Cost-effectiveness of interventions to improve hand hygiene in healthcare workers in middle-income hospital settings: a model-based analysis

Summary Background Multi-modal interventions are effective in increasing hand hygiene (HH) compliance among healthcare workers, but it is not known whether such interventions are cost-effective outside high-income countries. Aim To evaluate the cost-effectiveness of multi-modal hospital interventions to improve HH compliance in a middle-income country. Methods Using a conservative approach, a model was developed to determine whether reductions in meticillin-resistant Staphylococcus aureus bloodstream infections (MRSA-BSIs) alone would make HH interventions cost-effective in intensive care units (ICUs). Transmission dynamic and decision analytic models were combined to determine the expected impact of HH interventions on MRSA-BSI incidence and evaluate their cost-effectiveness. A series of sensitivity analyses and hypothetical scenarios making different assumptions about transmissibility were explored to generalize the findings. Findings Interventions increasing HH compliance from a 10% baseline to ≥20% are likely to be cost-effective solely through reduced MRSA-BSI. Increasing compliance from 10% to 40% was estimated to cost US$2515 per 10,000 bed-days with 3.8 quality-adjusted life-years (QALYs) gained in a paediatric ICU (PICU) and US$1743 per 10,000 bed-days with 3.7 QALYs gained in an adult ICU. If baseline compliance is not >20%, the intervention is always cost-effective even with only a 10% compliance improvement. Conclusion Effective multi-modal HH interventions are likely to be cost-effective due to preventing MRSA-BSI alone in ICU settings in middle-income countries where baseline compliance is typically low. Where compliance is higher, the cost-effectiveness of interventions to improve it further will depend on the impact on hospital-acquired infections other than MRSA-BSI.


Introduction
Hospital-acquired infections (HAIs) are a major cause of morbidity and mortality among hospitalized patients [1]. HAIs are also associated with a substantial economic burden due to longer hospital stays and additional antibiotic costs [2]. The risk of infection in developing countries is two to 20 times higher than in developed countries [3]. In Thailand, among hospitalized patients, the point prevalence of nosocomial infection has been estimated at 6.5% and w250,000 patients are believed to have an HAI each year [4].
Direct patient contact with healthcare workers (HCWs) transiently contaminated with nosocomial pathogens is believed to be the primary route of transmission. Improving HCW hand hygiene compliance can minimize the impact of this transmission route and reduce the incidence of nosocomial infection [5]. A multi-modal intervention including system change, training and education, observation and feedback, reminders, and a hospital safety climate has been developed and promoted by the World Health Organization (WHO). This campaign (referred to as  has been shown to be effective in increasing hand hygiene compliance [5,6]. Hand hygiene promotion is also relatively easy to implement and requires a modest level of investment. Nevertheless, in many healthcare settings, particularly in lowand middle-income countries, compliance remains poor and reports of rates of <10% may be typical [7e9]. Transmission dynamic models are useful tools to help understand the likely impact of interventions to control communicable diseases. Moreover, their use in health-economic evaluations of interventions that reduce transmission is essential to fully capture intervention benefits. However, whereas several studies have used dynamic models to consider hospital infections [10,11], economic evaluations of hand hygiene interventions have used only static models and have largely neglected developing countries where the need for appropriate investment is greatest [12e16]. Whereas one previous cohort study in Vietnam concluded that a hand hygiene intervention was cost-saving (i.e. the reduction in costs from HAIs averted exceeded the intervention cost), there have been no systematic attempts to quantify the levels of investment in hand hygiene promotion under which it remains cost-effective or to explore how appropriate levels of investment depend on preintervention levels of hand hygiene compliance [16].
The aims of this study are to develop a dynamic modelbased framework for evaluating the cost-effectiveness of hand hygiene promotion interventions and use it to evaluate the cost-effectiveness of such interventions in a middleincome country. Our analysis is informed by data from a typical regional hospital in Thailand, a middle-income country with a gross domestic product (GDP) per capita approximately equal to the world median. We focus on MRSA-BSI as this is one of the most serious and best-studied types of infection in ICU patients, there is clear evidence of frequent patient-to-patient transmission of MRSA, and evidence that such transmission can be interrupted by improved hand hygiene [17e19]. Hand hygiene interventions should also reduce other types of MRSA infections and infections with other organisms. However, since these are harder to quantify, we take a conservative approach by focusing on MRSA-BSI alone and are therefore likely to underestimate the true health benefits of the intervention [20].

Overall description
Transmission dynamic and decision analytic models were combined to simulate the transmission dynamics and evaluate the impact and cost-effectiveness of hand hygiene interventions. Two ICU settings were considered: a paediatric intensive care unit (PICU) and an adult intensive care unit (adult ICU). Epidemiological and economic parameters were derived from detailed local data from a typical tertiary hospital in Northeast Thailand. Information about catchment area, staff:patient ratios, and further details about the ICUs have been described elsewhere [21,22]. Incidence of hospital-acquired and healthcare-associated BSI from this and other hospitals in the same region have also been reported in previous studies [22,23].

Transmission dynamic model
A previously described deterministic hostevector model was constructed to simulate MRSA transmission dynamics in an ICU ( Figure 1) [24]. Patients can be admitted to the ICU either uncolonized or colonized with MRSA. Uncolonized patients can become colonized or infected by contact with transiently colonized HCWs. HCWs can be decolonized by performing hand  Model parameters were obtained from various sources (Table I). MRSA carriage data were derived from a previous observational study in North-east Thailand [22,27]. Estimates of the per-contact transmission probabilities (HCW to patient and patient to HCW) were derived using these data, combined with previous estimates of the probability of transmission from a colonized patient to HCW (see Appendix A) [27,40]. The rates at which colonized patients acquired an MRSA-BSI were estimated from the average number of cases per year at each ward divided by the expected number of colonized bed-days (estimated from the carriage data). Risk of death due to MRSA-BSI was taken from an observation study in the same setting [22]. The number of beds, number of HCWs per shift, rates of ICU discharge, wardspecific contact rates, and the baseline hand hygiene compliance were directly observed from the same hospital.

Economic evaluation
Cost-utility analysis was performed from a healthcare provider's perspective. The cost of the hand hygiene intervention was estimated over a period of one year. Health benefits were measured with a lifetime horizon and a 3% discounting rate. Costs were adjusted to 2016 values with the exchange rate of 33 Thai Baht per US$1 [41]. This economic evaluation is in accordance with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement [42].
There were two main cost components: cost of hand hygiene promotion; and costs associated with MRSA-BSI. The latter includes costs of additional hospital stay and antibiotic treatment ( Table I). The cost of the hand hygiene campaign accounted for staff time and materials used. As there is limited information for the cost of hand hygiene interventions in Thailand, we derived this information from a survey conducted in Australia from the national hand hygiene campaign implementing interventions similar to WHO-5 assuming the same time per bed-day requirements but applying Thai pay-scale salaries for registered nurses with two and 10 years' experience [28,29]. Costs of alcohol hand rub (AHR) were included in the model; other material costs were assumed to be negligible. The intervention was assumed to increase AHR use 3.5-fold (range: 2e5) [30]. Baseline AHR use was directly observed in all local paediatric and adult ICUs. We found similar amounts of AHR used in both types of ICUs, therefore the average AHR use (98 L per ICU at 10% compliance) was applied in both wards. Associated costs assumed a market price of US$2.4 per litre provided by the national pharmaceutical supplier in Thailand [31]. Total hand hygiene intervention costs were estimated to be US$675 and US$720 per ward per year in the PICU and adult ICU, respectively, due to the difference in number of beds per ward (seven and 10 beds, respectively).
Costs associated with MRSA-BSI were estimated from additional stay and treatment. Hospitalization cost was calculated as the excess length of stay due to MRSA-BSI multiplied by the cost per bed-day. Retrospective data from routine clinical and microbiological laboratory databases at the local hospital (2003e2010) were used to identify MRSA-BSI cases. Additional stay due to infection was estimated with a multi-state model accounting for time-dependent bias using the 'etm' package within R [43,44]. The economic value of a bed-day should reflect the opportunity cost of an occupied bed, which can be quantified by asking healthcare providers for their willingness to pay (WTP) for an unoccupied bed-day [45]. This opportunity cost is typically much lower than the cost calculated with an accounting approach (hospital budget divided by the total patient bed-days) [32]. In the absence of WTP per ICU bed-day in Thailand, the accounting cost was estimated using local hospital financial data and this was multiplied by the ratio of bedday costs estimated with WTP and accounting approaches reported in a previous study [32].
Treatment for MRSA-BSI was assumed to require a 14-day course of vancomycin with dose regimens following treatment guidelines for hospital-acquired MRSA-BSI. Drug costs were obtained from the Drug Medical Supply Information Center [33].
Estimates of life expectancy among post-ICU patients were taken from a previous study in North-east Thailand [21].
Health-related quality of life among patients after ICU discharge who had MRSA-BSI during the ICU stay was assumed to be the same as general post-ICU patients. These utility values were taken from the literature [36e39]. The median utility of 0.72 was used in the base case with a range from 0.56 to 0.88 and assumed to be constant.

Analyses
Four scenarios with different baseline versus postintervention hand hygiene compliance values were considered: (a) 10% vs 20%; (b) 10% vs 40%; (c) 10% vs 60%; and (d) 40% vs 60%. These are consistent with results from a systematic review where odds ratios were estimated to be 6.5 and 11.8 for WHO-5 and WHO-5 plus other interventions among studies using an interrupted time-series design (with a baseline compliance of 10%, these would give post-intervention compliance values of 42% and 57%, respectively) [46].
In each comparison, point estimates of incremental costs (DC) and QALYs gained (DQ) due to the intervention and the incremental cost-effectiveness ratio (ICER; DC/DQ) were calculated. The threshold willingness to pay per QALY gained (l) was taken as GDP per capita (US$4848), and a threshold value of three times GDP per capita was considered in a scenario analysis [47]. The latter threshold corresponds to WHO criteria for a cost-effective intervention and the former to a highly costeffective intervention [34].  the chosen WTP threshold are, by definition, cost-effective. Probabilistic sensitivity analyses (PSA) were undertaken using 10,000 Monte Carlo iterations where parameters were sampled from specified distributions (Table II). Simulation results were used to calculate the monetary net benefit (MNB), which is defined as l*Q e C for each level of achieved hand hygiene compliance and the distribution of incremental monetary net benefits (IMNB) for each comparison (l*DQ e DC). In addition, the maximum level of investment in the intervention at which it would still be cost-effective was calculated as monetary incremental benefits (l*DQ) plus the saving in treatment costs from averted infections. A series of hypothetical scenarios with different assumptions about the transmissibility and prevalence of MRSA colonization at admission were considered. The ward reproduction number (R A ), the expected number of MRSA cross-transmissions resulting from a single colonized patient during a single ward stay, assuming all other patients on the ward are susceptible, was varied between 0.5 and 5, while prevalence of MRSA colonization on ICU admission was varied between 1% and 15%.
Changes in costs and health outcomes under different baseline compliance and improvement levels were calculated and combined to evaluate the cost-effectiveness of such interventions in terms of the IMNB. The maximum level of investment at which the intervention would still be costeffective, the prevalence reduction, and final prevalence in all scenarios were also determined.

Results
Under base case assumptions (with a pre-intervention hand hygiene compliance of 10%), a multi-modal hand intervention (WHO-5) is highly likely to be cost-effective in both PICU and adult ICU settings if it increases hand hygiene compliance to 20% (Table II and Appendix B). Conversely, if baseline compliance is 40%, the expected IMNB is likely to be negative, indicating that the intervention is unlikely to be cost-effective solely as a result of reducing MRSA-BSIs (Figure 2). Risk of death due to MRSA-BSI in our study hospital was estimated to be between two-and three-fold higher than in high-income countries [22,35]. However, a scenario analysis showed that the intervention is still highly cost-effective if a lower mortality risk estimated from high-income settings is used instead (Table III). When the WTP threshold was three times GDP per capita (US$14,545), under base case assumptions (with a pre-intervention hand hygiene compliance of 10% and post-intervention compliance of 40%), the IMNBs were positive for both PICU and adult ICU.
In hypothetical scenario analyses, the hand hygiene intervention was found to be cost-effective in most scenarios, especially when there was high transmissibility and a high prevalence of MRSA colonization on admission (Figure 3). In the situations where the transmissibility is low (R A ¼ 0.5), where prevalence of MRSA colonization at admission is 5%, and where baseline compliance is 20% in both the PICU and the adult ICU, the intervention is always cost-effective even with only a 10% compliance improvement using the cost estimates in Table I. When the baseline compliance is 20%, the intervention will always be cost-effective if the intervention cost per year is less than US$1557 in the PICU and US$888 in the adult ICU, provided that the intervention increases compliance by 10%. Prevalence reduction of MRSA carriage and the final prevalence of all hypothetical scenarios are shown in Figure 4.

Discussion
Hand hygiene promotion using the WHO multi-modal campaign is likely to be highly cost-effective for ICU settings in Thailand where baseline compliance is low (20%) solely as a result of preventing MRSA-BSI. Factors that tended to make the intervention more cost-effective were low baseline compliance, high prevalence of colonization at admission, and high rates of transmission. With higher baseline compliance, the intervention may often still be highly cost-effective as a result of reduced MRSA-BSI alone if rates of MRSA carriage on ICU admission or ICU transmission are sufficiently high.
Because we ignored impacts of the intervention on other types of HAI (other MRSA infections and infections with other pathogens) our analysis is conservative and likely to underestimate health benefits. MRSA-BSIs represent only 5.1% of hospital-acquired BSIs in North-east Thailand, whereas Gramnegatives account for 67.6% [23]. Evidence linking increased hygiene with reduced infection rates is less compelling for Gram-negative organisms than it is for MRSA, but there are credible reports that such an association exists [48,49]. Some of the strongest evidence concerns multidrug-resistant Acinetobacter spp., where a segmented regression analysis found that a hand hygiene intervention was associated with a substantial change in incidence of infections with extensively drug-resistant Acinetobacter spp. in Taiwan [13]. Acinetobacter spp. has also been reported to be a frequent contaminant of the hands of HCWs in endemic settings in South-east Asia, strengthening the evidence for a causal link between increased hand hygiene and reduced infections [50]. Since Acinetobacter spp. are the largest single cause of hospitalacquired BSI in North-east Thailand, if infections with these organisms can be reduced substantially by improved hand hygiene, the implications for our analysis could be substantial [22]. Whereas our work provides an analytical framework for such an evaluation, better data on the epidemiology of Acinetobacter spp. and the effects of hand hygiene are needed to inform it.
To the best of our knowledge, our study is the first economic evaluation of a hand hygiene intervention to make use of a dynamic model in a developing country [12e16]. In previous economic evaluations in high-income countries, Pittet et al. (in Switzerland), and Chen et al. (in Taiwan) used data from observational studies to estimate reductions in infections due to hand hygiene interventions. Pittet et al. concluded that if only 1% of the observed reduction was due to the intervention it would have been cost-saving. Chen et al. also concluded that their intervention was likely to be cost-saving. Huis et al. (in the Netherlands) used trial data to inform a cost-effectiveness analysis, assuming a linear relationship between hand hygiene compliance and reduced infections, concluding that the intervention was likely to be cost-effective if the willingness to pay for a 1% reduction in the HAI rate was about US$6000. A study in the UK also concluded that hand hygiene interventions were likely to be cost-saving even if the reduction in rates of HAI were as low as 0.1%. As in our study, this report explicitly calculated QALY gains. However, unlike the other studies, staff time was not accounted for when costing the intervention. A previous study in a middle-income country (Vietnam) concluded that a hospital-wide hand hygiene promotion was cost-saving [16].
Direct comparison of these findings with ours is difficult for three reasons. First, only one of the previous studies quantified benefits in terms of final health outcomes (QALYs) [12]. Second, bed-day costs are much greater in high-income settings and account for most of the costs associated with HAIs. In developing country settings, costs of antibiotics to treat infections are likely to be the dominant cost [16]. Third, there are important differences in aims and methodology. We focused on MRSA-BSI (where we have strong evidence that it can be reduced by hand hygiene), reasoning that if the intervention is cost-effective for this outcome alone then it should certainly be cost-effective overall. We also made use of important methodological advances, accounting for the expected nonlinear association between hand hygiene compliance and infection rates using a mathematical model, avoiding timedependent biases when estimating increased length of stay, valuing bed-days based on opportunity cost rather than using an accounting approach, and estimating life-years gained by preventing mortality using data from a large linked-database study [10,21,24,43,32]. These advances, combined with the much lower bed-day costs, help explain why we estimated the cost per infection to be a few hundred dollars, whereas studies in other countries in both high-and middle-income countries estimated it to be a few thousand [13e16].
Our study has some limitations, the most important of which is that data are not yet available that allow us to include other pathogens in the model. In addition, since local information on the resources used for the hand hygiene intervention is limited, we were dependent on estimates of resource used for hand hygiene interventions from studies outside Thailand. A further limitation is that we evaluated the intervention over only a one-year post-intervention period where the initial investment would be most. Additional costs to maintain the compliance in the later years are difficult to quantify but likely to be less costly than the first year.
In conclusion, effective multi-modal hand hygiene interventions are likely to be cost-effective in ICU settings in   typical middle-income countries due to reduced incidence of MRSA-BSI alone under a wide range of circumstances. When this is not so, the cost-effectiveness of interventions to further improve hand hygiene will depend on the impact on other infections and other pathogens. Further work is needed to quantify this.