Comparing performance between log-binomial and robust Poisson regression models for estimating risk ratios under model misspecification

Chen, Wansu; Qian, Lei; Shi, Jiaxiao; Franklin, Meredith

doi:10.1186/s12874-018-0519-5

Cited by 355 publications

(256 citation statements)

References 27 publications

Supporting

Mentioning

229

Contrasting

Unclassified

Order By: Relevance

“…Multivariate mixed-effect Poisson models were used to identify associations between selected covariates and the probability of having a drug resistant infection. Using a Poisson distribution and estimating a robust variance is an alternative to fitting log-risk models (that use a binomial distribution), which often do not converge (36,37). Full specification of the model is available in Supplemental Text 1.…”

Section: Comparison Of Allele Frequenciesmentioning

confidence: 99%

The changing landscape of Plasmodium falciparum drug resistance in the Democratic Republic of Congo

Deutsch-Feldman

Aydemir

Carrel

et al. 2019

Preprint

View full text Add to dashboard Cite

Background: Drug resistant malaria is a growing concern in the Democratic Republic of the Congo (DRC), where previous studies indicate that parasites resistant to sulfadoxine/pyrimethamine or chloroquine are spatially clustered. This study explores longitudinal changes in spatial patterns to understand how resistant malaria may be spreading within the DRC, using samples from nation-wide population-representative surveys. Methods: We selected 552 children with PCR-detectable Plasmodium falciparum infection and identified known variants in the pfdhps and pfcrt genes associated with resistance. We compared the proportion of mutant parasites in 2013 to those previously reported from adults in 2007, and identified risk factors for carrying a resistant allele using multivariate mixed-effects modeling. Finally, we fit a spatial-temporal model to the observed data, providing smooth allele frequency estimates over space and time. Results: The proportion of co-occurring pfdhps K540E/A581G mutations increased by 16% between 2007 and 2013. The spatial-temporal model suggests that the spatial range of the pfdhps double mutants expanded over time, while the prevalence and range of pfcrt mutations remained steady. Conclusions: This study uses population-representative samples to describe the changing landscape of SP resistance within the DRC, and the persistence of chloroquine resistance. Vigilant molecular surveillance is critical for controlling the spread of resistance.

show abstract

Section: Comparison Of Allele Frequenciesmentioning

confidence: 99%

The changing landscape of Plasmodium falciparum drug resistance in the Democratic Republic of Congo

Deutsch-Feldman

Aydemir

Carrel

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…We examined the associations between these preferences and age, sexual identity and gender identity characteristics using chi-squared tests with a correction for survey data [35]. As the outcome was prevalent, we used RDS-II weighted robust Poisson regression (also dropping seed participants) to obtain weighted prevalence ratios [36,37] to assess associations between testing in the last 6 months and sociodemographic, sexual behaviour and self-reported STI symptoms, examining first bivariate associations and then adjusting sexual behaviour and STI symptom models for any sociodemographic variables showing evidence for associations with the outcome (Wald test p<0.2).…”

Section: Discussionmentioning

confidence: 99%

HIV testing, care and viral suppression among men who have sex with men and transgender individuals in Johannesburg, South Africa

et al. 2020

View full text Add to dashboard Cite

Introduction Men who have sex with men and transgender individuals (MSM/TG) carry a disproportionately high burden of HIV, including in South Africa. However, there are few empirical population-representative estimates of viral suppression and the HIV care cascade including HIV testing among this population, nor of factors associated with these outcomes. Methods We conducted a respondent driven sampling (RDS) survey among 301 MSM/TG in Johannesburg in 2017. Participants gave blood samples for HIV testing and viral load. Participants self-completed a survey including sociodemographics, HIV testing history, and engagement in care. We calculated RDS-II weighted estimates of the percentage of HIV-negative MSM/ TG reporting HIV testing in the previous 6 months, their testing experience and preferences. Among those HIV-positive, we estimated the percentage status-aware, on ART, and virally suppressed (<50 viral copies/ml plasma). We conducted RDS-weighted robust Poisson regression to obtain weighted prevalence ratios of factors associated with 1) HIV testing among those HIV-negative; and 2) viral suppression among those HIV-positive. Results There were 118/300 HIV-positive MSM/TG, (37.5%). Of the HIV-negative MSM/TG, 61.5% reported that they had tested for HIV in the previous 6 months, which was associated with selling sex to men (Prevalence Ratio = 1.67, 95% CI 1.36-2.05). There were 76/118 HIVpositive MSM/TG (56.5%) who reported having previously tested positive for HIV and 39/ 118 (30.0%) who reported current ART. There were 58/118 HIV-positive MSM/TG with viral loads <50 copies/ml plasma (46.9%). Viral suppression was associated with older age

show abstract

“…18 With this model, the unbiased prevalence ratio (PR) can be measured directly. [18][19][20] We con-ducted both unadjusted bivariate analyses to assess the relation between the outcome and exposure variables, and a multi variate adjusted model. All models used robust error variances, and we used survey weights for all statistical analyses to account for the survey sampling design.…”

Section: Discussionmentioning

confidence: 99%

Income-related disparities in private prescription drug coverage in Canada

Bolatova

Law

2019

cmajo

View full text Add to dashboard Cite

Background: Canada does not have universal public coverage for prescription drugs, which leaves an important role for private insurance plans. However, we do not have recent data on the characteristics of Canadians who report holding such coverage, particularly differences based on household income. We performed a study to examine the relation between household income and private drug insurance coverage in Canada. Methods: We used data from the 2015-2016 cycle of the Canadian Community Health Survey to investigate the relation between household income and holding private drug insurance. We constructed modified multivariate Poisson regression models with robust error variances, including several potential confounders. Results: Overall, 59.4% of respondents reported having private drug insurance. We found a strong dose-response relation between household income level and private drug insurance coverage: 19.8% of those with a household income less than $20 000 reported private coverage, compared to 76.2% of those with a household income of $80 000 or more. Interpretation: Higher-income households are much more likely to hold private drug insurance coverage in Canada. This likely contributes to differential access to medicines and health outcomes by different income groups.

show abstract

Comparing performance between log-binomial and robust Poisson regression models for estimating risk ratios under model misspecification

Cited by 355 publications

References 27 publications

The changing landscape of Plasmodium falciparum drug resistance in the Democratic Republic of Congo

The changing landscape of Plasmodium falciparum drug resistance in the Democratic Republic of Congo

HIV testing, care and viral suppression among men who have sex with men and transgender individuals in Johannesburg, South Africa

Income-related disparities in private prescription drug coverage in Canada

Contact Info

Product

Resources

About