Antibiotic prescriptions rates for acute respiratory tract infections in the United States ambulatory settings, 1995–2006

Abstract

Context

During the 1990s, antibiotic prescriptions for acute respiratory tract infections (ARTI) declined in the United States. The sustainability of those changes is unknown.

Objective

To assess trends in antibiotic prescriptions for ARTI

Design, Setting and Participants

National Ambulatory Medical Care Survey and National Hospital Ambulatory Medical Care Survey data (1995–2006) were used to examine trends in antibiotic prescription rates by antibiotic indication and class. Annual survey data and census denominators were combined in 2-year intervals for rate calculations.

Main outcome

National annual visit and antibiotic prescription rates for ARTI including otitis media (OM) and for non-ARTI.

Results

Among children aged <5 years, annual ARTI visit rates declined 17% (95% CI: 9–24) from 1883 per 1000 population in 1995–1996 to 1560 per 1000 in 2005–2006, primarily due to a 33% (95% CI: 22–43) decline in OM visit rates (from 950 to 634 per 1000, respectively). This decline was accompanied by a 36% (95% CI: 26–45) decline in ARTI-associated antibiotic prescriptions (from 1216 to 779 per 1000). Among persons aged ≥5 years, ARTI visit rates remained stable but associated antibiotic prescription rates declined by 18% (95% CI: 6–29), from 178 to 146 per 1000. Antibiotic prescription rates for non-OM ARTI for which antibiotics are rarely indicated declined by 41% (95% CI: 22–55) and 24% (95% CI: 10–37) among persons aged <5 and ≥5 years, respectively. Overall, ARTI-associated prescription rates for penicillin, cephalosporins, sulfonamides and tetracyclines declined. Prescription rates for azithromycin increased and it became the most commonly prescribed macrolide for ARTI and OM (10% of OM visits). Among adults, quinolone prescriptions increased.

Conclusion

Overall antibiotic prescription rates for ARTI declined, associated with fewer OM visits in children aged <5 years and with fewer prescriptions for ARTI for which antibiotics are rarely indicated. However, prescription rates for broad spectrum antibiotics increased significantly.

INTRODUCTION

Infections caused by antibiotic-resistant microorganisms are associated with increased morbidity, mortality and a substantial economic burden.¹ Antibiotic use creates selective pressure for the emergence of antibiotic-resistant bacteria.^2–4 During the past decade, initiatives in the United States promoted the judicious use of antibiotics,^{5, 6} particularly for acute respiratory tract infections (ARTI). ARTI are a common cause of healthcare encounters and antibiotic prescriptions, especially in young children. In the late 1990s, antibiotic prescription rates in both children and adults declined.^7–10 Notably, these declines were initially accompanied by increased prescription of broad spectrum antibiotics.^9–11

Other interventions may also have reduced antibiotic prescriptions and limited the spread of antibiotic resistance. Routine infant immunization with a seven-valent pneumococcal conjugate vaccine (PCV7) in the United States resulted in declines in rates of invasive disease due to antibiotic-resistant Streptococcus pneumoniae.¹² Data from CDC’s laboratory and population-based surveillance, indicates that declines in pneumococcal isolates resistant to penicillin, sulfonamides and selected cephalosporins has been sustained, but the proportion of erythromycin-resistant pneumococcus increased after the initial declines.¹³ In the late 1990s, increasing rates of macrolide-resistant invasive pneumococcal disease were temporally associated with increasing use of macrolides, especially azithromycin.¹⁴

Recent measurements of antibiotic prescription patterns in the United States are unavailable. This article assesses national trends in antibiotic prescriptions for ARTI in ambulatory settings.

METHODS

Sources of information

We analyzed annual data from the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS) from 1995 to 2006. NAMCS gathers data on a nationally representative sample of visits to office-based physicians engaged in direct patient care, whereas NHAMCS collects data on a representative sample of visits to hospital-based emergency departments and outpatient clinics. These surveys record patient demographics, symptoms, procedures, diagnoses, and prescribed medications for a systematic sample of visits. During the study period, physician participation rates in NAMCS ranged from 59%–73%, and the un-weighted item non-response rates were typically ≤5%.¹⁵ Rates of hospital participation in NHAMCS ranged from 90%–98% and the un-weighted item non-response rates were typically ≤5%.¹⁶

Up to three diagnoses per visit are recorded in NAMCS and NHAMCS, using ICD9-CM codes. ARTI visits included those with any listed diagnosis of otitis media (OM), bronchitis, bronchiolitis, pneumonia, sinusitis, pharyngitis, tonsillitis, laryngitis, tracheitis, nasopharyngitis and mastoiditis (ICD9-CM: 381.*–383.*, 460.*–466.*, 480.*–487.*, 490.*). Rates were also calculated for OM, the most common reason for physician visits and antibiotic prescriptions in young children⁸ (ICD9-CM: 381.*–382.*). For specific analyses, we also defined mutually exclusive groups: 1) ARTI for which antibiotics are often indicated (OM, mastoiditis, sinusitis, pharyngitis, tonsillitis and non-viral pneumonia);^17–19 and, 2) ARTI for which antibiotics are rarely indicated (acute nasopharyngitis, laryngitis, unspecified upper respiratory infection, bronchitis, bronchiolitis, viral pneumonia and influenza).^17–19 We also explored the concurrent presence of common non-ARTI diseases that could result in an antibiotic prescription such as urinary tract infections (ICD9-CM: 599.0, 595.0, 595.9) and soft tissue infections (ICD9-CM: 680.*–682.*).^{8, 20}

Antibiotic classes included in the study were penicillins, cephalosporins, sulfonamides, tetracyclines, macrolides (including azalides), lincosamides and quinolones. Antibiotic prescriptions were identified using the National Drug Code directory, generic names, drug names, and the Multum classification of therapeutic classes.^{15, 16} Visits that resulted in the prescription of ≥2 antibiotic classes contributed one observation to each class and rates of specific antibiotic prescriptions during ARTI visits were estimated. Since use of some antibiotics was rare, estimates for sulfonamides and tetracyclines, and lincosamides and macrolides were analyzed together. Quinolone exposures were too rare to estimate rates for children. Separate estimates were obtained for azithromycin, amoxicillin and amoxicillin/clavulanic acid. Azithromycin exposures were too rare to estimate prescription rates for OM for children <5 years until 1997.

Statistical analysis

For a comprehensive assessment of United States ambulatory settings, NAMCS and NHAMCS data were combined, and to obtain robust estimates, annual survey data were combined in 2-year intervals.⁸ All reported estimates were based on ≥30 un-weighted observations and relative standard errors <30%.^{15, 16}

Since children aged <5 years had the highest visit rates for ARTI and antibiotic prescriptions⁸ and early activities promoting judicious use of antibiotics were primarily focused on them,²¹ our analyses were stratified as children aged <5 years and older persons. To assess differential effects of age and the aging of the United States population²², persons aged ≥5 years were further stratified into 5–17, 18–39, 40–64, 65–79 and 80 or more years.

Survey data were weighted to produce national estimates, with weights accounting for selection probabilities, adjustment for non response, population ratio adjustments, and weight smoothing.^{15, 16} Weighted observations represented rate numerators and denominators were obtained from the Census Bureau. For each two-year period, the population was calculated by averaging the two year population estimates. To estimate average annual rates and rate ratios, we fitted Poisson regression models to the survey data. Census population estimates for each time period and age groups provided an approximation to the person-time for each period and were distributed according to the aggregated weight factor of the national surveys, representing the offset term in the models.^{22, 23}

Our models included terms for age, time and an age-time interaction term and accounted for the complex survey sampling design. Per our planned analyses, we compared the first (1995–1996) with the most recent (2005–2006) observation period, using rate ratios (RR). RR were estimated using linear predictors from the regression models. Rate differences (RD), and their 95% confidence intervals were also calculated.

We also assessed changes in prescribing practices by calculating the proportion of visits that resulted in antibiotic prescriptions.⁸ Proportions were estimated accounting for the survey design and trends in proportions were tested using weighted least square regressions.⁸ All reported p values were two-tailed and accounted for the complex survey design. A p value of <0.05 or the exclusion of 1 from the RR 95% confidence intervals indicated statistical significance. Data were analyzed using SAS 9.1 and Stata 10.0 survey packages. This study was considered exempt from review by the Institutional Review Boards of Vanderbilt University and CDC.

RESULTS

Visit characteristics

During the 12-year study period, the surveys estimated 6.2 billion ambulatory visits in the United States. Overall, 59% of patients were female and 84% were white. Approximately 82% of visits were to physician practices, 8% to hospital-based clinics, and 10% to emergency departments. Overall, antibiotics were prescribed in 13% of all ambulatory visits. ARTI visits accounted for 10% of all visits and for 44% of all antibiotic prescribing.

Children aged <5 years accounted for 9% of all visits and 34% of their visits were ARTI-associated. In this group, antibiotics were prescribed in 26% of all visits and in 56% of ARTI visits. Persons aged ≥5 years accounted for 91% of all visits and ARTI represented 8% of their visits. In this group, antibiotics were prescribed in 12% of all visits and in 58% of ARTI visits.

Children aged <5 years

ARTI visits

Overall annual visit rates were 4773 per 1000 population in 1995–1996 and remained stable during the study period. While ARTI-associated visit rates declined from by 17% (RR: 0.83, 95% CI: 0.76–0.91), non-ARTI visit rates increased by 25% (RR: 1.25, 95% CI: 1.19–1.31) (Table 1 and Figure 1). Concomitant conditions as a possible reason for antibiotic prescriptions associated with ARTI visits such as urinary tract infections (0.26% of ARTI visits) or soft tissue infections (0.06%) were uncommon.

Table 1. Rates and rate ratios of ambulatory visits and antibiotic prescriptions, United States 1995–2006.

Rates (per 1,000 population) and rate ratios of Ambulatory visit and antibiotic prescriptions, United States 1995–2006

All visits	1995–1996	2005–2006	RD 95% CI	RR 95% CI
<5 years	4773	5174	402 (−149 − 1017)	1.08 (0.97 − 1.21)
5 to 17 years	2085	2436	351 (174 − 541)	1.17 (1.08 − 1.26)
18 to 49 years	2689	2977	287 (129 − 455)	1.11 (1.05 − 1.17)
50 to 64 years	3793	4627	834 (615 − 1065)	1.22 (1.16 − 1.28)
65 to 79 years	5746	7166	1420 (1022 − 1841)	1.25 (1.18 − 1.32)
80 or more years	6103	7705	1602 (1086 − 2156)	1.26 (1.18 − 1.35)
ARTI visits
<5 years	1883	1560	−322 (−461 − −171)	0.83 (0.76 − 0.91)
5 to 17 years	490	459	−31 (−84 − 29)	0.94 (0.83 − 1.06)
18 to 49 years	220	206	−14 (−40 − 14)	0.93 (0.82 − 1.07)
50 to 64 years	207	246	39 (2 − 83)	1.19 (1.01 − 1.4)
65 to 79 years	282	266	−15 (−66 − 46)	0.95 (0.77 − 1.16)
80 or more years	286	232	−54 (−111 − 22)	0.81 (0.61 − 1.08)
Non ARTI visits
<5 years	2890	3614	724 (547 − 910)	1.25 (1.19 − 1.31)
5 to 17 years	1595	1977	382 (317 − 449)	1.24 (1.2 − 1.28)
18 to 49 years	2469	2771	302 (273 − 331)	1.12 (1.11 − 1.13)
50 to 64 years	3586	4381	795 (755 − 836)	1.22 (1.21 − 1.23)
65 to 79 years	5464	6899	1435 (1373 − 1498)	1.26 (1.25 − 1.27)
80 or more years	5817	7473	1656 (1570 − 1743)	1.28 (1.27 − 1.3)
ATB in all visits
<5 years	1552	1128	−424 (−551 − −281)	0.73 (0.64 − 0.82)
5 to 17 years	552	495	−57 (−112 − 5)	0.9 (0.8 − 1.01)
18 to 49 years	385	396	10 (−25 − 49)	1.03 (0.93 − 1.13)
50 to 64 years	362	445	82 (30 − 141)	1.23 (1.08 − 1.39)
65 to 79 years	461	635	174 (87 − 274)	1.38 (1.19 − 1.59)
80 or more years	543	696	153 (27 − 307)	1.28 (1.05 − 1.57)
ATB in ARTI visits
<5 years	1216	779	−437 (−542 − −315)	0.64 (0.55 − 0.74)
5 to 17 years	314	219	−95 (−127 − −58)	0.7 (0.6 − 0.82)
18 to 49 years	149	124	−25 (−45 − −1)	0.83 (0.7 − 0.99)
50 to 64 years	120	141	21 (−6 − 54)	1.17 (0.95 − 1.45)
65 to 79 years	145	134	−11 (−42 − 30)	0.93 (0.71 − 1.21)
80 or more years	154	110	−44 (−77 − 4)	0.71 (0.5 − 1.02)
ATB in non ARTI visits
<5 years	336	349	13 (−47 − 84)	1.04 (0.86 − 1.25)
5 to 17 years	238	276	38 (2 − 80)	1.16 (1.01 − 1.34)
18 to 49 years	236	271	35 (12 − 61)	1.15 (1.05 − 1.26)
50 to 64 years	243	304	62 (22 − 106)	1.25 (1.09 − 1.44)
65 to 79 years	316	501	185 (105 − 279)	1.58 (1.33 − 1.88)
80 or more years	389	586	197 (77 − 347)	1.51 (1.2 − 1.89)

^*ARTI, acute respiratory tract infections; ATB, antibiotics; RD, Rate difference; RR, rate ratio; CI, confidence interval.

Data are from NAMCS and NHAMCS. Rates are average annual rates for each time period.

Figure 1.

National rates of ambulatory visits and antibiotic prescriptions, United States, 1995–2006

Footnote: ARTI, acute respiratory tract infections, ATB, antibiotics. White markers represent children aged <5 years, black markers represent persons aged ≥5 years. Error bars represent 95% confidence intervals. Data are from NAMCS and NHAMCS. Rates are average annual rates for each time period.

ARTI-associated antibiotic prescriptions

Annual antibiotic prescription rates declined by 27% or 424 per 1000 children (95% CI: 281–551). This decline was due to a 36% (RR: 0.64, 95% CI: 0.55–0.74) reduction in ARTI-associated antibiotic prescriptions. Antibiotic prescription rates for non-ARTI visits remained stable (Table 1). The overall proportion of visits resulting in antibiotic prescriptions declined from 33% in 1995–1996 to 22% in 2005–2006 (p=0.008), due to reductions from 65% to 50% (p=0.011) in the proportion of ARTI visits resulting in antibiotics. The proportion of non-ARTI visits resulting in antibiotic prescriptions remained stable (p=0.157).

In physician practices, the proportion of ARTI visits that resulted in antibiotics declined from 65% to 48% (p=0.009); no significant declines were observed in emergency department visits (from 63% to 55%, p=0.115) or in outpatient clinics (from 60% to 54%, p=0.124).

Rates of ARTI visits resulting in penicillin prescriptions declined by 31% (Table 2). Amoxicillin/clavulanic acid prescription rates remained stable whereas amoxicillin prescription rates declined by 37%. Cephalosporin and sulfonamide/tetracycline prescription rates declined by 47% and 91%, respectively. Azithromycin prescription rates for ARTI increased 9-fold from 1995–1996 to 2005–2006 but overall prescription rates for macrolides/lincosamides remained stable (Table 2) mainly due to declines in erythromycin use (data not shown because of unreliable estimates during last periods).

Table 2. Rates, rate differences and rate ratios of antibiotic prescriptions during ARTI visits, United States, 1995–2006.

Rates (per 1,000 population), rate differences and rate ratios of antibiotic prescriptions during ARTI visits, United States 1995–2006

<5 years	1995– 1996	2005– 2006	RD 95% CI	RR 95% CI
Penicillins	653	448	−204 (−274 – −121)	0.69 (0.58 – 0.81)
Amoxicillin	556	351	−205 (−260 – −139)	0.63 (0.53 – 0.75)
Amoxicillin / Clavulanic Acid	81	92	11 (−19 – 55)	1.13 (0.76 – 1.68)
Cephalosporins	315	167	−148 (−192 – −88)	0.53 (0.39 – 0.72)
Lyncosamides/macrolides	160	161	1 (−41 – 57)	1 (0.74 – 1.36)
Azithromycin	16	140	124 (64 – 230)	8.78 (4.99 – 15.44)
Sulfonamides/tetracyclins	204	18	−186 (−193 – −174)	0.09 (0.05 – 0.15)
≥5 years	1995– 1996	2005– 2006	RD 95% CI	RR 95% CI
Penicillins	82	47	−34 (−42 – −25)	0.58 (0.48 – 0.69)
Amoxicillin	64	33	−31 (−37 – −24)	0.51 (0.42 – 0.63)
Amoxicillin / Clavulanic Acid	8	11	3 (1 – 7)	1.44 (1.08 – 1.93)
Cephalosporins	38	22	−15 (−20 – −10)	0.59 (0.47 – 0.74)
Lyncosamides/macrolides	47	58	10 (0 – 23)	1.22 (1.01 – 1.48)
Azithromycin	6	37	31 (22 – 44)	6.2 (4.63 – 8.3)
Sulfonamides/tetracyclins	15	8	−7 (−9 – −3)	0.53 (0.35 – 0.81)
Quinolones	4	21	17 (11 – 24)	5.39 (3.95 – 7.36)

^*ARTI, acute respiratory tract infections; RD, Rate difference; RR, rate ratio; CI, confidence interval.

Data are from NAMCS and NHAMCS. Rates are average annual rates for each time period.

OM in children aged <5 years

Annual OM rates declined by 33% from 950 to 634 per 1000 children from 1995–1996 to 2005–2006 (RR: 0.67, 95% CI: 0.57–0.78). Similarly, rates of antibiotic prescriptions for OM declined by 36% from 760 to 484 per 1000 (RR: 0.64, 95% CI: 0.54–0.75). The proportion of OM visits that resulted in antibiotic prescriptions remained about 80% throughout the study years (p=0.142) (Figure 2A).

Figure 2.

National rates of otitis media and non-otitis media ARTI visits, antibiotic prescriptions and proportion of visits resulting in antibiotic prescriptions, United States, 1995–2006

Footnote: ARTI, acute respiratory tract infections, ATB, antibiotics. White markers represent children aged <5 years, black markers represent persons aged ≥5 years. Error bars represent 95% confidence intervals. ARTI for which antibiotics are often indicated (Column B) included mastoiditis, sinusitis, pharyngitis, tonsillitis and non-viral pneumonia. ARTI for which antibiotics are rarely indicated (Column C) included acute nasopharyngitis, laryngitis, unspecified upper respiratory infection, bronchitis, bronchiolitis, viral pneumonia and influenza. Data are from NAMCS and NHAMCS. Rates are average annual rates for each time period.

Penicillin prescriptions rates for OM declined from 390 to 307 per 1000 6 (RR: 0.79, 95% CI: 0.65–0.96; amoxicillin rates declined from 326 to 233 per 1000 (RR: 0.71, 95% CI: 0.58–0.88); and amoxicillin/clavulanic acid prescription rates remained stable. Cephalosporin prescription rates declined from 211 to 103 per 1000 in 2005–2006 (RR: 0.49, 95% CI: 0.35–0.67). Sulfonamide/tetracycline prescription rates declined from 143 to 11 per 1000 (RR: 0.08, 95% CI: 0.04–0.15), whereas macrolide/lincosamide prescription rates remained stable. Azithromycin prescription rates remained stable from 1997–1998 to 2005–2006 (RR: 1.10, 95% CI: 0.70–1.73); in 2005–2006 it was the most commonly prescribed macrolide for OM with 65 (95% CI: 46–90) prescriptions per 1000 children (approximately 10% of OM visits).

Non-OM ARTI

Overall, rates of antibiotic prescriptions associated with non-OM ARTI visits declined by 35% (RR: 0.65, 95% CI: 0.52–0.81). Rates of non-OM ARTI for which antibiotics are often indicated; rates of antibiotic prescription associated with these visits; and the proportion of these visits resulting in an antibiotic prescription did not change significantly throughout the study years. (Figure 2B). Rates of ARTI visits for which antibiotics are rarely indicated remained stable, whereas associated antibiotic prescription rates declined by 41% (RR: 0.59, 95% CI: 0.45–0.78). The proportion of these non-OM ARTI visits resulting in antibiotic prescription also declined from 41% in 1995–1996 to 23% in 2005–2006 (p=0.038) (Figure 2C).

Persons aged ≥5 years

ARTI visits

Overall annual visit rates increased by 19% from 3154 to 3758 per 1000 population (RR: 1.19, 95% CI: 1.15–1.23). This was primarily due to a 21% increase in non-ARTI visit rates (RR: 1.21, 95% CI: 1.20–1.23). ARTI rates remained stable and similar changes were observed in the age-stratified analyses (Table 1 and Figure 1). Although declines in rates of otitis media and related antibiotic prescriptions in this age group were similar to those among children aged <5 years, these rates were substantially lower (Figure 2A). As with children, concomitant conditions during an ARTI visit that could result in antibiotic prescriptions such as urinary tract infections (0.66%) or soft tissue infections (0.15%) were uncommon.

ARTI-associated antibiotic prescriptions

Although overall antibiotic prescription rates remained stable among subjects aged ≥5 years, the stratified analysis showed significant increases among persons aged 50 years or older. Overall antibiotic prescriptions associated with ARTI visits declined by 18% (RR: 0.82, 95% CI: 0.71–0.94) but significant declines occurred only among persons aged <50 years. Non-ARTI visit associated antibiotic prescriptions increased in all groups aged ≥5 years (Table 1). Overall, about 12% of all visits for persons aged ≥5 years consistently resulted in antibiotic prescriptions. The proportion of ARTI visits that resulted in antibiotic prescriptions was 63% in 1995–1996 and 54% in 2005–2006 (p=0.072). The proportion of non-ARTI visits resulting in antibiotic prescriptions remained stable (p=0.795).

In physician practices, the proportion of ARTI visits that resulted in antibiotic prescriptions among persons aged ≥5 years declined from 63% to 52% (p=0.025), whereas no significant changes were observed in emergency department visits (from 63% to 64%, p=0.510) or in outpatient clinics (from 61% to 60%, p=0.885).

Overall rates of ARTI visits resulting in penicillin prescriptions declined by 42% (Table 2). Amoxicillin prescription rates declined by 49% whereas amoxicillin/clavulanic acid prescription rates increased by 44%. Cephalosporin prescription rates declined by 41% and sulfonamide/tetracycline prescription rates declined by 47%. Macrolide/lincosamide prescription rates increased by 22%, mainly due to a 6-fold increase in azithromycin prescription rates. Quinolone prescription rates increased 5-fold. Similar changes in prescriptions of antibiotic classes were observed across age-groups (Table 3).

Table 3. Rates, rate differences and rate ratios of antibiotic prescriptions during ARTI visits, United States 1995–2006.

Rates (per 1,000 population), rate differences and rate ratios of antibiotic prescriptions during ARTI visits, United States 1995–2006

Penicillins	1995–1996	2005–2006	RD 95% CI	RR 95% CI
<5 years	653	448	−204 (−274 – −121)	0.69 (0.58 – 0.81)
5 to 17 years	190	122	−68 (−91 – −39)	0.64 (0.52 – 0.8)
18 to 49 years	62	33	−29 (−36 – −21)	0.53 (0.43 – 0.66)
50 to 64 years	38	26	−12 (−21 – 4)	0.7 (0.44 – 1.1)
65 to 79 years	44	22	−21 (−32 – −3)	0.51 (0.28 – 0.94)
80 or more years	33	18	−16 (−25 – 5)	0.53 (0.24 – 1.16)
Cephalosporins
<5 years	315	167	−148 (−192 – −88)	0.53 (0.39 – 0.72)
5 to 17 years	65	39	−26 (−36 – −12)	0.6 (0.44 – 0.82)
18 to 49 years	29	17	−12 (−17 – −5)	0.59 (0.43 – 0.82)
50 to 64 years	28	16	−12 (−18 – −2)	0.59 (0.37 – 0.94)
65 to 79 years	38	25	−13 (−22 – 0)	0.65 (0.42 – 1.01)
80 or more years	48	24	−24 (−34 – −4)	0.5 (0.28 – 0.91)
Lyncosamides/macrolides
<5 years	160	161	1 (−41 – 57)	1 (0.74 – 1.36)
5 to 17 years	53	54	1 (−12 – 18)	1.02 (0.78 – 1.33)
18 to 49 years	47	58	12 (−1 – 28)	1.25 (0.99 – 1.59)
50 to 64 years	42	63	21 (4 – 45)	1.51 (1.09 – 2.08)
65 to 79 years	46	54	8 (−13 – 41)	1.17 (0.73 – 1.88)
80 or more years	46	46	0 (−23 – 44)	0.99 (0.5 – 1.95)
Sulfonamides/tetracyclines
<5 years	204	18	−186 (−193 – −174)	0.09 (0.05 – 0.15)
5 to 17 years	21	8	−13 (−17 – −1)	0.37 (0.15 – 0.93)
18 to 49 years	13	7	−6 (−9 – −1)	0.53 (0.3 – 0.93)
50 to 64 years	9	12	3 (−2 – 12)	1.34 (0.75 – 2.4)
65 to 79 years	15	5	−10 (−13 – −1)	0.36 (0.14 – 0.9)
80 or more years	23	5	−18 (−22 – −3)	0.23 (0.06 – 0.86)
Quinolones
18 to 49 years	4	16	13 (7 – 23)	4.63 (2.88 – 7.43)
50 to 64 years	7	30	24 (10 – 49)	4.56 (2.49 – 8.36)
65 to 79 years	6	36	31 (13 – 65)	6.41 (3.28 – 12.52)
80 or more years	8	37	29 (10 – 70)	4.58 (2.18 – 9.65)

^*ARTI, acute respiratory tract infections; RD, Rate difference; RR, rate ratio; CI, confidence interval.

Data are from NAMCS and NHAMCS. Rates are average annual rates for each time period.

Non-OM ARTI

Overall antibiotic prescription rates for non-OM ARTI among subjects aged ≥5 years declined by 15% (RR: 0.85, 95% CI: 0.73–0.98), due primarily to declines in antibiotic prescriptions for ARTI for which antibiotics are rarely indicated. Rates of non-OM ARTI visits for which antibiotics are often indicated showed a non-significant increase (RR: 1.12%, 95% CI: 0.97–1.30) and rates of antibiotic prescription associated with these visits remained stable. The proportion of these visits resulting in antibiotic prescriptions declined by 13% (p=0.004) and was 59% in 2005–2006 (Figure 2B). Rates of non-OM ARTI visits for which antibiotics are rarely indicated had a non-significant decline of 9% (RR: 0.91, 95% CI: 0.80–1.04) whereas antibiotic prescription rates associated with these visits declined by 24% (RR: 0.76, 95% CI: 0.63–0.90). The proportion of non-OM ARTI visits that resulted in an antibiotic prescription did not change significantly, and was 48% in 2005–2006 (Figure 2C).

DISCUSSION

Our analysis of 12 years of national ambulatory survey data showed sustained declines in ARTI-associated antibiotic prescription rates. These declines were primarily observed in physician practices and were associated with a decline in OM visits (children aged <5 years) and antibiotic prescriptions associated with ARTI for which antibiotics are rarely indicated (all age groups). Although these observations are encouraging, the prescription of broad spectrum antibiotics, namely azithromycin and quinolones, increased substantially during the study period.

The declines observed in antibiotic prescriptions in children aged <5 years were largely related to declines in OM visit rates. However, the proportion of OM visits that resulted in antibiotic prescriptions remained stable, suggesting that clinician’s practices for antibiotic prescribing given a diagnosis of OM did not change. Although the American Academy of Pediatrics practice guidelines for the management of OM were updated in early 2004 to encourage the initial observation of non-severe OM cases in selected children (i.e. “watchful waiting”),^{24, 25} the largest observed declines preceded these recommendations. Moreover, recent evidence suggests low adherence to the updated guidelines.²⁶

Although visit rates for non-OM ARTI for which antibiotics are rarely indicated remained stable in children aged <5 years, rates of antibiotic prescriptions associated with these visits declined. Furthermore, the proportion of visits that resulted in antibiotic prescriptions declined. Similarly, among persons aged ≥5 years, visit rates for ARTI for which antibiotics are rarely indicated remained stable and there was a significant reduction in antibiotic prescriptions associated with these visits. These findings suggest improvements in antibiotic prescribing practices and also that previous observed declines in antibiotic prescriptions for these conditions were sustained through 2006.^{9, 10, 27, 28}

Starting in the mid-1990’s, the CDC and several State and local health departments and other organizations launched interventions to promote appropriate antibiotic prescriptions.⁶ In 1995, the CDC launched the Campaign for Appropriate Antibiotic use in the Community and working with the American Academy of Pediatrics and members of the American Academy of Family Physicians, this initiative led to the publication of The Principles for appropriate use of antibiotic for pediatric upper respiratory tract infections.⁶ Furthermore, a number of projects launched at local or State levels have shown significant improvements in antibiotic prescribing, including controlled interventions in Denver (1997–1998), Boston/Seattle (1997–1998), Tennessee (1997–1998), Wisconsin (1997), and Alaska (1998–2000) among others.⁶ The CDC also launched the media campaign “Get Smart: Know when antibiotics work” in 2003.

Although the specific reasons for decreasing ARTI visit rates cannot be established by this study, several explanations could be hypothesized. First, the application of stricter diagnostic criteria for specific ARTI, a central component of the campaigns promoting the judicious use of antibiotics, could have increased the threshold to diagnose these conditions²¹ resulting in subsequent reductions in antibiotic prescriptions.²⁹ Second, parents might be less likely to bring their children to a clinician for mild ARTI.³⁰ Healthcare encounters can be prevented by educating parents to identify and relieve ear pain in their children.^{25, 31} Moreover, the CDC Get Smart campaign was associated with declines in ARTI visits and related antibiotic prescriptions in Colorado.³² Third, routine childhood immunization with PCV7 since 2000 reduced the incidence of both OM and pneumonia during the study period. The declines in OM visits we report here are consistent with other estimates of the effect of this intervention^33–37 Influenza vaccines, recommended for use among young children since 2004 and among all children aged ≥6 months since 2008, could further reduce the incidence of ARTI and associated antibiotic prescriptions.

Although ARTI-associated prescriptions for most antibiotic classes declined, prescription of selected macrolides and quinolones increased substantially. In the early 1990s, the recognition of the role of atypical microorganisms in ARTI and the emergence of penicillin-resistant bacteria led to the recommendation of macrolides as a first line empiric treatment for community-acquired pneumonia.^{3, 38} Macrolides have since been used widely to treat ARTI^{3, 39} and although declines in invasive pneumococcal disease due to non-susceptible pneumococci were documented after PCV7 introduction, concerns remain about the increasing use of azithromycin in the United States.^{7, 9, 12} Young children are commonly colonized with pneumococcus. Azithromycin is a bacteriostatic antibiotic and the least potent macrolide against pneumococcus. Its long half-life of approximately 3 days may result in prolonged exposure of bacteria to relatively inactive concentrations of the antibiotic and selection of resistant strains.^{40, 41} Although the ease of administration is an advantage, a recent meta-analysis of randomized and quasi-randomized controlled trials showed that azithromycin efficacy against bronchitis or pneumonia was not superior to amoxicillin or amoxicillin/clavulanate.⁴² Recent guidelines for community-acquired pneumonia in ambulatory adults suggest that macrolide monotherapy should be restricted to selected patients (i.e. previously healthy and without risk factors for drug-resistant S. pneumoniae).⁴³

After introduction of PCV7 in the United States, the incidence of pneumococcal disease due to non-vaccine serotypes increased. These increases have been most notable for serotype 19A, a type that frequently is resistant to multiple antibiotics.⁴⁴ Increasing use of antibiotics, in particular selected macrolides, may contribute to the emergence of these antibiotic-resistant non-vaccine serotypes. Increases in disease due to antibiotic-resistant serotype 19A were reported among Alaskan Native, Bedouin⁴⁵ and South Korean children, populations where azithromycin is widely used, even though the latter two groups were not routinely immunized with PCV7. Interestingly, no increases in disease due to serotype 19A have been described among White Mountain Apache,⁴⁶ a population exposed to PCV7 and in which azithromycin use is approved only for the treatment of sexually transmitted diseases and for children with documented penicillin allergy.⁴⁷ Our findings indicate that azithromycin prescriptions rates increased substantially during recent years. The effects of these increases on the emergence of antibiotic-resistant pneumococcal serotypes not covered by PCV7 remain to be determined.

We also observed increasing prescription rates for quinolones for ARTI in persons aged ≥5 years. Increases in quinolone prescriptions and disease caused by quinolone-resistant bacteria have been reported in the United States and abroad.^48–50 Although the reported proportion of pneumococcal disease that is quinolone-resistant is still small (i.e. <1.8%)⁴⁹, the increasing trends observed are of concern and warrant close monitoring.^{48, 49}

Our study results must be interpreted in light of a number of limitations. First, we used ICD9-CM codes to identify ARTI and could not confirm diagnoses or appropriateness of antibiotics prescribed. Second, although our study evaluated antibiotic prescriptions, the adherence to these regimens is unknown. Moreover, additional antibiotic prescriptions or changes made over the telephone were not captured by the surveys. Third, available information did not allow the distinction between initial and follow-up visits throughout the study years. The inclusion of follow-up visits when no new antibiotic was prescribed would result in an underestimation of the proportion of ARTI that resulted in an antibiotic prescription. Fourth, survey participation rates have declined and required adjustments in weighting strategies over time. Nevertheless, national survey data provided consistent information to assess changes in antibiotic prescription patterns. Fifth, federal institutions such as the Veteran Affairs ambulatory services are not included in the national surveys.

There is little doubt that antibiotic use leads to the emergence of antibiotic-resistant bacteria.² Even though prescriptions for most classes of antibiotics declined, the prescription of quinolones and azithromycin increased. The potential role of these increases on the emergence of antibiotic-resistant microorganisms, in particular S. pneumoniae, has been described.^{2, 3, 14, 51} Our results indicate that overall antibiotic prescription rates have declined significantly. These changes coincided with efforts to reduce inappropriate antibiotic prescribing and the initiation of routine infant immunization with pneumococcal conjugate vaccine. Further efforts to improve antibiotic selection are urgently needed.