Today, we're going to talk about what is a good coverage indicator. First, we want to go back to this question of what is the definition of a coverage indicator. Remember, we defined intervention coverage as the proportion of those in need of a service or intervention, who receive that service or intervention. What this looks like in terms of an actual indicator is that, again, we have a numerator divided by a denominator. The denominator is a measure of everybody who needs the service or intervention, and the numerator is people who need the service or intervention who receive it. To reiterate then, a coverage indicator must be a proportion. Some other characteristics of a good coverage indicator, the intervention needs to be well-defined. What is the service or intervention that they are supposed to be receiving? If it's something like breastfeeding promotion, what does that mean? Does that mean that they got counseling? From whom? At what point? Does it mean they were exposed to a mass media campaign? So really defining the intervention very precisely so that you can ask respondents about it and they will understand what you're asking about. The denominator should include everybody in the population in need of the intervention, which sounds pretty straightforward, but we'll talk in a minute about some cases where that can be more complicated to do. The numerator needs to include everybody in the population in need of the intervention, who then actually received it. This means the numerator must be a subset of the denominator. The numerator and the denominator should be bounded by a reference period. What I mean by that is whenever we're measuring something, we're measuring it over some period of time. It might be on the day of the survey, on the day of the interviewer's visit to the household. You ask, "In your home today, do you have an insecticide-treated bed net? Can I see it?" It might be the last 24 hours, so you might ask what a child ate or drank over the last 24 hours. It might be the last three years. You might ask a woman if she had a pregnancy in the last three years, and if so, whether she received antenatal care during that pregnancy. But you need to specify what the reference period is that you're asking about. Then both the numerator and the denominator need to be measurable. We can define very beautiful coverage indicators that are not things that we can actually measure, given the currently available methods. If it's something that you actually want to use, it has to be measurable with the available methods. Let's talk about some common indicators that you may encounter with coverage indicators. The first of these is that the indicator measures service contact rather than receipt of interventions. We've actually talked about this earlier in this lesson but I want to come back to it here. This idea that just because you had contact with a health service, doesn't mean that you actually got the interventions you needed. The example here, skilled birth attendant, is defined as the proportion of live births attended by a skilled health personnel over some period of time; three years, five years. But just because a woman delivered attended by a doctor, a nurse or a midwife, doesn't mean that she actually got the interventions she needed during labor and delivery, and we don't know what she got during labor and delivery just from this indicator. There are lots of similar indicators; antenatal care, one or more visits, four more visits, care-seeking of a child illness, any indicator that just measures whether somebody had contact with a particular health service but doesn't ask about what they received during that contact. Again, we measure these, but you want to be aware of the limitations of these indicators. Another common issue, the denominator, includes individuals not in need of the intervention. This happens sometimes where need is difficult to accurately measure using population-based methods, so using a household survey. An example of this is coverage of C-sections, so the proportion of live births delivered via cesarean section. Not all births need to occur via a cesarean section. In fact, only a small fraction do. But we don't have a valid way of assessing in a household survey interview with a woman, whether during her delivery which happened a year or two years or three years ago, she was in need or not of a cesarean section. In order to know that, you would have to go to the records of her birth at the health facility, assuming those existed and were of good-quality. The way we measure this is we just measure C-sections among all live births. We have some idea; approximately 5-15 percent of women, or of births need to occur via cesarean section. What we don't know when we measure this though, is if we measure a C-section rate of 15 percent, are all 15 percent of those women who were actually in need of a cesarean section, or are some of them elective C-sections? We have no way of knowing that. This is the limitation of including individuals not in need in the denominator, and therefore the numerator is you don't know whether the people who got the intervention actually needed it. Another common issue, somewhat related, is that the denominator is not representative. What do I mean by that? I mean it doesn't represent everybody in the population in need of the intervention. Where this commonly comes up is where denominators are estimated in a health facility settings. There, you're only seeing people in contact with the health facility or where the denominator is based on a diagnosis. People with a particular diagnosis, and of course, there are people who have a condition, but who have not been diagnosed with that condition. If you are using diagnosis as a denominator, you will only be measuring among those with a confirmed diagnosis and you'll be missing everybody who's undiagnosed. An example of this is coverage of PMTCT, prevention of mother-to-child transmission of HIV. This is defined as a proportion of HIV positive pregnant women who receive a complete course of ART prophylaxis. But that's only measured among women with an HIV diagnosis, who have been tested and found to be HIV positive. That is only done in health facility settings, in antenatal care settings. You're missing all those women who may be HIV positive, but have not been diagnosed and therefore, may not have received prophylaxis during their pregnancy. Then an issue that you see among basically, every indicator out there, every coverage indicator and most other indicators as well is information error and misclassification. This can affect the numerator and/or the denominator. It's affected by the type of intervention, but again, you will see some degree of misclassification or information error, basically, every indicator out there. What I mean by that is, the respondent giving inaccurate responses to the question. If you ask a respondent how many times they went for ANC and maybe they went six times, but it was three years ago, they don't quite remember and so they might report that they went five times. A common question is, for how many days during the pregnancy they took iron containing supplements? That's a hard thing to remember years after the fact and so women may give an inexact, and inaccurate response. Another example of this is antibiotic treatment of childhood pneumonia. This was something that was a measurement priority for awhile because pneumonia is a leading cause of death for children under five, antibiotic treatment is the primary curative intervention. This was defined as the proportion of children 0-59 months with suspected pneumonia receiving appropriate treatment with antibiotics. But actually, measuring that is very difficult. Measurement of antibiotic treatment of childhood pneumonia involves a number of different components. For a child to be effectively treated for pneumonia, first of all, they have to have pneumonia. The caregiver needs to realize that they are ill and take them for care. The child needs to be correctly diagnosed and prescribed an antibiotic. The caregiver has to actually give the antibiotic to the child. They have to give the correct dose. The child needs to complete the full course of antibiotic. It also needs to be an effective antibiotic, meaning both that there's no resistance against that particular antibiotic and also that it's not a counterfeit drug. If you really wanted to measure this the caregiver would need to be able to know all these things, remember all these things, and report on them accurately. You would have to measure at each of these steps, ask questions about each of these steps. But we can't for a number of reasons. For one thing, just at the diagnosis stage, there have been several studies now that have shown that questions asked to caregivers during household surveys cannot determine whether a child had pneumonia or not. Many of the children who are identified as having suspected pneumonia in household surveys actually don't have pneumonia and therefore, shouldn't be included in either the numerator or denominator. We're measuring antibiotic treatment among kids who don't need it. The other thing is that the components of the quality of the intervention, whether the child received the correct dose, whether they completed the full course, whether the antibiotic was effective, are things that are very difficult to measure via household surveys, either because they're not things that a caregiver would know. They wouldn't necessarily know whether it's ineffective antibiotic or they're difficult to ask about and to recall and so they're not typically things that we ask about. We don't typically ask about the dosing. We don't ask for how many days they received it. Really the way this was being asked was, does your child have signs of pneumonia? Did they get an antibiotic? Because of the issues with the denominator now, the way this is asked is, does your child have signs of pneumonia? Were they taken for care? We just look at care-seeking because every child with signs of acute lower respiratory infection should be taken for care but not all of them need antibiotics because not all of them have pneumonia. That gives you an idea of some of the challenges that we run up against in measuring some of these coverage indicators in household surveys. To summarize, coverage indicators need to have a denominator. There were proportion. The denominator needs to reflect those who need the service or those who would benefit from a particular behavior or practice. The denominator needs to be representative of the population in need and not just represent those with a diagnosis or those seeking services at a health facility. Coverage indicators also need to have a numerator that reflects receipt of an intervention rather than just contact with the service provider. They should specify a recall or reference period. Finally, coverage indicators are subject to information error or misclassification. When you are defining an indicator, when you're measuring an indicator, you need to be thinking about what survey respondents can accurately report on.
