Machine learning-based prediction of low-value care for hospitalized patients

Objective

Low-value care (i.e., costly health care treatments that provide little or no benefit) is an ongoing problem in United States hospitals. Traditional strategies for reducing low-value care are only moderately successful. Informed by behavioral science principles, we sought to use machine learning to inform a targeted prompting system that suggests preferred alternative treatments at the point of care but before clinicians have made a decision.

Methods

We used intravenous administration of albumin for fluid resuscitation in intensive care unit (ICU) patients as an exemplar of low-value care practice, identified using the electronic health record of a multi-hospital health system. We divided all ICU episodes into 4-h periods and defined a set of relevant clinical features at the period level. We then developed two machine learning models: a single-stage model that directly predicts if a patient will receive albumin in the next period; and a two-stage model that first predicts if any resuscitation fluid will be administered and then predicts albumin only among the patients with a high probability of fluid use.

Results

We examined 87,489 ICU episodes divided into approximately 1.5 million 4-h periods. The area under the receiver operating characteristic curve was 0.86 for both prediction models. The positive predictive value was 0.21 (95% confidence interval: 0.20, 0.23) for the single-stage model and 0.22 (0.20, 0.23) for the two-stage model. Applying either model in a targeted prompting system could prevent 10% of albumin administrations, with an attending physician receiving one prompt every 4.2 days of ICU service.

Conclusion

Prediction of low-value care is feasible and could enable a point-of-care, targeted prompting system that offers suggestions ahead of the moment of need before clinicians have already decided. A two-stage approach does not improve performance but does interject new levers for the calibration of such a system.

A voice-based digital assistant for intelligent prompting of evidence-based practices during ICU rounds

OBJECTIVE

To evaluate the technical feasibility and potential value of a digital assistant that prompts intensive care unit (ICU) rounding teams to use evidence-based practices based on analysis of their real-time discussions.

METHODS

We evaluated a novel voice-based digital assistant which audio records and processes the ICU care team's rounding discussions to determine which evidence-based practices are applicable to the patient but have yet to be addressed by the team. The system would then prompt the team to consider indicated but not yet delivered practices, thereby reducing cognitive burden compared to traditional rigid rounding checklists. In a retrospective analysis, we applied automatic transcription, natural language processing, and a rule-based expert system to generate personalized prompts for each patient in 106 audio-recorded ICU rounding discussions. To assess technical feasibility, we compared the system's prompts to those created by experienced critical care nurses who directly observed rounds. To assess potential value, we also compared the system's prompts to a hypothetical paper checklist containing all evidence-based practices.

RESULTS

The positive predictive value, negative predictive value, true positive rate, and true negative rate of the system's prompts were 0.45 ± 0.06, 0.83 ± 0.04, 0.68 ± 0.07, and 0.66 ± 0.04, respectively. If implemented in lieu of a paper checklist, the system would generate 56% fewer prompts per patient, with 50%±17% greater precision.

CONCLUSION

A voice-based digital assistant can reduce prompts per patient compared to traditional approaches for improving evidence uptake on ICU rounds. Additional work is needed to evaluate field performance and team acceptance.

Associations of past-year stigma and psychosocial syndemic conditions: Considerations for intersectional stigma measures among Black Sexual Minority Men

This secondary analysis of a mixed serostatus sample of Black sexual minority men (BSMM) used conditional inference tree methods to explore associations of past-year experienced stigma and psychosocial syndemic conditions. Experienced stigmas were attributed to race, sexuality, socioeconomic status, HIV status or some "other" reason. Psychosocial syndemic conditions studied included physical assault, intimate partner violence, polysubstance use, and depression symptomology. Data are from Promoting Our Worth, Equality and Resilience (POWER), a serial, cross-sectional study conducted between 2014-2017 (N=4430). Experiences of multiple stigmas were reported by n=938 (22.1%) of BSMM. Conditional inference tree results revealed that HIV-related stigma and its intersection with "other" stigma showed the greatest variance in psychosocial condition prevalence. Our findings suggest that when developing intercategorical intersectional analyses with BSMM, there are important stigmas for BSMM beyond those attributed to race, sexuality, and SES, particularly intersecting with HIV-related stigma. Conditional inference tree analysis shows promise in quantitative explorations of intersectional stigma with BSMM, but will benefit from the inclusion of additional forms of stigma, which should be considered by the field moving forward.

Intensivist physician-to-patient ratios and mortality in the intensive care unit

PURPOSE

A high daily census may hinder the ability of physicians to deliver quality care in the intensive care unit (ICU). We sought to determine the relationship between intensivist-to-patient ratios and mortality among ICU patients.

METHODS

We performed a retrospective cohort study of intensivist-to-patient ratios in 29 ICUs in 10 hospitals in the United States from 2018 to 2020. We used meta-data from progress notes in the electronic health record to determine an intensivist-specific caseload for each ICU day. We then fit a multivariable proportional hazards model with time-varying covariates to estimate the relationship between the daily intensivist-to-patient ratio and ICU mortality at 28 days.

RESULTS

The final analysis included 51,656 patients, 210,698 patient days, and 248 intensivist physicians. The average caseload per day was 11.8 (standard deviation: 5.7). There was no association between the intensivist-to-patient ratio and mortality (hazard ratio for each additional patient: 0.987, 95% confidence interval: 0.968-1.007, p = 0.2). This relationship persisted when we defined the ratio as caseload over the sample-wide average (hazard ratio: 0.907, 95% confidence interval: 0.763-1.077, p = 0.26) and cumulative days with a caseload over the sample-wide average (hazard ratio: 0.991, 95% confidence interval: 0.966-1.018, p = 0.52). The relationship was not modified by the presence of physicians-in-training, nurse practitioners, and physician assistants (p value for interaction term: 0.14).

CONCLUSIONS

Mortality for ICU patients appears resistant to high intensivist caseloads. These results may not generalize to ICUs organized differently than those in this sample, such as ICUs outside the United States.

Characterizing intensive care unit rounding teams using meta-data from the electronic health record

PURPOSE

Teamwork is an important determinant of outcomes in the intensive care unit (ICU), yet the nature of individual ICU teams remains poorly understood. We examined whether meta-data in the form of digital signatures in the electronic health record (EHR) could be used to identify and characterize ICU teams.

METHODS

We analyzed EHR data from 27 ICUs over one year. We linked intensivist physicians, nurses, and respiratory therapists to individual patients based on selected EHR meta-data. We then characterized ICU teams by their members' overall past experience and shared past experience; and used network analysis to characterize ICUs by their network's density and centralization.

RESULTS

We identified 2327 unique providers and 30,892 unique care teams. Teams varied based on their average team member experience (median and total range: 262.2 shifts, 9.0-706.3) and average shared experience (median and total range: 13.2 shared shifts, 1.0-99.3). ICUs varied based on their network's density (median and total range: 0.12, 0.07-0.23), degree centralization (0.50, 0.35-0.65) and closeness centralization (0.45, 0.11-0.60). In a regression analysis, this variation was only partially explained by readily observable ICU characteristics.

CONCLUSIONS

EHR meta-data can assist in the characterization of ICU teams, potentially providing novel insight into strategies to measure and improve team function in critical care.

Intersectional Experienced Stigma and Psychosocial Syndemic Conditions in a Sample of Black Men Who Have Sex with Men Engaged in Sex Work (BMSM-SW) from Six US Cities

Black men who have sex with men (BMSM) in the United States experience a disproportionate burden of violence, substance use, physical and mental health conditions relative to other racial groups. BMSM who engage in sex work (BMSM-SW) experience a high burden of psychosocial conditions, sexually transmitted infections, including HIV, and intersectional stigma. This analysis characterizes remuneration and client typologies for BMSM-SW, documents intersectional stigma experienced by BMSM-SW relative to other BMSM, and explores the impact of experienced intersectional stigma on the relationship between sex work engagement and psychosocial syndemic conditions (violence, polydrug use, and depression symptoms). Results show that a majority of BMSM-SW in the sample had female clients and that sex workers were more likely than other BMSM to hire another sex worker. BMSM-SW were more likely than other BMSM to report stigma attributed to race; sexuality; HIV status; socioeconomic status; and "other" attributes, and were more likely to report experiencing stigma across all settings assessed (schools; healthcare; employment; housing; police/courts; and in public/community). Intersectional stigma mediated the relationship between sex work engagement and psychosocial syndemic conditions, accounting for 49% (95% CI: 47.6-50.0%) of the relationship. Interventions for BMSM-SW should include resilience-building components to counteract the effects of intersectional stigma.

Characterizing Biomedical HIV Prevention Awareness and Use Among Black Transgender Women in the United States

Black transgender women (BTW) in the United States experience disproportionate rates of HIV despite biomedical prevention interventions such as pre-exposure prophylaxis (PrEP) and non-occupational post-exposure prophylaxis (nPEP). Using a sample of 490 BTW collected from 2014 to 2017, bivariate, multivariable, and multinomial analyses were conducted to determine factors associated with awareness and use of PrEP and nPEP. BTW living with HIV were more aware of PrEP than HIV-negative BTW. Structural, demographic, and trans-specific factors (e.g., experiences of homelessness, violence, and current hormone use) related to HIV risk were associated with PrEP and nPEP awareness. PrEP use was associated with behavioral HIV risks (e.g., STI diagnosis, having an HIV-positive partner, and needle-sharing) and may demonstrate risk recognition among BTW. Knowing someone using PrEP was significantly positively associated with PrEP use. Development of guidelines for PrEP and nPEP use for BTW should leverage the strengths of guidelines for other populations, while also acknowledging the unique risks for this population.

Syndemic Health Disparities and Sexually Transmitted Infection Burden Among Black Men Who Have Sex with Men Engaged in Sex Work in the U.S

Black men who have sex with men (MSM) engaged in sex work (BMSM-SW) experience elevated HIV and sexually transmitted infection (STI) prevalence. Further, BMSM-SW have been shown to have higher rates of syndemic psychosocial health conditions which contribute to HIV risk behavior and incidence, and poorer care outcomes than other groups of men who have sex with men. However, syndemic perspectives have not been applied to understanding past-year STI burden among BMSM-SW in the U.S. Sexually active Black MSM ≥ 18 years old were recruited from Black Pride events in six U.S. cities (n = 4421) between 2014 and 2017. Multivariable logistic regressions assessed correlates of past-year sex work engagement; whether BMSM-SW had higher odds of syndemic conditions; and whether BMSM-SW had higher odds of self-reported, past-year STI diagnoses. Structural equation models assessed relationships between sex work engagement, syndemic conditions, and STI controlled for sociodemographics and number of sexual partners. A total of 254 (5.7%) Black MSM reported past-year sex work, of whom 45.3% were HIV positive. BMSM-SW were significantly more likely to be Hispanic, to report past-year bisexual behavior, and to report annual income < $10,000. In multivariable models, BMSM-SW were significantly more likely to report intimate partner violence, assault victimization, polydrug use, and depression symptoms; they were also more likely to report past-year gonorrhea, chlamydia, and syphilis. Syndemic conditions mediated the relationship between past-year sex work and past-year STI burden, constituting a significant indirect effect. BMSM-SW in the U.S. face severe biopsychosocial health disparities. Interventions developed for BMSM engaged in sex work are lacking. Our results suggest that interventions containing safer sex work education and sex-positive biobehavioral HIV/STI prevention alongside substance use, mental health, employment, and education components will be most effective.

Barriers and facilitators to past six-month HIV testing among men who have sex with men in Belize

Prevalence of HIV in Belize is high, and men who have sex with men (MSM) are disproportionately impacted by HIV. HIV testing is critical in curbing the epidemic; however, little is known about factors associated with testing among MSM in Belize. Working with a non-governmental organization in a large, urban city within Belize, snowball sampling was applied to recruit Belizean MSM to complete a self-administered survey. Multivariable logistic regression analysis was employed to understand associations with HIV screening behavior. Access to healthcare, HIV knowledge, and reporting having heard of Section 53 of the Criminal Code of Belize (once outlawing same-sex sexual behavior), but not experiencing any negative impact from Section 53 were significantly positively associated with having received an HIV test in the past six months. Healthcare maltreatment (lifetime), depression symptomology, and shame were significantly negatively associated with having received a HIV test in the past six months. Findings suggest that multiple factors associated with stigma and discrimination negatively affect testing strategies among MSM in Belize.

Black Men Who Have Sex with Men and Lifetime HIV Testing: Characterizing the Reasons and Consequences of Having Never Tested for HIV

HIV testing remains a critical point of entry to HIV treatment services and now biomedical prevention as well. Yet despite the high HIV prevalence among Black men who have sex with men (MSM), insufficient attention has been given to factors associated with those Black MSM in the United States who have never received an HIV test in their lifetime. Promoting Our Worth, Equality, & Resilience (POWER) is a cross-sectional observational study that recruited Black MSM at Black Pride events across six cities in the United States from 2014 to 2017. Participants completed an anonymous questionnaire and were offered free, confidential HIV testing. Of the 4174 Black MSM without a prior HIV diagnosis, 404 (9.68%) had never tested for HIV (mean age = 31.03 years). Lower education and greater internalized homophobia were associated with never having tested for HIV. Higher age (AOR = 1.05, 95%, 1.02-1.07) and assumption of HIV-positivity (AOR = 3.24, 95% CI 1.53-6.84) were both associated with increased odds of an HIV-positive test result (n = 119; 36%). To compare, HIV prevalence among Black MSM who had received at least one HIV test before study participation was 23%. While a minority of Black MSM had never received an HIV test, this group had a significantly higher likelihood of HIV infection. Alternative HIV testing strategies are needed to facilitate HIV testing initiation among Black MSM for whom conventional HIV testing modalities are insufficient.

Examining the Impact of a Psychosocial Syndemic on Past Six-Month HIV Screening Behavior of Black Men who have Sex with Men in the United States: Results from the POWER Study

Syndemic production theory has been used to explore HIV transmission risk or infections but has not been used to investigate prevention behavior, or with large samples of non-Whites. This analysis is the first to explore the impact of syndemic factors on previous six-month HIV screening behavior among US Black MSM. Data from Promoting Our Worth, Equality and Resilience (POWER) were analyzed from 3294 participants using syndemic variable counts and measures of interaction/synergy. Syndemic variables included: past three-month poly-drug use, depression, last year intimate partner violence, HIV risk and problematic binge drinking. BMSM reporting two syndemic factors were more likely to report screening (AOR = 1.37, 95% CI 1.04-1.80; p = 0.028) with no significant associations for three or more conditions. Measures of joint effect revealed that there were synergies among depression, problematic binge drinking and poly-drug use but these psychosocial factors cannot entirely explain testing patterns and excess disease burden among BMSM.

HIV Testing and ART Adherence Among Unstably Housed Black Men Who Have Sex with Men in the United States

Black men who have sex with men (BMSM) have the highest HIV incidence rate among all MSM in the United States (US), and are also disproportionately affected by homelessness and housing instability. However, little is known about the effects of homelessness on the HIV testing and care continuum for BMSM. Between 2014 and 2017, the Promoting Our Worth, Equality, and Resilience (POWER) study collected data and offered HIV testing to 4184 BMSM at Black Pride events in six US cities. Bivariate analyses were used to assess differences in sociodemographics and healthcare access between BMSM who self-reported homelessness and those who did not. Multivariable logistic regression models were used to assess differences in HIV testing by homelessness status. Finally, bivariate and multivariable models were used to assess differences in HIV care continuum and treatment adherence outcomes by homelessness status. 615 (12.1%) BMSM in our sample experienced homelessness in the last 12 months. BMSM who self-reported homelessness had higher odds of receiving an HIV test in the past 6 months compared to their stably housed counterparts. BMSM who self-reported homelessness had higher odds of reporting difficulty taking ART and of missing a dose in the past week compared to stably housed BMSM. Findings suggest that HIV testing outreach and treatment-related services targeting unstably housed BMSM may be effective. Future community-based research is needed to investigate how homelessness and housing instability affect ART adherence, and how this population may experience success in HIV testing and adherence despite economic and social marginalization.

Intimate Partner Violence and Depression among Black Transgender Women in the USA: The Potential Suppressive Effect of Perceived Social Support

Depression among Black transgender women (BTW) in the USA is an often understudied mental health concern with far-reaching consequences for overall physical and mental health at both the individual and community levels. Intimate partner violence (IPV) among BTW is also a frequently understudied and important social determinant of health in need of further exploration. This study sought to address the gap in research on the relationship between IPV and depression among BTW using a time- and location-based community sample of BTW from six US cities. In addition, it sought to explore the potential protective or suppressive effect of perceived social support on this relationship. Generalized structural equation models were used to assess conditional direct and indirect effects of IPV on depression via the suppression effect of perceived social support. Evidence was found of a statistically significant conditional direct effect of IPV on depression as well as a statistically significant suppression effect for perceived social support. Specifically, there was a 20% lower likelihood of increased depressive symptomatology for every 1-unit increase in perceived social support reported by participants. These findings indicated that perceived social support may be an important intervention point for helping to improve the mental health and well-being of BTW.

Prevalence and Correlates of PrEP Awareness and Use Among Black Men Who Have Sex with Men and Women (MSMW) in the United States

Men who have sex with men and women (MSMW), including those who are Black, experience HIV-related disparities compared to men who have sex with men only (MSMO). Few studies have assessed the prevalence and correlates of pre-exposure prophylaxis (PrEP) awareness and use among Black MSMW. We recruited MSM ≥ 18 attending Black Gay Pride events between 2014-2017. We conducted multivariable logistic regressions to assess differences in PrEP awareness and use among HIV-negative Black MSM (n = 2398) and within Black MSMW (n = 419). MSMW were less likely than MSMO to report PrEP awareness (p < 0.001). Among PrEP-aware MSM, MSMW were more likely than MSMO to report PrEP use (p < 0.05). MSMW receiving gay community support were more likely to be PrEP-aware (p < 0.01). MSMW reporting any past-year STI diagnoses were more likely to report PrEP use (p < 0.01). Findings suggest that PrEP awareness campaigns tailored for Black MSMW, concomitant with STI-to-PrEP interventions, will facilitate greater PrEP uptake in this population.

Understanding the impact of a syndemic on the use of pre-exposure prophylaxis in a community-based sample of behaviorally PrEP-eligible BMSM in the United States

HIV Pre-Exposure Prophylaxis (PrEP) has shown great promise in reducing HIV transmission among affected populations; however, PrEP uptake among Black men who have sex with men (BMSM) has stalled. This study compares BMSM using PrEP and BMSM at risk for HIV not using PrEP based on differences in behavior, psychosocial conditions and the presence of a syndemic (n = 1,411). BMSM reporting PrEP use were significantly more likely to report three of five HIV risk behaviors and three of four psychosocial conditions. Odds of reporting PrEP use increased as the number of psychosocial conditions increased such that BMSM with three psychosocial conditions (AOR = 5.65, 95% CI: 3.17, 10.08) and four conditions (AOR = 18.34, 95% CI: 5.01, 67.20) demonstrated significantly greater odds of PrEP use compared to BMSM reporting one or less conditions. While BMSM at greatest risk are using PrEP, strategies are still needed for men at varying risk levels.

Characterizing the HIV care continuum among a community sample of black men who have sex with men in the United States

Black men who have sex with men (BMSM) have the highest HIV incidence rates among all men who have sex with men (MSM) and have been less likely than other MSM to be diagnosed, linked or retained in care for HIV. The Promoting Our Worth, Equality, and Resilience (POWER) study administered a behavioral health survey and HIV test to BMSM to estimate the HIV continuum of care among a community-derived sample. Of the N = 1,680 BMSM living with HIV, n = 956 (56.9%) were aware of their HIV-positive status. BMSM who had been previously diagnosed reported progression through the care continuum at greater than 80% for linkage to medical care (97.6%), retention in medical care (97.3%), prescription of antiretroviral therapy (93.6%) and viral suppression (86.4%). Results of this analysis reveal the urgent need to ensure BMSM are being screened and diagnosed. Additionally, community-driven samples are a promising method offering the opportunity to build on the resilience of BMSM progressing though the continuum as well as engage BMSM not yet in care. Lastly, such methods can assist in communicating with those lost to retention.

Elevated HIV Prevalence and Correlates of PrEP Use Among a Community Sample of Black Men Who Have Sex With Men

BACKGROUND

The HIV epidemic among black men who have sex with men (BMSM) demands urgent public health attention. Pre-exposure prophylaxis (PrEP) is a highly efficacious option for preventing HIV, but characteristics of PrEP use among community samples of BMSM are not well-understood.

METHODS

A serial cross-sectional survey assessment (N = 4184 BMSM reporting HIV-negative/unsure status) and HIV testing were conducted at Black Gay Pride events in 6 US cities in 2014, 2015, 2016, and 2017.

RESULTS

HIV prevalence was higher among BMSM self-reporting current PrEP use (1 of 3 participants) than BMSM not self-reporting current PrEP use (1 of 5 participants) [32.3%, N = 103/319 vs. 20.0%, N = 639/3,193, adjusted odds ratio (aOR) = 1.68, 95% confidence interval (CI): 1.31 to 2.15]. BMSM reporting current PrEP use (N = 380) were more likely to report having a greater number of male sex partners (aOR = 1.02, 95% CI: 1.01 to 1.03), a sexually transmitted infection diagnosis (aOR = 2.44, 95% CI: 1.88 to 3.16), and stimulant drug use (aOR = 2.05, 95% CI, 1.21 to 3.47) when compared with BMSM not reporting current PrEP use (N = 3804). PrEP use increased from 4.7% (2014) to 15.5% (2017) (aOR = 1.19, 95% CI: 1.13 to 1.25). Among PrEP users, inability to afford health care coverage was associated with testing HIV-positive (aOR = 2.10, 95% CI: 1.24 to 3.56).

CONCLUSIONS

The high prevalence of HIV infection among BMSM reporting PrEP use is concerning. It does not, however, challenge the efficacy of PrEP itself but rather the uptake of the surrounding preventative package including behavioral risk reduction support, sexually transmitted infection treatment, and medication adherence counseling. Further research to understand barriers to fully effective PrEP is needed to guide operational and behavioral interventions that close the gap on incident infection.

Differences in alcohol use and alcohol-related problems between transgender- and nontransgender-identified young adults

BACKGROUND

Little is known about differences in alcohol use and alcohol-related problems between transgender- and nontransgender-identified populations. Using data from a large-scale health survey, we compare the drinking patterns and prevalence of alcohol-related problems of transgender-identified individuals to nontransgender-identified males and females. For transgender-identified people, we examine how various forms of victimization relate to heavy episodic drinking (HED).

METHODS

Cross-sectional surveys were completed by 75,192 students aged 18-29 years attending 120 post-secondary educational institutions in the United States from 2011 to 2013. Self-reported measures included alcohol use, alcohol-related problems, victimization, and sociodemographics, including 3 gender-identity groups: transgender-identified individuals; nontransgender-identified males; and nontransgender-identified females.

RESULTS

Compared to transgender-identified individuals, nontransgender-identified males were more likely to report HED in the past 2 weeks (relative risk=1.42; p=0.006); however, nontransgender-identified males and females reported HED on fewer days than transgender-identified people (incidence-rate ratios [IRRs] ranged from 0.28 to 0.43; p-values<0.001). Compared to transgender-identified people, nontransgender-identified males and females had lower odds of past-year alcohol-related sexual assault and suicidal ideation (odds ratios ranged from 0.24 to 0.45; p-values<0.05). Among transgender-identified people, individuals who were sexually assaulted (IRR=3.21, p=0.011) or verbally threatened (IRR=2.42, p=0.021) in the past year had greater HED days than those who did not experience those forms of victimization.

CONCLUSIONS

Compared to transgender-identified people, nontransgender-identified males and females: have fewer HED occasions (despite nontransgender-identified males having greater prevalence of HED); and are at lower risk for alcohol-related sexual assaults and suicidal ideation. Experiences of sexual assault and verbal threats are associated with greater HED occasions for transgender-identified people.