Understanding adherence to therapeutic guidelines: a multilevel analysis of statin prescription in the Skaraborg Primary Care Database

Identifying factors explaining practice variation in secondary stroke prevention in primary care: a cohort study based on all patients with ischaemic stroke in the Stockholm region

OBJECTIVES

The aim of this study was to describe the practice variation in dispensation of secondary stroke preventive drugs among patients at different primary care centres (PCCs) in Stockholm region and to identify factors that may explain the variation.

DESIGN

Cohort study using administrative data from the Stockholm region.

SETTING

Stockholm Health Care Region, Sweden, serving a population of 2.3 million inhabitants, hospital and PCC data.

PARTICIPANTS

All patients (n=9761) with ischaemic stroke treated in hospital from 1 July 2009 to 30 June 2014 were included. Of these, 7562 patients registered with 187 PCCs were analysed. Exclusion criteria were; deceased patients, age <18, haemorrhagic stroke and/or switching PCC.

PRIMARY AND SECONDARY OUTCOME MEASURES

As primary outcome the impact of PCC organisation variables and patient characteristics on the dispensation of statins, antiplatelets, antihypertensives and anticoagulants were analysed. Secondarily, the unadjusted practice variation of preventive drug dispensation of 187 PCCs is described.

RESULTS

There was up to fourfold practice variation in dispensation of all secondary preventive drugs. Factors associated with a lower level of dispensed statins were privately run PCCs (OR 0.91 (95% CI 0.82 to 1.00)) and the patient being woman. Increased statin use was associated with a higher number of specialists in family medicine (OR 1.03 (95% CI 1.01 to 1.05)) and a higher proportion of patients registered with a specific physician (OR 1.37 (95% CI 1.11 to 1.68)). Women had on average a lower number of dispensed antihypertensives.

CONCLUSIONS

A high practice variation for dispensation of all secondary preventive drugs was observed. Patient and PCC level factors indicating good continuity of care and high level of general practitioner education were associated with higher use of statins. Findings are of importance to policymakers as well as individual providers of care, and more research and actions are needed to minimise inequality in healthcare.

Cross-classified Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA) to evaluate hospital performance: the case of hospital differences in patient survival after acute myocardial infarction

OBJECTIVE

To describe a novel strategy, Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA) to evaluate hospital performance, by analysing differences in 30-day mortality after a first-ever acute myocardial infarction (AMI) in Sweden.

DESIGN

Cross-classified study.

SETTING

68 Swedish hospitals.

PARTICIPANTS

43 247 patients admitted between 2007 and 2009, with a first-ever AMI.

PRIMARY AND SECONDARY OUTCOME MEASURES

We evaluate hospital performance by analysing differences in 30-day mortality after a first-ever AMI using a cross-classified multilevel analysis. We classified the patients into 10 categories according to a risk score (RS) for 30-day mortality and created 680 strata defined by combining hospital and RS categories.

RESULTS

In the cross-classified multilevel analysis the overall RS adjusted hospital 30-day mortality in Sweden was 4.78% and the between-hospital variation was very small (variance partition coefficient (VPC)=0.70%, area under the curve (AUC)=0.54). The benchmark value was therefore achieved by all hospitals. However, as expected, there were large differences between the RS categories (VPC=34.13%, AUC=0.77) CONCLUSIONS: MAIHDA is a useful tool to evaluate hospital performance. The benefit of this novel approach to adjusting for patient RS is that it allowed one to estimate separate VPCs and AUC statistics to simultaneously evaluate the influence of RS categories and hospital differences on mortality. At the time of our analysis, all hospitals in Sweden were performing homogeneously well. That is, the benchmark target for 30-day mortality was fully achieved and there were not relevant hospital differences. Therefore, possible quality interventions should be universal and oriented to maintain the high hospital quality of care.

Trends and Geographical Variability in Osteoporosis Treatment After Hip Fracture: A Multilevel Analysis of 30,965 Patients in the Region of Valencia, Spain

Despite improvements in the therapeutic arsenal and the recommendations of guidelines, low rates of prescribing osteoporosis medications are being reported worldwide for patients surviving a hip fracture, and important geographical variation remain. We aimed to describe trends in the proportion of patients that receive osteoporosis medication after hip fracture and to analyze the geographical variation in the prescription of drug therapy and its associated factors in the region of Valencia, Spain. We studied a population-based retrospective cohort of 30,965 patients aged 65 years and older, discharged from hospital after a hip fracture from January 2008 to December 2015, who were followed up for 3 months after discharge to identify the presence of any prescription of osteoporosis medication. We conducted a multilevel multiple logistic regression analysis with two levels (individuals and health departments [HD]) to determine which individual covariates were associated with receiving a prescription of osteoporosis medication in the 3 months after discharge, as well as the importance of the HD of hospitalization. The percentage of patients treated in the region decreased from a maximum of 28.9% in 2009 to 16.4% in 2015. By sex, the proportion of women treated reached a maximum of 33.4% in 2009 and declined to 19% in 2015, while the proportion of men reached a maximum of 14% in 2011 and reduced to 8.1% in 2015. By health department, there was a noticeable variability in the rate of patients treated, ranging from 40.9% to 11.1% in the whole period (intraclass correlation coefficient [ICC] = 7.54%; median odds ratio [MOR] = 1.64). Proportion of treated patients decreased in 20 of the 24 HDs. Variability could be also observed with regard to choice of medication by HD. This situation pressingly demands action (both at the organizational and professional levels) focused on populations at a higher risk (such as hip fracture patients) that particularly address underutilization and unwarranted variation.

Geographical and sociodemographic differences in discontinuation of medication for Chronic Obstructive Pulmonary Disease - A Cross-Classified Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA)

BACKGROUND

While discontinuation of COPD maintenance medication is a known problem, the proportion of patients with discontinuation and its geographical and sociodemographic distribution are so far unknown in Sweden. Therefore, we analyse this question by applying an innovative approach called multilevel analysis of individual heterogeneity and discriminatory accuracy (MAIHDA).

PATIENTS AND METHODS

We analysed 49,019 patients categorized into 18 sociodemographic contexts and 21 counties of residence. All patients had a hospital COPD diagnosis and had been on inhaled maintenance medication during the 5 years before the study baseline in 2010. We defined "discontinuation" as the absolute lack of retrieval from a pharmacy of any inhaled maintenance medication during 2011. We performed a cross-classified MAIHDA and obtained the average proportion of discontinuation, as well as county and sociodemographic absolute risks, and compared them with a proposed benchmark value of 10%. We calculated the variance partition coefficient (VPC) and the area under the receiver operating characteristics curve (AUC) to quantify county and sociodemographic differences. To summarize the results, we used a framework with 15 scenarios defined by the size of the differences and the level of achievement in relation to the benchmark value.

RESULTS

Around 18% of COPD patients in Sweden discontinued maintenance medication, so the benchmark value was not achieved. There were very small county differences (VPC=0.35%, AUC=0.54). The sociodemographic differences were small (VPC=4.98%, AUC=0.57).

CONCLUSION

Continuity of maintenance medication among COPD patients in Sweden could be improved by reducing the unjustifiably high prevalence of discontinuation. The very small county and small sociodemographic differences should motivate universal interventions across all counties and sociodemographic groups. Geographical analyses should be combined with sociodemographic analyses, and the cross-classified MAIHDA is an appropriate tool to assess health-care quality.

Disentangling the contribution of hospitals and municipalities for understanding patient level differences in one-year mortality risk after hip-fracture: A cross-classified multilevel analysis in Sweden

Background

One-year mortality after hip-fracture is a widely used outcome measure when comparing hospital care performance. However, traditional analyses do not explicitly consider the referral of patients to municipality care after just a few days of hospitalization. Furthermore, traditional analyses investigates hospital (or municipality) variation in patient outcomes in isolation rather than as a component of the underlying patient variation. We therefore aimed to extend the traditional approach to simultaneously estimate both case-mix adjusted hospital and municipality comparisons in order to disentangle the amount of the total patient variation in clinical outcomes that was attributable to the hospital and municipality level, respectively.

Methods

We determined 1-year mortality risk in patients aged 65 or above with hip fractures registered in Sweden between 2011 and 2014. We performed cross-classified multilevel analysis with 54,999 patients nested within 54 hospitals and 290 municipalities. We adjusted for individual demographic, socioeconomic and clinical characteristics. To quantify the size of the hospital and municipality variation we calculated the variance partition coefficient (VPC) and the area under the receiver operator characteristic curve (AUC).

Results

The overall 1-year mortality rate was 25.1%. The case-mix adjusted rates varied from 21.7% to 26.5% for the 54 hospitals, and from 18.9% to 29.5% for the 290 municipalities. The VPC was just 0.2% for the hospital and just 0.1% for the municipality level. Patient sociodemographic and clinical characteristics were strong predictors of 1-year mortality (AUC = 0.716), but adding the hospital and municipality levels in the cross-classified model had a minor influence (AUC = 0.718).

Conclusions

Overall in Sweden, one-year mortality after hip-fracture is rather high. However, only a minor part of the patient variation is explained by the hospital and municipality levels. Therefore, a possible intervention should be nation-wide rather than directed to specific hospitals or municipalities.

Geographic Disparity in Deceased Donor Liver Transplant Rates Following Share 35

BACKGROUND

The Organ Procurement and Transplantation Network implemented Share 35 on June 18, 2013, to broaden deceased donor liver sharing within regional boundaries. We investigated whether increased sharing under Share 35 impacted geographic disparity in deceased donor liver transplantation (DDLT) across donation service areas (DSAs).

METHODS

Using Scientific Registry of Transplant Recipients June 2009 to June 2017, we identified 86 083 adult liver transplant candidates and retrospectively estimated Model for End-Stage Liver Disease (MELD)-adjusted DDLT rates using nested multilevel Poisson regression with random intercepts for DSA and transplant program. From the variance in DDLT rates across 49 DSAs and 102 programs, we derived the DSA-level median incidence rate ratio (MIRR) of DDLT rates. MIRR is a robust metric of heterogeneity across each hierarchical level; larger MIRR indicates greater disparity.

RESULTS

MIRR was 2.18 pre-Share 35 and 2.16 post-Share 35. Thus, 2 candidates with the same MELD in 2 different DSAs were expected to have a 2.2-fold difference in DDLT rate driven by geography alone. After accounting for program-level heterogeneity, MIRR was attenuated to 2.10 pre-Share 35 and 1.96 post-Share 35. For candidates with MELD 15-34, MIRR decreased from 2.51 pre- to 2.27 post-Share 35, and for candidates with MELD 35-40, MIRR increased from 1.46 pre- to 1.51 post-Share 35, independent of program-level heterogeneity in DDLT. DSA-level heterogeneity in DDLT rates was greater than program-level heterogeneity pre- and post-Share 35.

CONCLUSIONS

Geographic disparity substantially impacted DDLT rates before and after Share 35, independent of program-level heterogeneity and particularly for candidates with MELD 35-40. Despite broader sharing, geography remains a major determinant of access to DDLT.

Albuminuria measurement in diabetic care: a multilevel analysis measuring the influence of accreditation on institutional performance

Background

Studies assessing institutional performance regarding quality of care are frequently performed using single-level statistical analyses investigating differences between provider averages of various quality indicators. However, such analyses are insufficient as they do not consider patients’ heterogeneity around those averages. Hence, we apply a multilevel analysis of individual-patient heterogeneity that distinguishes between ‘general’ (‘latent quality’ or measures of variance) and ‘specific’ (measures of association) contextual effects. We assess general contextual effects of the hospital departments and the specific contextual effect of a national accreditation programme on adherence to the standard benchmark for albuminuria measurement in Danish patients with diabetes.

Methods

From the Danish Adult Diabetes Database, we extracted data on 137 893 patient cases admitted to hospitals between 2010 and 2013. Applying multilevel logistic and probit regression models for every year, we quantified general contextual effects of hospital department by the intraclass correlation coefficient (ICC) and the area under the receiver operating characteristic curve (AUC) values. We evaluated the specific effect of hospital accreditation using the ORs and the change in the department variance.

Results

In 2010, the department context had considerable influence on adherence with albuminuria measurement (ICC=21.8%, AUC=0.770), but the general effect attenuated along with the implementation of the national accreditation programme. The ICC value was 16.5% in 2013 and the rate of compliance with albuminuria measurement increased from 91.6% in 2010 to 96% in 2013.

Conclusions

Parallel to implementation of the national accreditation programme, departments’ compliance with the standard benchmark for albuminuria measurement increased and the ICC values decreased, but remained high. While those results indicate an overall quality improvement, further intervention focusing on departments with the lowest compliance could be considered.

The median hazard ratio: a useful measure of variance and general contextual effects in multilevel survival analysis

Multilevel data occurs frequently in many research areas like health services research and epidemiology. A suitable way to analyze such data is through the use of multilevel regression models (MLRM). MLRM incorporate cluster‐specific random effects which allow one to partition the total individual variance into between‐cluster variation and between‐individual variation. Statistically, MLRM account for the dependency of the data within clusters and provide correct estimates of uncertainty around regression coefficients. Substantively, the magnitude of the effect of clustering provides a measure of the General Contextual Effect (GCE). When outcomes are binary, the GCE can also be quantified by measures of heterogeneity like the Median Odds Ratio (MOR) calculated from a multilevel logistic regression model. Time‐to‐event outcomes within a multilevel structure occur commonly in epidemiological and medical research. However, the Median Hazard Ratio (MHR) that corresponds to the MOR in multilevel (i.e., ‘frailty’) Cox proportional hazards regression is rarely used. Analogously to the MOR, the MHR is the median relative change in the hazard of the occurrence of the outcome when comparing identical subjects from two randomly selected different clusters that are ordered by risk. We illustrate the application and interpretation of the MHR in a case study analyzing the hazard of mortality in patients hospitalized for acute myocardial infarction at hospitals in Ontario, Canada. We provide R code for computing the MHR. The MHR is a useful and intuitive measure for expressing cluster heterogeneity in the outcome and, thereby, estimating general contextual effects in multilevel survival analysis. © 2016 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.

Bad apples or spoiled barrels? Multilevel modelling analysis of variation in high-risk prescribing in Scotland between general practitioners and between the practices they work in

OBJECTIVES

Primary care high-risk prescribing causes significant harm, but it is unclear if it is largely driven by individuals (a 'bad apple' problem) or by practices having higher or lower risk prescribing cultures (a 'spoiled barrel' problem). The study aimed to examine the extent of variation in high-risk prescribing between individual prescribers and between the practices they work in.

DESIGN, SETTING AND PARTICIPANTS

Multilevel logistic regression modelling of routine cross-sectional data from 38 Scottish general practices for 181,010 encounters between 398 general practitioners (GPs) and 26,539 patients particularly vulnerable to adverse drug events (ADEs) of non-steroidal anti-inflammatory drugs (NSAIDs) due to age, comorbidity or co-prescribing.

OUTCOME MEASURE

Initiation of a new NSAID prescription in an encounter between GPs and eligible patients.

RESULTS

A new high-risk NSAID was initiated in 1953 encounters (1.1% of encounters, 7.4% of patients). Older patients, those with more vulnerabilities to NSAID ADEs and those with polypharmacy were less likely to have a high-risk NSAID initiated, consistent with GPs generally recognising the risk of NSAIDs in eligible patients. Male GPs were more likely to initiate a high-risk NSAID than female GPs (OR 1.73, 95% CI 1.39 to 2.16). After accounting for patient characteristics, 4.2% (95% CI 2.1 to 8.3) of the variation in high-risk NSAID prescribing was attributable to variation between practices, and 14.2% (95% CI 11.4 to 17.3) to variation between GPs. Three practices had statistically higher than average high-risk prescribing, but only 15.7% of GPs with higher than average high-risk prescribing and 18.5% of patients receiving such a prescription were in these practices.

CONCLUSIONS

There was much more variation in high-risk prescribing between GPs than between practices, and only targeting practices with higher than average rates will miss most high-risk NSAID prescribing. Primary care prescribing safety improvement should ideally target all practices, but encourage practices to consider and act on variation between prescribers in the practice.

Measuring prevalence, reliability and variation in high-risk prescribing in general practice using multilevel modelling of observational data in a population database

Background

High-risk primary care prescribing is common and is known to vary considerably between practices, but the extent to which high-risk prescribing varies among individual general practitioners (GPs) is not known.

Objectives

To create prescribing safety indicators usable in existing electronic clinical data and to examine (1) variation in high-risk prescribing between patients, GPs and practices including reliability of measurement and (2) changes over time in high-risk prescribing prevalence and variation between practices.

Design

Descriptive analysis and multilevel logistic regression modelling of routine data.

Setting

UK general practice using routine electronic medical record data.

Participants

(1) For analysis of variation and reliability, 398 GPs and 26,539 patients in 38 Scottish practices. (2) For analysis of change in high-risk prescribing, ≈ 300,000 patients particularly vulnerable to adverse drug effects registered with 190 Scottish practices.

Main outcome measures

For the analysis of variation between practices and between GPs, five indicators of high-risk non-steroidal anti-inflammatory drug (NSAID) prescribing. For the analysis of change in high-risk prescribing, 19 previously validated indicators.

Results

Measurement of high-risk prescribing at GP level was feasible only for newly initiated drugs and for drugs similar to NSAIDs which are usually initiated by GPs. There was moderate variation between practices in total high-risk NSAID prescribing [intraclass correlation coefficient (ICC) 0.034], but this indicator was highly reliable (> 0.8 for all practices) at distinguishing between practices because of the large number of patients being measured. There was moderate variation in initiation of high-risk NSAID prescribing between practices (ICC 0.055) and larger variation between GPs (ICC 0.166), but measurement did not reliably distinguish between practices and had reliability > 0.7 for only half of the GPs in the study. Between quarter (Q)2 2004 and Q1 2009, the percentage of patients exposed to high-risk prescribing measured by 17 indicators that could be examined over the whole period fell from 8.5% to 5.2%, which was largely driven by reductions in high-risk NSAID and antiplatelet use. Variation between practices increased for five indicators and decreased for five, with no relationship between change in the rate of high-risk prescribing and change in variation between practices.

Conclusions

High-risk prescribing is common and varies moderately between practices. High-risk prescribing at GP level cannot be easily measured routinely because of the difficulties in accurately identifying which GP actually prescribed the drug and because drug initiation is often a shared responsibility with specialists. For NSAID initiation, there was approximately three times greater variation between GPs than between practices. Most GPs with above average high-risk prescribing worked in practices which were not themselves above average. The observed reductions in high-risk prescribing between 2004 and 2009 were largely driven by falls in NSAID and antiplatelet prescribing, and there was no relationship between change in rate and change in variation between practices. These results are consistent with improvement interventions in all practices being more appropriate than interventions targeted on practices or GPs with higher than average high-risk prescribing. There is a need for research to understand why high-risk prescribing varies and to design and evaluate interventions to reduce it.

Funding

Funding for this study was provided by the Health Services and Delivery Research programme of the National Institute for Health Research.

Use of Fibrates Monotherapy in People with Diabetes and High Cardiovascular Risk in Primary Care: A French Nationwide Cohort Study Based on National Administrative Databases

Background and Aim

According to guidelines, diabetic patients with high cardiovascular risk should receive a statin. Despite this consensus, fibrate monotherapy is commonly used in this population. We assessed the frequency and clinical consequences of the use of fibrates for primary prevention in patients with diabetes and high cardiovascular risk.

Design

Retrospective cohort study based on nationwide data from the medical and administrative databases of French national health insurance systems (07/01/08-12/31/09) with a follow-up of up to 30 months.

Methods

Lipid-lowering drug-naive diabetic patients initiating fibrate or statin monotherapy were identified. Patients at high cardiovascular risk were then selected: patients with a diagnosis of diabetes and hypertension, and >50 (men) or 60 (women), but with no history of cardiovascular events. The composite endpoint comprised myocardial infarction, stroke, amputation, or death.

Results

Of the 31,652 patients enrolled, 4,058 (12.8%) received a fibrate. Age- and gender-adjusted annual event rates were 2.42% (fibrates) and 2.21% (statins). The proportionality assumption required for the Cox model was not met for the fibrate/statin variable. A multivariate model including all predictors was therefore calculated by dividing data into two time periods, allowing Hazard Ratios to be calculated before (HR_<540) and after 540 days (HR_>540) of follow-up. Multivariate analyses showed that fibrates were associated with an increased risk for the endpoint after 540 days: HR_<540 = 0.95 (95% CI: 0.78–1.16) and HR_>540 = 1.73 (1.28–2.32).

Conclusion

Fibrate monotherapy is commonly prescribed in diabetic patients with high cardiovascular risk and is associated with poorer outcomes compared to statin therapy.

Prescriber variation in potentially inappropriate prescribing in older populations in Ireland

Background

Health care policy-makers look for prescribing indicators at the population level to evaluate the performance of prescribers, improve quality and control drug costs. The aim of this research was to; (i) estimate the level of variation in potentially inappropriate prescribing (PIP) across prescribers in the national Irish older population using the STOPP criteria; (ii) estimate how reliably the criteria could distinguish between prescribers in terms of their proportion of PIP and; (iii) examine how PIP varies between prescribers and by patient and prescriber characteristics in a multilevel regression model.

Methods

1,938 general practitioners (GPs) with 338,375 registered patients’ ≥70 years were extracted from the Health Service Executive Primary Care Reimbursement Service (HSE-PCRS) pharmacy claims database. HSE-PCRS prescriptions are WHO ATC coded. Demographic data for claimants’ and prescribers’ are available. Thirty STOPP indicators were applied to prescription claims in 2007. Multilevel logistic regression examined how PIP varied between prescribers and by individual patient and prescriber level variables.

Results

The unadjusted variation in PIP between prescribers was considerable (median 35%, IQR 30-40%). The STOPP criteria were reliable measures of PIP (average >0.8 reliability). The multilevel regression models found that only the patient level variable, number of different repeat drug classes was strongly associated with PIP (>2 drugs v none; adjusted OR, 4.0; 95% CI 3.7, 4.3). After adjustment for patient level variables the proportion of PIP varied fourfold (0.5 to 2 times the expected proportion) between prescribers but the majority of this variation was not significant.

Conclusion

PIP is of concern for all prescribers. Interventions aimed at enhancing appropriateness of prescribing should target patients taking multiple medications.

Increased registration of hypertension and cancer diagnoses after the introduction of a new reimbursement system

OBJECTIVE

To investigate the impact on ICD coding behaviour of a new case-mix reimbursement system based on coded patient diagnoses. The main hypothesis was that after the introduction of the new system the coding of chronic diseases like hypertension and cancer would increase and the variance in propensity for coding would decrease on both physician and health care centre (HCC) levels.

DESIGN

Cross-sectional multilevel logistic regression analyses were performed in periods covering the time before and after the introduction of the new reimbursement system.

SETTING

Skaraborg primary care, Sweden.

SUBJECTS

All patients (n = 76 546 to 79 826) 50 years of age and older visiting 468 to 627 physicians at the 22 public HCCs in five consecutive time periods of one year each.

MAIN OUTCOME MEASURES

Registered codes for hypertension and cancer diseases in Skaraborg primary care database (SPCD).

RESULTS

After the introduction of the new reimbursement system the adjusted prevalence of hypertension and cancer in SPCD increased from 17.4% to 32.2% and from 0.79% to 2.32%, respectively, probably partly due to an increased diagnosis coding of indirect patient contacts. The total variance in the propensity for coding declined simultaneously at the physician level for both diagnosis groups.

CONCLUSIONS

Changes in the healthcare reimbursement system may directly influence the contents of a research database that retrieves data from clinical practice. This should be taken into account when using such a database for research purposes, and the data should be validated for each diagnosis.

Trimethoprim and ciprofloxacin resistance and prescribing in urinary tract infection associated with Escherichia coli: a multilevel model

OBJECTIVES

Individual and group level factors associated with the probability of antimicrobial resistance of uropathogenic Escherichia coli were analysed in a multilevel model.

METHODS

Adult patients consulting with a suspected urinary tract infection (UTI) in 22 general practices over a 9 month period supplied a urine sample for laboratory analysis. Cases were patients with a UTI associated with a resistant E. coli. Previous antimicrobial exposure and other patient characteristics were recorded from the medical files.

RESULTS

Six hundred and thirty-three patients with an E. coli UTI and a full record for all variables were included. Of the E. coli isolates, 36% were resistant to trimethoprim and 12% to ciprofloxacin. A multilevel logistic regression model was fitted. The odds that E. coli was resistant increased with increasing number of prescriptions over the previous year for trimethoprim from 1.4 (0.8-2.2) for one previous prescription to 4.7 (1.9-12.4) for two and 6.4 (2.0-25.4) for three or more. For ciprofloxacin the ORs were 2.7 (1.2-5.6) for one and 6.5 (2.9-14.8) for two or more. The probability that uropathogenic E. coli was resistant showed important variation between practices and a difference of 17% for trimethoprim and 33% for ciprofloxacin was observed for an imaginary patient moving from a practice with low to a practice with high probability. This difference could not be explained by practice prescribing or practice resistance levels.

CONCLUSIONS

Previous antimicrobial use and the practice visited affect the risk that a patient with a UTI will be diagnosed with an E. coli resistant to this agent, which was particularly important for ciprofloxacin.