Average daily kept office visits, September 11 to December 31, 2001, compared with 2000 and 2002. Asterisk indicates P<.05 for comparison by 2-tailed t test (September 2001 vs September 2002); dagger, P<.01 for comparison by 2-tailed t test (October-December 2001 vs October-December 2002).
Weekly patient contacts and prescriptions, September 11 to December 31, 2001.
Hupert N, Chege W, Bearman GML, Pelzman FN. Antibiotics for AnthraxPatient Requests and Physician Prescribing Practices During the 2001 New York City Attacks. Arch Intern Med. 2004;164(18):2012-2016. doi:10.1001/archinte.164.18.2012
Copyright 2004 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2004
Little is known about patient encounters with primary care physicians and prescribing practices during the 2001 US anthrax attacks.
We retrospectively reviewed the electronic medical record of outpatient telephone and clinic visits at a large primary care practice in New York City from September 11 to December 31, 2001, to identify physician- and patient-related factors that were associated with prescribing antibiotics for anthrax prophylaxis.
Average daily patient volume from October to December was higher in 2001 (221.2 patients per day) compared with 2000 (199.1; P<.01) and 2002 (215.8; P = .14). Patient-initiated discussion about anthrax or smallpox were involved in 244 patient contacts with 63 physicians, including 92 (0.6%) of 14917 telephone contacts and 152 (1.0%) of 15 539 office visits. Fifty patients (21%) requested antibiotics or vaccines and 52 (22%) received antibiotics: 39 received ciprofloxacin; 12, doxycycline; and 1, both drugs. Independent predictors of receiving anthrax prophylaxis included requesting medication (odds ratio [OR], 8.1; 95% confidence interval [CI], 3.5-18.6), reporting powder or workplace exposure (OR, 4.5; 95% CI, 2.1-10.0), having an abnormal physical examination finding (OR, 3.9; 95% CI, 1.4-11.0), and being asymptomatic (reporting any illness symptoms was associated with an OR of 0.3 [95% CI, 0.1-0.6]).
Primary care physicians played an important and heretofore underdocumented role in responding to the 2001 anthrax attacks. Prescription of prophylactic antibiotics for anthrax was uncommon and appears to have been selective among concerned patients. These results highlight the importance of including primary care physicians in community-wide bioterrorism response planning.
The official public health response to the 2001 US anthrax attacks has been well documented at the federal, state, and local levels, but little is known about ambulatory patient encounters with primary care physicians during and after the attacks.1- 6 Nationwide, more than 10 000 affected workers and others were given 3.75 million prophylactic antibiotic tablets through official dispensing campaigns from October 2001 through January 2002, but media reports at the time suggested even more widespread prescribing of antibiotics by individual physicians.7- 14 However, to our knowledge, there has been no systematic study of the actions of primary care physicians in response to the anthrax attacks, leaving their contributions unclear and their proper role in any future bioterrorism response plans uncertain.15
To explore these issues, we examined the clinical experience of a large academic general internal medicine practice in New York City in the months following the September 11 terrorist attacks on the World Trade Center and the subsequent anthrax attacks in October 2001. Using a keyword-searchable outpatient electronic medical record (EMR), we identified all bioterrorism-related patient contacts after September 11, 2001. Our goals were to describe bioterrorism-related patient encounters, to quantify requests and prescriptions for prophylactic antibiotics for anthrax, and to determine patient-related factors associated with receiving antibiotics.
The study site is an academic general internal medicine practice located on the Upper East Side of Manhattan in New York City. Patient care is provided to more than 21 500 unique patients per year (more than 56 000 patient visits) by 28 full-time attending physicians and approximately 120 internal medicine residents. The clinic uses a proprietary keyword-searchable EMR for records such as clinic notes (both office visits and telephone contacts) and medication prescriptions. At the time of the study, there were approximately 125 000 unique patient records in this EMR.
The study period was defined from September 11 to December 31, 2001. We obtained the daily census of kept visits during this time for the index year (2001) as well as for 2000 and 2002 for comparison. To identify encounters involving patient-initiated discussion and concerns about bioterrorism, we performed a record review of telephone contacts and office visits during the study period using keyword searches for both "anthrax" and "smallpox." Charts were abstracted if they contained documented discussion of either disease (eg, "Patient concerned about potential anthrax exposure . . . "). Twenty-four records that included only brief physician-initiated comments, such as "Denies concern about anthrax," were identified by the keyword search but were excluded from further review. Disagreements about data elements were resolved by re-review and consensus. We collected information on type of physician consulted, patient demographics, potential exposure to anthrax or other biologic agents, symptoms, physical examination findings, laboratory testing for anthrax, and requests for and prescriptions of antibiotics or vaccines. We also identified all new medication records for ciprofloxacin, doxycycline, and amoxicillin prescriptions (both generic and tradenames) written in the clinic during the study period. Patients who had been prescribed any of these antibiotics were then cross-referenced with the list of patients with a documented discussion of bioterrorism in the clinic encouter.
The 2-tailed t test was used to analyze differences in daily clinic volume for each of the 4 study months between the index year and the preceeding and subsequent years and differences in patient age. Univariate associations of categorical variables with the outcome of recorded prescription of an antibiotic were analyzed using the χ2 test. To determine independent predictors of receiving antibiotics, we developed a multivariable logistic regression model by manually adding all variables that were associated with receipt of a prescription at P ≤ .10, retaining variables that remained significantly associated with the outcome of interest at P<.05 or that were confounders of other variables (causing >10% change in the point estimate of the odds ratio [OR]). SAS version 8.0 software (SAS Institute, Cary, NC) was used for all statistical calculations. This study was approved by the institutional review board of Weill Medical College and New York Presbyterian Hospital, New York City.
Daily clinic patient volume from October to December was higher in 2001 compared with 2000 (mean, 221.2 patients per day vs 199.1; P<.01) and 2002 (221.2 vs 215.8; P = .14) (Figure 1). In contrast, there was a significant drop off in patient visits in September 2001 directly following the attacks on the World Trade Center compared with the preceding and subsequent years (mean, 151.6 patients per day in 2001 vs 189.2 patients per day in 2000 [P = .09] and 203.6 patients per day in 2002 [P = .01]).
Of 30 456 total patient contacts between September 11 and December 21, 2001, 244 involved patient-initiated discussion about bioterrorism: 92 (0.6%) of 14917 telephone contacts and 152 (1.0%) of 15 539 office visits. Figure 2 shows the timeline of patient contacts and antibiotic prescriptions for the 244 study encounters. Also shown is the timing of 3 relevant events in the unfolding anthrax attacks: recognition of the first confirmed case in Florida on October 4, recognition of the first (cutaneous) case in New York City on October 12, and report of the single inhalational anthrax case in New York City on October 29. The first telephone contact occurred on September 24 and the first office encounter on September 25, over 1 week before the first case in Florida. The last office visit related to anthrax or smallpox was December 17, and the last telephone call was December 20. Peak bioterrorism-related contacts at the study site occurred during the week following the discovery of the single inhalational anthrax case in New York City.
Three patients discussed bioterrorism both in the office and on the telephone, for a final study sample size of 241 unique patients (Table 1). Women comprised 167 patients (69%), and the mean age was 43.7 years (range, 20-83 years). Sixty-three physicians were involved in these encounters, with a median of 2 patients per physician during the study period (mean, 3.75 patients per physician [range, 1-32]). Attending physicians cared for 166 (69%) of the study patients, whereas both attending and resident physicians interacted with 70 patients (29%); physician type was not available for 5 patients (2%).
Potential workplace or incidental exposure to anthrax was reported by 97 people (40% of all cases, including 25 postal workers); 110 patients (46%) had symptoms including sore throat (73% of symptomatic patients), cough (45%), subjective fever (40%), dyspnea/chest pain (16%), and nausea/vomiting (15%). Potentially exposed patients, defined by reported exposure to white powder or working in a mailroom, were not more likely to be symptomatic than unexposed patients (44 [45%] of 97 patients vs 66 [46%] of 144 patients). Thirty (12% of the total sample and 27% of all symptomatic patients) of these 110 symptomatic patients had abnormal results on physical examination, including 7 who had an abnormal lung examination results (3% of the total sample and 6% of symptomatic patients). Physical examination abnormalities were more likely to be present among symptomatic patients (20 [18%] of 110) compared with asymptomatic patients (10 [8%] of 131; P = .02). Twenty seven patients (11%) had laboratory blood analyses, and 59 (24%) had nasal swabs; all laboratory test results were within normal limits.
Fifty patients (21%) were recorded as specifically requesting antibiotics or vaccines, of whom 8 (3% of all patients) had been told to seek treatment by employers or health authorities. Patients who requested medications were less likely to have symptoms (10 [20%] of 50 vs 100 [52%] of 191; P<.001) and were no more likely to have reported potential exposure than were those who did not request medications. Physicians documented prescribing an antibiotic for 52 patients (22%): 39 received ciprofloxacin; 12, doxycycline; and 1, both drugs. Half of the patients who requested antibiotics (25 patients) received 1 of these prescriptions; the other 27 were given to patients who were not documented as requesting prophylactic antibiotics. Although telephone contacts were less likely than office contacts to report an exposure (P<.001) or abnormal symptoms (P<.001), patients who called in were significantly more likely to request medications (29 requests [32%] of 92) than were patients who came to the office (21 [14%] of 149; P = .02).
To verify the accuracy of physician documentation of antibiotic prescriptions, we searched for all computer order-entry records of new ciprofloxacin, doxycycline, or amoxicillin prescriptions during the study period. This yielded a total of 224 new prescriptions for ciprofloxin, 53 for doxycycline, and 92 for amoxicillin. All 52 patients who received antibiotics in our study sample were included among these groups, although the breakdown differed slightly from the totals obtained from clinical notes (35 receiving ciprofloxacin, 12 receiving doxycycline, 4 receiving both, and 1 receiving amoxicillin). As shown in Figure 2, all but 1 of the prescriptions were given out between the weeks of September 24 and November 5.
The last step in our analysis was to construct a multivariable logistic regression model to determine independent predictors of the clinical encounter ending in the prescription of prophylactic antibiotics for anthrax. Table 2 shows the final regression model, which had a c-statistic of 0.8. Three factors independently increased the probability that a patient would receive an antibiotic prescription: documentation that the patient requested antibiotics (OR, 8.1; 95% confidence interval [CI], 3.5-18.6), self-reported exposure to anthrax powder or possible workplace contamination (OR, 4.5; 95% CI, 2.1-10.0), and documentation of any abnormality on physical examination (OR, 3.9; 95% CI, 1.4-11.0). Report of symptoms, on the other hand, was significantly associated with a decreased likelihood of receiving antibiotics (OR, 0.3; 95% CI, 0.1-0.6). Factors that were considered in preliminary models but were not significant predictors nor confounders of the final model included older age, sex, insurance coverage, type of contact (ie, telephone vs office), being told to seek treatment, and type of physician consulted (ie, attending vs resident).
Despite widespread popular concern about bioterrorism and speculation about increased patient requests and physician prescribing of antibiotics for anthrax prophylaxis, only 1 in 5 patients who initiated discussion about anthrax or smallpox with physicians at this internal medicine practice in the wake of the 2001 terrorist attacks either requested antibiotics or received them. Patient encounters prompted by concerns about bioterrorism preceded the discovery of the first case of inhalational anthrax by over 1 week and persisted until mid-December, but they nevertheless made up only a small fraction of all patient visits in the study period. The timing of these encounters correlated with revelations about cases in the New York City metropolitan area.16
While we cannot comment on the New York City population as a whole, our results do not suggest widespread antibiotic abuse in the aftermath of the 2001 terrorist attacks, as measured by patient requests and physician prescribing in this academic outpatient practice. Prescription of antibiotics appropriate for anthrax prophylaxis (ie, ciprofloxacin, doxycycline, or amoxicillin) was most highly associated with patient requests, followed by report of potential exposure and abnormal findings on physical examination.17 Given the clinical and scientific uncertainty that accompanied the unfolding 2001 anthrax attacks, each of these factors can be considered at least reasonable grounds for therapeutic action. The study sample included a large number of postal workers and office workers from Midtown Manhattan, where documented cutaneous anthrax cases had been diagnosed by mid-October. Acceding to requests by these patients for prophylactic antibiotics may have appeared to be a low-risk, high-potential benefit clinical decision, especially if there was report of potential exposure. Lacking familiarity with the clinical presentation of inhalational anthrax, clinicians may have taken any abnormal finding on physical examination as sufficient impetus to err on the side of caution in these cases.18
However, this reasoning does not help to explain why patients reporting symptoms were less likely than others to receive antibiotics. The most compelling explanation for this seemingly counterintuitive result is that these patients' clinical presentations may have provided their physicians with sufficient grounds to make alternate diagnoses (eg, viral upper respiratory tract infection or bacterial sinusitis). While telephone contact encounters were more likely to involve requests for antibiotics and less likely to involve discussion of abnormal symptoms (thus potentially raising the probability that these patients would receive prescriptions), type of patient-physician encounter alone was not a significant predictor of whether the patient received prophylactic antibiotics.
Clinicians and public health officials responding to the 2001 attacks found themselves on new and uncertain ground in the diagnosis, prophylaxis, and treatment of inhalational and cutaneous anthrax.17- 21 Viewed in light of changing guidelines for screening and uncertainty about the clinical presentation or appropriate management of anthrax exposure during this time, the clinicians we studied appeared to have notable therapeutic restraint (eg, in the 28% of telephoned requests for antibiotics that were denied). The lower rate of prescribing for symptomatic patients suggests that physicians may have used clinical judgment (ie, not just reflex responses) in making treatment decisions in this setting, strengthening calls for increased clinical training on the recognition and management of victims of bioterrorism for general practitioners.
One limitation of this study pertains to its limited scope. We examined the postattack experience of only 1 academic practice, located in close proximity to media and health care facilities where cutaneous and inhalational anthrax cases occurred in 2001. Patient selection bias due to this proximity may cause our results to overstate prescribing in response to bioterrorism generally. Our findings do, however, provide a window onto the actual experience of an outpatient clinic and the response by primary care physicians to an ongoing bioterrorist attack.
An additional limitation is the reliance on clinic notes recorded in an EMR by practicing physicians and on automated prescription records. It is possible that some physicians used traditional written presciption pads for which there is no EMR record. This could bias our results toward underestimating the true rate of antibiotic dispensing in the aftermath of the attacks. Surveys of physicians and patients involved in the 2001 attacks may provide more insight into these matters.
To our knowledge, this study is the first systematic analysis of the primary care response to the 2001 anthrax attacks. Our results show that local primary care physicians who were not part of organized disaster response activities provided care to concerned and potentially exposed patients. While most patients did not request antibiotics, many of those who did appear to have received them for reasons that seem at least logical in hindsight. These findings may provide some reassurance to public health planners that both the public and the general medical community is capable of responding in a reasonable and pharmacotherapeutically measured fashion to future bioterrorist events, especially if official mass prophylaxis activities are carried out in a prompt fashion as was the case in New York City.22 These findings support greater inclusion of primary care physicians in comprehensive planning for the public health and medical response to such terrorist events. Given estimates of the tremendous personnel requirements for rapid large-scale mass prophylaxis campaigns to counter bioterrorism, effective mobilization of primary care practitioners may prove essential to their ultimate success.23
Correspondence: Nathaniel Hupert, MD, MPH, Assistant Professor of Public Health and Medicine, Weill Medical College of Cornell University, 411 E 69th St, New York, NY 10021 (email@example.com).
Accepted for publication February 10, 2004.
Dr Hupert was supported by contract No. 290-00-0013 from the Department of Health and Human Services, Agency for Healthcare Research and Quality, Rockville, Md, during this study.
We thank Erica Goodwin for her assistance with data collection.