Edit this paragraph II. Research purposes 1. Describe the three-compartment distribution of a disease or health condition. Through the investigation of a certain area or population, we can get the distribution of a certain disease in time, area and population, so as to find the high-risk population or find the relevant etiological clues, and provide the basis for disease prevention and control.
2. Describe the relationship between certain factors or characteristics and diseases, and determine the risk factors. For example, through the investigation of coronary heart disease and its risk factors, the relationship between hypertension, hyperlipidemia, overweight, smoking and related occupations and coronary heart disease was found, thus providing a basis for reducing risk factors and reducing the occurrence of coronary heart disease.
3. Provide valuable information for evaluating prevention and control measures and their effects. For example, after taking measures for several periods, cross-sectional studies are repeated, and the effect of measures implemented in the previous period can be evaluated according to the comparison of prevalence differences.
4. Provide basic data for disease surveillance or other types of epidemiological research.
Edit the third paragraph. Research purposes 1. Describe the distribution of diseases or health conditions.
2. Evaluate the health level of a country or place.
3. Study the factors that affect people's health and are related to diseases.
4. Used for the research of health service demand.
5. Used to evaluate medical or preventive measures and their effects.
6. Used for the formulation and inspection of relevant health standards.
7. Used to check and measure the quality of past data.
8. For the formulation and evaluation of community health planning.
Edit the fourth paragraph. Classification of Cross-sectional Studies (1) Census refers to the investigation or examination of every member of a certain range of people in a certain period of time in order to understand the epidemic situation of a certain disease or the health status of a certain population. The emphasis here is "every member of a certain range of people". For example, all residents of residential areas. A certain time can be 1-2 days or 1-2 weeks, and a large-scale census can also be completed within 2-3 months. The time of the census should not be delayed for too long, so as to avoid changes in the diseases or health status of the population and affect the quality of the census.
China has rich experience in conducting population census. A large-scale survey was conducted on tumors, cardiovascular diseases, goiter, hepatitis B and tuberculosis. Through early treatment and repeated prevention and treatment, some of these diseases have been controlled or basically controlled, and remarkable results have been achieved.
The main purpose of the census is to find cases early and give them timely treatment. It is best to have a high disease prevalence rate in the general survey, so that enough cases can be obtained in a short time.
Because the census is to investigate all the members of a certain population, it is relatively simple to determine the object of investigation; Moreover, the obtained data can understand the three distribution characteristics of the disease, so it can give some enlightenment to the epidemic factors of the disease.
But for diseases with short course of disease, low prevalence rate or complicated examination methods, it is not suitable for general survey. Due to the large number of census objects, it is inevitable to miss diagnosis and misdiagnosis; Due to the large number of staff involved in the census, their proficiency in investigation techniques and inspection methods is different, and the quality of investigators is difficult to control; At the same time, due to the heavy workload, it is difficult to conduct in-depth and detailed investigations.
(2) Sampling survey In actual work, some observation units (statistically called samples) are usually randomly selected from the population for investigation, which is called sampling survey. Sampling survey is to estimate some characteristics of the population represented by the sample according to the results of sampling survey, so sampling survey must follow the principle of randomization in order to obtain better representative samples. Sampling survey can save manpower, material resources and time. Because of its small scope of investigation, the investigation work is easy to do in detail. However, the design, implementation and data analysis of sampling survey are complicated, and duplication and omission are not easy to find, so it is not suitable for the research objects with excessive variation.
Commonly used random sampling methods are as follows:
1. Simple random sampling means that all observation units in the survey population are numbered first, and then some observation units are randomly selected to form a sample through a random number table or lottery.
At present, in the cross-sectional study, because there are too many observation units, it is difficult to number all the observation units, so there are not many opportunities to use simple random sampling, but it is the basis for implementing other sampling methods.
2. Systematic sampling is also called equidistant sampling or mechanical sampling. That is to say, the whole observation unit is divided into n parts according to a certain sequence number, and then k observation units are randomly selected from the first part, and then one observation unit is mechanically extracted from each part at equal intervals to form a sample.
Example 1 To know the HBSAg positive rate of employees in a certain unit, there are 1000 employees in this unit. Try to take a sample of 100 according to the systematic sampling method.
The current population is N= 1000, the sample number is = 100, and the sampling interval is =100/10. First, randomly determine a number between 1 and 10, such as 4.
3. Stratified sampling is also called classified sampling. That is to say, the population is divided into several types or groups (called "strata" statistically) according to a certain characteristic that has a great influence on the observed values, and then a certain number of observation units (which can be determined by proportion or optimal distribution) are randomly selected from each layer to form samples.
For example, the people surveyed are divided into different levels according to their age, gender or disease severity, and then randomly sampled at each level. Stratified sampling can reduce the sampling error caused by different characteristics of each layer.
4. Cluster sampling In cluster sampling, not an individual is sampled, but several groups (groups) composed of an individual. Cluster sampling is to divide the crowd into k "clusters" (such as k regions, etc.). ), each cluster includes several observation units. Then several groups are randomly selected from K groups, and all observation units in each group are sampled.
For example, the survey of hookworm disease randomly selected all villagers in several towns of a county. When doing a family planning survey, check all the residents of several neighborhood committees in the city.
Because the sampling survey randomly selects some observation units from the whole population as the survey objects, it will inevitably produce sampling errors, and the size of sampling errors varies with different sampling methods. Generally speaking, the order of sampling error from small to large is stratified sampling, systematic sampling, simple random sampling and cluster sampling.
(III) Sample size of sampling survey The method of finding the sample number n required for estimating the overall mean (or rate) is as follows (for hypothesis testing, see the experimental design for the estimation method of n).
1. For the average sampling of measured data, if you want to estimate the overall average, you can use the formula to estimate the number of samples.
n=[(tαs)/d]2 ( 15- 1)
Where n is the number of samples, d is the allowable error, that is, the allowable limit of the difference between the sample mean and the population mean (specified in advance), and s is the estimated standard deviation, which can be obtained through previous data or small-scale investigation. When α is determined, tα in the formula can be found in the T boundary value table (take both sides). In general, if the sample is large, α is 0.05 and tα= 1.96, which is about 2, then
n=(4s2)/d2 ( 15-2)
Example 2 It is planned to investigate the hemoglobin content of primary school students. The estimated standard deviation is 3g/dl, and it is hoped that d will not exceed 0.5g/dl, and α=0.05. How many people need to be investigated?
In this case, s = 3, d = 0.5, and α = 0.05, calculated according to the formula 15-2:
N=(4×32)/0.52= 144 (person)
That is, we need to investigate 144 people.
2. Ratio sampling For counting data, when you want to estimate the overall ratio, you can use the formula to estimate the number of samples.
n =[tα2P( 1-P)]/D2( 15-3)
Where p is the estimated rate, and other symbols have the same meanings as before.
Let d=rP(r is the allowable error coefficient of p), then the formula becomes:
n =(tα2/R2)[( 1-P)/P]( 15-4)
It is generally assumed that α=0.05 and r=0. 1, which is converted into:
n =(4/0. 12)[( 1-P)/P]= 400×[( 1-P)/P]( 15-5)
The third case is to investigate the infection rate of schistosomiasis. According to previous data, the estimated prevalence rate is P=30%. Let α = 0.05 and d = 0.1p. How many people need to be investigated?
In this case, P=0.30, α=0.05, r=0. 1, calculated according to the formula 15-5:
N=400×[( 1-0.3)/0.3]=933 (person)
That is, 933 people need to be investigated.
(4) Data analysis of cross-sectional study The data obtained from cross-sectional study can be collated and analyzed according to the following steps.
1. Check and check the original data, check the accuracy and completeness of the original data, fill in the blanks, delete duplicates and correct errors.
2. According to the clearly defined criteria of diseases or certain health conditions, all the interviewees were grouped and classified.
3. Compare the original data in groups to understand the distribution of diseases or a certain health state in different regions, different times and different people.
(V) Common bias in cross-sectional studies When the results of a study (observation) are different from their true values, the phenomenon or result of such differences is called bias.
1. The proportion of respondents who have no answer bias in the interview survey or communication survey is called the response rate. The factors that affect the response rate are: 1) people's understanding of the survey; 2) Whether the investigation method or content is appropriate; 3) The subjects are healthy, so they don't care about the disease investigation; 4) The respondent is in poor health or refuses to investigate at an advanced age; 5) Respondents did not meet when they went out. If the proportion of non-responders is high, such as 30% in the sample survey, it will cause bias.
2. Memory bias or reporting bias This is the bias caused by the respondent. For example, patients with certain diseases can often recall the past exposure history, while healthy people often forget the past exposure history. When investigating some sensitive questions, the respondents may be unwilling to give correct answers, which leads to biased reports.
3. Investigators are biased, for example, investigators consciously investigate objects with certain characteristics, not others. Sometimes, in order to get the content or answer you need, you have to ask leading questions.
4. Measurement bias Due to inaccurate instruments, inconsistent reagents and different experimental conditions, the measurement results are incorrect, which can make the investigation results deviate from the true value.
(6) Quality control in cross-sectional study Quality control is biased control, which is the key to the success or failure of current investigation. The above deviation should be considered in advance in the design stage to prevent it from happening. The main control methods are as follows.
1. Adhere to the principle of randomization and select the research objects in strict accordance with the sampling design scheme. Analyze the causes of non-response in time, make up for leaks and improve the detection rate.
2. Select precise instruments and equipment and calibrate them in advance to ensure the accuracy and reliability of the test results. 3. Strictly train investigators, supervise and control their quality, and unify investigation procedures and methods.