Unraveling the Impact of Pneumococcal Conjugate Vaccines on Ambulatory Antibiotic Drug Consumption in Young Children: An Interrupted Time-Series Analysis

Abstract

Background

Pneumococcal conjugate vaccines (PCVs) reduce respiratory infections in young children, the main antibiotic consumers. Following PCV implementation, dispensed antibiotic prescription (DAP) rates in young children were expected to decline.

Methods

Computerized data on DAP for children <5 years were examined during a 13-year period (including 4 pre-PCV years). All DAPs from clinics with ≥50 insured children, active both pre– and post–PCV implementation were included. Interrupted time-series with segmented regression was applied to analyze monthly DAP rate trends, adjusted for age, ethnicity, and season. Incidence rate ratios (IRRs) of DAPs during the late PCV13 period versus 4 years pre-PCV were calculated both as absolute rate ratios (aIRRs) and relative to expected rates (rIRRs).

Results

Of 1 090 870 DAPs, 57% were in children <2 years. All-DAP rates peaked in the cold season. Post–PCV7/PCV13 implementation, all DAP rates abruptly and significantly declined, reaching a plateau within 5 years. This was largely driven by amoxicillin/amoxicillin-clavulanate (75% of DAPs). Age <2 years and Bedouin ethnicity were significantly associated with higher pre-PCV DAP rates but with faster and greater decline post-PCV, achieving near elimination of gaps between ages and ethnic groups. Overall reduction (95% CIs) in DAP rates per 1000 was estimated between aIRR (344.7 [370.9–358.4]) and rIRR (110.4 [96.9–123.7]) values.

Conclusions

Shortly following PCV implementation, overall DAP rates showed an abrupt, steep decline, stabilizing within 5 years, in parallel to post-PCV respiratory infection trends previously described in this population, suggesting causality. The variable patterns of certain drug categories suggest additional influences beyond PCV.

pneumococcal conjugate vaccines, antibiotic consumption, community, children, impact

Young children are the main consumers of antibiotics [1], mostly for respiratory tract infections (RTIs) [2, 3]. Dispensed antibiotic prescription (DAP) rates peak yearly during the respiratory season [4]. Overprescribing for children is common [5]. Community stewardship programs have resulted in only limited success [6].

Pneumococcal conjugate vaccine (PCV) implementation resulted in a rapid reduction in pneumococcal disease in children. The impact of PCV on RTI burden has been many-fold higher than on invasive pneumococcal diseases (IPDs) [7]. In Israel, a marked reduction in RTIs following 7-valent/13-valent PCV (PCV7/PCV13) implementation was observed, including otitis media [8], pneumonia [9], all-cause lower RTIs [10], and conjunctivitis [11]. Given the relationship between antibiotic consumption and RTIs, an impact of PCVs on antibiotic consumption in young children is plausible; however, so far, only a very short postimplementation period was reviewed, with no attempt to study detailed drug-specific dynamics.

We assessed the impact of the sequential PCV7/PCV13 introduction into the National Immunization Plan (NIP) on ambulatory antibiotic consumption in children younger than 5 years, by determining the detailed DAP rate dynamics before and after implementation.

In southern Israel, 2 distinct ethnic groups, the Jewish and the Bedouin populations, live side by side. Prior to PCV implementation, crowding, pneumococcal carriage, and RTIs were more prevalent among Bedouin children [13–15] and DAP rates were higher [16]. Therefore, we compared the DAP rate dynamics between these 2 populations.

METHODS

Study Population and Design

This retrospective analysis covered a 13-year period (July 2005 through June 2018) in southern Israel (Negev district). The Soroka University Medical Center Ethics Committee approved the study.

Setting

Two ethnic groups inhabit the Negev: the Jewish population, generally comparable to a Western population, and the Bedouin population, in transition from semi-nomadism to settlement, generally comparable to a developing population. Compared with the Jewish population (~410 000 in 2006), the Bedouins (~150 000 in 2006) are characterized by overcrowding, lower income and educational level, larger family size [17, 18], and higher rates of pneumococcal diseases and RTIs [19]. In 2005 and 2018, 71 508 and 93 000 children younger than 5 years, respectively, lived in the Negev. Social contacts between Bedouin and Jewish children are limited. Community medical services are provided mostly in separate neighborhood clinics. Still, 11% of children receive medical services in mixed Jewish-Bedouin clinics.

In 2005 and 2018, 76% and 83% of all children younger than 5 years, respectively, received all medical services from the largest Health Maintenance Organization (HMO) in Israel, the Clalit HMO (C-HMO; data provided by C-HMO).

Vaccine Uptake

PCV7 was licensed in Israel in mid-2007 and used rarely in the private market until July 2009 when it was introduced into the NIP, with a catch-up campaign for children younger than 2 years. PCV13 has gradually replaced PCV7 since November 2010, with no further catch-up. The proportion of 2–4-year-old children having received 3 or more doses of PCV7 reached 86.0% by mid-2012; the proportion who received 3 or more PCV13 doses has gradually increased to ~90% since 2013 [13]. PCV uptake was similar in Jewish and Bedouin children.

Study Design

All children younger than 5 years insured by C-HMO being served by clinics with 50 or more children, active both before and after PCV implementation, were included. These comprised 84.5% and 95.2% of all Jewish and Bedouin children insured by C-HMO, respectively (Supplementary Table 1).

All DAPs were computerized. Monthly DAP rates were calculated by age, ethnic group, season (warm months: July–September; cold months: November–March), and antibiotic category (amoxicillin/amoxicillin-clavulanate, azithromycin, oral second-generation cephalosporins [OSGCs], and all-other antibiotics [trimethoprim/sulfamethoxazole, erythromycin, clarithromycin and clindamycin]). We did not study parenteral antibiotics. Amoxicillin and amoxicillin-clavulanate were consolidated into a single category, since during the study period, it was more frequently advised to use high-dose amoxicillin or amoxicillin-clavulanate, yet high-dose amoxicillin-clavulanate preparations had not yet been licensed, resulting in interchangeable use.

Statistical Analysis

To compare between ethnic groups, clinics caring solely for either ethnic group were chosen. Those caring for mixed Jewish-Bedouin children were included when all children were analyzed together.

We used an interrupted time-series analysis, a quasi-experimental design, to evaluate the DAP rate trends before and after PCV implementation. A transitional period between July 2009 and June 2010 (between PCV7 and PCV13 introduction) was defined. Segmented linear regression with autoregressive error models was designed, taking into account autocorrelation, seasonality, and the pre- and post-PCV trends. The effect of PCV introduction (the intervention) on DAP rates was estimated by inserting dummy variables in the model at the start of the early PCV period and at the start of the late PCV period. Correlograms used for assessing model adequacy are presented in Supplementary Figure 1.

We used a linear forecast model with time as covariate on the pre-PCV data to estimate DAP rates during the early and late PCV periods, if PCVs had not been introduced. The slopes calculated using this “naive” modeling approach were assigned a value of zero from post-PCV year 4 onward, assuming stabilization of any pre-PCV trends by then, as per Feikin et al [20]. The quantitative DAP rate reductions were defined as actual incident rate ratio (aIRR; mean 2015–2018 vs 2005–2009) and relative IRR (rIRR; mean 2015–2018 vs expected mean 2015–2018). All statistical tests were 2-sided; P < .05 was considered significant. Trends and ratios were considered significant when the null value was not contained within the 95% confidence interval.

All statistical analyses were done using R version 4.0.2 (R Foundation for Statistical Computing; http://www.R-project.org).

RESULTS

During the 13 study years, 1 090 870 prescriptions were dispensed; 617 983 (57%) were for children younger than 2 years.

Dispensed Antibiotic Prescription Rates Before PCV Implementation (2005–2009)

Mean all-DAP rates (per 1000 child-years) were 2590.34, 1393.40, and 1852.10 for ages younger than 2, 2–4, and younger than 5 years, respectively (Table 1). The overall rates were 1.6-fold higher among Bedouin than Jewish children. Higher rates among Bedouin children were observed for each antibiotic category (except for azithromycin) and age group.

Table 1.

Mean Yearly Prescription Rates per 1000 Child-Years During the Pre-PCV Period (July 2005–June 2009) and Late-PCV Period (July 2015–June 2018)

	Mean Yearly Prescription Rate Pre-PCV Period				Mean Yearly Prescription Rate Late-PCV Period
	Bedouin	Jewish	Pa	Total	Bedouin	Jewish	P	Total
Children <5 years
All antibiotics	2045.8 ± 106.5	1628.6 ± 25.9	.003	1852.1 ± 64.7	1523.6 ± 21.9	1507 ± 65.7	.712	1507.4 ± 30.8
Amoxicillin/amoxicillin-clavulanate	1566.9 ± 90.0	1223.1 ± 9.1	.004	1398.8 ± 56.5	1110.5 ± 16.6	1087.5 ± 61.4	.589	1098.6 ± 32.7
Azithromycin	234.6 ± 10.4	252.3 ± 24.8	.258	254.2 ± 14.3	293.8 ± 11.2	305.4 ± 12.4	.296	294.4 ± 5.6
OSGCs	171.9 ± 8.9	114.5 ± 3.0	<.001	142.7 ± 5.0	95.2 ± 11.8	99 ± 8.8	.676	96.1 ± 9.9
Other antibiotics	72.4 ± 20.7	38.8 ± 11.1	.039	56.4 ± 16.1	24.1 ± 1.8	15 ± 1.2	.003	18.3 ± 1.1
Children <2 years
All antibiotics	3122.1 ± 144.8	2048.7 ± 13.3	.001	2590.3 ± 91.5	2025.7 ± 44.6	1771.4 ± 61.4	.006	1876.1 ± 47.8
Amoxicillin/amoxicillin-clavulanate	2395.3 ± 119.9	1590.1 ± 26.2	.001	1991.1 ± 76.5	1410.2 ± 66.6	1270 ± 57.2	.052	1337.9 ± 51.9
Azithromycin	404.5 ± 27.9	278.9 ± 24.8	.001	348.0 ± 21.7	481.4 ± 17.2	370.9 ± 13.8	.001	408.5 ± 4.6
OSGCs	204.3 ± 16.6	123.5 ± 3.3	.002	163.2 ± 8.3	103.0 ± 8.0	113.2 ± 9.9	.244	107.2 ± 8.4
Other antibiotics	118.0 ± 34.1	56.1 ± 17.0	.027	88.0 ± 25.1	31.1 ± 3.1	17.3 ± 0.8	.012	22.4 ± 1.1
Children 2–4 years
All antibiotics	1397.6 ± 91.3	1356.8 ± 47.1	.467	1393.4 ± 56.2	1196.6 ± 54	1333.6 ± 74.8	.068	1267.8 ± 39.9
Amoxicillin/amoxicillin-clavulanate	1067.9 ± 73.6	985.6 ± 20.1	.108	1030.7 ± 47.4	915.6 ± 35.5	967.8 ± 68.1	.324	943.1 ± 34.3
Azithromycin	132.2 ± 4.8	235 ± 25.8	.003	195.8 ± 12.6	171.4 ± 9.6	262.5 ± 12.7	.001	220.1 ± 10.0
OSGCs	152.5 ± 6.0	108.6 ± 4.2	<.001	130.1 ± 4.3	90.1 ± 14.3	89.7 ± 8.1	.975	88.9 ± 11.0
Other antibiotics	44.9 ± 13.3	27.7 ± 7.8	.077	36.7 ± 11.0	19.6 ± 1.4	13.5 ± 1.7	.010	15.7 ± 1.3

Abbreviations: OSGC, oral second-generation cephalosporin; PCV, pneumococcal conjugate vaccine.

^aP value mean rates in Bedouin vs Jewish children (Student’s t test).

The most frequently prescribed antibiotic category was amoxicillin/amoxicillin-clavulanate (75% of DAPs) followed by azithromycin (14%) and OSGCs (8%); all-other antibiotics constituted only 3%. The proportions were similar in all ages and ethnic groups.

In general, DAP rates were higher during the cold season than the warm season; however, these differences were less accentuated among Bedouin than among Jewish children (Figures 1 and 2, Supplementary Figures 2–5). Among Bedouin children, azithromycin rates were similar in both seasons; OSGC rates were higher in the warm months among Bedouin but not among Jewish children.

Figure 1.

Mean monthly dispensed oral antibiotics prescription rates (per 1000 child-years), warm season (June–September) vs cold season (November–February), in children <5 years (July 2005–June 2018). The numbers in each group represent incidence rate ratios (and 95% CIs) for the late PCV vs pre-PCV periods. Abbreviations: CI, confidence interval; PCV, pneumococcal conjugate vaccine.

Figure 2.

Mean yearly dispensed oral antibiotics prescription rates (per 1000 child-years), warm season (June–September) vs cold season (November–February), in Bedouin and Jewish children <5 years old (July 2005–June 2018). The numbers in each group represent incidence rate ratios (and 95% CIs) for the late PCV vs pre-PCV periods. Abbreviations: CI, confidence interval; PCV, pneumococcal conjugate vaccine.

In the pre-PCV period, all-DAP, amoxicillin/amoxicillin-clavulanate, and OSGC rates showed slight, but significant negative trends (Figures 3 and 4, Table 2, Supplementary Table 2, Supplementary Figures 2–5). This was mainly derived from the trends in Bedouin children, without clear trends in Jewish children. No significant trend was observed for azithromycin, except for a slight but significant increase among Bedouin children younger than 2 years. For all-other antibiotics, a sharp and significant reduction was observed, reaching very low rates.

Figure 3.

Monthly dispensed oral antibiotic prescription rates (per 1000 child-years) in all children <5 years (July 2005–June 2018). Abbreviations: CI, confidence interval; PCV, pneumococcal conjugate vaccine.

Figure 4.

Monthly dispensed oral antibiotic prescription rates (per 1000 child-years) in Jewish and Bedouin children <5 years (July 2005–June 2018). Abbreviations: CI, confidence interval; PCV, pneumococcal conjugate vaccine.

Table 2.

Trends of Antibiotic Prescription Rates in the Pre-PCV (July 2005–June 2009), Early-PCV (July 2010–June 2015), and Late-PCV (July 2015–June 2018) Periods in All Children Younger Than 5 Years Old

	Pre-PCV Trend (95% CI)	Early-PCV Trend (95% CI)	P Valuea : Early-PCV vs Pre-PCV	Late-PCV Trend (95% CI)	P Valuea: Late-PCV vs Early-PCV13
All childrenb
All antibiotics	−.281 (−.479; −.083)	−.474 (−.906; −.043)	.102	−.057 (−.809; .695)	.011
Amoxicillin/amoxicillin-clavulanate	−.189 (−.354; −.024)	−.42 (−.778; −.061)	.02	−.114 (−.74; .511)	.025
Azithromycin	0.026 (−.023; .074)	.011 (−.094; .115)	.605	.032 (−.15; .215)	.589
OSGCs	−.026 (−.040; −.011)	−.051 (−.083; −.019)	.004	.032 (−.024; .088)	<.001
Other antibioticsc	−.087 (−.096; −.078)	−.007 (−.026; .011)	<.001	−.016 (−.049; .017)	.221
Bedouin children
All antibiotics	−.441 (-0.679; -0.203)	-0.824 (-1.342; -0.306)	.008	.013 (−.891; .916)	<.001
Amoxicillin/amoxicillin-clavulanate	−.335 (−.534; −.136)	-0.659 (-1.092; -0.226)	.007	−.180 (−.936; .575)	.004
Azithromycin	0.043 (−.008; .094)	−.047 (−.158; .064)	.003	.165 (−.028; .359)	<.001
OSGCs	−.049 (−.071; −.028)	−.088 (−.134; −.041)	.003	.015 (−.067; .096)	<.001
Other antibiotics	−.113 (−.126; −.1)	−.012 (−.041; .016)	<.001	−.010 (−.06; .04)	.809
Jewish children
All antibiotics	−.010 (−.269; .248)	−.092 (−.654; .47)	.597	−.144 (−1.125; .836)	.805
Amoxicillin/amoxicillin-clavulanate	.039 (−.163; .242)	−.168 (−.608; .272)	.087	−.136 (−.904; .632)	.847
Azithromycin	.002 (−.064; .068)	.085 (−.059; .228)	.037	−.024 (−.275; .226)	.045
OSGCs	−.003 (−.024; .018)	−.011 (−.056; .035)	.538	.041 (−.038; .12)	.003
Other antibiotics	−.058 (−.068; −.049)	.000 (−.020; .021)	<.001	−.027 (−.063; .009)	.001

Abbreviations: CI, confidence interval; OSGC, oral second-generation cephalosporin; PCV, pneumococcal conjugate vaccine.

^aP value for change in trends (segmented regression).

^bIncluding Jewish, Bedouin, and mixed clinics.

^cClarithromycin, clindamycin, erythromycin, trimethoprim/sulfamethoxazole.

Dispensed Antibiotic Prescription Rates Following PCV Implementation

Shortly following PCV implementation, all-DAP and amoxicillin/amoxicillin-clavulanate rates declined, generally reaching a new reduced and stable level within 5 years (Figures 3 and 4, Table 2, Supplementary Table 2, Supplementary Figures 6–9). These trends reached statistical significance for Bedouin children younger than 2 years. Notably, the higher the initial all-DAP rate was, the steeper the decline. The post–PCV implementation trends were similar in the cold and warm seasons (Figure 1).

The trends for azithromycin rates were complex. Among Bedouin children, the pre-PCV increasing trend was initially reversed in children younger than 2 years. However, 5 years later, an increasing trend was observed among both age groups. In contrast, among Jewish children, in whom no pre-PCV trend was observed, an increasing trend was observed during the first 5 post-PCV years, reaching significance for children younger than 2 years. Five years post–PCV implementation, the increasing trend was halted and rates were stabilized (significantly in children <2 years). In terms of seasonality, unlike the similarity between cold and warm seasons for all-DAP and amoxicillin/amoxicillin-clavulanate rates, in the warm season, azithromycin rates showed neither a significant increase nor decrease during the entire study duration (except for an increase in Jewish children <2 years) (Figures 1 and 2, Supplementary Figures 2–5). Thus, the trends described above were almost exclusively due to those occurring during the cold season.

The second complex category was OSGCs. The pre-PCV decreasing trend observed in Bedouin children continued in the first 5 years following implementation. In Jewish children, no trend was observed either in the pre-PCV period or in the first 5 years postimplementation. However, thereafter, a moderately increasing trend was observed among both populations (except for Bedouin children <2 years, whose rates were stable). Although an overall reduction was seen in both seasons, the main post-PCV reduction occurred during the warm season, resulting in similar rates during both seasons for all ages and ethnic groups.

The steep pre-PCV decline in all-other DAPs continued after implementation, resulting in rates of less than 3 DAPs per 1000 in all groups.

Importantly, despite significantly higher pre-PCV DAP rates among Bedouin versus Jewish children, no, or only slight, differences in rates were found between the populations in the late PCV13 period, thus closing most of the gap between the 2 groups (Table 1).

Estimated Quantitative Changes in Dispensed Antibiotic Prescriptions Following PCV Implementation

We compared mean annual antibiotic DAP rates (per 1000 child-years) in the last 3 years of the study (a period with generally stable rates) with the mean annual rates before PCV. Each comparison was done in 2 ways: (1) aIRR, DAP rate ratio in 2015–2018 versus 2005–2009, and (2) rIRR, DAP rate ratio during 2015–2018 versus expected rate during 2015–2018, taking into account pre-PCV trends (Table 3, Supplementary Table 3).

Table 3.

Mean Yearly Actual and Relative Prescription Rate Ratios and Their Respective Change in Number of Prescriptions per 1000 Child-Years in Children Younger Than 5 Years Old

	aIRRa (95% CI)	Actual Mean Yearly Change (95% CI)	rIRRb (95% CI)	Relative Mean Yearly Change (95% CI)
All childrenc
All antibiotics	.81 (.81; .82)	−344.7 (−358.4; −330.9)	.93 (.92; .94)	−110.4 (−123.7; −96.9)
Amoxicillin/amoxicillin-clavulanate	.79 (.78; .79)	−300.2 (−311.8; −288.6)	.89 (.88; .89)	−142.3 (−153.6; −130.9)
Azithromycin	1.16 (1.13; 1.18)	40.2 (33.6; 46.9)	1.06 (1.04; 1.08)	16.5 (10.5; 22.7)
OSGCs	.67 (.65; .7)	−46.6 (−49.9; −43.3)	.80 (.77; .83)	−24.4 (−27.6; −21)
Other antibioticsd	.33 (.3; .35)	−38.0 (−39.3; −36.7)	NA	NA
Bedouin children
All antibiotics	.74 (.74; .75)	−522.2 (−541.8; −502.4)	.91 (.9; .93)	−144.5 (−164.7; −124.1)
Amoxicillin/amoxicillin-clavulanate	.71 (.7; .72)	−456.5 (−473; −439.7)	.87 (.86; .88)	−165.7 (−182.7; −148.4)
Azithromycin	1.25 (1.21; 1.3)	59.3 (49.6; 69.3)	1.08 (1.05; 1.12)	22.4 (13.3; 31.9)
OSGCs	.55 (.53; .58)	−76.8 (−81.3; −72.0)	.74 (.70; .78)	−33.3 (−38.1; −28.4)
Other antibiotics	.33 (.3; .36)	−48.3 (−50.3; −46)	NA	NA
Jewish children
All antibiotics	.93 (.91; .94)	−121.6 (−143.2; −99.7)	.92 (.91; .94)	−124.3 (−143.8; −104.5)
Amoxicillin/amoxicillin-clavulanate	.89 (.87; .9)	−135.5 (−153.7; −117.1)	.86 (.85; .87)	−180.1 (−196.3; −163.5)
Azithromycin	1.21 (1.17; 1.25)	53.2 (42.7; 64)	1.20 (1.16; 1.23)	50.3 (41.1; 59.9)
OSGCs	.87 (.82; .91)	−15.4 (−20.8; −9.8)	.88 (.84; .93)	−13.0 (−17.9; −7.9)
Other antibiotics	.39 (.34; .44)	−23.8 (−25.5; −21.9)	NA	NA

Abbreviations: aIRR, actual incidence rate ratio; CI, confidence interval; NA, ; OSGC, oral second-generation cephalosporin; PCV, pneumococcal conjugate vaccine; rIRR, relative incidence rate ratio.

^aaIRR: mean rate 2015–2018 vs mean rate 2005–2009.

^brIRR: 2015–2018 vs expected mean 2015–2018, extrapolated from the pre-PCV trends.

^cIncluding Jewish, Bedouin, and mixed clinics.

^dClarithromycin, clindamycin, erythromycin, trimethoprim/sulfamethoxazole.

For both age and ethnic groups, all aIRRs were less than 1.0, except for azithromycin. For all-DAPs, amoxicillin/amoxicillin-clavulanate, and OSGCs, the aIRRs were significantly lower among Bedouin children than Jewish children. As expected, since the rIRR depended on pre-PCV trend, the results were more complex. For all-DAP and amoxicillin/amoxicillin-clavulanate, the rIRR was less than 1.0 in children younger than 2 years and more than 1.0 in children aged 2–4 years in both populations; for azithromycin, the rIRR was more than 1.0, except in Bedouin children younger than 2 years; for OSGCs, all rIRRs were greater than 1.0; and for all-DAP and amoxicillin/amoxicillin-clavulanate, rIRRs were significantly lower among Bedouin than Jewish children younger than 2 years.

Among all children younger than 5 years, the annual mean reduction (per 1000 child-years) of all-DAPs ranged between 344.7 (aIRR) and 110.4 (rIRR). For amoxicillin/amoxicillin-clavulanate, the respective reductions were 300.2 and 142.3 (for children <2 years, 653.2 and 387.2). For children 2–4 years, the aIRR resulted in a reduction of 125.6 and 87.7 of all-DAP and amoxicillin/amoxicillin-clavulanate prescriptions, respectively. However, the rIRR resulted in an increase of 72.6 and 39.3 of all-DAP and amoxicillin/amoxicillin-clavulanate prescriptions, respectively (Table 3, Supplementary Table 3).

DISCUSSION

Following PCV7/PCV13 implementation, a significant reduction in outpatient DAP rates for young children was observed. During the early postimplementation period, a decreasing trend was generally observed, stabilizing after approximately 5 years (late PCV13 period). This pattern was expected, given the previously observed dynamics of RTIs in young children following PCV implementation in Israel [8–11, 13, 14] and the fact that RTIs are the main reason for antibiotic prescriptions [4]. Supporting the association between PCV impact on RTIs and antibiotic consumption is their common seasonality, their highest rates in children younger than 2 years, and the stabilization of the trends after an initial reduction during early postimplementation [8–11, 13, 14]. Furthermore, the most commonly used antibiotics for RTIs in young children are amoxicillin and amoxicillin-clavulanate [21, 22]. Indeed, amoxicillin/amoxicillin-clavulanate contributed 75% of all DAPs in our study, with the highest rates during the cold (respiratory) season, and they showed the most impressive reduction after PCV implementation. Because of this predominance, the all-DAP trends closely followed this category. Despite the clear pattern of all-DAP and amoxicillin/amoxicillin-clavulanate rate trends, the behaviors of other drug categories were more complex and deserve further discussion.

Azithromycin was the only drug with increasing trends following implementation. Furthermore, this occurred only during the cold season. However, as with most drugs, azithromycin rates tended to stabilize. We were not able to explain these findings and need to accept that there is no proof that PCV implementation had a major impact on azithromycin prescription rates. Although azithromycin constituted only 14% of all DAPs, its increasing consumption rate is worrisome.

The only category without higher consumption during cold months was OSGCs. Among Jewish children, no seasonality was observed, while among Bedouin children, consumption was markedly higher during the warm months. This could be explained, at least in part, by the common use of OSGCs for skin and soft tissue infections, which are more prevalent in the warm months [23]. Nevertheless, the use of OSGCs for respiratory infections is not uncommon. The reduction in OSGC consumption, being already significant before PCV implementation in Bedouin children, the unique seasonality pattern, and the marked post-PCV reduction in the warm months among Bedouin children suggest that PCVs might not have played a major role in OSGC reduction.

The pre-PCV “all-other DAP” category rates were already very low, and further decreased after implementation, reaching negligible levels. Trimethoprim/sulfamethoxazole was almost eliminated due to high nonsusceptibility rates among most pathogens; clarithromycin and erythromycin were replaced by azithromycin; and clindamycin was only rarely used.

Because of the lower socioeconomic and living conditions, and the higher rates of pneumococcal carriage and RTIs among Bedouin compared with Jewish children, higher pre-PCV DAP rates among Bedouin children were expected. However, following PCV implementation, reductions were more accentuated among Bedouin children in all age groups and with most drugs, resulting in almost complete closure of the gap between the 2 ethnic groups.

A post-PCV reduction in antibiotic use was observed in previous studies using different methods and PCVs [12], examining only short-term post–PCV implementation, held in different populations, and examining various outcomes, such as overall antibiotics [2, 24], otitis media–associated antibiotics [25–27], and antibiotics for overall RTIs [28]. Furthermore, currently, all studies that reported incidence rates measured aIRRs by comparing post- with pre-PCV rates [2, 24, 26, 27] and most did not take into consideration the frequently already declining trends before the intervention.

The absolute DAP rate reduction (aIRR) is relatively easy to calculate for each drug category and for all-DAP rates, and resulting figures can be extrapolated. However, attempting to determine the causative role of PCV implementation is more complex and, consequently, any results will be imprecise due to potential biases. We attempted to overcome this, at least in part, by considering the pre-PCV trends (rIRR). However, the main difficulty is the inability to predict how, and for what duration, the pre-PCV trend will persist. We therefore arbitrarily considered the value of the trend slope to be zero from year 4, as was done by others [20]. We believe, therefore, that any attempt to evaluate the absolute impact of PCVs on DAP rates is at least imprecise, if not inaccurate.

It is therefore difficult to compare our results with those obtained by others. Still, using the aIRR, the reduction observed in our study seems somewhat greater than in most published studies, potentially because of higher baseline rates. We estimate that the overall DAP reduction in children younger than 5 years lies between 110.4 (rIRR) and 344.7 (aIRR) per 1000 child-years. The respective figures for children younger than 2 years were 346.2 to 714.2.

Continuous antibiotic pressure in the community deriving from excessive consumption, coupled with the nasopharyngeal pneumococcal replacement, resulted in increasing rate of nonsusceptibility among non-PCV pneumococcal serotypes [29, 30]. The overall DAP reduction rate translates into reduced pressure on the entire body microbiota. However, consumption still remains high. Therefore, although PCV implementation is an important tool against antimicrobial resistance, efforts to develop other tools, such as improvement in stewardship programs, must be continued.

The current study has several strengths. First, it is a long-term, population-based study that includes several pre–PCV implementation cohorts and both early and late postimplementation cohorts. Second, our analysis is the first to include adjustments for time postvaccination, in addition to age and season. Third, we studied 2 ethnic populations differing in many aspects, with minimal contact between them; the generally similar post-PCV DAP trends add strength to our findings. Furthermore, we documented a reduction in gaps between the populations following vaccination. Fourth, we were able to compare the post-PCV dynamics with those in published prospective studies on IPDs and acute RTIs in the same population [7–11], adding plausibility to the potential role of PCV in the reduction in DAP rates.

Our study also has several limitations. First, we did not have individual diagnoses attached to each prescription, making it impossible to choose appropriate controls by comparing DAP rates with nonrespiratory diagnoses. However, the ability to analyze the data by monthly rates, season rates, and by specific drug category made it plausible that most DAP rate reductions were for RTIs based on the following arguments: (1) the main reduction was in amoxicillin/amoxicillin-clavulanate, the first-line antibiotics for RTIs, and the reduction was much more pronounced during the cold season; (2) the rates of azithromycin, a drug used for many indications beyond RTIs, were not reduced; and (3) the rates of OSGCs, a category often used for skin and soft tissue infections, were higher in the warm season, during which RTIs are at the lowest rates. Second, we could not rule out other factors that could potentially play a role in the post–PCV implementation period. Although no significant changes in guidelines for RTI occurred, and no limitations for the use of specific antibiotics were imposed by the healthcare providers, factors such as improvement in hygienic conditions and economic status and implementation of the rotavirus vaccine in 2011 could have played a role to some degree.

In summary, DAP rates declined abruptly and significantly in children younger than 5 years after PCV implementation in Israel, reaching a new reduced and stable level within 5 years. Trends were similar in both Jewish and Bedouin children and in all age groups, largely driven by the dynamics of amoxicillin/amoxicillin-clavulanate, the main antibiotic category recommended for RTIs. The declining trends were observed in parallel to the post-PCV reductions in RTIs, supporting the causative role of PCVs.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Disclaimer. The funding source had no role in any aspect of the submitted article.

Financial support. This work was supported in part by Pfizer Inc. (grant number 0887X1–4603).

Potential conflicts of interest. D. G. has received grants from Merck, Sharp & Dohme, paid to their institution; has received consulting fees as a scientific consultant for GlaxoSmithKline, Merck, Sharp & Dohme and Pfizer; and has received personal fees as a speaker for GlaxoSmithKline, Merck, Sharp & Dohme, and Pfizer. S. B.-S. has received speakers’ fees and a grant from Pfizer (grant to their institution, Ben-Gurion University of the Negev); has received consulting fees as a scientific consultant for Merck, Sharp & Dohme; and has received personal fees as a speaker for GlaxoSmithKline, and Merck, Sharp & Dohme. R. D. has received grants/research support from Pfizer, Merck, Sharp & Dohme, and Medimmune (to their institution, Ben-Gurion University of the Negev); has received consulting fees as a scientific consultant for Pfizer, MeMed, Merck, Sharp & Dohme, and Biondvax; has received payment for expert testimony from Pfizer; has served on advisory boards of Pfizer, Merck, Sharp & Dohme, and Biondvax; and has been a paid speaker for Pfizer. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.