An Assessment Of The Measurement Methods And Quality Of A Pilot Study Of The World Health Survey In China

[BACKGROUND]For many years two challenges in the research field of health measurement have existed: whether the measuring instruments could capture the multiple dimensional profiles of health status and whether the measurement is comparable among different regions or cultures. In 2000, the World Health Organization (WHO) evaluated the health system performance of all member countries for the first time by using three overall objectives. However, the final rank list was questioned by many countries. After taking extensive consultations with member countries and assembling international expertise to carry out several turns of discussion, WHO initiated the World Health Survey (WHS) in 2001 in order to develop a set of instrument to gather data and methods of integrated evaluation. WHO expected the data and analytical results derived from this project would be more comparable among countries. Considering cognition and acculturation tests under a multilingual environment, the WHS instrument took a modularized structure and used the latest psychometric technique to enhance the cross-population comparability. Given the opportune context of the WHS, the Ministry of Health (MOH) of China cooperated with WHO and followed international standards to administer a survey in several provinces. It expected to strengthen the research capability of health measurement and health services surveys in China by testing the WHS instrument in a Chinese environment.[OBJECTIVE]To provide empirical data for WHO and establish valid, reliable and comparable measurement instrument of health and health system responsiveness.The specific aims are:●To localize and test the WHS instrument and validate the applicability of its modules in Chinese environment.●To analyze and systematically evaluate the reliability, validity and completeness of measurement of the WHS China pilot study.To establish a latent variable evaluation model of China's urban and rural residents with the WHS China pilot study data and assess the performance of ●the model.●To gain experience of the standard operation procedure on sampling, quality control and data management in a large scale household health survey of WHO and to improve the technique for health-related social surveys in China.[METHODS]Data Collection MethodMultiple stage random sampling method was used to draw 4000 households as household interview sample at 10 surveyed sites, which were selected at random from the Chinese Health Surveillance System, and 10% of households were selected at random to perform a retest. During the survey process, standardized WHS instruments and interviews were used by interviewers to collect household and individual health care information from interviewed households.Data Analysis MethodCohen's kappa, concordance correlation coefficient and intraclass correlation coefficient were used to analysis the test-retest reliability of measurement of the WHS China pilot study. The overall test-retest reliability of each module of the WHS instrument was compared with an integrated test-retest indicator. Cronbach's alpha and item-test correlation were used to evaluate the internal consistency of the cumulative Likert scale.Moreover, the dissimilarity index, sample-population deviation index, Myer index and Gini concentration ratio were used to analyze the sampling validity of household sample and of individual sample respectively. Correlation and factor analyses were used to evaluate the construct validity for several key scales of the WHS instrument.Item response rate and item responsiveness spectrum were used to analysis the completeness of measurement, and the misreporting of every item was also analyzed. A multiple imputation method was used to fill in the missing items in order to improve the efficiency of data utilization.In the procedure of health assessment, special statistical methods for survey data was first used to analysis the individual self-reported health. The result was then calibrated by the latent-variable-theory-based model.