Small count data presents challenges for analytics and requires significant data inspection. Outliers may become a problem for small count data and are often overlooked. A single outlier may have a strong effect on small count data because of the difficulty in identifying the nature of the outlier. In the healthcare industry, small count data often represent life-threatening incidences, such as rare diseases or “never events” in surgeries. The focus of this research is on data sets that report retained surgical devices.
| Authors | Pelaez, A. | Hofstra University |
|---|---|---|
| Winston, E. | Hofstra University | |
| Nejati, N. | 5E Analytics, LLC |
Abstract: Small count data presents challenges for analytics and requires significant data inspection. Outliers may become a problem for small count data and are often overlooked. A single outlier may have a strong effect on small count data because of the difficulty in identifying the nature of the outlier. In the healthcare industry, small count data often represent life-threatening incidences, such as rare diseases or “never events” in surgeries. The focus of this research is on data sets that report retained surgical devices. Any single outlier may have an impact on performance. Accurate identification of outliers is essential to healthcare providers. This paper examines two approaches for outlier identification of retained surgical devices data. A proposed method is based on the impact a potential outlier has on the variance of the data. The results of this method are compared to a chi-square distribution to identify potential outliers. The method reported a significantly lower number of outliers in comparison to Tukey fences. The results show that using the variance difference method provides a lucid and conservative approach to outlier identification. In the case of retained surgical devices, outliers may represent variation in the quality of surgical care provided at hospitals. The variance difference method has the potential to help with the accurate representation of small count data.
Citation:
Pelaez, A., Winston, E., & Nejati, N. (2019). Small Count Data and Outlier Analysis: An Exploratory Study of Patient Safety. Journal of Applied Quantitative Methods 14(2).
.
