Factors influencing completion of multi-dose vaccine schedules in adolescents: a systematic review

Background Completion of multiple dose vaccine schedules is crucial to ensure a protective immune response, and maximise vaccine cost-effectiveness. While barriers and facilitators to vaccine uptake have recently been reviewed, there is no comprehensive review of factors influencing subsequent adherence or completion, which is key to achieving vaccine effectiveness. This study identifies and summarises the literature on factors affecting completion of multi-dose vaccine schedules by adolescents. Methods Ten online databases and four websites were searched (February 2014). Studies with analysis of factors predicting completion of multi-dose vaccines were included. Study participants within 9–19 years of age were included in the review. The defined outcome was completion of the vaccine series within 1 year among those who received the first dose. Results Overall, 6159 abstracts were screened, and 502 full texts were reviewed. Sixty one studies were eligible for this review. All except two were set in high-income countries. Included studies evaluated human papillomavirus vaccine, hepatitis A, hepatitis B, and varicella vaccines. Reported vaccine completion rates, among those who initiated vaccination, ranged from 27 % to over 90 %. Minority racial or ethnic groups and inadequate health insurance coverage were risk factors for low completion, irrespective of initiation rates. Parental healthcare seeking behaviour was positively associated with completion. Vaccine delivery in schools was associated with higher completion than delivery in the community or health facilities. Gender, prior healthcare use and socio-economic status rarely remained significant risks or protective factors in multivariate analysis. Conclusions Almost all studies investigating factors affecting completion have been carried out in developed countries and investigate a limited range of variables. Increased understanding of barriers to completion in adolescents will be invaluable to future new vaccine introductions and the further development of an adolescent health platform. PROSPERO reg# CRD42014006765. Electronic supplementary material The online version of this article (doi:10.1186/s12889-016-2845-z) contains supplementary material, which is available to authorized users.


Background
In the past decade there has been an increase in the number of new vaccines licensed worldwide [1,2] and in the accessible funding for vaccine introduction to lowresource settings through the founding of Gavi, The Vaccine Alliance, in 2000 [3]. Multi-dose vaccines in the WHO recommended immunization schedule for adolescents are listed in Table 1; WHO defines adolescence as age 10-19 years inclusive. National vaccine schedules can depart from WHO recommendations, e.g. 2 doses of hepatitis A and meningococcal conjugate vaccines (MCV4) are offered to adolescents in the USA [4]. Although recommended for administration at birth, hepatitis B vaccine (HBV) is routinely offered to older children and adolescents if not previously immunised [5]. In settings where varicella is seen as a public health priority WHO recommends 2 doses of varicella vaccine, with the first dose at 12-18 months and up to 4 month interval between doses [6,7]. The most recently licensed multidose vaccines are the human papillomavirus (HPV) vaccines. In 2014, HPV vaccine recommendations were revised by WHO SAGE from a schedule of 3 doses [8], to 2 doses at a 6 month interval in girls less than or equal to 15 years of age [9,10] based on evidence of non-inferior immunogenicity [11,12].
At present, evidence suggests multiple doses of HBV, HPV, and varicella vaccines are needed for efficacious protection against disease in adolescents [5,9,13,14]. However; completion of the vaccine dose series, defined as receipt of the final dose within 1 year of the first dose, has proven challenging in some settings. Completion rates of HPV vaccine were lower than 30 % in the first years of introduction in the USA [15,16]. Addressing specific difficulties in administering vaccines to adolescents will be invaluable for implementation of future adolescent vaccines and further developing adolescent health services.
The currently available reviews of factors influencing completion focus solely on selected developed countries [1,[17][18][19], have non-systematic searches [20,21] or need updating [22]. This systematic review describes factors which have been investigated for their effect on multi-dose vaccine adherence in adolescents to aid development of interventions to improve adherence.

Search strategy
A comprehensive set of search terms was built around: 1) childhood/adolescence; 2) vaccination/immunisation; 3) adherence/completion. Articles with at least one term from each topic were identified. Search terms were informed by the Cochrane Child Health Group recommended terms for adolescents or school children [23] and included international spelling variations (Additional file 1: Table S1). Multi-dose vaccines administered to adolescents were identified through the Centers for Disease, Control, and Prevention (CDC) [6,24], and the WHO list of prequalified vaccines [25]; however, search terms were not limited to these vaccines. Medline, Embase, Global Health (Ovid SP), Popline, Web of Science, Africa Portal, Africa-wide information, ADOLEC, Cochrane, Open Grey databases, and PATH, Gavi, and WHO websites were searched in February 2014. No publication date restriction was set. Publications, abstracts and conference proceeding were eligible for inclusion. All texts were collated and reviewed using Endnote X7 (Thompson Reuters); automated and manual de-duplication was performed.

Inclusion criteria
Inclusion criteria for consideration of studies were outlined in a protocol a-priori as per PRISMA guidelines [26] ( Table 2). The title and/or abstract of each article were reviewed in the first instance by a single reviewer (KG). Modelling studies, immunogenicity/efficacy trials were excluded. Two reviewers (KG, SM/EK) screened the abstracts. Any study including a vaccine for which more than one dose was administered to persons 9-19 years old in a routine setting within 1 year was considered for inclusion. The WHO definition of adolescent (10-19 years old) was widened to include 9 year olds to include all participants in HPV vaccine studies (WHO recommended for 9-13 year olds). Inclusion criteria were independently applied to full texts by 2 authors (KG, SM/EK) ( Fig. 1) [26].

Data extraction
Data were extracted by 2 authors (KG, EK) into separate forms on Microsoft Excel 2010. Article selection and data extraction discrepancies were resolved through discussion. Data on the study population, setting, vaccine, rates of completion and the factors investigated were extracted alongside descriptive, univariate and multivariate results, as applicable.

Assessment of bias
An assessment of bias was recorded on the data extraction form for each study using criteria outlined in the Cochrane tool for assessment of bias in intervention and epidemiological studies [27]. Selection bias and information bias were assessed alongside the potential for confounding.

Data synthesis
Heterogeneity in study methods, population, context and classification of exposure variables, led to a descriptive synthesis. Groups which initiated and completed the recommended vaccine schedule within 1 year were compared to non-completers who only initiated.

Individual level factors Age
The association between age and completion was investigated in 31 articles. Multivariate analyses of at least 2 age categories within the age range of 9-19 years were conducted in 20 studies, in the USA (n = 17), Canada [33], France [34], and Australia [35]. Age recommendations vary across countries; results must be interpreted in context. There is some evidence that completion rates decrease as age of vaccine initiation increases for HPV vaccine, HAV, and HBV [15,31,[36][37][38]. In the USA, the HPV vaccine recommended age range is between 11 and 26 years; five studies state similar results among Medicaid enrolees, adjusting for insurance, race, region and year, 17 year olds were 0.84 times less likely to complete HPV vaccine compared to 11 year olds (95 % CI 0.74-0.95) [15]. Among attendees of an urban hospital, in adjusted analyses, 14-17 year olds had 0.71 the odds of completion HPV vaccine when compared to 9-13 years olds (95 % CI 0.59-0.98) [37].
In the USA five further studies found no association [33,[39][40][41][42][43] and two studies report the converse association, increased likelihood of completion with age between 13 and 17 years controlling for year, race, insurance status; this perhaps reflects the perception that it was an 'STI vaccine' in 2007-8 [44,45]. No association between age and HPV vaccine completion was found in multivariate analyses in Canada although only one school grade was targeted [33,[39][40][41][42][43].

Race
Racial or ethnic identity was analysed in 31 studies from the USA, Australia and Greece; 18 conducted multivariate analyses. Analysis of >100,000 women in North Carolina adjusted for location, clinic, insurance, and age found Black (aOR 0. 55 [43]. Race was the only significant predictor of completion in the NIS-Teen household survey in USA [44]. Ten additional large database studies in the USA with multivariate analyses corroborate this association for both HPV and HBV vaccines [15,16,36,37,39,40,42,43,[46][47][48]. However, no association between race and completion was found in 5 studies when controlling for gender, insurance and health clinic characteristics [29,38,45,49,50]. Hispanic adolescents were underrepresented in one survey with a low response rate [29]. Greek non-nationals had lower completion rates (33 %) than nationals (60 %) for 2 doses of HAV [51]. In the northern territories of Australia, 3 dose coverage of HPV vaccine was lower in indigenous compared to nonindigenous groups (54 % vs. 64 %), but completion rates were the same (84 %) [52].

Insurance
Many countries have supplied HPV vaccine free-ofcharge. In the USA, although the vaccine was not initially eligible for reimbursement in some health insurance plans, after it was recommended by the Advisory Committee on Immunization Practices it was included in the Vaccines for Children (VFC) programme which provides for underinsured and uninsured children [1]. Insurance status was investigated as a risk factor in 25 articles, 16 conducted multivariate analyses (15 USA, 1 France). In 2011, insurance status remained a significant predictor of HPV series completion in the USA; those publicly insured (Medicaid) were 2.08 times (95 % CI 1.16-3.7) more likely to complete compared to those with no insurance; those privately insured were not significantly more likely to complete than those on public insurance (aOR 1.16; 95 % CI 0.97-1.38) controlling for age, race, contraception use [42]. The association between insurance status and completion was stronger in 2006-8 reflecting policy changes [16,43]. In France completion rates were lower among recipients of complimentary social welfare compared to those with private insurance (aRR 0.88; 95 % CI 0.83-0.93) [34].
Longer enrolment on an insurance plan (>12 years) was associated with a 1-14 % increase in likelihood of completion of 2 doses of varicella vaccine; 9-12 % increased likelihood for HAV and 21-23 % for HBV in the vaccine safety database of almost 600,000 people in the USA between 1998 and 2004 [31] controlling for age, gender, healthcare utilisation and provider characteristics.
Across the USA there are substantial differences across states in beliefs, policy, and the rapidity of implementation of changes made at the national level. In Oregon state in 2008, HPV vaccine was offered free of charge and no difference was found in completion rates between publicly and privately insured participants [53]. In Maryland in 2006-10 private insurance was found to be a risk factor for noncompletion compared to those publicly insured (aOR 0.76; 95 % CI 0.59-0.98), controlling for race and age [37]. No association in multivariate analyses was seen in 5 studies in the USA [29,39,40,44,50].

Gender
Gender was assessed in seven articles; no correlation between completion of HBV and gender was seen in unadjusted results from Greece [51], nor in adjusted results in Australia [35]. In the USA, controlling for delivery site, age, insurance, year, chronic conditions and Education -USA (7), Greece (1) prior healthcare utilization, male gender was marginally associated with lower completion for varicella (aOR 0.93; 95 % CI 0.90-0.96), HAV (aOR 0.98; 0.97-0.99) and HBV (aOR 0.97; 0.96-0.98) [31]. Included studies did not report completion of HPV vaccine in boys, recommendations to vaccinate boys were issued in 2015 in the USA; however, clinics in the USA with higher female:male ratios obtained higher completion rates of HPV vaccine among females (aRR 2.16; 1.13-4.13) [49].
The effect of SES may differ by delivery strategy; in Canadian public schools with in-school HPV vaccine delivery, completion increased as SES decreased, in Catholic schools in which the pupils relied on community delivery, completion decreased as SES decreased [58]. A linear trend with the Scottish multiple index of deprivation was found with completion but not with initiation; however, the difference between the most and least deprived groups was small (8 %) and disappeared with the administration of a catch-up dose 1 year later [59]. Girls in Canada in 2007-8 living in lower income neighbourhoods were significantly less likely to complete HPV vaccine than girls living in middle income neighbourhoods (aOR 0.45; 0.28-0.72) [33]. In France compliance with the HPV vaccine schedule was lower in social welfare recipients compared to nonrecipients (aRR 0.88, 0.83-0.93) [34].

Healthcare utilization
History of health care utilization was inconsistently associated with completion. Seventeen articles from the USA, France and Australia analysed an individual's prior use of health care (defined by receipt of other recommended vaccines, or the number of prior visits to a primary health care provider) and completion of a multi-dose series of varicella, HPV or HBV vaccines. In adjusted analyses in the USA, >10 visits to a health care provider in the last year was associated with 15 % increased likelihood of HPV vaccine completion and a 4-6 % increase in HBV completion [31]. Similar findings were reported in France where compliance with the HPV vaccine regimen was 10 % higher if a girl had >6 consultations with a family physician in the past year [34]. The magnitude of the effect is supported by reports of a 2 % increased likelihood of completing the HPV vaccine series with every primary care provider visit in the past year [36].
A further eight studies found no association between vaccine completion and the number of visits to a primary healthcare in the preceding 2 years [60], non-acute care in the year preceding initiation [29,33,39,44,55], previous prescriptions [47] or receipt of tetanus, diphtheria, and pertussis booster (Tdap) and meningococcal vaccines [38].
Recorded contraceptive use (DMPA) at any time in the medical records by HPV vaccine recipients was associated with a two-fold increase (95 % CI 1.72-2.47) in the odds of HPV vaccine completion [42]. In Canada, HBV vaccination conferred 16.9 times higher odds (95 % CI 14.8-19.2) for HPV vaccine completion in comparison with those who had not received HBV. However, the association could be confounded by the differing vaccination policies and delivery strategies by school [58]. In Australia in an area with a high risk population, including young sex workers and drug users, a shorter time interval (<2 weeks) between first contact with the health care provider and initiation of vaccine series correlated with better HBV completion [35].

Vaccine related knowledge
Three American studies examined knowledge in relation to completion in multivariate analysis [45,47,55]. The ability to correctly identify the number of required doses remained associated with series completion (aRR 1.38; 95 % CI 1.08-1.76) [55]. Parents who remember receiving a provider recommendation for vaccination were more likely to have daughters who had completed the regimen (aOR 2.71; 1.99-3.70) [45]. However, general knowledge of HPV and HBV vaccine was not associated with completion in adjusted analyses [45,47].

Adverse events
Three studies assessed whether experience of adverse events following HPV vaccination affected series completion in the USA. Parents of daughters who had completed the three dose series reported pain or discomfort as often as parents whose daughters were late for their second or third dose (OR 0.76; 95 % CI 0.33-1.77) [61]. In a survey of over 3000 vaccine recipients [55] (response rate 27 %), multivariate analysis controlling for age, socio-economic status, health care utilization, showed reports of bruising or swelling at first dose did not affect completion of the series (aRR 0.88; 0.7-1.00). An association was not apparent for those reporting pain, syncope or dizziness [55]. A qualitative study of 18 women in the USA who did not complete the HPV vaccine series found none of them mentioned adverse events as a reason [62].

Risk behaviour
A variety of risk behaviours in seven studies were assessed in relation to completion of HBV or HPV vaccine schedules; no associations were found. Drug use, history of sexually transmitted infections (STIs), or alcohol use was not associated with completion in the USA [38,47,63]. In multivariate analysis in Australia, intravenous drug use, sex work, or hepatitis C status did not correlate with likelihood of completion of HPV in a health unit serving at-risk populations [35].

Concomittant healthcare
Three articles assessed the effect of concomitant health service delivery on adherence to HPV in the USA. Receipt of another vaccine at the time of HPV vaccination was not associated with odds of HPV completion controlling for socio-demographic and provider characteristics [55]. However, if the first dose was given at a health care provider visit which was attended primarily for another reason other than HPV, the odds of a mistimed 3rd dose were almost double (aOR 1.97; 95 % CI 1.39-2.80) than that if the first dose was at a vaccine only visit, controlling for age and race [48]. Type of visit was not associated in analysis investigating the effect of age and healthcare utilization [41].

Access
Access to vaccination sites was assessed in two studies in the USA. Compliance to the schedule and completion of the series were not governed by proximity or mode of transportation to the clinical site [47]. Distance from home to clinic was not associated with completion controlling for age, race, and healthcare utilization [40].

Qualitative studies
One qualitative study investigated why 9-26 year olds did not return for the final dose of the HPV vaccine series, in non-exclusive responses: 33 % claimed they didn't know they were meant to obtain further doses, 23 % claimed they were too busy, 15 % cited inconvenience, 38.5 % claimed they were too busy or forgot, 7.7 % claimed they were too busy and times were inconvenient [62]. Two additional surveys of partially vaccinated university students in the USA and Australia indicated the potential problems with vaccinating older age groups who have competing priorities; reasons focused on inconvenience and lack of time [64,65].

Maternal characteristics
Three studies in the USA analyzed the relationship between maternal preventative behavior (cervical screening) and their daughter's HPV vaccine series completion. In multivariate analysis, controlling for demographic, socioeconomic, family, and health plan characteristics, all three studies found that girls whose mothers had received a pap smear in the past three years were more likely to complete the HPV vaccine series (aOR 1.07, 95 % CI 1.06-1.08) [ [66]).
The relationship between maternal education and vaccine series completion was assessed in eight studies conducted in the USA (n = 7) and in Greece [51]. Adolescents whose mothers had less than high school education were less likely to complete the vaccine series in multivariate analysis [44,66]; both studies controlled for adolescent age, SES, and mother's health characteristics and found similar effect estimates (aOR 0.68; 95 % CI 0.56-0.84) [44]; aOR 0.60; 0.41-0.87 [66]). No association between maternal education and HPV or HAV vaccine series completion in multivariate analysis was found in three studies [39,45,50].
Maternal age and marital status were found to have no or very slight associations with vaccine series completion in four of the five included studies [39,44,45,66]. In unadjusted analysis, one study found daughters with mothers aged over 40 years were more likely to complete the HPV vaccine series compared to mothers who were less than 40 years old [56].

Provider/ organisational characteristics Delivery model
There is strong evidence for high completion rates with school-delivery in high -income and low-middle income countries. Canadian in-school HPV vaccination completion rates were 75 % (95 % CI 74.7-75.8) compared to 36 % (95 % CI 35.3-37.2) for girls provided with a community-delivery model [58]. In-school vaccinations conferred 1.8 times the odds of completing the HBV series compared to if adolescents had to go off-site (95 % CI 1.15-2.8) in a parent survey in the USA controlling for age, race, insurance, SES, prior healthcare utilization [29].
Only 2 articles included data from low and middleincome countries (LMIC); descriptive results are available regarding the success of different delivery strategies [67]. In Uganda, a school-based strategy appeared more successful (94 % completion) than a strategy in which the vaccine was given in the community with a child health programme (79-87 % completion year 1-year 2) although the delivery strategies had slightly different target populations. Peru's school-based strategy achieved 98.7 % completion, whilst combined school-based and health centre strategies in Vietnam achieved >99 % completion. In India, very similar completion rates were achieved in campaign and routine delivery approaches (97-98 %) [67]. Differences in completion rates achieved in 21 demonstration projects in 14 countries implementing different models of delivery were insignificant [68]; however, the mixed model (school based delivery with mop-up activities at health centres) seemed to confer marginally higher completion (96.6 %), the school-only model was intermediate (88.6 %) and the health facilityonly model was least effective (79.7 %)(p = 0.39) [69].
In Australia, high-risk groups, benefited from an accelerated schedule (0, 7, 21 days and 12 months), which increased the likelihood of HBV vaccine completion 1.35 times (1.01-1.80) controlling for drug use, and length of contact with the health facility [35].

Provider characteristics
Vaccine schedule completion was higher in an American school based programme when students returned the consent forms to their teacher compared to the school nurse [70]. In 1994-5 in Canada, initial parental consent was lower at private schools compared to public schools; however, private and public schools did not differ in completion rates. Different education providers (teachers or public health nurses) did not have an effect on completion, although education from teachers was associated with higher consent [30].
A further 17 studies investigated health provider characteristics, of which 12 reported adjusted analyses. There was no evidence that the speciality of an adolescent girl's primary care physician influenced HPV series completion in multivariate analyses [15,16,36,42,49,55]. However, for women >17 years of age in the USA, those with a paediatric/internal medicine physician were slightly less likely to complete the HPV regimen than those with a family medicine physician. Female providers were not significantly associated with completion (male primary care provider aRR0.93 0.85-1.01) [36]. In an American datalink study, those vaccinated at paediatric clinics had the highest completion (61 %) compared to family care practices (53 %; aOR 0.78; 95 % CI 0.76-0.80) and the local health departments (39 %; aOR 0.48; 0.47-0.50) [43].

Discussion
We present a comprehensive review of the available literature on factors influencing adherence to multi-dose vaccine schedules among adolescents. The majority of studies took place in the USA (n = 49), the remainder included Canada (n = 3), France (n = 3), Australia (n = 2), Greece (n = 1), the UK (n = 1) and 2 multi-country studies including LMICs. The two studies including LMICs focused on organisational level factors and reported high adherence to HPV vaccine [68], therefore our summaries of individual level factors are limited in generalizability to developed settings. The high level of variation in the definitions, number and selection of factors investigated in each study limits the comparability of study results and prevented conduct of a meta-analysis. The overall impact of the identified characteristics on vaccine adherence is likely to be dependent on the mix of other factors present, as well as the programmatic and local context.
Good adherence to multi-dose vaccines appears to be higher in early adolescence (9)(10)(11)(12) year olds) when compared to later adolescence (>14 years old). It is unclear whether this is linked to adolescent health seeking behaviour, which was inconsistently associated with completion, or whether it is governed by logistical reasons as cited in qualitative results. It could reflect factors which are not explored in the available literature such as which groups were most targeted with communication materials or the general decrease in utilization of health services through adolescence [71]. In some populations in the USA, there is evidence that Black or Hispanic girls are disproportionately prone to low completion rates when compared to White girls after adjustment for socioeconomic status and insurance, despite some reports of similar rates of initiation. Adolescent females may have a slightly elevated likelihood of vaccine completion compared to males; this association may be a symptom of increased opportunity whilst accessing contraception at the health centre. Higher household income, maternal education and maternal preventative health behaviour were associated with higher completion rates when compared to lower socio-economic families and those mothers who rarely sought screening. Insurance status may have a decreasing effect on completion over time as knowledge spreads that both HBV and HPV vaccines are eligible for reimbursement on any insurance plan in the USA. Experience of adverse events and general knowledge about the vaccine did not affect completion rates. School-based delivery alongside supplying vaccine to health centres for out-of-school girls appears to be a successful approach in countries with relatively high school attendance, including some LMICs [69], the UK [59] and Canada [58].

Conclusions
The factors which affect rates of vaccine completion are context and time specific. Providers and programme planners should be aware that obtaining good consent and initiation rates is not enough; sub-groups within the population may need more help than others to complete