ABSTRACT
The capture-recapture method is a common tool used in epidemiology to estimate the size of "hidden" populations and correct the underascertainment of cases, based on incomplete and overlapping lists of the target population. Log-linear models are often used to estimate the population size yet may produce implausible and unreliable estimates due to model misspecification and small cell sizes. A novel targeted minimum loss-based estimation (TMLE) model developed for capture-recapture makes several notable improvements to conventional modeling: "targeting" the parameter of interest, flexibly fitting the data to alternative functional forms, and limiting bias from small cell sizes. Using simulations and empirical data from the San Francisco, California, Department of Public Health's human immunodeficiency virus (HIV) surveillance registry, we evaluated the performance of the TMLE model and compared results with those of other common models. Based on 2,584 people observed on 3 lists reportable to the surveillance registry, the TMLE model estimated the number of San Francisco residents living with HIV as of December 31, 2019, to be 13,523 (95% confidence interval: 12,222, 14,824). This estimate, compared with a "ground truth" of 12,507, was the most accurate and precise of all models examined. The TMLE model is a significant advancement in capture-recapture studies, leveraging modern statistical methods to improve estimation of the sizes of hidden populations.
Subject(s)
HIV Infections , HIV , Humans , San Francisco/epidemiology , Linear Models , Bias , HIV Infections/epidemiologyABSTRACT
Decarceration policies, enacted for SARS-CoV-2 mitigation in carceral settings, potentially exacerbated barriers to care for people living with HIV (PWH) with criminal legal involvement (CLI) during Shelter-in-Place (SIP) by limiting opportunities for engagement in provisions of HIV and behavioral health care. We compared health care engagement for PWH with CLI in San Francisco, California before and after decarceration and SIP using interrupted time series analyses. Administrative data identified PWH booked at the San Francisco County Jail with at least one clinic encounter from 01/01/2018-03/31/2020 within the municipal health care network. Monthly proportions of HIV, substance use, psychiatric and acute care encounters before (05/01/2019-02/29/2020) and after (03/01/2020-12/31/2020) SIP and decarceration were compared using Generalized Estimating Equation (GEE) log-binomial and logistic regression models, clustering on the patient-level. Of 436 patients, mean age was 43 years (standard-deviation 11); 88% cisgender-male; 39% white, 66% homeless; 67% had trimorbidity by Elixhauser score (medical comorbidity, psychotic disorder or depression, and substance use disorder). Clinical encounters immediately dropped following SIP for HIV (aOR = 0.77; 95% CI: 0.67, 0.90) and substance use visits (aRR = 0.83; 95% CI: 0.70, 0.99) and declined in subsequent months. Differential reductions in clinical encounters were seen among Black/African Americans (aRR = 0.93; 95% CI: 0.88, 0.99) and people experiencing homelessness (aRR = 0.92; 95% CI: 0.87, 0.98). Significant reductions in care were observed for PWH with CLI during the COVID-19 pandemic, particularly among Black/African Americans and people experiencing homelessness. Strategies to End the HIV Epidemic must improve engagement across diverse care settings to improve outcomes for this key population.
Subject(s)
Criminals , HIV Infections , Substance-Related Disorders , Humans , Male , Adult , San Francisco/epidemiology , Emergency Shelter , HIV Infections/epidemiology , HIV Infections/therapy , Pandemics , Delivery of Health Care , Substance-Related Disorders/epidemiology , Substance-Related Disorders/therapyABSTRACT
Respondent-driven sampling (RDS) is a potential strategy for addressing challenges in accurate estimation of abortion incidence, but it relies on often untested assumptions. We conducted an RDS study to estimate the cumulative lifetime incidence of abortion in Soweto, Johannesburg, South Africa (April-December 2018), to evaluate whether RDS assumptions were met, and to compare RDS estimates of incidence with estimates adjusted for employment and age based on census data. A total of 849 participants were recruited from 11 seed participants between April and December 2018. The assumption that individuals can identify target population members and the assumption of approximation of sampling with replacement was met. There were minor violations of the assumptions of seed independence from the final sample and reciprocity of ties. Assumptions of accurate degree reporting and random recruitment were not met. Failure to meet assumptions yielded a study sample with different employment characteristics than those of the target population; this could not be resolved by standard RDS methods. The RDS estimate of cumulative lifetime abortion incidence was 12.1% (95% confidence interval: 9.8, 14.3), and the employment-adjusted estimate was 16.9% (95% confidence interval: 12.8, 22.1). We caution researchers in using RDS for representative estimates of abortion incidence. Use of postsurvey weights to adjust for differences in characteristics between the sample and the target population may yield more representative results.
Subject(s)
Abortion, Induced , Female , Pregnancy , Humans , Incidence , South Africa/epidemiology , Employment , Surveys and Questionnaires , Sampling StudiesABSTRACT
Arterial blood oxygen saturation as measured by pulse oximetry (peripheral oxygen saturation (SpO2)) may be differentially less accurate for people with darker skin pigmentation, which could potentially affect the course of coronavirus disease 2019 (COVID-19) treatment. We analyzed pulse oximeter accuracy and its association with COVID-19 treatment outcomes using electronic health record data from Sutter Health, a large, mixed-payer, integrated health-care delivery system in Northern California. We analyzed 2 cohorts: 1) 43,753 non-Hispanic White (NHW) or non-Hispanic Black/African-American (NHB) adults with concurrent arterial blood gas oxygen saturation/SpO2 measurements taken between January 2020 and February 2021; and 2) 8,735 adults who went to a hospital emergency department with COVID-19 between July 2020 and February 2021. Pulse oximetry systematically overestimated blood oxygenation by 1% more in NHB individuals than in NHW individuals. For people with COVID-19, this was associated with lower admission probability (-3.1 percentage points), dexamethasone treatment (-3.1 percentage points), and supplemental oxygen treatment (-4.5 percentage points), as well as increased time to treatment: 37.2 minutes before dexamethasone initiation and 278.5 minutes before initiation of supplemental oxygen. These results call for additional investigation of pulse oximeters and suggest that current guidelines for development, testing, and calibration of these devices should be revisited, investigated, and revised.
Subject(s)
COVID-19 Drug Treatment , COVID-19 , Dexamethasone , Health Equity , Adult , Humans , COVID-19/therapy , Dexamethasone/therapeutic use , Oximetry/methods , Oxygen/therapeutic use , Healthcare Disparities , Electronic Health RecordsABSTRACT
The big data revolution presents an exciting frontier to expand public health research, broadening the scope of research and increasing the precision of answers. Despite these advances, scientists must be vigilant against also advancing potential harms toward marginalized communities. In this review, we provide examples in which big data applications have (unintentionally) perpetuated discriminatory practices, while also highlighting opportunities for big data applications to advance equity in public health. Here, big data is framed in the context of the five Vs (volume, velocity, veracity, variety, and value), and we propose a sixth V, virtuosity, which incorporates equity and justice frameworks. Analytic approaches to improving equity are presented using social computational big data, fairness in machine learning algorithms, medical claims data, and data augmentation as illustrations. Throughout, we emphasize the biasing influence of data absenteeism and positionality and conclude with recommendations for incorporating an equity lens into big data research.
Subject(s)
Big Data , Public Health , Algorithms , Bias , Humans , Machine LearningABSTRACT
Inequitable gender norms and beliefs contribute to increased sexual risk behavior, and, among adolescent girls and young women (AGYW), risk of HIV acquisition. We investigated the longitudinal measurement properties of the Gender Equitable Men's Scale (GEMS) when applied to a cohort of AGYW in rural South Africa (2011-2015). We used item response theory [Person-Item maps, Differential Item Functioning (DIF)] and measurement invariance confirmatory factor analysis models to assess the validity and reliability of the GEMS instrument. Item difficulty and endorsement of gender equitable beliefs both shifted over time. DIF analysis identified item bias for over half of the items; influenced by age, pregnancy, sexual debut, and intimate partner violence. Measurement invariance models revealed strong longitudinal invariance properties. GEMS is a reliable longitudinal measurement of gender equitable beliefs, with notable bias for specific items when administered to subgroups. Additional items specific to the adolescent experience are warranted for a more stable assessment of gender equitable beliefs in a population facing shifting norms as they mature.
Subject(s)
HIV Infections , Adolescent , Female , HIV Infections/epidemiology , HIV Infections/prevention & control , Humans , Male , Men , Reproducibility of Results , Sexual Behavior , South AfricaABSTRACT
Objectives. To examine differences in HIV prevalence and experiences of discrimination within the trans women community in California's San Francisco Bay Area.Methods. Intersectional positions were constructed on the basis of race/ethnicity (non-Hispanic White, non-Hispanic Black, Latina) and gender identity (female identifying, transgender identifying). We used baseline data from the Trans*National study (2016-2017) to construct regression models that estimated racial/ethnic differences in the attribution of discrimination experienced and, along with surrogate measures for intersectionality, estimated risk among those who were dually marginalized (racial/ethnic minority and transgender identifying). Margins plots were used to visually compare absolute risk across all intersectional positions.Results. Black and Latina trans women were more likely to be HIV positive than non-Hispanic White trans women. In several of the study domains, we estimated a lower risk of reporting discrimination among dually marginalized trans women than among White female-identifying trans women.Conclusions. Quantitative intersectionality methods highlight the diversity of experiences within the trans women community and reveal potential measurement challenges. Despite facing multiple forms of systemic marginalization, racial/ethnic minority trans women report less discrimination than White trans women. Subjective reporting of discrimination likely undercounts risks among racial/ethnic minorities.
Subject(s)
Black or African American/statistics & numerical data , HIV Infections/psychology , Hispanic or Latino/statistics & numerical data , Homophobia/statistics & numerical data , Transgender Persons/statistics & numerical data , White People/statistics & numerical data , Adult , Female , HIV Infections/epidemiology , Health Services Accessibility/statistics & numerical data , Homophobia/psychology , Humans , Middle Aged , Residence Characteristics , San Francisco , Social Perception , Socioeconomic Factors , Transgender Persons/psychology , Young AdultABSTRACT
BACKGROUND: Coronavirus disease 2019 (COVID-19) can lead to acute respiratory distress syndrome (ARDS) but it is unknown whether prone positioning improves outcomes in mechanically ventilated patients with moderate to severe ARDS due to COVID-19. METHODS: A cohort study at a New York City hospital at the peak of the early pandemic in the United States, under crisis conditions. The aim was to determine the benefit of prone positioning in mechanically ventilated patients with ARDS due to COVID-19. The primary outcome was in-hospital death. Secondary outcomes included changes in physiologic parameters. Fine-Gray competing risks models with stabilized inverse probability treatment weighting (sIPTW) were used to determine the effect of prone positioning on outcomes. In addition, linear mixed effects models (LMM) were used to assess changes in physiology with prone positioning. RESULTS: Out of 335 participants who were intubated and mechanically ventilated, 62 underwent prone positioning, 199 met prone positioning criteria and served as controls and 74 were excluded. The intervention and control groups were similar at baseline. In multivariate-adjusted competing risks models with sIPTW, prone positioning was significantly associated with reduced mortality (SHR 0.61, 95% CI 0.46-0.80, P < 0.005). Using LMM to evaluate the impact of positioning maneuvers on physiological parameters, the oxygenation-saturation index was significantly improved during days 1-3 (P < 0.01) whereas oxygenation-saturation index (OSI), oxygenation-index (OI) and arterial oxygen partial pressure to fractional inspired oxygen (PaO2: FiO2) were significantly improved during days 4-7 (P < 0.05 for all). CONCLUSIONS: Prone positioning in patients with moderate to severe ARDS due to COVID-19 is associated with reduced mortality and improved physiologic parameters. One in-hospital death could be averted for every 8 patients treated. Replicating results and scaling the intervention are important, but prone positioning may represent an additional therapeutic option in patients with ARDS due to COVID-19.
Subject(s)
COVID-19/complications , COVID-19/therapy , Prone Position , Respiration, Artificial , Respiratory Distress Syndrome/therapy , Respiratory Distress Syndrome/virology , Respiratory Physiological Phenomena , Adult , Aged , COVID-19/mortality , COVID-19/physiopathology , Female , Hospital Mortality , Humans , Intensive Care Units , Male , Middle Aged , New York City , Oxygen/blood , Respiratory Distress Syndrome/mortality , Respiratory Distress Syndrome/physiopathology , SARS-CoV-2 , Severity of Illness IndexABSTRACT
BACKGROUND: Trans women experience high rates of gender-based violence (GBV)-a risk factor for adverse health outcomes. Transphobic hate crimes are one such form of GBV that affect trans women. However, little is understood about factors that shape transphobic hate crimes and racial/ethnic variation in these experiences. To contextualize GBV risk and police reporting, we examined self-reported types and correlates of transphobic hate crimes by racial/ethnic group of trans women in the San Francisco Bay Area. METHODS: From 2016 to 2018, trans women participated in a longitudinal cohort study of HIV. Secondary data analyses (N = 629) examined self-reported experiences of transphobic hate crimes (i.e., robbery, physical assault, sexual assault, and battery with weapon) by race/ethnicity, and whether hate crimes were reported to the police. Chi-square tests and simple logistic regression examined demographic, sociocultural, and gender identity factors associated with transphobic violence experiences and police reporting. RESULTS: About half (45.8%) of participants reported ever experiencing a transphobic hate crime; only 51.1% of these were reported to the police. Among those who reported a hate crime experience, Black (47.9%) and Latina (49.0%) trans women reported a higher prevalence of battery with a weapon; White (26.7%) and trans women of "other" race/ethnicities (25.0%) reported a higher prevalence of sexual assault (p = 0.001). Having one's gender questioned, history of sex work, homelessness as a child and adult, and a history incarceration were associated with higher odds of experiencing a transphobic hate crime. Trans women who felt their gender identity questioned had lower odds of reporting a hate crime to the police compared to those did not feel questioned. CONCLUSIONS: A high proportion of trans women experienced a transphobic hate crime, with significant socio-structural risk factors and racial differences by crime type. However, crimes were underreported to the police. Interventions that address structural factors, especially among trans women of color, can yield violence prevention benefits.
Subject(s)
Crime Victims , Transgender Persons , Adult , Child , Crime , Female , Gender Identity , Hate , Humans , Longitudinal Studies , Male , San Francisco/epidemiologyABSTRACT
Current population size estimates of communities at risk for HIV are needed to adequately plan interventions and care. We update the estimated number of men who have sex with men (MSM) living in San Francisco. Multiple data sources and approaches, including population growth, registered HIV cases, wisdom of the crowd, and a published estimate on the proportion of adults who are MSM were used for estimation. The several estimates were synthesized into a final figure with acceptable bounds using a Bayesian method. We project 69,974 (acceptable bounds 65,523-74,323) MSM are living in San Francisco as of the beginning of 2017. The population of MSM in San Francisco has increased by 19.4% since 2014.
Subject(s)
Homosexuality, Male/statistics & numerical data , Adult , Bayes Theorem , Health Surveys , Humans , Male , Population Density , Prevalence , San Francisco/epidemiologyABSTRACT
HIV prevalence in Alameda County (including Oakland) is among the highest in California, yet the case registry may under-appreciate the full burden of disease. Using lists from health care facilities serving socioeconomically diverse populations and the HIV surveillance list, we applied capture-recapture methods to evaluate the completeness of the surveillance system by estimating the number of diagnosed people living with HIV and seeking care in Alameda County in 2013. Of the 5376 unique individuals reported from the lists, 397 were missing from the surveillance list. Models projected the total population size to be 5720 (95% CI 5587-6190), estimating the surveillance system as 87% complete. Subgroup analyses identified groups facing a disproportionate burden of HIV as more likely to be detected by the surveillance list. The Alameda County HIV surveillance system reports a high proportion of persons diagnosed with HIV within the jurisdiction. Capture-recapture analysis can help track progress towards maximizing engagement in HIV care.
Subject(s)
HIV Infections/epidemiology , Models, Statistical , Population Surveillance/methods , Registries/statistics & numerical data , Adolescent , Adult , California , Female , Humans , Male , Middle Aged , Prevalence , Risk Factors , Young AdultABSTRACT
Background: Invasive community-associated methicillin-resistant Staphylococcus aureus (MRSA) incidence in the United States is higher among black persons than white persons. We explored the extent to which socioeconomic factors might explain this racial disparity. Methods: A retrospective cohort was based on the Centers for Disease Control and Prevention's Emerging Infections Program surveillance data for invasive community-associated MRSA cases (isolated from a normally sterile site of an outpatient or on hospital admission day ≤3 in a patient without specified major healthcare exposures) from 2009 to 2011 in 33 counties of 9 states. We used generalized estimating equations to determine census tract-level factors associated with differences in MRSA incidence and inverse odds ratio-weighted mediation analysis to determine the proportion of racial disparity mediated by socioeconomic factors. Results: Annual invasive community-associated MRSA incidence was 4.59 per 100000 among whites and 7.60 per 100000 among blacks (rate ratio [RR], 1.66; 95% confidence interval [CI], 1.52-1.80). In the mediation analysis, after accounting for census tract-level measures of federally designated medically underserved areas, education, income, housing value, and rural status, 91% of the original racial disparity was explained; no significant association of black race with community-associated MRSA remained (RR, 1.05; 95% CI, .92-1.20). Conclusions: The racial disparity in invasive community-associated MRSA rates was largely explained by socioeconomic factors. The specific factors that underlie the association between census tract-level socioeconomic measures and MRSA incidence, which may include modifiable social (eg, poverty, crowding) and biological factors (not explored in this analysis), should be elucidated to define strategies for reducing racial disparities in community-associated MRSA rates.
Subject(s)
Community-Acquired Infections/epidemiology , Ethnicity , Healthcare Disparities , Methicillin-Resistant Staphylococcus aureus , Staphylococcal Infections/epidemiology , Adult , Aged , Community-Acquired Infections/microbiology , Comorbidity , Female , Humans , Incidence , Male , Middle Aged , Socioeconomic Factors , Staphylococcal Infections/microbiology , United States/epidemiologyABSTRACT
Worldwide, the HIV epidemic is concentrated among hidden populations (i.e., female sex workers, men who have sex with men, and people who inject drugs). To understand the true scope and scale of the HIV epidemic, estimates of the sizes of these populations are needed. Various methods are available to enumerate hidden populations, but the degree of agreement between these methods has not been formally evaluated. We systematically reviewed the peer-reviewed literature to assess the extent to which different population size estimation methods provide the same estimate of a target population. Of the 341 studies identified from our search, 25 met our eligibility criteria. Twenty-one unique methods were documented. The service multiplier method was the most common in the review. Eighty target populations were estimated, covering 16 countries. We observed variable population size estimates, with little agreement between methods. We note trends in the relative performance of individual methods.
Subject(s)
Drug Users/statistics & numerical data , HIV Infections/epidemiology , Health Services Needs and Demand , Population Density , Sex Workers/statistics & numerical data , Sexual and Gender Minorities/statistics & numerical data , Substance Abuse, Intravenous , Female , Humans , MaleABSTRACT
African-American men who have sex with men (AA MSM) have been disproportionately infected with and affected by HIV and other STIs in San Francisco and the USA. The true scope and scale of the HIV epidemic in this population has not been quantified, in part because the size of this population remains unknown. We used the successive sampling population size estimation (SS-PSE) method, a new Bayesian approach to population size estimation that incorporates network size data routinely collected in respondent-driven sampling (RDS) studies, to estimate the number of AA MSM in San Francisco. This method was applied to data from a 2009 RDS study of AA MSM. An estimate from a separate study of local AA MSM was used to model the prior distribution of the population size. Two-hundred and fifty-six AA MSM were included in the RDS survey. The estimated population size was 4917 (95% CI 1267-28,771), using a flat prior estimated 1882 (95% CI 919-2463) as a lower acceptable bound, and a large prior estimated 6762 (95% CI 1994-13,863) as an acceptable upper bound. Point estimates from the SS-PSE were consistent with estimates from multiplier methods using external data. The SS-PSE method is easily integrated into RDS studies and therefore provides a simple and appealing tool to rapidly produce estimates of the size of key populations otherwise difficult to reach and enumerate.
Subject(s)
Black or African American/statistics & numerical data , HIV Infections/epidemiology , Homosexuality, Male/statistics & numerical data , Sexual Behavior/statistics & numerical data , Adolescent , Adult , Age Factors , Aged , Bayes Theorem , Humans , Male , Middle Aged , Sampling Studies , San Francisco/epidemiology , Socioeconomic Factors , Young AdultABSTRACT
People with schizophrenia are at increased risk for contracting HIV and face higher mortality rates compared with the general population. Viral suppression is key to HIV care, yet little is known about this metric among people with HIV and schizophrenia. A chart review was conducted among people with HIV/AIDS and schizophrenia living in San Francisco who had received inpatient mental health services between 2010 and 2016. Demographic, laboratory, medication, encounter, and discharge data were collected, and were compared with all people living with HIV in San Francisco (PLWH-SF). Among 153 people living with HIV and comorbid schizophrenia, 77% were virally suppressed, compared to 67% for all PLWH-SF. Viral suppression for people with comorbid HIV and schizophrenia living in San Francisco appears higher than PLWH-SF. Further research is needed to confirm the association and mechanisms behind better treatment outcomes for people living with HIV and comorbid schizophrenia.
Subject(s)
HIV Infections , Schizophrenia , Humans , San Francisco/epidemiology , Schizophrenia/epidemiology , HIV Infections/epidemiology , HIV Infections/drug therapy , HIV Infections/complications , Male , Female , Retrospective Studies , Adult , Middle Aged , Inpatients/statistics & numerical data , Inpatients/psychology , Comorbidity , Viral LoadABSTRACT
PURPOSE: Capture-recapture methods estimate the size of hidden populations by leveraging the proportion of overlap of the population on independent lists. Log-linear modeling relaxes the assumption of list independence, but best model selection criteria remain uncertain. Incorrect model selection can deliver incorrect and even implausible size estimates. METHODS: We used simulations to model when capture-recapture methods deliver biased or unbiased estimates and compare model selection criteria. Simulations included five scenarios for list dependence among three incomplete lists of population of interest. We compared metrics of log-linear model selection, accuracy, and precision. We also compared log-linear model performance to the decomposable graph approach (a Bayesian model average), the sparse multiple systems estimation (SparseMSE) approach that accounts for zero or low cell counts, and the Sample Coverage approach. RESULTS: Log-linear models selected by Akaike's information criterion (AIC) calculated accurate population size estimates in all scenarios except for those with sparse or zero cell counts. In these scenarios, the decomposable graph and the Sample Coverage models produced more accurate size estimates. CONCLUSIONS: Conventional capture-recapture model selection fails with sparse cell counts. This naïve approach to model selection should be replaced with the implementation of multiple different models in order triangulate the truth in real-world applications.
Subject(s)
Records , Humans , Bayes Theorem , Linear Models , Population DensityABSTRACT
Purpose: The purpose of our study was to examine the effects of mental distress (depression, anxiety, and post-traumatic stress disorder [PTSD]), incarceration, and hate crime on stimulant use (methamphetamine, crack, and cocaine) among transgender women. Methods: We conducted a secondary analysis of longitudinal data collected from 2016 to 2018 with 429 transgender women in the San Francisco Bay Area. Generalized estimating equation log-binomial regressions were used to calculate relative risks of stimulant use associated with mental distress, incarceration, and hate crime. Results: At baseline, transgender women experienced transphobic hate crime (46.4%), incarceration (53.0%), mental distress (69.2%), and stimulant use (28.4%). Transgender women who used stimulants reported lower education (45.1%, χ2 = 14.3, p = 0.001) and significantly more had been incarcerated (62.3%, χ2 = 5.9, p = 0.015), and reported diagnoses of depression (67.8%, χ2 = 6.1, p = 0.014), anxiety (62.8%, χ2 = 4.3, p = 0.039), and PTSD (43.8%, χ2 = 6.7, p = 0.010). Longitudinal multivariate analysis found that depression (adjusted relative risk [aRR] = 1.46, 95% confidence interval [CI] 1.09-1.95), anxiety (aRR = 1.42, 95% CI = 1.05-1.93), and PTSD (aRR = 1.38, 95% CI = 1.02-1.87) were associated with methamphetamine use but not with crack or cocaine use. Incarceration was associated with methamphetamine use and crack use, whereas experiencing hate crime was associated with crack use. Conclusions: Mental distress, incarceration, and hate crime were key exposures of stimulant use among transgender women. Intervention targets for reducing stimulant use should consider working upstream by addressing underlying stressors impacting mental health for transgender women, including laws to protect transgender women from hate crime and to reduce their disproportionate representation in the criminal justice system.
Subject(s)
Cocaine , Methamphetamine , Transgender Persons , Humans , Female , Crime , Mental HealthABSTRACT
BACKGROUND: Estimation of abortion incidence, particularly in settings where most abortions occur outside of health facility settings, is critical for understanding information gaps and service delivery needs in different settings. However, the existing methods for measuring out-of-facility abortion incidence are plagued with methodological challenges. Respondent-driven sampling (RDS) may offer a methodological improvement in the estimation of abortion incidence. OBJECTIVE: In this study, we tested the feasibility of using RDS to recruit participants into a study about abortion and estimated the proportion of people who ever attempted abortion as well as 1-year and 5-year incidence of abortion (both in-facility and out-of-facility settings) among women of reproductive age in Soweto, South Africa. METHODS: Participants were eligible if they identified as a woman; were aged between 15 and 49 years; spoke English, Tswana, isiZulu, Sotho, or Xhosa; and lived in Soweto. Working with community partners, we identified 11 seeds who were provided with coupons to refer eligible peers to the study. Upon arrival at the study site, the recruits completed an interviewer-administered questionnaire that solicited information about demographic characteristics, social network composition, health behaviors, sexual history, pregnancy history, and experience with abortion; recruits also received 3 recruitment coupons. Recruitment was tracked using coupon numbering. We used the RDS-II estimator to estimate the population proportions of demographic characteristics and our primary outcome, the proportion of people who ever attempted abortion. RESULTS: Between April 4, 2018, and December 17, 2018, 849 eligible participants were recruited into the study. The estimated proportion of people who ever attempted abortion was 12.1% (95% CI 9.7%-14.4%). A total of 7.1% (95% CI 5.4%-8.9%) reported a facility-based abortion, and 4.4% (95% CI 3.0%-5.8%) reported an out-of-facility abortion. CONCLUSIONS: The estimated proportion of people who ever attempted abortion of 12% (102/849) in our study likely represents a substantial underestimation of the actual proportion of abortion attempts among this study population-representing a failure of the RDS method to generate more reliable estimates of abortion incidence in our study. We caution against the use of RDS to measure the incidence of abortion because of persistent concerns with underreporting but consider potential alternative applications of RDS with respect to the study of abortion.
Subject(s)
Sampling Studies , Humans , Female , Adolescent , Young Adult , Adult , Middle Aged , South Africa/epidemiologyABSTRACT
Introduction: COVID-19 has caused tremendous death and suffering since it first emerged in 2019. Soon after its emergence, models were developed to help predict the course of various disease metrics, and these models have been relied upon to help guide public health policy. Methods: Here we present a method called COVIDNearTerm to "forecast" hospitalizations in the short term, two to four weeks from the time of prediction. COVIDNearTerm is based on an autoregressive model and utilizes a parametric bootstrap approach to make predictions. It is easy to use as it requires only previous hospitalization data, and there is an open-source R package that implements the algorithm. We evaluated COVIDNearTerm on San Francisco Bay Area hospitalizations and compared it to models from the California COVID Assessment Tool (CalCAT). Results: We found that COVIDNearTerm predictions were more accurate than the CalCAT ensemble predictions for all comparisons and any CalCAT component for a majority of comparisons. For instance, at the county level our 14-day hospitalization median absolute percentage errors ranged from 16 to 36%. For those same comparisons, the CalCAT ensemble errors were between 30 and 59%. Conclusion: COVIDNearTerm is a simple and useful tool for predicting near-term COVID-19 hospitalizations.