Between 1997 and 1999, a prospective hospital surveillance study in Vellore reported Hib to be the cause of 44% (8 of 18) of confirmed cases of bacterial meningitis, with ~7 confirmed cases per 100 000 children aged less than 5 years.¹⁶ Recent multicenter surveillance estimates from Vellore, Chennai, Lucknow, and New Delhi suggest that 70% of bacterial meningitis may be caused by Hib.¹⁷ We used the lower and more conservative estimate of 44%, but made adjustments to account for cases of Hib that were not detected in the laboratory and cases that did not have access to care. Because 57% (24 of 42) of the purulent (probable) cases had no confirmed pathogen in the Vellore study, we assumed the percent of unconfirmed cases attributable to Hib to be the same as the percent of confirmed cases attributable to Hib (44%), based on WHO guidelines ¹⁸ and following methods used in other bacterial meningitis etiology studies.¹⁹ In addition, we assumed that 23% of the cases would not have had access to formal medical care during this study based on the proportion of children living in Vellore who, according to the NFHS, did not seek care for acute lower respiratory infection (ALRI) in 2006.¹² Hence, the final adjusted Hib meningitis incidence was 22 per 100 000 children <5 years [7.12/(1–57%)]/(1–23%). We assumed this incidence estimate for all states.

State-level Hib meningitis case fatality ratios (CFRs) were calculated by adjusting the 11% CFR reported in multicenter bacterial surveillance sites in Chennai, Vellore, Lucknow, and New Delhi¹⁷ by the state-level proportion of children without access to medical care. For children with no access to care, a 100% CFR was assumed.¹⁰ State-level estimates of access to care were based on the NFHS 2006 survey with care seeking for ALRI considered to be a proxy for meningitis. The median state-level estimate of access to a medical provider was 71% and ranged from 44%–89% across the states.¹² After adjusting for access to care, the median meningitis CFR was 36% and ranged from 21%–61% across states.

A proportion of survivors of Hib meningitis suffer lifelong disabilities. Studies reporting the risk of sequelae following bacterial meningitis have been conducted in Chandigargh,²⁰ Kerala,²¹ Varanasi,²² Hyderabad,²³ and Pune.²⁴ Hib-specific data were presented in 3 of the studies,^20,23,24 but the study in Hyderabad only assessed hearing deficit and, therefore, was excluded. The pooled risk of major sequelae following Hib meningitis from the 2 remaining studies in Pune (5 of 13) and Chandigarh (4 of 13) was 35%.^20,24 These proportions are similar to the proportion of children who suffered sequelae in the US and other countries prior to the use of Hib conjugate vaccines.

Various studies of clinical pneumonia incidence have been conducted among children aged less than 5 years at the community-level in India. In these studies, the incidence per child per year was reported to be 0.54 (Haryana),²⁵ 0.29 (Pune),²⁶ 0.86 (Rajasthan),²⁷ 0.31 (Delhi),²⁸ 0.07 (Maharastra),²⁹ 0.10 (Lucknow),³⁰ 0.4 (Tripura),³¹ 0.67 (Delhi),³² and 0.53 (Karnataka).³³ These studies vary in terms of geographical location, study design, period of reporting, and definition of ALRI. We assumed the IQR from all reported estimates to give a plausible national range (ie, 0.29–0.54 episodes per child per year). This is broadly consistent with previous estimates for India.³⁴

We assumed that the fraction of ALRI caused by Hib would be equal to the fraction of ALRI prevented by Hib conjugate vaccination (3.8%) in the only randomized controlled trial conducted in Asia (Lombok Island, Indonesia).⁷ The IQR of Hib pneumonia incidence was therefore 1102 (3.8% × 29 000) to 2052 (3.8% × 54 000) per 100 000 children <5 years of age. The state-level distribution of Hib pneumonia incidence was assumed to be the same as the state-level distribution for the prevalence of underweight children,¹² a consistently reported risk factor for pneumonia in India.^35,36 Underweight children have weight-for-age that is 2 or more SDs lower than the median weight-for-age of an international reference population, which includes healthy children from Brazil, Ghana, India, Norway, Oman, and the US.³⁷

Regional estimates of the proportion of mortality of children <5 years of age caused by pneumonia in children aged 1–59 months were reported in the Million Death Study: West (13%); South (8%); East (18%); North (15%); Northeast (16%); and, Central (17%).² These were multiplied by total deaths in children <5 years of age in each state to estimate pneumonia deaths in this age group.

The fraction of pneumonia deaths in children aged 1–59 months caused by Hib is a contentious parameter and an important driver of cost-effectiveness results.¹⁰ We derived this fraction by combining: (1) state-level estimates of Hib pneumonia incidence described above; (2) recent multisite estimates of hospitalised pneumonia incidence and CFRs, extrapolated to parent regions⁵; and (3) state-level estimates of access to medical care from the 2005–2006 NFHS survey.¹² We further assumed a 23.6% CFR for all severe untreated pneumonia based on a community-based trial among children 0–4 years in Gadchiroli, central India.³⁸ Using this approach (Figure 2), we estimated 7% of pneumonia deaths in children aged 1–59 months to be caused by Hib nationally, ranging from 1%–11% across states. The 7% estimate is broadly consistent with the pooled 5% reduction in radiologic pneumonia found when combining the results from the only 2 Hib conjugate vaccine studies conducted in Asia. Using inverse variance meta-analysis, 64% weight was assigned to the randomized controlled trial in Indonesia (vaccine effectiveness –10%; 95% CI −33%, 9%) and 36% weight to the case control study in Bangladesh (vaccine effectiveness 32%; 95% CI −2%, 54%).³⁹ This pooled estimate is, however, subject to large heterogeneity (I-squared 82%) and is far lower than previously reported global estimates. We, therefore, evaluated a separate scenario assuming Hib to be the cause of 20% of pneumonia deaths 1–59 months across all states. The Hib global burden of disease project estimated 21% globally in children aged 1–23 months based on a global meta-analysis using studies from Bangladesh, Indonesia, Chile, and The Gambia.¹³ This is similar to the fraction reported by a hospital etiology study conducted in New Delhi around 20 years ago. In this study, Hib was estimated to be the cause of 19% (21 of 110) of hospitalized pneumonia in children <5 years,⁴⁰ although the study did not define whether the positive results of latex agglutination were found in urine (lower validity) or serum (higher validity). Two other studies from New Delhi and Chandigarh, estimated Haemophilus influenzae to be the cause in 16% (20 of 122)⁴¹ and 13% (6 of 46)⁴² of hospitalized pneumonia cases respectively, but did not distinguish type b from nontypeable or other Haemophilus influenzae types.

In the multicenter Invasive Bacterial Infections Surveillance Project (Chennai, Lucknow, Nagpur, New Delhi, Thiruvananthapuram, Vellore) one case of invasive NPNM Hib disease was confirmed for every 5.5 cases of Hib meningitis.⁴³ We applied this ratio to the meningitis incidence rate to derive an NPNM incidence rate of 4 per 100 000 children <5 years of age. We assumed 4.3% CFR for those with access to medical care based on the CFR reported by the Invasive Bacterial Infections Surveillance Project for Hib-positive cases without associated meningitis. We doubled the CFR for children without access to medical care (8.6%), but this conservative assumption is likely to underestimate the true mortality burden for invasive NPNM diseases. The adjusted state-level CFRs ranged from 5%–7%.

Please see Table I for a full summary of the disease burden parameters used for each state.

For cases of Hib meningitis and Hib NPNM, we assumed that children with access to medical care would be admitted to hospital and also have 1 outpatient consultation, either as a referral or follow-up visit. For all cases of pneumonia, we assumed 1 outpatient visit per case for those with access to medical care and that a fraction of these would also be admitted to hospital. Severe clinical hospitalized pneumonia per 100 000 per year in children aged <2 years was recently reported in Chandigarh (2717), Vellore (3075), and Kolkata (7890). To avoid over-estimation of incidence, we converted incidence at age <2 years into incidence at age <5 years using national age distribution assumptions. We then used state-level estimates of clinical pneumonia incidence at age <5 years and access to medical care to calculate the fraction of cases with access to medical care who were hospitalized. We estimated that 3%, 6%, and 13% were hospitalized in each of the 3 sites. On this basis, we assumed 3% for all North and North Eastern States, 6% for the South and Western States, and 13% for the Central and Eastern States. We further assumed that these all-cause pneumonia fractions would apply to Hib pneumonia, to give median state-level incidence of admissions of 55 per 100 000 per year among children aged <5 years, ranging from 22–185 across states.

Costs for outpatient care differed according to whether the child was taken to a public or private facility, a traditional healer, or whether only drugs were purchased at a pharmacy. Costs for inpatient admissions varied with the level and type of hospital. The NFHS was used to estimate the proportion of children accessing each type of provider according to State (Table II).¹² The NFHS only includes a primary and a secondary/tertiary hospital level category for the public sector. To account for differences in costs between secondary and tertiary levels, we crudely assumed that tertiary-level care accounted for 5% of the reported visits/admissions in the combined secondary/tertiary category.

Treatment cost estimates are summarized in Table III. Two published sources were used. Household expenses were collected from the 60th round of the Government of India National Sample Survey Organization socioeconomic survey, conducted in 2004.⁴⁴ Questions on morbidity and health care were incorporated in the survey, including detailed questions about medical expenditures according to type of disease. Information was collected for every inpatient admission and outpatient visit for each member of the sample household during the 365 and 15 days preceding the survey, respectively. Household expenditures were categorized according to medicines, user fees, lodging, transport, etc. Results of the survey on respiratory ailments in children <5 years old were used for approximating household costs of pneumonia treatment. For this disease category, the nationwide sample sizes were 644 outpatient episodes and 238 inpatient admissions. The cost of a traditional healer consultation was assumed to be one-half the cost of a public sector clinic visit.

To estimate household costs of meningitis treatment, the pneumonia costs were adjusted upwards in accordance with the additional lengths of stay in hospital as reported in multicenter surveillance in Lucknow, Chennai, Vellore, and New Delhi in preparation for the Hib probe study.⁵ This was 6 days for pneumonia/NPNM and 10 days for meningitis. Opportunity costs, in terms of time spent while looking after a sick child, were included in the sensitivity analysis by assuming the minimal wage rate of US$3.4 per day,⁴⁵ multiplied by the average length of stay in hospital. There is no empirical evidence from India on the costs of treating a case of sequelae or the life-time earnings lost because of caring for a child with permanent disability, so we did not include sequelae costs in the base case. However, in a scenario analysis based on a recent sequelae costing study from Senegal, we assumed undiscounted lifetime costs of US$53 165 per child with sequel.⁴⁶ To allow discounting of future costs, these costs were converted into annual costs spread over the entire life expectancy of each state.

Government costs of meningitis and pneumonia treatment were derived from a micro-costing study by Krishnan et al undertaken in the State of Haryana in children aged <5 years.^47,48 Data were collected from 2 primary health centers, 6 secondary hospitals, and 2 tertiary hospitals; 6 of these were government facilities and 2 were private. Another study on costs of severe pneumonia from 2 non-government organization hospitals in Vellore was used as a comparator.⁴⁹

The age distribution of Hib disease also was derived from the multicenter surveillance study. Among children <2 years of age with confirmed Hib disease, 24% were aged <3 months, 20% 3–5 months, 21% 6–8 months, 11% 9–11 months, and 23% 12–23 months old.⁵ We assumed that 6% of children <5 years of age would occur between 24 and 59 months,⁵⁰ so proportions for <24 months were adjusted accordingly.

The original DALY formula and disability weights of 0.279 for pneumonia and 0.616 for meningitis were used.⁵¹ Because there are no standard disability weights available for any of the NPNM diseases, the pneumonia weight was used. The weighted average disability weight for meningitis sequelae was 0.34 based on the reported global distribution of sequelae syndromes⁵² and their respective disability weights.⁵¹ The most common types of single sequela from Hib meningitis are hearing loss and seizures, comprising 33% and 16% of sequelae cases, respectively. Multiple sequelae are seen in approximately 20% of sequelae cases.⁵²

To account for gradual or phased Hib conjugate vaccine introduction, we assumed 50% and 75% of diphtheria-tetanus-pertussis (DTP) coverage levels in the first 2 years and full DTP coverage thereafter. Coverage of the first 3 doses of DTP were based on the 2009 State-Level Coverage Evaluation Survey.⁵³ In the base case, we assumed no improvement in coverage over time, but this assumption was varied in scenario analysis to allow for annual improvements (reductions in the unvaccinated) over the period 2012–2030. Many children do not receive their vaccines according to the recommended 6-, 10-, and 14-week schedule. We, therefore, estimated the timeliness of vaccination (age-specific coverage) for each state using previously described methods.⁵⁴

Simple multiplication of vaccine efficacy and vaccine coverage is likely to overestimate the impact of vaccination because children who receive the vaccine may not be at the highest risk of mortality. To account for this relative coverage effect, we estimated DTP2 coverage of underweight children relative to the total DTP2 coverage reported for the cohort,¹² the implication being that underweight children are likely to be at higher risk of death, and that DTP2 would broadly represent the relative coverage for all three doses. Relative coverage multipliers (coverage in underweight children divided by coverage in the cohort) were calculated for each State (Figure 3).

To estimate the percent reduction in disease, the base case estimates accounted for state-level timeliness of vaccination, dose-specific coverage, and relative coverage. Vaccine efficacy was determined from a recent meta-analysis of controlled clinical Hib conjugate vaccine trials. In this analysis, efficacy against all forms of invasive Hib disease was 93% (95% CI 83%, 97%) following 3 doses, 92% (95% CI, 69%, 98%) following 2 doses, and 59% (95% CI 0%, 86%) following 1 dose.⁵⁵

Herd immunity and waning dose protection were not considered in the base case, but were included in scenarios. We assumed that vaccine protection could wane at a fixed rate of up to 5% per year, and that herd immunity could increase overall impact by up to 20%.⁵⁶

Four Indian companies produce and market Hib conjugate vaccine: Serum Institute of India (Pune), Panacea Biotec (New Delhi), Bharat Biotech (Hyderabad), and Biological E (Hyderabad). The presentations are monovalent Hib conjugate vaccine and Hib combined with DTP and hepatitis B vaccines (“pentavalent vaccine”). Although Indian vaccine procurement is normally processed between the government and the manufacturers directly, Global Alliance for Vaccines and Immunization Alliance funded vaccines are purchased through United Nations Children’s Fund, which only accepts WHO prequalified vaccines. The vaccine used in Tamil Nadu and Kerala is the 10-dose vial pentavalent vaccine produced by Serum Institute of India procured at a United Nations Children’s Fund price of US$1.75 per dose.⁵⁷ For the base case we assumed a cost of $1.82 per dose (including 4% tax) and no decline in dose price over time. A declining price trend was evaluated in scenario analysis.

When estimating the incremental costs of Hib conjugate vaccine, we calculated the cost difference between a schedule with pentavalent vaccine and with DTP and hepatitis B vaccines. Phased introduction of monovalent hepatitis B vaccine in a 10-dose vial started in 2002 with nationwide uptake in 2011. The 2010 prices per dose of DTP and hepatitis B vaccines were US$0.04 and US$0.11, respectively.⁵⁸ Because a 10-dose pentavalent vaccine vial is used, there is no need to allow for cold chain expansion.

First, we varied each parameter in turn by ±10% to establish the parameters with the greatest influence on the cost-effectiveness results (univariate 1-way sensitivity analysis). Second, we ran 19 alternative scenarios (10 favorable and nine unfavorable) to evaluate how sensitive the results were when we changed combinations of influential parameters (multivariate scenario analysis) (eg, given the uncertainty around the incidence of Hib disease in India), the most unfavorable scenario assumed a dramatically reduced incidence rate for both Hib pneumonia (50% of the base value) and Hib meningitis (32% of the base value) combined with several other unfavorable assumptions.

Between 2012 and 2031, Hib conjugate vaccination is estimated to prevent 207 859 undiscounted child deaths (<1% of deaths in children <5 years of age) assuming no benefit from herd immunity. Undiscounted lives saved were 127 869 for Hib pneumonia, 77 840 for Hib meningitis, and 2150 for Hib NPNM. The median reduction in total deaths of children <5 years of age was 0.9% and ranged from 0.3%–2.0% across states (Table IV and Figure 4).

The incremental costs of introducing Hib conjugate vaccination would be approximately US$127 million per year based on current vaccine prices (Table V). Without Hib conjugate vaccination, the cost of a fully vaccinated child (including monovalent hepatitis B) is US$2.19. Introduction of Hib conjugate vaccine increased annual costs four-fold, leading to costs per fully vaccinated child of US$8.81. The estimated total incremental cost for 2012–2031 was US$2006 million after discounting at 3% per year.

Around 15% of the vaccine costs would be offset by health care cost savings due to reduced cases of Hib disease. This percentage varied considerably by state; in the Central region around a one-quarter of the vaccine costs were offset by health care costs (Table IV). Total costs avoided over the 2012–2031 period would be US$310 million after discounting at 3% per year. Around 77% (US$240 million) would be avoided by households, with the remaining US$70 million by the government.

From a government perspective, state-level cost-effectiveness ranged from US$192-US$1033 per DALY averted after discounting costs and benefits at 3% per year. With the inclusion of household health care costs, cost-effectiveness ranged from US$155-US$939 per discounted DALY averted. The vaccine would be most cost-effective in the Central and Eastern States (Figure 4). States with the highest percentage reductions in under-5 deaths were as diverse as Madhya Pradesh, Jharkhand, West Bengal, and Kerala.

For a 10% change in each parameter, in all states the parameter with the largest percent impact on the discounted cost per DALY averted was relative coverage (12%–17% effect across states), a parameter rarely included in cost-effectiveness studies of vaccines. In all states, the parameter with the second largest influence was vaccine dose price (11%–16% effect). The influence and rank of other parameters varied by state, but the incidence of Hib pneumonia deaths and the efficacy of the vaccine against Hib pneumonia had important effects (7%–14%) in all states.

A variety of assumptions, favorable and unfavorable to Hib conjugate vaccine introduction, were considered in scenario analysis (Figure 5). Costs per DALY averted ranged from a cost saving scenario, in which health care cost savings are greater than the cost of the vaccine program itself, to an unfavorable scenario costing US$1830 per discounted DALY averted. Figure 5 shows the cost per DALY averted of the base case scenario (US$331) and the cumulative effect of introducing favorable and unfavorable assumptions in sequence. Hence, the most unfavorable scenario (US$1830) combines all unfavorable assumptions listed above the base case scenario, and the cost saving scenario combines all favorable assumptions listed below the base case scenario.