Development and validation of an automated HIV prediction algorithm to identify candidates for pre-exposure prophylaxis: a modelling study

Krakower, Douglas S.; Gruber, Susan; Hsu, Katherine; Menchaca, John T.; Maro, Judith C.; Kruskal, Benjamin A.; Wilson, Ira B.; Mayer, Kenneth H.; Klompas, Michael

doi:10.1016/s2352-3018(19)30139-0

Cited by 91 publications

(73 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Instead of relying on 10‐fold cross validation, we evaluated the loss on an external validation set. The dataset contained information on patients with at least one risk factor seen in Atrius in 2016 (n = 245,475, 16 of whom were cases) 3 . Two percent of these 2016 patients were also in the 2007‐2015 dataset used to fit the model, albeit with different values of time‐dependent covariates.…”

Section: Resultsmentioning

confidence: 99%

“…These patients had at least one of 180 clinician‐specified risk factors for HIV acquisition and/or were diagnosed with incident HIV. Data on an additional n =755 579 patients who had 0 recorded risk factors were not available to us (details were previously published 3 ). A patient with at least one risk factor contributed one observation to our dataset for each year there was an encounter with the health care system, for a total of 2.3 million observations.…”

Section: Discussionmentioning

confidence: 99%

“…Choosing a risk score threshold to identify patients who should be further evaluated as PrEP candidates involves trading off sensitivity and specificity. This trade‐off should take into acount the relative costs of false positives and false negatives, and the availability of local resources to act on the information (see Reference 3).…”

Section: Discussionmentioning

confidence: 99%

“…One challenge is that HIV acquisition is extremely rare. A second challenge is that although covariates in EHR data are readily available, they are imperfect proxies for behavioral patterns indicative of exposure to the virus 3 . CDC criteria for PrEP prescribing rely on behavioral characters such as receptive anal intercourse or number of shared needles that are not routinely captured in the EHR 4 .…”

Section: Introductionmentioning

confidence: 99%

“…Observations on each patient consisted of demographic information, diagnosis codes, procedure codes, drug prescriptions, laboratory tests, and results. For clinical aspects of this work and details on the data source, we refer the interested reader to a companion paper in the medical literature 3 . In this article, we describe our methodologic approach to risk prediction that harnesses both clinical expertise and the tools of machine learning.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Using electronic health records to identify candidates for human immunodeficiency virus pre‐exposure prophylaxis: An application of super learning to risk prediction when the outcome is rare

Gruber¹,

Krakower

Menchaca

et al. 2020

Statistics in Medicine

Self Cite

View full text Add to dashboard Cite

Human immunodeficiency virus (HIV) pre‐exposure prophylaxis (PrEP) protects high risk patients from becoming infected with HIV. Clinicians need help to identify candidates for PrEP based on information routinely collected in electronic health records (EHRs). The greatest statistical challenge in developing a risk prediction model is that acquisition is extremely rare. Methods: Data consisted of 180 covariates (demographic, diagnoses, treatments, prescriptions) extracted from records on 399 385 patient (150 cases) seen at Atrius Health (2007‐2015), a clinical network in Massachusetts. Super learner is an ensemble machine learning algorithm that uses k‐fold cross validation to evaluate and combine predictions from a collection of algorithms. We trained 42 variants of sophisticated algorithms, using different sampling schemes that more evenly balanced the ratio of cases to controls. We compared super learner's cross validated area under the receiver operating curve (cv‐AUC) with that of each individual algorithm. Results: The least absolute shrinkage and selection operator (lasso) using a 1:20 class ratio outperformed the super learner (cv‐AUC = 0.86 vs 0.84). A traditional logistic regression model restricted to 23 clinician‐selected main terms was slightly inferior (cv‐AUC = 0.81). Conclusion: Machine learning was successful at developing a model to predict 1‐year risk of acquiring HIV based on a physician‐curated set of predictors extracted from EHRs.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Using electronic health records to identify candidates for human immunodeficiency virus pre‐exposure prophylaxis: An application of super learning to risk prediction when the outcome is rare

Gruber¹,

Krakower

Menchaca

et al. 2020

Statistics in Medicine

Self Cite

View full text Add to dashboard Cite

show abstract

Preexposure Prophylaxis to Prevent Acquisition of HIV

Barry,

Nicholson,

Silverstein

et al. 2023

JAMA

View full text Add to dashboard Cite

ImportanceAn estimated 1.2 million persons in the US currently have HIV, and more than 760 000 persons have died of complications related to HIV since the first cases were reported in 1981. Although treatable, HIV is not curable and has significant health consequences. Therefore, effective strategies to prevent HIV are an important public health and clinical priority.ObjectiveThe US Preventive Services Task Force (USPSTF) commissioned a systematic review to evaluate the benefits and harms of preexposure prophylaxis with antiretroviral therapy for the prevention of HIV acquisition, and the diagnostic accuracy of risk assessment tools to identify persons at increased risk of HIV acquisition.PopulationAdolescents and adults who do not have HIV and are at increased risk of HIV.Evidence AssessmentThe USPSTF concludes with high certainty that there is a substantial net benefit from the use of effective antiretroviral therapy to reduce the risk of acquisition of HIV in persons at increased risk of acquiring HIV.RecommendationThe USPSTF recommends that clinicians prescribe preexposure prophylaxis using effective antiretroviral therapy to persons at increased risk of HIV acquisition to decrease the risk of acquiring HIV. (A recommendation)

show abstract

Preexposure Prophylaxis for the Prevention of HIV

Chou,

Spencer,

Bougatsos

et al. 2023

JAMA

View full text Add to dashboard Cite

ImportanceA 2019 review for the US Preventive Services Task Force (USPSTF) found oral preexposure prophylaxis (PrEP) associated with decreased HIV infection risk vs placebo or no PrEP in adults at increased HIV acquisition risk. Newer PrEP regimens are available.ObjectiveTo update the 2019 review on PrEP, to inform the USPSTF.Data SourcesOvid MEDLINE, Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials, and Embase (January 2018 to May 16, 2022); surveillance through March 24, 2023.Study SelectionRandomized clinical trials of PrEP vs placebo or no PrEP or newer vs older PrEP regimens and diagnostic accuracy studies of instruments for predicting incident HIV infection.Data Extraction and SynthesisDual review of titles and abstracts, full-text articles, study quality, and data abstraction. Data were pooled using the DerSimonian and Laird random-effects model.Main Outcomes and MeasuresHIV acquisition, mortality, and harms; and diagnostic test accuracy.ResultsThirty-two studies were included in the review (20 randomized clinical trials [N = 36 575] and 12 studies of diagnostic accuracy [N = 5 544 500]). Eleven trials in the 2019 review found oral PrEP associated with decreased HIV infection risk vs placebo or no PrEP (n = 18 172; relative risk [RR], 0.46 [95% CI, 0.33-0.66]). Higher adherence was associated with greater efficacy. One new trial (n = 5387) found oral tenofovir alafenamide/emtricitabine (TAF/FTC) to be noninferior to tenofovir disoproxil fumarate/emtricitabine (TDF/FTC) in men who have sex with men (RR, 0.53 [95% CI, 0.23-1.26]). Two new trials found long-acting injectable cabotegravir associated with decreased risk of HIV infection vs oral TDF/FTC (RR, 0.33 [95% CI, 0.18-0.62] in cisgender men who have sex with men and transgender women [n = 4490] and RR, 0.11 [95% CI, 0.04-0.31] in cisgender women [n = 3178]). Discrimination of instruments for predicting incident HIV infection was moderate in men who have sex with men (5 studies; n = 25 488) and moderate to high in general populations of persons without HIV (2 studies; n = 5 477 291).Conclusions and RelevanceIn adults at increased HIV acquisition risk, oral PrEP was associated with decreased risk of acquiring HIV infection compared with placebo or no PrEP. Oral TAF/FTC was noninferior to oral TDF/FTC, and injectable cabotegravir reduced the risk of HIV infection compared with oral TDF/FTC in the populations studied.

show abstract

Development and validation of an automated HIV prediction algorithm to identify candidates for pre-exposure prophylaxis: a modelling study

Cited by 91 publications

References 30 publications

Using electronic health records to identify candidates for human immunodeficiency virus pre‐exposure prophylaxis: An application of super learning to risk prediction when the outcome is rare

Using electronic health records to identify candidates for human immunodeficiency virus pre‐exposure prophylaxis: An application of super learning to risk prediction when the outcome is rare

Preexposure Prophylaxis to Prevent Acquisition of HIV

Preexposure Prophylaxis for the Prevention of HIV

Contact Info

Product

Resources

About