Telephone calling patterns could provide a very early indication of a change in health status of a population. For example, an increase in telephone call volume emanating from a given neighborhood or apartment building—especially in the very early hours of the morning or late at night when levels are typically low—may indicate an emerging infection within that location. An increase in calls from one region of a city or building to medical clinics or pharmacies might also provide an early signal that requires further investigation.
Li and colleagues used correlation analysis to study the relationship between daily telephone call volume from the Watson Research Center to medical facilities and pharmacies and the same reference time series of respiratory illness as described in the previous section (Li and Aggarwal, 2003, Wagner et al., 2004). They compiled a list of approximately 5000 phone numbers that corresponded to medical offices in the area surrounding the Watson Research Center. They obtained these numbers from various sources including both online and hardcopy phone directories. The Watson Research Center tracks outgoing phone calls for billing purposes. They obtained a daily report of the total number of outgoing calls to the numbers on the compiled list of medical offices, as well as the number of callers (extensions) making those calls.
They studied call daily volumes during the period October 2001 through May 2002. The analysis ignored weekends, as call volumes dropped dramatically on non-working days. The maximum correlation between the call volume and disease activity was 0.72, which occurred when telephone calls led office visits by four days.
There have been no other studies of volumetric telephone usage. In our attempts to obtain volumetric data from phone companies, we have encountered concerns that even highly de-identified monitoring may be perceived as a "big brother'' invasion of privacy by phone company subscribers. It would perhaps be easier to enlist the phone companies' participation in such monitoring if there were additional studies showing ability to detect outbreaks, but without their participation, it will be difficult to further develop the scientific evidence. The most promising setting for future studies will likely be that of permissive environments (see Section 5 of this chapter).
Was this article helpful?