Date of Award

1-1-2011

Document Type

Campus Access Dissertation

Department

Epidemiology and Biostatistics

Sub-Department

Epidemiology

First Advisor

Jim Burch

Abstract

Introduction: Working together with SC DHEC and PPC, this study created and evaluated a technique for developing new syndrome categories from poison control center data, ability to detect outbreaks, and tested the reliability of the PPC call codings.

Methods: Clinical effects and the case notes fields were utilized. The clinical effects variable contains a one-word symptom category classified by the PPC staff. Three clinical effects were of evaluated: respiratory, GI, and dermal. Six new syndrome variables, one general (GNS) and one specific (SNS), were created for each of three syndromes by scanning the case notes field for key terms. Poisson regression was used to test whether the number of CUSUM alerts detected by each variable were significantly different. The signal-to-noise ratio (SNR - testing seasonality) was computed for each variable. Aberration and outbreak sensitivity and specificity tests were conducted for data capture variables. Questionnaires were administered to eligible PPC specialists (74% of staff) to test the reliability call codings. Kappa statistics and linear regression models were calculated to evaluate significance in agreements.

Results: GNS and SNS variables captured more calls in respiratory syndrome missed by the clinical effects variable. A significant reduction in C123 alerts were detected by the GNS and SNS variables compared to clinical effect variable (GNS p-value: 0.0334, SNS p-value: 0.0079). The GNS variable in respiratory syndrome had most ideal sensitivity (67%) and specificity (52%) of the three variables and syndromes. Several potential clusters were detected using the PPC variables missed by SC DHEC. For intra-rater reliability, the majority (83%) of the PPC specialists had moderate to substantial agreement (kappa values range: 0.41 - 0.80).

Conclusion: Best practice would be to use both the clinical effects and case notes fields of PPC data. The new variables (GNS and SNS) displayed the flexibility of syndromic surveillance systems to adapt to new public health concerns. The PPC specialists reliability was mostly moderate to substantial which illustrated a high agreement in coding calls, although still room for improvement. This study strengthens the role of poison control centers in public health to aid in identifying potential outbreaks missed by the health departments.

Share

COinS