Burden of micronutrient deficiencies by socio-economic strata in children aged 6 months to 5 years in the Philippines

Background Micronutrient deficiencies (MNDs) are a chronic lack of vitamins and minerals and constitute a huge public health problem. MNDs have severe health consequences and are particularly harmful during early childhood due to their impact on the physical and cognitive development. We estimate the costs of illness due to iron deficiency (IDA), vitamin A deficiency (VAD) and zinc deficiency (ZnD) in 2 age groups (6–23 and 24–59 months) of Filipino children by socio-economic strata in 2008. Methods We build a health economic model simulating the consequences of MNDs in childhood over the entire lifetime. The model is based on a health survey and a nutrition survey carried out in 2008. The sample populations are first structured into 10 socio-economic strata (SES) and 2 age groups. Health consequences of MNDs are modelled based on information extracted from literature. Direct medical costs, production losses and intangible costs are computed and long term costs are discounted to present value. Results Total lifetime costs of IDA, VAD and ZnD amounted to direct medical costs of 30 million dollars, production losses of 618 million dollars and intangible costs of 122,138 disability adjusted life years (DALYs). These costs can be interpreted as the lifetime costs of a 1-year cohort affected by MNDs between the age of 6–59 months. Direct medical costs are dominated by costs due to ZnD (89% of total), production losses by losses in future lifetime (90% of total) and intangible costs by premature death (47% of total DALY losses) and losses in future lifetime (43%). Costs of MNDs differ considerably between SES as costs in the poorest third of the households are 5 times higher than in the wealthiest third. Conclusions MNDs lead to substantial costs in 6-59-month-old children in the Philippines. Costs are highly concentrated in the lower SES and in children 6–23 months old. These results may have important implications for the design, evaluation and choice of the most effective and cost-effective policies aimed at the reduction of MNDs.

With p = prevalence of deciency z in socioeconomic group i and age group j, i ∈ {SES1, SES2, ... , SES10} and j ∈ {6m-11m, 12m-23m, 24-59m}. F N orm = cdf of normal distribution at the threshold for deciency th given the mean serum level µ Se ij and standard deviation of serum level σ Se ij . Prevalence for IDA is adjusted by multiplying p anemia ij with the share of anemia attributable to iron deciency. The number of decient children in each group is calculated by multiplying the p ij prevalence rate with the population size of each group (pop ij ).

Population Attributable Fractions
The share of illness that is caused by a deciency is determined by the population attributable fraction (PAF). The PAF depends on the prevalence of the deciency p z ij and the relative risk of the illness (r k ). PAFs were calculated according to equation 2.
This is Levin's classical formula for the attributable fraction [31].

Number of Deaths
The number of deaths is calculated as follows With mort being the mortality rate.
With n mod.ane. = Number of children with moderate anemia, n sev.ane. = number of children with severe anemia, dw phys.act. mod.ane. = DALY weight of reduced physical activity for the moderately anemic and dw phys.act. sev.ane. = DALY weight of reduced physical activity for the severely anemic. Impairement of physical activity is assumed to be temporary.

Mental Development
The calculation of rst year DALYs due to impaired mental development follows the same pattern as for physical activity (5).
With dw ment.dev. mod.ane. = DALY weight of reduced mental development for the moderately anemic and dw ment.def. sev.ane.
= DALY weight of reduced mental development for the severely anemic Future DALYs accrue until the end of the expected lifetime for a period of lif e exp.−age.

Stunting
From the literature we know the eect of ZnD on the height for age z-score (HAZ). The prevalence of stunting is calculated analogously to the prevalence of MNDs. The share of stunting (dended as HAZ < -2) attributable to ZND is calculated by estimating the mean HAZ score with and without deciency and comparing prevalence rates of stunting at the two values (6)- (9).
The dierent µ ij are calculated as follows (7) HAZ This equation can be solved for µ def ij and µ ndef ij With v stunt = prevalence of stunting, th stunt threshold of HAZ-score for stunting, µ def = mean value of HAZ score for the decient, µ ndef = mean value of HAZ score for the non-decient and e stunt znd = eect of ZnD on HAZ-score. DALYs are then calculated as follows (10).  (11).
With v k = prevalence of illness k, dur = duration of illness k. Number of DALYs is then calculated by dividing the number of days by 365 and multiplying by the DALY weight. From DHS data we know the number of illess episodes in the last 14 days. Thus any illness with onset in a period of 14 + dur k days is registered. By dividing with this term and multiplying by 365 the number of yearly episodes is calculated. In order for this calculation to be valid, we must assume that disease prevalence and duration is constant over any two consecutive weeks in the year. We also assume that any child is sick at most once in the observed two weeks period. DALYs are calculated as follows.
with dw k = disability weight of illness k.

Measles
For measles no group specic prevalence rates are available. These are estimated using the prevalence for the general population and the group specic mealses vaccination rates.
With v mea j = measles prevalence s j = share of population in age group j. Based on these age specic prevalence rates, the group specic prevalence is calculated by weighting the prevalence rate with the measles vaccination rate.
This simplies to Days of illness and DALYs are then calculated as follows With dw mea = disability weight for measles, ndays mea = number of days with measles, dur mea = average duration of a measles episode. Given that patients are immune to future measles infections after illness, we assume that the number of measles cases corresponds to the number of infections of new patients.

Current
Calculation of current yearly income losses is implemented as described in formula (18).
With wage i = Daily wage in dollars in SES i , dur k care = number of days o work to care for a child with illness k. w part = share of work force participation. The term 5 7 * (w part −0.5) * 2 adjusts the wage rate for the probability that a day of illness falls on a weekend and for the share of work force participation. We assume that only in (w part − 0.5) * 2 share of all households are both parents working. This depends on the assumption that all men are employed.
For the model a function has been dened that calculates the n-th partial sum of a geometric progression and subtracts the (b-1)-th partial sum. This allows for losses that only accrue from time b to time n. The formula looks as follows (20) a 0 with a 0 = future loss, q = 1+g 1+r , g = growth rate of loss, r = discount rate, b = rst period of loss, n = last period of loss.