Point-of-Care Testing as an Influenza Surveillance Tool: Methodology and Lessons Learned from Implementation

Abstract

Objectives. Disease surveillance combines data collection and analysis with dissemination of findings to decision makers. The timeliness of these activities affects the ability to implement preventive measures. Influenza surveillance has traditionally been hampered by delays in both data collection and dissemination. Methods. We used statistical process control (SPC) to evaluate the daily percentage of outpatient visits with a positive point-of-care (POC) influenza test in the University of Utah Primary Care Research Network. Results. Retrospectively, POC testing generated an alert in each of 4 seasons (2004-2008, median 16 days before epidemic onset), suggesting that email notification of clinicians would be 9 days earlier than surveillance alerts posted to the Utah Department of Health website. In the 2008-09 season, the algorithm generated a real-time alert 19 days before epidemic onset. Clinicians in 4 intervention clinics received email notification of the alert within 4 days. Compared with clinicians in 6 control clinics, intervention clinicians were 40% more likely to perform rapid testing (P = 0.105) and twice as likely to vaccinate for seasonal influenza (P = 0.104) after notification. Conclusions. Email notification of SPC-generated alerts provided significantly earlier notification of the epidemic onset than traditional surveillance. Clinician preventive behavior was not significantly different in intervention clinics.

Figure 1

Full-season charting of ILI cases and positive rapid influenza tests using statistical process control* of daily proportions in the UUPCRN population for the 2007-08 influenza season. *Solid horizontal lines define the center line, and dashed lines define the upper control limit. Letters on the chart denote particular events: “A” is 1 point above 3 sigma; “C” is 9 points in a row above the center line; “D” is 9 points in a row below the center line; “E” is 6 points in a row that are increasing (or decreasing); “F” is 14 points in a row alternating up and down. ^†7-day smoothed graphs have 2 gaps of 7 days each, corresponding to single day clinic closures for Thanksgiving (Nov 22) and Christmas (Dec 25). ^‡Early alert (Jan 19) was the date of the modeled real-time surveillance alert for positive rapid tests; epidemic onset (Feb 3) was the date when 4 of 5 consecutive days exceeded 1 sigma for positive rapid tests.

Figure 2

Statistical process control chart* of modeled real-time surveillance of positive influenza rapid tests demonstrating an early alert in the UUPCRN population for the 2007-08 season. *Solid horizontal line defines the center line. Dashes above each data point represent the upper control limit calculated for that day. Points exceeding the upper control limit are noted by “A.” The modeled real-time surveillance alert (Jan 10) was defined as the date for the data point following an exceedance of the upper control limit (A) that does not subsequently return to the center line.

Figure 3

Distribution of notification dates* of increased influenza activity relative to the UUPCRN epidemic onset by notification source^† for seasons 2004-05 through 2007-08. *Notification dates were based on the date that Sentinel Physician ILI reports were posted to the CDC and UDOH websites and the theoretical notification by email to UUPCRN clinicians 2 days following the early alert signal. ^†Notification source: CDC—Centers for Disease Control and Prevention. The national ILI threshold ranged from 2.1% to 2.5% during the study years. UDOH—Utah Department of Health. The historical Utah-specific ILI threshold was 3.6% during the study years and is shown as the first UDOH group in the graph. The threshold was recalculated in 2008-09 to 1.7% and is shown as the second UDOH group. The 1.7% threshold demonstrates the shift in notification date if this threshold had been used in the study years. UUPCRN—University of Utah Primary Care Research Network. The notification date is based on a theoretical email notification sent to clinicians 2 days after the modeled real-time surveillance alert.

Figure 4

Statistical process control chart* of the percent of visits with a positive rapid influenza test in UUPCRN clinics during the 2008-09 season. *Circled portion of the graph represents part (Dec 25–Jan 27) of the prealert period and includes the early alert signal. The signal occurred on Jan 26 (data point that did not return to baseline following a 3 sigma exceedance), with the data download on Jan 27, making the latter date the signal available to the analyst. Symbols used in the graph are A—designates a data point exceeded the upper control limit (3 sigma), C—designates the 9th data point in a row above the center line, and D—designates the 9th data point in a row below the center line.

Figure 5

The percentage of influenza-associated outpatient visits with particular clinician behaviors in 2008-09, by intervention status for the prealert and early alert time periods*. *The odds ratio (OR) and associated P value are reported for the interaction term (intervention status and time period) in the logistic regression model.