^{1}

^{*}

^{2}

^{1}

^{3}

^{2}

^{4}

^{4}

^{5}

^{4}

The authors have declared that no competing interests exist.

Analyzing data from Thailand's 72 provinces, Derek Cummings and colleagues find that decreases in birth and death rates can explain the shift in age distribution of dengue hemorrhagic fever.

An increase in the average age of dengue hemorrhagic fever (DHF) cases has been reported in Thailand. The cause of this increase is not known. Possible explanations include a reduction in transmission due to declining mosquito populations, declining contact between human and mosquito, and changes in reporting. We propose that a demographic shift toward lower birth and death rates has reduced dengue transmission and lengthened the interval between large epidemics.

Using data from each of the 72 provinces of Thailand, we looked for associations between force of infection (a measure of hazard, defined as the rate per capita at which susceptible individuals become infected) and demographic and climactic variables. We estimated the force of infection from the age distribution of cases from 1985 to 2005. We find that the force of infection has declined by 2% each year since a peak in the late 1970s and early 1980s. Contrary to recent findings suggesting that the incidence of DHF has increased in Thailand, we find a small but statistically significant decline in DHF incidence since 1985 in a majority of provinces. The strongest predictor of the change in force of infection and the mean force of infection is the median age of the population. Using mathematical simulations of dengue transmission we show that a reduced birth rate and a shift in the population's age structure can explain the shift in the age distribution of cases, reduction of the force of infection, and increase in the periodicity of multiannual oscillations of DHF incidence in the absence of other changes.

Lower birth and death rates decrease the flow of susceptible individuals into the population and increase the longevity of immune individuals. The increase in the proportion of the population that is immune increases the likelihood that an infectious mosquito will feed on an immune individual, reducing the force of infection. Though the force of infection has decreased by half, we find that the critical vaccination fraction has not changed significantly, declining from an average of 85% to 80%. Clinical guidelines should consider the impact of continued increases in the age of dengue cases in Thailand. Countries in the region lagging behind Thailand in the demographic transition may experience the same increase as their population ages. The impact of demographic changes on the force of infection has been hypothesized for other diseases, but, to our knowledge, this is the first observation of this phenomenon.

Every year, dengue infects 50–100 million people living in tropical and subtropical areas. The four closely related viruses that cause dengue are transmitted to people through the bites of female

Historically, dengue has mainly affected young children but, recently, its age distribution has shifted towards older age groups in several Southeast Asian countries, including Thailand. In addition, the interval between large increases in incidence (epidemics) of dengue hemorrhagic fever has lengthened. It is important to know why these changes are happening because they could affect how dengue infections are dealt with in these countries. One idea is that an ongoing shift towards lower birth and death rates (the demographic transition; this occurs as countries move from a pre-industrial to an industrial economy) is reducing dengue transmission rates by reducing the “force of infection” (the rate at which susceptible individuals become infected). As birth and death rates decline, immune individuals account for more of the population so mosquitoes are more likely to bite an immune individual, which reduces the force of infection. Similarly, because susceptible individuals enter the population by being born, changing the birth rate alters the interval between epidemics. In this study, the researchers test whether the demographic transition might be responsible for the changing pattern of dengue infection in Thailand.

The researchers retrieved data on dengue infection, demographic data (the population's age structure and birth and death rates), socioeconomic data, and climatic data for Thailand from 1980 to 2005 from various sources. They then fitted the data on dengue cases to several mathematical models to estimate the force of infection for each year. This analysis suggested that the force of infection has declined by 2% every year since the early1980s. Next, the researchers used statistical methods to show that the strongest predictor of this decline is the increase in the median age of the population (a measure of the average age of the population). Finally, using mathematical simulations of dengue transmission, they showed that a reduced birth rate and a shift in the population's age structure are sufficient to explain the recent shift in the age distribution of dengue cases, the reduction of the force of infection, and the increased interval between epidemics of dengue hemorrhagic fever.

The findings of all modeling studies depend on how the mathematical models are built and the accuracy of the data fed into them. Nevertheless, these findings suggest that recent changes in birth and death rates in Thailand are sufficient to produce the observed changes in the age distribution of dengue and periodicity of dengue outbreaks. One implication of these findings is that other countries in Southeast Asia that follow Thailand in the demographic transition may experience similar shifts in the pattern of dengue infections as the age structure of their populations changes. This means that clinical guidelines for the management of dengue infections in Southeast Asia will need to be adjusted to allow for the increasing age of dengue cases. Finally, although the researchers' calculations show the force of infection has fallen substantially over the past two decades, they also show that when a dengue vaccine becomes available, it will still be necessary to vaccinate most of the population to halt dengue transmission.

Please access these Web sites via the online version of this summary at

This study is further discussed in a

The US Centers for Disease Control and Prevention provides detailed information about

The World Health Organization provides information about

Links to additional information about dengue are provided by

Wikipedia has a page about the

In Thailand, dengue fever (DF) and dengue hemorrhagic fever (DHF) have traditionally been diseases of childhood, causing substantial morbidity and mortality among children <15 y of age. Recently, an increasing number of DHF cases among older individuals has been observed in Thailand

There are several potential reasons for this shift. One hypothesis is that the force of infection (the per capita rate at which susceptible individuals become infected) is declining because of reductions in vector abundance or contact between human and vector, possibly as a result of control efforts or economic development

Another hypothesis is that the age distribution of dengue cases has not changed at all, but reporting has shifted over time. The national reporting system for dengue receives reports of DHF, dengue shock syndrome (DSS) and, since 1994, DF. There has been a shift in the Thai surveillance system to report a larger number of cases of DF. Differences in the age distribution of this newly reported class of clinical cases could increase the apparent average age of dengue cases.

Here, we hypothesize that the shift in the age distribution of dengue cases is due to a shift in the age structure of the population, and its impact on the force of infection. Over the last 20 y the age structure of Thailand has changed dramatically. In 1980, 40% of the population was under 15 y; in 2000, 24% were under 15 y (

(A) Age structure of the Thai population 1980 (black), 1990 (red), and 2000 (blue). (B) Birth rates per 1,000 individuals 1970–2000. (C) Mean, median, and modal age of age-standardized dengue incidence data in Thailand 1981–2005. Age-specific incidence in all years was standardized to age distribution of Thailand's population in 2000 using direct age adjustment. Mean (filled circles), mode (unfilled circles), and median (plus sign) age of standardized age-specific incidence of dengue disease reported to the national surveillance system for each year are shown. (D) Mean age of DHF (red), DSS (green), DF cases (blue), and total (black) cases reported to the national surveillance system 1999–2005.

The age distribution of reported dengue cases provides information on the level of transmission occurring in the population. The age at which individuals become infected can be used to estimate the hazard of infection and the age-specific risk of infection

In this paper, we investigate changes in the age distribution of dengue disease reported to the Thai national surveillance system from 1985 to 2005. We test whether the change in the age distribution of dengue cases is attributable to changes in the distribution of dengue disease among the three clinical presentations (DHF, DF, DSS) of dengue infection or due to a change in the age distribution of just one of these clinical presentations.

We estimate the force of infection of dengue in each of the provinces of Thailand and look for associations between the spatial and temporal distribution of the force of infection and socio-demographic characteristics including age structure, birth rate, climate factors, and economic indices. An association between the force of infection and demographic structure supports our hypothesis that the demographic transition is the cause of the observed increase in the average age of dengue cases.

Changes in birth rate have been shown to be an important driver of changes in the dynamics of childhood infectious diseases

A reporting system for communicable disease that includes dengue was established in Thailand by the Ministry of Public Health in the early 1970s

Population data for each province of Thailand were obtained from censuses in 1980, 1990, and 2000 performed by the National Statistical Office of Thailand

We characterize the level of transmission in each province using the force of infection at time _{i}_{i}_{i}_{i}

Using age-stratified case reports, as opposed to serological data, to estimate

We fit the age distribution of dengue cases in each province in each year using multiple models. We assumed that the cumulative proportion of age-specific incidence per individual occurring up until each age provides an estimate of the proportion of individuals who have experienced two infections as a function of age. We then used catalytic models that have been applied to age-specific serological data to estimate the force of infection in each year for which we have age-specific dengue incidence (1985–2005), as well as the 20 y preceding

We used inverse variance–weighted linear regression to estimate the association of the mean force of infection in each province (1985–2005) and socioeconomic, demographic, and climatic factors. Similar methods were used to associate the change in force of infection in each province from 1985 to 2005 with changes in socioeconomic, demographic, and climatic factors.

We used wavelet analysis to estimate the multiannual periodicity of incidence in each of the provinces of Thailand. Wavelet transforms give an estimate of the period of oscillations at each point in a time series

We developed deterministic models of the transmission dynamics of four dengue serotypes to examine the impact of reducing birth rates on the period of oscillations in incidence. We considered three models: one with immune enhancement of transmission

We developed an age-specific model of dengue transmission to explore the impact of secular demographic trends (i.e., changing birth and death rates) on transmission dynamics and the age distribution of cases

Standardized age-specific incidence rates (see

(A) Estimates of the force of infection (per year) for each year, 1980–2005 in Thailand estimated using age-specific case data aggregated at the national level and model 3 (see

The best fitting model of the four estimated was model 4, which includes multiplicative age factors that modify the force of infection. The estimates of the force of infection generated by this model appear in

Age-specific incidence is shown by circles and fits from model shown by lines. Model 2 includes time-specific but not age-specific forces of infection. Model 3 includes additive age-specific forces of infection in addition to time-specific forces. Model 4 includes multiplicative age-specific forces of infection in addition to time-specific forces.

In Rayong province both age-specific DHF incidence and age-stratified serological data for a single year (1980) are available

Time-specific estimates of the force of infection exhibit multiannual oscillations as characterized by Fourier spectra. 48 of 72 provinces had significant peaks in Fourier spectra of time series of estimates of the force of infection generated by model 4 between a period of 2.5 and 6 y. The most common peak in Fourier spectra in this range was 2.7 y. This estimate is consistent with mean multiannual periodicities of DHF incidence data in Thailand

Including increased susceptibility of previously exposed individuals significantly improved the model fit to country-wide data (chi-square test of difference in likelihoods, each

Our central hypothesis is that increases in the median age of the total population lead to decreases in the force of infection. Province-specific mean forces of infection from 1985 to 2005 generated using model 4 were significantly associated with the provincial median age, with an estimated decrease of 1.5×10^{−3} (95% CI 8.7×10^{−4}, 2.2×10^{−3}) per year increase in the median age of provincial population. Mean provincial force of infection was significantly associated with mean household size, birth rate, proportion of the population under 15 y of age, percent of homes with sanitation, latitude, and average monthly rainfall (

Independent Variable | Regression coefficients (95% CI) |

Median age | −1.5e−3 (−2.2e−3 to −8.7e−4) |

Mean household size | 9.7e−3 (4.8e−3 to 1.5e−2) |

Birth rate (per 1,000) | 4.2e−4 (8.3e−6 to 8.3e−4) |

Proportion of population under 15 y of age | 6.8e−2 (3.7e−2 to 1.0e−1) |

Percent of homes with sanitation | −2.1e−4 (-3.3e−4 to −8.3e−5) |

Average rainfall (mm/mo) | 4.2e−5 (3.1e−6 to 8.1e−5) |

Latitude (degrees) | −1.3e−3 (-1.6e−3 to −8.1e−4) |

School attendance, percent of homes constructed with permanent materials, percent of provincial population in urban areas, gross provincial product (Baht per capita), and average temperature were not statistically significantly associated with the mean force of infection. Covariates significant in multiple linear regression: median age −3.2e−3 (95% CI −5.9e−3 to −5.1e−4), latitude −1.1e−8 (95% CI −2.0e−8 to −1.4e−9), and birth rate −1.4e−3 (95% CI −2.3e−3 to −4.1e−4).

The change in force of infection was significantly associated with the changes in two covariates, median age and the percentage of homes constructed with permanent materials (see

Independent Variable | Regression Coefficients (95% CI) |

Change in median age | −4.6e−2 (−7.7e−2, −1.5e−2) |

Change in percentage of homes constructed with permanent materials | −1.5e−2 (−2.6e−2, −3.6e−3) |

Change in school attendance, household size, percent of provincial population in urban areas, birth rate, gross provincial product, average rainfall, and average temperature were not statistically significantly associated with the change in the mean force of infection. Both median age and change in percentage of homes constructed with permanent materials significant in model that includes both.

Only median age was found to be a statistically significant predictor of both the mean force of infection and the change in the force of infection in each province between 1985 and 2005.

Between 1985 and 2005, there was a small but statistically significant decline in the national incidence of DHF (

(A) Incidence of dengue disease in Thailand, 1981–2005. The solid line shows the incidence per 1,000 individuals per year of DHF and DSS together, whereas the grey line shows the incidence of DHF, DSS, and DF. DF was included in case reports starting in 1993. (B) Change in incidence (per 1,000 per year) 1985–2005 with 95% CI versus change in force of infection for each of the provinces of Thailand. A linear association between these two variables is not statistically significant.

Each line is the period in months of the incidence in one province. Period presented is the mean period of power in a period band of 18 to 60 mo reconstructed using the continuous wavelet transform (see

Independent Variable | Regression Coefficients (95% CI) |

Change in median age | 0.09 (0.02–0.15) |

Change in school attendance | 0.02 (0.002–0.03) |

Change in mean household size, percent of provincial population in urban areas, birth rate, percentage of homes constructed with permanent materials, gross provincial product, average rainfall, and average temperature were not statistically significantly associated with the change in period of multiannual oscillations.

Using the forces of infection, we estimated the basic reproductive number (_{0}) of dengue in each province. _{0} characterizes the intrinsic transmissibility of the pathogen in a setting, as opposed to the force of infection, which depends upon the number of individuals currently infectious _{0} is to estimate the fraction of the population that must be vaccinated to eliminate transmission, the critical vaccination fraction, _{0}). We estimate the decline in the mean critical vaccination fraction in Thailand from 1980 to 2005 to be 5%, from 85% to 80% (i.e., _{0} has declined from 6.7 to 5.2). The critical vaccination fraction depends on the fraction susceptible at equilibrium. We estimate that the fraction of the population that is susceptible to any dengue serotype has changed little over the last 25 y, increasing from 14% to 19% of the population.

The period of multiannual oscillations in simulations of dengue transmission are dependent on assumed birth and death rates. _{0} of 4 to 12. At an _{0} of 6 (i.e., our estimate for Thailand) a change in the birth rate similar to that observed in Thailand increases the period of the system from just over 2 y to close to 4 y. This is consistent with the observed change in the period of multiannual oscillations in dengue and DHF from a mean of 33 mo to 49 mo (

(A) Simulations include immune enhancement of transmission (second infections are 1.85 times as transmissible as primary infections)

Simulations using a two-serotype model with ten age classes showed a steady increase in the average age of primary and secondary dengue infections when birth and death rates were reduced (

Average age of (A) primary dengue and (B) secondary dengue cases in a two-serotype transmission model. In both plots, age is indicated by the color blue with legend at right. The birth and death rate varies from 10 per 1,000 to 40 per 1,000 and the transmission coefficient varies from 200 to 500 (_{0}≈4–10). The average ages of both primary and secondary cases rise with decreasing birth/death rate and decreasing transmission coefficient. The average age of secondary cases rises faster with each unit decrease in birth/death rates (range of average ages in (B), 4–16 y, range of average ages in (A) 2–5 y.

Analysis of case data showed only a slight decrease in absolute incidence. Both the age-specific and non-age-specific simulations show large decreases in incidence as birth and death rates are reduced. However, these results are dependent upon including age-independent mortality, which is commonly used in transmission models. If instead, age-dependent mortality that is more consistent with observed age-specific rates of mortality (in the form of higher mortality among older individuals) is used, the reductions in incidence are smaller.

Recent changes in the average age of DHF in Southeast Asia have been attributed to control measures

If the demographic transition is responsible for the reduction in the force of infection of dengue, a similar reduction might be seen in other diseases. The average age of chickenpox, a directly transmitted respiratory infection, has increased in Thailand over the same period, from 10.2 y in 1985 to 12.5 y in 2005 (estimated using data from the national surveillance system). The impact of demographic changes may be more easily observed in dengue than chickenpox because severe manifestations of dengue are associated with secondary exposure. The average age of secondary exposure would increase more quickly than the average age of a primary exposure with reductions in hazard because the reduced hazard delays the primary and secondary infection. Chickenpox transmission is also mediated by social interactions more strongly than dengue, as there is no vector. Thus, interactions between children sufficient to transmit chickenpox may be less affected by increases in the overall prevalence of immunity if the majority of those immune are older. The transmission models that we have used assume that contact processes are frequency dependent rather than density dependent. Density-dependent models do not show the same reductions in the force of infection because the density-dependent contact assumes that the number of immune individuals has no impact on the contact between susceptible and infectious individuals

We find no evidence for the increases in annual DHF incidence between 1985 and 2000 reported in

The shift in the average age of infection might have public health impacts not reflected in total incidence. Total deaths due to DHF have declined steadily over the last 20 y in Thailand

There are several limitations to our study. Data from the passive surveillance describe predominantly severe, hospitalized cases. We do not have full knowledge of the biases in reporting that may be present. Ideally, our study would be based on incidence of infection as evidence by serology rather than incidence of disease. Age-stratified serological surveys performed in several locations in Thailand could help determine if the conclusions presented here are the same when considering infections rather than cases of severe disease. Another limitation of our study is that we do not have good information on reporting practices over time. To the extent that the data allow, we have found no secular trends in the types of facilities reporting or the ratio of urban to rural facilities reporting. However, we have limited ability to detect changes in reporting that could affect our results. In addition, we do not know the extent to which primary or tertiary infections may be present among the observed cases. Though they are thought to be rare among symptomatic cases, their presence in substantial numbers would bias our results.

We have not addressed a number of alternative hypotheses that could explain the observed increase in age of dengue cases. Control measures could be reducing vector densities or contact between vectors and humans. Migration may be moving adults from areas of low transmission to areas of greater transmission, thus increasing the average age of cases. Though we have identified a mechanism that can of itself explain the increase in average age and the lengthening of the interepidemic interval, we do not know what role these other factors might play in increasing the average age of dengue cases.

Several dengue vaccine candidates are under development. We estimate that both the proportion of the population susceptible and the critical vaccination fraction have changed very little (∼5%) over the last 20 y, although the force of infection has fallen substantially.

If the demographic transition is responsible for the reduction in the force of infection as proposed, the declines will not be limited to Thailand. In fact, an increase in the average age of dengue has been reported in several other countries in Southeast Asia that are experiencing reductions in birth and death rates

Thai translation of the abstract by SI.

(0.08 MB PDF)

Detailed methods and supplemental results.

(1.15 MB DOC)

confidence interval

dengue fever

dengue hemorrhagic fever

dengue shock syndrome