Chlamydia is the most prevalent notifiable infectious condition in the United States with 1,708,569 cases reported in 2017.1 That same year, 555,608 gonorrhea cases were reported, and rates continue to steeply increase. Rates of both diagnoses in St. Louis, the region where this study was conducted, are also high; in 2016, the city of St. Louis reported rates of 1279 per 100,000 persons and 750 per 100,000 persons for chlamydia and gonorrhea, respectively.2 The cost of diagnosing, providing care for, and managing the burden of these infections is considerable. One estimate places the annual direct health care toll of these infections at $516.7 million for chlamydia and $162.1 million for gonorrhea (in 2010 dollars).3 Those costs can be up to 80% higher when individuals are seen in the emergency department (ED) versus more traditional sexually transmitted disease (STD) clinical care settings.4
The ED is often not well equipped to manage the patient follow-up that STD testing and treatment protocols recommend5 and is also a costly place to receive partner services and/or treatment. However, individuals continue to seek care in the ED for STD-related conditions, potentially as a way to find care in a landscape where safety net providers are steadily decreasing. Perhaps because of this reduction in services, STD-related visits at EDs are on the rise,6 outpacing the general rise in all-cause ED visits.7 Of particular concern to many hospital administrators, clinicians and public health practitioners are individuals who visit the ED repeatedly for costly nonacute concerns, including STD-related care, because these visits are often potentially avertable.8
Individuals who use the ED repeatedly for many types of care have been identified with various terminologies, including “super users,” “frequent users,” or even “frequent fliers.” The definition of a repeat user of the ED can loosely be framed as “an individual who visits the ED more than once,” but the exact number of times necessary to qualify as a repeat user within a specified timeframe is subject to the research questions being asked and has not been officially determined by authorities or the literature.9 A recent article from an urban university hospital defines a frequent user as someone who made 2 visits in a month or 4 in a year.10
Highly frequent ED users have been found to use the ED in place of other health service providers.9 A recent systematic review of the literature on frequent users of the ED (individuals with as few as 2 and up to as many as 12 visits in a year) found that these users made up 4.5% to 8% of all ED patients but comprised 21% to 28% of ED visits.11 Another more recent systemic review found that repeat users could be up to 16% of ED patients and account for up to 47% of visits.12
Because the literature lacks information on the use of the ED by repeat users in reference to STD-related visits, aims of this article include describing the extent to which this happens and describing how individuals who use the ED more than once and receive chlamydia or gonorrhea care differ from individuals who are single users of the ED, as well as assessing what makes an individual more likely to be a repeat user of the ED receiving chlamydia or gonorrhea care.
This study uses clinical data on visits associated with STD-related care collected from 4 EDs in the St. Louis metropolitan region. All 4 EDs were within the same health system, and there were no written policies at the time of the study regarding the practice of testing for STDs. Three of the EDs saw patients of all ages, and the fourth was a pediatric ED. All visits where a test for chlamydia and gonorrhea was administered or an STD-related International Classification of Diseases, Ninth Revision/Tenth Revision (ICD) diagnosis code was given from January 1, 2010, to May 31, 2016, were included, although one hospital did not open until 2012. The following elements were extracted for each visit: patient demographics (date of birth, sex, and race), visit date and time, test results and type of test administered, all ICD diagnosis codes, and insurance type. We used the ICD codes to determine pregnancy and sexual assault status.
We applied the following exclusion criteria to the sample: (1) visits where the patient was aged 13 years, as minors 13 years and older in Missouri are allowed to receive STD care services without the permission of a parent per Missouri State Statute 431.061.1(4) (b) 3; (2) victims of sexual assault because Centers for Disease Control and Prevention protocol stipulates that those individuals should be treated but not necessarily tested5; (3) visits where a test other than a urogenital nucleic acid amplification test (NAAT) was administered because other test types, such as culture tests, often are incidental findings in this system; and (4) visits without any ICD codes because we could not determine sexual assault status.
Each patient's age at the time of each visit was calculated by applying the difference between the patient's date of admission to the ED and date of birth. Race had 11 designations, which were collapsed into 3 categories (black, white, and additional) because of small sample sizes of <1% for the original categorizations other than the 2 designations of “black” or “Caucasian.” Black was recoded into black. Caucasian was recoded into white. Alaska Native–Caucasian, Asian, Asian-black, Asian-other, Asian–Pacific Islander, black-Caucasian, black-other, black–Pacific Islander, black-unknown, Caucasian-other, Caucasian-unknown, Hispanic, Native American, other, other-unknown, Pacific Islander, and unknown were recoded into “multirace/additional.”
Possible test results for a visit included the following: positive for chlamydia, positive for gonorrhea, positive for both (dual infection), or negative for both. During the visit, each patient was seen by an ED health care professional who then assigned one or more ICD codes to the patient's record for the visit. For this study, we used these codes to ensure that patients were not victims of sexual assault and to determine pregnancy status.
Insurance status was categorized into 6 groups: (1) private insurance; (2) self-pay; (3) Medicaid; (4) Medicare; (5) other, comprising “other insurance” and government insurance, which was not able to be further distinguished; and (6) “missing,” which is a combination of 0 and NA. We were not able to obtain insurance information from the pediatric ED or for the years of 2010 to 2012 for 2 of the 3 adult EDs.
We used a unique patient identifier that was stable across the 4 EDs to count the number of visits made by each patient. Fifteen patients made multiple visits within the same day and received an NAAT at each visit. For these patients, the first visit where insurance was attached to the medical record was included in the analysis.
Demographic and clinical characteristics for each patient were selected from their first visit. First visit was chosen because this is the initial encounter of the patient with the ED in our data set, and as such, we used it as a leading indicator to determine whether the patient may return for additional visits in the analytic period.
We used a multivariable zero-inflated negative binomial regression to estimate the relationship between patient demographics and clinical characteristics and the total number of visits made over the study time frame after the initial visit. Because of the majority of patients with only a single visit (or because our outcome variable contained an excessive number of zeros), the zero-inflated model allowed us to model the likelihood of making any future visits alongside the binomial model, which can model the total number of visits. We coded our outcome variable based on user status (number of visits during the study period after the initial visit). We assigned patients who made 1 visit a 0, and patients who made more than 1 visit were assigned the total number of visits they made minus 1 (e.g., if a 0 indicates a single visit user, then a 2 would indicate a user who made 3 visits). Using the likelihood ratio test and examining the overdispersion parameter, we first tested to see if the negative binomial model fit better than a Poisson, and then tested to see if a zero-inflated model improved on the negative binomial model. We included the same predictors (age, sex with pregnancy status, race, payor information, and infection status of gonorrhea and chlamydia at first visit) in both the zero-inflated and the negative binomial parts of the model to better understand how they operate as predictors of user status. We additionally stratified the models by sex to look for differences in predictors, as screening patterns may differ by sex.
IBM SPSS 25 and Stata 15.0 were used to conduct this analysis. Institutional review board approval was received from the Washington University School of Medicine Institutional Review Board and the independent hospitals to conduct this study.
There were 30,475 individuals in the study who made a total of 46,964 visits to the 4 EDs (Table 1). Of the 30,475 patients, 8390 (27%) were repeat users of the ED, and these users made 53% (24,879/46,964) of the visits. Of those 46,964 visits, 35% (16,504) were repeat visits made by repeat users. The maximal number of visits made by any one individual in the entire study period was 39, and the mean number of visits made by individuals who came more than once was 2.97. Male patients comprised 23.7% of the sample (7222), and female patients comprised 76.3% (23,253). The mean (SD) age of men in the study at their first visit was 30.16 (12.48) years, and the mean (SD) age of women was 27.38 (10.17) years. Most male and female patients reported black as their race (86.2% and 83.2%). The total number of visits missing insurance was 12,860 (27.4%).
Most patients were negative for infection at their first visit, 72.7% of men (5249) and 86.1% of women (20,030). For men at their first visit, 11.0% had a gonorrhea infection (793) and 11.6% (838) had a chlamydia infection. For women at their first visit, 2.5% (585) had a gonorrhea infection and 9.4% (2179) had a chlamydia infection. For dual infection at first visit, 4.7% of men tested positive for both infections (342), whereas 2.0% of women had a dual infection (459). Self-pay was the most common insurance status for men (41.0%; 2959), and Medicaid was the most common insurance for women (30.6%; 7117).
At first visit, patients who eventually made multiple visits were 3 to 4 years younger on average than single users (Table 2). Male patients made up a larger percentage of single users (26.5%) compared with individuals who made 2, 3, or 4 or more visits. Individuals identifying as white made up a significantly larger percentage of single users (16.5%) compared with users who made 2, 3, or 4 or more visits. Self-pay was the most common insurance for single users (29.6%), whereas Medicaid was the most common insurance for users who made multiple visits (28.5% for 2 visits, 29.2% for 3 visits, 26.6% for 4 or more visits).
We first used the likelihood ratio test to compare the negative binomial model with a Poisson, and results show the estimated overdispersion parameter of 2.4 (P < 0.001). We then used the Vuong test to compare a zero-inflated model to a standard negative binomial, which was also significant (α = 1.65, P < 0.001). Based on these results, the zero-inflated negative binomial model was determined to best describe the data, and results from this model are presented hereinafter. The results of the 3 regression models (all patients, male patients only, and female patients only) are presented in Table 3. Incident rate ratios (IRRs) are presented for the negative binomial model, which models the count of visits, and odds ratios (ORs) are presented for the zero-inflated part, which models whether patients made any future visits. The interpretation of the OR here is the likelihood of being a zero, or the likelihood of not having the outcome, which is making multiple visits.
In the results from the zero-inflated part of the model, race was the strongest predictor of making future visits here. Black patients were far more likely to make multiple visits compared with white patients in the all-patient zero inflated model (OR, 0.17; 95% confidence interval [CI], 0.11–0.26) holding other predictors constant. The literal interpretation is that black patients had 83% decreased odds of not making future visits—or being a single user—compared with white patients. Race was also a significant predictor in the zero-inflated female-only model, but the estimate for race is not significant in the male-only model. Having an infection at first visit was not a significant predictor of making future visits in the all patient, female-only, or male-only models. Having Medicaid or self-pay insurance at first visit was a stronger predictor of making future visits in the female-only model than in the all-patient model.
The results from the negative binomial aspect of the model addressed those patients who had the opportunity to make multiple visits. In these 3 models (all-patient, male, and female), we found that insurance was the strongest predictor in the all-patient model as well as in the female-only model: having public or no insurance was associated with more future visits. Race was the strongest significant predictor among male patients for making future visits; the IRR for black male patients compared with white male patients was 6.1.
In some cases, the 2 estimates for a single covariate went in different directions, but the results were not significant (the 95% CI crossed 1) indicating the possibility of no difference.
Our results demonstrate that patients who repeatedly receive care for chlamydia and gonorrhea in EDs are more often younger and female, identify as black, and use either self-pay, Medicaid, or Medicare insurance as compared with private insurance, similar to the systematic review of frequent ED users by Giannouchos et al.12 Notably, the percentage of visits in this analysis where the individual presented to the ED more than once in the entire study period and received STD-related care (35%) is higher than the upper range of LaCalle and Rabin's11 recent systematic review on the subject for all ED morbidity visits (28%) as well as the 2018 review (16%) by Giannouchos et al.12 This indicates the possibility that not only are STD-related visits to the ED rising faster than all-cause visits,7 but that repeat STD-related visits to the ED may also be outpacing repeat all-cause visits to the ED as well. Given that a quarter of the patient population in our study was responsible for repeat visits, which accounted for slightly more than a third of the visits, this indicates that identifying repeat users early in their trajectory of ED use could be of considerable benefit to the ED and the health system.
Women made up a significantly large percentage of the entire study population, as well as the repeat users, and this is consistent with their high rates of infection of chlamydia in general.1 There are several potential implications for this higher repeated utilization of the ED associated with chlamydia and gonorrhea NAAT. First, women with gonorrhea or chlamydia infections often present for non–STD-specific complaints (pelvic pain, vaginal bleeding, etc.) and will be tested for gonorrhea and chlamydia as part of their evaluation resulting in higher volumes of testing in women. In addition, some of these women may be visiting repeatedly because their infection went undiagnosed or was undertreated.13,14 Second, women can have various gynecological and obstetrical sequelae, which may cause them to be more concerned about emergent issues at stake (risk of pelvic inflammatory disease, risk to pregnancy, etc.). For women younger than 25 years and women who feel that they may be at increased risk of contracting either chlamydia or gonorrhea, the Centers for Disease Control and Prevention recommends annual testing.5 Women may be seeking STD care in the ED as repeat users relying on the ED to fulfill the need for STD screening. It is also possible that some men may be using the ED in this fashion as well.
Repeat users of the ED have been demonstrated to be in worse health than patients who use the ED infrequently.15 Identifying these individuals early in the course of their ED use could be of clinical importance for their disease burden as well as relevant to the hospital for reducing the cost of subsequent visits. Although we are not sure of their overall health, users who frequented the ED multiple times did have higher rates of chlamydia and gonorrhea at their first visit. At the time of the study, no standard procedures for follow-up treatment of STDs were in place at these EDs and loss to follow-up was common. In their future visits, by seeking STD care in a non-ED, clinic-based venue, patients could receive better follow-up and case management services at a substantially lower cost.16
There is the possibility that some single users were actually repeat users, but additional visits were outside the scope of the study period. Although this is a limitation for the study, the implication of these potential additional visits is that the actual proportion of individuals who make multiple STD-related visits to the ED may be even higher than we report. It was also a limitation of our data set that we did not have the insurance status of all the visits in our study; 12,860 visits of 46,979 total visits, or 27.4% overall, lacked insurance status. The implication of this missing insurance data is that insurance status may play an even larger role in the choice to repeatedly visit the ED and receive STD care than our results indicate. Another limitation of the study was the lack of symptoms and chief complaint in the data set. These variables would have been very useful to describe what potentially prompted patients to seek care at the ED. In addition, it should be noted that our study only investigated 4 EDs in the region of a total of 14, and as a result, we may cannot report the number of visits to other EDs made by the repeat users who received chlamydia and gonorrhea testing.
We do not know what factors led these individuals to seek care in the ED and be tested for chlamydia and gonorrhea, but our results show that race and insurance status are factors at play. Social factors, such as poverty, lack of primary care, and lack of transportation may have contributed to a patient's decision to attend the ED for treatment and services. In addition, the reduction in free or reduced-cost service providers testing and treating STD-related conditions may have prompted some of the patients in our study to seek services at a facility such as the ED, which must provide care regardless of a patient's ability to pay. Because our results show higher rates of repeat use for STD care than reported in recent reviews,11,12 perhaps decreased public health funding along with difficulty accessing competent care is playing a larger role in STD care than other medical conditions. With these findings, there is the potential for enhanced patient education on the signs and symptoms of chlamydia and/or gonorrhea. Patient education could increase the individual's understanding of what to look for regarding these conditions in the future as well as what testing and treatment options exist.
Individuals who visit the ED and receive chlamydia and gonorrhea testing may benefit from information and guidance about clinical alternatives for seeking STD-related care at both their first and subsequent visits. It has been demonstrated that modeling and machine learning17 can effectively predict hospital use based on the history of the patient and other factors. In the case of individuals receiving STD-related care in the ED, a strong potential solution might be for the electronic medical record18 to proactively19 flag these individuals on their second visit as patients who might benefit from additional information about other nearby locations providing more robust STD screening services. In summary, as STD rates continue to rise in the United States, individuals frequenting the ED more than once who receive STD-related care such as the treatment of chlamydia and gonorrhea have the potential to be identified proactively and referred to more appropriate STD clinical care locations for future care.
1. Braxton J, Davis DW, Emerson B, et al. Sexually transmitted disease surveillance 2017. Centers for Disease Control and Prevention. 2018. Available at: https://www.cdc.gov/std/stats17/2017-STD-Surveillance-Report_CDC-clearance-9.10.18.pdf
. Accessed November 1, 2019.
2. Missouri Department of Health and Senior Services. Epidemiologic Profiles of HIV, STD, and Hepatitis in Missouri 2016. Jefferson City, MO: Missouri Department of Health and Senior Service, 2016.
3. Owusu-Edusei K Jr., Chesson HW, Gift TL, et al. The estimated direct medical cost of selected sexually transmitted infections in the United States, 2008. Sex Transm Dis 2013; 40:197–201.
4. Owusu-Edusei K, Patel CG, Gift TL. Does place of service matter? A utilisation and cost analysis of sexually transmissible infection testing from 2012 claims data. Sex Health 2016; 13:131–139.
5. Workowski KA, Bolan GA. Sexually transmitted diseases treatment guidelines, 2015. MMWR Recomm Rep 2015; 64(RR-03):1–137.
6. Batteiger TA, Dixon BE, Wang J, et al. Where do people go for gonorrhea and chlamydia tests: A cross-sectional view of the Central Indiana population, 2003–2014. Sex Transm Dis 2019; 46:132–136.
7. Pearson WS, Peterman TA, Gift TL. An increase in sexually transmitted infections seen in US emergency departments. Prev Med 2017; 100:143–144.
8. Galarraga JE, Pines JM. Costs of ED episodes of care in the United States. Am J Emerg Med 2016; 34:357–365.
9. Doupe MB, Palatnick W, Day S, et al. Frequent users of emergency departments: developing standard definitions and defining prominent risk factors. Ann Emerg Med 2012; 60:24–32.
10. Lucas RH, Sanford SM. An analysis of frequent users of emergency care at an urban university hospital. Ann Emerg Med 1998; 32:563–568.
11. LaCalle E, Rabin E. Frequent users of emergency departments: the myths, the data, and the policy implications. Ann Emerg Med 2010; 56:42–48.
12. Giannouchos TV, Kum HC, Foster MJ, et al. Characteristics and predictors of adult frequent emergency department users in the United States: A systematic literature review. J Eval Clin Pract 2019; 25:420–433.
13. Anaene M, Soyemi K, Caskey R. Factors associated with the over-treatment and under-treatment of gonorrhea and chlamydia in adolescents presenting to a public hospital emergency department. Int J Infect Dis 2016; 53:34–38.
14. Jenkins WD, Zahnd W, Kovach R, et al. Chlamydia and gonorrhea screening in United States emergency departments. J Emerg Med 2013; 44:558–567.
15. Vinton D, Capp R, Rooks S, et al. Frequent users of US emergency departments: Characteristics and opportunities for intervention. Emerg Med J 2014; 31:526.
16. Poon SJ, Schuur JD, Mehrotra A. Trends in visits to acute care venues for treatment of low-acuity conditions in the United States from 2008 to 2015. JAMA Intern Med 2018; 178:1342–1349.
17. Hong WS, Haimovich AD, Taylor RA. Predicting hospital admission at emergency department triage using machine learning. PLoS One 2018; 13:e0201016.
18. Poole S, Grannis S, Shah NH. Predicting emergency department visits. AMIA Jt Summits Transl Sci Proc 2016; 2016:438–445.
19. Kuehn B. A proactive approach needed to combat rising STIs. JAMA 2019; 321:330.