Cost-Effectiveness Analysis of Integrated Bite Case Management and Sustained Dog Vaccination for Rabies Control

ABSTRACT. The successful prevention, control, and elimination of dog-mediated rabies is challenging due to insufficient resource availability and inadequate placement. An integrated dog bite case management (IBCM) system plus dog vaccination can help address these challenges. Based on data from the IBCM system in Haiti, we conducted a cost-effectiveness evaluation of a newly established IBCM system plus sustained vaccination and compared it with 1) a no bite-case management (NBCM) and 2) a non–risk-based (NRB) program, where bite victims presenting at a health clinic would receive post-exposure prophylaxis regardless of risk assessment. We also provide cost-effectiveness guidance for an ongoing IBCM system and for sub-optimal dog vaccination coverages, considering that not all cost-effective interventions are affordable. Cost-effectiveness outcomes included average cost per human death averted (USD/death averted) and per life-year gained (LYG). The analysis used a governmental perspective. Considering a sustained 5-year implementation with 70% dog vaccination coverage, IBCM had a lower average cost per death averted (IBCM: $7,528, NBCM: $7,797, NRB: $15,244) and cost per LYG (IBCM: $152, NBCM: $158, NRB: $308) than NBCM and NRB programs. As sensitivity analysis, we estimated cost-effectiveness for alternative scenarios with lower dog-vaccination coverages (30%, 55%) and lower implementation costs. Our results suggest that better health and cost-effectiveness outcomes are achieved with the continued implementation of an IBCM program ($118 per life-year saved) compared with a newly established IBCM program ($152 per life-year saved). Our results suggest that IBCM is more cost-effective than non-integrated programs to eliminate dog-mediated human rabies.

Notes. For 30% vaccination coverage the probability that the offending dog was rabid is 6.3%, and the probability of developing rabies in the absence of PEP is 19%. For 55% and 70% vaccination coverage the probability that the offending dog was rabid is reduced from 6.3% to 1%, and probability of developing rabies in the absence of PEP is 19% ΩIBCM where the assumption that 100% of bite victims seeking health care after HARSP advice, and capital costs for surveillance and diagnostic components of IBCM are removed (scenario 2) i. Puppies born to rabid dogs will not survive and therefore exposed and infected dogs are not included in the birth rate.
ii. Adjustment factors = dog rabies/reported dog rabies = 1 iii. Adjusted bites = reported dog rabies* adjustment factor. Where reported dog rabies included confirmed, probable, and suspected dogs.

Health Seeking Behaviour Assumptions
i. Estimated human rabies infections were calculated from the proportion of people bitten, categorized as confirmed, probable, or suspected, the probability that the offending dog was rabid (Wallace et al., 2015), and the probability of acquiring rabies if exposed with no PEP (Hampson et al., 2009), and the vaccination level being assessed.
ii. To calculate the number of fatal infections, the estimated human rabies infections were combined with the probability that a bite victim sought medical care and received PEP.
iii. This study assumed that those patients who sought medical care and received PEP did not develop rabies, regardless of the reported overall compliance (i.e., completing the five-dose course).

v.
Fatal rabies infection=(share of patients who seek care)*adjusted human rabies infections*(1-% of confirmed, probably, suspect individuals who receive PEP)+(1-(share of patients who seek care + share of patients who seek care with active bite investigation))*adjusted human rabies infections.

Program-Based Assumptions /Cost Effectiveness Assumptions
i. This study uses generalized WHO-Choice (Choosing Interventions that are Cost-Effective) thresholds as a willingness-to-pay indicator. ii.
WHO-CHOICE defines a threshold as three times gross domestic product (GDP) per capita, to identify the cost-effectiveness of interventions. GDP was adjusted to 2018 rate. iii.
Net monetary benefit (NMB) was estimated as a function of WHO-CHOICE for an additional unit of effectiveness, as defined by year of life gained (compared to the situation in Haiti), and calculated as: NMB = WHO-CHOICE * Δ effectiveness -Δ cost iv.
An ideal life expectancy, taken from GBD 2010 was used, to avoid attributing higher weights to deaths in more affluent communities (Murrey et al.,2012). v.
YLLs were calculated by multiplying deaths at each age by the reference standard life expectancy at that age as estimated by BDB 2010 (Murrey et al., 2012). vi.
We assumed that the age distribution of fatal human rabies infections in Haiti is equivalent to that of Tanzania, which has already been characterized (Cleaveland et al. 2002). vii.
Estimations were given for years of useful life for the equipment. Annual capital costs (US$) estimated show the equivalent annual cost for the capital outlay, assuming that the resale value is zero. Costs were estimated using constant dollars (with no inflation) and used a real discount rate of 3% (Drummond et al., 2005). viii.
To complete robustness checks of IBCM, we assumed that all vaccination coverages explored (30, 55, 70%) were achieved and maintained, ix.
To complete a robustness check of IBCM we assumed that as the HARSP continues and vaccination at 30% is achieved year on year over ten years, the probability of being bitten by a rabid dog should reduce as a function of the 30% vaccination coverage, coupled with the counseling component of HARSP.

Accounting for uncertainty
The share of PEP paid for by the government was varied at 0, 50, and 100%, with the probability that patients who seek medical care was kept at 54% (HARSP baseline data), and upper (85%) and lower (15%) bounds applied for comparison.
Uncertainty was accounted for by adjusting the share of PEP paid by the Haitian government, (lower bound 0%, expected value 50%, and upper bound 100%) and health indicators as a function of this uncertainty, assessed.
To complete a robustness check of HARSP, with a reduction in the number of rabid dogs in the population, we assumed that the probability of being bitten by a rabid dog would decrease and in turn so too would the probability of developing rabies as a result.
We assumed that the continued implementation of HARSP should result in fewer rabid dogs diagnosed and more non-rabid dog bites. Our assumption of a lower vaccination rate of 30% will mean that initially, there is a higher probability that fewer dogs will be immune to rabies, when compared to a higher vaccination rate, keeping all other variables constant. We assume that over time a reduction in the probability of the offending dog being rabid will occur and amend the values from 6.3 to 1.3 to reflect this. The probability that the offending dog is rabid is estimated based on empirical data and the aggregate value of all dog investigations available per year, as described by Wallace et al. 2015 and estimated at 6.3%. A reduction to 1.3% probability is computed using an estimate by year: A confirmed exposure = confirmed exposure*number of confirmed in study per year + probable exposure*number probable exposures in stud per year / number of dog investigations/year. Figure S1. (Scenario 1). Sensitivity analysis for the new implementation of an IBCM, by varying the proportion of patients who seek medical care after a bite incident out of the total number of patients who reported to IBCM (lower bound:15%, baseline: 54%, upper bo the probability that the offending dog was rabid (lower bound: 1%, baseline: 6.3%, upper bound: 36%), to assess the impact on fatal number of rabies infections. dog was rabid (b) 6.3% probability tha offending dog was rabid.

Supplementary Figures
1). Sensitivity analysis for the new implementation of an IBCM, by varying the proportion of patients who seek medical care after a bite incident out of the total number of patients who reported to IBCM (lower bound:15%, baseline: 54%, upper bound: 85%) and varying the probability that the offending dog was rabid (lower bound: 1%, baseline: 6.3%, upper bound: 36%), to assess the impact on fatal number of rabies infections. (a) 1 % probability that offending dog was rabid (b) 6.3% probability that the offending dog was rabid (c) 36% probability that 8 1). Sensitivity analysis for the new implementation of an IBCM, by varying the proportion of patients who seek medical care after a bite incident out of the total number of nd: 85%) and varying the probability that the offending dog was rabid (lower bound: 1%, baseline: 6.3%, upper bound: % probability that offending offending dog was rabid (c) 36% probability that the