GetaKit is a University of Ottawa study to evaluate the outcomes of a mail-

out HIV self-testing program. Here's what we found.

Automated STI/HIV risk assessments: Testing an

online clinical algorithm in Ottawa, Canada
There are a few different ways of HIV testing in Ontario. You can do a blood test with your family
doctor, a point-of-care (rapid) test at your local sexual health clinic, or order an HIV self-test. Each
test detects HIV infection differently. This means that a blood test can detect HIV at 6 weeks after
exposure, but both point-of-care and self-tests should be used after a window period of 3-months.

This is why GetaKit uses the self-assessment to make sure that each HIV self-test that is sent is
appropriate. The self-assessment questions we ask participants are based on the types of questions
you would be asked if you were visiting a clinic in person. Our system takes your answers and
calculates your level of risk. If your risk is low, it might be better to get a blood test. If your risk is
higher, an HIV self-test is a good option.

Here are some examples:

Eric is HIV-negative, black, cis-gendered, 40 year-old
man who has sex with men. He does not use
condoms and is not taking PrEP. He last tested for HIV
and STIs about 6-months ago and has had new
sexual partners since, but he's unsure of their risk.
GetaKit will recommend an HIV self-test for Eric, as
well as testing for syphilis, gonorrhea and chlamydia.

Sarah is HIV-negative, white, cis-gendered, 20 year-old woman

who has sex with men. She doesn’t always use condoms when
having sex. Sarah and her partners don't use injection drugs. She
was tested 6 months ago for HIV and other STIs and has had new
sexual partners since. GetaKit will recommend vaginal
gonorrhea/chlamydia testing, but would also recommend a blood
test instead of an HIV self-test.

Want to know what GetaKit would recommend for you? Try the HIV/STI Screener here.

We strongly support open access, which is why you can read the
full article here.
Automated STI/HIV risk assessments: Testing an online clinical algorithm in Ottawa, Canada
Patrick O'Byrne, Alexandra Musten, Lauren Orser and Scott Buckingham
International Journal of STD and AIDS (32), 1365-1373 (2021) https://doi.org/10.1177/09564624211031322
Miscellaneous paper

International Journal of STD & AIDS

2021, Vol. 32(14) 1365–1373
Automated STI/HIV risk assessments: © The Author(s) 2021

Testing an online clinical algorithm in Article reuse guidelines:

Ottawa, Canada sagepub.com/journals-permissions
DOI: 10.1177/09564624211031322
journals.sagepub.com/home/std

Patrick O’Byrne1 , Alexandra Musten2, Lauren Orser1,  and

Scott Buckingham3

Abstract
Despite the ongoing transmission of sexually transmitted infections (STIs) and HIV, many people became unable to access
testing due to COVID-19. To address this, we created a mail-out HIV self-test kit, which could be delivered without
restrictions in our region. The uptake and feedback from this project made us realize that comprehensive STI testing was
being sought. To ensure testing occurred correctly—that is, it would be targeted at the persons most affected by STIs/
HIV—we automated clinical decision-making. We built this model based on a 2-by-2 matrix that plots the risk of STI/HIV
transmission and risk of STI/HIV exposure. The intercept of these two measures classifies a person as low, medium, or high
risk. After automating this logic, 16 expert clinicians in STI/HIV care tested this system with over 400 test patient cases and
refined the algorithm until it yielded the exact outcomes that these clinicians would offer patients based on guidelines.
Findings of interest are that the scale of the y-axis is exponential, in that risk factors for exposure do not climb cumulatively
but do so according to a quadratic equation. This helps ensure that testing services are targeted at those who are most
inequitably burdened by these infections.

Keywords
Screening, HIV, sexually transmitted infections, self-testing, online testing

Introduction bacterial STIs, treatment eradicates infection, while for HIV,

treatment can yield viral load suppression and a state of
Across Canada and the United States, over the last 10 years, virtual non-infectivity.11 Following a status neutral ap-
the rates of most bacterial sexually transmitted infections proach,12 when testing yields negative results, other pre-
(STIs) have increased, while those for HIV have remained vention techniques, such as counseling, pre-exposure
relatively stable.1 These infections, with little change, also prophylaxis (PrEP), and condoms, can mitigate risk.
continue to disproportionately affect the same, often mi- Many people, however, do not access care for reasons
nority, subgroups. In Ottawa, Canada, where this study is related to distance, timing of services, wait times, cost, or
based, for HIV, the most affected populations include gay, fear of stigma (by clinicians and related to being seen
bisexual, and other men who have sex with men (gbMSM); seeking testing).13–15 Research also consistently identifies
persons of African, Caribbean, or Black (ACB) ethnicities; that one-third to one-half of gbMSM do not disclose their
persons who use drugs; trans persons; and members of sexual orientation and practices to their primary care providers
Indigenous communities.2 For chlamydia, most infections due to concerns about how practitioners might react.16,17
are diagnosed in persons under 30 years of age, more often
in females than males.3 gbMSM are also unequally bur-
dened by gonorrhea,4 with research suggesting that up to 1
School of Nursing, University of Ottawa, Ottawa, ON, Canada
2
70% of such infections in gbMSM are in the oropharynx and Ontario HIV Treatment Network, Toronto, ON, Canada
3
rectum, not the genital tract.5 Syphilis, lastly, is mostly Machine, Ottawa, Canada
diagnosed among gbMSM,6 although recent increases
Corresponding author:
among females7 suggest changes in epidemiology. Patrick O’Byrne, School of Nursing, University of Ottawa, 451 Smyth
One prevention technique for STIs/HIV is testing, which Road, Ottawa, ON K1H 8M5, Canada.
identifies infections and connects people with care.8–10 For Email: [email protected]
1366 International Journal of STD & AIDS 32(14)

Another barrier is that clinicians may be unaware of how or partners.8 This includes inquiries about oral, vaginal, and
when to offer testing, resulting in missed opportunities for anal sex; about prevention strategies (e.g., condoms and
earlier diagnoses or failed identification of extragenital in- PrEP); about the sex/gender and ethnicity of partners;
fections.18,19 The outcome is that some patients either do not and other risk practices (e.g., injection drug use and sex
obtain testing or receive incomplete testing when they seek work). (Table 1.) Then, clinicians analyze collected data to
care—due to both personal apprehensions and health systems determine risk.
failures. As always, these barriers are more pronounced for Figure 1 illustrates this clinical risk assessment process
minority groups.20 Unsurprisingly, the COVID-19 pandemic with a two-by-two matrix, which has a person’s reported
worsened access for STI/HIV care. In our jurisdiction, sexual risk practices and associated risk of transmission on the x-
health clinics closed walk-in services and screening for axis and their probability of exposure1 to a given STI on
asymptomatic persons, which broadly resulted in an ap- the y-axis. As one moves along the x- and y-axes, risk varies.
proximately 75% reduction in HIV testing.
To address this myriad of access issues, we developed
GetaKit.ca,21 where persons can create an account, com- Table 1. STI/HIV risk assessment questions.
plete an STI self-assessment, and obtain STI/HIV screening Question category Sub-questions
based on reported practices. Our hope was that this system
would promote STI/HIV testing both related to the COVID- Demographics Age
19 pandemic and ongoing barriers to care, and would build Sex
on research that computer-assisted interviews yield more Sexual orientation
truthful answers regarding STI risk practices, compared to Gender
Ethnicity
clinician-obtained histories.22,23 We also hoped that GetaKit
Country of birth
would promote testing among visible and sexual minorities,
Sex practices Oral, vaginal, anal, sex toys
although research24–26 suggests that online systems may be Sex partner characteristics Sex and gender
under-utilized by members of racialized communities. ACB
Nevertheless, we hoped that, even if a targeted outreach and Born in countries where HIV
a simple interface could not address this barrier, then GetaKit is endemic
might at least streamline services for other groups, thus Bisexual
freeing up limited in-person clinician-time to provide services Injection drug use
to minority groups. To implement GetaKit, we obtained Sex work
funding from the Ontario HIV Treatment Network and re- Other risk practices Personal injection drug use
search ethics approval from the University of Ottawa (H-02- Sex work
20-5518). All participants who have used GetaKit have STI/HIV history When last tested
New partners since last tested
provided expressed consent for research and online services.
Prior diagnoses
Unique to our project was that we created an algorithm
which (1) stratifies participants based on reported risk STI: sexually transmitted infection; ACB: African, Caribbean, and Black.
practices and (2) recommends testing based on clinical
guidelines.27,28 That is, we created an algorithm that au-
tomates STI/HIV clinical decision-making and which rec-
ommends specific tests to individual participants based on
their reported information. While other online risk calcu-
lators exist, these often only stratify persons’ level of risk,
whereas our system recommends and provides direct access
to relevant STI/HIV testing.29,30 In this article, we report on
our algorithm and demonstrate its functionality using five
archetypal patients which show the algorithm’s re-
sponsiveness to varying risk profiles. These cases also
highlight how our algorithm could help ensure that mem-
bers of the groups most affected by STIs/HIV receive
comprehensive testing.

Understanding STI/HIV risk assessments

The task of STI/HIV clinicians is to determine which parts
of patients’ anatomy were involved in sexual contact, when Figure 1. STI/HIV risk matrix. STI: sexually transmitted
contacts occurred, and what were the characteristics of infection.
O’Byrne et al. 1367

To determine risk, clinicians must obtain information about

each axis and plot the intercept.
To explain further, we can use the case of a gay man who
neither uses condoms nor PrEP and engages in receptive
anal sex. For HIV, these practices put him at the far right of
the x-axis and his same-sex partners put him at the top of y-
axis. Plotting this makes him “high risk.” (Figure 2(a).)
If, however, this same person only engages in oral sex
with male partners, the risk of HIV transmission is low, even
if the probability of having an HIV-positive partner is high.
(Figure 2(b).)
Similarly, a white Canadian-born 20-year-old female
who engages in condomless vaginal sex with white male
partners is low risk for HIV (because while the risk of
transmission is high, the probability of exposure is low—
Figure 2(c)); she would, however, be high risk for chlamydia
(as the prevalence is high among this demographic—
Figure 2(a)).
This matrix can, therefore, determine overall risk by
plotting exposure probability and risk of transmission. (For
more information on this matrix, please see O’Byrne
et al.38). The limitation though is that clinicians must know
the risk of transmission for all sexual practices and the
prevalence of STIs/HIV in many populations. While this
may be possible in STI clinics, it is less possible in primary
care. Other barriers to testing (as noted above) further
undermine access to testing and limit the effectiveness of
testing for STI/HIV prevention initiatives.

The algorithm
Building the algorithm
Due to higher-than-expected uptake21 2 for our HIV self-
testing project and profound restrictions on access to STI/
HIV testing due to the COVID-19 pandemic, we worked to
offer full STI testing via our online platform. This involved
three steps. First, we engaged in community consultations
with local gbMSM, ACB, Indigenous, and trans organ-
izations to create culturally sensitive and trauma-informed
questions that would be non-stigmatizing for participants.
As part of this, we built a 20-question STI self-assessment
that participants could complete via GetaKit.ca. These
questions inquired about all items in Table 1. As part of this
self-assessment, participants were encouraged to seek in-
person care if they reported symptoms, were a contact of an
STI or HIV, or required post-exposure prophylaxis (PEP) or
emergency contraception. We also reviewed STI/HIV Figure 2. (a) Examples of STI/HIV risk assessments.
testing window periods and encouraged retesting based
on reported timelines. The logic to our model regarding
window periods was to “test and retest” to identify in- Second, we reviewed the GetaKit self-assessment with
fections that pre-existed the last reported sexual contact. We stakeholders who worked in the field of STI/HIV testing and
identified the utility of this approach in our PEP study,31 prevention and modified the language and questions ac-
where some participants who presented for PEP had un- cordingly. This phase did not involve research participants,
diagnosed HIV infections. but peers who worked for partner agencies. Our goal was to
1368 International Journal of STD & AIDS 32(14)

have knowledgeable peers in community organizations For the y-axis, we weighted populations based on local
refine our self-assessment. prevalence, with low prevalence groups having a score ≤0,
Third, we converted Figure 1 into an algorithm that medium prevalence groups having a score ranging between 1
a computer could use to impute a risk score based on the and 11, and high prevalence groups having a score ≥12. For
reported data from Table 1 for the following infections: HIV, population variables, we included age, sex, gender, ethnicity,
syphilis, hepatitis C, and gonorrhea and chlamydia for all sex of partners, characteristics of sex partners (ACB, HIV-
anatomical sites where one could acquire these infections positive, IDU), personal use of illicit drugs, engagement in
(oropharynx, rectum, vagina, and urethra). This involved sex work, last time tested, and if the participant reported new
creating scores for both the x- and y-axes and numerical sexual partners since their last STI/HIV testing. For each
thresholds for tests to be recommended. Our calculation was infection, the scores attached to each of the foregoing items
simple: sum the risk score, sum the population score, were combined to generate a final population score. Figure 4
multiply these scores, and determine if these outcome shows an example of this weighting for HIV.
values breached the test threshold. We created this algorithm
in Google Sheets, which was sufficiently robust for our
needs. Indeed, because our model stratifies participants
Testing the algorithm
based on their risk profiles and risk practices, we were not To operationalize the algorithm, we had the computer
predicting and did not need more advanced software. sum the risk scores for practices (x-axis) and exposure (y-
For the x-axis, we used established risk levels from the axis) and multiply these to determine the final score. This
research for the probability of STI/HIV transmission for means that, were a person’s reported practices to include
different practices to stratify practices as low, medium, or condomless receptive and penetrative anal sex, the for-
high risk.32–34 We determined that the scores for a low-risk mula dictated that the computer would add the score for
practice were <1, for medium-risk practice ranged from 1 to engaging in receptive anal sex (a high-risk score for HIV)
9, and for a high-risk practice were ≥10. Items such as with the score for penetrative anal sex (a high-risk score
HIV-status, PrEP, and condom use further adjusted the risk for HIV) to yield the final score for HIV testing. As can be
scores. (Figure 3.) The variation in the assigned scores seen from Figure 3, the cumulative risk score for HIV
allowed our algorithm to trigger specific testing for certain serology for the foregoing practices in someone with
practices in isolation (e.g., injection drug use and hepatitis C a penis would be 75 points. The same process occurred
testing) or only when a set of risk practices were reported in for all reported practices for each infection. The output of
combination (e.g., receptive anal sex and male gender for this calculation was then multiplied by the total pop-
rectal gonorrhea/chlamydia testing). ulation score and compared to the appropriate test

Figure 3. STI/HIV risk scores.

Figure 4. Population scores.

O’Byrne et al. 1369

threshold. The algorithm then recommended tests when drug use and does not report doing so himself, nor does he
the test threshold was exceeded. While the risk matrix report sex work. Based on these practices, clinical guide-
classified risk as low, medium, or high, the testing lines would indicate that this person should receive HIV
threshold was set at a medium risk. testing (serology and/or a point-of-care test), syphilis se-
At first, this process did not yield results that corre- rology, and gonorrhea and chlamydia testing by urine and
sponded with recommended STI/HIV testing. We at- by oral and rectal swabs. The algorithm correctly identified
tempted to correct this by varying the test thresholds, but these recommended tests. See Box 1.
this did not yield the desired outcomes. Through further
development, we realized that increasing population risk Box 1.
scores exponentially rectified this situation. That is, we
Test Final Pop. Pop. Question
determined that two population risk factors did not sum Testv required score multiplier score score
as 1 + 1 = 2, but rather, increased in larger steps based
on ranges. A population score of ≤2 was multiplied by HIV Yes 1125 15 19 75
0 and a score of ≥3 was transformed using the following serology
equation: HIV self- Yes 450 15 19 30
test
y ¼ 0:0001x3 þ 0:356x2 þ 0:0978x  0:3615 Oral Yes 85.455 105.5 53 0.81
Rectal Yes 1899 105.5 53 18
Transformation of the population scores according to
Vaginal No 0 105.5 53 0
this formula corresponded with testing recommendations
Urine Yes 3270.5 105.5 53 31
that more closely aligned with guidelines. To further refine Syphilis Yes 1411.8 12 17 117.65
this, we adjusted the weights assigned to each sexual Hep C No 0 0 9 6
practice until appropriate testing was recommended in all
fictitious evaluation cases that were input by our evaluation
team of specialized healthcare professionals who worked
Of note, if this person were to report being HIV-positive,
in our local STI clinic. Indeed, we refined the algorithm by
then HIV testing is removed. See Box 2.
having a team of three physicians, three nurse practi-
tioners, and ten registered nurses input over 400 fictitious
test cases to ensure results corresponded with clinical
guidance documents for STI/HIV testing. After including Box 2.
the exponential multiplication of population risk scores
Test Final Pop. Pop. Question
before multiplying the test score by the population score
Test required score multiplier score score
and slight adjustments in the assigned risk score for
varying sexual practices, the outcome was perfect align- HIV No 28875 15 19 1925
ment between the algorithm’s recommendations and serology
Public Health Agency of Canada and Public Health On- HIV self- No 29550 15 19 1970
tario clinical practice guidelines.27,28 test
Oral Yes 85.455 105.5 53 0.81
Rectal Yes 1899 105.5 53 18
Clinical examples Vaginal No 0 105.5 53 0
Urine Yes 32705 105.5 53 31
To demonstrate the STI/HIV algorithm, we will show Syphilis Yes 1423.8 12 17 118.65
the scoring process for five exemplar cases. These en- Hep C No 0 0 9 11.9
compass a range of screening situations and highlight the
responsiveness of the algorithm to varying practices and
participant characteristics.
Case 2
The second case involves an HIV-negative, white, cis-
Case 1 gendered, 20-year-old female with male partners. She re-
The first case is an HIV-negative, white, cis-gendered, 40- ports condomless vaginal sex and oral sex (receive/
year-old male who engages in oral sex (received/performed) perform). She does not report sex work or injection drug
and anal sex (received/performed) with male partners. He use. She reports that her partners are white and do not use
does not use condoms or PrEP. He was last tested for STIs/ injection drugs either. She was last tested 6 months ago for
HIVabout 6 months ago and had a new sexual partner since. STIs/HIV and has had new sexual partners since. Based on
He is unsure if any of his sexual partners engage in injection the guidelines, this female should have a vaginal gonorrhea/
1370 International Journal of STD & AIDS 32(14)

chlamydia test only. While oral testing could be indicated, Notably, if this same person reports that she was last tested
local guidelines do not recommend this in the absence of less than 3 months ago, but that she is now outside the testing
a “clinical indication.” HIVand syphilis testing are also only window for HIV, the algorithm removes the gonorrhea and
recommended annually for this person. The algorithm cor- chlamydia testing while retaining the HIV testing to rule out
responded exactly with such recommendations. See Box 3. infection at the appropriate time. See Box 5.

Box 5.

Box 3. Test Final Pop. Pop. Question

Test required score multiplier score score
Test Final Pop. Pop. Question
Test required score multiplier score score HIV Yes 92.4 16.5 20 5.6
serology
HIV No 4.2 1.5 6 2.8 HIV self- Yes 44.55 16.5 20 2.7
serology test
HIV self- No 1.35 1.5 6 0.9 Oral No 0 166 0.14
test Rectal No 0 166 0.4
Oral No 8.925 63.75 39 0.14 Vaginal No 0 166 5
Rectal No 229.5 63.75 39 3.6 Urine No 0 166 0
Vaginal Yes 348.7125 63.75 39 5.48 Syphilis No 0 0 61 5.15
Urine No 0 63.75 39 0 Hep C No 0 0 41 3
Syphilis No 12.425 3.5 9 3.55
Hep C No 0 0 11 1

Case 4
Case 3 The fourth case is an HIV-negative, white, 32-year-old
trans-male who has internal genitals. They engage in oral
The third case is an HIV-negative, Black, cis-gendered, 26-
sex (performs only) and anal sex (receptive only) with
year-old heterosexual female with male partners who are
partners who have external genitals. They report injection
also Black. She reports condomless vaginal and anal sex and
drug use and no sex work. They were last tested between 3
performing oral sex. She was last tested 6 months ago with
and 6 months ago and have had new partners since. Ac-
new partners since. She does not report injection drug use
cording to current guidelines,25,26 trans-males are dispro-
for herself or her partners. Per local guidelines, this female
portionately affected by STIs/HIV and warrant comprehensive
should receive a vaginal gonorrhea/chlamydia test. Due to
testing. Based on the identified risk practices, this person
elevated HIV incidence among Black women, serology for
should receive gonorrhea and chlamydia testing (oral and
HIV and a rapid HIV test should also be offered. No in-
rectal), syphilis, HIV, and hepatitis C serology, and an HIV
creased screening is warranted for syphilis. The algorithm
rapid test. The algorithm identified such testing appropri-
yielded these exact recommendations for this test case.
ately. See Box 6.
See Box 4.

Box 4. Box 6.

Test Final Pop. Pop. Question Test Final Pop. Pop. Question
Test required score multiplier score score Test required score multiplier score score

HIV Yes 210 37.5 30 5.6 HIV Yes 295.4 105.5 75 2.8
serology serology
HIV self- Yes 101.25 37.5 30 2.7 HIV self- Yes 189.9 105.5 75 1.8
test test
Oral No 8.925 63.75 39 0.14 Oral Yes 68.575 105.5 142 0.65
Rectal No 25.5 63.75 39 0.4 Rectal Yes 422 105.5 142 4
Vaginal Yes 348.7125 63.75 39 5.47 Vaginal No 0 105.5 142 0
Urine No 0 63.75 39 0 Urine No 0 105.5 142 0
Syphilis No 19.425 3.5 9 5.55 Syphilis Yes 169.2 70.5 41 2.4
Hep C No 0 0 11 3 Hep C Yes 80 40 31 2
O’Byrne et al. 1371

Case 5 to those who are most affected by STIs/HIV. This accuracy

also ensures that persons are offered testing based on
The last case involves an HIV-negative, Indigenous, cis- provincial and federal health guidelines,27,28 which has
gendered 35-year-old female with male partners. She been a barrier to implementation of previous online testing
engages in vaginal and oral sex (receives/performs). She programs in our jurisdiction. While access to testing is
was last tested over 12 months ago and had a new partner important, indiscriminate delivery can also produce un-
6 months ago. She does not report injection drug use for wanted outcomes. For example, testing without appropriate
herself or her partners, nor does she report sex work. Per assessment can exacerbate inequitable healthcare delivery
local guidelines,27,28 this person should receive gonorrhea/ access. In the context of ongoing testing supply shortages,
chlamydia testing (vaginal only), plus serology for HIV providing these resources to members of groups not bur-
and syphilis. Due to elevated HIV and hepatitis C prev- dened by STIs/HIV means that the members of the groups
alence among members of Indigenous populations in with the greatest burden might not have access to testing.
Canada, this person should also be offered a rapid HIV test Another issue is that all testing has performance limitations,
and hepatitis serology. The algorithm yields these out- which are magnified by prevalence. As the positive pre-
comes. See Box 7. dictive value of a test decreases in tandem with prevalence,
test accuracy decreases when persons who are unlikely to
Box 7.
have infections are tested. In addition to issues regarding
Test Final Pop. Pop. Question resources, such inappropriate testing can also generate
Test required score multiplier score score ethical issues related to the consequences of false positive
results, such as requiring treatment or partner follow-up, and
HIV Yes 295.4 105.5 53 2.8 any distress associated with diagnosis. Our automated STI/
serology
HIV risk assessment and screening algorithm helps rectify
HIV self- Yes 94.95 105.5 53 0.9
this situation by appropriately targeting testing.
test
Oral No 14.77 105.5 59 0.14
A second noteworthy point about our algorithm is its
Rectal No 379.8 105.5 59 3.6 malleability. Figures 3 and 4 show that each test and each
Vaginal Yes 577.085 105.5 59 5.47 population is weighted individually, allowing for the al-
Urine No 0 105.5 59 0 gorithm to be tailored based on changing epidemiology
Syphilis Yes 84.3125 23.75 24 3.55 and research. If evidence emerges showing differing es-
Hep C Yes 23.75 23.75 24 1 timates of STI/HIV transmission risk or about new pre-
vention technologies, the algorithm can be easily updated
by re-allocating risk scores; the logic and calculations
within the algorithm, however, remain unchanged. The
Discussion layout we developed for the algorithm similarly allows it to
In response to a higher-than-expected uptake of our HIV function in diverse geographic settings, provided that the
self-test study22 plus profound reductions in access for STI/ population scores are updated to reflect local epidemiol-
HIV testing due to COVID-19, we expanded Getakit.ca to ogy. The cumulative calculations for population scores
include automated clinical decision-making for STI/HIV also allow new risk groups or new risk factors to be added
risk assessments. We believe this is the first of such algo- to score calculations, again without requiring major
rithms that operates through a website and consistently modifications to the automated process. To the best of our
recommends testing with a high degree of sophistication knowledge, no similar automated algorithms exist.
Third, the malleability and accuracy of our algorithm
according to local guidelines27,28 based on participants’
means that it could be used to increase STI/HIV testing
reported risk practices. This algorithm raises a few note-
uptake among those who most commonly avoid testing.
worthy points for discussion.
While we would always consider in-person healthcare to
First, the accuracy our algorithm is notable. Sixteen
be optimal, we know such interactions are problematic for
healthcare professionals from nursing and medicine who
many due to geography, wait times, and concerns about
work in our STI clinic tested the system by inputting over being seen.13–16 We also know that many persons avoid
400 fictitious test cases with different risk practices and STI/HIV testing due to concerns about stigmatization,
profiles and found that the algorithm output matched local and that such concerns increase among persons who are
clinical guidelines27,28 perfectly. The importance of this trans, ACB, Indigenous, gbMSM, etc.17,35–37 In other
finding cannot be overstated. While accuracy ensures that words, accessing STI/HIV testing is often most difficult
people who seek testing can obtain such services, both for the persons who are most affected by these infections.
generally and during pandemics, accuracy of testing dis- As such, a possible utility of our algorithm is that it can
tribution also ensures equitable allocation of finite resources provide care in non-stigmatizing ways using culturally
1372 International Journal of STD & AIDS 32(14)

sensitive, non-judgmental language for persons who his- both most burdened by STIs/HIV and often most vic-
torically have had negative experiences with the health- timized by the healthcare system. Mass roll-out and up-
care system regarding their ethnicities, skin color, sexual take evaluation will determine if this assertion holds true.
orientation, or gender identities. This is a major strength to
this algorithm. Acknowledgments
Another strength of our algorithm is that it ensures that POB would like to thank the Ontario HIV Treatment Network
the members of the groups that are most affected by STIs (OHTN) for his endowed Research Chair in Public Health & HIV
and HIV can obtain full services. While research24–26 has Prevention. LO would like to thank the Canadian Institutes of
shown that some algorithms formalize ethnic biases and Health Research (CIHR) for her Vanier Scholarship.
consequently impede access to care for racialized com-
munities, we designed ours so that the thresholds to Declaration of conflicting interests
qualify for care were more easily surpassed by members of The author(s) declared no potential conflicts of interest with re-
these groups. In opposition to what has been found in spect to the research, authorship, and/or publication of this article.
some previous healthcare algorithms, therefore, we en-
sured easier access to care for minority and racialized Funding
groups; notably, this approach was supported by our The author(s) disclosed receipt of the following financial support
community consultations. for the research, authorship, and/or publication of this article: This
work was supported by the Ontario HIV Treatment Network
(OHTN
Limitations
The development and utility of our STI/HIV screening ORCID iDs
algorithm is not without limitations. Our work was based on Patrick O’Byrne  https://orcid.org/0000-0002-0587-1409
Canadian guidelines only, with a specific focus on STI/HIV Lauren Orser  https://orcid.org/0000-0001-7732-2866
epidemiology in Ottawa. Its applicability more broadly has
not been tested, although the algorithm is sufficiently Notes
nimble to allow for adjustments based on local prevalence
data. Another limitation is that the algorithm requires field 1. The logic for the y-axis is that, as STI/HIV prevalence increases
testing. Trained clinicians with in-depth understanding of within a person’s sexual network, so does the risk of exposure
the subject material completed the validation using fictitious (i.e., of having a sexual partner with an STI or HIV).
patient scenarios. If such a high degree of alignment with 2. For our initial HIV self-test study, we can intended to test
guidelines will occur when the system is used by the lay a minimum of 150 participants in the first 6 months of im-
public is yet to be determined. A reassuring finding though plementation. By 7 months, we had distributed 444 test kits.
is that the clinicians who tested the system were not trained
in how to register, navigate, or complete the self-assessment References
and test ordering, and all were able to complete the process. 1. Public Health Agency of Canada [PHAC]. Update on
Sexually Transmitted Infections. 2019. Retrieved from
https://www.canada.ca/content/dam/phac-aspc/documents/
services/publications/diseases-conditions/update-sti/STI_
Conclusion
update_2016_Jan_10-EN.PDF.
In this article, we presented an STI/HIV risk assessment 2. Haddad N, Weeks A, Robert A, et al. HIV in Canada –
algorithm that we developed in Ottawa, Canada and showed surveillance report, 2019. Can Commun Dis Rep 2021; 47(1):
the logic we automated to stratify participants’ STI/HIV 77–86.
risk to ensure they were offered appropriate testing, in 3. Choudhri Y, Miller J, Sandhu J, et al. Chlamydia in Canada,
accordance with local guidelines. Our robust pilot testing 2010-2015. Can Commun Dis Rep 2018; 44(2): 49–54.
with over 400 test patient cases showed the accuracy of our 4. Choudhri Y, Miller J, Sandhu J, et al. Gonorrhea in Canada,
2010-2015. Can Commun Dis Rep 2018; 44(2): 37–42.
algorithm to recommend the same testing that our 16 expert
5. Friedman DS and O’Byrne P. Extragenital testing increases
STI/HIV clinicians would have offered. We believe this is
case detection of gonorrhea and chlamydia: the impact of
the first of such algorithms to exist and posit that a major implementing nucleic acid amplification testing. Can Com-
strength of our system is its ease of modification based on mun Dis Rep 2020; 46(9): 285–291.
changing epidemiology and scientific evidence about STI/ 6. Choudhri Y, Miller J, Sandhu J, et al. Infectious and congenital
HIV transmission and prevention. While we believe that syphilis in Canada, 2010-2015. Can Commun Dis Rep 2018;
in-person testing is ideal, we think this automated online 44(2): 43–48.
system might overcome some barriers to STI/HIV testing, 7. PHAC. Syphilis in women and congenital syphilis in Canada,
especially for minority and marginalized persons who are 2019. Can Commun Dis Rep 2020; 46(10): 366.
O’Byrne et al. 1373

8. PHAC. Canadian guidelines on sexually transmitted infections: 23. Fairley CK, Sze JK, Vodstrcil LA, et al. Computer-assisted self
Primary care and sexually transmitted infections. 2013. Re- interviewing in sexual health clinics. Sex Transm Dis 2010;
trieved from: https://www.canada.ca/en/public-health/services/ 37(11): 665–668. doi:10.1097/OLQ.0b013e3181f7d505.
infectious-diseases/sexual-health-sexually-transmitted- 24. Gasmelsid N, Moran BC, Nadarzynski T, et al. Does online
infections/canadian-guidelines/sexually-transmitted-infections/ sexually transmitted infection screening compromise care?
canadian-guidelines-sexually-transmitted-infections-17.html. A service evaluation comparing the management of chla-
9. Taylor MM, Frasure-Williams J, Burnett P, et al. Interventions mydial infection diagnosed online and in the clinic. Int J
to improve sexually transmitted disease screening in clinic- STD AIDS 2021; 32(6): 528–532. doi:10.1177/09564624
based settings. Sex Transm Dis 2016; 43(2Suppl 1): S28–S41. 20980929.
10. Steen R, Wi TE, Kamali A, et al. Control of sexually trans- 25. Wiens J, Price WN and Sjoding MW. Diagnosing bias in data-
mitted infections and prevention of HIV transmission: driven algorithms for healthcare. Nat Med 2020; 26(1): 25–26.
mending a fractured paradigm. Bull World Health Organ 26. Challen R., Denny J., Pitt M., et al. Artificial intelligence, bias
2009; 87(11): 858–865. and clinical safety. BMJ Qual Saf 2019; 28(3): 231–237.
11. Eisinger RW, Dieffenbach CW and Fauci AS. HIV viral load 27. PHAC. Canadian guidelines on sexually transmitted in-
and transmissibility of HIV infection: undetectable equals fections. 2020. Retrieved from https://www.canada.ca/en/
untransmittable. JAMA 2019; 321(5): 451–452. doi:10.1001/ public-health/services/infectious-diseases/sexual-health-
jama.2018.21167. sexually-transmitted-infections/canadian-guidelines/sexually-
12. Myers JE, Braunstein SL, Xia Q, et al. Redefining prevention transmitted-infections.html.
and care: a status-neutral approach to HIV. Open Forum Infect 28. Public Health Ontario. Ontario Gonorrhea Testing and
Dis 2018; 5(6): 1–4. Treatment Guide. 2nd Edition. 2018. Retrieved from https://
13. Traversy GP, Austin T, Ha S, et al. An overview of recent www.publichealthontario.ca/-/media/documents/s/2018/
evidence on barriers and facilitators to HIV testing. Can summary-gonorrhea-testing-treatment.pdf?la=en.
Commun Dis Rep 2015; 41(12): 304–321. 29. Centre for Disease Control and Prevention (CDC). HIV risk
14. Myerson B, Barnes P, Emetu R, et al. Institutional and reduction tool: Estimate the HIV risk. 2019. Retrieved from
structural barriers to HIV testing: elements for a theoretical https://hivrisk.cdc.gov/risk-estimator-tool/.
framework. AIDS Pt Care 2014; 8(1): 22–27. doi:10.1089/ 30. Melbourne Sexual Health Centre. Check your risk. 2017.
apc.2013.0238. Retrieved from https://checkyourrisk.org.au/.
15. Wong JPH, Chan KBK, Boi-Doku R, et al. Risk discourse and 31. O’Byrne P, MacPherson P, Roy M, et al. Community-based
sexual stigma: barriers to STI testing, treatment, and care nurse-led post-exposure prophylaxis: results and implications.
among young heterosexual women in disadvantaged neigh- Int J Std/aids 2017; 28(5): 505–511. doi:10.1177/
bourhoods in Toronto. Can J Hum Sex 2012; 21(2): 75–89. 0956462416658412.
16. Brooks H., Llewellyn C. D., Nadarzynski T., et al. Sexual 32. Galvin SR and Cohen MS. The role of sexually transmitted
orientation disclosure in health care: a systematic review. Br diseases in HIV transmission. Nat Rev Microbiol 2004; 2(1):
J Gen Pract 2018; 68(668): e187–e196. doi:10.3399/ 33–42. doi:10.1038/nrmicro794.
bjgp18X694841. 33. Tan DHS, Hull MW, Yoong D, et al. Canadian guideline on
17. Scheim AI and Travers R. Barriers and facilitators to HIVand HIV pre-exposure prophylaxis and nonoccupational post-
sexually transmitted infections testing for gay, bisexual, exposure prophylaxis. CMAJ 2017; 189(47): E1448–E1458.
and other transgender men who have sex with men. AIDS doi:10.1503/cmaj.170494.
Care 2017; 29(8): 990–995. doi:10.1080/09540121.2016. 34. Stoley JE and Cohen SE. Syphilis transmission: a review of
1271937. the current evidence. Sex Health 2015; 12(2): 103–109. doi:
18. Cassell J. A., Brook MG, Mercer CH, et al. Treating sexually 10.1071/SH14174.
transmitted infections in primary care: a missed opportunity? 35. Fisher CB, Fried AL, Macapagal K, et al. Patient-provider
Sex Transm Infec 2003; 79: 134–136. communication barriers and facilitators to HIV and STI
19. Marcus JL, Bernstein KT, Kohn RP, et al. Infections missed by preventive services for adolescent MSM. AIDS Behav 2018;
urethral-only screening for chlamydia or gonorrhea detection 22: 3417–3428, doi:10.1007/s10461-018-2081-x.
among men who have sex with men. Sex Transm Dis 2011; 36. Djiadeu P, Nguemo J, Mukandoli C, et al. Barriers to HIV care
38(10): 922–924. doi:10.1097/OLQ.0b013e31822a2b2e. among Francophone African, Caribbean and Black immigrant
20. Harling G, Subramanian S, Bärnighausen T, et al. Socio- people living with HIV in Canada: a protocol for a scoping
economic disparities in sexually transmitted infections among systematic review. BMJ Open 2019; 9: e027440. doi:10.1136/
young adults in the United States: examining the interaction bmjopen-2018-027440.
between income and race/ethnicity. Sex Transm Dis 2013; 37. Harb CYW, Pass LE, De Soriano IC, et al. Motivators and
40(7): 575–581. doi:10.1097/OLQ.0b013e31829529cf. barriers to accessing sexual health care services for transgender/
21. O’Byrne P, Musten A, Orser L, et al. At-hone HIV self-testing genderqueer individuals assigned female sex at birth. Trans-
during COVID: implementing the GetaKit project in Ottawa. gender Health 2019; 4(1): 58–67. doi:10.1089/trgh.2018.0022.
Can J Public Health 2021; 112: 587–594. 38. O’Byrne P, Orser L, Jacob JD, Bourgault A and Lee SR Re-
22. Richens J, Copas A, Sadiq ST, et al. A randomised controlled sponding to critiques of the Canadian PrEP guidelines: In-
trial of computer-assisted interviewing in sexual health clinics. creasing equitable access through a nurse-led active-offer PrEP
Sex Transm Infections 2010; 86: 310–314. doi:10.1136/sti. service (PrEP-RN). Canadian Journal of Human Sexuality
2010.043422. 2019; 28(1): 5–16.