Estimating caseload of pregnant and lactating women

This question was posted the Assessment forum area and has 14 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Anonymous 721

Normal user

8 Aug 2011, 18:09

Hi,
Please could someone tell me how I work out the caseload for pregnant and lactating women? Do you use GAM, MAM or SAM?
Is the incidence worked out with multiplication of 1.6 or 2 as with children under 5 years?

We will be screening and admitting using MUAC (<210mm).
Should we only admit to SFP or should be admit to OTP as well?

Mark Myatt

Technical expert

15 Aug 2011, 12:53

Caseload will depend upon the case-definition that is used for program admission. If you admit all pregnant and lactating women then you will need some estimate of the number of pregnant women in the population (P), the number of lactating women in the population (L), and your coverage target (C). Caseload is then:

C * (P + L)

If you have different admission criteria then you need some estimate of numbers in each eligible group.

The "1.6 or 2" is a correction factor to turn a prevalence estimate into an annual cumulative incidence. This will apply to children only and, even then, is pretty rough and ready since prevalence changes with time so the annual incidence estimate will depend on the time of the survey estimating prevalence.

You have to ask something like "Do I expect my caseload to change much over time?". I imagine that the number of lactating women will not change much over time (I amy be wrong) so you only have to account for new pregnancies. This requires a fertility estimate.

If you are admitting using (e.g.) MUAC < 210 mm then you will need to factor in prevalence / incidence in to the calculations above. It will all be very approximate.

Regarding the The SFP / OTP question ... I think something like:

MUAC < 160 mm : Admit to inpatient stabilisation

MUAC < 185 mm : Admit to OTP

MUAC < 210 mm : Admit to SFP

Might be suitable.

I hope this is of some help.

filippo dibari

Valid International.

Normal user

15 Aug 2011, 13:31

Hi,
in case of HIV/TB nut programmes, you might want to consider that there is an increase of high risk of mortality in adults with BMI<17 starting ARV - (Paton et al. 2006).

BMI<17 corresponds to a MUAC< 210 (based on Ferro-Luzzi et al. 1996). MUAC will allow you to have a better community-based active case finding for enrolment purposes. In other words, you do not want to wait to have HIV/TB adults wasted to enrol them in nutrition rehab.

MSF in Kenya (HIV/TB) were using MUAC 160 but the the patients were far too sick. BMI (weight/(height x height)) did not often work because the patients were often bedridden and could not stand properly due to gross weakness.

Contact me if you need more info on the papers mentioned here.

Cheers.

fil

Mark Myatt

Technical expert

17 Aug 2011, 08:52

I agree with Filippo. MUAC does facilitate case-finding in the community. This is well-known from CTC / CMAM programs.

You have to be wary of BMI and (W/H in general) as it is very influenced by body-shape and means very different things in different populations (and different individuals). For example, warm-climate Lowland pastoralists have lower BMIs than colder-climate upland agrarians. The Ferro-Luzi et al. thresholds hold only for a subset of populations.

It is important to realise that the MUAC < 160 mm threshold was proposed in 2000 for severe wasting requiring inpatient therapeutic feeding. If we translate this to CTC / CMAM terms this would correspond to severe wasting with complications requiring inpatient stabilisation. Things have moved on since 2000. The wide availability of ready-to-use therapeutic products allowing at-home treatment means that we are no longer restricted by bed availability and do not need to ration services as we had to ten years ago. Making it up as I go along (nothing new there!) we might have:

MUAC < 160 : Inpatient stabilisation
160 <= MUAC < 185 : Outpatient therapeutic using RUTF
185 <= MUAC < 210 : Supplementary feeding.

It is important to note that inpatient stabilisation of adults will require appropriate therapeutic diets (e.g. low protein to energy ratios).

Ezekiel

Normal user

17 Aug 2011, 10:02

I am confused. Can someone tell me the expected death rates in SC's based on sphere standards?

Rajesh Narwal

Normal user

22 Sep 2011, 11:56

Dear Ezikiel,

The SPHERE standards do not segregate the death rates by OTP or SC. The overall death rates of those under therapeutic care should be <10%.

Hope this helps

Mark Myatt

Technical expert

9 Nov 2011, 09:05

SFP is not usually considered as "therapeutic care". Mortality in the MAM groups (i.e. MUAC between 115 and 125 mm is not much above baseline of 1/ 10,000 / day which would be 0.3% in any month - you may want to check my arithmetic).

Anonymous 1059

Normal user

21 Dec 2011, 13:18

I would like to return to the question of estimating beneficiary numbers for pregnant and lactating women (PLW) for programs where an admission criterion is used (as opposed to admitting all PLW).

We are currently supporting a supplementary feeding program for PLW. We are using a relatively “generous” admission criterion of MUAC <230mm. Survey data suggests that 70% of PLW in the program catchment area have a MUAC <230mm, and program admission data has shown accordingly high numbers of admissions.

We are planning to expand the current program geographically in 2012, and therefore need to estimate the likely caseload. We have tried using the formula discussed in other threads for estimating beneficiary numbers in programs addressing child malnutrition (with PLW% replacing U5%) to help us factor in prevalence, incidence and coverage:

EXPECTED = POP * U5% * EP% * IC + (POP * U5% * EP% * CFI * AC)

POP = 33,000
PLW% = 0.07
EP = 0.7
IC = 0.5 (current program data)
CFI = 1.6
AC = 0.6

EXPECTED = 33,000*0.07*0.7*0.5 + (33,000*0.07*0.7*1.6*0.6) = 2,361

Total population of PLW = 33,000*0.07 = 2,310

As you can see, the expected number of beneficiaries is greater than the total population of PLW. This makes sense mathematically, but perhaps not in reality. Also, I note that Mark Myatt suggests above that the CFI of 1.6 applies only to children.

I would be grateful for any advice on the most appropriate way to estimate our likely beneficiary numbers in this situation.

Mark Myatt

Technical expert

22 Dec 2011, 10:23

I'll have a go at this ...

I have just done a similar calculation for a different application (sample size for a coverage survey of ANC services) and will work with those numbers (which are approximate - regional and national averages taken from DHS, Census, and the CIA World Factbook and rounded).

I think that we first have to work out how many pregnant women there will be on any one day:

    Population (Both sexes) = 110,000
    Birthrate = 44 / 1,000 population
    Births per year = (44 / 1000) * 110,000 = 4,840
    Births per day = 4,840 / 365.25 = 13.25
    Mean gestation period (days) = 280
    Number pregnant women on any day = 13.25 * 280 = 3,710

You're probably only interested in women in the second and third trimesters:

    Number T2/T3 on any day = 3,710 * (2/3) = 2,473

The next step requires you to specify the length of intervention. If (e.g.) you plan to intervene for one year after birth then need will be something like:

	((12 + 6) / 6) * 2,473 = 7,419

You may want to factor in mortality (at U5MR = 1 / 10,000 / day that would be about 3.7% per year) and applies only to the births:

    7,419 - (2 * 2,473 * 0.037) = 7236

Factoring in your eligibility criteria (70%) we get:

    7,236 * 0.7 = 5065

Factoring in coverage (including defaulting) you may have less than half of that:

    5,193 * 0.5 = 2,532

So, on any one day you might be supporting about 2,500 PLWs.

Someone should check my reasoning and my arithmetic.

Does this make sense?

Is it any help?

By the way ... (1) With 70% eligibility it usual to give the intervention to all PLWs as this saves on screening and removes the problem of rejected referrals, and (2) I agree with using the 230 mm threshold for PLWs - It's not generous since it is evidence-based (the 210 mm threshold is OK for non-PLW adults).

Anonymous 1059

Normal user

3 Jan 2012, 17:19

Happy New Year, and thank you for your helpful reply. I hope you won’t mind if I come back with some further questions.

As you suggest, we are targeting pregnant women in the second and third trimesters. However, we are intervening for up to six months after delivery rather than twelve. Following on from your calculations above, which give the estimated daily caseload for our programme, how would you recommend estimating the total number of pregnant and lactating women who would be admitted over a given project period (e.g. one year or 18 months)? Do we need to take account of the new/incident cases occurring over this period, and if so, how?

If the data is available, would it be more appropriate to use the infant mortality rate instead of the U5MR to factor in mortality? (If not, please could you explain the rationale?)

And lastly, I am a little confused about the overall approach here as compared to that used to estimate caseloads for children. I think I follow the logic of your calculations above, but the approach seems different to the “incidence plus prevalence” used as a basis for estimating caseloads for children. Would you mind explaining a little more?

Thanks again for your help.

Mark Myatt

Technical expert

3 Jan 2012, 18:59

Happy New Year!

Q1 : Following from the calculations above you would support about 2,500 PLWs on any one day. The method used to arrive at this caseload estimate takes into account new / incident cases. Using the numbers above, I think we have:

    Incidence per year = 3,710 * 0.7 = 2597

We'd need to factor in coverage:

    Admissions per year = 3,710 * 0.7 * 0.5 = 1299

This is for continuous enrolment of incident cases. There should be a spike of admissions of prevalent cases at program start.

Q2 : It would be better to use infant mortality rate but we would expect this to be lower than historical estimates as the mother is in a MCH program which (if it is any good) will reduce infant mortality.

Q3 : The approach is "incidence plus prevalence" but in this case we have reasonable estimates of both incidence and prevalence from the birthrate, population size, U5MR, and gestation period and these are usually stable over the medium term. With the child case we have a point prevalence (reflecting an unknown incidence and duration) and a correction factor that is derived from an informed guess about incidence and duration.

Again ... please take a look at my reasoning and arithmetic ... It may be very wrong.

Ernest Guevarra

Valid International

Frequent user

4 Jan 2012, 16:13

This discussion is indeed very interesting as the calculations that Mark presents here were something we were playing with as we found ourselves asking similar questions as you were while preparing a document on sample sizes for an assessment of a programme that included pregnant and lactating women as one of their beneficiaries.

I agree with the first set of calculations Mark presented on how to arrive at number of potential PLW beneficiaries. I just want to make the following comment (which doesn't really affect the calculation that much):

Mean Gestation Period: 280 days (or 40 weeks) is a sort of the general average that is used for gestation based on Naegele's rule. This is still continually used because of population-based studies that showed that the mean gestation period is 281 for first-time mothers and 280 for others. However, there is a more recent study (Patel et al, 2003) that showed that there seems to be differences in mean gestation age by ethnic group with 40 weeks (280 days) for white Europeans and 39 weeks (273) for Blacks and Asians. This is not too big a difference to critically affect the calculations and a little bit of an overestimate would be good anyway for planning purposes. I guess I just wanted to bring in this possibility that for African women, mean gestation age might be shorter.

As for the next set of calculations that Mark presented to answer the question on incidence per year, I would admit that I'm not that clear about some variables that Mark used. The incidence per year formula used by Mark was:

Incidence per year = 3,710 * 0.7 = 2597

Mark, is the 3,710 the total number of pregnant women in any one day and not just those in their 2nd and 3rd trimester? If that is the case, shouldn't it be 2,473 which is the number of pregnant women in their second and third trimester?

I tried calculating this in a similar way to the first set of calculations Mark used but calculating for a one year project with continued support for only up to 6 months after childbirth.

Number of T2/T3 on any day = 2,473

Next will be the length of intervention e.g., in this case up to 6 months after childbirth

((6 + 6) / 6) * 2,473 = 4,946

Factoring mortality (under 5) of about 37% (for births only)

4,946 - (1 * 2,473 * 0.037) = 4854

Factoring eligibility criteria of 70%

[code]4854 * 0.7 = 3,398

Factoring coverage of 50%

3,398 * 0.5 = 1,699

Using this calculation, I'm getting 1,699 compared to 1,299 which Mark came up with.

I'm not sure if I approached this correctly but I'm thinking that in theory, both calculations should have resulted in the same.

Mark Myatt

Technical expert

6 Jan 2012, 15:18

I fear that I may have confused myself and others ... sorry about that.

Let's go back to the original example population data:

    Population (Both sexes) = 110,000
    Birthrate = 44 / 1,000 population

It seems to me that incidence is the same as:

    Births per year = (44 / 1000) * 110000 = 4840

All we do is "lag" this back 6 months. This gives us an estimate of the the number of women entering T2 in any given year. That's our annual incidence estimate.

If we have 70% eligibility and 50% coverage we would recruit:

    per year = (44 / 1000) * 110000 * 0.7 * 0.5 = 1694

Which is (pretty much) what Ernest gets. I think the difference is just rounding error.

Is this any better?

Anonymous 1059

Normal user

20 Jan 2012, 13:53

Thank you Mark and Ernest for all of your input. I am hoping you will have the time and energy for one (last?) question on this, to check if I have understood correctly. Based on the discussion above, does it follow to say…

The number of individual pregnant and lactating women (PLW) who will benefit from a project will be the sum of the PLW eligible for admission at the start of the project (prevalent cases) and the PLW who become eligible during the project period (incident cases), multiplied by the proportion of PLW the project succeeds in reaching (project coverage).

Prevalent cases (those eligible for admission on day one of the project) will = births per day x average duration of admission, expressed in days x estimated prevalence of MUAC below threshold (if MUAC is used for admission).

Incident cases (those becoming eligible for admission during the project period) will = births per year x duration of project, expressed in years x estimated prevalence of MUAC below threshold (if MUAC is used for admission).

To continue with the example project:

Project duration: 3 yrs
Admission criteria: PLW with MUAC <230 mm
Admission duration: pregnancy (in practice T2 & T3) and 6 months of lactation
Total catchment population: 110,000
Birthrate: 44 per 1,000 population per year
Prevalence of MUAC <230 mm in PLW: 70%
Coverage: 50%

Births per day = 44/1000 * 110000 /365.25 = 13.25
Average duration of T2+T3+6 months lactation = (280*2/3 + 365/2) = 369.17 days
Prevalent cases = 13.25*369*0.7 = 3422.48

Births per year = 44/1000 * 110000 = 4840
Incident cases = 4840 * 3 * 0.7 = 10164

Estimated number of beneficiaries = (3422.48+ 10164) x 0.5 = 6793 women

(Note: I have not factored in infant/child mortality, to try to keep things a little simpler.)

I would be very grateful for your feedback! Thank you.

Mark Myatt

Technical expert

23 Jan 2012, 15:04

Yes ...

    case load = (prevalent cases + incident cases) * coverage

All we really need is an estimate of incidence since:

    prevalence = incidence + entries - exits

And we expect entries and exits to be the same (ignoring deaths) as T2 + T3 + six months is about one year ... it's a special case with duration = 1 year and it makes our calculations simpler.

We have

    incident cases = birth rate * population * eligible

Using the example data:

    incident cases = (44 / 1000) * 110000 * 0.7
                   = 3388 per year

For the first year we have:

    case load = (prevalent cases + incident cases) * coverage
              = (3388 + 3388) * 0.5
              = 3388

In subsequent years we only have incident cases:

	case load = incident cases * coverage
	          = 3388 * 0.5
	          = 1694

For three years we have:

    case load = [(years + 1) * incident cases] * coverage
              = [(3 + 1) * 3388] * 0.5
              = 6776

Which is about what you get.

I hope this helps.

» Post a reply