Characteristics and prognostic factors of HIV-positive adults in the emergency department of a university hospital in South Korea, 2013–2022

2.1 Study design

This single-center retrospective study investigated characteristics of HIV-positive adult patients presenting to the ED of an 895-bed university hospital between 01 January 2013 and 31 December 2022. Per annual ED census data, the hospital serves approximately 60,000 patients annually. Referrals are unnecessary for ED access in South Korea, with governmental subsidization of all medical expenses for HIV care upon registration at a public health center.

Visits were categorized into direct hospital admissions from the ED or discharges to home to evaluate admission-associated factors. Additionally, visits in the poor prognosis group were compared with the remainder to evaluate prognostic indicators. The poor prognosis group was defined as admitted visits with intensive care unit (ICU) placement, visits with in-hospital mortality, or both.

2.2 Study population

Records were retrospectively reviewed for patients aged ≥18 years diagnosed with HIV infection who visited the ED between January 2013 and December 2022. Exclusion criteria encompassed the unawareness of positive HIV serostatus upon initial ED presentation, death prior to arrival, discharge against medical advice, and minor procedural indications including lumbar punctures, intravenous albumin-paired paracenteses, antibiotic injections, or wound dressing.

2.3 Data collection and definitions

The following variables were extracted from medical records of HIV-positive ED patients: age, sex, Korean Triage and Acuity Scale (KTAS) level, ambulance transport, reason for visit, comorbidities, duration of HIV diagnosis and highly active ART (HAART), vital sign measurements, laboratory results, supplemental oxygen necessity in the ED, ancillary department consults, ED length of stay, and cumulative ED presentations over the 10-year study period. For admitted cases, the assigned location (general ward versus ICU), hospital length of stay, and in-hospital mortality were additionally documented.

Overlapping reasons for visit were grouped with the five most prevalent indications utilized for statistical evaluations, as follows: (1) abdominal pain and/or nausea/vomiting; (2) fever; (3) injury or poisoning; (4) general weakness or poor oral intake; and (5) skin rashes. HIV-associated variables including the most recent cluster determinant 4 (CD4) count, and HIV viral load were also collected via chart review. The number of years since HIV diagnosis was counted as one from the next year of HIV diagnosis.

All parameters were recorded per visit, incorporating multiple presentations by given patients over the 10-year timeframe, with total ED visits per patient also recorded. Select variables such as age, KTAS level, CD4 count, and viral load were further grouped for analytical purposes.

2.4 Statistical analyses

Data with a normal distribution were expressed as mean ± standard deviation, whereas data with a skewed distribution were expressed as median and interquartile range (IQR). First, patients were divided into the discharge and admission groups, and the characteristics of each group were compared. Given that patients could contribute to multiple visits at different time points in our study, we used the generalized estimating equation (GEE) method with an exchangeable correlation structure to produce regression estimates when analyzing repeated measures [9]. Statistical significance was set at p < 0.05. Second, the poor prognosis group was compared to the remainder, and variables with a p value of < 0.10 in the simple model were candidates for the multiple models to investigate independent prognostic factors. The p value was relaxed to p < 0.10 to prevent exclusion of potentially meaningful clinical factors. All the statistical analyses were performed using SAS (version 9.4, SAS Institute, Cary, NC, USA).

3.1 Overall characteristics

During the study period, a total of 369 ED visits corresponding to 183 discrete HIV-positive patients were included in the analysis. Fig. 1 depicts temporal trends in annual ED visit frequency and admission rates among HIV-positive patients over the 10-year period. An average of 8.9 per 10,000 ED visits are made by PWH annually, with an overall admission rate of 49.8%.

Fig. 1.Annual changes of visit numbers of HIV-positive patient and their admission rates via ED. HIV, human immunodeficiency virus; ED, emergency department.

Baseline features of the study population’s ED presentations are summarized in Tables 1 and 2. Most visits (88%) occurred among male patients, with predominant reasons for presentation encompassing abdominal pain and/or nausea/vomiting (17%), fever (13%), injury or poisoning (11%), general weakness or poor oral intake (11%), and skin rashes (5%). Suicide attempts were made in 10 visits (2.7%). Psychiatric and malignant comorbid conditions affected 33% and 21% of visits, respectively. The median interval from initial HIV diagnosis was 8 years; over 90% of cases demonstrated current HAART receipt and CD4 counts ≥200 cells/mL within one year of the documented visits.

As shown in Table 2, half of ED presentations culminated in hospital admission with a median ensuing stay of 9 days. ICU-level care was necessitated for 26 visits (7%) directly from the ED, while 18 registered in-hospital deaths amounted to a 5% visit-based mortality rate. Over the 10-year study period, the median number of ED visits per patient was 3 (range, 1–19 visits).

3.2 Comparison between the admission and discharge group

Parameters exhibiting statistically significant differences between the admission and discharge cohorts appear in Table 3. Regarding comorbidities, chronic renal failure (CRF) increased the odds of admission by 17.7 times (p = 0.001). Additionally, patients in the admission group were more likely to present to the ED with fever, general weakness, or poor oral intake, whereas skin rashes were more common in the discharge group. Specifically, presenting with fever had a threefold increase in the odds of admission (p = 0.001), while presenting with skin rash had a twofold increase in the odds of discharge (p = 0.001) compared with other reasons for visit, except for the five main complaints. In addition, the use of oxygen and consultation with other departments increased the odds of admission by 4.3 and 4.7 times, respectively (p = 0.0004 and p < 0.0001).

3.3 Independent prognostic factors

We compared the poor prognosis group (n = 37), who were admitted directly to the ICU and/or died during the hospital course, with the remainder of the visits (n = 332). The poor prognosis group consisted of 18 patients died during hospitalization (11 died in the ED and 7 in the ICU) and 19 visits who survived after ICU care. In a simple model analysis, the following variables were statistically significant (p < 0.0001) in the poor prognosis group; arrived by ambulance (odds ratio (OR) = 5.78, 95% confidence interval (CI) = 2.92–11.42), KTAS level 1, 2 over KTAS level 4, 5 (OR = 8.255, 95% CI = 4.29–428.79), and use of oxygen at ED (OR = 9.558, 95% CI = 4.49–20.34). Multiple model analysis was performed with significant variables (with a p value of < 0.10 in the simple model) only from comorbid conditions and laboratory findings (Table 4). The odds of poor prognosis were increased 1.9 and 1.6 times for a 10-unit increase in BUN and a 1-unit increase in lactic acid levels, respectively (p = 0.0002 and p < 0.0001).