depression and of these 75.3% (n = 229) were significantly depressed (PHQ-9 score > or = 10). Of these, 31.0% were diagnosed by a physician with a depressive disorder. A total of 57.6% (n = 175) of study participants had both significant depression symptoms and functional impairment. Of these 37.7% were diagnosed by physicians as depressed. Cohen's Kappa analysis, used to determine the agreement between depression symptoms elicited using the PHQ-9 and physician documentation of these symptoms showed only slight agreement (0.001-0.101) for all depression symptoms using standard agreement rating scales. Further analysis showed that only suicidal ideation and hypersomnia or insomnia were associated with an increased likelihood of physician depression diagnosis (OR 5.41 P sig < .01 and (OR 2.02 P sig < .05 respectively). Other depression symptoms and chronic medical conditions had no affect on physician depression diagnosis. CONCLUSION: Two-thirds of individuals with depression are undiagnosed in primary care settings. While functional impairment increases the rate of physician diagnosis of depression, the agreement between a structured assessment and physician elicited and or documented symptoms during a clinical encounter is very low. Suicidality, hypersomnia and insomnia are associated with an increase in the rate of depression diagnosis even when physician and self report of the symptom differ. Interventions that emphasize the use of routine structured screening of primary care patients might also improve the rate of diagnosis of depression in these settings. Further studies are needed to explore depression symptom assessment during physician patient encounter in primary care settings.