Table 3.

Methodological details of studies

Study	Analysis type	Methods employed	Details of method and any model used	Key limitations
Alleweldt et al. 2021	Cost–benefit analysis of WGS for pathogen surveillance across different public health laboratories in different countries	Case study of eight reference laboratories routinely using WGS, of which the five using WGS for bacterial pathogens are considered here Costs of WGS compared to next-best conventional method calculated Break-even analysis	Costs of WGS and conventional methods considered from laboratory perspective: equipment, consumables, staff and other costs, and included estimated lifespans, rate of use and maintenance Cost of illness included healthcare utilization, productivity loss and premature death (value of statistical life method) Break-even analysis considered salmonellosis	Underdiagnosis/underreporting multiplier not used, therefore total cost of illness likely underestimated, and break-even point likely overestimated
Brown et al. 2021	Cost–benefit analysis of WGS for foodborne illness source tracking in the USA	Theoretical model using social welfare maximization framework Tested in an empirical model using data from FDA and NCBI databases	Theoretical model included: net value of food production minus the total burden of foodborne illness associated with food production, minus the implementation costs of the programme Empirical models included: pathogen, year, average illnesses per outbreak and the no. of sequences for the pathogen in NCBI in given year, food vehicle implicated, indicators to establish effects of Food Safety Modernization Act in different years Cost–benefit model: reduction in illness with increases in WGS isolates in NCBI, underreporting/underdiagnosis multiplier and estimated burden of illness related to each pathogen	Model may not account for complete societal costs of foodborne illness cases
Elliott et al. 2021	Cost-effectiveness/cost–utility analysis of WGS and environmental metagenomics for surveillance and management of CRAB in an Australian hospital	A hybrid agent-based and discrete-event model looking at use of WGS and environmental metagenomics in three scenarios modelled using observed outbreak data Healthcare costs and health utility considered	Used AnyLogic (AnyLogic, Chicago, IL, USA) to model the burns unit over 32 months Scenario 1 used actual time point – WGS of CRAB isolates commencing shortly after the outbreak was detected, and environmental metagenomics being introduced more than 1 year later. Scenario 2 assumed WGS use prior to the start of the outbreak and delayed introduction of environmental metagenomics, and scenario 3 assumed prior WGS and metagenomics use	Model based on a single outbreak, single pathogen and single unit, limiting generalizability
Ford et al. 2021	Cost analysis of WGS for public health surveillance of non-typhoidal Salmonella enterica in Australia	Costs per case calculated for notified cases Cost per sample calculated based on billing by five reference laboratories Break-even analysis for WGS vs serotyping and MLVA Costs modelled in three simulated outbreak scenarios	Cost per case included direct and indirect healthcare costs, productivity lost, premature mortality Modelled scenarios: point source outbreak, prolonged outbreak without peak, prolonged outbreak with late peak. In lieu of an effectiveness estimate, intervention at 30, 60 or 90 days earlier on the epidemiological curve was modelled	Effectiveness estimate not used – unclear how intervention points in modelled outbreak scenarios may apply in real outbreaks Underreporting/underdiagnosis multiplier not used, therefore societal costs likely underestimated
Gordon et al. 2021	Cost–benefit/cost-effectiveness of surveillance for six common multidrug-resistant bacteria across hospitals in Queensland, Australia	National data on HAI (9.9%) subdivided by organism and site used to estimate colonizations, infections and deaths in Queensland (16 hospitals) Sequencing data derived from WGS surveillance project performed for research purposes locally (1783 isolates). An SNP threshold <5/Mb used to define a cluster, of which 2–18 clusters identified per pathogen involving 5–13 patients Five-year budget impact analysis – costs for provision of care	Model accounted for changes in resistance rates over the time period Sensitivity analysis performed	Assumption that transmission is successfully broken when WGS data are acted upon, with turnaround time of 7 days for data Sequencing set-up costs not included
Kumar et al. 2021	Cost-effectiveness study of WGS for infection prevention and control in a US hospital	Transmission network for 11 outbreaks (1 hospital) involving 89 patients built Transmissions averted by WGS (compared to standard of care) estimated from model Deaths averted estimated Change in cost calculated	Estimates of transmissions averted based on effectiveness of intervening against the suspected transmission route, time from transmission to positive culture and time from time to obtain and act on WGS results Deaths averted calculated by attributable mortality to infections and different anatomical sites Costs included sequencing, staffing for the IPC team and treatment of infections	Deaths averted not considered in economic terms (value of statistical life) Sequencing cost per isolate considered $72.13 (70), which is the lowest in this review Higher acuity of WGS in identifying outbreaks means likely underestimate of benefits due to detection of occult outbreaks/transmission
Dymond et al. 2020	Cost–utility study of WGS surveillance of Staphylococcus aureus in hospitalized inpatients in the UK	Estimation of MRSA cases averted Costs of treatment of infection or colonization, costs of sequencing and collecting samples were calculated Mortality and QALYs estimated	MRSA case numbers taken from previous prospective cohort study (one laboratory covering three hospitals) Probability of MRSA acquisition assumed 0.5 % and reduction in MRSA due to genome sequencing assumed 90% Cost of WGS $134.29 (£100)/isolate, cost of ITU care and cost of colonization only considered Mortality estimate from Klein et al. 2019, QALY decrement 0.35	Probability of MRSA acquisition and the estimated efficacy of WGS are both hypothetical figures, with efficacy of WGS assumed 90%, although addressed to some extent in sensitivity analysis Cost of treating infection assumed intensive care admission, resulting in overestimation. As cost of treatment savings the major driver of cost savings in model, this may have important implications
Lee et al. 2020	Cost–benefit analysis of WGS surveillance for MRD E. coli in a hospital outbreak in Queensland, Australia	Stochastic hybrid discrete-event, agent-based model using AnyLogic software Evaluating early vs no WGS in different outbreak scenarios (also includes delayed use of WGS): actual outbreak data, simulated outbreak with environmental transmission and outbreak with an increased virulence pathogen	Considered healthcare costs including environmental cleaning, bed closure costs and outbreak team costs	Costs for treating infection (and death) only included in scenario of increased virulence The authors note that the outbreak is modelled on an actual OXA-181 outbreak that did not cause clinical infection in any patients
Jain et al. 2019	Cost–benefit analysis of WGS for public health surveillance of non-typhoidal Salmonella enterica from specific food vehicles (fresh produce, poultry and eggs) in Canada	Salmonellosis cases attributable to fresh produce, poultry and eggs estimated Cost of illness estimated for PFGE and WGS Three scenarios modelled DALYs and QALYs calculated for PFGE but NOT for WGS	Notified cases, underdiagnosis/underreporting multiplier, and likely percentage attributable to food vehicles of interest used to estimate case numbers – severity of illness calculated with estimates from Hoffman et al. 2012 Cost of illness included direct and indirect costs, premature mortality, costs of surveillance calculations use estimates from Thomas et al. 2015, QALYs (derived from Hoffman et al. 2012) and DALYs (derived from [5]) Modelled scenarios: two historic outbreaks used to model effect of (1) 70 % reduction in cases, (2) 50 % reduction in cases, (3) net benefits for all notified cases in Canada	Considered food vehicles account for 49 % of salmonellosis cases in Canada Model does not account for any direct cost difference between the two technologies compared (equipment, personnel, etc), and only mentions in the discussion that expert opinion estimates consumable costs to be lower for WGS than PFGE $115.12 vs $219.45 per isolate (135.6 vs 258.49), with $257 098.95 (302 837) in equipment costs DALYs and QALYs calculated for PFGE but not for WGS