Nephrology Comanagement and the Quality of Antibiotic Prescribing in Primary Care for Patients With Chronic Kidney Disease: A Retrospective Cross-sectional Study

A Retrospective Cross-Sectional Study

Justin X.G. Zhu; Danielle M. Nash; Eric McArthur; Alexandra Farag; Amit X. Garg; Arsh K. Jain

Disclosures

Nephrol Dial Transplant. 2019;34(4):642-649. 

In This Article

Materials and Methods

Study Overview and Setting

We conducted a population-based, propensity score–matched and retrospective cross-sectional study in the province of Ontario, Canada from 1 April 2003 to 31 March 2014. Ontario currently has 13 million citizens and all residents have universal health care access, including physician services, hospital care and laboratory investigations. Ontario residents ≥65 years of age also have universal prescription drug coverage via the Ontario Drug Benefit program. This study was conducted at the Institute for Clinical Evaluative Sciences (ICES) Western site in London, Ontario, Canada. Our study was approved by the research ethics board at Sunnybrook Health Sciences Centre. The reporting of this study follows recommended guidelines for routinely collected health care data (Supplementary data, Table S1).[15]

Data Sources

We ascertained baseline characteristics, physician visits and prescription data using eight linked health care databases. These datasets were linked using unique encoded identifiers. Demographic and vital status information on all Ontario residents who have ever been issued a health card is recorded in the Ontario Registered Persons Database. Detailed diagnostic and procedural information on all hospital admissions is recorded in the Canadian Institute for Health Information's Discharge Abstract Database and all emergency department (ED) visits in the National Ambulatory Care Reporting System database. Health claims for inpatient and outpatient physician services are recorded in the Ontario Health Insurance Plan database. Outpatient prescription drug information, including the dispensing date, quantity of pills and number of days supplied, is accurately recorded in the Ontario Drug Benefit database for all individuals ≥65 years of age, with an error rate of <1%.[16] Daily drug dose is calculated using the strength of medication multiplied by the quantity of tablets divided by the number of days supplied. The ICES Physician Database contains information on physicians in Ontario such as medical specialty, education, practice location and demographics. We obtained baseline serum creatinine values from two-linked laboratory databases: Dynacare, a large outpatient provincial laboratory provider, and Cerner (Kansas City, MO, USA), an electronic medical record database containing inpatient, outpatient and ED laboratory values for 12 hospitals in southwestern Ontario. These data sources have been used in previous population-based renal drug studies.[5,14,17–24]

Identification of Patients and Study Prescriptions

We identified all patients in Ontario ≥65 years of age who filled a prescription between 1 April 2003 and 31 March 2014 prescribed by a primary care physician for one of the following study antibiotics: cephalexin, ciprofloxacin, clarithromycin, nitrofurantoin, trimethoprim/sulfamethoxazole, levofloxacin, cefprozil, amoxicillin/clavulanic acid, cefixime, tetracycline or ofloxacin. It is recommended that the daily dose of these antibiotics be reduced in the presence of CKD (see 'Outcome' section). The date of prescription filling served as the index date (study entry date). To be eligible for this study, all patients were required to have a baseline value of renal function, and we identified patients' renal function using their estimated glomerular filtration rate (eGFR), calculated from their most recent outpatient serum creatinine value in the 1 year prior to their index date [a median of 57 (25th–75th percentile 15–143) days prior to the index date] using the CKD Epidemiology Collaboration (CKD-EPI) formula.[25]

The best equation to estimate kidney function for the purposes of drug adjustment continues to be controversial. The US Kidney Disease Education program indicates that equations that express results in mL/min/1.73 m2 or mL/min are both appropriate for this purpose. In this study we estimated glomerular filtration rate (GFR) using the CKD-EPI equation, which when <30 mL/min/1.73 m2 would also generally identify a patient with a Cockcroft–Gault result <30 mL/min. In addition, we found that outpatient serum creatinine tests in our region are generally stable.[26] Therefore we also allowed patients with only a single eligible outpatient serum creatinine in the study to prevent reductions to the sample.

We made the following exclusions: patients with missing key variables (age, gender, database ID, non-Ontario residents and patients who are deceased prior to prescription date); patients in the first year of eligibility for prescription drug coverage (age <66 years) were excluded to prevent incomplete medication records; patients with eGFR ≥30 mL/min/1.73 m2 (to include only patients with Stage 4 or 5 CKD); patients discharged from the hospital or ED in the 7 days prior to the index date to ensure that prescriptions were new outpatient prescriptions; patients with any study medications in the 180 days prior to index date to ensure new antibiotic use and to eliminate prescriptions used for chronic infections; patients on chronic dialysis or with evidence of a prior kidney transplant; patients with multiple study prescriptions on the same date; or patients with prescriptions not prescribed by a primary care physician. If there were multiple eligible prescriptions available, we restricted to the first prescription (i.e. one prescription per patient).

Exposure

The primary exposure was comanagement, which was defined as having at least one outpatient visit with a nephrologist in the year prior to the index date (see Supplementary data, Table S2). In Ontario, a specialist consultation (including nephrology) will always result in a letter back to the referring physician. It is also common practice for specialists to send a copy of the assessment to a patient's primary care physician even when the referral is made by another type of physician involved in the patient's care. Patients were categorized into those with and without evidence of nephrology comanagement.

Outcome

Our primary outcome was whether the antibiotic prescription was appropriately dosed for a patient's given eGFR. Dosing recommendations were in accordance with UpToDate and the Compendium of Pharmaceuticals and Specialties (CPS) in November 2014, focusing specifically on Canadian recommendations. A prescription was labeled as inappropriate if the daily dose was above the acceptable cutoff as defined in Table 1.

We included the following antibiotics in our study: cephalexin, ciprofloxacin, clarithromycin, nitrofurantoin, trimethoprim/sulfamethoxazole, levofloxacin, cefprozil, amoxicillin/clavulanic acid, cefixime, tetracycline and ofloxacin. We selected common antibiotics across multiple classes that are prescribed in the outpatient setting.[27] These antibiotics have all been associated with side effects, which may be exacerbated in patients with decreased drug clearance.[28]

Statistical Analysis

Variables for baseline characteristics were identified a priori and were compared between noncomanaged and comanaged groups using standardized differences. This metric describes differences between group means relative to the pooled standard deviation (SD) and is considered a meaningful difference if >10%.[29] Continuous variables were described as mean with SD and median with interquartile range (IQR). Categorical and binary variables were described as a proportion.

We used propensity score matching to achieve balance on a large number of measured baseline characteristics in the two groups defined by nephrology comanagement. A propensity score for the predicted probability of receiving nephrology comanagement was derived from a logistic regression model in which treatment status was regressed on >35 variables that were potentially associated with comanagement or the outcome (Supplementary data, Table S3).[29] We used greedy matching to match each comanaged patient to a noncomanaged patient based on the following characteristics: the logit of the propensity score (±0.2 SD), CKD stage (Stage 4 versus Stage 5) and year of index date (pre-2006 versus 1 January 2006 and onwards). We applied matching without replacement, where patients could only be selected once for inclusion in the study. Greedy matching without replacement has previously been demonstrated to produce less biased estimates than other algorithms.[30] We used conditional logistic regression to obtain the conditional odds ratio (OR) of the association between nephrology comanagement and appropriately dosed prescriptions, with noncomanaged patients as the referent group. As there may have been clustering by primary care physician, we addressed this in a sensitivity analysis. Specifically, we reran the logistic regression model, accounting for correlation by primary care physician using generalized estimating equations. Using model statistical interaction terms, we also performed subgroup analyses to determine whether the association between comanagement and appropriately dosed prescriptions was modified by the introduction of mandatory eGFR reporting in Ontario (pre-2006 versus post-2006) or CKD stage [Stage 4 (eGFR 15–<30 mL/min/1.73 m2) versus Stage 5 (eGFR <15 mL/min/1.73 m2)].

The studied variables included age, gender, eGFR, albumin:creatinine ratio (ACR) (where available), hematuria (where available), rural residence (population <10 000), neighborhood income quintile, long-term care placement and year of index prescription date; the number of health care encounters in the last year, including hospitalizations, ED visits, primary care physician visits and internal medicine visits; time since last nephrology visit; number of unique medications within the last 180 days; prescriptions within the previous 180 days, including antihypertensive medications, diabetic medications and immunosuppressive medications; primary care prescriber characteristics, including age, gender, practice location, country of graduation and time since graduation; and patient comorbidities in the past 5 years, including Charlson comorbidity index, hypertension, diabetes, coronary artery disease, congestive heart failure, myocardial infarction, chronic lung disease, major cancers, atrial fibrillation, stroke, chronic liver disease and peripheral vascular disease. See Supplementary data, Table S3 for administrative codes used to define baseline characteristics.

All statistical analyses were performed using Statistical Analysis Software (SAS) version 9.4 (SAS Institute, Cary, NC, USA). A two-sided P-value <0.05 was defined as statistically significant.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....