Pediatric Electrolytes

Hogan et al. (2003) used correlation analysis and the detection algorithm method to study the sales of pediatric electrolyte products during outbreaks in children. Pediatric electrolyte products are solutions of salts and water that are indicated for the treatment of dehydration in children ages five years and under. They studied the sales of pediatric electrolytes during annual winter outbreaks due to diseases such as rotavirus gastroenteritis and influenza in children ages five and under. Their study involved six urban regions (which we refer to as cities) over the course of three winters.

The researchers obtained a historical data set of sales of pedi-atric electrolytes from Information Resources, Incorporated. The data set contained weekly sales of pediatric electrolyte products sold for the six cities for the three-year study period. The sales data represented nearly 100% of all sales. They formed the sales data into time series for each city. To create a gold-standard reference time series, they obtained all hospital ICD-9-coded discharge diagnoses for the same time period for each of the cities, grouped hospital ICD-9 diagnoses into respiratory and diarrheal groups, and created time series of weekly counts (Figure 22.2). When forming the weekly counts, they used the date of admission of the patient to hospital (the earliest date at which a biosurveillance organization could possibly detect an outbreak from hospital data) to bias the

figure 22.2 Correlation of weekly sales of electrolyte products and hospital diagnoses of respiratory or diarrheal illness in children aged 0 to 5. (Data courtesy of IRI, Utah Department of Health, Indianapolis Network for Patient Care, and PA HC4. From Hogan, W. R., Tsui, F.-C., Ivanov, O., et al. (2003). Early detection of pediatric respiratory and diarrheal outbreaks from retail sales of electrolyte products. J Am Med Inform Assoc 10:555-62, with permission.)

figure 22.2 Correlation of weekly sales of electrolyte products and hospital diagnoses of respiratory or diarrheal illness in children aged 0 to 5. (Data courtesy of IRI, Utah Department of Health, Indianapolis Network for Patient Care, and PA HC4. From Hogan, W. R., Tsui, F.-C., Ivanov, O., et al. (2003). Early detection of pediatric respiratory and diarrheal outbreaks from retail sales of electrolyte products. J Am Med Inform Assoc 10:555-62, with permission.)

analysis against finding that detection from OTC monitoring was earlier than detection from more traditional data used in public health surveillance. They performed both correlation and detection algorithm analyses on the time series for each city. The hospital diagnosis data showed a total of three winter respiratory/diarrhea outbreaks in children age five and under for each of the six cities; thus, there were a total of 18 outbreaks to study.

The results of the correlation analysis showed a consistently high correlation between sales of pediatric electrolytes and hospital diagnoses of respiratory and gastrointestinal illness, with a mean correlation of 0.90 (95% CI, 0.87-0.93) for the 18 outbreaks. They found that the time lag between sales and diagnoses at which the correlation was maximal was on average 1.7 weeks earlier for sales of pediatric electrolytes than hospital admissions (95% CI, 0.5-2.9).The results by outbreak ranged from an OTC lead of eight weeks to an OTC lag of one week relative to the date of hospital admission.

The results of the detection-algorithm analysis were similar. The researchers used the exponentially weighted moving average algorithm (see Chapter 14) to detect the 18 outbreaks from sales of pediatric electrolytes and from hospital admissions. Detection from pediatric electrolytes occurred on average 2.4 weeks earlier than detection from hospital diagnoses. There was, however, significant variability in the timing among cities and years. The authors used regression analysis to estimate that 25% of the variation was explained by differences in signal strength (the ratio of the signal strength of pediatric electrolytes to the signal strength of hospital diagnoses varied across winters and cities) caused by differences in the severities of outbreaks and differences in the underlying variability in how people use pediatric electrolytes and medical services in different cities and different years. They hypothesized that the remaining variability might be explained by nosocomial outbreaks of pediatric diseases, which would appear only in the hospital diagnosis data set. The organisms causing these outbreaks—rotavirus, respiratory syncytial virus, adenovirus, and influenza virus—are important causes of nosocomial infections in pediatric hospitals. They were not able to control for this factor because the hospital discharge data sets did not specify which diagnoses were nosocomial. They called for additional studies using outpatient diagnoses to control for this potential confounder.

This study identified pediatric electrolytes as an important OTC category to monitor for outbreaks in pediatric populations. This study was the first study of the information value of biosurveillance data to include a sufficient number of outbreaks to compute a confidence interval on the measurement of timeliness of detection. By studying a sufficiently large number of similar outbreaks, this study revealed that there is variability from year to year and city to city in when outbreaks may be detected, and it sounded a cautionary note on the interpretation of studies of single outbreaks.

0 0

Post a comment