So far, you have heard about life expectancy, number of deaths and disease patterns around the world. But how do we know what we know about health in different countries? Where did the data come from? And more fundamentally, do we have reliable data for this statements we make about the state of world's health?. To start at the most basic level, many countries do not have complete information on how many people died because registering births and deaths which is called vital registration is not a universal practice. But even when the occurrence of death is recorded, its medical cause may not really correlate at all or at least correctly. Sometimes, this happens because there is limited information about the medical history of those who die. For example, a child who dies after an episode of fever in a remote area without reaching a hospital, may have died of diarrhea or malaria or pneumonia. Sometimes, a percent, especially in very young and very old ages, may have multiple condition and it's not clear which one led to their death. This map shows the overall quality of death registration and cause of that assignment for different countries based on work at the World Health Organization. As you can see, the quality of death registration varies across the world. The great majority of high-income countries in the Americas, Europe, and the Pacific have high-quality data as do some middle-income countries like Brazil, Uzbekistan, and Cuba. These countries regularly collect complete data on deaths and its causes. Death registration is of medium or low quality in many parts of Asia, the Middle East, and Latin America. Here, the data suffer from incompleteness or limited information on causes of death. In Sub-Saharan Africa and South Asia, most countries have low or very low quality death registration. The data may not be usable. You previously heard that some of the highest levels of mortality and the lowest life expectancies in the entire world are in Sub-Saharan Africa. This map shows that this information is based on weak data on actual deaths. In the absence of high-quality death registration, researchers have come up with other ways to estimate the number of deaths and their causes. For example, demographers have developed methods to use information about the woman's reproductive history such as number of pregnancies, live births, and survival of her children to measure child mortality. Similarly, survey data about number and survival of siblings can be used to measure adult mortality. These survey data typically come from a sample of the population of a country and in some places even from the national census. To obtain information on causes of death in places without medical certification, people have developed a method called verbal autopsy. In a verbal autopsy, a family member of the deceased is ask questions about medical history and some symptoms in days and weeks prior to death. Then, an algorithm uses the information to assign a cause of death. The reliability of course of death using verbal autopsy can vary based on the condition. For example, it is relatively straightforward to assign deaths from road traffic crash or drowning, or sudden chest pain followed by death most likely means a heart attack. For other conditions, symptoms are subtle and may go across diseases. For example, someone who has had a lengthy period of cars before dying may have had tuberculosis, chronic lung disease, or lung cancer. Even in those countries were cause of death data are medically certified, the data can still have errors. One of the reasons is the issue of multiple conditions that we saw earlier. Physicians, especially when a death happens outside the hospital may not have access to all of the diseases medical history or even have time to use it in a busy hospital. Finally, different physicians may use the information differently because of their own training or because of increasing awareness about some conditions. For example, there has been a large rise in high-income countries in the number of deaths that are assigned to dementias in older ages. Some of these rights is probably because of a real increase in dementia deaths. But it may also be the case that over time, physicians have become more aware of the existence of dementia and have more likely to write it on the death certificate. Of course, health policy and health care are not just about avoiding death but also reducing morbidity. So we also want information about non-fatal conditions and morbidity. There are certain diseases where some people will die and others will survive. For example, many people die of cancer, stroke and heart attack, while many others will be diagnosed by survive, and the survivors are likely to have been seen in a hospital. For these kinds of conditions, where there is good hospital records system or a registry, information on disease levels is available. Unlike cancers and heart attacks, for other diseases, some people may see a health care worker and others may not at all or only after years of having the condition. This could happen because they do not affiliate the symptoms with a disease like hearing difficulties or dementia. It could be because the condition is not covered by insurance like back pain and dental conditions, or because they're concerned about stigma for example for depression or alcohol and drug dependence. It could also be because they don't have immediate symptoms which is seen in hypertension and early diabetes. In this graph, you can see the prevalence of hypertension and whether people had received a diagnosis through a health worker or not in a random sample of England's population. As you can see, even in a place with a strong healthcare system and notice that we'll share of those with hypertension, have not been diagnosed and therefore are not recorded in routine health data. Therefore, for many diseases, routine health statistics cannot be used without at least some adjustment of the raw data. Many of these diseases can be detected in what is called the Health Examination Survey, which gathers information through a health-related questionnaire, physical exam, blood analysis, and other tests in a random sample of the population. But health surveys are costly and often have poor uptake levels. People who agree to participate may have better or worse health than the general population, which will affect the outcome of the survey. You can now begin to understand the challenges of estimating mortality given our data limitations. Improving death and birth registration should be a priority because it helps improve decisions about health and social care. You can also see that, in many places and for many conditions, data on morbidity are limited. Even where data exists and are complete, there may be errors. Next, you will have a closer look at some detailed examples of how estimates are generated for a specific conditions.
