Pharmaceutical sales - a method of disease surveillance?

Introduction

Disease surveillance is among the core functions of public health agencies. Although disease surveillance satisfies a variety of public health needs, the most visible function is outbreak detection. Most state and city/county epidemiology programs obtain disease reports from clinical laboratories and health care providers. Analysis of these data and direct calls from alert citizens or health care providers are often the initial indications of a disease outbreak (1, 2). The limitations of these traditional disease surveillance efforts have long been recognized and include underreporting of illnesses and delays in receiving information (3, 4).

An early indication of a diarrheal-illness problem in Milwaukee, Wisconsin, was noted by a pharmacist. On Thursday, April 1, 1993, he noted a dramatic increase in sales of over-the-counter antidiarrheal and anticramping medications. Normally, his pharmacy sold $30 a day of such medications. Starting that Thursday, drug sales increased to approximately $500 to $600 a day, or 17 to 20 times the normal sales. These increased sales continued on Friday, as a result of which, the pharmacy sold most of its supply of antidiarrheal medications. The pharmacist called the health department to inquire about excessive reports of diarrhea or intestinal illness. On Saturday, the increased sales continued, and by Sunday the pharmacist had contacted the three local television stations to report what he believed to be a major occurrence of diarrheal disease in the city. One television station covered his story, on Sunday night and by Monday morning, April 5, the extent of the outbreak that ultimately affected 400,000 people was starting to become clear. On Wednesday, April 7, the outbreak was confirmed by the Milwaukee Health Department.

The success with which drug sales indicated the Milwaukee outbreak has prompted considerable discussion about use of drug sales as a disease surveillance tool, including surveillance recommendations to several county health departments (5, 6). This study was undertaken to summarize the experiences of agencies that have attempted to use drug sales surveillance and to evaluate its potential value and limitations.

Two previous studies indicated the potential of using medication sales for public health surveillance. Collin et al. reported using gastrointestinal medication sales to evaluate the effect of water supplies on enteric illness in Meurthe-and-Moselle, France, in 1981 (7). Using drug sales together with other epidemiological information, they successfully detected 1,000 cases of enteric illness caused by contaminated water during five different outbreaks. In 1991, Maggini et al. reported using the Italian National Health Service records of drug prescriptions to estimate the prevalence of tuberculosis cases in Italy. Analysis of the information, routinely stored in a computer, showed tuberculosis to be seven times more prevalent than official notifications indicated (8).

More recently, several health departments began using drug sales for surveillance of diarrheal illnesses. The City of New York Department of Environmental Protection (NYCDEP) identified two sources of drug sales data: 1) a local distributor that serves several hundred stores in New York City, New Jersey, and Long Island and 2) a chain of small drug stores (9). After signing proprietary agreements, NYCDEP began receiving data on the sales of three different antidiarrheal medications. Sales are recorded each day and analyzed in terms of the number of units and amount of active ingredients of a particular drug sold. One major advantage of this system is that the data are donated and the expense of analysis, once the [TABULAR DATA FOR TABLE 1 OMITTED] system has been developed, is minimal. Because of the area covered, information on a large number of drug sales is available. Analysis of the data over more than a year has allowed health officials to detect sales of promotional items at given drug stores. Interestingly, no other major peaks or seasonal variations in sales have been detected.

The State of Maryland Health Department also began following sales of certain nonprescription antidiarrheal medications as a way of monitoring diarrheal-disease occurrence. They initially approached several local pharmacies and grocery stores. Two large pharmacy chains agreed to supply data on the weekly sales of Imodium[R] (loperamide) and Pepto-Bismol[R] (bismuth subsalicylate) by store and by zip code. Weekly sales data are electronically transmitted by one pharmacy chain and monthly sales data by the other. To date, no general or seasonal trends have been identified (10).

The Milwaukee Department of Health collected some information on drug sales from local retailers for less than a year after the 1993 outbreak. They discontinued their system because of difficulties in obtaining sales data on a regular basis and in compiling the detailed information on a wide range of drugs and drug preparations. Given the lack of computer systems available in local pharmacies or of software appropriate for translating the information, the data were particularly difficult to analyze (11).

Approach

Investigations of Outbreaks

To determine whether increased sales of antidiarrheal medications occurred during diarrheal-disease outbreaks other than the one in Milwaukee, state and provincial epidemiologists who investigated recent major outbreaks in five cities were contacted (Table 1). They were questioned about the availability of data on antidiarrheal medication sales and whether an increase was noted. Since minor outbreaks that affected only a small fraction of the population would not be expected to significantly increase sales of antidiarrheal medications, only municipal waterborne diarrheal-disease outbreaks were tracked.

Survey of State Health Departments for Drug Sales Surveillance

A telephone survey was initiated to determine the numbers of states that were currently using, considering using, and not attempting to use drug sales surveillance. The appropriate administrative unit of each state health department was contacted by telephone and asked three questions:

1. Do you use drug sales as a surveillance method for diarrheal disease or any other disease?

2. Do you monitor nurses' hotline calls, and

3. Do you use computerized health maintenance organization (HMO) information as a surveillance method for diarrhea or any other disease?

The answers were recorded as "yes," "no," or "considering." Responses from all 50 states were tabulated.

Developing a Statewide Surveillance System

To evaluate the availability of over-the-counter drug sales data in New Mexico and to identify the potential limitations of the information, pharmaceutical sales data were requested from retailers and wholesale distributors operating in the state of New Mexico and from national database companies, marketers, and drug manufacturers. Retailers were identified from the telephone directory and contacted by telephone. The remaining companies were identified by referral from pharmacies or wholesalers and also contacted by telephone. Sales information on Imodium[R] and its generic form, loperamide, (also sold as GNP Anti-Diarrhea, Anti-Diarrhea, and BL Anti-Diarrhea were obtained from one wholesaler. The data were reported, by area, as the number of units sold per month, and the percent of yearly sales that occurred each month was determined.

Analysis of Nurse Hotline

To compare drug sales with another indication of diarrheal illness, nurse hotline data were obtained from the Lovelace Health Hotline (LHH). LHH serves the Lovelace Health Plan, a large New Mexico HMO with approximately 150,000 members. When contacted by telephone, nurses identify symptoms based on specific approved guidelines and triage calls for immediate or next-day medical care. In addition, they advise patients about home remedies when appropriate. The records of all calls and symptoms are computerized and forwarded to the member's physician. Since the hotline is used after clinic hours and on holidays, the period of operation and number of calls varied by month. Calls with the major symptom of diarrhea were identified for adults and children (under the age of 17 years) and reported as a percent of all hotline calls received for the month.

Results

Sales of Antidiarrheal Medications During Cryptosporidiosis Outbreaks

Information on nonprescription antidiarrheal drug sales was requested from investigators of four outbreaks of cryptosporidiosis, which occurred in Las Vegas, Nevada; Collingwood, Ontario; Kelowna, British Columbia, and Cranbrook, British Columbia (Table 1). An outbreak of cryptosporidiosis in Las Vegas, Nevada, during spring 1994 was investigated by both the Clark County Health District and the Centers for Disease Control and Prevention (CDC). With the exception of several cases of cryptosporidiosis among children, all cases of diagnosed cryptosporidiosis occurred among persons who were extremely immunosuppressed. Because of the national implications of a waterborne outbreak in a community sewed by a state-of-the-art, well-managed water treatment system and because of the absence of any detectable water quality problems, an advisory panel was assembled by the American Water Works Association (6, 12). [TABULAR DATA FOR TABLE 2 OMITTED] The panel recommended tracking nonprescription, antidiarrheal drug sales for the period before and after the outbreak. Efforts were made by Las Vegas health officials to obtain drug sales information from local sources. Unfortunately, information was not available because of a lack of computerized checkout systems and because of concerns over the proprietary nature of the data. Attempts to obtain sales data from a national database were unsuccessful because only six of 84 pharmacies were included in an existing national database. These database companies will only release aggregate sales information from 16 or more pharmacies (13).

In early 1996, residents of Collingwood, a small Ontario community located on the southern shore of Georgian Bay, Lake Huron, experienced an outbreak of cryptosporidiosis. The outbreak was initially detected on March 5. This outbreak is believed to have originated from contaminated surface-derived drinking water. As part of the outbreak investigation, health department investigators requested sales data for nonprescription antidiarrheal medications from 12 area pharmacies. The intent was to validate the occurrence of an outbreak and better estimate its magnitude. Sales data were requested for two time periods: the first quarter of 1995 (one year before the outbreak) and the same period of 1996 (the time of the outbreak). Responses were received from three pharmacies (a 25 percent response rate). Two Collingwood pharmacies noted a major increase in sales in 1996, as compared with 1995 (Table 2). Pharmacy 1 submitted more detailed data. In February 1996, the total sales of antidiarrheal drugs were 26.4 times those of February 1995. In March 1996, total sales were 5.6 times those of March 1995. Based on Poisson distribution, the sales in each drug group from February to April were significantly higher (p [less than] .001) in 1996 than in 1995. For each month of 1996, the total sales of all antidiarrheal drugs were higher than in the respective month of 1995 (p [less than] .001). Pharmacy 2 reported an approximately threefold sales increase over the previous year but was not able to provide sales data. Pharmacy 3 reported that no differences in drug sales were noted between 1995 and 1996, but did not provide sales data.

In both Kelowna, British Columbia, and Cranbrook, British Columbia, outbreaks of cryptosporidiosis occurred in late spring of 1996. All pharmacies - 10 to 12 in each city - were contacted by telephone, and investigators learned that significantly more antidiarrheal drugs had been sold during the outbreaks than expected. Unfortunately, pharmacy reports of the magnitude of increased sales were not available.

TABLE 3

Drug Sales Surveillance by States

Number        States               Use of Drug Sales as
of States                          a Surveillance Method

1             Maryland             Diarrheal illness
2             Florida, Missouri    Bacterial disease
2             Kansas, Michigan     Chronic illness such as diabetes
1             Montana              HIV and AIDS

                                   Attempted Use of Drug Sales

1             Massachusetts        Gastrointestinal illness
1             Colorado             Communicable diseases
1             Arizona              Considering using drug sales
1             New Jersey           Considering using HMO data
40            All others           None

The health departments were contacted by telephone for information
about their use of drug sales for surveillance of illness. Note
that New York City's surveillance program is a municipal program
and not included in this table.

Survey of Disease Surveillance Using Drug Sales

To evaluate the current status of disease surveillance using drug sales, a telephone survey of all 50 states was conducted. The appropriate health officials were asked if they were currently using drug sales as a surveillance method or considering it for the future. The majority (38 states) responded that they were not currently using drug sales for disease surveillance and not considering using it (Table 3). In 10 states, health departments had attempted drug sales surveillance or were considering using it. Only six states had set up drug sales surveillance for studying illnesses such as HIV or infections. Only the state of Maryland was using drug sales to study diarrheal illness. Arizona was seriously considering setting up such a system. No state obtained nurse-hotline call data or health care utilization information from HMOs.

Establishing an Antidiarrheal Drug Sales Surveillance System in New Mexico

To evaluate the feasibility of establishing a drug sales surveillance program, a pilot surveillance system using drug sales was initiated in New Mexico. Five possible sources for obtaining information on drug sales were investigated:

1. manufacturers, 2. national commercial databases, 3. marketers, 4. retail stores, and 5. wholesalers (distributors).

Two manufacturers of the over-the-counter drugs, Imodium[R] and Kaopectate[R], were contacted for sales data. One company stated that sales data are proprietary and thus not released. The other manufacturer did not directly collect sales data but purchased this information from a national database company.

The two national database companies that routinely collect various types of information and provide analyses of the data were contacted for antidiarrheal-drug sales data for New Mexico. These companies purchase sales information from local pharmacies and then compile the information on a regional or national basis. They primarily serve a limited number of large markets and provide sales information to manufacturers, as well as to other purchasers. Unfortunately, smaller markets, such as Albuquerque and the state of New Mexico, are not sufficiently large for routine collection of sales information. Special geographic areas could be studied at a cost of $10,000 to $30,000 per year. If the number of participating stores is small, the analysis could be used only for internal purposes and could not be released.

Three drug marketers that track drug sales were also contacted. These companies purchase sales data from drugstores. Most of their sales data are from national chains and cover large regions. One marketer was willing to create a customized local market (area of analysis) for Albuquerque at the price of $20,000 to $30,000 per year. Part of this fee would cover the cost of paying individual pharmacies or grocery stores approximately $4,000 per year for their sales data.

Eight major local pharmacies and four independent drug chains were identified from the telephone directory out of a total of 23 independent pharmacies and six chains in the Albuquerque area. They were contacted by telephone. Although half of the stores were initially enthusiastic about the project, all of them ultimately chose not to participate in the study. The reasons differed from store to store and included the proprietary nature of drug sales, time constraints that limited the pharmacists' ability to provide data, and the inability to collect appropriate data. Even stores with electronic checkout systems did not possess software that would enable them to report sales of specific nonprescription drugs.

The source of information that proved to be both useful and accessible was a drug distributor, one of three initially contacted. This distributor serves the state of New Mexico, as well as Durango, Colorado, and El Paso, Texas, and had the software to provide monthly sales of Imodium[R] and its generic form, loperamide. Sales data from November 1995 to October 1996 were obtained at a cost of $100. The statewide sales information was categorized by the zip code of the drug store. The total number of drugs sold per month in New Mexico was proprietary and the number of drug stores per zip code was not available. The size of the container and the preparation of these two drugs were reported but counted as one unit of sale for our purposes. We assumed that most individuals would require one package of any number of capsules for a single illness. This greatly simplified analysis since the sale of an item rather than the quantity or preparation purchased was of primary interest. Sales in this case refer to sales by the distributor to the pharmacy or grocery store, rather than direct customer sales, and reflect restocking activity. Retail sales may have occurred one or two weeks before restocking activity.

The sales data were edited to remove entries from 14 zip codes. In those zip codes, sales were absent for six months or more for unknown reasons, perhaps because of store closures, infrequent purchases, or changing distributors. Ultimately, the analysis was based on 4,960 sales from 43 zip codes grouped into larger geographical areas. Monthly sales for the 12-month period showed temporal variation in each geographic area [ILLUSTRATION FOR FIGURE 1 OMITTED]. In the Albuquerque zip codes, the percentages showed a sales peak during May [ILLUSTRATION FOR FIGURE 1A OMITTED]. Sales in the central Rio Grande valley south of Albuquerque to Socorro also peaked during April to June 1996 [ILLUSTRATION FOR FIGURE 1C OMITTED]. A minor peak during the same period appears in El Paso, Texas [ILLUSTRATION FOR FIGURE 1B OMITTED], but is more difficult to distinguish because of the ascending curve. Northwestern New Mexico/Durango, Colorado, did not show a similar spring peak in sales activity [ILLUSTRATION FOR FIGURE 1C OMITTED]. In each area, a small peak in sales was seen during one of two winter months, (December or January) and most likely represents expected seasonal diarrheal diseases.

To ascertain whether the major peak in sales of antidiarrheal medications observed in Albuquerque from April to June 1996 corresponded to an increase in the occurrence of diarrhea in the community, pharmacy sales data were related to nurse hotline calls concerning diarrhea. Out of a total of 22,585 calls received for the period November 1995 to August 1996, monthly diarrhea-related calls from adults peaked in May [ILLUSTRATION FOR FIGURE 2A OMITTED]. This result is consistent with increased drug sales during May and supports the notion that an increase in diarrheal illness occurred in Albuquerque during that period. Diarrhea-related calls for pediatrics [ILLUSTRATION FOR FIGURE 2B OMITTED] showed a different pattern, perhaps because of distinctive treatment modes for children and adults. LHH recommendations in nurses' hotline guides do not include antidiarrheal medications for children, and pediatrician methods advise avoiding antidiarrheal medications for children (14).

Discussion

In addition to the Milwaukee cryptosporidiosis outbreak, two other outbreaks manifested increased sales of nonprescription antidiarrheal medications. Specific information was not available for another two outbreaks. For the Collingwood, Ontario, outbreak, the sales increase was dramatic at Pharmacy 1, with sales increasing 5.5 times over those of the previous March. This information indicates that monitoring drug sales is likely to be useful for retrospectively validating that an outbreak occurred. Results from the New Mexico surveillance systems also indicate that prospectively monitoring over-the-counter drug sales may detect important variations in diarrheal-disease occurrence. Although a number of operational problems were experienced in establishing a drug sales surveillance program in Maryland, New York, and Albuquerque, information was eventually obtained inexpensively in each area. The information in each case is readily monitored and, together with other surveillance tools such as records of nurse hotline calls and routine disease surveillance programs, can enhance the monitoring of diarrheal diseases. Temporary price reductions and the beginning of the foreign travel season may falsely indicate an increase in diarrhea activity of the population. Illnesses that do not affect a significantly large segment of the population may not be detectable because of background levels of different forms of diarrhea. Multiple surveillance systems can, however, provide a means of validating the findings. Ultimately, it is essential that public health agencies use a variety of novel means for monitoring the health of the population. Although health care providers offer one source of disease reports, it is increasingly suspected that significant outbreaks can occur in the absence of dramatic increases in health care utilization (5). Although a perfect yet inexpensive disease surveillance tool probably does not exist, the increasing availability of computerized drug sales information should be considered.

Acknowledgements

The authors appreciate the contribution of Sonia Freeman in data processing. The research described in this article has been supported by U.S. EPA through assistance agreement CR 824227 to the Lovelace Institutes.

Disclaimers

The views expressed in this article are those of the individual authors and do not necessarily reflect the views and policies of U.S. EPA. They have not been subject to the agency's peer and administrative review, and no official endorsement should be inferred. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

Corresponding author: Floyd Frost, Ph.D., The Lovelace Institutes, 2425 Ridgecrest Drive SE, Albuquerque, New Mexico 87108.

REFERENCES

1. Frost, F.J., R.L. Calderon, and G.F. Craun (1996), "Waterborne Disease Surveillance," Journal of the American Water Works Association, September:66-78.

2. Berkelman, R.L., R.T. Bryan, J.M. Hughes, J.W. LeDuc, and M.T. Osterholm (1994), "Infectious Disease Surveillance: A Crumbling Foundation," Science, 264(5157):368-370.

3. Frost, F.J., R.L. Calderon, and G.F. Craun (1995), "Waterborne Disease Surveillance: Findings of a Survey of State and Territorial Epidemiology Programs," journal of Environmental Health, 58:6-11.

4. Foster, L.R. (1990), "Surveillance for Waterborne Illness and Disease Reporting: State and Local Responsibilities," Methods for the Investigation and Prevention of Waterborne Disease Outbreaks, G.F. Craun, ed., EPA/600/1-90/005a, Washington, D.C.: U.S. EPA, pp, 39-43.

5. Craun, G.F. (Oct. 7, 1994), Report of the New York City's Advisory Panel on Waterborne Disease Assessment, New York: New York City Department of Environmental Protection.

6. Las Vegas Cryptosporidiosis Review Panel (1995), Report to the California-Nevada Section of the American Water Works Association, Southern Nevada Water Authority, Las Vegas Valley Water District, and AWWA Research Foundation, Denver, Co: AWWA.

7. Collin, J.F., J.M. Foliguet, J.J. Melet, and M. Morlot (1981), "Eau d'Adduction Publique et Gastro-Enterites en Meurthe-et-Moselle," Journal Francais d'Hydrobologie, 35:155-174.

8. Maggini, M., S. Alegiani, B. Caffari, R. Raschetti, and S. Salmaso (1991), "Epidemiological Use of Drug Prescriptions as Markers of Disease Frequency: An Italian Experience," Journal of Clinical Epidemiology, 44:1299-1307.

9. Mykol, Y., A. Ashendorff, M. Layton, J. Miller, and A. Seeley (1996), "Anti-Diarrheal Medicine Sales and Clinical Laboratories Surveillance for Gastrointestinal Illnesses Outbreak Detection," Proceedings of the Annual Meeting of the American Public Health Association, p. 242.

10. Taylor, Jean (May 8, 1997), Maryland State Department of Health, Baltimore, Maryland, personal communication.

11. Blair, Kathy (May 29, 1996), City of Milwaukee Department of Health, Milwaukee, Wisconsin, personal communication.

12. Goldstein, S.T., A.J. Bryant, S.D. Griffiths, B.L. Herwaldt, A.W. Hightower, DD. Juranek, D.G. Martin, J.L. Mesnik, O. Ravenholt, and R.R. Reich (1996), "Cryptosporidiosis: An Outbreak Associated with Drinking Water Despite State-of-the-Art Water Treatment," Annals of Internal Medicine, 124(5):459-468.

13. Shunney, Jane, R.N. (Sept. 25, 1996), Clark County Department of Health, Las Vegas, Nevada.

14. Walker-Smith, J.A., J. Richard Hamilton, and W. Allan Walker, eds. (1996) Practical Pediatric Gastroenterology, 2nd ed., Hamilton, Ontario: B.C. Decker, Inc.