pharmacoepidemiology and drug safety 2015; 24: 113–120 Published online 31 July 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/pds.3686
ORIGINAL REPORT
Antibiotic prescribing by telephone in primary care Edward Ewen1*, Vincent J. Willey2, Paul Kolm3, William F. McGhan4 and Marci Drees1 1
Department of Medicine, Value Institute, Christiana Care Health System, Newark, DE, USA HealthCore, Wilmington, DE, USA 3 Value Institute, Christiana Care Health System, Newark, DE, USA 4 University of the Sciences in Philadelphia, Philadelphia, PA, USA 2
ABSTRACT Objectives Little is known about the contribution of telephone-based prescribing on overall antibiotic utilization. The objective of this study was to determine the extent and characteristics of telephone-based antibiotic prescribing in teaching and non-teaching primary care practices. Methods This retrospective cohort study included all patients (n = 114 610) cared for by teaching and non-teaching internal medicine, pediatrics, family practice, and obstetrics/gynecology practices (n = 19) affiliated with a large US healthcare system during 2006–2010 and using a common electronic medical record. Rates and types of antibiotics prescribed by teaching and non-teaching practices via telephone contact and office visit were compared among the overall cohort. All telephone-related prescriptions during 2008 underwent chart review to determine indications for antibiotic prescribing. Results Overall, 28.9 antibiotic prescriptions were issued per 100 patient-years, with 63 418 total antibiotic prescriptions and 7876 (12.4%) generated after telephone contact. Telephone-based prescribing increased steadily from 2.2 to 4.2 per 100 patient-years during the study period. Both telephone-based and office-based antibiotic prescribing were higher in non-teaching practices. Of 1790 antibiotics prescribed by telephone during 2008, the majority were for urinary tract infection (28.3%), sinusitis (20.1%), and unspecified upper respiratory infection (URI, 15.0%). Conclusions Overall, one in every eight antibiotics was prescribed via telephone encounter. These data highlight the need to include the impact of this practice in analysis of outcomes associated with outpatient antibiotic prescribing and to incorporate telephonic prescribing into guidelines facilitating appropriate antibiotic use. Copyright © 2014 John Wiley & Sons, Ltd. key words—infectious disease; ambulatory care; physician behavior; practice variation; pharmacoepidemiology Received 7 March 2014; Revised 18 June 2014; Accepted 7 July 2014
INTRODUCTION Overuse of antibiotics in outpatient settings is well documented. Despite numerous randomized controlled trials finding no meaningful clinical benefit of antibiotic treatment for illnesses such as acute bronchitis,1–4 antibiotics are prescribed for ≥50% outpatients presenting with acute upper respiratory infections5–7 and 72–90% of patients with acute bronchitis.8–10 Despite recent decreases in overall utilization, antibiotic prescribing rates remain high11 and are often inappropriate,12 and an increasing proportion are relatively broad-spectrum.13–16 Outpatient antibiotic use has been linked to increased resistance among group A streptococci and S. pneumoniae.17–19 *Correspondence to: E. Ewen, Department of Medicine, Christiana Care Health System, 4755 Ogletown-Stanton Road, John H. Ammon Medical Education Center, Suite 2E70, Newark, DE 19718, USA. Email: [email protected]
Copyright © 2014 John Wiley & Sons, Ltd.
Antibiotic prescriptions may be generated during face-to-face or telephonic patient–provider interactions. Little is known about antibiotic prescribing by telephone because telephone calls between patients and providers are rarely billed and documentation is frequently suboptimal. Older studies of telephone contacts in internal medicine practices found 45% of calls to be symptom related, and nearly half of these were managed entirely by telephone.20–22 Up to two-thirds of telephone calls resulted in prescriptions, and antibiotics were among the most frequently prescribed new medications.23,24 A comprehensive study conducted in Norway in the late 1980s found that antibiotics comprised 26% of all new medications prescribed by “indirect contact,” which included telephone-only interactions, and one-fourth of all antibiotics were prescribed by indirect contact.25,26 Factors that may influence antibiotic prescribing behavior by either method include provider type, stage in training, and practice setting. Although guidelines for
114
e. ewen et al.
prudent use of antibiotics have been developed,27,28 adherence to and acceptance of medical guidelines are significantly less in non-teaching or private settings compared with academic settings, in a variety of medical situations.29–31 Studies demonstrate that resident physicians prescribe antibiotics less frequently than practicing physicians or non-physician clinicians,32,33 and pediatricians34 and general internists35 less frequently than other primary care providers. Increasing years from medical school graduation is also linked to high prescribing behavior.34 Healthcare reform in the USA, including adoption of emerging care models such as the patient-centered medical home and the anticipated shift from fee-for-service payment to population-based arrangements such as Accountable Care Organizations, seeks to improve care and lower costs by transforming primary care delivery. These models intend to enhance access to care and will likely result in greater non-visit-based care delivery, including telephonic patient encounters.36 This approach may however encourage medication overprescribing,37 which may be particularly true for antibiotics given the lack of simple office-based diagnostic tests to rule out infection. To what extent telephone prescribing contributes to overall antibiotic use is unknown. This study aims to examine the prevalence of and the patterns associated with telephone antibiotic prescribing in a large primary care network. PATIENTS AND METHODS Study setting and population We conducted a retrospective cohort study of antibiotic prescribing activity in 19 primary care practices affiliated with Christiana Care Health System from 1 January 2006 to 31 December 2010. These general internal medicine (n = 4), family medicine (n = 10), women’s health (n = 4), and pediatric (n = 1) practices were included on the basis of their use of a common office electronic medical record (EMR) over the study period, and included 767 unique faculty and resident physicians. Resident teaching practices, defined as practices where residents functioned as primary care providers, comprised 5 of these 19 practices. Patients with any activity in these selected practices during the study period were included in the study population. Data source Demographic information, prescribing activity, office visit, and telephone activity were obtained directly from the office EMR. All medications prescribed are entered into the EMR in a structured (codified) manner Copyright © 2014 John Wiley & Sons, Ltd.
and are linked to the clinical documents describing the interactions resulting in the prescriptions. The document type is defined by the type of provider–patient interaction (e.g., office visit, telephone call) responsible for generating the document. In this way, we were able to identify the antibiotic prescribed and precisely what type of interaction (face-to-face office visit or telephone call) led to the prescription. This study was approved by the Christiana Care Health System and the University of the Sciences institutional review boards. Assigning patient-time to a practice To calculate antibiotic prescribing rates, patient time was allocated to each practice. Unlike claims-based studies where membership duration is clearly identified as a function of insurance administration, our study was carried out from a provider perspective. In this setting, patient assignment to a practice must be inferred from patient interactions with that practice. As these interactions may be sporadic with long intervals of inactivity, we considered patients to be active within a practice if they had any interactions with that practice in the form of office visits, telephone calls, or refills of any medication within a 24-month period. Person-time accrual for a given patient within a practice began at the date of the first office visit. Patient assignment to a practice was calculated on a day-to-day basis, and patients could contribute fractions of patient-years based on the number of days they were active in that year. Patients could only be active and accrue person-time in one primary care and/or one obstetrics/gynecology practice at a time. In some cases, patients were active in both a primary care and an obstetrics/gynecology practice. In these cases, the fractional time assigned to a given practice was determined by the proportion of interactions spent with that practice for that year. For example, if a patient had 10 contacts over the year with eight internal medicine and two gynecology visits, we assigned 80% of the patient-year to the internal medicine practice and 20% to the gynecology practice. The prescriptions were always assigned to the specific practices that generated them. Each practices’ patient census was calculated as a 3-year running average, and patient data were censored on date of death, date of last activity plus 24 months, date that they became active in another practice of similar type, or 01/01/11, whichever came earliest. To determine a stable estimate of practice size, we only included practices after they had been actively using the EMR for at least 1 year to account for the lag in patient data entry at the onset of EMR implementation. Pharmacoepidemiology and Drug Safety, 2015; 24: 113–120 DOI: 10.1002/pds
115
antibiotic prescribing by telephone
Longitudinal analysis
class, infection type, and reason for antibiotic treatment (new, ongoing, or recurring infection, treatment failure, and adverse reaction to prior treatment). A randomly selected 5% sample of chart notes were reviewed by both researchers, and inter-rater reliability was calculated using Cohen’s kappa.38 The reviewers were unaware which notes were selected for inter-rater reliability testing. Hierarchical logistic regression was performed using Stata version 12; all other analyses were performed using SPSS version 15.
Antibiotic prescriptions were categorized as initiated by telephone or office visit and by antibiotic classifications, which included narrow spectrum beta-lactams (penicillin, amoxicillin, ampicillin, and first-generation cephalosporins), extended spectrum beta-lactams (amoxicillin/clavulanate, second-generation and third-generation cephalosporins), macrolides, second-generation fluoroquinolones (ciprofloxacin, norfloxacin, ofloxacin), thirdgeneration fluoroquinolones (levofloxacin, moxifloxacin), sulfonamides, tetracycline, metronidazole, and nitrofurantoin. Non-oral and refill antibiotics, as well as minocycline (because of its chronic use for acne), were excluded. Practices were categorized as teaching or non-teaching. Telephone antibiotic prescriptions were calculated by patient-year and characterized by practice type and antibiotic classification, and the temporal relationship between telephone prescribing and office visits was examined. Descriptive statistics were compared using Chi-square and Mann–Whitney U tests. Trends and comparison of trends for teaching and non-teaching practices were examined overall and for each year separately using a hierarchical logistic regression clustered by patient and weighted inversely by years active within each practice setting, with antibiotic prescriptions as the outcome variable and year, practice type, year by practice type interaction, and patient demographics as predictor variables.
RESULTS During 2006 to 2010, the EHR included 114 610 patients, representing 219 282 patient-years of followup. Family medicine practices contributed the greatest proportion of patient-years (37.1%), followed by combined internal medicine-pediatrics and family medicine (21%), internal medicine (17.9%), women’s health (18.6%), and pediatrics (5.4%). Teaching practices contributed 84 100 (34.8%) patient-years. Among the total patient population, 63 418 antibiotics were prescribed during 61 707 encounters (face-to-face or telephonic) to 31 302 individuals (28.9 antibiotics per 100 patient-years). Prescribing via the telephone accounted for 12.4% of all antibiotic prescriptions; however, 39.0% of patients with telephone-based prescriptions had an office visit within the previous 7 days. Patients who received telephone antibiotic prescriptions were older and more likely to be female and Caucasian compared with those prescribed antibiotics only at office visits (p < 0.001; Table 1).
Chart review analysis To better understand the rationale behind telephone prescribing and the types of infections treated during these interactions, researchers (E. E. or V. J. W.) performed manual review of all telephone documents resulting in antibiotic prescriptions during 2008. Each note was categorized according to antibiotic, antibiotic
Prescribing trends Overall antibiotic prescribing showed a statistically significant increase over the 5-year period (p < 0.001).
Table 1. Patient demographics by prescribing method and practice setting* Practice setting†
Age, mean (SD) Age, n (%)
Sex, n (%) Race, n (%)
≤17 years 18–45 years 46–64 years 65+ years Female White Black Hispanic Other
Prescribing method
Total n = 114 610 (100%)
Non-teaching n = 71 533 (62.4%)
Teaching n = 29 860 (26.1%)
Office only n = 27 487 (24.0%)
Phone n = 6617 (5.8%)
38.1 (20.6) 18234 (15.9) 54567 (47.6) 29291 (25.6) 12518 (10.9) 73200 (63.9) 65006 (56.7) 31554 (27.5) 4463 (3.9) 13587 (11.9)
40.1 (20.3) 10083 (14.1) 32501 (45.4) 20372 (28.5) 8577 (12) 44742 (62.5) 48721 (68.1) 13622 (19) 1939 (2.7) 7251 (10.1)
34.1 (21.4) 6974 (23.4) 13798 (46.2) 6532 (21.9) 2556 (8.6) 17206 (57.6) 10705 (35.9) 12199 (40.9) 1769 (5.9) 5187 (17.4)
36.5 (20.8) 5083 (18.5) 13008 (47.3) 6751 (24.6) 2645 (9.6) 18106 (65.9) 16444 (59.8) 7719 (28.1) 958 (3.5) 2366 (8.6)
42.5 (20.5) 697 (10.5) 3019 (45.6) 1942 (29.3) 959 (14.5) 4986 (75.4) 4713 (71.2) 1359 (20.5) 154 (2.3) 391 (5.9)
*p < 0.001 for all variables when comparing teaching versus non-teaching and office only versus phone. † 13 217 (11.5%) of patients had visits in both a teaching and non-teaching facility over the study time period (not shown).
Copyright © 2014 John Wiley & Sons, Ltd.
Pharmacoepidemiology and Drug Safety, 2015; 24: 113–120 DOI: 10.1002/pds
116
e. ewen et al.
The frequency of antibiotic prescribing, regardless of method, rose steadily from 23.6 prescriptions in 2006 to a peak of 32.5 in 2009, then leveled off in 2010 at 30.3 (all per 100 patient-years). Significant increases were observed in both telephone-based and officebased antibiotic prescribing (each p < 0.001). Antibiotic prescribing via telephone increased from 2.2 in 2006 to 4.2 in 2010, whereas office-based antibiotic prescribing rose from 21.4 in 2006 to 26.1 in 2010 (all per 100 patient-years; Figure 1). Teaching versus non-teaching practices Patients within teaching practices were younger, less likely to be female, and more likely to list their race as Black, Hispanic, or other compared with patients in non-teaching practices (p < 0.001; Table 1). The proportion and total number of patients within teaching and non-teaching practices changed during the study period, with the number of teaching patients remaining relatively stable while the number of nonteaching patients increased more than sevenfold. This reflects the later adoption of EMR technology by non-teaching practices in our health system. Teaching practices prescribed antibiotics less frequently overall than non-teaching practices (20.2 vs. 34.4 per 100 patient-years, p < 0.001). Rates of both telephonebased and office-based antibiotic prescribing were higher in non-teaching practices over the 5-year period (Figure 2), with 1.6 vs. 4.8 telephone prescriptions and 18.5 vs. 29.6 office prescriptions from teaching and non-teaching practices, respectively (all per 100 patient-years; both p < 0.001). When adjusted for patient demographics, teaching practices were less likely overall than non-teaching
Figure 2. Frequency of antibiotic prescribing by prescribing method and practice setting. The proportion and total number of patient-years contributed by teaching and non-teaching practices changed over the course of the study period. The total patient-years by setting are provided in the table below the graph. p < 0.001 for the increasing trends over time for antibiotic prescribing via the telephone or office (and for decreasing trend of office-based prescribing for teaching practices) using hierarchical logistic regression. NT, nonteaching; T, teaching
practices to prescribe antibiotics both during an office visit (adjusted OR 0.51, 95%CI 0.51–0.52) and by telephone (adjusted OR 0.32, 95%CI 0.31–0.33). An analysis of prescribing trends over the study period revealed an increase office-based prescribing in nonteaching practices, whereas it declined in teaching practices. Telephone prescribing over the study period increased in both practice settings (Table 2). Antibiotics prescribed The most commonly prescribed antibiotics using either method were macrolides and narrow spectrum beta-
Table 2.
Adjusted trends in antibiotic prescribing by practice setting Office prescribing
Practice setting Non-teaching Calendar year Sex (Female) Race (White) Teaching Calendar Year Sex (Female) Race (White)
Figure 1. Frequency of antibiotic prescribing by prescribing method. p 0.001 for the increasing trends over time for antibiotic prescribing via the telephone and office using hierarchical logistic regression
Copyright © 2014 John Wiley & Sons, Ltd.
Telephone prescribing
AOR
95%CI
AOR
95%CI
1.33 0.98 2.36
1.31–1.33 0.96–0.99 2.33–2.40
1.28 1.49 2.92
1.26–1.30 1.43–1.56 2.79–3.05
0.90 1.04 0.95
0.89–0.91 1.01–1.06 0.93–0.97
1.23 1.56 1.30
1.19–1.26 1.43–1.71 1.20–1.39
AOR, adjusted odds ratio. Office prescribing increased over the study period in non-teaching practices, whereas it declined in teaching practices. Telephone prescribing increased over time in both practice settings. Those of Caucasian race and female sex were more likely to receive antibiotics by telephone in both settings.
Pharmacoepidemiology and Drug Safety, 2015; 24: 113–120 DOI: 10.1002/pds
117
antibiotic prescribing by telephone
lactams (Table 3). Of the macrolides, 99.4% comprised azithromycin or clarithromycin, and 11.3% were generated by telephone. Both second-generation and third-generation fluoroquinolones were more frequently prescribed via the telephone compared with office-based prescribing (p < 0.001). In total, 23.9% and 16.6% of all second-generation and third-generation fluoroquinolone prescriptions, respectively, were prescribed via the telephone. In examining all fluoroquinolones together, approximately one out of every five prescriptions (20.1%) for this antibiotic class was generated via the telephone. Treatment reasons and indications for telephone prescribing All telephone notes generated in 2008 and linked to antibiotic prescriptions were reviewed (n = 1921). Twenty-three of these telephone notes were excluded due to lack of information. Among the remaining 1898 telephone notes, 47 (2.5%) were prescriptions generated as a result of administrative reasons, such as dosing changes. An additional 44 (2.3%) telephone notes were generated as a result of financial issues related to the original antibiotic, and 17 (0.9%) telephone antibiotic prescriptions were generated in response to proactive drug utilization review by the pharmacist in response to a potential drug interaction or allergy. The reasons and indications for treatment were evaluated for the remaining 1790 telephone notes. The majority of telephone antibiotic prescribing was characterized as empiric (75.6%) although 13.0% was directed by objective clinical data such as cultures or imaging results.
The notes were then further characterized by treatment reason and indication (kappa statistic for treatment reason = 0.64 and indication = 0.97). The majority (63.9%) of antibiotic telephone notes were in response to new infections, whereas an additional 13.3% related to recurring infections (Table 4). The most common indications for telephone antibiotic use were urinary tract infections, sinusitis, and various upper respiratory tract infections. The most commonly prescribed antibiotics for urinary tract infections, representing 85.1% of all telephone antibiotics for these infections, were second-generation fluoroquinolones (36.2%), sulfonamides (28.1%), and nitrofurantoin (20.8%). For sinusitis, the most commonly prescribed antibiotics were macrolides (36.7%), Table 4. Treatment reason and indication for telephone antibiotic prescribing (n = 1790) Treatment reason New infection Recurring infection Treatment failure Extending treatment Adverse reaction to prior treatment No treatment reason listed Indication Urinary tract infection Sinusitis Unspecified upper respiratory infection Bronchitis Pharyngitis Genitourinary Cellulitis Gastroenteritis/Intra-abdominal Pneumonia Lyme disease No indication listed
n (%) 1144 (63.9) 238 (13.3) 174 (9.7) 105 (5.9) 81 (4.5) 48 (2.7) n (%) 507 (28.3) 360 (20.1) 268 (15.0) 160 (8.9) 157 (8.8) 90 (5.0) 87 (4.9) 46 (2.6) 36 (2.0) 19 (1.1) 60 (3.4)
Table 3. Antibiotic classes by prescribing method Antibiotic class Beta-lactams—narrow spectrum Beta-Lactams—extended spectrum Clindamycin Macrolides Metronidazole Nitrofurantoin nd Quinolones –2 Generation Quinolones—third generation Sulfa Tetracycline Other Total
Total Rx n (%)
Office Rx n (%)
Phone Rx n (%)
p-value*
14548 (22.9) 7539 (11.9) 960 (1.5) 16268 (25.7) 4095 (6.5) 1978 (3.1) 4057 (6.4) 4322 (6.8) 6259 (9.9) 3374 (5.3) 18 (0.03) 63418
13255 (23.9) 6683 (12) 864 (1.6) 14430 (26) 3652 (6.6) 1584 (2.9) 3082 (5.5) 3605 (6.5) 5360 (9.7) 3012 (5.4) 15 (0.02) 55542
1293 (16.4) 856 (10.9) 96 (1.2) 1838 (23.3) 443 (5.6) 394 (5) 975 (12.4) 717 (9.1) 899 (11.4) 362 (4.6) 3 (0.04) 7876
ORIGINAL REPORT
Antibiotic prescribing by telephone in primary care Edward Ewen1*, Vincent J. Willey2, Paul Kolm3, William F. McGhan4 and Marci Drees1 1
Department of Medicine, Value Institute, Christiana Care Health System, Newark, DE, USA HealthCore, Wilmington, DE, USA 3 Value Institute, Christiana Care Health System, Newark, DE, USA 4 University of the Sciences in Philadelphia, Philadelphia, PA, USA 2
ABSTRACT Objectives Little is known about the contribution of telephone-based prescribing on overall antibiotic utilization. The objective of this study was to determine the extent and characteristics of telephone-based antibiotic prescribing in teaching and non-teaching primary care practices. Methods This retrospective cohort study included all patients (n = 114 610) cared for by teaching and non-teaching internal medicine, pediatrics, family practice, and obstetrics/gynecology practices (n = 19) affiliated with a large US healthcare system during 2006–2010 and using a common electronic medical record. Rates and types of antibiotics prescribed by teaching and non-teaching practices via telephone contact and office visit were compared among the overall cohort. All telephone-related prescriptions during 2008 underwent chart review to determine indications for antibiotic prescribing. Results Overall, 28.9 antibiotic prescriptions were issued per 100 patient-years, with 63 418 total antibiotic prescriptions and 7876 (12.4%) generated after telephone contact. Telephone-based prescribing increased steadily from 2.2 to 4.2 per 100 patient-years during the study period. Both telephone-based and office-based antibiotic prescribing were higher in non-teaching practices. Of 1790 antibiotics prescribed by telephone during 2008, the majority were for urinary tract infection (28.3%), sinusitis (20.1%), and unspecified upper respiratory infection (URI, 15.0%). Conclusions Overall, one in every eight antibiotics was prescribed via telephone encounter. These data highlight the need to include the impact of this practice in analysis of outcomes associated with outpatient antibiotic prescribing and to incorporate telephonic prescribing into guidelines facilitating appropriate antibiotic use. Copyright © 2014 John Wiley & Sons, Ltd. key words—infectious disease; ambulatory care; physician behavior; practice variation; pharmacoepidemiology Received 7 March 2014; Revised 18 June 2014; Accepted 7 July 2014
INTRODUCTION Overuse of antibiotics in outpatient settings is well documented. Despite numerous randomized controlled trials finding no meaningful clinical benefit of antibiotic treatment for illnesses such as acute bronchitis,1–4 antibiotics are prescribed for ≥50% outpatients presenting with acute upper respiratory infections5–7 and 72–90% of patients with acute bronchitis.8–10 Despite recent decreases in overall utilization, antibiotic prescribing rates remain high11 and are often inappropriate,12 and an increasing proportion are relatively broad-spectrum.13–16 Outpatient antibiotic use has been linked to increased resistance among group A streptococci and S. pneumoniae.17–19 *Correspondence to: E. Ewen, Department of Medicine, Christiana Care Health System, 4755 Ogletown-Stanton Road, John H. Ammon Medical Education Center, Suite 2E70, Newark, DE 19718, USA. Email: [email protected]
Copyright © 2014 John Wiley & Sons, Ltd.
Antibiotic prescriptions may be generated during face-to-face or telephonic patient–provider interactions. Little is known about antibiotic prescribing by telephone because telephone calls between patients and providers are rarely billed and documentation is frequently suboptimal. Older studies of telephone contacts in internal medicine practices found 45% of calls to be symptom related, and nearly half of these were managed entirely by telephone.20–22 Up to two-thirds of telephone calls resulted in prescriptions, and antibiotics were among the most frequently prescribed new medications.23,24 A comprehensive study conducted in Norway in the late 1980s found that antibiotics comprised 26% of all new medications prescribed by “indirect contact,” which included telephone-only interactions, and one-fourth of all antibiotics were prescribed by indirect contact.25,26 Factors that may influence antibiotic prescribing behavior by either method include provider type, stage in training, and practice setting. Although guidelines for
114
e. ewen et al.
prudent use of antibiotics have been developed,27,28 adherence to and acceptance of medical guidelines are significantly less in non-teaching or private settings compared with academic settings, in a variety of medical situations.29–31 Studies demonstrate that resident physicians prescribe antibiotics less frequently than practicing physicians or non-physician clinicians,32,33 and pediatricians34 and general internists35 less frequently than other primary care providers. Increasing years from medical school graduation is also linked to high prescribing behavior.34 Healthcare reform in the USA, including adoption of emerging care models such as the patient-centered medical home and the anticipated shift from fee-for-service payment to population-based arrangements such as Accountable Care Organizations, seeks to improve care and lower costs by transforming primary care delivery. These models intend to enhance access to care and will likely result in greater non-visit-based care delivery, including telephonic patient encounters.36 This approach may however encourage medication overprescribing,37 which may be particularly true for antibiotics given the lack of simple office-based diagnostic tests to rule out infection. To what extent telephone prescribing contributes to overall antibiotic use is unknown. This study aims to examine the prevalence of and the patterns associated with telephone antibiotic prescribing in a large primary care network. PATIENTS AND METHODS Study setting and population We conducted a retrospective cohort study of antibiotic prescribing activity in 19 primary care practices affiliated with Christiana Care Health System from 1 January 2006 to 31 December 2010. These general internal medicine (n = 4), family medicine (n = 10), women’s health (n = 4), and pediatric (n = 1) practices were included on the basis of their use of a common office electronic medical record (EMR) over the study period, and included 767 unique faculty and resident physicians. Resident teaching practices, defined as practices where residents functioned as primary care providers, comprised 5 of these 19 practices. Patients with any activity in these selected practices during the study period were included in the study population. Data source Demographic information, prescribing activity, office visit, and telephone activity were obtained directly from the office EMR. All medications prescribed are entered into the EMR in a structured (codified) manner Copyright © 2014 John Wiley & Sons, Ltd.
and are linked to the clinical documents describing the interactions resulting in the prescriptions. The document type is defined by the type of provider–patient interaction (e.g., office visit, telephone call) responsible for generating the document. In this way, we were able to identify the antibiotic prescribed and precisely what type of interaction (face-to-face office visit or telephone call) led to the prescription. This study was approved by the Christiana Care Health System and the University of the Sciences institutional review boards. Assigning patient-time to a practice To calculate antibiotic prescribing rates, patient time was allocated to each practice. Unlike claims-based studies where membership duration is clearly identified as a function of insurance administration, our study was carried out from a provider perspective. In this setting, patient assignment to a practice must be inferred from patient interactions with that practice. As these interactions may be sporadic with long intervals of inactivity, we considered patients to be active within a practice if they had any interactions with that practice in the form of office visits, telephone calls, or refills of any medication within a 24-month period. Person-time accrual for a given patient within a practice began at the date of the first office visit. Patient assignment to a practice was calculated on a day-to-day basis, and patients could contribute fractions of patient-years based on the number of days they were active in that year. Patients could only be active and accrue person-time in one primary care and/or one obstetrics/gynecology practice at a time. In some cases, patients were active in both a primary care and an obstetrics/gynecology practice. In these cases, the fractional time assigned to a given practice was determined by the proportion of interactions spent with that practice for that year. For example, if a patient had 10 contacts over the year with eight internal medicine and two gynecology visits, we assigned 80% of the patient-year to the internal medicine practice and 20% to the gynecology practice. The prescriptions were always assigned to the specific practices that generated them. Each practices’ patient census was calculated as a 3-year running average, and patient data were censored on date of death, date of last activity plus 24 months, date that they became active in another practice of similar type, or 01/01/11, whichever came earliest. To determine a stable estimate of practice size, we only included practices after they had been actively using the EMR for at least 1 year to account for the lag in patient data entry at the onset of EMR implementation. Pharmacoepidemiology and Drug Safety, 2015; 24: 113–120 DOI: 10.1002/pds
115
antibiotic prescribing by telephone
Longitudinal analysis
class, infection type, and reason for antibiotic treatment (new, ongoing, or recurring infection, treatment failure, and adverse reaction to prior treatment). A randomly selected 5% sample of chart notes were reviewed by both researchers, and inter-rater reliability was calculated using Cohen’s kappa.38 The reviewers were unaware which notes were selected for inter-rater reliability testing. Hierarchical logistic regression was performed using Stata version 12; all other analyses were performed using SPSS version 15.
Antibiotic prescriptions were categorized as initiated by telephone or office visit and by antibiotic classifications, which included narrow spectrum beta-lactams (penicillin, amoxicillin, ampicillin, and first-generation cephalosporins), extended spectrum beta-lactams (amoxicillin/clavulanate, second-generation and third-generation cephalosporins), macrolides, second-generation fluoroquinolones (ciprofloxacin, norfloxacin, ofloxacin), thirdgeneration fluoroquinolones (levofloxacin, moxifloxacin), sulfonamides, tetracycline, metronidazole, and nitrofurantoin. Non-oral and refill antibiotics, as well as minocycline (because of its chronic use for acne), were excluded. Practices were categorized as teaching or non-teaching. Telephone antibiotic prescriptions were calculated by patient-year and characterized by practice type and antibiotic classification, and the temporal relationship between telephone prescribing and office visits was examined. Descriptive statistics were compared using Chi-square and Mann–Whitney U tests. Trends and comparison of trends for teaching and non-teaching practices were examined overall and for each year separately using a hierarchical logistic regression clustered by patient and weighted inversely by years active within each practice setting, with antibiotic prescriptions as the outcome variable and year, practice type, year by practice type interaction, and patient demographics as predictor variables.
RESULTS During 2006 to 2010, the EHR included 114 610 patients, representing 219 282 patient-years of followup. Family medicine practices contributed the greatest proportion of patient-years (37.1%), followed by combined internal medicine-pediatrics and family medicine (21%), internal medicine (17.9%), women’s health (18.6%), and pediatrics (5.4%). Teaching practices contributed 84 100 (34.8%) patient-years. Among the total patient population, 63 418 antibiotics were prescribed during 61 707 encounters (face-to-face or telephonic) to 31 302 individuals (28.9 antibiotics per 100 patient-years). Prescribing via the telephone accounted for 12.4% of all antibiotic prescriptions; however, 39.0% of patients with telephone-based prescriptions had an office visit within the previous 7 days. Patients who received telephone antibiotic prescriptions were older and more likely to be female and Caucasian compared with those prescribed antibiotics only at office visits (p < 0.001; Table 1).
Chart review analysis To better understand the rationale behind telephone prescribing and the types of infections treated during these interactions, researchers (E. E. or V. J. W.) performed manual review of all telephone documents resulting in antibiotic prescriptions during 2008. Each note was categorized according to antibiotic, antibiotic
Prescribing trends Overall antibiotic prescribing showed a statistically significant increase over the 5-year period (p < 0.001).
Table 1. Patient demographics by prescribing method and practice setting* Practice setting†
Age, mean (SD) Age, n (%)
Sex, n (%) Race, n (%)
≤17 years 18–45 years 46–64 years 65+ years Female White Black Hispanic Other
Prescribing method
Total n = 114 610 (100%)
Non-teaching n = 71 533 (62.4%)
Teaching n = 29 860 (26.1%)
Office only n = 27 487 (24.0%)
Phone n = 6617 (5.8%)
38.1 (20.6) 18234 (15.9) 54567 (47.6) 29291 (25.6) 12518 (10.9) 73200 (63.9) 65006 (56.7) 31554 (27.5) 4463 (3.9) 13587 (11.9)
40.1 (20.3) 10083 (14.1) 32501 (45.4) 20372 (28.5) 8577 (12) 44742 (62.5) 48721 (68.1) 13622 (19) 1939 (2.7) 7251 (10.1)
34.1 (21.4) 6974 (23.4) 13798 (46.2) 6532 (21.9) 2556 (8.6) 17206 (57.6) 10705 (35.9) 12199 (40.9) 1769 (5.9) 5187 (17.4)
36.5 (20.8) 5083 (18.5) 13008 (47.3) 6751 (24.6) 2645 (9.6) 18106 (65.9) 16444 (59.8) 7719 (28.1) 958 (3.5) 2366 (8.6)
42.5 (20.5) 697 (10.5) 3019 (45.6) 1942 (29.3) 959 (14.5) 4986 (75.4) 4713 (71.2) 1359 (20.5) 154 (2.3) 391 (5.9)
*p < 0.001 for all variables when comparing teaching versus non-teaching and office only versus phone. † 13 217 (11.5%) of patients had visits in both a teaching and non-teaching facility over the study time period (not shown).
Copyright © 2014 John Wiley & Sons, Ltd.
Pharmacoepidemiology and Drug Safety, 2015; 24: 113–120 DOI: 10.1002/pds
116
e. ewen et al.
The frequency of antibiotic prescribing, regardless of method, rose steadily from 23.6 prescriptions in 2006 to a peak of 32.5 in 2009, then leveled off in 2010 at 30.3 (all per 100 patient-years). Significant increases were observed in both telephone-based and officebased antibiotic prescribing (each p < 0.001). Antibiotic prescribing via telephone increased from 2.2 in 2006 to 4.2 in 2010, whereas office-based antibiotic prescribing rose from 21.4 in 2006 to 26.1 in 2010 (all per 100 patient-years; Figure 1). Teaching versus non-teaching practices Patients within teaching practices were younger, less likely to be female, and more likely to list their race as Black, Hispanic, or other compared with patients in non-teaching practices (p < 0.001; Table 1). The proportion and total number of patients within teaching and non-teaching practices changed during the study period, with the number of teaching patients remaining relatively stable while the number of nonteaching patients increased more than sevenfold. This reflects the later adoption of EMR technology by non-teaching practices in our health system. Teaching practices prescribed antibiotics less frequently overall than non-teaching practices (20.2 vs. 34.4 per 100 patient-years, p < 0.001). Rates of both telephonebased and office-based antibiotic prescribing were higher in non-teaching practices over the 5-year period (Figure 2), with 1.6 vs. 4.8 telephone prescriptions and 18.5 vs. 29.6 office prescriptions from teaching and non-teaching practices, respectively (all per 100 patient-years; both p < 0.001). When adjusted for patient demographics, teaching practices were less likely overall than non-teaching
Figure 2. Frequency of antibiotic prescribing by prescribing method and practice setting. The proportion and total number of patient-years contributed by teaching and non-teaching practices changed over the course of the study period. The total patient-years by setting are provided in the table below the graph. p < 0.001 for the increasing trends over time for antibiotic prescribing via the telephone or office (and for decreasing trend of office-based prescribing for teaching practices) using hierarchical logistic regression. NT, nonteaching; T, teaching
practices to prescribe antibiotics both during an office visit (adjusted OR 0.51, 95%CI 0.51–0.52) and by telephone (adjusted OR 0.32, 95%CI 0.31–0.33). An analysis of prescribing trends over the study period revealed an increase office-based prescribing in nonteaching practices, whereas it declined in teaching practices. Telephone prescribing over the study period increased in both practice settings (Table 2). Antibiotics prescribed The most commonly prescribed antibiotics using either method were macrolides and narrow spectrum beta-
Table 2.
Adjusted trends in antibiotic prescribing by practice setting Office prescribing
Practice setting Non-teaching Calendar year Sex (Female) Race (White) Teaching Calendar Year Sex (Female) Race (White)
Figure 1. Frequency of antibiotic prescribing by prescribing method. p 0.001 for the increasing trends over time for antibiotic prescribing via the telephone and office using hierarchical logistic regression
Copyright © 2014 John Wiley & Sons, Ltd.
Telephone prescribing
AOR
95%CI
AOR
95%CI
1.33 0.98 2.36
1.31–1.33 0.96–0.99 2.33–2.40
1.28 1.49 2.92
1.26–1.30 1.43–1.56 2.79–3.05
0.90 1.04 0.95
0.89–0.91 1.01–1.06 0.93–0.97
1.23 1.56 1.30
1.19–1.26 1.43–1.71 1.20–1.39
AOR, adjusted odds ratio. Office prescribing increased over the study period in non-teaching practices, whereas it declined in teaching practices. Telephone prescribing increased over time in both practice settings. Those of Caucasian race and female sex were more likely to receive antibiotics by telephone in both settings.
Pharmacoepidemiology and Drug Safety, 2015; 24: 113–120 DOI: 10.1002/pds
117
antibiotic prescribing by telephone
lactams (Table 3). Of the macrolides, 99.4% comprised azithromycin or clarithromycin, and 11.3% were generated by telephone. Both second-generation and third-generation fluoroquinolones were more frequently prescribed via the telephone compared with office-based prescribing (p < 0.001). In total, 23.9% and 16.6% of all second-generation and third-generation fluoroquinolone prescriptions, respectively, were prescribed via the telephone. In examining all fluoroquinolones together, approximately one out of every five prescriptions (20.1%) for this antibiotic class was generated via the telephone. Treatment reasons and indications for telephone prescribing All telephone notes generated in 2008 and linked to antibiotic prescriptions were reviewed (n = 1921). Twenty-three of these telephone notes were excluded due to lack of information. Among the remaining 1898 telephone notes, 47 (2.5%) were prescriptions generated as a result of administrative reasons, such as dosing changes. An additional 44 (2.3%) telephone notes were generated as a result of financial issues related to the original antibiotic, and 17 (0.9%) telephone antibiotic prescriptions were generated in response to proactive drug utilization review by the pharmacist in response to a potential drug interaction or allergy. The reasons and indications for treatment were evaluated for the remaining 1790 telephone notes. The majority of telephone antibiotic prescribing was characterized as empiric (75.6%) although 13.0% was directed by objective clinical data such as cultures or imaging results.
The notes were then further characterized by treatment reason and indication (kappa statistic for treatment reason = 0.64 and indication = 0.97). The majority (63.9%) of antibiotic telephone notes were in response to new infections, whereas an additional 13.3% related to recurring infections (Table 4). The most common indications for telephone antibiotic use were urinary tract infections, sinusitis, and various upper respiratory tract infections. The most commonly prescribed antibiotics for urinary tract infections, representing 85.1% of all telephone antibiotics for these infections, were second-generation fluoroquinolones (36.2%), sulfonamides (28.1%), and nitrofurantoin (20.8%). For sinusitis, the most commonly prescribed antibiotics were macrolides (36.7%), Table 4. Treatment reason and indication for telephone antibiotic prescribing (n = 1790) Treatment reason New infection Recurring infection Treatment failure Extending treatment Adverse reaction to prior treatment No treatment reason listed Indication Urinary tract infection Sinusitis Unspecified upper respiratory infection Bronchitis Pharyngitis Genitourinary Cellulitis Gastroenteritis/Intra-abdominal Pneumonia Lyme disease No indication listed
n (%) 1144 (63.9) 238 (13.3) 174 (9.7) 105 (5.9) 81 (4.5) 48 (2.7) n (%) 507 (28.3) 360 (20.1) 268 (15.0) 160 (8.9) 157 (8.8) 90 (5.0) 87 (4.9) 46 (2.6) 36 (2.0) 19 (1.1) 60 (3.4)
Table 3. Antibiotic classes by prescribing method Antibiotic class Beta-lactams—narrow spectrum Beta-Lactams—extended spectrum Clindamycin Macrolides Metronidazole Nitrofurantoin nd Quinolones –2 Generation Quinolones—third generation Sulfa Tetracycline Other Total
Total Rx n (%)
Office Rx n (%)
Phone Rx n (%)
p-value*
14548 (22.9) 7539 (11.9) 960 (1.5) 16268 (25.7) 4095 (6.5) 1978 (3.1) 4057 (6.4) 4322 (6.8) 6259 (9.9) 3374 (5.3) 18 (0.03) 63418
13255 (23.9) 6683 (12) 864 (1.6) 14430 (26) 3652 (6.6) 1584 (2.9) 3082 (5.5) 3605 (6.5) 5360 (9.7) 3012 (5.4) 15 (0.02) 55542
1293 (16.4) 856 (10.9) 96 (1.2) 1838 (23.3) 443 (5.6) 394 (5) 975 (12.4) 717 (9.1) 899 (11.4) 362 (4.6) 3 (0.04) 7876
