The Cost Effectiveness of Pandemic Influenza Interventions: A Pandemic Severity Based Analysis

Background The impact of a newly emerged influenza pandemic will depend on its transmissibility and severity. Understanding how these pandemic features impact on the effectiveness and cost effectiveness of alternative intervention strategies is important for pandemic planning. Methods A cost effectiveness analysis of a comprehensive range of social distancing and antiviral drug strategies intended to mitigate a future pandemic was conducted using a simulation model of a community of ∼30,000 in Australia. Six pandemic severity categories were defined based on case fatality ratio (CFR), using data from the 2009/2010 pandemic to relate hospitalisation rates to CFR. Results Intervention strategies combining school closure with antiviral treatment and prophylaxis are the most cost effective strategies in terms of cost per life year saved (LYS) for all severity categories. The cost component in the cost per LYS ratio varies depending on pandemic severity: for a severe pandemic (CFR of 2.5%) the cost is ∼$9 k per LYS; for a low severity pandemic (CFR of 0.1%) this strategy costs ∼$58 k per LYS; for a pandemic with very low severity similar to the 2009 pandemic (CFR of 0.03%) the cost is ∼$155 per LYS. With high severity pandemics (CFR >0.75%) the most effective attack rate reduction strategies are also the most cost effective. During low severity pandemics costs are dominated by productivity losses due to illness and social distancing interventions, while for high severity pandemics costs are dominated by hospitalisation costs and productivity losses due to death. Conclusions The most cost effective strategies for mitigating an influenza pandemic involve combining sustained social distancing with the use of antiviral agents. For low severity pandemics the most cost effective strategies involve antiviral treatment, prophylaxis and short durations of school closure; while these are cost effective they are less effective than other strategies in reducing the infection rate.


Influenza transmission model
In the simulation model we assumed that infectious transmission could occur when an infectious and susceptible individual came into contact during a simulation cycle. Following each contact a new infection state for the susceptible individual (either to remain susceptible or to become infected) was randomly chosen via a Bernoulli trail. Once infected an individual progressed through a series of infection states according to a fixed timeline.
The probability that a susceptible individual would be infected by an infectious individual was calculated according to the following transmission function, which takes into account the disease infectivity of the infectious individual Ii and the susceptibility of susceptible individual Is at the time of contact.
The baseline transmission coefficient β was initially chosen to give an epidemic with a final attack rate of 17.4% which is consistent with seasonal influenza as estimated in Table 3 of [3]. To achieve simulations under a range of basic reproduction numbers (R0), β was increased from this baseline value to achieve epidemics of various R0 magnitudes; details of the procedure for estimating β and R0 are given in [2]. A pandemic with a reproduction number of 1.5 corresponds to some estimations of the basic reproduction number of the 2009 H1N1 pandemic [4--7], while a reproduction number of 2.5 corresponds to some estimates of what may have occurred in the 1918/1919 pandemic [8,9].
The disease infectivity parameter Inf(Ii) was set to 1 for symptomatic individuals at the peak period of infection and then to 0.5 for the rest of the infectivity period The infectiousness of asymptomatic individuals is also assumed to be 0.5 and this applies to all infected individuals after the latent period but before onset of symptoms. The infection profile of a symptomatic individual was assumed to last for 6 days as follows: a 0.5 day latent period (with Inf(Ii) set to 0) is followed by 1 day asymptomatic and infectious, where Inf(Ii) is set to 0.5; then 2 days at peak infectiousness (with Inf(Ii) set to 1.0); followed by 2.5 days reduced infectiousness (with Inf(Ii)set to 0.5). For an infected but asymptomatic individual the whole infectious period (of 5.5 days) is at the reduced level of infectiousness with Inf(Ii) set to 0.5. This infectivity profile is a simplification of the infectivity distribution found in a study of viral shedding [10]. As reported below in the results section for the unmitigated no intervention scenario, these assumptions regarding the duration of latent and infectious periods lead to a mean generation time (serial interval) of 2.47 days which is consistent with that estimated for H1N1 2009 influenza [4].
Following infection an individual is assumed to be immune to re--infection for the duration of the simulation. We further assume that influenza symptoms develop one day into the infectious period [10], with 20% of infections being asymptomatic among children and 32% being asymptomatic among adults. These percentages were derived by summing the age--specific antibody titres determined in Table 5 of [11]. Symptomatic individuals will withdraw into the home with the following probabilities; adults 50% and children 90%, which is in keeping with the work of [12,13].
The susceptibility parameter Susc(Is) is a function directly dependent on the age of the susceptible individual. It captures age--varying susceptibility to transmission due to either partial prior immunity or age--related differences in contact behaviour. To achieve a realistic age specific infection rate, the age-specific susceptibility parameters were calibrated against the serologic infection rates for seasonal H3N2 in 1977--1978 in Tecumseh, Michigan [3]. The resulting age--specific attack rates are consistent with H1N1 2009 influenza [14,15], with a higher attack rate in children and young adults (details of the calibration procedure may be found in [2]).
The antiviral efficacy factor AVF(Ii,Is) = (1 --AVEi)*(1 --AVEs) represents the potential reduction in infectiousness of an infected individual (denoted by AVEi) induced by antiviral treatment, and the reduction in susceptibility of a susceptible individual (denoted by AVEs) induced by antiviral prophylaxis. When no antiviral intervention was administrated the values of both AVEi and AVEs were assumed to be 0, indicating no reduction in infectiousness or susceptibility. However, when antiviral treatment was being applied to the infectious individual the value of AVEi was set at 0.66, capturing a reduction in infectiousness by factor of 66% [16]. Similarly, when the susceptible individual was undergoing antiviral prophylaxis the value of AVEs was set to 0.85 indicating a reduction in susceptibility by a factor of 85% [16]. This estimate is higher than most previous modelling studies, which assume an AVEs of 30%. This common assumption appears to stem from an estimate made in [17] based on 1998--1999 trial data. Our higher value is based on a more comprehensive estimation process reported in [16], which also incorporated data from an additional study performed in 2000--2001 [18]. It is also in line with estimates of 64%--89% reported in [19].

Intervention Strategies
We examined a comprehensive range of intervention strategies including school closure, antiviral drugs for treatment and prophylaxis, workplace non-attendance (workforce reduction) and community contact reduction. These interventions were considered individually and in combination and social distancing interventions were considered for either continuous periods (that is, until the local epidemic effectively ceased) or periods of fixed duration (2, 4 or 8 weeks). Antiviral drug interventions and social distancing interventions were initiated when specific threshold numbers of symptomatic individuals were diagnosed in the community, and this triggered health authorities to mandate the intervention response. This threshold was taken to be 0.1% of the population. It was assumed that 50% of all symptomatic individuals were diagnosed, and that this diagnosis occurred at the time symptoms appeared.
For continuous school closure, all schools were closed simultaneously once the intervention trigger threshold was reached. For fixed duration (e.g. 2 weeks or 8 weeks) school closure, schools were closed individually as follows: for a primary school the whole school was closed if 1 or more cases were detected in the school; in a high school only the class members of the affected class were isolated (sent home and isolated at home) if no more than 2 cases were diagnosed in a single class; however if there were more than 2 cases diagnosed in the entire high school the school was closed. Note that these school closure policies were only activated after the community--wide diagnosed case threshold was reached; cases occurring in schools before this time did not result in school closure. This policy of triggering school closure based on epidemic progression avoids premature school closure which can reduce the effectiveness of limited duration school closure; see [20] for a detailed description of proposed school closure strategies.
Three antiviral drug strategies have been examined; antiviral drugs used solely for treatment of symptomatic cases (strategy T), T plus prophylaxis of all household members of a symptomatic case (strategy T+H), and T+H plus prophylaxis applied to the extended contact group (such as school or workplace contacts) of a symptomatic case (strategy T+H+E). Antiviral treatment (and prophylaxis for household or work / school group contacts) was assumed to begin 24 hours after the individual became symptomatic. It was assumed that an individual would receive at most one prophylactic course of antiviral drugs. Further details of antiviral interventions are given in [21,22].
Workforce reduction (WR) was modelled by assuming that for each day the intervention was in effect each worker had a 50% probability of staying at home and thus did not make contact with co--workers. Community contact reduction (CCR) was modelled by assuming that on days when the intervention was in effect all individuals made 50% fewer random community contacts. The most rigorous social distancing interventions considered in this study, which we denote as strict social distancing, involve the combined activation of school closure with workforce reduction and/or community contact reduction, and for this to occur for significant time periods; continuous and 8 weeks duration were considered.

Extended Prophylaxis and Rigorous Social Distancing
For strategies including rigorous social distancing (continuous school closure and community contact reduction), adding extended prophylaxis to antiviral treatment and household prophylaxis results in a slightly higher attack ratethis can be seen in Table 2 of the main paper.
This occurred because it was assumed that each individual would receive at most one prophylactic antiviral course: many prophylactic courses were used early on as a result of infections of school and work peers (which did not present a high transmission risk due to rigorous social distancing) and were then unavailable later when a household member (presenting a high transmission risk) became infected.
The result of extending prophylaxis to school and workplace contact groups in this scenario was that more infections occurred overall as prophylactic courses were used less efficiently, but that a slightly lower number of life years were saved as the infection (and mortality) burden was shifted slightly away from working adults to children. At lower severity categories this shift also resulted in a slightly lower total cost, as household and extended prophylaxis prevented additional lost work days compared to household prophylaxis.

Cost Effectiveness without Death Related Productivity Losses
The costing model used for the analysis in the main paper included future productivity losses arising from deaths caused by the pandemic. An alternative cost analysis for all interventions and severity categories omitting this cost component was also conducted. Figure S1.1 shown below is a counterpart to Figure 1 in the main paper, showing the cost effectiveness of interventions for severity categories 0, 1, 3 and 5. Table  S1.1 presents the total cost and cost per life years saved for each intervention strategy and all six severity categories.

Figure S1.1 Intervention cost effectiveness plain without death--related productivity losses
Cost effectiveness for each intervention strategy is plotted as a point in a two-dimensional plane, with points coloured by severity category: blue points --category 0 (case fatality ratio 0.03%), orange points --category 1 (case fatality ratio 0.1%), green points --category 3 (case fatality ratio 0.75%), red points --category 5 (case fatality ratio 2.5%). Horizontal axis represents the cost of the intervention strategy as a difference in total cost between two scenarios; an outbreak with the intervention in place and an outbreak with no interventions, expressed as dollars per member of the population. Vertical axis represents the number of life years saved by each strategy: the difference in life years lost for an outbreak with and without the strategy in place, expressed as life years saved per 10,000 population and also discounted. Compared to the costing methodology used in the main paper, the total cost of pandemics is lower when productivity losses due to pandemic--related deaths (DRPL) are excluded. The overall cost is reduced by a factor ranging from 2.9 at category 1 to 5.5 at category 5, for unmitigated pandemics. Tables S1.2 and S1.3 present the overall outcomes (LYS, total costs with/without DRPL, cost breakdowns, and cost per LYS with/without DRPL) for both costing methodologies and all simulated intervention strategies for severity category 1 and 5 respectively.
If DRPL costs are excluded, the strong correlation for high severity pandemics between intervention effectiveness (manifested as a reduction in deaths and an increase in life years saved) and the reduction in total pandemic cost becomes less pronounced. This can be seen by comparing the red (category 5) set of interventions in the cost effectiveness planes in Figure  S1.1 with those in Figure  1 in the main paper. For categories 3 and above the relative cost effectiveness of strategies within a given severity category remains unchanged, and the results are thus independent of which cost analysis method is used. Hence for high severity pandemics, the inclusion or exclusion of death related productivity losses (DRPL) leaves the qualitative outcome of the analyses unchanged.
For both cost effectiveness methodologies, and for all pandemic severity categories, the most cost effective intervention strategies are the same, namely continuous school closure and community contact reduction coupled with the use of antiviral agents for treatment and prophylaxis.
For lower severity pandemics (categories 1 and 2) the relative cost effectiveness of interventions differs slightly to that in the main paper, where DRPL costs are included. Strategies which combine school closure and antiviral measures are the most cost effective. Within this class of interventions, increasing the duration of school closure (from 2 weeks to 8 weeks to continuous closure) saves an increasing number of lives while slightly increasing the total pandemic cost. This contrasts with the analyses in the main paper where DRPL were included. In this case continuous social distancing interventions were significantly more cost effective (judged in terms of cost per LYS) than those of limited duration.
Although certain interventions with different durations of social distancing have similar cost effectiveness ratios for low severity pandemics, interventions with continuous social distancing are found to be more effective at reducing the attack rate and consequential mortality rates. For example, for category 1 pandemics a strategy of 2 weeks school closure with antiviral measures has a similar cost per LYS ratio to continuous school closure, community contact reduction and antiviral measures; however the latter results in 62% more life years saved. Life--years saved (LYS) expressed as years per 10,000 members of population. Total costs with death related productivity losses (bold) and total costs without death related productivity losses (bold shaded column) are expressed as dollars per member of population. Cost breakdowns and cost per LYS with and without (shaded column) death related productivity losses are also expressed as dollars per member of population for each intervention strategy. Both costs (death related productivity losses) and LYS are discounted. Both costs (death related productivity losses) and LYS are discounted. Cost categories are abbreviated as follows: HCC - health care costs (GP visits, hospitalisation and ICU usage), AVC - antiviral costs (cost of drugs, dispensing and stockpile renewal), ISDPL -illness and social distancing productivity losses, DRPL -death--related productivity losses. Interventions abbreviated as: SC -school closure; CCR -50% community contact reduction; WR -50% workforce reduction; 4, 8 -intervention duration in weeks; cont -continuous duration; T -antiviral treatment of diagnosed symptomatic cases; H -antiviral prophylaxis of household members of diagnosed symptomatic cases, E -antiviral prophylaxis of school class or workplace contacts of diagnosed symptomatic cases. Pure social distancing interventions are marked by *. Life--years saved (LYS) expressed as years per 10,000 members of population. Total costs with death related productivity losses (bold) and total costs without death related productivity losses (bold shaded column) are expressed as dollars per member of population. Cost breakdowns and cost per LYS with and without (shaded column) death related productivity losses are also expressed as dollars per member of population for each intervention strategy. Both costs (death related productivity losses) and LYS are discounted. Cost categories and interventions are abbreviated as for Table S1.2.

Sensitivity Analyses Transmissibility
The transmissibility of a future influenza pandemic will strongly influence the final attack rate (and consequent mortality rate), the effectiveness of interventions, and the total cost of the pandemic. We repeated our cost effectiveness analysis for all intervention strategies and severity categories for pandemic strains having unmitigated transmission characteristics both lower and higher than our primary assumption of a basic reproduction number R = 1.8, with low transmissibility setting having R = 1.5 and high transmissibility having R = 2.5. Figure S1.2 shows cost effectiveness planes for these alternative transmissibility settings, for all interventions that reduced the final attack rate by at least 50%, for severity categories 0, 1, 3 and 5.

Voluntary Household Isolation of Symptomatic Individuals
An underlying assumption of the simulation model is that 50% of adults (and 90% of children) who become symptomatically infected will withdraw to their home for the duration of their illness. Sensitivity analysis results for this parameter are given in Table S1. 4

. Increasing self--isolation decreases infection transmission and the final attack rate (and vice versa). These results indicate that cost effectiveness (in terms of cost per life year saved) is relatively insensitive to the voluntary isolation percentage.
The largest effect occurred at high severity if 25% rather than 50% of adults self-isolated. This rendered interventions less effective and consequently less cost effective. At low severity, decreased self--isolation caused an increased in productivity losses due to workplace absenteeism, but avoids productivity losses due to illness, resulting in relatively little effect on total cost and cost effectiveness.
If 75% instead of 50% of adults self--isolated, total pandemic costs are lower, but the number of life years saved by interventions is also lower, due to the fact that the total number of life years lost lower, resulting in little change to cost effectiveness.

Intervention--Related Parameters
Sensitivity analyses were conducted to assess the impact of variation of key intervention parameters. Alternative higher and lower parameter values for antiviral efficacy, compliance to home isolation during school closure, degree of workforce reduction and degree of community contact reduction were examined and the effect which these parameter settings have on the resulting attack rate, cost and cost effectiveness of interventions were determined. Results of these analyses are presented in Tables S1.5 to S1.8 and are summarised in the main paper. Final attack rates (%), total costs (shaded row) and cost per Life Year Saved (below shaded rows) are given for 4 intervention strategies (columns), for the baseline assumption that 50% of symptomatic adults (and 90% of symptomatic children) would stay at home, and for 2 alternate parameter values. Both costs (death related productivity losses) and LYS are discounted. Results are presented for category 1 (upper  table) and category 5 (lower table). Interventions are abbreviated as for Table S1.1. Cost and cost effectiveness values are given for 7 intervention strategies (columns), for the baseline best--estimate antiviral efficacy values and for 2 alternative efficacy parameter settings. Final attack rates (bold face) and LYS per 10,000 member of population (in bracket) are given above shaded cells. For no intervention total cost without death related productivity losses is given below the final attack rate. For each intervention, antiviral costs are given in shaded cell followed by combined health care cost, and illness and social distancing productivity losses.     Final attack rates, life years saved (LYS) per 10,000 population, total costs per person and cost per LYS are given for 5 intervention strategies (row blocks). Results are given assuming that susceptibility to infection is uniform for all ages (grey columns), and for a 2009--like age--specific susceptibility profile (unshaded columns). Results are presented for 6 severity categories (rows). Both costs (death related productivity losses) and LYS are discounted. Interventions are abbreviated as for Table S1.1 .  The table shows the lives saved by the most cost effective strategies for three different pandemic severity categories for Australia (population 23 million). Pandemic costs are shown with and without (shaded rows) death related productivity losses. Death related productivity losses are discounted. Interventions are abbreviated as for Table S1.1 . Table S1. 12 Intervention strategy, cost effectiveness (with death related productivity losses) and life years saved