Analysis of Research Report Results - Term Paper Example

1. Statistical Procedures used in the Study The study makes use of several statistical procedures in order to evaluate the medical errors in Pediatric Emergency Medicine. Random sampling was used to select the data set for analysis. For this 12 days were randomly selected from summer of 2000 and the charts of 1532 children treated in those days were analysed. Statistical reviews were performed by senior investigators in order to establish data accuracy. This is one of the best practices in statistical analysis. Also, the various errors were classified according to their significance/severity and the type of error. This ensured a meaningful analysis. K-statistic was used to remove any bias between various people and to ensure reliability of the ratings. K-statistic basically gives the unique symmetric unbiased estimator of the cumulative statistical distribution (Weisstein, 1996). Besides, descriptive statistics were used to analyse the occurrence of errors. Median, range and other measures were used for this purpose. Univariate analysis was conducted for all variables and the output/dependent variable (error). For the variables in which the relationship was significant, multivariate analysis was conducted. Logistic regression was the statistical procedure used for these analyses. 2. Conclusions of the Study Out of the 1549 visits to the Emergency Departments, charts were available for 98.9% or 1532 people. After review by research workers, it was concluded that 73.7% or 1129 cases had no potential errors. However, these were reviewed by senior investigators. After independent reviews by senior investigators as well as pediatricians, it was concluded that 10.1% or 154 cases had medication errors. These conclusions were drawn after a series of reviews and are justifiable. The K-statistic for initial agreement between reviewers on errors came out to be .79 while the K-statistic for ranking of errors came out as .70. As these values are close to 1, they indicate less bias and high degree of consensus between the researchers on ratings (Weisstein, 1996). Hence, the conclusions derived are quite appropriate. After this, the errors were classified on the basis of their types and severity. Since, these figures were based on facts and classification criteria decided beforehand, the conclusions made are appropriate. From the univariate analysis on the factors impacting the prescribing errors, a few important conclusions were made. Firstly, the medication errors were the highest between 4 AM and 8AM. Secondly, the proportion of errors was proportional to the severity of disease. Thirdly, the rate of errors was higher among trainees as compared to trained staff. Fourthly, the probability of error was greater during weekends tan during week days. Lastly, age group did not have a significant impact on the error rate or proportion. All these conclusions were derived from statistical tests using 95% significance level and can hence be justified. Moreover, the conclusions are in line with an intuitive perception. For instance, the concentration of a doctor is likely to be low in the early morning hours and week-ends leading to more errors. Similarly, it is more difficult to give accurate prescription for a high severity disease (that is the reason why it was high severity disease at the first place). Also, training is bound to reduce errors in any field. Hence, the conclusions can be well justified. After performing multivariate analysis, the severity of a disease and training were found out as the most important variables impacting the error rate. The study data was extrapolated for the entire population to come up with an estimate. This is the foundation of inferential statistics. Therefore, there is no fault in the methodology used in analysis. However, the conclusions are subjective based on the definitions and assumptions used in the beginning of the study. Hence, the conclusions may vary significantly from other similar studies. 3. Statistical Significance In both univariate and multivariate analyses, 95% confidence interval has been used to calculate the odds ratio. Therefore, those value variables for which p-value adheres to this confidence interval were used for final conclusion. For instance, in univariate analysis, there was no significant difference in the errors of different age groups from 12-18 years age group. Hence, the age group of the patients was not considered as a factor impacting the proportion of errors. For other variables i.e. Training, Severity of disease, Time of ordering, time within a week etc. there was a significant difference between the errors for various categories of these variables. Hence, it was concluded that these variables have a significant impact on the proportion of errors or the error rate. Therefore, the findings derived from this study are statistically significant. The process used to make this determination generally involves a few steps. Firstly, the hypothesis formulation is done. For instance, the hypothesis formulation for logistic regression for age group can be done as (Weaver, 2005): Ho (Null Hypothesis): The prescription error is independent of the age group Ha (Alternate Hypothesis): The prescription error is dependent on the age group After this the test statistic for logistic regression can be calculated (moore16.pdf). The same is compared with the actual values. Alternatively, p-value is calculated. If p-value < .05, null hypothesis is accepted which implies that there is no significant difference between the errors for various age groups. If p-value>.05, null hypothesis is rejected which implies that there is a significant difference between the errors for various age groups (Weaver, 2005). The analysis is generally performed using advanced statistical tools such as SPSS. In this case, the analysis has been performed using 95% confidence interval. This may be sufficient for many industries. However, in healthcare industry where the need for accuracy is greater, 95% confidence interval may not be sufficient. Hence, it would be prudent to use a 99% confidence interval. This would increase the accuracy of the analysis significantly though the size of confidence interval would increase. Also apart from logistic regression, multiple regression analysis can also be used effectively for this study. 4. References Logistic Regression. Chapter 16. Retrieved September 5, 2011 from Weisstein.Eric W. (1996). K-Statistic. Mathworld- A wolfram web resource. Retrieved September 5, 2011 from Weaver.B. (2005). Probaility and Hypothesis Testing. Retrieved September 5, 2011 from Read More