Welcome Reception & Poster Session I

This study aims to understand the effect of survey timing on the responses regarding experiences of advanced cancer patients by comparing those who were diagnosed six and three years ago. We hypothesized that the responses of cancer patient experiences may change over time because of the change in the patient's surroundings as well as the changes in types of patients who return the survey. We conducted the nationwide cancer patient experience survey in 2019, for the purpose of evaluating the progress of the Third Cancer Control Plan in Japan. Self-administered questionnaire was distributed to patients with six or three years after diagnosis (Group 1 and Group 2, respectively). We analyzed their answers using design-based weights. Most of the results did not differ between the two groups but responses were different in several aspects of their experiences. For example, overall assessment of treatment on a scale 0-10 tended to be better among the Group 2 (7.47 vs 7.26, p=0.19), and larger proportion of the Group 2 knew about second opinion or cancer support center (38.4% vs 29.9% p=0.03; 74.3% vs 65.6%, p=0.04, respectively). In contrast, more of the Group 1 answered they were able to go out their daily life than the Group 2 (72.4% vs. 63.7%, p=0.03). More respondents in the Group 1 answered that they felt discriminated due to cancer and physical distress compared to the Group 2 (5.8% vs. 2.4%, p<0.01; 59.3% vs. 41.5%, p<0.01, respectively). These differences may be attributable to several factors. From the policy perspective, the National Cancer Control Acceleration Plan was implemented in 2015. Also, some advanced cancer patients may have had quality of life adjustment. Further research is needed to understand the factors for change and differences in order to evaluate the progress of cancer control based on patient experience. 

Healthcare costs are increasing at alarming rates in the United States (US) putting a heavy burden to the healthcare reimbursement system. Cost and cost savings have become an important focus as health policy administrators are tasked with determining the most effective allocation of limited resources. The availability of large databases, such as administrative data, comes with many challenges for analyses, including: skewed data, inflated zero counts, and potential selection bias among comparison groups. Thus, it is imperative that they are evaluated correctly. There are many different methods currently being used to estimate costs including: generalized linear models with a log link, natural logarithm transformed costs, gamma distribution, median regression, two-part models, and Bayesian models. This review will identify which methods are statistically and mathematically appropriate for large claims data.
Scopus and Ovid were searched for potential statistical method papers using multivariable modelling of cost. Inclusion criteria required either a comparison of two or more statistical methods to analyze cost or one statistical method performed on two or more different types of cost data.
The review identified 1,048 potential papers, of which, 80 met the inclusion criteria for a full article review. There was a total of 9 papers included in the review; one paper looked at simulations and eight papers assessed real cost data. There were 28 models assessed across the nine papers with ordinary least squares (OLS) and generalized linear models (GLM) being the most common.
GLM using the gamma distribution was included in all but two of the comparisons. Most other models that were compared to the GLM Gamma distribution with log link found it to be the superior model in both simulated data and real administrative data. 

During the COVID-19 pandemic, there was a dramatic decrease in the percentage of patients admitted to the hospital through the emergency department. One study showed a 32% decrease in admissions from March 3, 2020, through September 8, 2020 (Nourazari, et al. 2021). The decreased hospitalizations affect patients' access to treatment for chronic conditions, such as gout.

Gout is a common hyperuricemic metabolic condition that leads to recurrent inflammatory arthritis and subsequently severe pain. High uric acid levels can form tiny crystals that lodge in joints, as well as accumulate in the kidneys, causing kidney stones. Patients affected by this condition experience a higher prevalence of metabolic syndrome and risk of cardiometabolic comorbidities (Choi, et al. 2022). According to a Global Burden of Disease analysis of 195 countries and territories from 1990 to 2017, the incidence, prevalence, and disability burden of gout have all increased worldwide in recent decades. In 2016, gout affected more than 9.2 million US adults (Chen, 2019).

Hospitalizations and emergency room visits related to the treatment of gout have increased over the past two decades. US hospitalization rates due to gout doubled from 1993 to 2011 (Lim, et al. 2016). From 2009 to 2012 alone, annual ED visits for gout in the US increased from 180,789 to 205,152 (Singh & Yu, 2016).

The COVID-19 pandemic had an impact on hospital utilization and how patients received care for chronic conditions such as gout. To better understand the impact of COVID-19 on hospital utilization and care received for gout, this study will use the National Hospital Care Survey (NHCS) to analyze hospital admissions and emergency department care data for patients with gout during 2020 and 2021.

NHCS is an annual administrative data collection of a year of inpatient and emergency department encounter-level data from a sample of non-institutional and non-federal hospitals with six or more staffed inpatient beds. NHCS provides statistics on health care utilization and care to answer key questions of interest to public health professionals, researchers, and health care policy makers. NHCS data are unweighted and are not nationally representative.

To meet the need for hospital data during the COVID-19 pandemic, NHCS released preliminary 2020 and 2021 inpatient and emergency department data from approximately 50 hospitals to develop a NHCS COVID-19 dashboard reporting on hospital care related to COVID-19. The preliminary NHCS data are not nationally representative, but exploratory analysis of hospital care can be used to provide insight on emerging health care trends. In the preliminary NHCS data from January 2020 through December 2021, there were 30,157 inpatient admissions and 34,376 emergency department visits for patients with a diagnosis of gout. This study will present an overview of the hospital admissions and emergency visits concerning gout during the COVID-19 pandemic, highlighting changes of gout-related hospital utilization compared to increases in COVID-19 hospitalizations. These data demonstrate the impact of COVID-19 on the treatment of chronic conditions, such as gout flares.

References:
Choi, H.K., et al. Excess comorbidities in gout: the causal paradigm and pleiotropic approaches to care. (2022).

Nourazari S, et al. Decreased hospital admissions through emergency departments during the COVID-19 pandemic. Am J Emerg Med. 2021

Lim SY, Lu N, Oza A et al. Trends in gout and rheumatoid arthritis hospitalizations in the United States, 1993–2011. JAMA 2016;315:2345–7.

Chen-Xu, M., et al. (2019), Contemporary Prevalence of Gout and Hyperuricemia in the United States and Decadal Trends: The National Health and Nutrition Examination Survey, 2007–2016

Singh JA, Yu S. Time Trends, Predictors, and Outcome of Emergency Department Use for Gout: A Nationwide US Study. J Rheumatol. 2016 Aug 

We are focusing on a type of data set that emerge recently by the advance of single-cell technologies - multi-modal data, which are usually refer to the data set with multiple feature sets pointing to the same individual. With the development of recent single-cell technologies, it has generated a great deal of excitement and interest in studying functional genomics at cellular resolution. For example, recent Patch-seq techniques enable measuring multiple characteristics of individual neuronal cells, including transcriptomics, morphology, and electrophysiology in the complex brains, also known as single-cell multi-modal data. More detailly, Patch-seq experiments profile the electrophysiological properties and transcriptome of the same individual neurons, with the goal of identifying the underlying relationships between gene expression and neuronal function. Additionally, recorded neurons can be backfilled with appropriate dies to evaluate cell morphology.

We used nonlinear manifold alignment to align multiple features of single-cell data, which are cell transcriptomics and electrophysiological features. It provided a manifold workflow that applies statistical and machine learning methods to reduce the high dimensional multi-modal data into a 3D manifold, followed by clustering by Gaussian Mixture Kernel, electrophysiological feature prediction, functional enrichment, and gene regulatory network analyses, we showed the underlying relationships between gene expression and neuronal function. The good performance of nonlinear manifold alignment, compared to other methods on single-cell multiple data, suggests that our method is interpretable, able to show the 3D trajectory that have not been shown by other studies, and have a strong prediction power between identified features, thus should be useful and appliable to all similar data types in different industries or application areas as well. 

Objective
In 2014, Berkeley, California, became the first US jurisdiction to pass a sugar sweetened beverage (SSB) excise tax (1 cent per ounce). Three other localities in California (Albany, Oakland, San Francisco) followed suit and approved similar 1-cent-per-ounce taxes in 2016. We aim to conduct the first large-scale intermediate evaluation of SSB excise tax in these cities. This is also the first time to include obesity as health outcome.
Research Design and Methods
We used California Health Interview Survey (CHIS) adult data from 2009 to 2021. We obtained SSB excise tax information from city council websites. We assumed that once the SSB excise tax policy was implemented it remained so unless explicitly repealed at some point. The policy was implemented in 2015 in Berkeley, 2018 in San Francisco, 2017 in Oakland and Albany. We estimated the number of times drinking soda per week as well as obesity prevalence. Data for soda consumption is available until 2018. A BMI ≥30 kg/m2 is obese for adults 20 years and older. Propensity score matched difference-in-difference (DID) was implemented. We selected controls from non-SSB tax cities. Controls were matched on age, gender, race/ethnicity, marital status, smoke, education, employment, poverty, general health conditions, diabetes, and high blood pressure.
Results
Soda tax decreased the number of times an individual consumes soda per week in Berkeley (ATT, average treatment effect among treated=-0.61, 95%CI -0.69, -0.54). Soda tax decreased obesity prevalence in Berkeley, Oakland and San Francisco (ATT=7%, 95%CI -8%, -6%; ATT=4%, 95%CI -5%, -3%; ATT=8%, 95%CI -9%, -7%. There was no change in Albany (ATT=0, 95%CI -2%, 1%).
Conclusion
SSB excise tax decreased soda consumption and obesity prevalence in most cities. It can be an effective policy response to reduce SSB consumption and build healthier beverage environments.
Abstract word count: 290 

Studies have indicated that consumers use rating and review data to inform their purchasing decisions (e.g., Amazon, Yelp). The ride sharing platform Uber even allows drivers to rate and review passengers - which can affect the time that a customer has to wait for a driver. In this paper, we argue that, given the unique distributional attributes of ratings data, the results are not being correctly interpreted. Here we explore how physicians and healthcare providers should be considering ratings data that patients are posting on platforms such as Vitals, RateMDs, and Healthgrades to provide a clearer picture of the patient experience. 

Background:
Network analyses describe complex patterns through graphical displays and may help detect novel associations using empirical data. Network science was first used to analyze social relationships, and is increasingly used to examine patterns of co-occurring health conditions or health service use. Networks are particularly useful for populations with complex health status or healthcare needs. Characterizing health condition co-occurrence networks by the number and strengths of connections (i.e., relationships) can aid in better understanding health needs among populations.

Dementia is an incurable neurodegenerative condition that impacts an individual's cognition, independence, and life-expectancy. Persons with dementia (PWD) residing in the community have different access to health resources than PWD in long-term care (LTC). Multimorbidity, the co-occurrence of multiple chronic health conditions, is common among PWD. As a consequence of multimorbidity, many PWD are cared for by multiple healthcare professionals. Understanding profiles of chronic health conditions will help identify areas where coordination between healthcare services is necessary.

Purpose and Objectives:
To compare patterns of multimorbidity among PWD residing in long-term care (LTC) and the community in the province of Manitoba, Canada, using network analysis. Demographic differences in multimorbidity patterns will be evaluated by stratifying networks by sex, age (67-74, ≥75 years).

Methods:
This study will use population-based administrative databases from the Manitoba Population Research Data Repository, including outpatient claims, inpatient records, community-based pharmaceutical dispensations, and LTC records. These data were extensively validated to investigate chronic health conditions and were used to generate multimorbidity networks. This retrospective cohort will consist of PWD ≥67 years with ≥2 other chronic health conditions residing in Manitoba from 2015 to 2020 in LTC and community settings. LTC in Manitoba are regulated residences where healthcare workers provide 24-hour care and are not used for sub-acute care or rehabilitation. Chronic conditions will be identified using the open-source Clinical Classification System, which encompasses 130 clinically relevant chronic condition categories.

Non-directional networks, consisting of nodes (chronic health conditions among PWD) connected by edges (cosine index, a metric that quantifies the strength of association between pairs of health conditions), will be generated based on residence location (LTC, community). Only statistically significant associations (edges) will be included in networks; determined by Pearson chi-square controlling for false discovery rate. Networks will be portrayed through graphical displays. Networks' properties will be described, including the number and distribution of: node, edge, and node degrees (i.e., associations with other diseases for a given condition), will be described. The Louvain community detection algorithm will be used to identify clusters of closely associated health conditions. Modularity (a measure of the extent to which a network divides into clusters) will be calculated.

Significance and Impact:
Our study will use unique visual and inferential techniques to describe the complexity of multimorbidity among PWD in LTC and the community. Characterizing profiles of chronic conditions is an important step towards developing policies or improving services aimed to provide high quality care for PWD. Health services that frequently interact due to co-occurring chronic conditions can be improved to avoid fragmentation and engage collaboratively in patient-centred decision making. Cross-specialty collaborative care models and multidisciplinary teams can be designed to address co-occurring conditions. Community and LTC programs can more efficiently allocate resources to meet intricate health needs. 

Nearest neighbor (NN) matching is a conceptually natural and practically well-used tool to align data sampled from different groups. In a landmark paper, Abadie and Imbens (2006) provided the first large-sample analysis of NN matching. Their theory, however, requires a crucial assumption that the number of NNs, M, is fixed. We reveal something new out of their study and show that, once allowing M to diverge with the sample size, an intrinsic statistic in their analysis actually constitutes a consistent estimator of the density ratio. Furthermore, we show that through selecting a suitable M, this statistic can attain the minimax lower bound of estimation over a Lipschitz density function class. Consequently, with a diverging M, the NN matching with Abadie and Imbens (2011)'s bias correction provably yields a doubly robust estimator of the average treatment effect and is semiparametrically efficient if the density functions are sufficiently smooth and the outcome model is appropriately specified. It can thus be viewed as a precursor of the recently proposed double machine learning estimators. 

Data analyses are predicated on assumptions, many of which are not directly testable. A key assumption underlying observational research methods for analysis of real-world data (e.g. health insurance claims or electronic health records) is that treatment assignment is independent of the potential outcomes. In other words, an unbiased estimate of the treatment effect is obtained only when there are no unmeasured confounders that are related to both who receives the treatment and the outcomes of interest. However, it is not possible to directly test for the impact of covariates that are unobserved to the analyst.

Our study uses Monte Carlo simulations to evaluate the impact of unobserved confounders on the treatment effect estimates and to evaluate the performance of the E-Value sensitivity test with the application of regression and propensity score methods under varying levels of unobserved confounding. Specifically, we compare observed and unobserved confounder balance, odds ratios of treatment vs. control, and E-Values sensitivity test statistics from GLM regression models, inverse-probability weighted models, and propensity score matching models, over correlations of increasing strength between observed and unobserved confounders. The E-Value sensitivity test to assess the effect of unmeasured confounding is notable for its ease of implementation and interpretation. The E-value reports the minimum strength of association between an unmeasured confounder and the treatment and outcome that would explain away the estimated treatment effect.

We find that, while propensity score methods balance observed confounders, they may exacerbate imbalances in unobserved confounders resulting in biased treatment effect estimates. Moreover, we find that E-values calculated after applying propensity score methods tend to be larger when unobserved confounders result in more biased treatment effect estimates. This result has important implications for the appropriate application and interpretation of common statistical methods and sensitivity testing. First, we confirm previous findings that the propensity score methods – matching or weighting – may increase the imbalance in unobserved confounders. The magnitude of the effect depends on the strength of correlation between the confounder, treatment, and outcomes. In all cases, this implies that propensity score methods are only appropriate to use when the underlying assumptions can be justified by knowledge of the treatment context and data source. Second, the E-Value may misrepresent the size the unobserved effect needed to change the magnitude of the association between treatment and outcome when propensity score methods are used.

Thus, caution is warranted when interpreting the E-Value. Sensitivity testing is an important element of any analysis but is not a substitute for a well-informed study design. 

Background: The purpose of the study was to explore patient recovery time association through inference of physical exercise mediator part and moderator role of COVID-19 vaccine.
Methods: Survey sample size was 659 COVID-19 positive patients who received treatment either at home or in hospitals in Dhaka city, Bangladesh. A cross-sectional study data was collected from May to August 2021. Statistical analysis was performed to assess the potential relationship, and the mediation effect was obtained from a bootstrap approach (n=5000) with SPSS version 26.
Result: Two process models (4 and 58) were tested to investigate the association. The mean age of the patients was 38.43 years (SD=13.89) counting males (39.8%) and females (60.2%). Mediation model 4, using physical exercise (M) as mediator variable, total effect (effect= 0.0778, t =2.1932, P < 0.05), and direct effect (effect= 0.0728, t =2.0643, P < 0.05) of existing comorbidity (X) on recovery time (Y) that was significant, but bootstrap indirect effect was insignificant [impact= (0.0050), BootLL 95%CI (-0.0007) - BootUL 95%CI (0.0128)]. Next, moderated process model 58, comorbidity (X), physical exercise (M) is associated significantly (p<0.05) but the moderator variable Covid-19 vaccine (W) was found insignificant (p>0.05), and interaction was significant [t=1.9873, p<0.05, 95%Confidence Interval (C.I.) =0.0005-0.0983]. Furthermore, conditional effects of focal predictor at values of moderator Covid-19 vaccine who not taken was significant [effect=-0.0494, P<0.05, 95% C.I.= -0.0815 to -0.0173]. Nevertheless, an indirect effect of the bootstrap re-sampling method included value of zero and moderated mediation negative index (-0.009), indicating insignificant association with moderated mediation model of the hypothesis.
Conclusion: Results suggest that moderated mediation model analysis may be effective to reduce curative time among Covid-19 affected patients by maintaining regular physical exercise and vaccination who did not receive. 

Suicide is a significant cause of premature death. Suicide is a low base-rate event, 10 suicides per 100,000 per year in Canada, and this makes research challenging. Its predictors exist commonly within people with mental illness, making it difficult to parse out those who will have elevated suicidal risk from those who will not. Many investigations focus on the more readily identifiable stages of the suicidal spectrum: suicidal ideation and suicide attempts rather than suicide deaths. Investigations on help-seeking behaviours of suicidal individuals largely focus on utilization of formal mental health services due to the prevailing conception of suicide as heavily linked to mental illness.
Suicide studies typically use clinical samples, patients with schizophrenia for example, or sample high-risk groups such as adolescents and veterans. In contrast, the studies in this proposal will focus on the general population, and use a large community-representative survey, the Canadian Community Health Survey (CCHS), which has been linked to other health administrative data as well as mortality data. With these large and linked datasets, a unique opportunity exists to study the rare phenomenon of suicidality: sourcing individuals who self-reported suicidal ideation and their rich sociodemographic characteristics from the CCHS, then following these subjects over time to examine their service utilization and suicide attempts patterns using health administration databases, and, if they have died during the follow-up periods, causes of mortality using the Canadian Vital Statistics Database.
Extant literature in this area is also largely cross-sectional in design which, although useful in assessing prevalence and identifying risk factors, do not allow causal inferences to be made. The proposed studies will add value due to their prospective cohort design and multiple follow-up periods (1 year for service utilization, and 1, 3, and 5 years for suicide attempts or suicides) which will provide information of what happens to people with suicidal ideation (PWI) as time passes.
The Poster will present the protocol for three papers: a Scoping Review will be conducted to examine the association between mental health service provision and suicide rates in the general population. The review is near completion and is revealing mixed results (both positive and negative correlations) on this association. Therefore, it is uncertain if mental health service provision makes a difference in improving the suicide rates in the general population.
Second, a prospective cohort study will be conducted on PWI and their use of mental health services by frequency and type over a one-year follow-up period. The study will explore whether PWIs of higher severity use more and specific types of services, compared to those of lesser severity, and compared to persons with no suicidal ideation (PWNI). The impact of mental and substance use disorders will be studied to see if these disorders have a moderating impact on the relationship between suicide ideation and mental health service utilization.
Third, a prospective cohort study will be conducted on PWI and their risk of attempting suicide or dying by suicide. The study will explore if PWIs of greater severity are at higher and earlier risk of suicidal behaviour (attempting suicide or suicide) over 1-, 3- and 5-year follow-up periods. The impact of mental and substance use disorders will be studied to see if these disorders have a moderating impact on the relationship between suicide ideation and suicidal behaviour.
Poisson regression analysis will be used to analyze health service utilization. Cox proportional hazard models will be used to explore suicide attempts and suicide outcomes. Additionally, competing risk analysis will be performed to take into account competing causes of mortality, using the Fine Gray model. 

Background. As patients' social determinants of health (SDoH) data are increasingly collected in electronic health records (EHRs), they can potentially help providers tailor clinical recommendations for individuals who are at risk of negative health outcomes, facilitating referrals to community services. The National Academy of Medicine released recommendations on core measures of SDoH that should be documented in EHRs. However, the risk categories from these SDoH measures have not been investigated for their associations with health outcomes. To support meaningful use of the collected SDoH information, our aim was to investigate the added value of tracking SDoH information beyond routinely collected demographic information.

Methods. The eligible sample was composed of all adult patients (age 18 or older) who had a primary care provider in Mayo Clinic and responded to the SDoH questionnaire from July to December 2019 across Mayo Clinic campuses (MN, AZ, FL) and Mayo Clinic Health System (SE MN, SW MN, SW WI, and NW WI). Outcomes included any hospitalization and any emergency department (ED) visit in 2020, which were pulled from the medical records. Other outcomes included patient-reported feelings of depression and anxiety in 2020, which were reported in the current visit information (CVI) questionnaire. Patients who answered the enterprise-system-triggered SDoH questionnaire in the latter half of 2019 and CVI questionnaire in 2020 were included in the final sample.

Logistic regression models were estimated for each combination of the four outcomes and nine SDoH domains (i.e., alcohol use, financial resource strain, food insecurity, intimate partner violence, physical activity, social connections, feeling stressed, tobacco use, and transportation needs). While there were no missing data on the outcomes, missing data rates on the SDoH domains ranged from 1.7% to 42.6% (median=11.9%). Missing data were addressed via multiple imputation. The covariates included age, female, white race, Hispanic, being in committed relationship, education level, Charlson comorbidity index, and geocoded SDoH (i.e., area deprivation index). To address a time effect due to observations being collected at different times, the amount of time between SDoH answer date and either the last date of 2020 or date of death was used as an offset variable. An α level of 0.001 was used. We calculated the odds ratios (ORs) and interpreted ORs of 1.46 or greater to be meaningful.
Results. In our sample (N=159,258), 8.1% were hospitalized, 18.0% had ED visit, 7.5% reported depression, and 13.6% reported anxiety in 2020. Among those who were not-at-risk for each SDoH domain, the average proportion of patients experiencing hospitalization was 7.3%, 16.3% for ED visit, 6.4% for feeling depressed, and 11.44% for feeling anxious. The strongest predictors for ED visit were high-risk financial resource strain, food insecurity, and unmet transportation needs. The strongest predictor for hospital admission was physical inactivity in the prior year. In addition, feeling stressed followed by intimate partner violence were most predictive of feeling depressed or anxious. Social isolation was also significantly associated with feeling depressed. Being a heavy drinker, being at medium risk for financial resource strain, insufficient physical activity, moderate social isolation, and being a former smoker were not associated with any of the studied outcomes at a meaningful level.
Conclusions. Some risk categories predicted health outcomes, while others did not. Implications for further actions would be to provide alerts to providers and recommendations or referrals to community resources for patients emphasizing those SDoH domains that were found to be predictive of adverse outcomes. Future studies should investigate whether providing these recommendations and connecting patients with community resources is associated with improved patient outcomes. 

Objective: To examine factors associated with racial inequities in discharge location, skilled nursing facility (SNF) utilization, and readmissions.
Data Sources: A 20% sample of longitudinal Medicare claims from 2016 to 2018.
Study Design: We present layered target matching, a method for studying sources of inequities. Layered target matching examines a fixed target population profile representing any race, ethnicity, or vulnerable population, sequentially adjusting for sets of characteristics that may contribute to inequities these groups endure. We demonstrate the method in a study of racial inequities in post-acute care use and readmissions, sequentially matching first to a demographics target, then a richer target adding reasons for admission and clinical characteristics on admission, and finally a third target further adding hospital characteristics and complications experienced during the hospitalization. This process helps clarify potential sources of differences in post-acute care use and readmissions, enabling policy makers to address them more efficiently. Using recent methods for describing the populations implicitly targeted by linear regression, we also investigate the implications of choice of adjustment method for disparities research in populations with limited covariate overlap; specifically, the highly segregated cities of Chicago and Detroit.
Data Collection/Extraction Methods: We studied Black and Non-Hispanic White fee-for-service Medicare beneficiaries aged 66+ admitted to short-term acute-care hospitals for qualifying diagnoses or procedures between 1/1/2016 and 11/30/2018.
Principal Findings: Admitted Black patients tended to be younger, had significantly higher rates of risk factors such as diabetes, stroke, or renal disease, and were much more frequently admitted to large or academic hospitals. Relative to demographically similar White patients, Black patients were significantly more likely to be discharged to SNFs (21.8% vs. 19.3%, difference=2.5%, P<0.0001) and to receive any SNF care within 30 days of discharge (25.3% vs. 22.4%, difference=2.9%, P<0.0001). Black patients were also significantly more likely to experience 30-day readmission (18.7% vs. 14.5%, difference=4.2%, P<0.0001). Differences in reasons for hospitalization and risk factors explained most of the differences in discharge location, post-acute care use, and readmission rates, while additional adjustment for differences in hospital characteristics and complications made little difference for any of the measures studied. Finally, the target populations implied by conventional regression models fit on the highly segregated cities of Chicago and Detroit appeared to be much healthier than the sample of Black patients in each city, with the implied means of some important risk factors such as emergent admission and renal disease being approximately 10% lower than those in the true samples.
Conclusions: We found significant Black-White differences in discharge to SNFs, SNF utilization, and readmission rates. Using layered target matching, we found that differences in risk factors and reasons for hospitalization explained most of these differences, while differences in hospitals did not materially impact the differences. These findings suggest that policies targeting preventive care and reducing differences in occurrences of admissions requiring more intensive post-acute care may possibly be more successful than policies targeting hospitals. 

As of June 20, 2022, approximately 540.963.088 cases of the disease and its variants have been reported in approximately 212 countries and territories worldwide according to the Johns Hopkins COVID monitoring database with a total of 6.323.007 deaths. On March 5, 2020, the first case of COVID-19 was reported in Costa Rica, which as of March 16, 2022, has evolved into a total of 827,426 confirmed cases (WHO, 2022). Government authorities have generated specific policies aimed at containing the spread of the disease, including the start of vaccination in December 2020, and the deployment of institutional action from various sides, in order to address the emergency, contain the spread of the disease and address the economic and social consequences. On March 16, 2020, through Executive Decree Nº42227-MP-S, a state of national emergency was declared throughout the territory of the Republic, and the phases of attention were established in accordance with the National Law on Risk Prevention and Attention to Risks and Attention to Emergencies[3] (Response, Rehabilitation and Reconstruction). According to the study conducted by Rojas and Romero (2021) which a Bayesian model revealed that as the number of COVID-19 cases increased, the measures generated the necessary counterweight to mitigate infection, while the diminished or slowed effect of stopping the growth was mainly noted in the epidemiological weeks when there was a strong acceleration preceded by a relaxation of measures. This complements a preliminary reading of the effect of the measures on the behavior of the pandemic, linked precisely to the intention of analyzing the political and institutional components and their impact on the decision-making processes. From the perspective of emergency attention and public administration, various analytical approaches have shown structural challenges in the face of the attention of high-impact emergency situations[3], linked to the articulation of public policy mechanisms, the multi-hazard scenario of the territory, the design of public policies and the use of follow-up and monitoring mechanisms, interlocution between actors, leadership[4], available resources, risk management system, coordination with international resources, institutional framework, and growing and changing threats and vulnerabilities in the world. Considering the lessons learned from one of the most relevant public health phenomena of the 21st century, which has generated a significant amount of knowledge, this work aims to collect it at the level of public administration and the management of emergencies at the national and local level. This narrative explores lessons learned in three areas: governance and public management; pandemic planning and management; and the risk management and emergency care system. From the perspective of governance and public management, an analysis of the variables of assessment of the decision-making process, inter-institutional coordination, transparency and accountability is incorporated, with special emphasis on two key aspects: monitoring and evaluation, as well as the use of evidence for decision-making.The use of a centralized system for data collection, the decision-making mechanisms and how a transitional organizational structure was set up to make decisions in an expeditious and concrete manner are highlighted. In this sense, the research contributes to the analysis of the use of data for decision making, highlighting fundamental elements around: the agents and actors involved, the impact on public policy and the identification of good practices from the institutional framework to strengthen national statistical systems. Investigation seeks to articulate a memory that will allow various sectors to have a starting point in the identification strengths and challenges in the the achievement of sustainable development and recovery in the medium and long term through efficient and resilient public administration. 

Objective: To study the association between a prescribing physician's position in a physician shared-patient network and their patients' receipt of risky drug combinations.

Data Sources: All medical encounters in Medicare fee-for-service for Ohio-residing beneficiaries were used to form a shared-patient network of physicians based in Ohio, and Part D prescription drug events for beneficiaries receiving opioids, benzodiazepines, or non-benzodiazepine sedative hypnotics (sedative hypnotics) in 2014 prescribed by these physicians.

Study Design: In this retrospective, observational study we assigned patient prescription receipt to time-varying drug states and linked each drug state transition to a 'responsible' prescribing physician. Outcomes of interest include transitions across drug states, particularly those resulting in combinations of increased risk (e.g., a benzodiazepine or sedative hypnotic with an opioid), and patients' time to discontinuation of overlapping prescriptions of an opioid, benzodiazepine, and sedative hypnotic.

Data Collection/Extraction Methods: An informal physician network (not restricted to a hospital or a health system) was constructed based on sharing patients between physicians reflected in face-to-face visits in Medicare claims. Transitions of risky drug states were related to characteristics of a prescriber's physician network position and compared between primary care physicians (PCPs) and specialists.

Principal Findings: Among beneficiaries receiving none of the three risky drug groups, patients seeing physicians with higher closeness centrality (shorter average path lengths to other physicians through the network) were less likely to transition to two or three risky drugs (OR of 2-drug = 0.923, 95% CI: [0.907, 0.939], p < 0.001; OR of 3-drug = 0.785, 95% CI: [0.657, 0.938], p = 0.008); and they were 4.4% more likely to discontinue overlapping prescriptions of an opioid, benzodiazepine, and sedative hypnotic (95% CI: [1%, 8.9%], p = 0.047). Compared to PCPs, psychiatrists appeared more likely to prescribe risky drug combinations, and their patients were less likely to discontinue overlapping three-drug prescriptions.

Conclusions: Characterizing physicians' prescribing behavior related to their position in shared-patient networks may reveal strategies for optimizing network-based interventions to improve prescribing quality.
 

Background: Patients taking immunosuppressive or immunomodulatory agents (IIAs) have been thought to be at greater risk for severe COVID-19-related health outcomes. These IIAs, such as conventional disease-modifying antirheumatic drugs (DMARDs), anti-tumor necrosis factor (TNF) biologics, non-anti-TNF biologics, and glucocorticoids, are often prescribed for disease management in patients with autoimmune rheumatic diseases, transplantations, and cancer.

Objective: to assess the risk for severe outcomes (hospitalization, intensive care unit (ICU) admission, and 60-day mortality) in patients prescribed IIAs.

Methods: A retrospective cohort analysis was conducted using administrative health data from British Columbia (BC), Canada. Cohort eligibility included all BC adults who tested positive on SARS-CoV-2 PCR tests from the provincial public health agency, between February 6, 2020 and August 15, 2021. Current IIA use was defined as use within the last 3 months. IIA exposure was divided into 8 medication classes: 1) antimalarials, 2) methotrexate, 3) leflunomide, 4) immunosuppressants (azathioprine, mycophenolate mofetil (MMF), cyclosporine, cyclophosphamide), 5) anti-TNF biologics (adalimumab, certolizumab, etanercept, golimumab, infliximab), 6) non-anti-TNF biologics (abatacept, anakinra, secukinumab, tocilizumab), 7) rituximab, and 8) glucocorticoids. Certain IIAs were assessed individually due to distinct mechanisms of action relative to their particular medication class. Hospitalization and ICU admission data were obtained from hospital discharge abstracts, and mortality within 60 days of a positive SARS-CoV-2 test from vital statistics. We used overlap weighted logistic regression models, with age, socioeconomic status, Romano modification of the Charlson comorbidity index, hypertension, rurality, and number of previous SARS-CoV-2 PCR tests as variables.

Results: Among 147,301 adults who tested COVID-19-positive, we included 307 patients prescribed antimalarials (mean age 57.4 years, 27.4% male), 373 prescribed methotrexate (mean age 55.2 years, 40.4% male), 60 prescribed leflunomide (mean age 60.3 years, 36.5% male), 409 prescribed immunosuppressants (mean age 54.3 years, 48.1% male), 282 prescribed anti-TNF biologics (mean age 45.0 years, 15.9% male), 110 prescribed non-anti-TNF biologics (mean age 50.3 years, 42.0% male), 43 prescribed rituximab (mean age 57.1 years, 33.6% male), and 1237 prescribed glucocorticoids (mean age 58.5 years, 49.2% male, median dose 250.0 mg), each with an equal number of comparators. Risk of hospitalization and ICU admission were increased in patients using immunosuppressants (any one of azathioprine, MMF, cyclosporine, or cyclophosphamide) (adjusted odds ratio (aOR): 2.08 and 2.88, respectively), MMF (aOR: 2.82 and 2.52, respectively), or glucocorticoids (aOR: 1.63 and 1.86, respectively), compared to non-users. Risk for ICU admission or 60-day mortality combined also increased for these groups (62% greater risk for immunosuppressants users, 93% greater risk for MMF users, and 69% greater risk for glucocorticoid users) compared to non-users. Only glucocorticoid users exhibited increased risk for 60-day mortality (aOR: 1.58) compared to non-users.

Conclusion: Public health data analysis demonstrated that, while patients exposed to immunosuppressants and glucocorticoids have significant increases in risk for severe COVID-19 outcomes, patients exposed to most other IIAs, including biologics, did not. These findings may inform patients, healthcare providers, and policymakers on the need and priority of personal infection-prevention measures, prescription alterations, or public health programs like booster vaccination campaigns. 

Background: Reports of firearm violence have become a daily fixture of the news cycle in the United States. Between 1981 and 2020, 1,357,504 Americans and 33,838 Arizonans were killed by firearms. Many more had non-fatal injuries. Firearm deaths now exceed motor vehicle deaths in both AZ and the US. About one-third of US adults report owning a firearm, while 42% reside in a household with firearms. In addition to the human toll, firearm violence carries a significant economic toll, both through medical costs and years of potential life lost, estimated at $557 billion nationally and $16 billion in AZ in 2019. Despite public support for reasonable and effective gun control laws, few preventative actions have been taken nationally or in most states to curb this uniquely American tragedy. AZ's very weak gun control laws have warranted a "failing" rating from multiple independent research organizations, clearly indicating the need for further legislation. Firearm violence in AZ takes many forms and varies depending on race, ethnicity, sex, and rurality, among other factors. This report was prepared for the Arizona Public Health Association to support policy and advocacy priorities.

Methods: This report was compiled using publicly accessible data sources including the CDC Multiple Cause of Death database (WONDER), the CDC Web-based Injury Statistics Query and Reporting System (WISQARS), and the CDC National Violent Death Reporting System. Several other non-governmental publicly available data sources were also utilized to characterize gun ownership, the strength of state gun safety laws, school shootings, police shootings, and other data. Annual trends were analyzed using Joinpoint software from the National Cancer Institute and a literature review was conducted to incorporate findings from analytic epidemiologic research and to identify policies in other states as possible models for AZ.

Results: Preliminary analysis revealed that AZ is performing worse than the national average on virtually all indicators of gun mortality. From 1999-2020, the age-adjusted firearm mortality rate per 100,000 among Arizonans was 42% higher than the national rate. Joinpoint trend analysis indicated that firearm deaths in AZ peaked in 1995, then declined until 2014 when rates began increasing by 3% per year. The leading cause of firearm death in AZ is suicide (65%), followed by homicide (31%). While firearm suicide rates have been slowly increasing (0.6% per year), firearm homicide rates declined sharply (7% per year) between 2005 and 2014, then increased by the same rate after 2014. Firearm deaths resulted in 28,049 YPLL before age 65 in 2020, over half of which were due to suicide. Stratification by several demographic indicators revealed statistically significant disparities in firearm mortality in AZ. From 1999-2020, the risk of firearm mortality among non-metro residents was 12% higher than the risk among metro residents. The firearm mortality rate among males was nearly 6-fold higher than among females. Analysis of racial and ethnic disparities showed that the risk of firearm mortality among non-Hispanic Black people was 35% higher than for non-Hispanic White people in AZ. Rates among Hispanics were 84% higher in AZ than in the US. Firearm mortality rate were two-fold higher in states with the weakest guns laws compared to states with the strongest gun laws.

Discussion: Initial data analysis highlights disparities that may be amenable to local legislative action. Discussed in light of the June 2022 Supreme Court ruling which stipulated a constitutional right to concealed carry and the passage of the federal Bipartisan Safer Communities Act, this report presents findings through the public health model of gun violence prevention. This framework is used to define the problem of gun violence in AZ, identify risk factors, and propose prevention strategies to promote a safer community for all Arizonans. 

Introduction
Understanding the factors contributing to mental wellbeing in youth is a public health priority. Self-reported enthusiasm for the future may be a useful indicator of wellbeing and has been shown to forecast social and educational success. Typically, cross-domain measures of ecological and health-related factors with relevance to public policy are analyzed either in isolation or in targeted models assessing bivariate interactions. Here, we capitalize on a large provincial dataset and cutting-edge interpretable machine learning to analyze dozens of measures simultaneously in a non-linear, interactive framework.

Objective
The aim of this study is to identify the sociodemographic, experiential, behavioral, and other health-related factors most strongly associated with levels of subjective enthusiasm for the future in a large sample of elementary and secondary school students using machine learning.

Methods
We analyzed data from 14,142 participants in the 2019 Ontario Student Drug Use and Health Survey (OSDUHS) (grades 7 to 12 inclusive, age range: 11 to 20 years old). Our primary outcome of interest was self-reported enthusiasm, measured by level of agreement with the statement "I am very enthusiastic about my future" (4-point Likert Scale). We used 50 variables derived from the OSDUHS survey to model this outcome, including demographics, perception of school experience (including aspects of school connectedness and academic performance), physical activity and sleep patterns, substance use, and physical and mental health indicators, among others. Models were built using a non-linear decision-tree based machine learning algorithm, called eXtreme Gradient Boosting (XGBoost), to classify students as indicating either high or low levels of enthusiasm. Cross-validated hyperparameter optimization and model training were performed on 80% of the sample, and unbiased model performance was evaluated on a 20% withheld test set. SHapley Additive exPlanations (SHAP) values were used to interpret the model, providing a ranking of feature importance and identifying interactions between model features.

Results
An 81% classification accuracy was achieved with our top performing model; precision and recall were 82% and 81%, respectively. The top three contributors to higher self-rated enthusiasm for the future from this model include: higher self-rated physical health (SHAP=0.62), feeling able to discuss problems or feelings with their parents (0.49), and having a sense of belonging in their school (0.32). In addition, subjective social status at school was a top feature and showed non-linear effects, with benefits to predicted enthusiasm only present in the mid-high range of values (from 0-10). These physical health and school participation variables also interacted within the model, where improvements in one area may mitigate deficits in the other.

Conclusion
Using machine leaning, we identified key factors related to self-reported enthusiasm for the future in a large sample of young students: perceived physical health, school social status and connectedness, and quality of family relationships. These factors were more important in determining enthusiasm than many common policy targets including social media, drug, and alcohol use. Together, our findings suggest that a focus on physical health and school connectedness should be central to impactful policy aimed at improving the wellbeing of youth, particularly when it comes to improving enthusiasm for the future. 

Despite the extensive use of network autocorrelation models in social network analysis, network autocorrelation models for binary dependent variables have received surprisingly scant attention. In this paper, we develop a binary network autocorrelation model using a Bayesian approach for model estimation. We develop multiple prior distributions for the focal peer effect parameter (ρ) designed to improve the performance of these estimators while introducing minimal information into the analysis. These priors include uniform priors with different ranges, Jeffreys priors, and a newly-developed uniform prior on a transformation of ρ. The performance of the proposed Bayesian approach and the sensitivity of results to the prior are assessed using a simulation study. Finally, we construct a New England region patient-sharing hospital network and apply our approaches to study the adoption of robotic surgery among hospitals using a cohort of Medicare beneficiaries in 2016 and 2017. 

Many medical decisions during the pandemic were made without the support of causal evidence obtained in clinical trials. We study the case of nebulized ibuprofen (NaIHS), a drug that was extensively used on COVID-19 patients in Argentina amidst wild claims about its effectiveness and without regulatory approval. Using data on 5,146 patients hospitalized in 11 health centers spread over 4 provinces (of which 19.8% received the treatment), we find a large, negative and statistically significant correlation between NaIHS treatment and mortality using inverse probability weighting estimators. We consider several threats to identification, including the selection of "low" risks into NaIHS, spillovers affecting patients in the control group, and differences in the quality of care in centers that use NaIHS. While the negative correlation appears robust, our results are best interpreted as emphasizing the benefits of running a randomized controlled trial and the challenges of incorporating information produced in other, less rigorous circumstances. 

It is well recognized that incorporating summary information from external sources may improve the efficiency of the analysis of a study of interest, whereas failure to recognize inconsistency or properly account for the discrepancy between the external sources and the internal study may introduce bias. Even in absence of these concerns, failure to properly account for uncertainty in the external information may also lead to invalid inference. Addressing all three potential sources of invalid inference, we propose a penalized combined likelihood approach that simultaneously selects and incorporate consistent external information in the internal study analysis, while properly accounting for uncertainty in the selected information. The proposed approach does not assume the homogeneity of the target population between the internal study and the external sources nor require a reference dataset to address the discrepancies. The proposed estimator is as efficient as an oracle estimator that incorporates only consist external information. For implementation, we develop an alternative optimization strategy in which each step is a convex optimization. Simulation studies show that our approach consistently selects the correct external information, and has a smaller mean square error than the maximum likelihood method that only analyzes the internal study. We illustrate our approach by an application of gestational weight gain study. 

Background:
Telehealth has long been touted for its promise to extend access to health care to those in rural and underserved communities. The exponential increase in telehealth utilization brought about by the pandemic has raised important policy considerations and are even more urgent as temporary policy provisions allowing telehealth during the public health emergency are soon to sunset. As we grapple with the future of telehealth and push for more equitable access to this modality, it is critical that the research community supply relevant, digestible data to facilitate informed decision and policy making. Applying various mapping techniques to claims data, telehealth administrative data, and social determinants of health data may prove helpful in speeding the data to telehealth policy pipeline.
The Medical University of South Carolina (MUSC)-one of two National Telehealth Centers of Excellence and the administrative headquarters for the state-funded South Carolina Telehealth Alliance (SCTA)-began using geographic heat mapping to inform its strategic planning and advocacy among policymakers. This year, three heat mapping methods were deployed to investigate a) gaps in access to outpatient specialty care in SC to inform targeted telehealth specialty services, b) hospital outmigration patterns to inform hospital-based telehealth consultative support to rural hospitals, and c) virtual urgent care (VUC) utilization among vulnerable populations to inform advocacy for payment by Medicaid.
Methods:
Outpatient Specialty Care: Using a 5% national Medicare data set from 2018-2019, E/M CPT codes were filtered by billing provider specialty and grouped by patient county. Specialty visit volumes per capita for each county were calculated, and counties were stratified to into quartiles for heat mapping based on county visit rates. Additionally, for the counties in the bottom 3 quartiles, the number of visits needed for each county to advance to the quartile above them was calculated and overlayed onto the county map. This helped visualize the scale of added telehealth specialty visits needed in each county to achieve more equitable access.
Hospital Outmigration: To visualize hospital outmigration, we analyzed the same 5% sample of Medicaid data, this time by inpatient specialty CPT codes and patient county to determine "demand" for inpatient specialty services per capita by county. This was then compared with HCUP hospital admission data during the same time period to determine county "supply" per capita. The difference between these numbers was calculated for each county, and this difference was used to place counties in quartiles to indicate levels of outmigration that might be mitigated by telehealth consultative support.
VUC Utilization: Finally, using MUSC's VUC database, MUSC calculated the rates of VUC encounters per capita by patient zip code. Zip codes were stratified into quartiles based on their rate and mapped. Utilization rate maps were then compared with zip code mapping of social vulnerability based on the CDC's social vulnerability index (SVI).
Results & Discussion: Heatmap results show low specialty care utilization and neuro admissions to local county hospitals among Medicare participants in rural counties in SC, particularly along the I-95 corridor which is a region known for low access to care and poorer health outcomes. Heatmaps of virtual urgent care visits per capita (that were free during COVID) within the large contiguous Charleston tri-county area in coastal SC show an inverse relationship to the areas with the lowest social vulnerability. This trend of low utilization among the most vulnerable population was also found in a New York study. Simple but powerful data driven methods are needed to ensure state policymakers understand and support efforts to reach the most vulnerable populations in order to realize the promise of increased access using telehealth particularly in rural and underserved states. 

Medical prescriptions/dispensing information in a pharmacy database can be used in conjunction with an electronic medical record/chart warehouse to predict medical outcomes—incident disease, recurrence and exacerbation of disease, death, etc. Drug studies with medical records may serve in assessing benefit on the target disease, in repurposing of the drug for another disease/outcome, in determining drug interactions—harmful or beneficial, or as time-varying covariates in other studies. It often happens in such studies that preparation and cleaning of the drug information for an individual is time consuming taking more than 50% of the time for the whole project. This proposal to extract, pre-process/clean drug courses from drug prescriptions/dispensing data will make drug prescription data usable as part of an analytics pipeline that facilitates future quality assurance, business analytical or research studies while meeting regulatory and governance issues.

Design of a drug course analytics platform and the rules of access and drug course definitions would be determined by the institution requirements and scientific/medical consultancy. The processing of additional drug courses or modification of existing drug courses can be requested by the user. These requests also contain definitions of cohort selection, of study dates, and choice of options for cleaning the drug courses, etc. Thus, the drug course analytics platform would maintain procedures for data extraction currently written in MS SQL and a number of utilities for data cleaning/pre-possessing currently written in SAS.
Such utilities and rules include
1. Clean the drug course information for a new drug with a rule maintaining courses of at least 30 days duration for chronic conditions/diseases.
2. Pool drugs by drug class; for example, pool pain medications by morphine equivalent doses.
3. Combine 2 (maybe 3) drugs into combinations; for example. 22 = 4 combinations for 2 drugs, or 23 = 8 combinations for 3 drugs which can be used to assess additive or interaction drug effects. The combinations can be computed as a refinement of partitions from the constituent drug courses while maintaining course durations of at least 30 days.
4. Define time-varying covariates; for example, body mass index (BMI) might be defined by courses of BMI that are constant up to a ±5% change.
5. Merge the main outcome of the time to event data with the combined drug courses. Intervening binary, medical events can be presented as only two courses.

The drug course analytical platforms with its pre-processing and rules provide data in a form that meet the requirements of business analytics, quality assurance or research for medical outcomes - many outcomes can be represented as time to event data or number of events data. The drug information can be represented as a sequence of drug courses which as mathematical constructs are both time-varying variables [time series] and partitions of the time interval to event. Cox's proportional hazards ratio, survival model is a powerful analysis tool for outcome data as it includes data from individuals who do not have the event under study during the time to the end of the study—called censored records/events. This semi-parametric model is available in SAS's PHREG procedure and R's survival package, both of which have provision for time varying variables such as drug courses. Other medical outcomes can be represented by Poisson number of events in a fixed time period: e.g. exacerbation of disease for two years following incident disease. The rate of such events occurring can be analyzed by Generalized Linear Modeling with log link and Poisson distribution. This model also allows time varying variables. Both statistical models require time-vary input as flat files such as precisely provided by the Drug Course Medical Analytics Platform proposed here-in.

Both models allow propensity scoring as weights which is an optimal method for reducing selection bias based on variables from the database.

The need for the proposed Platform is well illustrated by our recently published drug study about reducing risk of incident asthma [1]. The review process lead to adding a third drug requiring a refinement (merger) of three (3) drug time-courses as new input for recalculation of Propensity Scoring weights and for the survival analyses.

[1] . Sood A, Qualls C, Murata A, Kroth PJ, Mao J, Schade DS, Murata G. (2022) Potential for repurposing oral hypertension/diabetes drugs to decrease asthma risk in obesity. J Asthma. 2022 Jul
13:1-9. doi: 10.1080/02770903.2022.2097919. Online ahead of print. PMID: 35796615

 

Relevance to ICHPS: This abstract is relevant to contemporary health policy because it provides prevalence estimates and discussion points focused on a population with high healthcare utilization that requires special accommodations to meet their needs and to promote a positive experience with healthcare. The U.S. population analyses presented here lay the groundwork for analyses of healthcare quality, cost, and access in light of contextual policy and environmental factors (e.g., ACA, COVID-19 pandemic).

Intro: Having a disability is often associated with challenges in areas such as education, employment, housing, and healthcare. Specifically, in healthcare, disparities in access, cost, quality, and accessibility among those with disabilities are commonplace. Recent estimates of the prevalence of disability in the U.S. can inform private industry stakeholders, policymakers, and disability advocacy community about the relative frequency of each functional type of disability by important demographic characteristics (e.g., age, sex, race, insurance type). Such awareness will promote increased transparency of potential disparities in this population, guide healthcare stakeholder decision making, and inform advocacy efforts around access, quality, and cost of healthcare.

Methods: Two years of self-report responses (2017-2018) to disability questions from the U.S. Medical Expenditures Panel Survey (MEPS) were analyzed to estimate disability prevalence, overall and by functional type (e.g., Hearing, Seeing, Cognition), among the adult (18+) U.S. population in total and by demographic categories. Complex survey design was utilized to produce weighted prevalence estimates representing the U.S. non-institutionalized population. Crude and age-adjusted prevalence estimates were calculated.

Results: In 2018, approximately 18% (almost 1 in 5) of Americans reported have at least one functional disability, with the most common disabilities being mobility (~10%) and cognitive (~6%) difficulties. Rates of any disability by sex were similar (~19% among females & ~17% among males) and older Americans trended toward higher rates of disability with age (~20% among those 55-64 and >40% among those 65+). By race, White, Black, Multiracial, and American Indian Alaska Native (AIAN) people were relatively comparable in terms of any disability prevalence (ranging between ~17% to ~23%), but Asian Americans had a notably lower prevalence of any disability (~9%). The prevalence of any disability among those with public insurance highest (~37%) and lower among those with any private insurance (~12%) and those who were uninsured (~7%). Prevalence of disability by functional type and insurance type will be presented as well. In general, rates of disability between 2017 and 2018 were remarkably similar.

Discussion: Disability remains a source of healthcare inequities spanning cost, quality, access, and accessibility (either physical or virtual). Given relatively high rates of disability among adults in the U.S. (~1 in 5), healthcare and insurance providers should ensure facilities and services are accessible, disability-competent, and consistent with the health and ability status of the patient population. Disability prevalence may be lowest among the uninsured because of Medicaid's disability eligibility criteria and/or expanded insurance coverage under the ACA for those with disabilities. Future analyses may wish to examine disability prevalence by additional insurance categories, over time (e.g. pre & post ACA), and by state medicaid expansion/waiver status. Pay for performance reimbursement models are increasingly driven by outcomes as measured by quality metrics. Better, more equitable patient outcomes and metrics of quality can be simultaneously achieved if stakeholders collect, monitor, and address not only aggregate measures of quality, but quality among patients with the greatest challenges in accessing and affording healthcare. 

Covid-19 makes a drastic requirement of knowing the status of covid positive in hospitals, labs, clinics, health centers. A plethora of patients, deficit of lab technician, huge uses of the medical devices/accessories, experiment ingredients, serge time of life and death events made the hospital and health systems inconsistent. As such the aim of the current study is to explore a quick way to detect covid from easily obtained medical images. Several machine learning models like generalized linear, linear discrimination, principal components, factor analysis, neural network, convoluted neural network, augmented neural network, recurrent neural network, random forest models, etc have been found in detecting the status of covid. The accuracy rate of these models have also been found. 

Health policy researchers are often interested in the causal effect of a medical treatment in situations where randomization is not possible. Full matching on the propensity score (Gu & Rosenbaum, 1993) aims to emulate random assignment by placing observations with similar estimated propensity scores into sets with either 1 treated unit and one or more control units or 1 control unit and multiple treated units. Sets of the second type, with treatment units forced to share a comparison unit, can be unhelpful from the perspective of statistical efficiency. They are often necessary to achieve an experiment-like arrangement (as measured with observed covariates), but optimal full matching on estimated propensity scores is known to exaggerate the number of many-one matches that are truly necessary, generating lopsided matched sets and smaller effect sample sizes (Hansen, 2004).

In this presentation, we introduce an enhancement of the Hansen and Klopfer (2006) optimal full matching algorithm that counteracts this exaggeration by enabling analysis to permit treatment units to share a control while limiting the number that are permitted to do so. The result is a more well-balanced matching structure that prioritizes 1:1 pairs as opposed to matches with lopsided, many-to-one configurations of matched sets.

This enhanced optimal full matching is then illustrated in a pilot study on the effects of Extracorporeal Membrane Oxygenation (ECMO) for treatment of pediatric acute respiratory distress syndrome. Within this pilot study, in which data sample size was strictly limiting, existing methods for limiting the sharing of controls have already resulted in an increased effective sample size. The present enhancement of Hansen and Klopfer's optimal full matching algorithm provides an additional boost. Results indicate an increase in effective sample size at the small expense of covariate balance. Our enhancement of Hansen and Klopfer matching algorithm provides researchers with a new tool on how to manage this bias-variance tradeoff. 

Social networks of healthcare utilization and processes are often constructed using confidential Medicare administrative data. This confidentiality hinders researchers' ability to study and distribute data that can aid reproducible research. Network analysis provides a powerful approach to investigating the complex structure of relationships in various contexts, such as healthcare delivery and biology. In conjunction with graph topological structures, patterns of node properties or attributes across the network provide insights into the key (vertex, edge, and sub-network) features underlying the network. For example, in health services research, node properties, such as physician's specialty, may help explain the community structure in the network. In this work, we propose a graph simulation model which simulates networks using degree and property augmentation (GRANDPA) and include a flexible R package for users to generate healthcare networks that emulate the original but can be freely distributed and analyzed.

Our GRANDPA model is built upon the Property Graph Model (PGM) introduced by Sathanur et. al.[1] PGM simulates graphs by computing the probability distribution of all possible combinations of attribute categories, i.e. joint label assignment distribution, for vertices and the joint distribution over pairs of label categories for modeling the edge connectivity, with the option of adding a degree label by dividing the degree distribution into bins. Along with the degree distribution, community structure has previously been shown to be useful in recovering the network structure.[2] GRANDPA extends the PGM framework by allowing the additional augmentation of a community detection label as well as other topologically defined vertex characteristics.

To demonstrate GRANDPA's potential, we conducted two case studies. The first used Zachary's karate club network[3] included in the igraph package in R,[4] and we created two vertex labels associated with network sub-structures. In the second case study, we generated a unipartite patient-sharing network using the 2019 Medicare claims data. We used physician's primary medical specialty as a node attribute label. For both case studies, we augmented the generated graphs with a degree distribution label and a community membership label.[5] In addition, we added an attribute label by computing the linchpin score of each physician to identify their importance as a "one-of-a-kind" physician to their neighbors based on their specialty.[6]

To assess network recovery, we compared the original graph with generated graphs. In these case studies, both degree and community augmentation lead to generated networks that best resembled original networks. In the karate network, the normalized root mean square error (NRMSE) between the complementary cumulative distribution function (CCDF) of the original and the best-returned graph's degree-distribution is 0.0508. Visually, the original graph and the final generated graph have nearly identical topological structures. For the unipartite healthcare network, the NRMSE between the CCDF of the original and the best generate graphs degree-distribution is 0.0514. Likewise, the final graph has the same number of communities and similar internal community structures compared to the original one.

Reference:

[1] A. V. Sathanur, S. Choudhury, C. Joslyn, and S. Purohit, "When Labels Fall Short: Property Graph Simulation via Blending of Network Structure and Vertex Attributes," Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, pp. 2287–2290, Nov. 2017, doi: 10.1145/3132847.3133065.
[2] B. Karrer and M. E. J. Newman, "Stochastic blockmodels and community structure in networks," Phys. Rev. E, vol. 83, no. 1, p. 016107, Jan. 2011, doi: 10.1103/PhysRevE.83.016107.
[3] W. W. Zachary, "An Information Flow Model for Conflict and Fission in Small Groups," Journal of Anthropological Research, vol. 33, no. 4, pp. 452–473, Dec. 1977, doi: 10.1086/jar.33.4.3629752.
[4] G. Csardi and T. Nepusz, "The igraph software package for complex network research," p. 10.
[5] F. Radicchi, C. Castellano, F. Cecconi, V. Loreto, and D. Parisi, "Defining and identifying communities in networks," Proc. Natl. Acad. Sci. U.S.A., vol. 101, no. 9, pp. 2658–2663, Mar. 2004, doi: 10.1073/pnas.0400054101.
[6] M. D. Nemesure, T. M. Schwedhelm, S. Sacerdote, A. J. O'Malley, L. R. Rozema, and E. L. Moen, "A measure of local uniqueness to identify linchpins in a social network with node attributes," Appl Netw Sci, vol. 6, no. 1, p. 56, Dec. 2021, doi: 10.1007/s41109-021-00400-8.