Article Text

Appropriateness of peripherally inserted central catheter use among general medical inpatients: an observational study using routinely collected data
1. Amol A Verma1,2,3,
2. Alexander Kumachev2,
3. Sonam Shah2,
4. Yishan Guo1,
5. Hae Young Jung1,
6. Shail Rawal2,4,
7. Lauren Lapointe-Shaw2,4,
8. Janice L Kwan2,5,
10. Terence Tang2,7,
1. 1 Li Ka Shing Knowledge Institute, St Michael's Hospital, Toronto, Ontario, Canada
2. 2 Department of Medicine, University of Toronto, Toronto, Ontario, Canada
3. 3 Institute of Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada
4. 4 Department of Medicine, University Health Network, Toronto, Ontario, Canada
5. 5 Department of Medicine, Mount Sinai Hospital, Toronto, Ontario, Canada
6. 6 Department of Medicine, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada
7. 7 Institute for Better Health, Trillium Health Partners, Mississauga, Ontario, Canada
1. Correspondence to Dr Amol A Verma, Li Ka Shing Knowledge Institute, St Michael's Hospital, Toronto, ON M5B 1W8, Canada; amol.verma{at}mail.utoronto.ca

## Abstract

Background Peripherally inserted central catheters (PICC) are among the most commonly used medical devices in hospital. This study sought to determine the appropriateness of inpatient PICC use in general medicine at five academic hospitals in Toronto, Ontario, Canada, based on the Michigan Appropriateness Guide for Intravenous Catheters (MAGIC).

Methods This was a retrospective, cross-sectional study of general internal medicine patients discharged between 1 April 2010 and 31 March 2015 who received a PICC during hospitalisation. The primary outcomes were the proportions of appropriate and inappropriate inpatient PICC use based on MAGIC recommendations. Hospital administrative data and electronic clinical data were used to determine appropriateness of each PICC placement. Multivariable regression models were fit to explore patient predictors of inappropriate use.

Results Among 3479 PICC placements, 1848 (53%, 95% CI 51% to 55%) were appropriate, 573 (16%, 95% CI 15% to 18%) were inappropriate and 1058 (30%, 95% CI 29% to 32%) were of uncertain appropriateness. The proportion of appropriate and inappropriate PICCs ranged from 44% to 61% (p<0.001) and 13% to 21% (p<0.001) across hospitals, respectively. The most common reasons for inappropriate PICC use were placement in patients with advanced chronic kidney disease (n=500, 14%) and use for fewer than 15 days in patients who are critically ill (n=53), which represented 14% of all PICC placements in the intensive care unit. Patients who were older, female, had a Charlson Comorbidity Index score greater than 0 and more severe illness based on the Laboratory-based Acute Physiology Score were more likely to receive an inappropriate PICC.

Conclusions Clinical practice recommendations can be operationalised into measurable domains to estimate the appropriateness of PICC insertions using routinely collected hospital data. Inappropriate PICC use was common and varied substantially across hospitals in this study, suggesting that there are important opportunities to improve care.

• hospital medicine
• health services research
• quality measurement

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

## Introduction

Peripherally inserted central catheters (PICC) are among the most commonly used devices for vascular access in hospitals.1 2 In the USA, an estimated 350 million PICCs are placed each year.1 PICCs can be safely inserted in the arm to facilitate intravenous access, but they are not without complications. PICC-associated bloodstream infections range from 2.1 to 4.8 occurrences per 1000 catheter days and deep vein thromboses occur in 2.5%–4.9% of patients with a PICC.3 PICC placement also requires substantial resources including materials, human resources and, often, use of an angiography suite and fluoroscopy.4 5 Inserting and maintaining one PICC was estimated to cost US$690 per person compared with US$237 for peripheral catheters.4 Given these complications and costs, it is important to ensure that PICCs are used appropriately.

The Michigan Appropriateness Guide for Intravenous Catheters (MAGIC) describes appropriate and inappropriate criteria for PICC use based on an international expert panel.2 A large multicentre study in Michigan has highlighted that PICC use varies across hospitals,6 and PICC placement is often inappropriate, particularly because of short duration7 or use in patients with chronic kidney disease.8 An intervention based on MAGIC criteria modestly reduced inappropriate PICC use at one site.9 However, the appropriateness of PICC use outside of Michigan has not been well described, and patient characteristics associated with inappropriate use are not known.

The purpose of this study was to describe the appropriateness of inpatient PICC use in general medicine at five hospitals in Toronto, Ontario, Canada, and identify patient-level predictors of inappropriate use. We hypothesised that inappropriate PICC use would be common and would vary across hospitals.

## Methods

### Design, setting and participants

This was a retrospective cross-sectional study involving the general medicine services at five teaching hospitals participating in the General Medicine Inpatient Initiative (GEMINI)10 in Toronto, Ontario, Canada. The participating hospitals are large, urban, tertiary-care hospitals affiliated with the University of Toronto. They are publicly funded and serve diverse multiethnic populations. General medicine care at these hospitals is delivered primarily on clinical teaching units, with attending staff physicians supervising teams of residents and medical students. PICC placement at participating hospitals requires a physician order and cannot be performed solely at the discretion of nurses. PICC placement was reliably ascertained from radiology records, as PICCs were inserted by interventional radiology at participating sites (see online supplementary table 1 for further details). Midline catheters or ultrasound-guided peripheral intravenous catheters were not routinely used at study hospitals.

### Supplemental material

We included general medicine inpatients who were discharged between 1 April 2010 and 31 March 2015 and received a PICC in hospital. Medication data were needed to assess appropriateness of PICC use but were missing prior to 2012 at one institution. Therefore, we excluded 579 PICC placements from the institution during that time (representing 12% of all PICC placements). General medicine patients in participating hospitals have previously been described in detail.10–12 They represent nearly 40% of emergency department admissions and 25% of hospital bed days and are broadly similar to other general medicine inpatient populations in Canada, Europe and the USA.10

### Data collection

GEMINI collects administrative and clinical data from hospital information systems for all general medicine inpatients and has previously been described in detail.10 Patient characteristics, including demographics and comorbidities, were collected from hospitals as reported to the Canadian Institute for Health Information for the Discharge Abstract Database. Based on International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Canada diagnostic codes, each patient was assigned a Charlson Comorbidity Index score.13 Data pertaining to intensive care utilisation, interventional radiology procedures, laboratory testing and in-hospital medications were extracted from each hospital. Laboratory data were used to calculate the Laboratory-based Acute Physiology Score (LAPS), a measure of illness severity that predicts in-hospital mortality.14 15 We did not directly collect data about the indication for PICC placement, because this is not uniformly documented in medical records and could not be ascertained retrospectively.

### Measures and outcomes

The primary outcomes were the proportion of PICCs categorised as appropriate and inappropriate based on in-hospital use. We were not able to collect data about posthospital PICC use, and therefore focused our analysis on inpatient use. Because of poor documentation regarding PICC removal, we assumed that once inserted, PICCs remained in situ for the duration of the hospital admission. This likely resulted in an overestimation of the duration of PICC use, which would underestimate inappropriate use.

We examined variations between hospitals in appropriateness of PICC use. We limited our analysis to the first PICC insertion for all admissions and categorised this according to MAGIC criteria. We organised the MAGIC recommendations into four domains, related to different aspects of PICC use: (1) infused medications, (2) phlebotomy, (3) critical illness, and (4) chronic kidney disease (figure 1). The first three domains reflect indications for PICC placement. For each domain, MAGIC criteria were used to categorise every PICC placement as ‘appropriate’ or ‘inappropriate’. Some PICC placements could not be categorised as appropriate or inappropriate because either the MAGIC criteria were ambiguous (eg, MAGIC does not comment on appropriateness of PICC use for infrequent phlebotomy for 6 or more days) or relevant data were not available (eg, no creatinine value measured). In these cases, the PICC placement was categorised as ‘uncertain’. See online appendix table 2 for further details about categorisation. Thus, each PICC placement was assigned to one appropriateness category for each of the four domains. Finally, each PICC placement received a single appropriateness categorisation based on all four domains. PICCs were considered ‘appropriate’ if they met at least one appropriate criterion for the three domains pertaining to indications for PICC placement and were not inappropriate in any category. PICCs were considered ‘inappropriate’ if they met at least one inappropriate criterion. PICCs were considered ‘uncertain’ if they met no criteria for appropriate or inappropriate use in any domain.

Figure 1

Appropriateness of PICC placements, categorised according to Michigan Appropriateness Guide for Intravenous Catheter (MAGIC) criteria. Non-peripherally compatible medications were identified based on published recommendations of the Infusion Nurses Society (online supplementary appendix table 3).28 Frequent phlebotomy was defined as occurring every 8 hours or more frequently, as per MAGIC recommendations.2 Estimated glomerular filtration rate was calculated using the Modification of Diet in Renal Disease (MDRD) study equation.29 The PICC insertions in these categories were considered uncertain for the following reasons: (A) we could not determine how long the PICC was used after hospital discharge; (B) MAGIC does not comment on appropriateness of PICC use for infrequent phlebotomy for 6 or more days; (C) the PICC was not used for phlebotomy or the duration of use was uncertain; (D) we can only say PICC use was not inappropriate based on kidney function; (E) creatinine not available. CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; ICU, intensive care unit; PICC, peripherally inserted central catheter.

To examine patient predictors of inappropriate PICC use, we controlled for physician and hospital-level effects. Each hospital admission was attributed to a single ‘Most Responsible Physician’ as defined by the Canadian Institute for Health Information as the physician who is ‘responsible for the care and treatment of the patient for the majority of the visit to the health care facility’,16 a definition that has been widely used in health services research.17–20 Data regarding physician characteristics (gender, years of practice) were collected from the publicly available register of the College of Physicians and Surgeons of Ontario.21

### Statistical analysis

We report the proportion of PICCs that were categorised as appropriate, inappropriate and uncertain at each hospital. Two-sided 95% CIs for each proportion were calculated using the Wilson procedure with a correction for continuity.22 The statistical significance of hospital-level differences was tested using χ2 tests. We next examined patient factors associated with inappropriate PICC use, using a multivariable generalised estimating equation model to account for the clustering of multiple PICC placements by individual physicians and to control for measured physician characteristics. We accounted for hospital-level differences by adjusting for hospital fixed effects at the level of each PICC insertion. All analyses were performed using R V.3.5.0 (R Foundation for Statistical Computing).

## Results

There were 101 660 hospital admissions during the study period, and we included 3479 PICC placements. The mean patient age was 65 years (SD 18 years) and 46% were female (table 1). There were 88 physicians with at least 10 PICC placements, and on average they each placed 37 PICCs (SD 19).

Table 1

Patient characteristics in the study cohort, categorised by appropriateness of PICC placement

### Appropriateness of PICC placements

Based on infused medications, 2222 (64%, 95% CI 62% to 65%) PICCs were considered appropriate and 16 (0.5%, 95% CI 0.3% to 0.8%) were inappropriate (figure 1). PICC insertion during an intensive care unit (ICU) admission occurred in 389 admissions and 93 (24%, 95% CI 20% to 29%) of these PICC placements were appropriate based on use for 15 or more days in the ICU. There were 296 (76%, 95% CI 71% to 80%) PICC placements that were used for fewer than 15 days in the ICU, which would not be recommended by MAGIC. However, 243 of these PICC placements (62% of PICCs placed in ICU, 95% CI 57% to 67%) were categorised as appropriate based on their use after the patient was discharged from the ICU to a ward setting. The remaining 53 (14%, 95% CI 10% to 18%) PICC placements were considered inappropriate, and of these, 32 were used for fewer than 5 days in the ICU and 21 were used for 6–14 days in the ICU.

PICCs were appropriately used for frequent phlebotomy in 281 (8%, 95% CI 7% to 9%) cases and inappropriately used for phlebotomy in 34 (1%, 95% CI 0.7% to 1.4%) cases. The appropriateness of PICC placements was uncertain in 847 (24%, 95% CI 23% to 26%) cases, because MAGIC criteria do not comment on non-frequent phlebotomy for more than 5 days. PICCs were inappropriately placed in 500 (14%, 95% CI 13% to 16%) patients with stage 3b or greater chronic kidney disease. Six patients (0.2%, 95% CI 0.1% to 0.4%) did not have creatinine measured in hospital.

Overall, 1848 (53%, 95% CI 51% to 55%) PICC placements were appropriate according to MAGIC criteria, 573 (16%, 95% CI 15% to 18%) were inappropriate and 1058 (30%, 95% CI 29% to 32%) were of uncertain appropriateness. In total, four PICCs (<0.01%, 95% CI <0.01% to <0.01%) could not be categorised in any domain due to missing data.

The proportion of appropriate PICCs ranged from 44% to 61% across hospitals (p<0.001, table 2) and the proportion of inappropriate PICCs ranged from 13% to 21% (p<0.001, table 2).

Table 2

Appropriateness of PICC use at participating hospitals

### Predictors of inappropriate PICC placement

In a multivariable model, patients were more likely to receive an inappropriate PICC if they were older, female, had a Charlson Comorbidity Index score greater than 0, a greater LAPS illness severity score or were in the ICU at the time of PICC placement (table 3). The odds of inappropriate PICC placement were significantly higher at one hospital, after controlling for patient and physician factors.

Table 3

Predictors of inappropriate PICC placement, estimated using multivariable regression

## Discussion

This multicentre observational study involving nearly 3500 PICC placements across more than 100 000 general medicine hospital admissions at five hospitals in Toronto, Ontario, Canada, has several key findings. First, we found that only 53% of PICC placements were definitively appropriate and 16% were inappropriate according to the MAGIC guidelines.2 Second, we identified wide variation in the appropriateness of PICC placements across hospitals, with inappropriate use ranging from 13% to 21%. Third, we found that the most common causes of inappropriate PICC use were placement in patients with chronic kidney disease (stage 3b or greater) and placement for fewer than 15 days in the ICU. The most common indication for appropriate PICC placement was delivery of infused medications. Fourth, we identified patient predictors of inappropriate PICC use, which shed light on opportunities for future research. We were able to operationalise MAGIC recommendations into domains that could be measured using routinely collected data. The high rates of inappropriate use and hospital variation suggest that there are substantial opportunities to improve the appropriateness of PICC use in general hospital medicine.

Since the MAGIC guidelines were published in 2015,2 several papers have examined the appropriateness of PICC use based on a large multicentre study of hospitals participating in the Michigan Hospital Medicine Safety Consortium.6–8 23 The indications for PICC placement, patterns of use and PICC-associated complications varied across hospitals in Michigan (eg, PICC complications varied from 4% to 36%).6 Short-term use of PICCs was found to be common and associated with both patient and hospital factors.7 Advanced chronic kidney disease (stage 3b or greater, estimated glomerular filtration rate (eGFR) less than 45 mL/min/1.73 m2) affected 23% of patients who had PICCs placed.8 At one Michigan hospital, a resource-intensive multimodal intervention based on MAGIC recommendations led to a small reduction in inappropriate PICC use.9 Our research builds on this existing literature. To our knowledge, the appropriateness of PICC placements using MAGIC recommendations has not been studied outside of Michigan and patient characteristics associated with inappropriate use have not been described.

Inappropriate PICC placement among patients with advanced chronic kidney disease was lower in our study (14%) than in the Michigan cohort, but still occurred at a substantial rate. This is important because PICC placement is associated with lower likelihood of transition to a working fistula or graft for dialysis access.24 A portion of the observed difference between our study and previous work might be explained in the sampling strategy, as the Michigan-based study sampled patients directly from the ICU, whereas we only included ICU patients who were also cared for on a general medicine service before and/or after ICU admission.

Patients who were older, female, had higher levels of comorbidity and more severe illness were more likely to receive an inappropriate PICC. These patients may have more difficult-to-obtain vascular access or a more urgent perceived need for vascular access. Clinicians may therefore be more likely to recommend PICC insertions in these patients, even when their use is not considered strictly appropriate. An alternative explanation may be that women and older patients typically have lower muscle mass and therefore have lower eGFRs at creatinine levels that appear within the normal range or less profoundly elevated. Given that eGFR is not always reported along with creatinine measurements, some clinicians might systematically underestimate the degree of kidney disease among women and older patients, and thus may be more likely to inappropriately recommend PICC insertion in these patients. Further research should validate patient predictors of inappropriate PICC use and explore the reasons for inappropriate use in specific patient groups. This may shed light on opportunities for and barriers to quality improvement.

A major strength of our study is the use of routinely collected clinical and administrative data, extracted from hospital electronic information systems, to examine appropriateness of PICC use. By extracting granular clinical information regarding radiology procedures, laboratory test utilisation and medication use, we were able to provide data about the appropriateness of more than 70% of all PICC placements. Given the high costs of prospective data collection,9 and the proliferation of ‘big data’,25 the approach articulated in this manuscript may represent a scalable method of monitoring the appropriateness of PICC use to support quality improvement. This approach does have several important limitations. First, we were unable to collect data about the specific indication for PICC placement. Instead, this was inferred based on the way the PICC was used in hospital. It is possible that some PICCs were placed with the intention of appropriate use (eg, need for medication infusion for a prolonged duration), but retrospective analysis revealed use that was ultimately inappropriate (eg, the medication was discontinued earlier than originally expected). This would overestimate inappropriate use. However, the major cause of inappropriate use was chronic kidney disease, and this would have been knowable at the time of PICC placement and remains an important opportunity for quality improvement. Second, we only collected data during the hospitalisation, and therefore we do not comment on the appropriateness of PICC use after hospitalisation. Third, we were unable to collect data on PICC removal, as this is almost entirely performed by a clinician at the bedside rather than by the radiology department and is poorly documented. We assumed that PICCs were inserted for the duration of the hospital admission. In our study, only 35 (1%) PICC placements were identified as having a duration of use of 5 days or less and 36% of PICC placements occurred within 5 days of hospital discharge. Paje and colleagues previously reported that 25% of PICCs had a dwell time of 5 days or shorter and that nearly all of these (96%) were removed during hospitalisation.7 Therefore, our study almost certainly underestimates the proportion of PICCs inserted for a short duration and underestimates inappropriate PICC use. Fourth, we used the minimum inpatient creatinine value to calculate eGFR as a proxy for baseline renal function. This approach is reasonable in patients without prehospital data26 but may overestimate eGFR27 and underestimate chronic kidney disease. Finally, this study was conducted at five teaching hospitals in Toronto where the primary method of PICC insertion was through interventional radiology procedures. Other methods, such as bedside insertion by nurses, may be more common at other hospitals. This may limit the generalisability of our methods to identify PICC insertions but is unlikely to affect the generalisability of our estimates of appropriateness. We found a high rate of PICC use among patients with advanced chronic kidney disease similar to studies in Michigan, suggesting our findings may generalise to other practice settings.

Directions for future research include external validation of our methods to assess whether routinely available data can reliably measure appropriateness of PICC use. Further research is needed to explain why large variations in appropriateness were seen across hospitals and to characterise variation at the clinician and team levels. There appear to be important opportunities to improve the appropriateness of PICC use. Quality improvement initiatives may include promoting dissemination of MAGIC, expanding availability of alternatives to PICC use (such as midline catheters) and focusing on the most common causes of inappropriate use (chronic kidney disease and use in ICUs).

In conclusion, inappropriate PICC use was common and varied widely across five hospitals in Toronto, Canada. The most common reasons for inappropriate PICC use were placement in patients with advanced chronic kidney disease and use for short durations in the ICU. Routinely collected electronic clinical data may enable cost-efficient measurement of the appropriateness of PICC use. Given the costs and potential complications associated with PICC placement, targeted interventions to reduce inappropriate PICC use may represent important quality improvement opportunities.

## Acknowledgments

We thank Dr Faten Sallam and Dr Radha Koppula for data abstraction to identify PICC placements. Both were paid research assistants for this work.

• ## Supplementary Data

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

## Footnotes

• Contributors The study was designed by AAV, AK, SS and FR. YG and HYJ conducted the statistical analysis. AAV, YG, HYJ, SR, LLS, JLK, AW, TT and FR contributed to data collection. AAV wrote the first draft of the manuscript and all authors contributed to critical interpretation of the analysis and revised the manuscript for important intellectual content.

• Funding This study was funded by Green Shield Canada Foundation and Mak Pak Chiu and Mak-Soo Lai Hing Chair in General Internal Medicine.

• Competing interests None declared.

• Patient consent for publication Not required.

• Ethics approval This study was approved by the research ethics board at all participating hospitals.

• Provenance and peer review Not commissioned; externally peer reviewed.

• Data availability statement Data are available upon reasonable request. The study’s lead investigators will make data for this manuscript available upon request as possible in compliance with local research ethics board requirements and data sharing agreements.