Introduction
Reducing inappropriate antibiotic use in the outpatient setting is a global health priority and a core objective of the 2020–2025 US National Action Plan for Combating Antibiotic-Resistant Bacteria. Reference Blumenthal, Peter, Trubiano and Phillips1–Reference Sun, Manzanares and Foley5 Using antibiotics without medical guidance (non-prescription antibiotic use) can lead to unnecessary use of antibiotics for viral infections and likewise is a threat to patient safety and public health by increasing patients’ risk of adverse effects, superinfections, and disruption of the microbiome and may promote the development of antimicrobial resistance. Reference Blumenthal, Peter, Trubiano and Phillips1–4
Non-prescription antibiotic use is prevalent in both low- and high-income countries. Reference Morgan, Okeke and Laxminarayan3,Reference Sun, Yao and Zhou6–Reference Lescure, Paget, Schellevis and van Dijk9 Recent studies have shown that non-prescription antibiotic use is prevalent in the United States, ranging from 20% to 45%, depending on the patient populations and community groups surveyed. Reference Grigoryan, Paasche-Orlow and Alquicira10–Reference Essigmann, Aguilar and Perkison12 Non-prescription antibiotics were obtained from leftover/previously prescribed courses, friends/relatives and social networks, stores and markets in the United States (illegally sold), over-the-counter sales in other countries, and Internet sources. Reference Grigoryan, Paasche-Orlow and Alquicira10,Reference Grigoryan, Germanos and Zoorob13,Reference Laytner, Chen and Nash14
Previous studies have found that antibiotic misuse is influenced by patient and healthcare system determinants. Reference Lescure, Paget, Schellevis and van Dijk9,Reference Grigoryan, Paasche-Orlow and Alquicira10,Reference Grigoryan, Germanos and Zoorob13,Reference Laytner, Chen and Nash14 Patient-level factors include younger age, lower education levels, using antibiotics in the past, and storing antibiotics in their homes. Reference Lescure, Paget, Schellevis and van Dijk9,Reference Grigoryan, Paasche-Orlow and Alquicira10,Reference Grigoryan, Germanos and Zoorob13,Reference Laytner, Chen and Nash14 Healthcare system factors included having poor access to health care (eg, high healthcare costs, transportation problems) or difficulties with medical visits (eg, long clinic waits to be seen by the clinician or to get an appointment). Reference Grigoryan, Germanos and Zoorob13,Reference Laytner, Chen and Nash14 However, most of the US-based studies on the determinants of non-prescription antibiotic use have focused on Hispanic and Latinx immigrant populations with low income and education, minimal to no health insurance coverage, and limited English language proficiency. Reference Essigmann, Aguilar and Perkison12,Reference Mainous, Diaz and Carnemolla15–Reference Sanchez17
Large-scale surveys investigating the independent effects of social factors on patients’ decisions to use antibiotics without a prescription are lacking in socioeconomically and linguistically diverse US-based primary care settings. Our study aimed to (1) determine the prevalence of intended (future) use of non-prescription antibiotics for 13 predefined situations and (2) identify the social and demographic predictors associated with intended use for these types of situations in diverse outpatient settings.
Study methods
Participants
Design
A cross-sectional survey on non-prescription antibiotic use was conducted in Texas between January 2020 and June 2021. Reference Grigoryan, Paasche-Orlow and Alquicira10 Surveys were administered by trained research coordinators in the waiting rooms of 6 public, safety-net primary care clinics (eg, 3 continuity and 3 same-day/walk-in) and 2 private emergency departments (EDs) that serve racially, ethnically, and socioeconomically diverse patients. Reference Grigoryan, Paasche-Orlow and Alquicira10
Recruitment
Clinic personnel provided recruitment flyers to each participant who checked in for a primary care visit. Reference Grigoryan, Paasche-Orlow and Alquicira10 Interested adult patients volunteered to be surveyed anonymously. Individuals under 18 years old or adults who were unable to answer the survey questions were excluded. Reference Grigoryan, Paasche-Orlow and Alquicira10 All patients were offered a $15 incentive for their time. Reference Grigoryan, Paasche-Orlow and Alquicira10
Survey
The survey interviews (Appendix 1) were conducted in person (when permitted during the pandemic) or remotely (via teleconferencing) in the patients’ preferred language (English or Spanish). Additional survey details (eg, design, recruitment, and survey guide development) are published elsewhere. Reference Grigoryan, Paasche-Orlow and Alquicira10
Survey variables
To assess patients’ endorsement of non-prescription antibiotics (intended use) in various situations, we queried, “If you were feeling sick, would you take antibiotics in the following situations without contacting a doctor/nurse/dentist/clinic?” Each patient was then presented with 13 predefined items/situations, Reference Grigoryan, Burgerhof and Degener18 ranging from “You have leftover antibiotics at home from a previous prescription” to “Antibiotics are cheaper than over-the-counter cold and flu medications” (Appendix 1). Each situation/item was discussed and iteratively designed with input from the clinical staff, a consultant in linguistic and cultural competency for Hispanic communities, a health literacy expert, and our patient advisory board at the safety-net clinics to ensure applicability to this patient population. Reference Grigoryan, Paasche-Orlow and Alquicira10 Patients could respond with “yes,” “no,” or “I don’t know” after each situation was presented to them. Individuals who reported “yes” to any of the situations were scored as endorsing non-prescription antibiotic use.
Patients’ sociodemographic factors were also assessed (Table 1). These sociodemographic factors included patients’ age, gender, race and ethnicity, education level, health insurance status, healthcare system, health literacy, and language preference. Race and ethnicity categories included non-Hispanic Black/African American, Hispanic or Latinx, non-Hispanic White, and other. Education levels included less than high school, high school or GED, or some college and above. Health insurance status included private or Medicare, public or Medicaid or county financial assistance program (CFAP), and self-pay. The CFAP includes those patients who have benefits from Harris County, which allows access to public clinic providers at either very low or no cost. Health literacy was assessed using a brief screening tool by Chew and colleagues. Reference Chew, Bradley and Boyko19,Reference Chew, Griffin and Partin20
Table 1. Patient sociodemographic characteristics by healthcare system. Differences between the public safety-net and private healthcare systems (P < 0.05) are significant
a Boldface indicates statistical significance (P value <0.05).
b Other includes mixed race (Black/African American and Hispanic or “Afro-Latin”) and Asian.
c The county financial assistance program includes those patients who have benefits from a county, which allows access to public clinic providers at either very low or no cost.
d Individuals endorsing (answering “yes”) having difficulty understanding written information, confidence in filling out medical forms by themselves, and someone helps them read clinic or hospital materials. Reference Chew, Bradley and Boyko19,Reference Chew, Griffin and Partin20
Ethics
This study was approved by the Institutional Review Board for Baylor College of Medicine and Affiliated Hospitals (protocol H-45709).
Statistical methods
Comparison of prevalence of intended use between safety-net clinics and private EDs
χ2 tests for categorical variables were used to determine if there were any significant differences among 13 situations between the private and public safety-net healthcare systems. A P value of <0.05 was considered significant (Table 2).
Table 2. Patient-reported intended use for each situation by the healthcare system
a Boldface indicates statistical significance (P value <0.05).
Situational summary factors scores
We used principal components analysis (PCA) to demonstrate the dimensionality of the 13 situations (items). We used PCA as a data reduction technique and confirmed the factor structure underlying these situations, identifying distinct “summary factors.” Smaller composite summary factors were created for items with high correlation (item factor loading >0.6) (Table 3). Three distinct situational summary factors (consistent with the conceptual basis) emerged from the 13 situations/items entered, including (1) barriers to a doctor’s visit (Cronbach’s α = 0.96; mean inter-item correlation [IC] = 0.79), (2) accessibility of non-prescription antibiotics (Cronbach’s α = 0.81; mean IC = 0.48), and (3) previous symptom relief with antibiotics (Cronbach’s α = 0.95; mean IC = 0.9) (Table 3).
Table 3. Dimensionality of the 13 situations/items, including item factor loadings and reliability statistics per situational summary factor
a Patients could respond to each situation with “yes,” “no,” or “I don’t know.” Individuals who reported “yes” endorsed intended non-prescription antibiotic use, and those responding with “no” or “I don’t know” were categorized as non-endorsers of intended non-prescription antibiotic use.
b Loading scores >0.6 were included in the summary factor score.
Each patient was assigned a composite score for each situational summary factor by using the sum of all “yes” responses to each of the situations/items per summary factor. Scores >0 indicate that the patient answered yes to one or more situations within each summary factor. For example, a patient with a score of 0 for summary factor 1 would indicate that they did not experience any barriers to a doctor’s visit (Table 3). Furthermore, each situational summary factor and the patients’ composite scoring were determined as described below.
Summary factor 1. “Barriers to a doctor visit” included 6 items relating to the patients’ (1) work responsibilities, (2) family responsibilities, (3) access to transportation, (4) ability to get a convenient appointment time, (5) difficulty getting an appointment with the doctor when sick, and (6) the perceived cost of the visit (Table 3). Each patient was assessed on a scale of 0–6. For example, a patient’s summary factor 1 composite score of 0 indicated they did not answer “yes” to any situations/items related to the barriers to a doctor visit, and 6 indicated they responded “yes” to all the situations/items presented (Table 4).
Table 4. Range of patients’ composite summary scores for each situational summary factor
Summary factor 2. “Accessibility of non-prescription antibiotics” consisted of 5 items relating to patients’ having (1) leftover antibiotics in their possession, (2) non-prescribed antibiotics given to them by friends or relatives, (3) purchased antibiotics without a prescription in the United States, (4) purchased antibiotics abroad, and (5) beliefs that antibiotics are less expensive than over-the-counter cold and flu medications (Table 3). Each patient was assessed on a scale of 0–5. For example, a patient’s summary factor 2 composite score of 0 indicated they did not answer “yes” to any situations/items related to the accessibility of non-prescription antibiotics, and 5 indicated they responded “yes” to all the situations/items presented (Table 4).
Summary factor 3. “Previous symptom relief with antibiotics” included two items: (1) the patient got better by taking an antibiotic in the past, and (2) the doctor prescribed the patient an antibiotic in the past for the same or similar symptoms (Table 3). Each patient was assessed on a scale of 0–2. For example, a patient’s summary factor 3 composite score of 0 indicated they did not answer “yes” to any situations/items related to previous symptom relief with antibiotics, and 2 indicated they responded “yes” to all the situations/items presented (Table 4).
Linear regression models
Multivariate linear regressions were performed for each summary factor (outcome) to determine the patient-level factors associated with the situations that influence patients’ endorsement of non-prescription use. Predictors for each model included sociodemographic variables that were significant in univariate analyses at the P < 0.2 level (Table 1).
Results
Patient characteristics
Table 1 shows the patients’ sociodemographic characteristics across all clinics/EDs (overall) and by the safety-net and private healthcare systems. Of the 564 patients surveyed, most were female (72%), Hispanic or Latinx (47%), African American or Black (33%), and visited safety-net clinics (72%). Most patients were insured through Medicaid or the county financial assistance program (57%) (Table 1).
Table 4 shows the patients’ composite scores for each situational summary factor. Overall, the accessibility of non-prescribed antibiotics was cited by more than half of the respondents as a situation predisposing influencing their intention to use non-prescription antibiotics (54%). Over half of the patients surveyed reported experiencing at least one situation related to previous symptom relief with antibiotics (53%). More than a third of patients reported barriers to a doctor’s visit (37%) (Table 4).
Healthcare system differences
Table 2 shows the number and percentage of patients that endorsed intended use for the specific situations queried, grouped by the healthcare system. More patients expressed their intention to use non-prescribed antibiotics for each situation in the safety-net clinics than those seen in private EDs.
Many patients reported high intended use (51.3% for safety-net and 37.4% for private) for situations regarding symptom relief with antibiotics (summary factor 2) and accessibility of non-prescribed antibiotics (summary factor 3). For instance, “you got better by taking the antibiotic before” (public 51% vs private 37%; P = 0.012), “your doctor prescribed you the antibiotic for the same symptoms before” (public 53.8% vs private 38.7%; P < 0.001), and “you have leftover antibiotics at home from a previous prescription” (public 55.5% vs private 36.8%; P < 0.001) (Table 2). In addition, the situations related to summary factor 1, barriers to a doctor’s visit, were frequently reported across both healthcare systems. For example, “you cannot take time off work” (public 30.8% vs private 12.9%; P < 0.001), having “no time to go to the doctor because of family/caregiving responsibilities” (pubic 30.8% vs private 14.8%; P < 0.001), or “a visit with a doctor is too expensive” (public 35.5% vs private 20%; P = 0.007) increased patients’ intended use (Table 2).
Multivariate linear regression models
Table 5 shows the multivariate linear regression results for the 3 situational summary factors. Summary factor 1: Barriers to a doctor visit. Compared to private ED patients, the patients receiving care through the safety-net clinics had endorsed higher intended use due to barriers to a doctor’s visit (β [standard error, SE] = 1.34 [0.24]; P < 0.001). Younger patient age was associated with higher intended use for this summary factor (β [SE] = −0.02 [0.008]; P = 0.031). Patient gender, race and ethnicity, education level, insurance status, language preference, and health literacy level were not significant predictors in this model (Table 5).
Table 5. Multivariable linear regression results for each situational summary factor outcome
a All models were adjusted for age, sex, race/ethnicity, education, insurance, healthcare system, survey language, and health literacy. Patient sex, race/ethnicity, education, insurance, survey language, and health literacy were insignificant in the models.
Summary factor 2: Accessibility of non-prescribed antibiotics. Compared to private ED patients, the patients receiving health care through the safety-net clinics had higher intended use because of the accessibility of non-prescribed antibiotic sources (β [SE] = 0.75 [0.154]; P < 0.001). Younger age was associated with more intended use due to the accessibility of non-prescribed antibiotic sources (β [SE] = −0.016 [0.005]; P = 0.001). Patient gender, race and ethnicity, education level, insurance status, language preference, and health literacy level were not significant predictors in this model (Table 5).
Summary factor 3: Previous symptom relief with antibiotics. Safety-net clinic patients had higher intended use than private ED patients because of prior symptom relief with antibiotics for similar/same symptoms (β [SE] = 0.36 [0.096]; P < 0.001). Younger age was associated with higher intended use due to previous symptom relief using antibiotics (β [SE] = −0.01 [0.003]; P = 0.002). Patient gender, race and ethnicity, education level, insurance status, language preference, and health literacy level were insignificant predictors in this model (Table 5).
Discussion
Our study revealed alarmingly high proportions of patients endorsing intended non-prescription antibiotic use across all 13 predefined situations. Up to half (30–50%) of all patients surveyed expressed an intention to use antibiotics without a prescription for situations involving high doctor visit costs, possessing leftover prescription antibiotics, and having experienced prior symptom relief when using antibiotics. In addition, our adjusted results found that younger patient age and receiving care from the safety-net clinics was associated with increased intention to use non-prescription antibiotics across all summary factors (ie, perceived barriers to a doctor visit, accessibility of non-prescription antibiotics, and previous symptom relief with antibiotics).
The high percentage of patients intending to use leftover antibiotic courses (50.4%) or using antibiotics because their doctor had prescribed it previously (49.6%) highlights the ongoing, systemic problem of overprescribing. Clinicians often prescribe antibiotics for self-limiting symptoms that do not typically require antibiotic courses (eg, acute respiratory infections) or prescribe longer courses than recommended by guidelines. Reference Grigoryan, Zoorob, Wang and Trautner21–Reference Hersh, King and Shapiro23 Some of this prescribing behavior is driven by perceived patient expectations of receiving antibiotics. Reference Boiko, Gulliford and Burgess24
Additionally, the large percentage of patients reporting prior symptom relief with antibiotic treatment (47.5%) indicates a substantiated need for additional education on appropriate antibiotic use for patients. Other studies show patients’ trust in primary care professionals’ clinical advice, including willingness to consider alternative/nonantibiotic treatment options if counseled appropriately and involved in shared decision-making regarding antibiotics. Reference Spicer, Roberts and Hicks25,Reference Coxeter, Mar and Hoffmann26 In addition, patient-provider counseling focused on the individual (patient-level) adverse outcomes (eg, Clostridioides difficile infection and severe drug interactions/complications) may resonate with patients and may have more individual relevance than discourse about societal harms or antibiotic resistance resulting from inappropriate antibiotic use. Reference Sun, Manzanares and Foley5,Reference Miller, Carson and Keller27
Our findings also highlight that younger patients report higher intentions to use non-prescription antibiotics for situations related to healthcare system barriers, accessibility of non-prescribed antibiotics, and beliefs that antibiotics will alleviate symptoms experienced. These findings are consistent with other European and US-based studies that also found younger patient age groups were more likely to use non-prescription antibiotics. Reference Sun, Yao and Zhou6,Reference Petersen, Cosgrove and Quinn28–Reference Anderson30 Younger patients may have lower knowledge around antibiotics, resistance, and how to care for self-limiting infections than older patients. Additionally, younger patients may not have an established primary care doctor compared with older patients.
Limitations
Despite having a large, diverse outpatient sample with representation across healthcare systems, including public primary care clinics and private EDs, our study findings may not be generalizable to all US-based patient populations. However, other large US cities with demographics similar to that of Houston may have similar results. For instance, the Rice University Kinder Institute projected that by 2040, most US cities will resemble the sociodemographics of the greater Houston metroplex. Reference Binkovitz31 Additionally, our survey relied on self-reporting; thus, a social desirability response bias may have occurred despite our best efforts to phrase questions neutrally. Some patients may not disclose their antibiotic use without a prescription due to stigma, discrimination, or legalities surrounding the practice. Moreover, the coronavirus disease 2019 pandemic and quarantine may have introduced unintended contextual changes that may have impacted responses. Finally, because patients were surveyed in the healthcare setting, they may be more prone to seek healthcare advice, and these results may be an underrepresentation of the true burden of non-prescription antibiotic use. More research is needed to expand on these results and investigate these factors in other geographically, racially, ethnically, and linguistically diverse study populations in the United States.
Our results broadly align with the priorities of the United States Government Office of the Assistant Secretary for Planning and Evaluation (ASPE). Reference Whitman, De Lew and Chappel32 The ASPE reports that patients’ social factors can negatively impact over 50% of their health outcomes. Reference Whitman, De Lew and Chappel32 The ASPE encourages integrating patients’ social situations into public health, health care, social services, and other governmental institutions in collaboration to advance health equity and improve patient health outcomes and well-being. Reference Whitman, De Lew and Chappel32 In addition, in a recent survey of over 1,500 US physicians, social factors adversely impacted patients’ health behaviors and outcomes. 33 Approximately 77% of the surveyed physicians stated that all or many of their patients had reported at least 1 social determinant that adversely impacts their health care, with financial instability (34%) and transportation (24%) being the most frequently reported social factors, according to patients. 33
Conclusions
Awareness and recognition of patients’ situational circumstances may help healthcare practitioners and policymakers target the underlying reasons why patients intend to use non-prescription antibiotics in future interventions. Future policies and antibiotic stewardship interventions should consider improving patient access to health care and tailoring education focused on safe antibiotic use. Clinicians may choose to provide medically appropriate, alternative (nonantibiotic) treatment options and information to patients experiencing certain symptoms where antibiotic treatments may not be warranted.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ash.2024.361.
Acknowledgments
We would like to thank Osvaldo Alquicira and Juanita Salinas for their assistance in the survey data collection.
Financial support
This work is supported by grant no. R01HS026901 from the Agency for Healthcare Research and Quality (AHRQ). The content is solely the responsibility of the authors and does not necessarily represent the official views of the AHRQ. L. L.’s work is supported by a Ruth L. Kirschstein National Research Service Award (NRSA T-32 6T32HC10031). B. W. T.’s work is supported in part by the US Department of Veterans Affairs (VA) Health Services Research and Development Service (grant no. CIN 13-413) at the Center for Innovations in Quality, Effectiveness, and Safety. The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.
Competing interests
B. W. T. reports grants or contracts from the VA Health Services Research and Development Service, AHRQ R01 (R01 HS029489), AHRQ R18 (R18HS028776), Craig H. Neilson Foundation, Genentech, and Peptilogics; payment from George Washington ID Board for a Review Course; travel support for meeting attendance from the VA Office of Research and Development and the Infectious Diseases Society of America; and an unpaid role on a DSMB for CSP #2004. L. G. reports grants or contracts from the VA Health Services Research and Development Service, AHRQ R01 (R01 HS029489), AHRQ R18 (R18HS028776), Craig H. Neilsen Foundation, National Institute on Alcohol Abuse and Alcoholism (R18HS028776), and a research education grant (1R25AA028203-01) from the National Institute on Alcohol Abuse and Alcoholism Award. All other authors report no potential conflicts.
Open access
