Faculty Bibliography

The quality of a cluster analysis of unlabeled units depends on the quality of the between units dissimilarity measures. Data dependent dissimilarity is more objective than data independent geometric measures such as Euclidean distance. As suggested by Breiman, many data driven approaches are based on decision tree ensembles, such as a random forest (RF), that produce a proximity matrix that can easily be transformed into a dissimilarity matrix. A RF can be obtained using labels that distinguish units with real data from units with synthetic data. The resulting dissimilarity matrix is input to a clustering program and units are assigned labels corresponding to cluster membership. We introduce a General Iterative Cluster (GIC) algorithm that improves the proximity matrix and clusters of the base RF. The cluster labels are used to grow a new RF yielding an updated proximity matrix which is entered into the clustering program. The process is repeated until convergence. The same procedure can be used with many base procedures such as the Extremely Randomized Tree ensemble. We evaluate the performance of the GIC algorithm using benchmark and simulated data sets. The properties measured by the Silhouette Score are substantially superior to the base clustering algorithm. The GIC package has been released in R: https://cran.r-project.org/web/packages/GIC/index.html.

The United States Preventive Services Task Force is perhaps America's best-known source of evidence-based medicine (EBM) recommendations. This paper reviews aspects of the history of one such recommendation-screening for colorectal cancer (CRC)-to explore how the Task Force evaluates the best available evidence to reach its conclusions.Although the Task Force initially believed there was inadequate evidence to recommend CRC screening in the 1980s, it later changed its mind. Indeed, by 2002, it was recommending screening colonoscopy for those aged 50 and older, "extrapolating" from the existing evidence as there were no randomized controlled trials of the procedure. By 2016, due in part to the use of an emerging analytic modality known as modeling, the Task Force supported four additional CRC screening tests that lacked randomized data. Among the reasons the Task Force gave for these decisions was the desire to improve adherence for a strategy-screening healthy, asymptomatic individuals-that it believed saved lives.During these same years, the Task Force diverged from other organizations by declining to advocate screening otherwise healthy Black patients earlier than age 50-despite the fact that such individuals had higher rates of CRC than the general population, higher mortality from the disease and earlier onset of the disease. In declining to extrapolate in this instance, the Task Force underscored the lack of reliable data that proved that the benefits of such testing would outweigh the harms.The history of CRC screening reminds us that scientific evaluation relies not only on methodological sophistication but also on a combination of intellectual, cognitive and social processes. General internists-and their patients-should realize that EBM recommendations are often not definitive but rather thoughtful data-based advice.

More than 61,000 persons below the age of 18 are living on the streets in the Greater Accra region in Ghana. Street children is a hidden vulnerable population and a global public health issue in the world, but little is known about their mental health and health needs, and mechanisms that contribute to their poor health. With a lack of mental health research to guide intervention or psychoeducation programme and policy planning, this study aimed to address these research gaps by examining prevalence of mental health problems and a set of associated risk factors (i.e. Perceived quality of life, and social connection). In addition, we examined whether the associations between risk factors and mental health problems were moderated by demographic and contextual factors (i.e., gender, age, work status, reason for living on street, number of years in street). Two hundred and seven children between age 12 and 18 who lived on the street in three cities (Accra, Sekondi Takoradi, and Kumasi) were recruited. Data were gathered through adolescent survey/interviews. Multiple regression was utilized to examine risk factors and moderation effects. Results support high mental health needs among street children. Approximately 73% street children experienced moderate to severe mental health problems, and 90% experienced poor quality of life. Perceived quality/happiness of life was the strongest predictor for street children's mental health. Social connection was associated with children's mental health only in certain subgroups and contexts. This study adds new epidemiological evidence for street children, an extremely vulnerable population, in Ghana and global child and adolescent mental health.

The use of social media continues to increase in health care and academia. Health care practice, particularly the oncologic field, is constantly changing because of new knowledge, evidence-based research, clinical trials, and government policies. Therefore, oncology trainees and professionals continue to strive to stay up-to-date with practice guidelines, research, and skills. Although social media as an educational and professional development tool is no longer completely new to medicine and has been embraced, it is still under-researched in terms of various outcomes. Social media plays several key roles in professional development and academic advancement. We reviewed the literature to evaluate how social media can be used for professional development and academic promotion of oncology professionals.

INTRODUCTION:Maternal asthma in pregnancy is associated with adverse perinatal and child health outcomes; however, mechanisms are poorly understood. METHODS:The PRogramming of Intergenerational Stress Mechanisms (PRISM) prospective pregnancy cohort characterized asthma history during pregnancy via questionnaires and quantified placental DNAm using the Illumina Infinium HumanMethylation450 BeadChip. We performed epigenome-wide association analyses (n = 223) to estimate associations between maternal active or inactive asthma, as compared to never asthma, and placental differentially methylated positions (DMPs) and differentially variable positions (DVPs). Models adjusted for maternal pre-pregnancy body mass index, smoking status, parity, age and education level and child sex. P-values were FDR-adjusted. RESULTS:One hundred and fifty-nine (71.3%) pregnant women reported no history of asthma (never asthma), 15 (6.7%) reported inactive, and 49 (22%) reported active antenatal asthma. Women predominantly self-identified as Black/Hispanic Black [88 (39.5%)] and Hispanic/non-Black [42 (18.8%)]. We identified 10 probes at FDR<0.05 and 4 probes at FDR<0.10 characterized by higher variability in maternal active asthma compared to never asthma mapping to GPX3, LHPP, PECAM1, ATAD3C, and ARHGEF4 and 2 probes characterized by lower variation mapping to CHMP4A and C5orf24. Amongst women with inactive asthma, we identified 52 probes, 41 at FDR<0.05 and an additional 11 at FDR <0.10, with higher variability compared to never asthma; BMP4, LHPP, PHYHIPL, and ZSCAN23 were associated with multiple DVPs. No associations were observed with DMPs. DISCUSSION:We observed alterations in placental DNAm in women with antenatal asthma, as compared to women without a history of asthma. Further research is needed to understand the impact on fetal development.

BACKGROUND:Little is known about the epidemiology of diagnosis in primary care. METHODS:A prospective observational cohort study was conducted of adults presenting between August and December 2018 to primary care clinics across two health systems with an undiagnosed medical problem. Primary outcomes were (1) likelihood of a definitive diagnosis by 12 months and (2) time to diagnosis. Multivariate logistic regression was used to assess for factors associated with the likelihood of reaching a diagnosis, and multivariable Cox regression was used to assess for factors associated with time to diagnosis. Bivariate models were used to explore unadjusted relationships between the cases' organ systems and likelihood of and time to diagnosis. RESULTS:Among 410 cases in a diverse patient population, 206 (50.2%) reached a final diagnosis within 12 months, with a median time to diagnosis of 5 days (interquartile range = 0-46). Among these cases, 32.4% reached a diagnosis within the first month. A majority of cases not diagnosed within a month of the first presentation remained undiagnosed at 12 months. The likelihood of diagnosis and time to diagnosis did not differ by clinician or patient characteristics, clinicians' level of diagnostic uncertainty, chronicity of the medical issue, or visit type. There were no significant associations between organ system and likelihood of time to diagnosis. CONCLUSION/CONCLUSIONS:Patients presenting with new or unresolved problems in ambulatory primary care often remain undiagnosed after a year. There were no provider or patient-level variables associated with such lack of diagnosis. The causes, contributors, and consequences of lack of timely diagnosis and potential solutions require further research.

BACKGROUND:There is pressing need to improve hospital-based addiction care. Various models for integrating substance use disorder care into hospital settings exist, but there is no framework for describing, selecting, or comparing models. We sought to fill that gap by constructing a taxonomy of hospital-based addiction care models based on scoping literature review and key informant interviews. METHODS:Methods included a scoping review of the literature on US hospital-based addiction care models and interventions for adults, published between January 2000 and July 2021. We conducted semi-structured interviews with 15 key informants experienced in leading, implementing, evaluating, andpracticing hospital-based addiction care to explore model characteristics, including their perceived strengths, limitations, and implementation considerations. We synthesized findings from the literature review and interviews to construct a taxonomy of model types. RESULTS:Searches identified 2,849 unique abstracts. Of these, we reviewed 280 full text articles, of which 76 were included in the final review. We added 8 references from reference lists and informant interviews, and 4 gray literature sources. We identified six distinct hospital-based addiction care models. Those classified as addiction consult models include (1) interprofessional addiction consult services, (2) psychiatry consult liaison services, and (3) individual consultant models. Those classified as practice-based models, wherein general hospital staff integrate addiction care into usual practice, include (4) hospital-based opioid treatment and (5) hospital-based alcohol treatment. The final type was (6) community-based in-reach, wherein community providers deliver care. Models vary in their target patient population, staffing, and core clinical and systems change activities. Limitations include that some models have overlapping characteristics and variable ways of delivering core components. DISCUSSION/CONCLUSIONS:A taxonomy provides hospital clinicians and administrators, researchers, and policy-makers with a framework to describe, compare, and select models for implementing hospital-based addiction care and measure outcomes.