Abstract
To the Editor:
Regular engagement in HIV care is crucial for the health of people with HIV (PWH). PWH who are retained in care have improved health outcomes and are less likely to transmit HIV to others. 1 However, in the United States, only 55% of PWH are retained in care. 2
Various strategies exist to improve engagement in care among PWH. These range from simple appointment reminders to intensive case management. However, not all PWH require these strategies. Recent efforts have focused on preemptive identification of PWH at risk for disengagement from care, so that resources can be directed to prevent PWH from falling out of care. Researchers have developed models to predict PWH’s risk for future disengagement from care.3–5
While these methods have utilized both electronic medical record (EMR) data and patient-reported data to predict retention in care, they have not included data on provider intuition regarding PWH’s risk for falling out of care. Providers may have specialized knowledge of their patients and their unique barriers to appointment attendance, which may not be documented in the EMR or reported by patients. Therefore, we sought to assess the accuracy of provider intuition regarding retention in HIV care.
Between January 1, 2024, and March 31, 2024, we surveyed HIV care providers and staff at the University of Chicago Medicine Adult Ryan White HIV care clinic. Individuals were eligible to complete the survey if they were a physician, advanced practice provider (APP), or social worker. Informed consent was obtained, and participants were provided with an EMR-derived list of patients with HIV whom they had seen in the clinic in the prior year (for physicians and APPs) or a random list of 95 patients seen in the prior year (for social workers). They were then asked to answer questions for each patient regarding the likelihood of being retained in care in the next 12 months, using a 6-point Likert scale (very unlikely, somewhat unlikely, neutral, somewhat likely, very likely, don’t know). In the analysis, patients listed as “don’t know” were excluded. We defined lapses in care using the 12-month gap definition (i.e., patients were retained if they attended at least one visit at the HIV care clinic in the following 365 days and had lapsed in care if not). Participants could access the EMR as needed to assess each patient and skip patients with whom they were unfamiliar.
Participants were also asked to report specific barriers and facilitators to remaining retained in care for each of their patients. Free-text responses regarding barriers and facilitators were categorized into thematic groups. Missing data and “none” responses were excluded from the thematic analysis, as well as responses that indicated providers were unfamiliar with a patient or no longer provided care to a patient.
After the surveys were completed, patient-identifying information was removed, and survey data were matched to EMR data by an honest broker. De-identified EMR data regarding HIV appointment attendance were collected prospectively for 12 months to identify the outcome of lapsing in care.
For patients with more than one survey completed by multiple providers, scores were averaged for provider intuition about lapsing in care. Descriptive statistics were used to assess the distribution of patient and provider demographics. We created a univariate logistic regression model to assess the accuracy of provider intuition regarding the prospective outcome of lapsing in care. All data analyses were performed in R (version 4.3.0; R Core Team). This study was approved by the University of Chicago Institutional Review Board (IRB23-1172).
Ten providers completed 337 surveys regarding 294 unique patients. Of the providers, 7 (70.0%) were attending physicians; 1 (10.0%) was a fellow; 1 (10.0%) was an APP; and 1 (10.0%) was a social worker. They were 60.0% (6/10) female. In terms of race, 80.0% (8/10) were white, 10.0% (1/10) Asian, and 10.0% (1/10) Multiracial. Median age was 39 years (IQR: 37.2–42.5). The number of patients assessed by each provider ranged from 3 to 58 (median of 35.5).
Table 1 shows the demographic characteristics of patients. Among the 294 patients, 63 (21.4%) lapsed in care. Among the 337 surveys, 44 (13.1%) rated the likelihood of lapsing as very likely, 37 (11.0%) as likely, 39 (11.6%) as neutral, 91 (27.0%) as unlikely, 119 (35.3%) as very unlikely, and 7 (2.1%) as unknown.
Patient Demographics
Provider intuition was predictive of lapsing in care (Table 2), with an Area Under the Curve (AUC) of 0.67 (95% CI: 0.60–0.74). Notably, patients whom providers rated as likely to lapse in care had 6.27 times the odds of lapsing than those they deemed very unlikely to lapse. Patients rated as very likely to lapse in care had 5.22 times the odds of lapsing than those deemed very unlikely to lapse.
Odds of Lapsing in Care Based on Provider Intuition Survey Responses
OR, odds ratio; CI, confidence interval.
Among free-text responses, the most commonly identified factors contributing to lapses in care included prior history of lapsing in care, unstable life circumstances (e.g., a lack of housing or transportation), and mental health issues. The most commonly cited protective factors included a history of excellent adherence, clinic support staff, and connection with the provider.
We found that providers’ intuition was predictive of PWH lapsing in care. To our knowledge, this is the first study to examine providers’ intuition regarding retention in care among PWH. Several studies have examined physicians’ ability to predict and estimate patient adherence to antiretroviral therapy (ART). In general, providers have low accuracy in predicting ART adherence. For example, one study found that providers’ predictions of ART adherence were no different than chance. 6 Another study found that physicians’ predictions were less accurate for predicting future ART adherence compared with those of their patients. 7
The factors that providers identified as barriers or facilitators to lapsing in care were similar to prior literature. Other studies have shown that history of poor retention, social drivers of health, and mental health issues contribute to lapses in care among PWH.8,9 History of excellent adherence, both to ART and to appointments, was frequently cited as a protective factor. ART adherence and retention in care are closely correlated, 10 so it follows that patients with a history of prior excellent ART use are more likely to attend future HIV care appointments. The relationship between the patient and the provider, as well as the connection with other clinic staff, including social workers and case managers, has been observed as an important protective factor for retention in care in other studies as well.11,12
Our study has several limitations. It was conducted among a relatively small sample of providers at a Ryan White clinic within an academic medical center. Findings may not be generalizable to other settings. Some patients may have been inaccurately classified as lapsing in care when they transferred care to a different clinic. If this were the case, provider predictions of retention may have appeared less accurate than in reality. Finally, provider predictions may be subject to biases stemming from racial, economic, or educational disparities between providers and patients. 6
In conclusion, we found that providers’ intuition was prospectively associated with their patients’ future retention in HIV care. These results should not be interpreted as justification for accepting provider perceptions as an inevitability. Instead, they reinforce the need to proactively allocate additional resources and support to those patients highlighted by provider intuition as being at higher risk for disengagement from HIV care. 13 Future studies that develop predictive models for retention in HIV care may benefit from including providers’ assessments of patients’ future risk for lapsing in care and corroborating barriers and facilitators to retention in care with patients.
Authors’ Contributions
J.P.R. and A.M. conceived of the study, obtained funding, and supervised the study. E.E.F., J.A.M., and S.A.D. analyzed and interpreted the data. J.S. recruited participants and interpreted the data. J.P.R. drafted the article. All the authors revised the article, as well as they read and approved the final article.
Footnotes
Author Disclosure Statement
The authors declare that they have no competing interests.
Funding Information
This work was funded by the National Institute of Mental Health, award R21MH134756.