Iron deficiency anemia is so common around the world that it is a concern for almost every population in both developing and developed nations. Although iron deficiency has been studied in vast detail, there is no simple or reliable clinical measure for iron deficiency anemia. The signs of clinical deficiency include pallor, fatigue, breathlessness, dizziness, palpitations, and edema (when severe and chronic) are not used to assess deficiency for several reasons. Pallor (or paleness) of the tongue and the conjunctiva of the eye is difficult to detect (e.g. it may look different depending on the person’s natural skin tone) and takes a lot of clinical experience to recognize. Some of the other clinical manifestations such as fatigue, breathlessness, and palpitations, they are not specific to iron deficiency and therefore may result from other causes.

CLINICAL INDICATORS:
(not often used for assessment, most are not specific to anemia)

As a result of having non-specific clinical indicators (that are not diagnostic), assessment usually involves the use of of sub-clinical indicators.

SUB-CLINICAL INDICATORS:

Fortunately, the diagnostic test for anemia using hemoglobin as the indicator has become relatively inexpensive and may be done easily with blood from a finger prick using a small, portable device. The laboratory test for anemia is far easier and therefore more widely used than that for serum retinol (VAD) or urinary iodine (IDD). Data sets (especially large datasets, >1000 cases) rarely have sub-clinical indicators collected other than hemoglobin. It might be seen in a very specific field survey for assessment of IDA, but it is not very common. The down side to this is that hemoglobin measures a late stage of deficiency since hemoglobin does not drop until iron stores are already depleted in the body, whereas some of the other indicators (e.g. serum ferritin) reflect earlier stages of depletion. Even so, hemoglobin does reflect the iron status of the population albeit at a point when attention is desperately needed.

In a data set from Cambodia (cambodia.sav), data on iron deficiency has been collected (as well as vitamin A, and iodine), specifically hemoglobin. Run a few CLEANING EXERCISES FOR HEMOGLOBIN (frequencies, descriptives, and histograms). Look for distribution patterns in both children and mothers for hemoglobin and compare the mean and median values.

The two histograms produced from the Cambodia clean data illustrate the normal distribution of the hemoglobin outcome variable. If you were to see outlying values (say 25 g/l Hb) you might consider "cleaning" the case by setting to missing. The range for hemoglobin is quite wide, but it would be unlikely to see measurements below 3 g/l Hb and above 20 g/l Hb. The lower bound has a more obvious limit, because the total lack of hemoglobin would measure as 0 g/l. Even so when hemoglobin is 2 g/l or lower, there probably evidence that there is not enough oxygen to support the body, therefore the measurement might likely be invalid. The upper bound has a less distinct cut-off, considering some populations (especially certain European descendants) have a prevalent genetic disorder called hyperchromatosis, leading to excess iron storage (hence, higher hemoglobin scores). Considerations for local influences, including genetics and more importantly altitude, are necessary when collecting data.

In the Cambodia data, both maternal and child hemoglobin look similar, where the mean of the population of mothers is slightly higher (as expected) than that of the children 0-5 years. Both mean values of hemoglobin are lower than the cut-off for anemia (11 g/l Hb for children 0-5, and 12g/l Hb for non-pregnant women), which indicates you will find anemia prevalent in these populations. The causes of anemia have been discussed and will be revisited for this population in subsequent sections.

 Now, try using the following table to create a new variable, CATEGORIZING HEMOGLOBIN into anemic and not anemic. This will involve a simple recoding (into a different variable) routine. Name the new variable anemia, if you like.

 

Indicator	Individual level cut-off for mild/moderate anemia
Hemoglobin- 6 mo – <5 years ³ 5 years – 11 years ³ 12 years – 14 years Women (not pregnant) Pregnant women	<11.0 to ³ 7 g/l <11.5 to ³ 7 g/l <12.0 to ³ 7 g/l <12.0 to ³ 7 g/l <11.0 to ³ 7 g/l
Hematocrit- 6 mo – <5 years ³ 5 years – 11 years ³ 12 years – 14 years Women (not pregnant) Pregnant women	33% 34% 36% 36% 33%

To code for multiple levels of anemia --mild, moderate, and severe, these cut-points are as follows: mild – upper end of the listed range above down to 10g/l Hb; moderate- 7g/l Hb up to 10g/l Hb; severe - <7g/l Hb. This is performed using the recode, into a different variable routine again.

On a rare occasion, other outcome indicators will be used that provide some information about the individual’s iron stores such as serum ferritin. This is usually not collected in developing countries though, because of lack of analysis facilities and expense. Additionally, serum ferritin levels are influenced by infection, which is common in developing countries. Almost always, hemoglobin or hematocrit values will be collected as the only outcome indicator of IDA.

It could also be interesting to at the less specific outcomes of IDA such as still births or infant deaths. If anemia is prevalent in the population, then look and see how many still births are reported and if they are more common among the anemic mothers. This data set contains a variable called Alive (child is currently alive or has died) and a variable called Diedch (number of children <5 who have died in the previous year).

Process Indictors and other Independent Variables Related to IDA

In addition to the outcome variables of interest, cleaning may be more tedious when considering the process indicators collected for IDA since the responses are likely more transformable than that for a clinical measure. To get a better sense of what this means, look at the listing of possible process variables for IDA and then make a list from the Cambodia data set of the non-outcome indicators in that data set (use the codebook).

Micronutrient Indicators

 
Now that the variables of interest are listed out, it will be a little easier to begin running some associations and creating transformed variables that make comparison easier. For example, the nutrition variables WAZ, HAZ, and WHZ can be transformed into underweight, stunted and wasted respectively (using <-2 SD as the cut-off). With these you can look at the associations of the anthropometry measures of general malnutrition against IDA.