Keywords electronic health records and systems - clinical decision support systems - human
immunodeficiency virus - point-of-care diagnostic testing - dental clinic
Background and Significance
Background and Significance
At the conclusion of 2022, the Centers for Disease Control and Prevention (CDC) estimated
that 1.2 million individuals in the United States were living with HIV of which 156,000
remained unaware of their status.[1 ] Notably, 80% of new HIV infections are thought to stem from individuals who are
unaware of their HIV-positive status or are not receiving care.[1 ]
[2 ]
To reach the U.S.' Ending the HIV Epidemic (EHE) goals, according to The Joint United
Nations Program on HIV/AIDS's (UNAIDS) fast-track targets, the health system must
diagnose 95% of people with HIV as early as possible, provide immediate initiation
or reinitiation of antiretroviral therapy in 95% of patients diagnosed and sustain
antiretroviral therapy for HIV viral load suppression in 95% of those on treatment
by 2030.[3 ]
The standard of care is to offer initial HIV testing for individuals aged 13 and over
and periodic retesting for those with HIV-related risk behaviors, including unprotected
sex.[4 ] HIV and sexually transmitted infections (STI) such as syphilis, gonorrhea, and chlamydia
are linked epidemiologically and by synergistic transmission.[5 ]
[6 ] Interventions like pre-exposure prophylaxis (PrEP) for HIV prevention, immediate
antiretroviral therapy initiation, and programs to promote retention in care are recommended.[7 ]
[8 ]
It is important to note that individuals diagnosed with HIV can lead healthy lives
with antiretroviral therapy.[9 ] Additionally, those at high risk for HIV who test negative can reduce their risk
of sexual transmission by 99% through the use of these highly effective antiretroviral
medications when taken as prescribed.[9 ] In 2022, 36% of eligible individuals in the United States were prescribed PrEP;
therefore, increasing this coverage is a crucial strategy in the EHE initiative.[10 ]
People with HIV are particularly vulnerable to oral health issues. Common problems
include chronic dry mouth, gum disease (gingivitis), bone loss around the teeth (periodontitis),
canker sores, oral warts, fever blisters, thrush (oral candidiasis), hairy leukoplakia
(which causes rough, white patches on the tongue), and tooth decay.[11 ]
Expanding the workforce to health care providers within the dental clinics presents
an untapped opportunity to EHE, especially since it is a site of contact for patients
who may not be engaged in medical care.[2 ]
[4 ]
[12 ]
[13 ] Although HIV testing is well accepted in the dental clinic, there is concern regarding
documented barriers to HIV testing by dental care providers.[14 ]
[15 ]
[16 ]
[17 ] These barriers include patient acceptability, gaps in knowledge needed to provide
HIV testing and posttest counseling, time constraints during clinical encounters,
resources, and sustainability.[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ] Additionally, dentists may not feel comfortable addressing issues related to HIV
risk and/or sexuality with patients. They require training on how to inform their
patients about their HIV status and assist in establishing linkages with HIV-related
medical services.[16 ]
[21 ]
[22 ]
[23 ]
Approximately 70% of individuals who have never been tested for HIV have had contact
with a dental provider.[15 ] In fact, a significant subset of patients who have not seen a general health care
provider within a 2-year period have instead seen a dental provider.[24 ] The use of the electronic health record (EHR), Epic, has expanded into 1,800 dental
practices and eight dental schools, reflecting the growing integration of medical
and dental records. This unified platform allows clinical professionals to share data,
enhancing treatment and public health initiatives.[25 ] Columbia University's College of Dental Medicine (CDM) is an integral part of Columbia
University Irving Medical Center (CUIMC), located in New York City. This consortium
operates on Epic, encompassing a diverse pool of over 4 million individual patient
records. Throughout the fiscal years 2022 and 2023, the CDM dental clinics served
over 60,000 unique patients. An analysis at CDM reveals that 63% of patients seen
in the dental clinics have not utilized medical services within our consortium. Moreover,
less than 3% of these patients had a documented HIV test in the EHR.
EHR reminders for HIV testing have demonstrated improvement in screening and/or testing
for HIV.[26 ]
[27 ]
[28 ] Furthermore, clinical decision support systems (CDSS) have been shown to increase
HIV testing and can be programmed to prompt retesting if an individual has ongoing
HIV risk factors.[29 ]
[30 ] Like all CDSS tools, their effectiveness can be enhanced by incorporating the “five
rights” for effective decision support: the right information to the right people,
through the right channels, in the right intervention formats, and at the right points
in the workflow.[31 ]
[32 ]
In this paper, we will explain how our organization leveraged an untapped resource—dental
clinics and dental providers—to expand HIV testing and developed a CDSS to accurately
identify and target patients who would benefit most from HIV testing.
Objectives
The Division of Infectious Diseases at CUIMC initiated a collaborative study with
CDM to contribute to the broader goal of EHE. This study aims to develop a CDSS, referred
to as a BestPractice Advisory within the EHR, designed to accurately pinpoint patients
who would benefit from HIV testing. Additionally, we intend to introduce a streamlined
and intuitive workflow for health care providers, facilitating seamless ordering,
and execution of HIV point-of-care (POC) screening.
Methods
Building the Clinical Decision Support System
The first objective, developing the CDSS, involved identifying the target population
for HIV testing. In this study, we included patients who are 18 years old or older
with either (1) no history of HIV testing ever documented in the EHR or (2) a negative
HIV test and a positive STI test within the past 2 years. While age is a simple data
field, determining which HIV and STI tests to evaluate as well as their positive or
negative status required significant clinical input, data analysis, data cleansing,
and formulaic categorization of laboratory results.
The infectious diseases clinicians identified three STI's—chlamydia, gonorrhea, and
syphilis—to consider. Subsequently, they determined which particular HIV and STI tests
and their associated Logical Observation Identifiers Names and Codes (LOINC) to include
in the evaluation ([Table 1 ]).
Table 1
Reference table for HIV and sexually transmitted infection Logical Observation Identifiers
Names and Codes utilized within the clinical decision support system development
LOINC
Name
56888-1
HIV 1 + 2 Ab + HIV1 p24 Ag [presence] in serum or plasma by immunoassay
80387-4
HIV 1 + 2 Ab [presence] in serum, plasma, or blood by rapid immunoassay
75666-8
HIV 1 + 2 Ab and HIV1 p24 Ag [identifier] in serum, plasma, or blood by rapid immunoassay
29893-5
HIV 1 Ab [presence] in serum or plasma by immunoassay
30361-0
HIV 2 Ab [presence] in serum or plasma by immunoassay
68961-2
HIV 1 Ab [presence] in serum, plasma, or blood by rapid immunoassay
81641-3
HIV 2 Ab [presence] in serum, plasma, or blood by rapid immunoassay
31201-7
HIV 1 + 2 Ab [presence] in serum or plasma by immunoassay
48345-3
HIV 1 + O + 2 Ab [presence] in serum or plasma
7918-6
Cells.CD3 + CD4 + CD8+ (double-positive)/100 cells in blood
80203-3
HIV 1 and 2 Ab [identifier] in serum, plasma, or blood by rapid immunoassay
80387-4
HIV 1 + 2 Ab [presence] in serum, plasma, or blood by rapid immunoassay
20447-9
HIV 1 RNA [#/volume] (viral load) in serum or plasma by NAA with probe detection
25835-0
HIV 1 RNA [presence] in serum or plasma by NAA with probe detection
44871-2
HIV 1 proviral DNA [presence] in blood by NAA with probe detection
5017-9
HIV 1 RNA [presence] in blood by NAA with probe detection
70241-5
HIV 1 RNA [#/volume] (viral load) in plasma by probe and target amplification method
detection limit = 20 copies/mL
35437-3
HIV 1 Ab [presence] in saliva (oral fluid) by immunoassay
24111-7
Neisseria gonorrhoeae DNA [presence] in specimen by NAA with probe detection
21613-5
Chlamydia trachomatis DNA [presence] in specimen by NAA with probe detection
88718-2
Chlamydophila pneumoniae DNA [presence] in nasopharynx by NAA with probe detection
43304-5
Chlamydia trachomatis rRNA [presence] in specimen by NAA with probe detection
6357-8
Chlamydia trachomatis DNA [presence] in urine by NAA with probe detection
57288-3
Chlamydia trachomatis rRNA [presence] in nasopharynx by NAA with probe detection
50387-0
Chlamydia trachomatis rRNA [presence] in cervix by NAA with probe detection
11084-1
Reagin Ab [titer] in serum
31147-2
Reagin Ab [titer] in serum by RPR
Abbreviations: LOINC, Logical Observation Identifiers Names and Codes; STI, sexually
transmitted infection.
Normalizing data are a crucial step in data preprocessing and analysis, ensuring that
all features in a dataset have the same scale. This involves both manual and programmatic
cleansing of the data.[33 ]
[34 ] Upon querying the EHR database for results pertaining to these tests, we noticed
that actual results were not simply “positive” or “negative.” Some were stored as
discrete values; however, most were free-text, some with spelling errors and others
representing ambiguous interpretations. The dataset was extracted, formatted, and
sorted by HIV and STI tests, from the most to the least common, to facilitate detailed
review. The clinicians carefully reviewed and categorized the results as positive
or negative. The next step involved developing logic that would consistently label
such results in the same way. For example, a result with a value of “react” was determined
to be positive; this was shorthand for “reactive,” indicating a positive result. However,
the opposite of a reactive value is nonreactive, indicating a negative result. To
programmatically identify a positive result, logic needed to accommodate other types
of common result values or misspellings. For this reason, string-searching values
for “reactive” and assuming the test was positive would not be correct most of the
time—we would miss potential values “react” or incorrectly flag “nonreactive” as positive.
The team went through many iterations, and any result with text displaying “neg” (considered
to be “negative”) or “non” (considered to be “nonreactive”) or “nr” (also considered
to be “nonreactive”) would be negative. On the contrary, any result with text displaying
“pos” (considered to be “positive”), “react” but not containing “no” (considered to
be “reactive”) or entered as detectable numerical values would be positive ([Fig. 1 ]).
Fig. 1 This figure reveals a small sample of the laboratory values within the electronic
health record and the iterative work of interpreting the results with several combinations
of logic.
Once finalized, we translated the logic into “rules,” which represents “if-then” conditional
statements within the EHR's CDSS framework. Using “rules”-based programming, we ensured
that the CDSS would be able to identify our target population and any exclusions;
determined when and where the CDSS would fire; and configured the end user display,
the permissible actions, and the recommended follow through ([Fig. 2 ]).
Fig. 2 This figure shows the front-facing popup alert from the clinical decision support
system within the Epic electronic health record.
Developing the Workflow
The second objective in the study was to develop and implement a HIV POC testing workflow.
We selected CDM's Advanced Education General Dentistry (AEGD) residency program to
implement the workflow since these residents have already completed dental school
and have experience treating a diverse patient population.
Collaborating with AEGD faculty, we determined operationally how best to design the
workflow ([Fig. 3 ]). The workflow in the AEGD clinic begins when the dental resident opens the patient's
chart and the CDSS signals that the patient is a candidate for HIV testing. The resident
offers HIV testing, and if the patient agrees, a preconfigured order set including
the laboratory order and visit diagnosis, screening for HIV, is presented. If the
test is declined, predefined reasons for not testing will display with subsequent
delays in firing again depending on the selection.
Fig. 3 Diagram depicting end-to-end clinical workflow after clinical decision support system
and HIV testing is implemented.
The test utilized was the OraQuick Rapid Antibody Test Advanced HIV-1/2. It was chosen
due to its ease of use, and although a comparison of rapid POC tests found that sensitivity
of oral tests was slightly lower (98.03%) than blood based specimens (99.68%), specificity
was similar (99.74% oral vs. 99.91% blood).[35 ] As part of the testing workflow, a specimen is collected via a cheek swab, and results
are available before the visit's completion. The resident would then document the
results in the patient's chart and inform the patient of the results, along with information
and education on PrEP. If a positive result arises, a warm handoff occurs, during
which the patient is escorted to the HIV clinic for confirmatory testing and further
education ([Fig. 3 ]).
To help dental residents effectively implement the proposed workflow, they were provided
with trainings, presentations, tip sheets, and videos that offered detailed, step-by-step
instructions. For patient interactions, infectious diseases physicians coached the
residents, conducted role-playing scenarios, and educated them on how to offer tests,
communicate results, and perform warm handoffs. For EHR instructions, realistic training
patients were created in a simulation environment so residents could see the CDSS
fire and practice the proposed workflow, including ordering tests and documenting
results. Finally, a patient navigator (research assistant) who was well-versed in
the workflow and sensitive communication methods assisted residents during the initial
go-live.
Assessing the Objectives
The period of when the CDSS was functional and the associated workflow was fully implemented
was from September 2022 to June 2023. We planned to assess the effectiveness of the
CDSS and workflow in phases:
Phase 1: a ramp-up period when the study team was available to smooth out issues,
answer questions, and enforce workflow integrity (8 weeks).
Phase 2: a period when the patient navigator was available to assist the residents
(12 weeks).
Phase 3: a period when the patient navigator was no longer available (12 weeks).
While reporting on CDSS firing rates and HIV testing rates by residents helped assess
uptake, staying in constant communication with the residents and faculty of the AEGD
clinic was critical. The provider buy-in was vital in ensuring smooth execution. Periodic
Q&A sessions were held to both help residents and obtain qualitative feedback on the
CDSS and workflow.
Results
Once the CDSS was activated, during Phase 1 (ramp-up period) of implementation, of
the 1,613 dental residents' patient visits, 956 (or 59%) prompted the CDSS to alert
the provider about the potential need for HIV testing, with a testing rate of 3.1%
(n = 30; [Table 2 ]). This phase was characterized by dental providers acclimating to the new workflow,
fine-tuning their time management strategies, gaining confidence in discussing the
sensitive topic of HIV testing with patients, and familiarizing themselves with inventory
and materials.
Table 2
This table details the clinical decision support system data and testing effectiveness
in each of the three phases of the implementation
Time period
Ramp up (8 wk)
With navigator (12 wk)
Without navigator (12 wk)
Visits
1,613
2,231
2,595
CDSS alerts
956
893
838
% of CDSS alerts/visits
59%
40%
32%
HIV tests
30
113
96
% of HIV tests/CDSS alerts
3.1%
12.7%
11.5%
Abbreviation: CDSS, clinical decision support system.
During Phase 2, introduction of a patient navigator whose duties included facilitating
interactions with patients on the importance of HIV testing and preventions modalities
like PrEP had a remarkable impact, leading to a significant increase in HIV testing
rates, reaching 12.7% (n = 113; [Table 2 ]). The presence of the patient navigator played a pivotal role in facilitating and
enhancing the testing process. What is particularly encouraging is even after the
patient navigator's support was no longer available, the HIV testing rate was 11.5%
(n = 96; [Table 2 ]).
Throughout Phases 2 and 3, the percentage of times the CDSS fired decreased to 40%
(893 alerts/2,231 visits) and 32% (838 alerts/2,595 visits), respectively ([Table 2 ]). Once the CDSS fired, if its suggestion for HIV testing is not accepted, there
is a delay in subsequent firing as follows: 1 day for patient barriers, 7 days for
visit barriers, and 6 months for patient refused. Since a significant number of patients
return to clinics for multiple appointments, the proportion of CDSS alerts decreased.
Overall, 18% of the CDSS alert declinations were due to patient barriers, 52% were
due to visit barriers, and 30% were due to patient refusal ([Table 3 ]).
Table 3
This table details the reasons that dental residents noted for declining the clinical
decision support system when it fired an alert
Reasons for declination
Count
% Reason/Total
Patient barriers
387
18
Visit barriers
1,153
52
Patient refused
655
30
Total declinations
2,195
100
Discussion
The strategy to EHE is to increase testing and provide patients with information regarding
PrEP. With an initial 59% CDSS alert rate in our AEGD residency program, this validates
the premise for expanding HIV testing. Normalizing HIV POC testing in the dental ambulatory
care setting is a key aspect of this program and follows the guidelines recommended
by the American Dental Association, CDC, and NYSDOH for HIV testing.[1 ]
[36 ]
[37 ]
Even without the patient navigator in Phase 3, we observed that HIV testing maintained
a consistent level of performance, underscoring the sustained effectiveness of the
system in promoting and facilitating testing even in the absence of additional assistance.
Our testing rate of approximately 12% is encouraging since a National Hospital Ambulatory
Medical Care Survey sampling emergency department visit in the United States in 2018
revealed a testing rate of 1.05%.[38 ] Additionally, in a small study that did not utilize CDSS's, but instead offered
HIV testing based on patient responses to a questionnaire, a testing rate of 8.2%
(21/256) was observed.[18 ]
HIV testing was a new process for residents and the supervising faculty, and a few
concerns were identified early on; however, through collaborative efforts, we made
several modifications, improvements, and iterations in workflows, policies, and the
CDSS. For example, a dental assistant lacking the required training would open the
patient chart, potentially dismissing the CDSS. Modifying the triggers based on provider
role, subsequently prevented this occurrence as it was restricted to alert only residents.
With a more efficient protocol, we would expect improved testing rates in further
implementations of this CDSS and workflow at other partner dental sites. We identified
several limitations of this approach, including patient reluctance to discuss HIV
testing in a dental setting, which may lead to lower acceptance rates, as well as
challenges in integrating testing into existing workflows, potentially disrupting
routine operations and requiring additional time and resources. This was evidenced
by the relatively higher percentages of CDSS declinations due to patient refusals
and visit barriers, respectively.
Although we preemptively attempted to mitigate anticipated barriers such as resource/time
constraints and knowledge gaps, several additional barriers were identified through
over-the-shoulder observations, resident Q&A sessions, faculty feedback, and patient
navigator input. These included language barriers, availability and location of test
kits, adherence to the manufacturer's testing instructions, and faculty supervision.
Remediating such barriers would alleviate concerns in the future as well. Finally,
promoting HIV testing in dental setting, providing extra training, and coaching residents
would further enhance the overall experience.
In reflecting on the aforementioned lessons learned, we realized the complexity involved
with introducing HIV testing in the dental setting. These ranged from technical expertise
to staffing resources and comprehensive training. Resolving barriers and finalizing
CDSS and workflow modifications during this study helped us develop a more detailed
and thorough template for future implementations. This is especially important since
a new cohort of dental residents matriculate every year; therefore, this detailed
documented protocol will become even more valuable.
Conclusion
An integrated EHR like Epic allows dental providers to access medical records, including
diagnoses and key laboratory values, which can help determine the need for guideline-supported
HIV testing. EHR data and CDSS tools can effectively facilitate HIV testing by dental
providers and contribute to national HIV EHE initiatives. Although our study was limited
to a residency program, the dental setting provides an opportunity to expand testing
to patients who would otherwise not receive this service.
Clinical Relevance Statement
Clinical Relevance Statement
The implementation of HIV testing in the dental ambulatory care setting is significant
in that this protocol demonstrates the feasibility in expanding the workforce within
the health care ecosystem. As primary care networks continue to expand, seizing upon
integrated electronic health records and collaboration among health care providers
unveils a promising avenue for elevating patient care and health outcomes. Embedding
clinical decision support tools into workflows not only facilitates the identification
of standards of care but also enables the implementation of public health initiatives
within multidisciplinary health care organizations, thus elevating health care standards.
Multiple Choice Questions
Multiple Choice Questions
Which of the below were incorporated as part of the criteria for identifying patients
that would benefit from HIV screening?
Correct Answer: The correct answer is option c. The overarching logic was either “no history of HIV
testing” or a combination of a “negative HIV test” plus either a “positive gonorrhea/chlamydia/syphilis
test,” as long as the patient is “18 years old or older.”
Which of the below was a step involved in developing the CDS?
Identifying qualifying diagnoses codes
Utilizing artificial general intelligence
Establishing positive and negative laboratory value criteria
Disregarding laboratory values in the decision-making process
Correct Answer: The correct answer is option c. The main steps involved identifying qualifying LOINC
codes, establishing the correct criteria for interpreting positive and negative laboratory
values, and ultimately piecing everything together via rule-based programming.
