Nasopharyngeal Carriage of Streptococcus pneumoniae Among Young Children in Haiti Before Pneumococcal Conjugate Vaccine Introduction

Abstract

Background

Streptococcus pneumoniae, or pneumococcus, is a leading cause of morbidity and mortality in children worldwide. Pneumococcal conjugate vaccines (PCV) reduce carriage in the nasopharynx, preventing disease. We conducted a pneumococcal carriage study to estimate the prevalence of pneumococcal colonization, identify risk factors for colonization, and describe antimicrobial susceptibility patterns among pneumococci colonizing young children in Port-au-Prince, Haiti, before introduction of 13-valent PCV (PCV13).

Methods

We conducted a cross-sectional study of children aged 6–24 months at an immunization clinic in Port-au-Prince between September 2015 and January 2016. Consenting parents were interviewed about factors associated with pneumococcal carriage; nasopharyngeal swabs were collected from each child and cultured for pneumococcus after broth enrichment. Pneumococcal isolates were serotyped and underwent antimicrobial susceptibility testing. We compared frequency of demographic, clinical, and environmental factors among pneumococcus-colonized children (carriers) to those who were not colonized (noncarriers) using unadjusted bivariate analysis and multivariate logistic regression.

Results

Pneumococcus was isolated from 308 of the 685 (45.0%) children enrolled. Overall, 157 isolates (50.8%) were PCV13 vaccine-type serotypes; most common were 6A (13.3%), 19F (12.6%), 6B (9.7%), and 23F (6.1%). Vaccine-type isolates were significantly more likely to be nonsusceptible to ≥1 antimicrobial (63.1% vs 45.4%, P = .002). On bivariate analysis, carriers were significantly more likely than noncarriers to live in a household without electricity or running water, to share a bedroom with ≥3 people, to have a mother or father who did not complete secondary education, and to have respiratory symptoms in the 24 hours before enrollment (P < .05 for all comparisons). On multivariable analysis, completion of the pentavalent vaccination series (targeting diphtheria, pertussis, tetanus, hepatitis B, and Haemophilus influenzae type b) remained significantly more common among noncarriers.

Conclusions

Nearly a quarter of healthy children surveyed in Haiti were colonized with vaccine-type pneumococcal serotypes. This baseline carriage study will enable estimation of vaccine impact following nationwide introduction of PCV13.

Streptococcus pneumoniae, also known as pneumococcus, is a leading cause of bacterial pneumonia in young children and a common cause of other clinical syndromes including septicemia, meningitis, and otitis media [1–5]. In 2015, pneumococcus was estimated to cause >300 000 deaths globally in children aged <5 years [2]. More than 95 serotypes of pneumococcus have been identified, although certain serotypes are more likely than others to cause clinical illness [6, 7]. A 7-valent pneumococcal conjugate vaccine (PCV7) was introduced in the United States in 2000, and higher-valency vaccines (PCV10 and PCV13) are now available. These vaccines have dramatically reduced pneumonia and other complications of pneumococcal infection in regions where they have been introduced [2, 8]. The World Health Organization now recommends administration of either PCV10 or PCV13 for routine childhood immunization [9].

Nasopharyngeal colonization by pneumococcus typically precedes clinical illness by a period of days to months depending on the serotype [10]; colonization may also be asymptomatic and transient [11]. Young children are more likely to be colonized than older children or adults, and pneumococcal carriage is particularly high in low-income settings. Studies have shown that young children who have received a PCV are less likely than unvaccinated children to carry vaccine-type pneumococci and subsequently less likely to transmit these serotypes to unvaccinated children and adults. Assessing pneumococcal carriage among children is a useful way to determine the potential impact of the pneumococcal conjugate vaccine, particularly in regions where surveillance for invasive pneumococcal disease is limited.

In Haiti, pneumonia is the leading cause of death in children under age 5 years, accounting for up to 20% of all deaths in this age group [12, 13]. PCV13 became part of the routine immunization schedule in Haiti in late 2018 [14]. In this article, we describe pneumococcal carriage among children aged 6–24 months at an immunization clinic in Port-au-Prince prior to PCV introduction.

METHODS

The objectives of this investigation were to estimate the proportion of children 6–24 months old in Port-au-Prince who are colonized with pneumococcus, to describe the distribution of vaccine-type and non-vaccine-type serotypes, to assess epidemiological factors associated with carriage, to describe antimicrobial susceptibility patterns among pneumococcal isolates, and to establish a baseline for future comparison with pneumococcal colonization following PCV13 introduction in Haiti.

Study Design and Setting

Participants were recruited from an immunization clinic in a large, urban, government-run hospital in Port-au-Prince, Haiti. We recruited children from an immunization clinic because these patients were likely to be young children, to be healthy or only mildly ill, and to have immunization records available. The hospital was located adjacent to the National Laboratory, which allowed for the prompt transfer and storage of specimens. Using data from carriage studies conducted in other low-income countries, we determined that enrolling 600–700 children should enable detection of ≥30% decline in carriage of vaccine-type pneumococci following PCV13 introduction, assuming 30% carriage of vaccine-type pneumococci at baseline.

Enrollment and Study Procedures

Children were eligible to participate if they were between the ages of 6 and 24 months (inclusive) and current residents of Port-au-Prince. Children were excluded if a parent or guardian was not present to provide consent, if they were in obvious physical distress, or if they had been previously enrolled in the study. Children were enrolled from 3 September 2015 through 29 January 2016. Figure 1 shows the details of the enrollment process. Participants’ parents or guardians were interviewed in their preferred language (French or Kreyòl) about demographics and epidemiological factors that have been associated with pneumococcal carriage in prior studies. A nasopharyngeal (NP) specimen was obtained from each participant using pediatric-sized dacron-tipped flocked swabs. Three study nurses were trained in the technique of performing NP specimen collection prior to the start of the study. Specimens were transported in STGG (skim milk, tryptone, glucose, and glycerin) media at 4°C, frozen at –80°C within 4 hours of collection [15], and shipped to the Streptococcus Laboratory at the Centers for Disease Control and Prevention (CDC) in Atlanta, Georgia, for processing.

Figure 1.

Study enrollment.

Laboratory Methods

For each NP specimen, 200 µL of swab-inoculated STGG media was transferred to 5.0 mL Todd–Hewitt broth containing 0.5% yeast extract and 1 mL of rabbit serum and was incubated at 35°C–37°C for 6 hours. Cultured broth (10 µL) was plated on sheep blood agar and incubated in 5% carbon dioxide at 35°C–37°C. After 18–24 hours of incubation, plates were examined for the appearance of α-hemolytic colonies. Single colonies (for each morphology type) were selected and subcultured, and pneumococci were identified by susceptibility to optochin and bile solubility [15]. Serotypes for the pneumococcal isolates were determined using the Quellung reaction.

Confirmed pneumococcal isolates from each NP swab were tested for susceptibility to penicillin, amoxicillin, cefotaxime, ceftriaxone, cefuroxime, erythromycin, clindamycin, tetracycline, trimethoprim-sulfamethoxazole (TMP-SMX), rifampin, ciprofloxacin, levofloxacin, meropenem, quinupristin-dalfopristin, vancomycin, linezolid, and chloramphenicol using the broth microdilution method following Clinical and Laboratory Standards Institute (CLSI) guidelines. Antimicrobial nonsusceptibility (including both intermediate and resistant) was determined based on CLSI breakpoints [16] (using penicillin breakpoints for nonmeningitis pneumococcal infections).

Statistical Methods

To evaluate determinants of NP colonization, we first performed an unadjusted bivariate analysis to identify factors associated with pneumococcal colonization. Proportions were compared using χ ² (or Fisher exact test for cell counts <5) with P < .05 considered statistically significant. We then used a multivariable approach that combined use of both a directed acyclic graph (DAG) and multiple logistic regression to evaluate and adjust for confounders to the association. To do this, we created a DAG (using DAGitty version 2.3 http://dagitty.net/) to identify factors that might exist on the causal pathway between breastfeeding and pneumococcal colonization after reviewing results of bivariate analysis (Supplementary Figure). Additional contribution to colonization of other factors was evaluated through a backward logistic procedure using factors both detected by the DAG, and those factors identified as significant from the bivariate analysis. Factors were eliminated if Wald P values were >0.2. Analyses were performed using SAS version 9.3 or 9.4 (SAS Institute, Cary, North Carolina), EPI INFO version 7.2.2.6 (CDC, Atlanta, Georgia), and Microsoft Excel 2013 (Redmond, Washington) software.

Ethical Considerations

The study protocol was evaluated and approved by CDC’s institutional review board and the Haiti Ethics Committee prior to the start of the study. Written consent was obtained from the parent or guardian of each participant.

RESULTS

Pneumococcal Colonization

Among 2084 children approached, 704 were eligible and 685 children were enrolled (Figure 1). Pneumococcus was isolated from 308 (45%) specimens; 1 participant had 2 serotypes isolated from a single specimen, for a total of 309 pneumococcal isolates. Thirty-one specimens exhibited overall poor bacterial growth with no pneumococcus isolated, perhaps suggesting that the specimen was not obtained or preserved properly and potentially leading to an underestimate of the true carriage rate. One hundred fifty-seven (51%) isolates had PCV13-type serotypes by the Quellung reaction (Figure 2). The most common vaccine-type serotypes were 6A (n = 41 [13.3%]), 19F (n = 39 [12.6%]), 6B (n = 30 [9.7%]), and 23F (n = 19 [6.1%]). Two vaccine-type serotypes, 5 and 7F, were not isolated from any children. The most common non-vaccine-type serotypes were 11A (n = 32 [10.4%]) and 34 (n = 18 [5.8%]). Eleven (3.6%) isolates were nontypeable.

Figure 2.

Serotype distribution of pneumococci isolated from Haitian children aged 6–24 months (N = 309). One child’s nasopharyngeal swab was positive for 2 different serotypes; 309 serotypes were identified from 308 children. Abbreviations: NT, nontypeable; PCV13, 13-valent pneumococcal conjugate vaccine.

Factors Associated With Pneumococcal Colonization

Factors associated with colonization included sharing a bedroom with ≥3 people, breastfeeding, and having symptoms of respiratory infection (cough, difficulty breathing, or runny nose) in the 24 hours or cough in the 30 days prior to enrollment (P < .05 for all comparisons) (Table 1). Factors associated with not being colonized with pneumococci included living in a household with running water or electricity, having a father or mother who completed secondary school education, living in a household with indoor cooking facilities, using propane or other noncoal fuel sources for cooking, and completing the 3-part pentavalent vaccination series (targeting diphtheria, pertussis, tetanus, hepatitis B, and Haemophilus influenzae type b) or 2-part rotavirus vaccination series (P < .04 for all).

Table 1.

Epidemiological Factors Associated With Pneumococcal Carriage in Children Aged 6–24 Months—Port-au-Prince, Haiti, 2015–2016 (N = 685): Results of Unadjusted Bivariate Analysis

Epidemiologic Factors	Pneumococcus-Positivea, No. (%)	Pneumococcus-Negativeb, No. (%)	P Value
Total	308	377
Demographics
Male sex	157 (51)	181 (48)	.440
Age, mo, median (IQR)	11 (8–14)	10 (6–13)	.128
6–11	168 (55)	234 (62)
12–17	96 (31)	95 (25)
18–24	44 (14)	48 (13)
Maternal age at enrollment, y			.425
Median (IQR)	29 (25–33)	30 (27–34)
≤20	23 (7.5)	22 (5.8)
21–30	162 (53)	182 (48)
≥31	123 (40)	173 (46)
Socioeconomic status
Running water at home	83 (27)	149 (40)	<.001
Electricity at home	272 (88)	362 (96)	<.001
Household member with a permanent job	204 (66)	274 (73)	.067
Father completed secondary school	147 (48)	262 (70)	<.001
Mother completed secondary school	84 (27)	191 (51)	<.001
Household characteristics
Shared bedroom with ≥3 people	137 (44)	100 (27)	<.001
Indoor cooking facilities	138 (45)	210 (56)	<.001
Cooking done in room where child sleeps	16 (12)	14 (6.7)	.109
Household trash burned near household	30 (10)	41 (11)	.627
Tobacco smoker living in household	26 (8)	27 (7)	.544
Primary household cooking fuelc			<.001
Charcoal	213 (69)	188 (50)
Propane	70 (23)	148 (39)
Otherd or not specified	25 (8.1)	41 (11)
Medical history
Breastfeeding			.010
Never breastfed	17 (5.5)	44 (12)
Weaned	99 (32)	126 (34)
Still breastfed	192 (62)	204 (54)
Vaccination history
Pentavalent seriese complete (3 doses)	254 (87)	341 (95)	.002
Rotavirus series complete (2 doses)	223 (76)	299 (84)	.035
Recent symptoms
Fever in last 24 h	54 (18)	51 (14)	.147
Cough in last 24 h	136 (44)	101 (27)	<.001
Difficulty breathingf in last 24 h	54 (18)	45 (12)	.038
Runny nose in last 24 h	150 (49)	105 (28)	<.001
Any symptom in last 24 h	196 (63)	153 (41)	<.001
Fever in last 30 d	168 (55)	178 (47)	.056
Cough in last 30 d	211 (69)	200 (53)	<.001
Difficulty breathingf in last 30 d	103 (33)	101 (27)	.058
Diagnosis of pneumonia in last 30 d	6 (2)	3 (1)	.312
Any symptom in last 30 d	257 (83)	268 (71)	<.001
Antimicrobial use
Any antimicrobial use in last 7 d	24 (8)	43 (11)	.115
Any antimicrobial use in last 30 d	46 (15)	71 (19)	.203

Data are presented as No. (%) unless otherwise indicated. Statistically significant differences (P < .05) are shown in bold.

Abbreviation: IQR, interquartile range.

^aPneumococcus-positive = culture positive for pneumococcus (carrier)

^bPneumococcus-negative = culture negative for pneumococcus (noncarrier)

^cIncludes indoor and outdoor cooking locations.

^dOther fuel sources included kerosene (n = 23), natural gas (n = 16), electricity (n = 4), wood (n = 1), or other/not specified (n = 80).

^eThe pentavalent vaccination series targets diphtheria, pertussis, tetanus, hepatitis B, and Haemophilus influenzae type b.

^fParents or guardians were asked whether the child had difficulty breathing or rapid breathing.

The DAG supported inclusion of variables assessing child age, maternal age, maternal and paternal education, household crowding, and socioeconomic status in an adjusted analysis using logistic regression (Supplementary Figure). Factors initially included in the backward logistic procedure are shown in Table 2. We retained 2 variables (breastfeeding and vaccination status) in the model during the logistic procedure because the final model was used to calculate the adjusted effects of breastfeeding and vaccination on pneumococcal colonization.

Table 2.

Epidemiological Factors Associated With Pneumococcal Carriage in Children Aged 6–24 Months After Adjustment by Logistic Regression—Port-au-Prince, Haiti, 2015–2016 (n = 685)

Epidemiologic Factors	OR (95% CI)	P Value	aORa (95% CI)	P Value
Demographics
Age, mo, median (IQR)
6–11	Reference
12–17	1.41 (1.00–1.99)	.052	…
18–24	1.27 (.81–2.01)	.291	…
Maternal age, y, at enrollment, median (IQR)
≤20	Reference
21–30	0.85 (.45–1.59)	.258	…
≥31	0.68 (.36–1.28)	.227	…
Socioeconomic status
Running water at homeb	0.56 (.41–.78)	<.001	0.63 (.43–.94)	.022
Electricity at homeb	0.31 (.17–.58)	<.001	0.78 (.37–1.63)	.502
Father completed secondary schoolb	0.40 (.29–.55)	<.001	0.71 (.46–1.10)	.121
Mother completed secondary schoolb	0.37 (.26–.50)	<.001	0.60 (.40–.92)	.019
Household characteristics
Shared bedroom with ≥3 people	2.22 (1.61–3.06)	<.001	2.02 (1.36–2.98)	<.001
Indoor cooking facilitiesb	0.65 (.48–.87)	<.001	0.99 (.66–1.49)	.955
Primary household cooking fuel
Charcoal	Reference		Reference
Propane	0.42 (.30–.59)	<.001	0.87 (.54–1.42)	.583
Other or unknown	0.54 (.32–.92)	.022	1.08 (.55–2.12)	.829
Medical history
Breastfeeding
Never breastfed	Reference		Reference
Weaned	2.03 (1.10–3.77)	.023	1.55 (.74–3.25)	.242
Still breastfed	2.44 (1.36–4.41)	.003	1.66 (.81–3.38)	.167
Vaccination history
Pentavalent seriesc complete (3 doses)b	0.50 (.32–.78)	.002	0.35 (.17–0.73)	.005
Rotavirus series complete (2 doses)b	0.68 (.48–.97)	.035	0.95 (.55–1.64)	.853
Symptoms in 24 h before enrollmentb,d	2.56 (1.88–3.50)	<.001	2.03 (1.41–2.91)	<.001
Antimicrobial use
No antimicrobial use in last 30 d	Reference
Any antimicrobial use in last 7 d	0.65 (.39–1.10)	.109	…
Any antimicrobial use in last 30 d	0.76 (.50–1.14)	.177	…

Statistically significant differences (P < .05) are shown in boldface.

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; IQR, interquartile range; OR, odds ratio.

^aAdjusted for the following variables: father completed secondary education (yes/no), mother completed secondary education (yes/no), running water at home (yes/no), child shares a bedroom with at least 3 other people (yes/no), child had symptoms of fever, cough, difficulty breathing, or runny nose in the 24 hours before enrollment (yes/no).

^bReference group is participants without this characteristic.

^cThe pentavalent vaccination series targets diphtheria, pertussis, tetanus, hepatitis B, and Haemophilus influenzae type b.

^dIncludes fever, cough, difficulty breathing, or runny nose.

Factors that remained associated with colonization after backward elimination included the following: whether both the child’s father and the child’s mother completed secondary education; whether the child’s household had running water; whether ≥3 people slept in the same room as the child; and whether the child had symptoms of fever, cough, difficulty breathing, or runny nose in the 24 hours before enrollment. After adjustment for these factors, the odds ratios (ORs) associated with breastfeeding changed from 2.03 (weaned) and 2.44 (still breastfed) (P < .05 for both) to 1.55 and 1.66 (neither statistically significant), respectively. The association between completing the rotavirus vaccine series was also no longer significant (Table 2). However, completion of the pentavalent vaccination series remained significantly more common among noncarriers (adjusted OR, 0.35 [95% confidence interval, .17–.73]).

Antimicrobial Use and Susceptibility

Parents or guardians of 117 children (17%) reported antimicrobial use among their children within the past 30 days. β-lactam antimicrobials were the most commonly reported class (n = 77 [66%]), with amoxicillin as the most common agent (n = 63 [54%]) (Table 3). Children who had received any antimicrobial within 30 days prior to study enrollment were less likely to be colonized with pneumococcus than those who did not receive any antimicrobial in the prior 30 days (15% vs 19%), but this difference did not reach statistical significance.

Table 3.

Antimicrobial Use Within the Last 30 Days by Parental Report (n = 117)

Antimicrobial Class	Antimicrobial	No. (%)
β-lactams	Amoxicillin	63 (54)
	Ampicillin	1 (1)
	Augmentin	2 (2)
	Cephalexin	3 (3)
	Ceftriaxone	2 (2)
	Cloxacillin	6 (5)
Aminoglycosides	Gentamicin	1 (1)
Nitroimidazoles	Metronidazole	7 (6)
Tetracyclines	Tetracycline	1 (1)
Folate pathway antagonists	Trimethoprim-sulfamethoxazole	12 (10)
Multiple	>1 antimicrobial	4 (3)
Unknown	Unknown	15 (13)

Most pneumococcal isolates (n = 168 [54%]) showed nonsusceptibility to ≥1 antimicrobial (Figure 3); nonsusceptibility to tetracycline, TMP-SMX, or erythromycin was most frequently noted. PCV13-type serotypes were significantly more likely to show nonsusceptibility to at least 1 antimicrobial agent compared with non-vaccine-type serotypes (99/157 [63%] vs 69/152 [45%], P = 0.002). Among colonized children, use of a β-lactam antimicrobial in the 30 days prior to study enrollment was significantly associated with having an isolate showing β-lactam nonsusceptibility (8/30 [27%] vs 13/278 [4.7%], P < .001).

Figure 3.

Percentage of pneumococcal isolates (N = 309) showing nonsusceptibility (defined by resistant- or intermediate-range minimum inhibitory concentration values) to select antimicrobial agents. No isolates showed nonsusceptibility to quinolones (levofloxacin or ciprofloxacin), linezolid, rifampin, quinupristin-dalfopristin, or vancomycin. Abbreviations: PCV13, 13-valent pneumococcal conjugate vaccine; TMP-SMX, trimethoprim-sulfamethoxazole.

Discussion

Among children in an urban immunization clinic in Port-au-Prince, Haiti, approximately 45% of those aged 6–24 months carried pneumococcus in the nasopharynx, and more than half of isolates obtained were PCV13 vaccine-type serotypes. This 1-time, cross-sectional study performed over a 5-month period suggests that nationwide PCV13 implementation will result in a substantial public health benefit, given the experience from settings with similar carriage findings that have introduced PCV [2, 8]. These findings supported the decision by the Ministère de la Santé Publique et de la Population to introduce PCV13 in Haiti in 2018.

To our knowledge, this is the largest pneumococcal carriage study of healthy children in Haiti and serves as a baseline for future evaluation of the impact of nationwide PCV13 implementation. Earlier studies of pneumococcal carriage among young children in the Caribbean, prior to the introduction of pneumococcal vaccines in their respective countries, have focused on different patient populations and showed a range of carriage estimates. A 1999–2000 study in the Dominican Republic of children aged 6–59 months with respiratory disease requiring antimicrobial treatment found pretreatment carriage rates of 73%–75% [17]. In 2001, overall carriage was 77.5% among healthy children aged 1–6 years attending daycare in Havana City, Cuba [18], whereas overall carriage among healthy children aged 2–18 months in 2013 in Cuba was only 21.6% [19]. Most recently, as part of an 8-country study of children hospitalized with and without pneumonia in 2010–2014, investigators found a pneumococcal carriage rate of 86.1% among Haitian children with signs or symptoms of pneumonia compared with 60.7% among a control group of children from Haiti who were admitted for surgery or at a hospital clinic for outpatient appointments [20]. Among the control group of children in Haiti, 43.2% of pneumococcal isolates were vaccine-type serotypes. The lower overall rate of carriage in our study compared with this prior study of children in Haiti may be attributable to differences in patient age, health status, or seasonality. Also, the numbers of Haitian children enrolled in the earlier multisite study were small: 101 cases and 122 controls were included. Overall pneumococcal carriage in our study was also lower than in other developing countries in recent studies [21–25]. This discrepancy may reflect a genuine difference in carriage between populations or may be due in part to other factors such as characteristics of the specific population sampled or differences in laboratory procedures used for pathogen identification and serotyping between studies. For example, in the 8-country study [20], polymerase chain reaction–based methods were used for both detection and serotyping directly from the swab, which can increase sensitivity compared to culture-based procedures but may also overestimate pneumococcal carriage rates given the high prevalence of confounding nonpneumococcal species that can yield false positives [26, 27].

Serotype distribution is known to vary geographically and is altered by PCV implementation. The top 7 serotypes in our study (6A, 19F, 11A, 6B, 23F, 34, and 14) accounted for 61.8% of carriage strains and were also among the most prevalent in prevaccine populations in Haiti (cases and controls) [20], Cuba [19], and Brazil [28]. Consistent with other carriage study reports [21, 29–33], vaccine-type serotypes 1, 5, and 7F were very uncommon (0–0.3%); these serotypes are often common causes of invasive disease in unvaccinated children but are thought to have a shorter duration of carriage. Serotypes 9V and 19A were also rare in our study, whereas these were more common in other populations and slightly more common in children from Haiti with signs or symptoms of pneumonia [20, 30, 32].

Antimicrobial nonsusceptibility among recovered pneumococcal isolates varied by serotype. Nonsusceptibility was associated with the use of any antimicrobial within 30 days prior to specimen collection, which may be due in part to selective pressure generated by antimicrobial use. Nonsusceptibility to any antimicrobial was significantly more common among vaccine-type serotypes, suggesting that the introduction of PCV13 may reduce the proportion of resistant pneumococcal isolates carried by young children [34–36]. PCV introduction may also reduce the frequency of antimicrobial prescribing [37], which has been documented by other countries after introducing PCV. However, PCV implementation did not have a significant impact on resistance in Latin America [38], and other studies have documented a limited duration of resistance reduction depending on serotype replacement [39, 40]. Furthermore, baseline nonsusceptibility to β-lactam antimicrobials, the most frequently prescribed to children in our study, was low overall (<10%); however, nonsusceptibility to clindamycin, erythromycin, tetracycline, and TMP-SMX exceeded 10% among vaccine-type serotypes.

On bivariate analysis, most of the epidemiologic factors that we examined in our study showed expected associations with pneumococcal carriage. Lower socioeconomic status, higher household crowding, and presence of respiratory symptoms were significantly associated with carriage, echoing previous findings [19–21, 25, 30, 32, 41, 42]. A few findings were unexpected, however. Breastfeeding has been previously associated with lower rates of pneumococcal colonization [21, 43, 44] or has shown no relationship to colonization [19, 29, 30, 42, 45, 46]; however, in our bivariate analysis, children who were breastfed were more often colonized than those who were not breastfed. This association was no longer statistically significant after adjustment for multiple factors, including maternal education and running water in the home, suggesting that the family’s socioeconomic status or the mother’s opportunities for work outside the home (where breastfeeding may be inconvenient or impossible) might have led to this paradoxical relationship. The effects of breastfeeding on colonization may also be short-lived, and could be influenced by factors other than duration (such as frequency) that were not assessed on our questionnaire. Completion of rotavirus and pentavalent vaccination series each was associated with a lower likelihood of carriage on bivariate analysis. After adjustment, only the relationship to the pentavalent vaccine series remained statistically significant. Completion of the 3-part vaccination series could be a marker for other factors that we did not directly assess, such as access to care. Our study found no association with recent antimicrobial use despite low rates of resistance to amoxicillin, the antimicrobial reported by caregivers as the most commonly prescribed. Prior studies have linked recent antimicrobial use to decreased likelihood of colonization [47–49]. Finally, in our investigation, indoor cooking was significantly associated with lower likelihood of colonization. Cooking indoors or in the sleeping area has been previously associated with an increased risk of colonization [50], has shown a borderline increased risk [51], or has shown no effect on colonization [52]. However, cooking location was no longer statistically significant after adjustment for other factors.

Our investigation has several limitations. First, some pneumococcal serotypes may commonly colonize young children yet rarely cause clinically significant illness [53, 54], while others that are carried for a short duration (and thus unlikely to be detected on a cross-sectional study) may have a high propensity to cause invasive disease [53, 55, 56]. Thus, the distribution of serotypes seen on carriage studies do not precisely predict the distribution of serotypes causing severe disease. Furthermore, our results may not be generalizable to all young children in Haiti for several reasons. Our decision to enroll participants from a public immunization clinic results in a sampling bias, as children in the clinic may come from families who disproportionately value or have better access to preventive services than other portions of the population. For example, children in this study were more likely to have received the pentavalent vaccine—associated with a lower likelihood of pneumococcal carriage—than children in Haiti generally during 2015–2016 [57]. Based on questionnaire responses, we believe that children enrolled in our study were born to mothers who were older, had fewer siblings at home, and may have had higher socioeconomic status than children in the general population. Furthermore, carriage patterns are known to vary with geography, urbanicity, and seasonality; our study took place in a single urban center over a 5-month period. However, most (>95%) children had lived in Port-au-Prince for at least 4 months, suggesting that study results are generalizable to children seeking care at immunization clinics in Port-au-Prince.

Conclusions

In summary, nearly half of healthy children in Haiti in our study were colonized with S. pneumoniae. Five of the 7 most commonly observed serotypes are included in PCV13, suggesting that PCV13 will have an important impact in Haiti. This study establishes the baseline of vaccine-type and non-vaccine-type carriage among young, healthy children at an immunization clinic in Port-au-Prince, Haiti, prior to the introduction of PCV13. The high vaccination rates for pentavalent vaccines and the rotavirus vaccine that was recently introduced in Haiti suggest that the immunization clinic may be a favorable setting for a follow-up study after vaccine introduction to monitor the impact of PCV13 on serotype distribution and antimicrobial resistance.

Supplementary Data

Supplementary materials are available at The Journal of 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

Acknowledgments. The authors would like to thank Paul Adrien, Yves Gaston Deslouches, and Donald Lafontant from Ministère de la Santé Publique et de la Population; Jacques Boncy, Josiane Buteau, Marie Danielle Jean Baptiste, Marise Joseph, Gerard Joseph, Mentor Ali Ber Lucien, and Emmanuel Rossignol from Laboratoire National de Santé Publique; Marie Berthude Siméus, Barbara Villemenay, and the immunization clinic staff members from Hôpital Universitaire de la Paix; the study nurses Martense Bourdeau, Perside Dornévil, and Manite Garçon; Robert Burris, Cyrce Davis, Mark Katz, and Jean-Louis Yves Frantz from the Centers for Disease Control and Prevention (CDC)–Haiti; Hollis Walker from the CDC-Atlanta Streptococcus Laboratory for assistance with specimen processing and isolate characterization; and Barbara Marston, Rania Tohme, and Jennifer Verani from CDC-Atlanta for their invaluable contributions to this investigation.

Disclaimer. The findings and conclusions in this report are those of the authors and do not represent the official position of the CDC.

Supplement sponsorship. This supplement is sponsored by the World Health Organization and the U.S. Centers for Disease Control and Prevention.

Potential conflicts of interest. All authors: No reported conflicts of interest.

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.