Reducing Ophthalmic Antibiotic Use for Non-severe Conjunctivitis in Children

Abstract

Background

Antibiotics are often overprescribed for pediatric conjunctivitis. We implemented a system-level quality improvement (QI) intervention to reduce unnecessary ophthalmic antibiotic use.

Methods

The multi-faceted intervention in Denver, CO comprised a clinical care pathway, nurse protocol modifications, electronic health record (EHR) changes, parent education materials, and clinician education. We evaluated children aged 6 months–17 years with conjunctivitis seen between November 2018 and December 2022. A multi-interrupted time series model evaluated the effectiveness of the intervention over three time periods: Pre-COVID, Pre-Intervention (November 2018–February 2020), COVID, Pre-Intervention (March 2020–March 2021), and Post-Intervention (April 2021–December 2022). Fisher’s exact tests compared treatment failure and healthcare utilization rates between time periods and among children receiving or not receiving ophthalmic antibiotics.

Results

Among 6960 eligible encounters, ophthalmic antibiotic use was reduced by 18.8% (95% CI: 16.3, 21.3) from Pre-COVID, Pre-Intervention to Post-Intervention. During the Pre-Intervention period following the onset of COVID, a reduction of 16.1% (95% CI: 12.9, 19.3) was observed. Implementation of the intervention resulted in an additional 2.7% (95% CI: −0.4, 5.7) reduction in antibiotic prescribing, primarily in younger children (ages 6 months–5 years). The greatest reduction in prescribing occurred for nurse triage encounters with an 82.1% (95% CI: 76.8, 87.5) reduction in prescribing rates (92.6%–10.5%). Treatment failure occurred in 1301 (18.7%) children and was more common among children that received an ophthalmic antibiotic than those that did not (20.0 vs 17.9%; P = .03).

Conclusion

The QI intervention significantly reduced ophthalmic antibiotic prescribing for pediatric conjunctivitis without increasing treatment failure rates or health care utilization.

INTRODUCTION

Acute infectious conjunctivitis is one of the most common pediatric infections worldwide resulting in significant costs and impact on daycare, school, and work attendance [1–3]. In adults most cases of infectious conjunctivitis are caused by viruses, whereas, children show a wider range of etiology between viruses and bacteria [4–7]. There is no reliable way to clinically differentiate viral from bacterial infections and even in cases of bacterial infection ophthalmic antibiotics provide no or only modest benefit [7–11]. Thus, the American Academy of Ophthalmology recommends against antibiotics for viral infections and no or delayed antibiotics for acute conjunctivitis of unknown etiology [12].

Despite these recommendations, we previously reported that 60%–80% of children with infectious conjunctivitis are prescribed ophthalmic antibiotics [13]. Prescribing practices vary by specialty and clinical setting and prescribing by nurses and health system triage systems are typically driven by protocols that favor antibiotic prescribing [13, 14]. Clinicians are often unaware of standardized guidelines for treatment or return to daycare/school recommendations leading to high variability in care [15].

The overuse of ophthalmic antibiotics has numerous consequences including adverse drug reactions in up to 8% of children [16, 17], the development of antibiotic resistance [18–20], and the disruption of the normal protective microbiota of the eye [21, 22]. Importantly, many daycare and school policies and providers inappropriately exclude children until they have been treated with an ophthalmic antibiotic [23–26]. National guidelines do not recommend antibiotic treatment prior to return and recommend exclusion only for children with systemic symptoms (eg, fever) or in cases where close contact cannot be avoided [27, 28].

To address these challenges, we implemented a quality improvement (QI) intervention at Denver Health and Hospital Authority (DHHA) in Denver, CO to reduce unnecessary ophthalmic antibiotic use and standardize return to daycare/school guidance. We aimed to evaluate the effectiveness of the intervention in reducing ophthalmic antibiotic use and assessed treatment failure rates and healthcare utilization as balancing measures.

METHODS AND MATERIALS

The intervention took place at DHHA in Denver, CO and we evaluated data from November 2018 through December 2022. DHHA is a large, urban, academically affiliated, integrated health system comprised of 32 federally qualified health centers (FQHC) including 12 primary care clinics, 20 school-based health centers, 3 urgent care centers, a level one trauma center that includes a pediatric emergency department, and a 24-hour nurse triage line (NurseLine). The system is the region’s primary safety-net health system and 75% of patients served are at or below 150% of the federal poverty level [29].

We commenced an interdisciplinary team that included primary care, urgent care, infectious disease, and pediatric ophthalmology physicians, nursing and NurseLine leadership, and our antibiotic stewardship pharmacist. We previously identified key drivers of high prescribing which included: (1) lack of an institutional clinical care pathway, (2) poor understanding of national guidelines for management and return to school recommendations, (3) pressure from families, (4) regional school/daycare requirements, and (5) NurseLine and nurse protocols that recommended antibiotic prescribing for most children [15]. To address these drivers, we developed a multi-faceted intervention for all outpatient settings including outpatient clinics, NurseLine and emergency/urgent care departments. Similar to other health systems, standing orders at DHHA allowed nurses and nurse triage line providers to prescribe ophthalmic antibiotics over the phone with clinician oversite at baseline. The intervention included creation of an institutional clinical care pathway, changes to nurse and NurseLine conjunctivitis protocols, electronic health record (EHR) changes, parent education materials, and education for clinicians, nurses, and Denver Public School nurses (Table 1). All materials are available in the Supplementary material for modification and use. We launched the intervention in May of 2021 and all components were fully implemented by June of 2021. A washout or transition period was not used as many elements went into effect immediately.

Table 1.

Intervention Components

Intervention	Description
Clinical Care Pathway	Evidence-based algorithm for management of acute infectious conjunctivitis in children and adults
Update NurseLine triage protocols	Removed automatic antibiotic prescribing for most children and increased recommendations for supportive care Aligned return to school/daycare guidance with national guidelines
Updated clinic nurse conjunctivitis protocols	Removed automatic antibiotic prescribing for most children and increased recommendations for supportive care Aligned return to school/daycare guidance with national guidelines
Electronic decision support	EPIC (Verona, WI) Smartset that included diagnosis code, supportive care measures, medications (eg, artificial tears), and parent education materials
School/Work note	A pre-written return to school/work note in EPIC with return to school/daycare recommendations in alignment with state, national and institutional guidelines
Provider/nurse education	All NurseLine nurses as well as Denver Public School nurses and all providers in the pediatric, family medicine, internal medicine, urgent care, and emergency departments were provided education/training with a 1 hour session on the new acute conjunctivitis management protocol and return to school/day care recommendations
Parent Education	Electronic education materials in English and Spanish that could be added to the after-visit summary

We evaluated ophthalmic antibiotic prescribing over three time periods: Pre-COVID, Pre-Intervention (November 2018–February 2020), COVID, Pre-Intervention (March 2020–May 2021), and Post-Intervention (June 2021–December 2022). We included children who were aged 6 months–17 years at diagnosis date and had an encounter for conjunctivitis in a DHHA outpatient setting including: clinic nurse phone calls, NurseLine, outpatient clinics, urgent care, or emergency departments. Because care for conjunctivitis often occurs over the phone we included in-person and phone or video visits. We included encounters with an associated International Classification of Diseases (ICD)10 code [13] for acute infectious conjunctivitis or a reason for visit of conjunctivitis. We excluded children if they had a diagnosis of acute infectious conjunctivitis diagnosis within 30 days prior to the encounter, used ophthalmic antibiotics in the 30 days prior to the encounter, had a history of ophthalmic surgeries or chronic conjunctivitis in the past 12 months or if their conjunctivitis diagnosis was associated with an inpatient encounter.

Our primary outcome was the proportion of children with conjunctivitis who were prescribed an unnecessary ophthalmic antibiotic. We assumed all antibiotic prescriptions for children who met the inclusion criteria were unnecessary based on the current American Academy of Ophthalmology recommendations [11]. Balancing measures were treatment failure defined as (1) emergency department/urgent care visit for conjunctivitis within 14 days of the index encounter, (2) clinic follow-up telehealth/video or in-person for conjunctivitis within 14 days of the index encounter, and/or (3) new ophthalmic antibiotic prescription (same or different) within 14 days of the index encounter. Treatment failure included children who were initially evaluated, not prescribed an antibiotic, and subsequently required an antibiotic because they worsened or did not improve within 3–5 days. To ensure all subsequent visits for conjunctivitis were included, the team analyzed diagnoses of various ICD-10 codes for conjunctivitis (which were previously validated [13]) to assess frequency of conjunctivitis related follow-ups.

We used a multi-interrupted time series model to assess the impact of COVID and the intervention on reducing ophthalmic antibiotic use [30]. This model is ideal for observational analyses of longitudinal effects following events or interventions where randomization is not feasible. The model took the general form:

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{O}\mathrm{u}\mathrm{t}\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{e}=\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{h}+\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{o}{\mathrm{n}}_{\mathrm{C}\mathrm{O}\mathrm{V}\mathrm{I}\mathrm{D}\text{~}\text{-}\mathrm{P}\mathrm{r}\mathrm{e}}+\mathrm{t}\mathrm{i}\mathrm{m}{\mathrm{e}}_{\mathrm{C}\mathrm{O}\mathrm{V}\mathrm{I}\mathrm{D}\text{~}\text{-}\mathrm{P}\mathrm{r}\mathrm{e}}+\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{o}{\mathrm{n}}_{\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}}+\mathrm{t}\mathrm{i}\mathrm{m}{\mathrm{e}}_{\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}}}\end{array}}$$

where month is the incremental increase in months starting at the beginning of the Pre-COVID, Pre-Intervention period, intervention_COVID-Pre is a dummy variable (0/1) that is used to identify the COVID, Pre-Intervention period, time_COVID-Pre is the incremental increase in time (months) from the beginning of the COVID, Pre-Intervention period, intervention_Post is a dummy variable (0/1) that is used to identify the Post-Intervention period, and time_Post is the incremental increase in time (months) with respect to the beginning of the Post-Intervention period. We assessed changes in ophthalmic antibiotic prescriptions and trends (slope and level change) for each time frame.

We conducted Pearson’s chi-squared or Fisher exact tests (when cell size was less than 5) to assess association between time periods, including treatment failure as defined by utilization of primary care or emergency/urgent care, and by prescription of new ophthalmic antibiotic. The Pre-COVID, Pre-Intervention period and the COVID, Pre-Intervention period were each compared to the Post-Intervention period for balancing measures and the interrupted time series slope and level changes. Finally, we directly compared balancing measures between children who were and were not prescribed an ophthalmic antibiotic at their index encounter (regardless of time period). We considered an alpha of .05 or less using two-tailed tests to be statistically significant for all analyses.

The project was reviewed by the Quality Improvement Committee of DHHA, which is authorized by the Colorado Multiple Institutional Review Board at the University of Colorado, Denver, CO and the DHHA Ethics Committee, Denver CO, and was exempted as it was determined not to be human subjects’ research.

RESULTS

In total, there were 6960 eligible encounters including 2685 in the Pre-COVID, Pre-Intervention period, 1287 in the COVID, Pre-Intervention period, and 2988 in the Post-Intervention period (Table 2). Encounters occurred most often in primary care settings (71.2%) including 1340 (19.3%) in family medicine and 3615 (51.9%) in pediatrics. In total, there were 635 (9.1%) NurseLine encounters. Similar to the demographics of the DHHA population, patients tended to be Hispanic (4787, 68.8%), White (5171, 74.3%) and utilize public insurance, financial assistance, or self-pay (6409, 92.1%).

Table 2.

Demographics of Patients by Intervention Period

Feature	Pre-COVID, Pre-Intervention N (%) N = 2685	COVID, Pre-Intervention N (%) N = 1287	Post-Intervention N (%) N = 2988
Age (years; mean, SD)	7.0 (5.20)	8.5 (5.33)	7.5 (5.10)
Gender
Male	1371 (51.1)	666 (51.8)	1654 (55.3)
Female	1314 (48.0)	621 (48.2)	1334 (44.7)
Race
Black	364 (13.6)	217 (16.9)	475 (15.9)
White	2122 (79.0)	966 (75.1)	2083 (69.7)
Other	199 (7.4)	104 (8.1)	430 (14.4)
Ethnicity
Non-Hispanic	806 (30.0)	389 (30.2)	978 (32.7)
Hispanic	1879 (70.0)	898 (69.8)	2010 (67.3)
Insurance
Private	156 (5.8)	56 (4.4)	306 (10.2)
Public	1950 (72.6)	958 (74.4)	2328 (77.9)
Self-pay	551 (20.5)	256 (19.9)	260 (8.7)
Financial assistance	26 (1.0)	16 (1.2)	64 (2.1)
Other/unknown	2 (0.1)	1 (0.1)	30 (1.0)
Language
English	1819 (67.8)	828 (64.3)	1974 (66.1)
Spanish	747 (27.8)	406 (31.6)	864 (28.9)
Other/unknown	119 (4.4)	53 (4.1)	150 (5.0)
Clinic type
NurseLine	352 (13.1)	111 (8.6)	172 (5.8)
Urgent/emergency care	353 (13.1)	78 (6.1)	355 (11.9)
Family medicine	558 (20.8)	283 (22.0)	499 (16.7)
Pediatrics	1277 (47.6)	686 (53.3)	1652 (55.3)
Ophthalmology	136 (5.1)	96 (7.5)	14 (0.5)
Other	9 (0.3)	33 (2.6)	296 (9.9)

At baseline there were inequities in ophthalmic antibiotic prescribing by race and ethnicity. In total 6.3% more White children and 4.4% more non-Hispanic children were prescribed an antibiotic than African American/Black children and Hispanic children, respectively (Table 3). These inequities in prescribing by race and ethnicity appeared to decrease but were not significant between the Pre-COVID, Pre-Intervention and Post-Intervention period. The differences in the proportions of children that were prescribed an antibiotic fell from 6.3% to 4.9% (P = .07) between White and African American/Black children and 4.4%–1.3% (P = .41) between non-Hispanic and Hispanic children.

Table 3.

Ophthalmic Antibiotic Prescribing by Time Period, Demographics, and Clinical Setting

	Time Period			Absolute Difference
	Pre-COVID, Pre-Intervention	COVID, Pre-Intervention	Post-Intervention	Pre-COVID, Pre-Intervention to COVID, Pre-Intervention	Pre-COVID, Pre-Intervention to Post-Intervention	COVID, Pre-Intervention to Post-Intervention
	N (%)	N (%)	N (%)	% (95% CI)	% (95% CI)	% (95% CI)
	N = 2685	N = 1287	N = 2988
Total	1332 (49.6)	431 (33.5)	921 (30.8)	16.1 (12.9, 19.3)	18.8 (16.3, 21.3)	2.7 (−0.4, 5.7)
Age
6–23 months	325 (69.3)	118 (78.7)	181 (48.8)	−9.4 (−17.1, −1.6)	20.5 (13.9, 27.1)	29.9 (21.6, 38.2)
2–5 years	477 (60.8)	147 (50.2)	354 (39.1)	10.6 (4.0, 17.3)	21.7 (17.1, 26.4)	11.1 (4.5, 17.6)
6–10 years	290 (42.6)	97 (28.5)	231 (27.7)	14.1 (8.0, 20.1)	14.9 (10.1, 19.7)	0.8 (−4.9, 6.5)
10–17 years	240 (32.0)	69 (13.7)	155 (17.7)	18.3 (13.8, 22.8)	14.3 (10.1, 18.5)	−4.0 (−7.9, −0.0)
Race
Black	162 (44.5)	79 (36.4)	123 (25.9)	8.1 (−0.1, 16.3)	18.6 (12.1, 25.1)	10.5 (3.0, 18.0)
White	1077 (50.8)	317 (32.8)	642 (30.8)	17.9 (14.3, 21.6)	19.9 (17.0, 22.8)	2.0 (−1.6, 5.6)
Other	93 (46.7)	35 (33.7)	156 (36.3)	13.1 (1.7, 24.5)	10.4 (2.2, 18.7)	−2.6 (−12.8, 7.5)
Ethnicity
Non-Hispanic	425 (52.7)	144 (37.0)	310 (31.7)	15.7 (9.8, 21.6)	21.0 (16.5, 25.5)	5.3 (−0.3, 10.9)
Hispanic	907 (48.3)	287 (32.0)	611 (30.4)	16.3 (12.5, 20.1)	17.9 (14.9, 20.9)	1.6 (−2.1, 5.2)
Clinic type
NurseLine	326 (92.6)	99 (89.2)	18 (10.5)	3.4 (−3.0, 9.8)	82.1 (76.8, 87.5)	78.7 (71.3, 86.1)
Urgent/emergency care	223 (63.2)	51 (65.4)	198 (55.8)	−2.2 (−13.9, 9.5)	7.4 (0.2, 14.6)	9.6 (−2.1, 21.4)
Family Medicine	252 (45.2)	89 (31.5)	157 (31.5)	13.7 (6.9, 20.5)	13.7 (7.9, 19.5)	−0.0 (−6.8, 6.8)
Pediatrics	526 (41.2)	185 (27.0)	503 (30.4)	14.2 (9.9, 18.5)	10.7 (7.2, 14.2)	−3.5 (−7.4, 0.5)
Ophthalmology	3 (2.2)	1 (1.0)	2 (14.3)	1.2 (0.8, 1.5)	−12.1 (−30.6, 6.4)	−13.2 (−31.7, 5.2)
Other	2 (22.2)	6 (18.2)	43 (14.5)	4.0 (−26.1, 34.2)	7.7 (−19.7, 35.2)	3.7 (−10.1, 17.4)
Provider type
Physicians	421 (38.3)	98 (26.5)	273 (22.3)	11.8 (6.5, 17.1)	16.0 (12.3, 19.7)	4.2 (−0.9, 9.2)
Physician Assistants	184 (57.1)	47 (39.8)	151 (41.5)	17.3 (7.0, 27.7)	15.7 (8.2, 23.1)	−1.7 (−11.8, 8.5)
Nurse Practitioners	198 (60.2)	40 (59.7)	169 (59.7)	0.5 (−12.4, 13.4)	0.5 (−7.3, 8.3)	−0.0 (−13.1, 13.0)
Optometrists	2 (1.8)	0 (0.0)	2 (1.3)	1.8 (−0.6, 4.3)	0.6 (−2.5, 3.5)	−1.3 (−3.0, 0.5)
Nurses (NurseLine)	326 (92.6)	99 (89.2)	18 (10.5)	3.4 (−3.0, 9.8)	82.1 (76.8, 87.5)	78.7 (71.3, 86.1)
Other/unknown	201 (42.6)	147 (28.0)	308 (39.1)	14.6 (8.8, 20.5)	3.4 (−2.2, 9.1)	−11.1 (−16.3, −6.0)

There was a 18.8% (95% CI: 16.3, 21.3) reduction in ophthalmic antibiotic use from the Pre-COVID, Pre-Intervention period to the Post-Intervention period (Table 3, Figure 1). The emergence of the COVID-19 pandemic, prior the intervention, resulted in a 16.1% (95% CI: 12.9, 19.3) reduction in ophthalmic antibiotic prescriptions that was primarily driven by a reduction in prescribing among school-aged children (ages 6–17 years). Rates of prescribing among school-aged children remained low even after schools reopened and other social distancing measures (eg, mask wearing) were reduced. The implementation of the intervention resulted in an additional 2.7% (95% CI: −0.4, 5.7) reduction in ophthalmic antibiotic prescribing that was primarily driven by reduced prescribing for younger children (ages 6 months–5 years). The greatest reduction in prescribing occurred for nurses with NurseLine encounters which had an 82.1% (95% CI: 76.8, 87.5) reduction in prescribing rates (92.6%, 326/352 to 10.5%, 18/172). All outpatient settings, except ophthalmology (which already had low prescribing rates at baseline), saw a significant reduction in ophthalmic antibiotic prescribing rates. Aside from nurses, physicians had the next greatest reduction in prescribing rates, with a 16.0% reduction (95% CI: 12.3, 19.7). The slope change observed between the COVID, Pre-Intervention and Post-Intervention period was statistically significant (P < .01), indicating an increase in prescribing. This change coincided with the reopening of schools, the relaxation of social distancing measures and seasonality. There was no significant difference between slopes in the Pre-COVID, Pre-Intervention and Post-Intervention period (slope = 0.26, 1.07, respectively; P = .24) (Figure 1, Supplementary Table). Level change fell significantly between the Pre-COVID, Pre-Intervention and Post-Intervention period (−32.70; P < .01).

Figure 1.

Change in ophthalmic antibiotic prescribing by time-period. There is no significant difference in slope between the Pre-COVID, Pre-Intervention and Post-Intervention period (slope = 0.26, 1.07, respectively; P = .24). Level change between Pre-COVID, Pre-Intervention and the Post-Intervention period was significant (−32.7, P < .01). See Supplementary Table for more details.

Treatment failure occurred in 1301 (18.7%) children and was more common among children that received an ophthalmic antibiotic than those that did not (20.0 vs 17.9%; P = .03; Table 4). The most common type of failure was seeking primary care for a diagnosis of conjunctivitis within 14 days (1279, 18.4%). New prescriptions for ophthalmic antibiotics (7, 0.1%) and visits to emergency/urgent care within 14 days (22, 0.3%) were rare. There were no increases in overall treatment failure between the Pre-COVID, Pre-Intervention and the Post-Intervention period (19.1 vs 20.3%; P = .23; Table 5). No patients experienced serious complications (eg, corneal ulcerations) that required ophthalmology follow-up.

Table 4.

Treatment Failure by Initial Management Strategy

Clinical Outcome	Antibiotica	No Antibiotic	P-value
	N (%) N = 2684	N (%) N = 4276
Failure-any	536 (20.0)	765 (17.9)	.03
Emergency/urgent care visit (within 14 days)	10 (0.4)	12 (0.3)	.51
Primary care visit (within 14 days)	526 (19.6)	753 (17.6)	.04
New ophthalmic antibiotic prescribed (within 14 days)	4 (0.2)	3 (0.1)	.44

^aPrescribed ophthalmic antibiotics at index encounter.

Table 5.

Failure Rates by Intervention Period

Clinical Outcomes	Pre-COVID, Pre-Intervention	Post-Intervention	P-value	COVID, Pre-Intervention	Post-Intervention	P-value
	N (%) N = 2685	N (%) N = 2988		N (%) N = 1287	N (%) N = 2988
Failure-any	547 (20.3)	571 (19.1)	.23	183 (14.2)	571 (19.1)	<.01
Emergency or urgent care visit (within 14 days)	7 (0.3)	9 (0.3)	.77	6 (0.5)	9 (0.3)	.40
Primary care visit (within 14 days)	540 (20.1)	562 (18.8)	.22	177 (13.8)	562 (18.8)	<.01
New ophthalmic antibiotic prescribed (within days)	3 (0.1)	3 (0.1)	1.00	1 (0.1)	3 (0.1)	1.00

CONCLUSION

A low-cost, pragmatic QI intervention significantly reduced ophthalmic antibiotic prescribing for conjunctivitis in children and did not result in increased emergency/urgent care visits or subsequent ophthalmic antibiotic prescriptions. All clinical care settings except ophthalmology (whose baseline was already low), saw a reduction in prescribing and the greatest improvement was among children evaluated by NurseLine. Children who were prescribed an ophthalmic antibiotic were more likely to have treatment failure than those who were managed only with supportive care.

Importantly, children with conjunctivitis are managed across many settings and specialties within health systems. Conjunctivitis is often managed over the phone by nurse triage lines or clinic nurses. We previously reported that prescribing in these settings is high and is largely driven by protocols [13]. Modification of these protocols resulted in the greatest reduction in antibiotic use. Additionally, it was critical to ensure alignment between the care and return to school recommendations provided by different specialties, nurses and clinicians, and care settings. This likely reduced frustration and confusion among staff, parents, and school nurses. Standardized clinical care pathways, protocols, EHR tools, and school/work note templates were used throughout the organization. Formative cross-collaboration with different specialties and nursing leadership was important to identify blind spots, assure the interventions could be successfully implemented and get buy-in.

The differences in prescribing rates by age group are notable. The COVID-19 pandemic resulted in a substantial decrease in ophthalmic antibiotic use among school age children. This might have occurred because there was decreased pressure from schools for children to use an ophthalmic antibiotic prior to return because many children were attending virtual school, were already excluded for other respiratory symptoms, or were required to wear masks and social distance which can decrease transmission. In addition, this may be explained by a decrease in the incidence of viral upper respiratory infections early on in the pandemic due to quarantine [31, 32]. In contrast, the intervention had the greatest impact on prescribing for younger children. This could be because prescribing for older children was already low by the time the intervention was implemented or because clinicians were initially more likely to prescribe antibiotics to younger children. Although antibiotic prescribing increased over the Post-Implementation period, likely due to children returning to daycare and school and the easing of social distancing measures, in addition to seasonality, the slope did not differ significantly from the Pre-COVID, Pre-Intervention period. Additionally, the change in level of prescribing between the Pre-COVID, Pre-Intervention and Post-Intervention period was significant, demonstrating that even after children returned to Pre-COVID activities, antibiotic prescribing remained lower than the Pre-COVID, Pre-Intervention period. Inequities in prescribing have been frequently reported among children with common illnesses including conjunctivitis, acute otitis media [33], and other respiratory infections [34]. It is speculated that implicit bias around perceived parental expectations for antibiotics and lower likelihood of shared decision-making with non-white, non-Hispanic parents may account for these inequities [33, 34]. We previously demonstrated that a similar pragmatic, system-level approach to standardizing care reduced inequalities in care for sexually transmitted infections [35]. We suspect that standardization of practices resulted in reduced implicit bias by clinicians and staff though further work is needed to understand how to best sustain these changes.

We carefully monitored treatment failure rates after the intervention to ensure that children who were not initially prescribed an ophthalmic antibiotic did not simply seek care in our emergency department or urgent cares, have another clinician prescribe an ophthalmic antibiotic based on parent preference, or medically fail supportive care. In alignment with clinical trial data, overall failure rates were low and were not higher for children who were managed by supportive care compared to those prescribed an ophthalmic antibiotic [7–9, 11]. Few (<1%) children in our organization sought emergency/urgent care or required an ophthalmic antibiotic prescription after the index encounter. Nearly 20% of children were evaluated in a primary care setting for conjunctivitis within 14 days, which may have been partly due to scheduled follow-up visits or follow-up for an overdue well child visit or vaccines. However, rates of subsequent care decreased over time and these visits were not associated with ophthalmic antibiotic prescribing even for children initially managed by supportive care. Treatment failure rates were lower for children treated with supportive care vs ophthalmic antibiotics. Because this was not a randomized controlled trial it is not clear if antibiotics were associated with an increased need to seek care or if there were simply inherent differences between children that were and were not initially prescribed an antibiotic (eg, severity of symptoms). In alignment with our findings, a recent randomized clinical trial found that children initially treated with ophthalmic antibiotics were more likely to have relapse and require subsequent treatment than those managed by supportive care [11]. In any event, neither the intervention nor use of supportive care resulted in increased treatment failure or healthcare utilization.

Our evaluation has some limitations. Because the intervention needed to span numerous care areas including NurseLine we could not randomize clinics to either intervention or control. Thus, while the data suggest the reduction in antibiotic use was from the intervention, we cannot exclude the possibility that it could have been due to other factors. We also did not track the use of the interventions as a process measure due to feasibility and since many aspects were mandatory and unavoidable (such as EHR changes). Although, if possible, process measures could strengthen other/future studies with greater ability to attribute interventions to the outcome. The intervention took place in one health system; thus, the results may not be generalizable to other systems or different patient populations. There is a possibility that some relevant encounters were missed due to alternative diagnoses codes. We were, however, systematic in our process of identifying and including relevant conjunctivitis related ICD-10 codes. Based on current American Academy of Ophthalmology recommendations, we felt comfortable with the assumption that most ophthalmic antibiotic prescriptions were unnecessary, however, some occurrences may have warranted antibiotics. With the upward slope Post-Intervention, it is not clear whether the effect of the intervention is transient or if the reduction in antibiotic prescriptions will persist over time. Continual tracking of prescribing rates is needed to track sustainability of the intervention. Finally, we did not directly assess the impact of the intervention on parent satisfaction, length of symptoms, pediatric quality of life, or school/work absenteeism. This is an important next step in understanding how to best improve care for children with conjunctivitis.

Despite these limitations, our evaluation has several key strengths. The intervention was low-cost and pragmatic; it could be easily replicated by other health systems. We were able to evaluate the impact of the intervention across multiple specialties and clinical care settings using a large number of children. Because the population at DHHA is diverse, we could also evaluate for changes in prescribing inequities. Additionally, because DHHA patients tend to only seek care within the DHHA system we could reliably assess for treatment failure after the index episode.

In conclusion, a pragmatic multi-faceted intervention successfully reduced ophthalmic antibiotic use for children and did not increase treatment failure rates or healthcare utilization. Clinicians and health systems should carefully evaluate their current prescribing practices and return to school guidance. There are almost certainly opportunities to improve care for a diagnosis that affects a tremendous number of children.

Notes

Financial support. H.F. received salary support from the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under Award Number K23HD099925. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Role of funder/sponsor. The funder had no role in the design or interpretation of the study.

Potential conflicts of interest. The authors have no conflicts of interest to disclose.

Contributors statement. T.S. conceptualized and designed the study, provided subject matter support, analyzed and interpreted data, drafted the initial manuscript, and critically reviewed and revised the manuscript. J.D. performed all data collection, analyzed and interpreted data, critically reviewed and revised the manuscript. N.W. performed all data collection, analyzed and interpreted data, critically reviewed and revised the manuscript. J.J. conceptualized and designed the study, provided subject matter support, analyzed and interpreted data, drafted the initial manuscript, and critically reviewed and revised the manuscript. A.K. assisted with data collection and analysis, and critically reviewed and revised the manuscript. T.C.J. provided subject matter support, analyzed and interpreted data, and critically reviewed and revised the manuscript. H.M.F. conceptualized and designed the study, provided subject matter support, analyzed and interpreted data, drafted the initial manuscript, and critically reviewed and revised the manuscript. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.