Hand, foot and mouth disease (HFMD) affects millions of children across Asia annually, leading to an increase in implemented control policies such as surveillance, isolation and social distancing in affected jurisdictions. However, limited knowledge of disease burden and severity causes difficulty in policy optimisation as the associated economic cost cannot be easily estimated. We use a data synthesis approach to provide a comprehensive picture of HFMD disease burden, estimating infection risk, symptomatic rates, the risk of complications and death, and overall disability-adjusted life-year (DALY) losses, along with associated uncertainties.

Complementary data from a variety of sources were synthesised with mathematical models to obtain estimates of severity of HFMD. This includes serological and other data extracted through a systematic review of HFMD epidemiology previously published by the authors, and laboratory investigations and sentinel reports from Singapore’s surveillance system.

HFMD is estimated to cause 96 900 (95% CI 40 600 to 259 000) age-weighted DALYs per annum in eight high-burden countries in East and Southeast Asia, with the majority of DALYs attributed to years of life lost. The symptomatic case hospitalisation rate of HFMD is 6% (2.8%–14.9%), of which 18.7% (6.7%–31.5%) are expected to develop complications. 5% (2.9%–7.4%) of such cases are fatal, bringing the overall case fatality ratio to be 52.3 (24.4–92.7) per 100 000 symptomatic infections. In contrast, the EV-A71 case fatality ratio is estimated to be at least 229.7 (75.4–672.1) per 100 000 symptomatic cases. Asymptomatic rate for EV-A71 is 71.4% (68.3%–74.3%) for ages 1–4, the years of greatest incidence.

Despite the high incidence rate of HFMD, total DALY due to HFMD is limited in comparison to other endemic diseases in the region, such as dengue and upper respiratory tract infection. With the majority of DALY caused by years of life lost, it is possible to mitigate most with increased EV-A71 vaccine coverage.

Hand, foot and mouth disease (HFMD) is a high-incidence paediatric disease with notified cases exceeding a million each year. Although the number of mild cases vastly outnumbers severe cases, the potential for complications and death has led to an inflated perception of disease severity, which incited a stronger response than necessary in multiple jurisdictions in Asia.

Although the symptomatic case fatality rate (sCFR) of HFMD has been computed in the literature, it is a poor reflection of the general severity and disease burden because it is sensitive to the notification rates of the surveillance system.

Using a combination of Singapore’s surveillance data along with serological and epidemiological data found in the literature, we constructed a series of models and indices that describe the entire spectrum of HFMD severity. These findings were translated into an age-weighted disability-adjusted life-years (DALY) for comparison to other diseases.

Our results show that after correcting for non-notification sCFR is likely to be inflated many times. Also, disability plays a minor part in overall DALY and the majority of DALYs could be lowered if fatality rate is lowered, possibly with the EV-A71 vaccine. High asymptomatic rates have profound implications to the modelling studies previously conducted to estimate the reproduction number of EV-A71, which have been generally heterogeneous.

Policies directed at HFMD should be re-evaluated based on this updated estimate of disease severity and burden.

Hand, foot and mouth disease (HFMD) is a paediatric disease which, although usually self-limiting, can result in complications associated with the central nervous system (CNS) or death.

The main aetiological agents of HFMD are the human enterovirus A71 (EV-A71) and coxsackieviruses 10 and 16, with EV-A71 being associated with more severe outcomes. As with other usually mild diseases where infections vastly exceed severe cases, efforts to track all HFMD infections to determine indices for different levels of severity are impractical due to the large number of cases involved.

To quantify the extent of HFMD across Asia, its mortality risk and the economic burden of control measures against it,

The severity of infection was classified into tiers which require different levels of care: (i) clinically apparent, or symptomatic, cases; (ii) hospitalisations; (iii) complications, or severe cases when the child displays neurological or cardiopulmonary complications; and (iv) fatal cases. These manifestations were matched against similar definitions from the Global Burden of Disease 2013 (GBD2013) study

These severity tiers were arranged as a pyramid where lower echelons represent the majority of (milder) infections with decreasing frequency but increasing severity towards the apex

The severity pyramid structure for hand, foot and mouth disease (HFMD) (total, all aetiologies) and EV-A71 in particular. Bordered boxes are data sources, while boxes within the pyramids are estimated quantities. Pale yellow boxes represent the models and outcomes of submodel 1, the infection rate model. This model uses serological data and differential equations to estimate the rate at which infection must occur to match with observed seropositivity level of EV-A71. Blue boxes represent the data, model and outcomes of submodel 2, the symptomatic incidence rate model. This model uses surveillance data from Singapore to infer the incidence rate of HFMD and EV-A71-specific HFMD. Together with submodel 1, the estimated symptomatic rate of EVA71 (cyan box) can be calculated. The orange boxes represent the data and models of submodel 3: the hierarchical model, which is used to estimate indices that quantifies the higher-level severities of HFMD. CNS, central nervous system; MOH, Ministry of Health.

Quantifying the disease burden of HFMD requires an estimate of the total number of cases, which was calculated from the infection rate (both symptomatic and asymptomatic) obtained from seroepidemiological studies. From a systematic review previously published by the authors,

To estimate clinically apparent cases, we used data on notified cases from Singapore, where HFMD is a legally notifiable disease and where active surveillance in preschools is conducted via daily temperature taking; this was used to calculate a lower bound for the overall asymptomatic rate of HFMD, in submodel 2. Indices for higher severity levels of HFMD and EV-A71, which include deaths, complications, hospitalisations and symptomatic cases, used information from 27 papers

The infection rate model estimates the age-specific proportion of children who must be infected by the EV-A71 virus each year to reach the observed level of seroprevalence. Serological data from China,

Data and results from the infection rate model (submodel 1). Left: seropositivity levels of EV-A71 with fitted curve. Red data points are from China, blue from Taiwan and orange from Singapore, green from Vietnam and brown from Thailand. The black line is the fitted model with 95% Bayesian credible intervals. Middle: required and observed rate of EV-A71 infection. The brown line shows the age-specific infection rate of EV-A71 implied from the seropositivity curve and is the required rate to get the observed seropositivity levels; the blue line shows the calculated average EV-A71 infection rate based on Singapore’s data from submodel 2. The difference between the two lines are the unobserved cases. Right: asymptomatic infection rate. Calculated from the percentage difference between the required and observed rates of infection in the middle.

Posterior distributions of parameters were estimated in R

Since October 2000, Singapore has adopted a comprehensive approach for HFMD surveillance including a mandatory case reporting system by medical practitioners and childcare centres,

We constructed a matrix containing the incidence

The EV-A71-specific incidence is a proportion of all HFMD cases. We obtain

Age and year trend of hand, foot and mouth disease (HFMD) symptomatic infection from the Incidence rate model (submodel 2). Left: the age effect. The black line reflects the overall proportion of incidence for each year of age, and red line represents the age effect for individual years from 2005 to 2012. Middle: the year effect. The black line is an index which reflects the number of notified cases for 2005–2012. The year effect for 2013–2015 is estimated from data (notified cases) obtained from the Weekly Infectious Disease Bulletin released by the Ministry of Health, Singapore, using linear regression. The year effect of EV-A71 (blue) is mathematically derived from the proportion of cases due to EV-A71 estimated by laboratory data (right). Right: laboratory data from Singapore’s surveillance system. This shows the proportion of cases due to each virus in each year from 2005 to 2015.

The upper levels of the severity pyramid from across Asia were modelled as nested binomial distributions, with the probability of reaching state

To estimate the DALY lost across Asia to HFMD, we conservatively estimated the incidence of HFMD infection to be twice the incidence of EV-A71 using the incidence rate model (submodel 2), to allow for infection by other aetiological agents of HFMD which tend to be more common than EV-A71. The number of symptomatic cases was then calculated using the asymptomatic rate from the infection rate model (submodel 1). The numbers of more severe manifestations (hospitalised, CNS and deaths) were then estimated using the parameters from the hierarchical model (submodel 3). Again, posterior distributions were used to characterise uncertainty in these estimates.

Disability weights from GBD2013

We used the age-weighted approach

Under this framework, the formula for DALY for an individual is

where

Among the seven contesting models, model 7 with an unconstrained initial rate of infection and a decreasing rate of infection that reaches 0 at time

where

The fitted values are

The year effect,

The age effect

For Singapore, the required rate for new infections is much higher than the actual observed level, and the differences are due to asymptomatic infection. For the most vulnerable age group, around 70% of EV-A71 cases do not show sufficient symptoms to be identified either at presentation to their preschool or by their doctor (

For HFMD in general (ie, without reference to the specific viral agent), the symptomatic case hospitalisation ratio is estimated to be 6% (2.8%–14.9%). Among hospitalised cases, 18.7% (6.7%–31.5%) develop CNS complications, of which 5.0% (2.9%–7.4%) were fatal. Overall, the CFR for HFMD is estimated to be 52.3 (24.4–92.7) per 100 000 symptomatic infections. Among the papers which identified EV-A71 as the pathogen, the majority recruited children seeking treatment in hospitals. Of these, 36.9% (25.9%–48%) were associated with CNS complications, of which 10.5% (4.9%–17.8%) died. In contrast, there were no discernible differences in the fatality rate of HFMD patients with CNS complications between cases with EV-A71 and all HFMD infections, possibly because most severe cases were indeed caused by EV-A71. Estimates within studies and overall are presented in

Data and results from the hierarchical model (submodel 3). Top: hand, foot and mouth disease (HFMD) severity for total HFMD. Bottom: HFMD severity for lab-tested EV-A71 cases. These were modelled in two separate hierarchical pyramid models. The median and 95% credible intervals for each individual paper are shown by the black dot and lines, respectively. All estimates at the individual level are based from a hyperparameter or a higher-level parameter. The vertices of the diamond show the median (also the red line) and 95% Bayesian credible Intervals. These represent the overall estimate for each severity index, where the estimates are linked together with a multiplicative relationship: P(D|C) = P(D|S) × P(S|H) × P(H|C). All parameters were estimated using a Bayesian hierarchical approach using Markov chain Monte Carlo. Case hospitalisation ratio estimate for EV-A71 cannot be calculated as studies on EV-A71 cases are almost performed exclusively from hospital data, and thus the lowest level is ‘hospitalised’ and not ‘cases’.

Overall disease severity of hand, foot and mouth disease (HFMD)

Event | HFMD | Combined | EV-A71 | |||

Rate | 95% CI | Rate | 95% CI | Rate | 95% CI | |

Mortality (per 1000 CNS) | ||||||

Pr(Death | CNS) | 50.2 | (29.2 to 73.5) | – | – | 104.7 | (49.0 to 177.8) |

CNS (per 1000 hospitalisations) | ||||||

Pr(CNS | Hospitalisation) | 186.7 | (66.6 to 314.9) | – | – | 368.8 | (258·. to 480.0) |

Hospitalisation (per 1000 cases) | ||||||

Pr(Hospitalisation | Case) | – | – | 60.2 | (28.2 to 148.6) | – | – |

Mortality (per 1000 hospitalisations) | Product of mortality per CNS and CNS per hospitalisation | |||||

8.8 | (2.9 to 17.7) | 38.3 | (16.6 to 69.5) | |||

Mortality (per 100 000 cases) | Product of mortality per hospitalisation and hospitalisation per case | |||||

52.3 | (24.4 to 92.7) | 230.1 | (74.9 to 670.1) | |||

Symptoms (per 1000 infections) | ||||||

Pr(Case | Infection) | ||||||

Age 1 | – | – | 220.6 | (195.8 to 247.7) | – | – |

Age 2 | – | – | 334.6 | (299.7 to 369.7) | – | – |

Age 3 | – | – | 325.4 | (293.8 to 357.6) | – | – |

Age 4 | – | – | 263.1 | (236 to 291.5) | – | – |

Age 5 | – | – | 214.0 | (189.8 to 240.6) | – | – |

Age 6 | – | – | 166.6 | (146.6 to 191) | – | – |

Age 7 | – | – | 133.0 | (115.3 to 154.8) | – | – |

Age 8 | – | – | 111.2 | (95.1 to 131.9) | – | |

Age 9 | – | – | 100.3 | (84.8 to 121.2) | – | – |

Infection (per 1000 person-years) | ||||||

Pr(Infection) | ||||||

Age 1 | 44.9 | (36.4 to 53.4) | – | – | 22.4 | (18.2 to 26.7) |

Age 2 | 58.4 | (47.4 to 69·5) | – | – | 29.2 | (23.7 to 34.7) |

Age 3 | 49.0 | (39.8 to 58.3) | – | – | 24.5 | (19.9 to 29.1) |

Age 4 | 34.3 | (27.9 to 40.8) | – | – | 17.2 | (13.9 to 20.4) |

Age 5 | 24.3 | (19.8 to 28.9) | – | – | 12.2 | (9.9 to 14.5) |

Age 6 | 16.5 | (13.4 to 19.7) | – | – | 8.3 | (6.7 to 9.8) |

Age 7 | 11.6 | (9.4 to 13.8) | – | – | 5.8 | (4.7 to 6.9) |

Age 8 | 8.5 | (6.9 to 10.2) | – | – | 4.3 | (3.5 to 5.1) |

Age 9 | 6.8 | (5.5 to 8.1) | – | – | 3.4 | (2.8 to 4.1) |

CNS, central nervous system.

The disease burden estimated using our approach is 96 900 (40 600–259 000) DALYs per annum in eight high-endemic countries in East and Southeast Asia (estimates by country are provided in

Annual disability-adjusted life—year (DALY) losses in eight Asian countries/regions with 95% credible intervals (CI)

Country or region | DALY | 95% CI |

People’s Republic of China (excluding Hong Kong and Taiwan) | 75 881 | (31 835 to 202 591) |

Hong Kong special administrative region, People’s Republic of China | 285 | (115 to 767) |

Japan | 5456 | (2290 to 14 589) |

Malaysia | 2723 | (1138 to 7281) |

Singapore | 259 | (104 to 748) |

Taiwan, Republic of China | 1084 | (435 to 3052) |

Thailand | 3928 | (1644 to 10 536) |

Vietnam | 7248 | (3042 to 19 414) |

In this study, we used a statistical modelling approach to synthesise information on HFMD from countries in East and Southeast Asia where the disease is highly prevalent. To estimate a conservative lower bound for the asymptomatic rate of EV-A71, we used case notification data from Singapore. The small, clearly demarcated jurisdiction with policies that lead to comprehensive case finding of symptomatic HFMD cases—notification being currently mandatory for physicians and childcare centres—may permit a higher fraction of symptomatic cases to be notified than other settings. Supporting evidence comes from the higher reported incidence rate of HFMD in Singapore (around 6%–14% at age 1) than in other areas (eg, about 3–5 times higher than China where the incidence is around 3% at age 1,

Such a high asymptomatic rate for HFMD has implications on estimates of R_{0}, the basic reproduction number. A large proportion of unobserved infections imparts a downward bias on estimated infection rate parameters in compartmental models (such as susceptible-infected-removed variants) that assume all cases are symptomatic, and accounting for the high asymptomatic rate could be the key to reconciling the highly heterogeneous R_{0} estimates in the literature, which vary from almost unity to over 20.

When performing data synthesis, there is a particular challenge in the allocation of collected information into the severity tiers defined in this study. For instance, we were concerned that the identified EV-A71 cases in the literature were subject to selection bias as it is more likely for physicians to take samples from sicker children. We therefore excluded studies that reported clinically diagnosed cases of EV-A71 because of this potential for bias, and as a result, the lowest level on the severity pyramid we could estimate for EV-A71 was ‘hospitalisations’ rather than ‘symptomatic cases’. To obtain the CFR for EV-A71, therefore, required assuming the case hospitalisation rate is the same for EV-A71 and other aetiologies (ie, 6%, 95% CI 2.8% to 14.9%,

Despite the high incidence rate of HFMD, the relative scarcity of severe symptoms and the generally short disability durations led to estimated annual DALY losses in eight high-incidence Asian countries of only 96 900 (40 500–258 400), considerably below the disability caused by upper respiratory tract infections (603 000) but comparable to dengue (100 000 in the same countries) which has low incidence in China.

The challenge in calculating the disease burden of HFMD lies in integrating and synthesising available data from decades of research into useful quantities for public health actions. In doing so, many assumptions were made. In particular, we integrated data from serological and epidemiological studies from a mixture of country-level and city-level studies across Asia to develop an overall ‘Asian’ average which summarises the severity across affected parts of the continent, which differ in economic development and healthcare systems. These differences cannot be easily adjusted using statistical methods due to low amounts of information from each individual country. Differences between and within countries, however, mean that these overall estimates may not apply to local decision-making. Second, the age-specific asymptomatic rate estimates were based on surveillance data from Singapore, which we believe has a notification system for symptomatic infections presenting at primary care or in education that is close to complete, but with room for misdiagnosis. Third, societal impacts—such as the indirect costs that result from caregivers taking time off work or other impacts of control policies

Seye Abimbola

WMK conceptualised the study, developed the analytical strategy, collected data, performed the statistical analysis, interpreted the results and wrote the first draft of the report. HB contributed to obtaining of data and interpretation of the results. HL and MI-CC contributed to the analysis and interpretation of results. ARC contributed to the conceptualisation of the study, analytical strategy, interpretation of results, performed critical revisions of the report and secured funding. All authors contributed to revising the paper and approved the final version.

The work was funded by Singapore’s Ministry of Health, Health Services Research grant number HSRG12MAY023, Communicable Disease Public Health Research Grant number CDPHRG12NOV021, the Singapore Population Health Improvement Centre, the Centre for Infectious Disease Epidemiology and Research and the President’s Graduate Fellowship to WMK.

None declared.

Not commissioned; externally peer reviewed.

Requests for data access should be directed to the corresponding author.