key: cord-0861540-jrhkrbf6
authors: Newsome, Andrea Sikora; Murray, Brian; Smith, Susan E; Brothers, Todd; Al-Mamun, Mohammad A; Chase, Aaron M; Rowe, Sandra; Buckley, Mitchell S; Murphy, David; Devlin, John W
title: Optimization of critical care pharmacy clinical services: A gap analysis approach
date: 2021-06-01
journal: Am J Health Syst Pharm
DOI: 10.1093/ajhp/zxab237
sha: efb216920e5c53e1b55cbe6ae724f78989048e86
doc_id: 861540
cord_uid: jrhkrbf6

In an effort to expedite the publication of articles related to the COVID-19 pandemic, AJHP is posting these manuscripts online as soon as possible after acceptance. Accepted manuscripts have been peer-reviewed and copyedited, but are posted online before technical formatting and author proofing. These manuscripts are not the final version of record and will be replaced with the final article (formatted per AJHP style and proofed by the authors) at a later time.

A c c e p t e d M a n u s c r i p t However, this ratio is not supported by rigorous research; ratios as low as 1:8 and as high as 1:30 have been proposed as being better. [5] [6] [7] [8] [9] In the United States, 2,800 board-certified CCPs are currently available to provide care to the 100,000 ICU beds that are filled with 5 million patients per year. There is currently a substantial shortage of qualified CCPs to provide an optimal level of pharmacotherapeutic care to critically ill Americans. [10] [11] [12] When surveyed, 27% of CCPs self-reported working at a CCP-to-ICU patient ratio of 1:18, and another 25% reported working at a ratio greater than 1:30. 13 In a more recent survey of 441 US hospitals, a median ratio of 1:17 (interquartile range, 1:12 to 1:26) was reported. 2 In this same survey, only 70.8% of institutions had direct clinical CCP services in the ICU, CCP coverage was limited on nights and weekends, and CCPs frequently had nondirect patient care responsibilities. 2 Indeed, few institutions provide 24-7-365 clinical CCP coverage, despite the ever-increasing complexity of ICUs and calls for greater clinical CCP services during the current global pandemic. 14 As a result, general practice pharmacists, without formal CCP training, may be placed in the ICU despite evidence suggesting that a lack of ICU experience may be associated with fewer clinically important interventions and fewer optimal-level clinical CCP activities. [15] [16] [17] [18] Research surrounding the current staffing challenges faced by critical care physicians, advanced practice providers, and nurses, the risk for clinician burnout that may result from this, and the impact on care quality is informative for CCP staff justification. [19] [20] [21] [22] [23] [24] [25] Increased nursing workload as a result of reduced staffing is associated with greater patient mortality, particularly in those patients who are most seriously ill. 25 These reports have resulted in calls for federal mandates to standardize ICU nurse-to-patient care ratios. 26 Up to 60% of CCPs report experiencing burnout; increased roles and responsibilities are the most commonly reported factors. 15, [27] [28] [29] However, few studies exist regarding CCP staffing ratios and burnout. Survey results suggest that CCPs perceive higher CCP-to-ICU patient ratios to be associated with worse quality of care. 13 This moral distress has previously been shown to be associated with clinician burnout.

A c c e p t e d M a n u s c r i p t Optimal CCP clinical services. The first step to addressing the current CCP staffing gap is to more clearly define what is meant by "optimal" with respect to CCP clinical services in ICUs. We propose the optimal staffing model to be a function of identifying a CCP-to-ICU patient ratio that maximizes a composite of 3 key domains: patient outcomes, healthcare costs, and CCP welfare ( Figure 1 ). Within this model, individual pharmacist time is appropriately allocated among ICU responsibilities, including direct patient care, quality improvement, teaching, and outside (the ICU) leadership and service activities. Clearly delineated relationships between ICU census, patient need for CCP services, and the CCP staffing time required to deliver these services would allow for institutions to develop staffing models that fit their unique needs. Globally, as our knowledge of CCP-to-ICU patient ratio individualization continues to evolve and CCP staffing models improve, reduced medication-associated harm, healthcare waste, and CCP burnout will ideally ensue.

Gaps to achieving optimal critical care pharmacy clinical services. No study has formally evaluated how the CCP-to-ICU patient ratio affects the number or quality of medication interventions, patient outcomes, healthcare costs, or pharmacist well-being. Indeed, the relationships that define the optimal CCP-to-ICU patient ratio, the quality of CCP care, the time required to deliver that care, and the resulting ICU patient-related outcomes remain poorly characterized, yet, a priori, these causative relationships are likely present. Among ICUs in the United Kingdom, research focused on pharmacist staffing resources, time spent by pharmacists in the ICU conducting direct patient activities, and the quality and impact of CCP interventions achieved support the hypothesis that the CCP-to-ICU patient ratio and quality of care are interrelated. 16, 18, 30, 31 However, the application of this research to US CCPs is limited given that pharmacist practice models are different, healthcare reimbursement is distinct, and CCP training is more structured. While many before-and-after studies have clearly shown that the addition of a CCP to the ICU interprofessional team improves patient outcomes, the intensity at which clinical CCP services should be delivered to optimize patient care outcomes remains unclear. 32-39 In other words, while the presence of nearly any CCP service is better than the lack of a CCP presence altogether, the A c c e p t e d M a n u s c r i p t incremental relationship of workload to outcome is unknown. Key evidence gaps to address include the following:

 If a CCP has fewer ICU patients for whom to provide clinical care, does their effectiveness improve?

 If CCP clinical services are delivered on evenings and/or weekends, are ICU patient outcomes further optimized?

 What factors most influence the optimal clinical CCP-to-ICU patient ratio that should be delivered (eg, severity of illness, number and/or type of medications)?

 Are there additional CCP responsibilities with equally positive effects on patient outcomes that might influence this staffing ratio (eg, teaching, quality improvement, or medication order validation)?

 Does a synergistic relationship exist between the delivery of direct patient care activities, indirect patient care responsibilities, service roles (eg, institutional committees, supervisor roles, national organizations), and education of students, trainees, and colleagues? Indeed, while it has been established that clinician-scientists fill a vital translational gap with regard to scientific discovery owing to the nature of the "practice-oriented" questions they can ask, this concept is less formally established for clinicians but may have similar ramifications with regard to quality improvement. 40 These knowledge gaps prevent pharmacy administrators from being able to individualize CCP scheduling to a specific ICU patient ratio goal that will maximize return on investment.

Importantly, a failure to be able to fully optimize the efficiency of this expensive resource may serve as a barrier to CCP service expansion and have the unintended consequence of dictating CCP practice models by available resources as opposed to clinically oriented patient outcomes guiding the need for new resource allocation. This approach to CCP modeling precludes allowing desired ICU clinical outcomes to drive, refine, and optimize the CCP model best suited to a particular ICU. A backwards-design approach may be more optimal; one such approach is depicted in Figure 2 .

A c c e p t e d M a n u s c r i p t Formalization of on-and off-service periods (as in many intensivist staffing models) may be beneficial to consider for CCPs. With many CCPs having multiple roles in the context of their primary direct ICU patient care role, a reasonable CCP-to-patient ratio would allow for both high-quality patient care and the completion of other nondirect care activities. Key knowledge gaps to optimizing CCP staff models and global resolution strategies are presented in Table 1 .

approach. To effectively conduct a meaningful trial comparing clinical CCP practice models, outcome measures and covariates must be thoughtfully defined. A reliable metric able to define and incorporate the key domains and outcomes of CCP performance (eg, the institutional resources to hire and support CCPs and the influence of these pharmacists on patient outcome and medicationrelated healthcare costs) has a strong potential to nurture the development of robust CCP practice models. However, the tools and modes that have been developed to date each have notable limitations ( Table 2) .

Medication regimen complexity has been proposed as a quantifiable metric designed to connect all components of the optimal CCP practice model, including ICU patient outcomes, healthcare costs, pharmacist welfare, and pharmacist resources. 3, [68] [69] [70] [71] [72] [73] [74] [75] The MRC-ICU scoring tool is the first validated, objective method to measure medication regimen complexity solely in critically ill adult patients. 71, 76 Adapted from methods used for the medication regimen complexity index (MRCI), the MRC-ICU scoring tool was developed and validated in a single-center study of medical ICU patients and demonstrated an appropriate construct, a convergent, discriminant, and internal validity. 71 Subsequently, a prospective study in both medical and surgical ICU populations was conducted at an academic medical center and a community teaching hospital to confirm initial external validity. 75 The MRC-ICU score has also been evaluated in a multicenter study of 4,052 patients and 28 institutions, and preliminary analysis from the published abstract showed promise, with each increase of 1 point in MRC-ICU demonstrating 2% increased risk of mortality following Cox A c c e p t e d M a n u s c r i p t regression. 70 An abbreviated version of MRC-ICU, with 17 (vs 39) line items was then validated in an effort to simplify future information technology (IT) builds. 68 To objectively quantify pharmacist activity, MRC-ICU at 24 hours has also been correlated with CCP interventions at 24 hours and cumulative interventions at discharge. 74 Drug-drug interactions (DDIs) correlated with both the MRC-ICU score at 24 hours. 27 Notably, correlation with potential DDIs was an expected finding as the rate of DDIs is known to increase as the total number of medications increases, thus serving as another validity confirmation but also suggesting a potential clinical use to identify high-risk patients. 77, 78 For a construct such as medication regimen complexity to be effective, evaluation of the relationship with "real-time" events in which a CCP can intervene is necessary. In a multicenter, retrospective study, linear regression controlling for gender, age, and weight demonstrated that the MRC-ICU scoring tool has an association with fluid overload (as defined as a positive cumulative fluid balance that would be expected to produce weight gain of over 10% from baseline). 73 This finding is promising as fluid overload is a common adverse drug event in ICU patients associated with poor outcomes but also has a documented role for CCP intervention and benefit. 41, 79, 80 Potential roles for a metric like MRC-ICU include providing resource predictions from a hospital administrative perspective for clinical CCP position justification, real-time guidance to establish optimal CCP-to-ICU patient ratios, and clinician-oriented information for prioritization of those critically ill adults who are at greatest risk for unfavorable medication-related outcomes (eg, fluid overload). Ideally, by applying this tool in a wide variety of institution types and ICU settings without losing translatability, a common language to investigate and establish best pharmacy practice models can be established. A potential construct for applying such a metric may be where resources are objectively predicted by an algorithm and patient outcomes are evaluated as a result of this staffing model, allowing for future site-specific modifications. In such a construct, patientspecific data and a pharmacy-based metric would be fed into a predictive model that provides both A c c e p t e d M a n u s c r i p t predictions for patient outcomes (eg, mortality, fluid overload) and pharmacist resources per patient (eg, number of interventions, time per intervention, etc). At this juncture, institution-specific needs (eg, unit census, non-patient care responsibilities of the pharmacist) would be incorporated to identify a reasonable pharmacist-to-patient ratio and associated staffing model. This staffing model would then be evaluated by comparing actual patient outcomes and pharmacist activities against the tool's original predictions to allow for further optimization (Figure 3) .

A proof-of-concept model was developed using machine learning (ML) methods to validate the MRC-ICU score, and, despite a small dataset, the results suggest that prediction of patient outcomes can be improved with inclusion of medication regimen complexity data and the MRC-ICU score in comparison to the traditional predictor of acute physiology and chronic health evaluation (APACHE) III. 72 ML methods have shown promise in the domains of both critical care and pharmacy, but full application of a pharmacy-based metric for critical care pharmacy model optimization within a clinical decision support system (CDSS) requires rigorous prevalidation utilizing real-time scenarios. 81, 82 Indeed, use of artificial intelligence and ML predictive algorithms facilitates the interpretation and application of complex information to real-time data and may be the future of predicting CCP resource needs. ML tools have transformed healthcare data to predict, diagnose, and treat disease. Recently, CDSS improvements have begun to support complex decision-making with the assistance of IT, data mining, and ML methods. These methods have been developed and used to predict severe sepsis and septic shock, early warning of sepsis, 83,84 unplanned transfer to the ICU, 85 ICU discharge, 86 acute injury risk, 87 complications in the critical care setting, 88 and hospital mortality. 89 While prediction model interpretation provided by CDSS requires substantial testing and validation, including integration of expert knowledge, to improve its overall accuracy, application of CDSS utilizing pharmacy-based metrics has the potential to provide much-needed insight as an alternative analytic approach to handle complex interactions and may elucidate hidden patterns that generate clinically justified predictions. 90, 91 A c c e p t e d M a n u s c r i p t Quality improvement approach. Institution-based evaluation and quality improvement efforts focused on potential clinical CCP staffing gaps are important. Available resources to justify additional CCP resources are reviewed elsewhere. 3 In the only study of its kind, one institution employed a mixed methods approach to justify the addition of a second full-time CCP in a medical ICU. 92 While the complexity of interventions (as defined by a perceived quality ranking) was not observed to increase, the absolute number of desirable and optimal clinical activities increased from baseline and a multiprofessional focus group of medical ICU team members qualitatively described the additional coverage as beneficial. Figure 4 provides recommendations for institution-based evaluation of current CCP services.

which exposes critically ill patients to worse outcomes and increased healthcare costs. Regardless of an institution's current status, incremental improvements to optimize current and future resources are possible. Use of a quantifiable and externally validated metric that allows for cross-institution and cross-patient population evaluation of patient outcomes, healthcare costs, pharmacist welfare, and pharmacist resources has strong potential for optimizing CCP clinical services. While closing the gap between current and optimal CCP clinical services will require deliberate action and a coordinated effort, CCP-delivered care for every critically ill patient is a deeply worthwhile vision.

Dr. Newsome has received research funding through the National Center for Advancing Translational A c c e p t e d M a n u s c r i p t Following the dictum "begin with the end in mind," pharmacy practice models would be dictated by clinically oriented patient needs; however, the ability to predict these needs with regard to critical care pharmacist resources is a vital knowledge gap. M a n u s c r i p t that likely prevent the need for direct "interventions."  Value tracking supports the value of pharmacist services, but substantial limitations are present, notably the difference between cost avoidance and cost savings and low-quality data. [45] [46] [47] [48] The metric of "cost avoidance to pharmacist salary" may have more quickly diminishing ROI than using patient-centered outcomes (eg, length of stay).

 Real-time documentation of CCP interventions that occurs concurrently with patient care and includes both direct and indirect interventions  While tracking the value of CCPs is essential for staffing justification, more robust foundational data and not "widget counting" are needed to lay the groundwork for more comprehensive and reliable evaluations of the holistic nature of clinical pharmacy services.

The potential for CCP workload to result in pharmacist burnout not considered  Practice models do not account for the high practice standards set by guidelines and may contribute to CCP BOS, which has higher prevalence across the whole critical care discipline and contributes to worse patient safety outcomes. 29,49-52  Investigation of factors associated with CCP BOS (eg, lack of off-service time, vacation coverage, patient volume) 50,53  The link between CCP practice models, CCP-to-ICU patient ratios, ICU patient outcomes, and BOS requires exploration. Limited ability to compare CCP practice models among institutions

Position paper on critical care pharmacy services: 2020 update

Characterizing critical care pharmacy services across the United States

Justification of the value of critical care pharmacists: still a work in progress?

Action plan for successful implementation of optimal ICU pharmacist activities: next steps for the critical care pharmacist position paper

Effects of a hospitalwide pharmacy practice model change on readmission and return to emergency department rates

Faculty of Intensive Care Medicine, Intensive Care Society. Core Standards for Intensive Care Units

Pharmacy practice model transformation from medication focus to patient centered care. Presented at: ASHP Midyear Clinical Meeting

SHPA Committee of Specialty Practice in Clinical Pharmacy. SHPA standards of practice for clinical pharmacy

The critical care clinical pharmacist: evolution of an essential team member

Trends in critical care beds and use among population groups and Medicare and Medicaid beneficiaries in the United States

Board of Pharmacy Specialties. Find a board certified pharmacist

A survey of critical care pharmacists to patient ratios and practice characteristics in intensive care units

Health care delivery and pharmacists during the COVID-19 pandemic. Patient Safety Network

Examination of critical care pharmacist work activities and burnout

Reducing medication errors in critical care patients: pharmacist key resources and relationship with medicines optimisation

Measuring the impact of a pharmacist in the intensive care unit-are all pharmacists created equal?

Patient acuity indicators as predictors of pharmacy workload

Using a nursing-workload index to validate hospital pharmacy productivity

Method to determine allocation of clinical pharmacist resources

An evaluation of acuity adjustment metrics to track medication expense over time

Effective use of workload and productivity monitoring tools in health-system pharmacy, part 2

Effective use of workload and productivity monitoring tools in health-system pharmacy, part 1

Development and validation of the medication regimen complexity index

Medication regimen complexity measured by MRCI: a systematic review to identify health outcomes

Characterization of changes in medication complexity using a modified scoring tool

Pharmacists are associated with reduced mortality in critically ill patients: now what?

Pharm-crit: medication regimen complexity in the ICU (MRC-ICU) as a predictor of inpatient mortality

Development and validation of a medication regimen complexity scoring tool for critically ill patients

Medication regimen complexity score as an indicator of fluid balance in critically ill patients

Medication regimen complexity is associated with pharmacist interventions and drug-drug interactions: a use of the novel MRC-ICU scoring tool

Multi-center validation of a novel medication regimen complexity scoring tool

Polypharmacy and potential drug-drug interactions in emergency department patients in the Caribbean

The impact of drug interactions and polypharmacy on antimicrobial therapy in the elderly

Fluid stewardship during critical illness: a call to action

Association of hidden fluid administration with development of fluid overload reveals opportunities for targeted fluid minimization

Demystifying artificial intelligence in pharmacy

Artificial intelligence in the intensive care unit

Multicenter comparison of machine learning methods and conventional regression for predicting clinical deterioration on the wards

Clinician perception of a machine learningbased early warning system designed to predict severe sepsis and septic shock

Predicting unplanned transfers to the intensive care unit: a machine learning approach leveraging diverse clinical elements

Towards a decision support tool for intensive care discharge: machine learning algorithm development using electronic healthcare data from MIMIC-III and Bristol

The development of a machine learning inpatient acute kidney injury prediction model

Machine learning for real-time prediction of complications in critical care: a retrospective study

Predicting hospital mortality for intensive care unit patients: time-series analysis

M a n u s c r i p t  No universal ICU pharmacy practice model can exist given the unique nature of each ICU and each institution (ie, there is no "one size fits all")  Although scores like APACHE III and case-mix indices can provide general comparisons, the nature of the ICU patient population, specialty or focus area of the institution, geographic region, institution size, and other factors can lead to comparisons "between apples and oranges" when discussing the correct CCP-to-ICU patient ratio. A c c e p t e d M a n u s c r i p t key components upon which to base CPS allocation.  After applying this methodology to each medical service, the service receives a score from 1 to 5, with 5 indicating the highest need for a CPS, at which time pharmacists were reallocated. While a broad staffing model for a large academic medical center with an associated school of pharmacy, it is not specific to a critical care population.  Not linked to patient outcomes  Bases patient acuity on DRG  Limited external validity due to its design oriented toward a single academic medical center  Not specific to CCPs PIS: resource-based relative value intensity grouping system that utilizes pharmaceutical resource consumption data to allocate pharmacy personnel  Calculated by multiplying the number of patients with a specific DRG by a specific PIW, giving insight into pharmacy cost and patient acuity  PIW is calculated by comparing the median pharmacy cost for a given DRG with the median pharmacy cost for all DRGs to determine the "intensity" of pharmacy resource use in relation to other diagnoses.  Gives insight into medication expense at an institution and potentially patient acuity at that site 63  Assumes patient acuity is correlated with DRG, which has been shown to not always hold true 64,65  Has only been used to predict expenditure on resources and has not been shown to improve patient outcomes or determine optimal pharmacist-to-patient ratio  Not specific to CCPs MRCI: measure of patient-level MRC through 3 key components: dosage form, dosage frequency, and additional medication directions 66  Developed from the MCI from 134 chronic obstructive pulmonary disorder patient regimens by using an expert panel of 5 researchers scoring 6 regimens to demonstrate construct-and contentrelated validity  Two researchers scored the same 6 regimens to determine interrater and test-retest reliability  Is a patient-oriented outpatient tool intended to screen for community pharmacist clinical services (vs those in the ICU)  Not intended to be related to patient acuity or patient outcomes 67Abbreviations: CCP, critical care pharmacist; CPS, clinical pharmacy specialist; DRG, diagnosis-related group; ICU, intensive care unit; MCI, medication complexity index; MRCI, medication regimen complexity index; pCATCH, census, patient acuity, teaching services, medication cost, and use of high-priority medications; PIS, pharmacy intensity score; PIW, pharmacy intensity weight.A c c e p t e d M a n u s c r i p t A c c e p t e d M a n u s c r i p t A c c e p t e d M a n u s c r i p t A c c e p t e d M a n u s c r i p t Study Objective To explore the benefit of reduced CCP-to-ICU patient ratios at a single institution Intervention Institution-specific change in service coverage (eg, plan to go from 1 to 2 CCPs in a unit, addition of evening services, etc) Primary Outcome Options 

Recommended to include outcomes from multiple categories to allow for exploratory relationships between workload, care quality, patient outcomes, and wellness Parameters  Patient demographics (eg, age, sex, admission diagnosis)  Patient acuity measures (eg, sequential organ failure assessment)  Primary and secondary outcomes as noted above  Institution-specific metrics (eg, medication regimen complexity)  Pharmacist staffing (eg, patients cared for per shift, on-and off-service patients, ICU vs non-ICU responsibilities, shift type) Data Analysis  Descriptive statistics and basic inferential statistics (eg,  2 )  Recommended to include multivariate modeling to be able to adjust for noted confounders such as patient acuity or length of stay