Abstract
Objective
To investigate the scale of antimicrobial prescribing without a corresponding visit, and to compare the attributes of patients who received antimicrobials with a corresponding visit to those who did not have a visit.
Design
Retrospective cohort
Methods
We followed 185,010 Medicare patients for one year after an acute myocardial infarction. For each antimicrobial prescribed, we determined if the patient had an inpatient, outpatient or provider claim in the 7 days prior to the antimicrobial prescription being filled. We compared the proportions of patient characteristics for those prescriptions associated with a visit and without a visit (i.e., phantom prescriptions). We also compared the rates at which different antimicrobials were prescribed without a visit.
Results
We found that of 356,545 antimicrobial prescriptions, 14.75% had no evidence of a visit in the week prior to the prescription being filled. A higher percentage of patients without a visit were identified as white (p<0.001) and female (p<0.001). Patients without a visit had a higher likelihood of survival and fewer additional cardiac events (AMI, cardiac arrest, stroke, all p<0.001). Among the antimicrobials considered, amoxicillin, penicillin, and agents containing trimethoprim and methenamine were much more likely to be prescribed without a visit. In contrast, levofloxacin, metronidazole, moxifoxacin, vancomycin, and cefdinir were much less likely to be prescribed without a visit.
Conclusions
Among this cohort of patients with chronic conditions, phantom prescriptions of antimicrobials are relatively common and occurred more frequently among those patients who were relatively healthy.
Introduction
The emergence of antimicrobial resistance is a major public health concern and has major clinical implications.1–4 One approach for controlling antimicrobial resistance is to minimize inappropriate antimicrobial use.5–8 To achieve this goal, stewardship campaigns and programs have been established in both inpatient and outpatient settings.6,9–13 Stewardship programs have included a variety of strategies.9,10,14–16 In inpatient settings, programs have incorporated, for example, monitoring and feedback and prescriber education.9,17,18 In outpatient settings, programs have used feedback to prescribers, delayed prescribing, and attempts to educate patients,10,19–22 because patient and family expectations are thought to be major drivers of inappropriate antimicrobial use.23–25
Regardless of the intervention proposed, most stewardship approaches are designed to be implemented in traditional practice settings. However, not all antimicrobial prescribing is based upon office visits; prescribing may occur without a direct, face-to-face clinical encounter, e.g., by telephone. Thus, for non-face-to-face (i.e. phantom) encounters, standard approaches to improve antimicrobial prescribing may not be effective. Given the interest in reducing antimicrobial use across the spectrum of care, the purpose of this project is to determine the frequency of antimicrobial prescribing in a large cohort of patients with comorbidities, describe the types of antimicrobials prescribed to outpatients via phantom encounters, and to compare patient characteristics in antimicrobial prescribing between phantom and face-to-face healthcare encounters.
Methods
Study Cohort
To study outpatient antimicrobial prescribing among patients with chronic diseases, we used a cohort of Medicare recipients with a prior history of a myocardial infarction for whom we have complete claims data. Our cohort consists of 185,010 Medicare recipients who suffered an acute myocardial infarction (AMI) in 2007 or 2008 (an inpatient stay with primary diagnosis code 410.x1). In the analyses, we utilized Part D prescription drug events from the Chronic Condition Data Warehouse (www.ccwdata.org) and claims and enrollment information from Medicare. Each patient’s index date was considered to be the date of hospital discharge for the visit corresponding to the AMI. Exclusion criteria for this study were: having had an AMI in the year before the index stay; not being enrolled in Medicare Parts A and B for a year before the index stay and either a year after the index stay or until their death; and not being enrolled in Medicare Part D during the 6 months before and the year after the index stay. The University of Iowa’s institutional review board approved this study.
Defining a Phantom Prescription
In the year following each patient’s index date, we recorded each instance a patient received an antimicrobial of interest (Appendix 1). We included the most commonly prescribed oral antimicrobials. If a patient filled a prescription for an antimicrobial of interest and also visited a medical professional in the 7 days preceding the date on which the prescription was filled, we labeled the prescription as an antimicrobial prescription with a visit. In contrast, if a prescription was filled but there was no visit recorded in the 7 preceding days, we labeled this prescription as a phantom prescription. We defined a visit to a medical professional as any visit that yielded an inpatient or outpatient Medicare institutional claim or a carrier Medicare non-institutional claim. Inpatient claims are hospital based. In addition to general outpatient visits, outpatient claims include hospital outpatient departments, rural health clinics, federally qualified health centers, renal dialysis facilities, outpatient rehabilitation facilities, comprehensive outpatient rehabilitation facilities, and community mental health centers. Carrier claims are provider claims for non-inpatient services. Carriers include physicians, physician assistants, clinical social workers, nurse practitioners, independent clinical laboratories, ambulance providers, and freestanding ambulatory surgical centers.26 Finally, to estimate the impact of refill prescriptions on the phantom antimicrobial prescribing rate, We defined a refill prescription when a subject fills another prescription for the same antimicrobial within 7 days of the former prescription’s supply ending.
Comparing Patient Attributes
To determine if patients who received phantom antimicrobial prescriptions were different from patients who received antimicrobials following a visit, we compared the phantom prescriptions to visit-associated prescriptions. We considered the following patient characteristics: age, gender, race, measures of socioeconomic status, measures of patient comorbidities in the year prior to the index AMI and measures of cardiac and survival outcomes in the year post-index. We compared the proportions by estimating odds ratios and by testing whether the odds ratios were equal to one.
Because many patients received multiple antimicrobials in the year following their AMI, we answered the question of whether there existed differences between the two groups in two different ways. First, we considered the patient as the unit of analysis. Here if a patient received any phantom prescription for antimicrobials, the patient was considered to be a patient with phantom prescriptions. If a patient received one or more antimicrobials and all the prescriptions were associated with a visit, then he or she was considered to be a patient without phantom prescriptions. If a patient did not receive any antimicrobial prescriptions in the study period, he or she was excluded from this analysis.
Second, we used the antimicrobial prescription as the unit of analysis. For this analysis, for each antimicrobial prescription, we recorded the attributes of the patient for whom the prescription was prescribed and filled. For this analysis, some individuals will be counted more than once and can even be counted in both the phantom and visit-associated cohorts. For instance, if a patient received 7 phantom and 3 visit-associated antimicrobial prescriptions, the patient’s attributes will be counted 7 and 3 times in the phantom and visit-associated cohort attributes, respectively. To account for the clustering effects related to some patients receiving multiple antimicrobial prescriptions, we employed a univariate GEE model with a binomial distribution, logit link and exchangeable correlation structure.
Comparing Antimicrobial Prescription Rates
We were primarily interested in the antimicrobials most frequently prescribed. Thus, we only considered antimicrobials that were one of the 20 most common antimicrobials prescribed to either patients with phantom prescriptions or patients with a visit. For each of these drugs, we computed: (1) the total number of antimicrobial claims, both phantom and visit-associated; (2) the percentage of all antimicrobial claims associated with each drug, both phantom and visit-associated; and (3) the percentage of claims for each drug that were prescribed without a visit.
Odds ratios and their 95% confidence intervals comparing the odds that the given antimicrobial prescription is phantom to those odds for all other antimicrobials listed in Appendix 1 are presented in Table 3. P-values corresponding to Wald tests comparing the odds ratios to one are also presented in Table 3. Bonferroni adjustments were made for multiple comparisons.
Table 3.
Comparing Phantom and Visit-Associated Prescription Rates for Various Antimicrobials
| Antimicrobial | Phantom Count |
Visit- Assoc. Count |
Percent of Given ABX RX that are Phantom |
OR | 95% CI | p-value |
|---|---|---|---|---|---|---|
| Amoxicillin | 7607 | 17936 | 29.78% | 2.6969 | (2.6209,2.7752) | <0.0001 |
| Azithromycin | 6575 | 33660 | 16.34% | 1.1475 | (1.1156,1.1803) | <0.0001 |
| Nitrofurantoin | 5599 | 20928 | 21.11% | 1.6115 | (1.5622,1.6623) | <0.0001 |
| Ciprofloxacin | 5070 | 41816 | 10.81% | 0.6689 | (0.6486,0.6898) | <0.0001 |
| Cephalexin | 4568 | 29076 | 13.58% | 0.8993 | (0.8704,0.9292) | <0.0001 |
| Levofloxacin | 4271 | 44823 | 8.70% | 0.5111 | (0.4945,0.5281) | <0.0001 |
| Sulfamethoxazole-Trimethoprim | 4173 | 27117 | 13.34% | 0.8800 | (0.8506,0.9104) | <0.0001 |
| Doxycycline | 2363 | 12269 | 16.15% | 1.1186 | (1.0693,1.1701) | <0.0001 |
| Clindamycin | 1862 | 6210 | 23.07% | 1.7600 | (1.6698,1.8851) | <0.0001 |
| Amoxicillin-Clavulanate | 1768 | 16051 | 9.92% | 0.6240 | (0.5936,0.6560) | <0.0001 |
| Penicillin | 1256 | 1802 | 41.07% | 4.1029 | (3.8154,4.4120) | <0.0001 |
| Metronidazole | 1097 | 11169 | 8.94% | 0.5585 | (0.5245,0.5947) | <0.0001 |
| Moxifloxacin | 824 | 8125 | 9.21% | 0.5796 | (0.5392,0.6230) | <0.0001 |
| Cefuroxime | 650 | 6151 | 9.56% | 0.6059 | (0.5586,0.6573) | <0.0001 |
| Vancomycin | 604 | 6343 | 8.69% | 0.5452 | (0.5013,0.5929) | <0.0001 |
| Trimethoprim | 561 | 997 | 36.01% | 3.2767 | (2.9534,3.6353) | <0.0001 |
| Clarithromycin | 549 | 2845 | 16.18% | 1.1166 | (1.0186,1.2240) | 0.0186 |
| Erythromycin | 418 | 1297 | 24.37% | 1.8697 | (1.6739,2.0885) | <0.0001 |
| Tetracycline | 411 | 1544 | 21.02% | 1.5429 | (1.3833,1.7209) | <0.0001 |
| Methenamine | 354 | 659 | 34.95% | 3.1192 | (2.7404,3.5504) | <0.0001 |
| Cefdinir | 263 | 2894 | 8.33% | 0.5229 | (0.4607,0.5934) | <0.0001 |
| Ampicillin | 248 | 1740 | 12.47% | 0.8230 | (0.7202,0.9404) | 0.0042 |
| Cefadroxil | 209 | 1348 | 13.42% | 0.8958 | (0.7741,1.0366) | 0.1397 |
Phantom (visit-associated) count for a specific antimicrobial is the total number of such prescriptions filled for that antimicrobial. The odds ratio for a specific antimicrobial compares the odds of a prescription for the given antimicrobial being phantom to those odds for all other antimicrobials listed in Appendix 1.
Because antimicrobials are often inappropriately prescribed for influenza, 27 we determined if the rate at which antimicrobials are prescribed without a prescription is significantly higher at times when influenza is more prevalent. We performed these analyses for all antimicrobials listed in Appendix 1, as well as certain antimicrobials and combinations of antimicrobials. Data from the CDC giving the national influenza-like illness (ILI) rate for each week in the years 2007–2009 were used to determine influenza activity. If the ILI rate was higher than 2.1% in a particular week, then we considered this a high influenza rate week, otherwise the week was considered a low influenza rate week. We determined if each antimicrobial prescription was filled during a high or low influenza rate week.
Finally, because urinary tract infections (UTIs) may be frequently self-diagnosed and in some cases appropriately treated without a visit, we assessed whether antimicrobials often prescribed for UTIs were prescribed without a corresponding visit at a higher rate than the other antimicrobials. We considered Sulfamethoxazole-Trimethoprim, Trimethoprim, Fosfomycin, Nitrofurantoin, Ciprofloxacin, and Levofloxacin as drugs often prescribed for UTIs. Using GEE models because some patients received multiple antimicrobial prescriptions, we modeled the odds of receiving one of the aforementioned UTI drugs without a visit compared to the odds for any other microbial studied. Note these models are univariate. We report Wald p-values testing whether the odds ratio equals one. All analyses were performed using SAS 9.4.
Results
Of the 185,010 patients in the cohort, 113,904 received an antimicrobial of interest in the year following their index acute myocardial infarction. In our cohort, a total of 356,545 prescriptions were filled for all antimicrobials of interest. Of these prescriptions, 52,587 (14.75%) were filled without a corresponding visit in the 7 days prior to the filling of the prescription. Of these 52,587 phantom prescriptions, 12,511 (23.79%) were what we defined as a refill prescription. If one were to eliminate all refill prescriptions, then the refill-adjusted phantom antimicrobial prescribing rate would be 13.04%.
As shown in Table 1, our cohort of 185,010 patients is predominately female (59.42%), white (82.83%) and from a low income area (61.68%). Furthermore, 16.27% of the cohort suffered another cardiac event in the year following their acute myocardial infarction; 76.37% survived for one year, and 65.81% survived one year without experiencing another cardiac event. The cohort of 113,904 antimicrobial receivers is also primarily female (61.56%), white (83.75%), 81.90% survived and 69.40% survived without experiencing a cardiac event for one-year post-discharge. We defined a cardiac event as an acute myocardial infarction, stroke or unstable angina.
Table 1.
Patient-level Analysis Comparing Patient Characteristics Associated with Phantom Prescriptions to Visit-Associated Prescriptions
| Number (%) | |||||||
|---|---|---|---|---|---|---|---|
| Attribute | All | ABX Receivers |
Phantom | Visit- Assoc. |
OR | 95% CI | p-val |
| Overall | 185010 | 113904 | 31,011 | 82893 | |||
| Male Sex | 75073 (40.58) | 43784 (38.44) | 11312 (36.48) | 32472 (39.17) | 0.892 | (0.868,0.916) | <0.0001 |
| Race | |||||||
| White | 153242 (82.83) | 95395 (83.75) | 26940 (86.87) | 68455 (82.58) | 1.396 | (1.344,1.449) | <0.0001 |
| Black | 15412 (8.33) | 8344 (7.33) | 1761 (5.68) | 6583 (7.94) | 0.698 | (0.661,0.737) | <0.0001 |
| Hispanic | 10781 (5.83) | 6842 (6.01) | 1496 (4.82) | 5346 (6.45) | 0.735 | (0.693,0.780) | <0.0001 |
| Other | 5575 (3.01) | 3323 (2.92) | 814 (2.62) | 2509 (3.03) | 0.864 | (0.797,0.936) | 0.0003 |
| Age | |||||||
| 66–70 | 33571 (18.15) | 20920 (18.37) | 5753 (18.55) | 15167 (18.30) | 1.017 | (0.984,1.052) | 0.3225 |
| 71–75 | 34313 (18.55) | 21606 (18.97) | 5774 (18.62) | 15832 (19.10) | 0.969 | (0.937,1.002) | 0.0652 |
| 76–80 | 37238 (20.13) | 23339 (20.49) | 6246 (20.14) | 17093 (20.62) | 0.971 | (0.940,1.003) | 0.0745 |
| 81–85 | 36002 (19.46) | 22186 (19.48) | 5986 (19.30) | 16200 (19.54) | 0.985 | (0.953,1.018) | 0.3635 |
| 85 or older | 43886 (23.72) | 25853 (22.70) | 7252 (23.39) | 18601 (22.44) | 1.055 | (1.023,1.088) | 0.0007 |
| Socioeconomic Factors | |||||||
| Low Income Area | 114123 (61.68) | 71012 (62.34) | 19166 (61.80) | 51846 (62.55) | 0.969 | (0.943,0.995) | 0.0213 |
| Non-English-speaking area | 52319 (28.28) | 31587 (27.73) | 8082 (26.06) | 23505 (28.36) | 0.891 | (0.865,0.917) | <0.0001 |
| Dually Eligible for Medicaid | 68498 (37.02) | 43502 (38.19) | 11105 (35.81) | 32397 (39.08) | 0.870 | (0.846,0.893) | <0.0001 |
| One-Year Pre-Index Comorbidities | |||||||
| Diabetes | 72146 (39.00) | 46852 (41.13) | 11671 (37.64) | 35181 (42.44) | 0.818 | (0.797,0.841) | <0.0001 |
| CKD | 37901 (20.49) | 24678 (21.67) | 5590 (18.03) | 19088 (23.03) | 0.735 | (0.711,0.760) | <0.0001 |
| COPD | 52324 (28.28) | 35856 (31.48) | 9677 (31.21) | 26179 (31.58) | 0.983 | (0.955,1.011) | 0.2235 |
| One-Year Post Discharge Cardiac and Survival Outcomes | |||||||
| AMI | 18287 (9.88) | 11961 (10.50) | 2712 (8.75) | 9249 (11.16) | 0.763 | (0.730,0.798) | <0.0001 |
| Cardiac Arrest | 7891 (4.27) | 3992 (3.50) | 771 (2.49) | 3221 (3.89) | 0.631 | (0.582,0.683) | <0.0001 |
| Stroke | 5508 (2.98) | 3305 (2.90) | 738 (2.38) | 2567 (3.10) | 0.763 | (0.702,0.829) | <0.0001 |
| Cardiac Event | 30105 (16.27) | 19699 (17.29) | 4601 (14.84) | 15098 (18.21) | 0.782 | (0.754,0.811) | <0.0001 |
| Survival | 141288 (76.37) | 93293 (81.90) | 26829 (86.51) | 66464 (80.18) | 1.586 | (1.529,1.645) | <0.0001 |
| Cardiac Free Survival | 121756 (65.81) | 79045 (69.40) | 23224 (74.89) | 55821 (67.34) | 1.446 | (1.404,1.490) | <0.0001 |
Comparing patients' race, age, SES, one-year pre-index comorbidities andone-year post-index outcomes for four cohorts: all patients, patients receiving any antimicrobial of interest, patients who received at least one antimicrobial and none of which were phantom, and patients receiving at least one phantom prescription. For patients with at least one antimicrobial prescription, the odds ratios compare the odds a patient with the given attribute will have a phantom prescription to those odds of a patient without the given attribute
Table 1 displays patient characteristics for patients who received at least one phantom antimicrobial and patients who received only visit-associated antimicrobial prescription(s). Of the patients who received an antimicrobial, the odds that at least one of the prescriptions was a phantom prescription was significantly higher for the following groups: females, white patients, patients 85 and older, patients who survived the year following their index AMIs, patients who did not have either diabetes or chronic kidney disease (CKD) in the year prior to their index AMIs, patients who did not have any cardiac events in the year after their index dates, those who lived in predominantly English-speaking areas and those who were not dually eligible for Medicare.
In Table 2 we show the results from our antimicrobial-level analysis. Note that the results are similar to the results of our patient-level analysis (Table 1). The following patient characteristics were associated with a higher likelihood that a prescription did not have a corresponding visit: female, white, 85 or older, from an area that predominantly speaks English, not dually eligible for Medicaid, without a diagnosis of diabetes, CKD or chronic obstructive pulmonary disease (COPD) in the year prior to their index AMI, and those who did not have any cardiac events and survived the year following their index AMI.
Table 2.
Antimicrobial-Level Analysis Comparing Patient Characteristics for Phantom Prescriptions to Visit-Associated Prescriptions
| Number (%) | ||||||
|---|---|---|---|---|---|---|
| Characteristic | All | Phantom | Visit- Associated |
OR | 95% CI | p-value |
| Overall | 356,545 | 52,587 | 303,958 | |||
| Male Sex | 128,786 (36.12) | 18,424 (35.04) | 110,362 (36.31) | 0.9702 | (0.9459,0.9951) | 0.0194 |
| Race | ||||||
| White | 302,733 (84.91) | 46,214 (87.88) | 256,519 (84.39) | 1.3089 | (1.2624,1.3572) | < 0.0001 |
| Black | 23,467 (6.58) | 2,696 (5.13) | 20,771 (6.83) | 0.7506 | (0.7122,0.7912) | < 0.0001 |
| Hispanic | 20,463 (5.74) | 2,342 (4.45) | 18,121 (5.96) | 0.7445 | (0.7032,0.7882) | < 0.0001 |
| Other | 9,882 (2.77) | 1,335 (2.54) | 8,547 (2.81) | 0.9505 | (0.8797,1.0270) | 0.1985 |
| Age | ||||||
| 66–70 | 66,360 (18.61) | 9,611 (18.28) | 56,749 (18.67) | 1.0047 | (0.9736,1.0368) | 0.7689 |
| 71–75 | 68,210 (19.13) | 9,785 (18.61) | 58,425 (19.22) | 0.9683 | (0.9384,0.9992) | 0.0444 |
| 76–80 | 72,481 (20.33) | 10,144 (19.29) | 62,337 (20.51) | 0.9386 | (0.9106,0.9674) | < 0.0001 |
| 81–85 | 69,602 (19.52) | 10,148 (19.30) | 59,454 (19.56) | 0.9732 | (0.9437,1.0037) | 0.0841 |
| 85 or Older | 79,892 (22.41) | 12,899 (24.53) | 66,993 (22.04) | 1.1105 | (1.0789,1.1430) | < 0.0001 |
| Socioeconomic Factors | ||||||
| Low Income Area | 224,739 (63.03) | 32,897 (62.56) | 191,842 (63.11) | 0.9484 | (0.9249,0.9726) | < 0.0001 |
| Non-English-speaking area | 95,002 (26.65) | 13,154 (25.01) | 81,848 (26.93) | 0.9244 | (0.8991,0.9505) | < 0.0001 |
| Dually Eligible for Medicaid | 147,973 (41.50) | 19,464 (37.01) | 128,509 (42.28) | 0.7919 | (0.7719,0.8123) | < 0.0001 |
| One-Year Pre-Index Comorbidities | ||||||
| Diabetes | 156,599 (43.92) | 19,575 (37.22) | 137,024 (45.08) | 0.7217 | (0.7038,0.7401) | <0.0001 |
| CKD | 85,351 (23.94) | 9,384 (17.84) | 75,967 (24.99) | 0.6460 | (0.6259,0.6667) | <0.0001 |
| COPD | 128,023 (35.91) | 16,971 (32.27) | 111,052 (36.54) | 0.8199 | (0.7986,0.8417) | <0.0001 |
| One-Year Post-Index Cardiac and Survival Outcomes | ||||||
| AMI | 38,946 (10.92) | 4,373 (8.32) | 34,573 (11.37) | 0.6968 | (0.6678,0.7271) | < 0.0001 |
| Cardiac Arrest | 11,338 (3.18) | 1,144 (2.18) | 10,194 (3.35) | 0.6654 | (0.6150,0.7198) | < 0.0001 |
| Stroke | 10,231 (2.87) | 1,194 (2.27) | 9,037 (2.97) | 0.7449 | (0.6875,0.8071) | < 0.0001 |
| Cardiac Event | 63,785 (17.89) | 7,476 (14.22) | 56,309 (18.53) | 0.7165 | (0.6926,0.7413) | < 0.0001 |
| Survival | 299,436 (83.98) | 46,061 (87.59) | 253,375 (83.36) | 1.3845 | (1.3357,1.4351) | < 0.0001 |
| Cardiac Free Survival | 250,728 (70.32) | 40,134 (76.32) | 210,594 (69.28) | 1.4241 | (1.3846,1.4646) | < 0.0001 |
Comparing patients' race, age, SES, one-year pre-index comborbidities one-year post-index survival and cardiac outcomes for phantom and visit-associated antimicrobial prescriptions. The odds ratios compare the odds that a prescription filled by a patient with the given attribute will be a phantom prescription to the odds of a patient without the given attribute.
The third set of analyses involved determining if specific antimicrobial agents had a higher chance of being prescribed without a corresponding visit and results are displayed in Table 3. We adjusted for multiple comparisons using the Bonferroni method using a level of significance of 0.05/23=0.0022. Even when adjusting for multiple comparisons, all the odds ratios are significantly different from one for all antimicrobials in Table 3 except Clarithromycin, Ampicillin and Cefadroxil. Specifically, prescriptions including Amoxicillin, Penicillin, Trimethoprim and Methenamine are much more likely than the typical antimicrobial studied to be prescribed without a visit. Conversely, Levofloxacin, Metronidazole, Moxifloxacin, Vancomycin, and Cefdinir are much less likely to be prescribed without a visit.
The odds of a prescription for various antimicrobials and combinations of antimicrobials being phantom in high influenza weeks compared to those odds during low influenza weeks is displayed in Table 4. For all antimicrobials in Appendix 1, we estimated that the odds a prescription was filled without a visit during a high influenza week is 1.0467 (1.0251–1.0687, p<0.0001) times that of the odds in a week with a lower influenza rate. For a combination of Clarithromycin, Azithromycin, Amoxicillin, Amoxicillin-Clavulanate, Levofloxacin, Moxifloxacin and Doxycycline, the odds ratio is 1.0413 (1.0103–1.0733, p=0.0087), yielding a result very similar to that of all antimicrobials. Both Amoxicillin and Moxifloxacin, each analyzed separately, yielded odds ratios which were not significantly different from one. Azithromycin, on the other hand, was significantly more likely to be prescribed without a visit when the influenza rate was high, with an odds ratio of 1.1123 (1.0517–1.1764, p=0.0002). Finally, prescriptions for common UTI drugs were less likely to be phantom prescriptions than non-UTI-drug prescriptions: the odds of receiving one of the commonly prescribed UTI drugs without a visit was 0.6931 (0.6783–0.7082, p<0.0001) times the odds of receiving a phantom prescription for any of the other antimicrobials studied.
Table 4.
Comparing Odds of Phantom Prescriptions during Influenza Season
| Drug or Drug Combo |
Odds Ratio |
95% CI | p-value |
|---|---|---|---|
| All ABX in Appendix 1 | 1.0467 | (1.0251,1.0687) | <0.0001 |
| Selected 7 ABX* | 1.0413 | (1.0103,1.0733) | 0.0087 |
| Azithromycin | 1.1123 | (1.0517,1.1764) | 0.0002 |
| Amoxicillin | 0.9604 | (0.9037,1.0206) | 0.1931 |
| Moxifloxacin | 1.0697 | (0.9151,1.2504) | 0.3976 |
The odds ratios compare the odds of a prescription for the given antimicrobial or combination of antimicrobials being phantom for those filled during a high influenza week to the odds for those filled during a low influenza week.
Note that the Selected 7 antimicrobials are: Clarithromycin, Azithromycin, Amoxicillin, Amoxicillin-Clavulanate, Levofloxacin, Moxifloxacin, and Doxycycline.
Discussion
Our results show that almost 15% of antimicrobials were prescribed without evidence of a corresponding encounter with a healthcare provider. Even removing refill prescriptions from consideration yielded a phantom antimicrobial prescribing rate of over 13%. These phantom prescriptions were more likely to be prescribed to patients who were white, female and/or forpatients who were relatively healthy. For example, having diabetes and CKD reduced the odds of phantom prescriptions. Finally, phantom antimicrobial prescriptions occurred more commonly in our study population during weeks with higher influenza rates, especially prescriptions for azithromycin.
Although we could link a majority of antimicrobial use to healthcare encounters, we believe that this rate of phantom prescribing is important to consider when designing programs to measure and control antimicrobial use. Among our cohort of approximately 185,000 patients, during a year of follow up, over 356,500 antimicrobial prescriptions were filled, and 52,587 of these were phantom prescriptions. In the United States, over 250,000 million prescriptions for antimicrobials are filled every year.28 Thus, given our results, it is conceivable that several tens of millions of antimicrobial prescriptions may be written without a corresponding face-to-face encounter with a healthcare provider. These phantom antimicrobial prescriptions are concerning because they are potentially immune from traditional office-based-stewardship interventions.
We think that the majority of these phantom prescriptions are filled on the basis of phone calls. While some of these antimicrobial prescriptions are appropriate, it is harder to evaluate the appropriateness of antimicrobial use. From the clinician’s perspective, it is substantially harder to make an appropriate diagnosis without the ability to do: a physical exam, a structured interview, or order appropriate diagnostic tests. Furthermore, over the phone, the history may not actually be taken by the prescriber, but instead may be taken by someone else (e.g., a nurse or answering service). From a stewardship perspective, without a clinical note, detailed diagnosis or lab values, it is difficult to evaluate prescriptions, feed back prescribing rates in a standardized fashion, or do accurate benchmarking.
Our results are especially important given the emergence of telehealth or electronic visits (e-visits). Substantial concerns have been raised for the potential for such visits to increase inappropriate antimicrobial use.29–31 Nevertheless; e-visits have advantages over telephone encounters. First, they are structured appointments and telephone-based prescribing is not. Insofar as telephone visits often occur at busy times, it decreases the opportunity for correcting patient expectations; a major driver of antimicrobial use.23–25Second, a scheduled e-visit has the potential to learn more about the patient’s condition. Finally, with e-visits, there is potentially more-structured data to enable stewardship efforts than with telephone prescribing.
Although our study population was older with more comorbidities, it was actually the healthier and more affluent patients that were more likely to have these phantom antimicrobial prescriptions. Initially this result was surprising to us. Our initial intuition was the clinicians would be more likely to prescribe antimicrobials to people that were not as healthy, as these patients may be more vulnerable and more likely to have worse outcomes if they were to suffer a condition requiring an antimicrobial. Given that phantom prescribing is also more common among people living in relatively more affluent areas, our findings may be related to patient expectations. In affluent areas, expectations for treatment may be higher. Indeed, other work has shown that where provider density is higher, inappropriate antimicrobial prescribing is higher.32,33
This work is subject to many limitations. Our population is made up entirely of Medicare fee-for-service beneficiaries who have suffered an AMI. Therefore our results may not be generalizable to healthier and younger populations. However, this is an important group to consider. Infections may have more serious outcomes in this population, and adverse drug events may be more common. Future work should focus on relatively healthy populations where inappropriate prescribing may actually be higher.
Our study has several limitations. First, some antimicrobial prescriptions labeled as phantom may have been associated with a visit. If patients went to a clinic and paid cash, we have a record of this visit. Likewise, some subjects may have paid cash for their antimicrobials and not submitted a claim. Second, our phantom prescribing estimates may, in part, be driven by appropriate antimicrobial prescribing for UTIs. However, the odds of receiving a prescription for an antimicrobial agent typically prescribed for UTIs without a corresponding visit were significantly lower than the odds of receiving a phantom prescription for other antimicrobials. Third, some prescriptions could be related to dental visits, which we do not capture, but dental prescriptions are only likely only to affect a subset of the antimicrobials we considered, and some antimicrobials highlighted in our analysis are much less likely to be prescribed in a dental setting (e.g., azithromycin). Finally, antimicrobials could have been filled in a nursing home where there was a provider encounter, but it was not charged to Medicare as such. To address this issue, we conducted a sensitivity analysis. We compared patients with at least one claim for a long-term-care facility to those without such a claim. We found that patients with nursing home claims were less likely to have a phantom prescription than patients without such a claim. (Data not shown.) We were not surprised by this finding because, overall, phantom prescriptions tended to occur in healthier patients.
Despite our limitations, we found that phantom antimicrobial prescribing was relatively common. As the population ages and clinics become busier, pressure and incentives for these types of encounters will likely increase. In addition, e-visits will undoubtedly continue to increase. Thus, new antimicrobial stewardship efforts are needed to target these types of prescribing. Also, our results suggest that the rate of encounter-free antimicrobial prescribing should be investigated in other populations, especially younger, healthier populations that might be more likely to have higher expectations for antimicrobial prescribing.
Acknowledgments
Financial Support. This work was supported by the National Heart, Lung and Blood Institute at the National Institutes of Health [grant number K25HL122305 to LAP] and the University of Iowa Health Venture’s Signal Center [to PMP].
Appendix 1: List of Antimicrobials Studied
| Levofloxacin | Ciprofloxacin | Azithromycin | Cephalexin | Sulfamethoxazole-Trimethoprim |
| Nitrofurantoin | Amoxicillin | Amoxicillin-Clavulanate | Doxycycline | Metronidazole |
| Moxifloxacin | Clindamycin | Vancomycin | Cefuroxime | Clarithromycin |
| Cefdinir | Penicillin | Ampicillin | Tetracycline | Erythromycin |
| Trimethoprim | Cefadroxil | Minocycline | Rifampin | Linezolid |
| Methenamine | Cefpodoxime | Cefaclor | Demeclocycline | Cefprozil |
| Dicloxacillin | Rifaximin | Dapsone | Gemifloxacin | Cefditoren |
| Norfloxacin | Fosfomycin | Cefixime | Ceftibuten | Ofloxacin |
| Telithromycin | Carbenicillin |
Footnotes
Presentations: An earlier version of this work was presented at the American Pharmacists’ Association Meeting on March 5, 2016 in Baltimore, MD.
Potential Conflicts of Interest. All authors report no conflicts of interest relevant to this article.
Contributor Information
Benjamin N. Riedle, University of Iowa, Department of Biostatistics, Iowa City, IA 52242, USA.
Linnea A. Polgreen, University of Iowa, Department of Pharmacy Practice and Science, Iowa City, IA 52242, USA.
Joseph E. Cavanaugh, University of Iowa, Department of Biostatistics, Iowa City, IA 52242, USA
Mary C. Schroeder, University of Iowa, Department of Pharmacy Practice and Science, Iowa City, IA 52242, USA
Philip M. Polgreen, University of Iowa, Departments of Internal Medicine and Epidemiology, Iowa City, 52242, USA.
References
- 1.Schwartz B, Bell DM, Hughes JM. Preventing the emergence of antimicrobial resistance. A call for action by clinicians, public health officials, and patients. JAMA. 1997;278(11):944–945. doi: 10.1001/jama.278.11.944. [DOI] [PubMed] [Google Scholar]
- 2.Spellberg B, Guidos R, Gilbert D, et al. The epidemic of antibiotic-resistant infections: a call to action for the medical community from the Infectious Diseases Society of America. Clinical infectious diseases. 2008;46(2):155–164. doi: 10.1086/524891. [DOI] [PubMed] [Google Scholar]
- 3.French GL. The continuing crisis in antibiotic resistance. International journal of antimicrobial agents. 2010;36(Suppl 3):S3–S7. doi: 10.1016/S0924-8579(10)70003-0. [DOI] [PubMed] [Google Scholar]
- 4.World Health Organization. Antimicrobial resistance: global report on surveillance 2014. 2014. [Google Scholar]
- 5.File TM, Jr, Srinivasan A, Bartlett JG. Antimicrobial stewardship: importance for patient and public health. Clinical infectious diseases. 2014;59(Suppl 3):S93–S96. doi: 10.1093/cid/ciu543. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Huttner B, Harbarth S, Nathwani D. Success stories of implementation of antimicrobial stewardship: a narrative review. Clinical microbiology and infection : the official publication of the European Society of Clinical Microbiology and Infectious Diseases. 2014;20(10):954–962. doi: 10.1111/1469-0691.12803. [DOI] [PubMed] [Google Scholar]
- 7.Fleming-Dutra KE, Hersh AL, Shapiro DJ, et al. Prevalence of Inappropriate Antibiotic Prescriptions Among US Ambulatory Care Visits, 2010–2011. JAMA. 2016;315(17):1864–1873. doi: 10.1001/jama.2016.4151. [DOI] [PubMed] [Google Scholar]
- 8.Costelloe C, Metcalfe C, Lovering A, Mant D, Hay AD. Effect of antibiotic prescribing in primary care on antimicrobial resistance in individual patients: systematic review and meta-analysis. BMJ (Clinical research ed) 2010;340:c2096. doi: 10.1136/bmj.c2096. [DOI] [PubMed] [Google Scholar]
- 9.Wagner B, Filice GA, Drekonja D, et al. Antimicrobial stewardship programs in inpatient hospital settings: a systematic review. Infection control and hospital epidemiology. 2014;35(10):1209–1228. doi: 10.1086/678057. [DOI] [PubMed] [Google Scholar]
- 10.Drekonja DM, Filice GA, Greer N, et al. Antimicrobial stewardship in outpatient settings: a systematic review. Infection control and hospital epidemiology. 2015;36(2):142–152. doi: 10.1017/ice.2014.41. [DOI] [PubMed] [Google Scholar]
- 11.Rohde JM, Jacobsen D, Rosenberg DJ. Role of the hospitalist in antimicrobial stewardship: a review of work completed and description of a multisite collaborative. Clinical therapeutics. 2013;35(6):751–757. doi: 10.1016/j.clinthera.2013.05.005. [DOI] [PubMed] [Google Scholar]
- 12.Johannsson B, Beekmann SE, Srinivasan A, Hersh AL, Laxminarayan R, Polgreen PM. Improving antimicrobial stewardship: the evolution of programmatic strategies and barriers. Infection control and hospital epidemiology. 2011;32(4):367–374. doi: 10.1086/658946. [DOI] [PubMed] [Google Scholar]
- 13.Hersh AL, Beekmann SE, Polgreen PM, Zaoutis TE, Newland JG. Antimicrobial stewardship programs in pediatrics. Infection control and hospital epidemiology. 2009;30(12):1211–1217. doi: 10.1086/648088. [DOI] [PubMed] [Google Scholar]
- 14.Tabah A, Cotta MO, Garnacho-Montero J, et al. A Systematic Review of the Definitions, Determinants, and Clinical Outcomes of Antimicrobial De-escalation in the Intensive Care Unit. Clinical infectious diseases. 2016;62(8):1009–1017. doi: 10.1093/cid/civ1199. [DOI] [PubMed] [Google Scholar]
- 15.Luyt CE, Brechot N, Trouillet JL, Chastre J. Antibiotic stewardship in the intensive care unit. Critical care (London, England) 2014;18(5):480. doi: 10.1186/s13054-014-0480-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Policy statement on antimicrobial stewardship by the Society for Healthcare Epidemiology of America (SHEA), the Infectious Diseases Society of America (IDSA), and the Pediatric Infectious Diseases Society (PIDS) Infection control and hospital epidemiology. 2012;33(4):322–327. doi: 10.1086/665010. [DOI] [PubMed] [Google Scholar]
- 17.Elligsen M, Walker SA, Pinto R, et al. Audit and feedback to reduce broad-spectrum antibiotic use among intensive care unit patients: a controlled interrupted time series analysis. Infection control and hospital epidemiology. 2012;33(4):354–361. doi: 10.1086/664757. [DOI] [PubMed] [Google Scholar]
- 18.Kaki R, Elligsen M, Walker S, Simor A, Palmay L, Daneman N. Impact of antimicrobial stewardship in critical care: a systematic review. The journal of antimicrobial chemotherapy. 2011;66(6):1223–1230. doi: 10.1093/jac/dkr137. [DOI] [PubMed] [Google Scholar]
- 19.Finkelstein JA, Davis RL, Dowell SF, et al. Reducing antibiotic use in children: a randomized trial in 12 practices. Pediatrics. 2001;108(1):1–7. doi: 10.1542/peds.108.1.1. [DOI] [PubMed] [Google Scholar]
- 20.Andrews T, Thompson M, Buckley DI, et al. Interventions to influence consulting and antibiotic use for acute respiratory tract infections in children: a systematic review and meta-analysis. PloS one. 2012;7(1):e30334. doi: 10.1371/journal.pone.0030334. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Gerber JS, Prasad PA, Fiks AG, et al. Effect of an outpatient antimicrobial stewardship intervention on broad-spectrum antibiotic prescribing by primary care pediatricians: a randomized trial. JAMA. 2013;309(22):2345–2352. doi: 10.1001/jama.2013.6287. [DOI] [PubMed] [Google Scholar]
- 22.Hallsworth M, Chadborn T, Sallis A, et al. Provision of social norm feedback to high prescribers of antibiotics in general practice: a pragmatic national randomised controlled trial. Lancet. 2016;387(10029):1743–1752. doi: 10.1016/S0140-6736(16)00215-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Nyquist AC, Gonzales R, Steiner JF, Sande MA. Antibiotic prescribing for children with colds, upper respiratory tract infections, and bronchitis. JAMA. 1998;279(11):875–877. doi: 10.1001/jama.279.11.875. [DOI] [PubMed] [Google Scholar]
- 24.Mangione-Smith R, McGlynn EA, Elliott MN, Krogstad P, Brook RH. The relationship between perceived parental expectations and pediatrician antimicrobial prescribing behavior. Pediatrics. 1999;103(4 Pt 1):711–718. doi: 10.1542/peds.103.4.711. [DOI] [PubMed] [Google Scholar]
- 25.Bauchner H, Pelton SI, Klein JO. Parents, physicians, and antibiotic use. Pediatrics. 1999;103(2):395–401. doi: 10.1542/peds.103.2.395. [DOI] [PubMed] [Google Scholar]
- 26.Carrier RIF. [Accessed May 8, 2016];Overview. Research Data Assistance Center (ResDAC) website. http://www.resdac.org/cms-data/files/carrier-rif. Published 2016.
- 27.Polgreen PM, Yang M, Laxminarayan R, Cavanaugh JE. Respiratory fluoroquinolone use and influenza. Infection control and hospital epidemiology. 2011;32(7):706–709. doi: 10.1086/660859. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Hicks LA, Bartoces MG, Roberts RM, Suda KJ, Hunkler RJ, Taylor TH, Jr, et al. US outpatient antibiotic prescribing variation according to geography, patient population, and provider specialty in 2011. Clinical infectious diseases. 2015;60(9):1308–16. doi: 10.1093/cid/civ076. [DOI] [PubMed] [Google Scholar]
- 29.Mehrotra A, Paone S, Martich GD, Albert SM, Shevchik GJ. A comparison of care at e-visits and physician office visits for sinusitis and urinary tract infection. JAMA internal medicine. 2013;173(1):72–4. doi: 10.1001/2013.jamainternmed.305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Uscher-Pines L, Mulcahy A, Cowling D, Hunter G, Burns R, Mehrotra A. Antibiotic Prescribing for Acute Respiratory Infections in Direct-to-Consumer Telemedicine Visits. JAMA internal medicine. 2015;175(7):1234–5. doi: 10.1001/jamainternmed.2015.2024. [DOI] [PubMed] [Google Scholar]
- 31.Uscher-Pines L, Mulcahy A, Cowling D, Hunter G, Burns R, Mehrotra A. Access and Quality of Care in Direct-to-Consumer Telemedicine. Telemedicine journal and e-health. 2016;22(4):282–7. doi: 10.1089/tmj.2015.0079. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Klein EY, Makowsky M, Orlando M, Hatna E, Braykov NP, Laxminarayan R. Influence of provider and urgent care density across different socioeconomic strata on outpatient antibiotic prescribing in the USA. The journal of antimicrobial chemotherapy. 2015;70(5):1580–7. doi: 10.1093/jac/dku563. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Masiero G, Filippini M, Ferech M, Goossens H. Socioeconomic determinants of outpatient antibiotic use in Europe. International journal of public health. 2010;55(5):469–78. doi: 10.1007/s00038-010-0167-y. [DOI] [PMC free article] [PubMed] [Google Scholar]