Antibiotic Prescribing for Acute Cough in Primary Care
Antibiotic Prescribing for Acute Cough in Primary Care
We performed a retrospective analysis to identify acute cough visits to a single primary care practice. We included visits by patients aged 18 to 64 years old, with a cough lasting up to 21 days, and without chronic lung disease. This analysis was part of an evaluation to inform an intervention targeted at reducing antibiotic prescribing for acute cough/acute bronchitis throughout our Practice Based Research Network.
The Phyllis Jen Center for Primary Care is a teaching practice in the Brigham and Women's Primary Care Practice Based Research Network. The practice has over 40 faculty physicians and 80 internal medicine residents with continuity clinics each academic year. About 40 additional internal medicine residents rotate through the practice for urgent care visits. There are also 3 Nurse Practitioners, 2 Registered Nurses, 2 Pharmacy Technicians, 2 Pharmacists, and 1 Social Worker. The practice serves a socioeconomically, racially, and ethnically diverse patient population.
The practice uses the Longitudinal Medical Record as the official electronic health record (EHR). The EHR was internally developed by Partners HealthCare, of which Brigham and Women's Hospital is a founding member. The EHR includes primary care and subspecialty notes, problem lists, medication lists, coded allergies, and laboratory test and radiographic study results. The EHR has clinical decision support, but not pertaining to the antibiotic treatment of acute respiratory infections.
Using the EHR schedule, we identified all visits made to the practice between March 1, 2011 and June 30, 2012. We included only patients aged 18 to 64 years old to match the age range of the HEDIS performance measure. We excluded patients who had made a visit to the practice in the prior 30 days, and patients with chronic lung disease on their problem list. Our list of chronic lung disease diagnoses included, but was not limited to, asthma, asthmatic bronchitis, chronic obstructive pulmonary disease, interstitial lung disease, and chronic bronchitis.
For patient visits that met our inclusion criteria, we reviewed the EHR note to identify patients with acute cough. We verified that patients met a guideline-consistent definition of acute cough/acute bronchitis: a cough lasting 21 or fewer days in a patient age 18 to 64 years old without chronic lung disease or immunosuppression. To be included in our analysis, patients did not have to have a chief complaint of cough documented in the visit note. If the treating clinician did not mention the duration of cough, study staff contacted the clinician to ascertain whether the cough was acute (≤ 21 days). We excluded visits at which the clinician did not address acute cough in their treatment plan.
To assess inclusion and exclusion criteria and to record information about acute cough visits, we imported and manually extracted acute cough visit data into a study-specific database (Microsoft Access 2003). We imported data including date, provider name, patient identification number, age, sex, race/ethnicity (from registration data) and vital signs. We extracted clinical documentation data about the patient's chief complaint or reason for visit, symptoms, physical exam findings, diagnoses, and treatments. For symptoms, we also noted whether the following eight symptoms were present, not present, or not mentioned: sore throat, phlegm/sputum, shortness of breath, headache, fevers, nasal symptoms, myalgias, and chest pain.
For the physical exam, we extracted information on the normality, abnormality, or lack of mention about general appearance, tympanic membranes, external auditory canals, oropharynx, lymphadenopathy, lung sounds, or wheezing. Finally, we recorded the presence or absence of a chest x-ray, and whether the results were normal or abnormal.
We extracted up to three clinician-assigned diagnoses as documented in the assessment and plan section of the visit note. For each visit, we assigned one of 16 common diagnoses associated with an acute cough. We considered sinusitis, pneumonia, streptococcal pharyngitis, otitis media, bacterial infection, and pertussis "antibiotic-appropriate diagnoses". We use the term "antibiotic-appropriate diagnoses" for diagnoses for which some patients may require antibiotics according to guidelines (e.g., sinusitis with severe symptoms) even though the majority of patients with "antibiotic-appropriate diagnoses" may not require antibiotics. We considered upper respiratory infection (URI), acute bronchitis, viral syndrome, post-nasal drip, non-streptococcal pharyngitis, allergies, reactive airway disease, gastroesophageal reflux disease, conjunctivitis, and influenza "non-antibiotic-appropriate diagnoses". For diagnoses not in one of these 16 categories, we classified them as "other" and considered "other" a non-antibiotic-appropriate diagnosis. Even though some of these diagnoses might more rightly be considered symptoms (e.g., post-nasal drip), these were the diagnoses assigned by the treating clinician.
We noted when the clinician expressed uncertainty when assigning the diagnosis, such as when providers used words like "maybe," "unclear," or used question marks in association with the diagnosis. As examples, we considered notes to express diagnostic uncertainty if they included text such as the patient "had a cough and sore throat with unclear etiology" or "presenting with ?acute bronchitis". We also considered clinicians' weighing of multiple possible diagnoses an expression of uncertainty (e.g., "pneumonia versus acute bronchitis" or "URI or possible underlying sinusitis"). We did not consider clinicians documenting two distinct diagnoses an expression of uncertainty (e.g. "pneumonia with acute bronchitis" or "URI with sinusitis").
We collected and documented up to four prescribed medications as recorded in the electronic health record, giving preference to antibiotics and cough-related treatments. We did not abstract information about non-cough-related diagnoses or associated antibiotics (e.g. amoxicillin given to a patient with a urinary tract infection). We did not distinguish between immediate antibiotic prescribing and "delayed antibiotic prescribing". Delayed antibiotic prescribing was very rare and guidelines and performance measures make no distinction or allowance for delayed antibiotic prescribing.
We extracted information about whether patients made any follow-up visits to an affiliated site or received any antibiotic prescriptions within the 30 days following their initial visit. For each follow-up visit made, we noted where the care was delivered (emergency department, primary care practice, specialist, etc.), the subject of the visit note, and whether the visit was related to acute bronchitis and/or antibiotic prescribing. If the follow-up visit was cough-related, we extracted up to three diagnoses, whether the clinicians expressed diagnostic uncertainty, and/or prescribed antibiotics.
For each visit, we assigned a principal diagnosis. We prioritized the principal diagnosis by giving preference to the most antibiotic-appropriate diagnosis mentioned in a given visit in the following order: pneumonia, sinusitis, streptococcal pharyngitis, otitis media, pertussis, non-streptococcal pharyngitis, acute bronchitis, upper respiratory infection, post nasal drip, allergies, and other. For example, if a visit was given both URI and pneumonia diagnoses, the visit was assigned a principal diagnosis of pneumonia.
We used standard descriptive statistics to describe and compare baseline characteristics of physicians and patients. We compared continuous variables using Student's t test and categorical variables using the X test. We used SAS version 9.2 (SAS Institute Inc, Cary, NC). We considered p-values < 0.05 significant.
The Partners HealthCare Human Research Committee approved the study protocol.
Methods
Overview
We performed a retrospective analysis to identify acute cough visits to a single primary care practice. We included visits by patients aged 18 to 64 years old, with a cough lasting up to 21 days, and without chronic lung disease. This analysis was part of an evaluation to inform an intervention targeted at reducing antibiotic prescribing for acute cough/acute bronchitis throughout our Practice Based Research Network.
Setting
The Phyllis Jen Center for Primary Care is a teaching practice in the Brigham and Women's Primary Care Practice Based Research Network. The practice has over 40 faculty physicians and 80 internal medicine residents with continuity clinics each academic year. About 40 additional internal medicine residents rotate through the practice for urgent care visits. There are also 3 Nurse Practitioners, 2 Registered Nurses, 2 Pharmacy Technicians, 2 Pharmacists, and 1 Social Worker. The practice serves a socioeconomically, racially, and ethnically diverse patient population.
The practice uses the Longitudinal Medical Record as the official electronic health record (EHR). The EHR was internally developed by Partners HealthCare, of which Brigham and Women's Hospital is a founding member. The EHR includes primary care and subspecialty notes, problem lists, medication lists, coded allergies, and laboratory test and radiographic study results. The EHR has clinical decision support, but not pertaining to the antibiotic treatment of acute respiratory infections.
Data Sources
Using the EHR schedule, we identified all visits made to the practice between March 1, 2011 and June 30, 2012. We included only patients aged 18 to 64 years old to match the age range of the HEDIS performance measure. We excluded patients who had made a visit to the practice in the prior 30 days, and patients with chronic lung disease on their problem list. Our list of chronic lung disease diagnoses included, but was not limited to, asthma, asthmatic bronchitis, chronic obstructive pulmonary disease, interstitial lung disease, and chronic bronchitis.
Data Extraction
For patient visits that met our inclusion criteria, we reviewed the EHR note to identify patients with acute cough. We verified that patients met a guideline-consistent definition of acute cough/acute bronchitis: a cough lasting 21 or fewer days in a patient age 18 to 64 years old without chronic lung disease or immunosuppression. To be included in our analysis, patients did not have to have a chief complaint of cough documented in the visit note. If the treating clinician did not mention the duration of cough, study staff contacted the clinician to ascertain whether the cough was acute (≤ 21 days). We excluded visits at which the clinician did not address acute cough in their treatment plan.
To assess inclusion and exclusion criteria and to record information about acute cough visits, we imported and manually extracted acute cough visit data into a study-specific database (Microsoft Access 2003). We imported data including date, provider name, patient identification number, age, sex, race/ethnicity (from registration data) and vital signs. We extracted clinical documentation data about the patient's chief complaint or reason for visit, symptoms, physical exam findings, diagnoses, and treatments. For symptoms, we also noted whether the following eight symptoms were present, not present, or not mentioned: sore throat, phlegm/sputum, shortness of breath, headache, fevers, nasal symptoms, myalgias, and chest pain.
For the physical exam, we extracted information on the normality, abnormality, or lack of mention about general appearance, tympanic membranes, external auditory canals, oropharynx, lymphadenopathy, lung sounds, or wheezing. Finally, we recorded the presence or absence of a chest x-ray, and whether the results were normal or abnormal.
We extracted up to three clinician-assigned diagnoses as documented in the assessment and plan section of the visit note. For each visit, we assigned one of 16 common diagnoses associated with an acute cough. We considered sinusitis, pneumonia, streptococcal pharyngitis, otitis media, bacterial infection, and pertussis "antibiotic-appropriate diagnoses". We use the term "antibiotic-appropriate diagnoses" for diagnoses for which some patients may require antibiotics according to guidelines (e.g., sinusitis with severe symptoms) even though the majority of patients with "antibiotic-appropriate diagnoses" may not require antibiotics. We considered upper respiratory infection (URI), acute bronchitis, viral syndrome, post-nasal drip, non-streptococcal pharyngitis, allergies, reactive airway disease, gastroesophageal reflux disease, conjunctivitis, and influenza "non-antibiotic-appropriate diagnoses". For diagnoses not in one of these 16 categories, we classified them as "other" and considered "other" a non-antibiotic-appropriate diagnosis. Even though some of these diagnoses might more rightly be considered symptoms (e.g., post-nasal drip), these were the diagnoses assigned by the treating clinician.
We noted when the clinician expressed uncertainty when assigning the diagnosis, such as when providers used words like "maybe," "unclear," or used question marks in association with the diagnosis. As examples, we considered notes to express diagnostic uncertainty if they included text such as the patient "had a cough and sore throat with unclear etiology" or "presenting with ?acute bronchitis". We also considered clinicians' weighing of multiple possible diagnoses an expression of uncertainty (e.g., "pneumonia versus acute bronchitis" or "URI or possible underlying sinusitis"). We did not consider clinicians documenting two distinct diagnoses an expression of uncertainty (e.g. "pneumonia with acute bronchitis" or "URI with sinusitis").
We collected and documented up to four prescribed medications as recorded in the electronic health record, giving preference to antibiotics and cough-related treatments. We did not abstract information about non-cough-related diagnoses or associated antibiotics (e.g. amoxicillin given to a patient with a urinary tract infection). We did not distinguish between immediate antibiotic prescribing and "delayed antibiotic prescribing". Delayed antibiotic prescribing was very rare and guidelines and performance measures make no distinction or allowance for delayed antibiotic prescribing.
We extracted information about whether patients made any follow-up visits to an affiliated site or received any antibiotic prescriptions within the 30 days following their initial visit. For each follow-up visit made, we noted where the care was delivered (emergency department, primary care practice, specialist, etc.), the subject of the visit note, and whether the visit was related to acute bronchitis and/or antibiotic prescribing. If the follow-up visit was cough-related, we extracted up to three diagnoses, whether the clinicians expressed diagnostic uncertainty, and/or prescribed antibiotics.
Data and Statistical Analysis
For each visit, we assigned a principal diagnosis. We prioritized the principal diagnosis by giving preference to the most antibiotic-appropriate diagnosis mentioned in a given visit in the following order: pneumonia, sinusitis, streptococcal pharyngitis, otitis media, pertussis, non-streptococcal pharyngitis, acute bronchitis, upper respiratory infection, post nasal drip, allergies, and other. For example, if a visit was given both URI and pneumonia diagnoses, the visit was assigned a principal diagnosis of pneumonia.
We used standard descriptive statistics to describe and compare baseline characteristics of physicians and patients. We compared continuous variables using Student's t test and categorical variables using the X test. We used SAS version 9.2 (SAS Institute Inc, Cary, NC). We considered p-values < 0.05 significant.
The Partners HealthCare Human Research Committee approved the study protocol.