RESEARCH “RECALL BIAS OF ILLNESS INTERVIEW SURVEY!’ Aroon
REPORTING
METHODS
IN A HEALTH
“DEVELOPMENT AND VALIDATION OF AN INSTRUMENT FOR MEASURING STRESS IN CHILDREN BETWEEN 3-5 YEARS? Dipty lain, C. Gore, and N. Uikey, CEU, Government Medical
Chirawatkul, Y. Thavonpitak, and S. Tassaniyom, et al., CEU, Khon Kaen University, Thailand. Objective: (1) To compare illness episodes of household members reported every day with two-week recall survey. (2) To identify social and demographic factors that may influence the difference of the two methods. Design: This is a cross-sectional analytical study. Cluster sampling was used to select household sample. All members (513 persons) of 66 households of a rural area and 66 households of the urban area of Khon Kaen province were target sample. One school child in each selected household was asked to report any illness of household member on a daily basis. These reported data were compared with those derived from the two-week recall survey. Another 570 persons were interviewed by two-week method that used as control group: Setting: Rural and urban community in Khon Kaen province. Participants: All members of sample households who are available at the time preceding interview. Main Outcome Measures: Percentage of illness episodes decreased by twoweek recall survey. Results: (1) Number of illness episodes obtained by two-week recall method is 34 percent less than every day reported of the same group. (2) The average illness episode per person per two weeks of two-week recall survey of control group is 47.5% less than everyday report of the everyday interview group. (3) The difference of illness episode reported by the two methods are not influenced by rural or urban setting or age, sex, education or occupation of household members. Conclusion: Health survey using two-week recall should improve recall bias by interviewing technique such as using tracer condition questions.
College, Nagpur, India. Objective: To develop and validate an instrument for measuring stress in children of ages 3-5 years learning in schools as perceived by parents and teachers in Nagpur. Design: Study done in two stages. Stage 1: we constructed a questionnaire based on identification of 75 stress situations, behavior indicators and consultations with teachers, parents and experts. Stage 2: cross sectional design to establish factor structure, internal consistency, test-retest reliability, construct and concurrent validity. In absence of reliable gold-standard, a surrogate reference, Draw-a-Man test was administered to children which measures stress by low scores. Setting: Six primary schools randomly selected from corporation school lists. Participants: Stage 1: Groups of teachers, parents, subject experts and interviewers. Stage 2: 251 randomly selected parents, teachers and children between 3 to 5 years. Results: The instrument for parents contained 30 items while that of teachers had 28 items to be answered on three point scale for stress perception. Reliability coefficient by split-half method was found to be 0.5714 and 0.6577 as perceived by parents and teachers respectively significant at 0.01 level (df = 249). The correlation coefficient between stress scores estimated by parents and teachers was 0.6 (p < 0.01). There was negative correlation r = -0.12 (df249) between stress scores as perceived by parents, teachers and scores on Draw-a-Man test. Conclusion: The instrument to measure stress among school children aged 3-5 years had significant reliability and validity. The instrument scores had inverse relationship with Draw-a-Man scores, which is expected in a child with stress. Implication: In India, children between 3-5 years are subjected to formal examinations for admissions to schools and for class promotions. These children are likely to experience stress and are not able to express it. The proposed instrument can be used to measure stress in these children.
“USE OF MARKOV WITH REPEATED
“CRITICAL APPRAISAL OF ECONOMIC EVALUATIONS PUBLISHED IN THE FIELD OF RHEUMATOLOGY AND RELATED DISCIPLINES.” Marcos Bosi Ferraz, A. Maetzel, and C. Bombardier, CEU, Escola Paulista de Medicina, Sao Paulo, Brazil. Objective: The relationship between the effectiveness of health care interventions and costs is increasingly being evaluated. This study aims to critically appraise the economic evaluations (EE) published in the field of rheumatology and related disciplines. Data Identification: A MEDLINE search was performed from 1966 to February 1995 to identify all EE. Several keywords and text words were used. The total number of papers retrieved was 1,435. Study Selection: All papers had their title and abstract reviewed independently by 2 assessors and 57 were classified by either assessor as being a definite or possible full EE. Full EE was defined as an analysis comparing 2 or more strategies involving the assessment of both costs and consequences. The reference list of these papers identified 6 additional definite or possible EE that were also systematically reviewed. Data Extraction: All 63 papers selected were critically appraised independently by the 2 assessors using a standardized form. Divergences between the assessors were documented. A consensus form was hlled for each study. Results: Thirty-six full EE were identified (33 cost-effectiveness and 3 costutility analysis). The majority of studies were published between 1987-1990 (28%) and 1991-1994 (61%). The main topic areas covered were methods of prevention (44%), treatment (30%) and treatment-prevention (22%). The disorders most frequently studied were OA (30%), Osteoporosis (22%) and RA (14%). Direct costs and indirect costs were measured or estimated in 100% and 30% of the EE, respectively. The viewpoint of the analysis was stated explicitly in just 12 (33%) studies. Incremental and sensitivity analysis were presented in only 17 (47%) and 23 (64%) of the studies. Improper use of economic terms was also documented. Conclusions: The EE literature reviewed only partially adheres to basic analytic methods. There is a need to improve the methods of EE in the field of rheumatology, so policy makers can interpret and make better informed and efficient resource allocation decisions.
18s
MODEL IN A LONGITUDINAL STUDY MEASURES OF ORDINAL OUTCOME!’
L. Jeyaseelan, B. Antonysamy, and J.K. John, CEU, Christian Medical College & Hospital, Vellore, India. Objectives: To demonstrate the use of Markov model in a longitudinal study with ordinal outcome by a longitudinal study on “Factors Associated with the Course and Outcome of Schizophrenia.” In chronic disease epidemiology, when the repeated measures are on continuous scale, statistical method of choice is ANOVA for repeated measures. But when the repeated measures are on ordinal scale, the choice of statistical method is not familiar to researchers. They often do a point analysis or multiple comparisons. Design: Longitudinal study with active surveillance conducted during March 1981 to September 1988. Setting: Referral hospital in South India. Participants: Outpatients from the district of North Arcot Ambedkar and inpatients from southern states. Main Outcome Measure: Overall outcome with best, intermediate and worst categories, derived from pattern of course, occupational and social outcome assessed every year. Statistical Method: Transition probability matrix (TPM) was constructed using the change in the patients category every year from the immediate previous year. The life expectancy (LE) matrix was estimated using TPM. Lambda was also calculated to predict the category at time t+ 1, knowing the category at Hypothesis testing after controlling for other variables was done using quasi loglinear model. Results: The LE was 37.9,36.3 and 24.6 years for patients with best, moderate and worst outcome respectively. Patients who had age at onset over 24 years and duration of illness less than one year had significantly longer LE. Patients with duration of illness more than one year had 50% reduction in LE. Conclusions: The LE matrix provides duration of survival in each disease state. These estimates could further be combined with disease state specific quality of life adjustments to derive QALY estimates for use in clinical decision analysis. But Markov model has limited use in hypothesis testing to compare group of patients. However, this can be done by loglinear model or survival analysis.