The use of informal household surveys for rapid health assessment

Although large household surveys are time-consuming exercises, smaller surveys can be carried out more quickly in emergencies. During the initial assessment of an emergency, limited surveys using non-probability sampling of affected populations may provide an estimate of the extent of damage and immediate health needs for guiding emergency decisions. However, the results of these surveys may be difficult to compare to those of subsequent, more statistically valid ones.

Larger, statistically valid household surveys are a valuable tool during later stages of the emergency, when there is more time available to refine the initial estimates, based on the rapid health assessment. Given the variety of situations in which rapid household surveys may be conducted, each one must be designed specifically. This annex does not provide assistance in deciding what information to collect, writing interview questions, choosing a representative sample, and analysing the resulting data, all of which require skilled personnel. It does give a broad overview of some issues involved in conducting surveys during rapid health assessment.

The process of selecting a sample

The purpose of conducting a survey is to describe key characteristics of the population under study, such as the proportion of houses damaged by an earthquake or the proportion of children vaccinated against measles. To derive an accurate estimate, the survey sample must be representative of the overall population. Therefore, if the affected population is very large or dispersed over a large area, the survey sample should be taken from as wide an area as is practical and not restricted to a small sub-area, which may not be typical of the population as a whole. Moreover, surveys should avoid sampling only the most accessible members of the affected population (e.g. those living along roads, near markets or in the centre of town).

The first step of any survey is to define the area under study. It is usually best to draw a rough map of the area that would include as much detail as possible about where people live, relative population concentrations, and major geographical features, such as roads and rivers. Use local informants to provide overall information about an area, as well as information on which areas are most and least affected. Investigators may wish to draw their sample from areas showing a wide range of severity of impact. It is a good idea to ask different people their opinion. The next step is to decide how to select the sample and its size. This decision depends on a number of factors including:

- the size of the area under study;
- the number of investigators available;
- the amount of time available for the survey;
- the availability of transport;
- the distribution of the affected population (e.g. isolated households, villages, and camps); and
- the different circumstances facing people in various parts of the emergency-affected area.

The simplest and quickest survey can be done by choosing a sample of 50 households at random. Data collection in this survey may take two or three people only one afternoon or less to complete. More extensive surveys may be necessary, but will require more people and time to complete.

Cluster sampling is a technique developed to save on survey costs. It involves selecting random starting points and then subsequently choosing systematically. For example, in a rural area, 30 villages are chosen randomly from a list of all the villages in the affected area. Then in each village, a house is chosen at random and subsequent houses chosen by selecting the houses closest to that house. To select households spread over a larger area of the village, the team may choose to select every fifth or every tenth house until the required number of houses in that village is surveyed.

The number of households visited in each cluster depends on what is to be assessed. In the standard cluster survey used to assess vaccination levels, seven households are chosen in each cluster. Such a survey may require three to five teams of interviewers and take three days or more to complete. In some assessments, as many as 700 children must be found to estimate health parameters with the necessary degree of precision. Depending on the amount of travel needed, these surveys may require 10 or more teams and more than a week to complete. These large surveys may not be appropriate for rapid health assessment in sudden emergencies, such as natural emergencies or sudden population displacement, where information about the population is needed very quickly.