key: cord-0022376-mi5q82ig
authors: Ghannam, Mousa; Malihi, Parasteh; Laudanski, Krzysztof
title: Examination of Electrolyte Replacements in the ICU Utilizing MIMIC-III Dataset Demonstrates Redundant Replacement Patterns
date: 2021-10-14
journal: Healthcare (Basel)
DOI: 10.3390/healthcare9101373
sha: ef490e16786bacec3cfa81e407072980c2cf76ab
doc_id: 22376
cord_uid: mi5q82ig

Electrolyte repletion in the ICU is one of the most ubiquitous tasks in critical care, involving significant resources while having an unclear risk/benefit ratio. Prior data indicate most replacements are administered while electrolytes are within or above reference ranges with little effect on serum post-replacement levels and potential harm. ICU electrolyte replacement patterns were analyzed using the MIMIC-III database to determine the threshold governing replacement decisions and their efficiency. The data of serum values for potassium, magnesium, and phosphate before and after repletion events were evaluated. Thresholds for when repletion was administered and temporal patterns in the repletion behaviors of ICU healthcare providers were identified. Most electrolyte replacements happened when levels were below or within reference ranges. Of the lab orders placed, a minuscule number of them were followed by repletion. Electrolyte repletion resulted in negligible (phosphate), small (potassium), and modest (magnesium) post-replacement changes in electrolyte serum levels. The repletion pattern followed hospital routine work and was anchored around shift changes. A subset of providers conducting over-repletion in the absence of clinical indication was also identified. This pattern of behavior found in this study supports previous studies and may allude to a universal pattern of over-repletion in the ICU setting.

Electrolyte repletion is a ubiquitous routine intervention carried out to maintain a homeostatic range of compounds in serum [1, 2] . An abnormal level of most abundant electrolytes (sodium, chloride, potassium, magnesium, phosphate, and calcium) is linked to heart, coagulation, muscle, and nervous system pathologies, as these electrolytes are critical for electrical signal conduction and enzyme activity [3] [4] [5] [6] [7] .

As with any other therapeutic maneuver, the goal of electrolyte repletion is to improve outcomes while reducing harm [8, 9] . The decision making is based in very significant way on a comparison between electrolyte levels and reference ranges, which are considered physiological [1, 2, 10] . These references ranges are not delineated but rather fuzzy, as there is only some evidentiary support that relative hyperkalemia (potassium level between 4 and 4.5 mEq/dL) can prevent atrial fibrillation after cardiac surgery, but this is not a uniform finding [3, 4, [11] [12] [13] . Relative hypermagnesemia can stabilize the heart's electrical activity, terminate atrial fibrillation, and decrease seizure likehood while relieving bronchospasms in certain conditions when a drug is given as bolus resulting in a rapid increase in electrolytes with normalization afterward [1, 5, 6, 13] . Consequently, medical professionals may take the laboratory value of electrolyte levels in serum in context or subject it to their interpretation. A lack of consensus regarding what the optimal electrolyte

The data set used in this study was from the MIMIC-III critical care database [37] . This database comprises de-identified health records for over forty thousand critical care patients admitted to the ICU in the Beth Israel Deaconess Medical Center (Boston, MA, USA) between 2001 and 2012. In addition, the database includes demographics, vital signs, laboratory test results, medication, and caregiver information for 2008-2012. The dataset was assembled into a relational database (PostgreSQL) and queried as necessary, and a statistical analysis was conducted using R-packages software [38, 39] .

The MIMIC-III database, limited to Metaview data, includes critical care entries containing information on 23,387 ICU stays. The screening process began by excluding ICU stays lacking electrolyte repletion or a lab value (n = 10,978) ( Figure 1 ). Then, similar data exclusion and inclusion criteria were used as in the prior study to minimize heterogeneity using the International Classification of Diseases (ICD-9) codes (Supplementary Table  S1 ) [19] . In this study, patients under 18 years old (n = 19) were excluded. The MIMIC-III database contains a list of diagnoses in order of relevancy for each patient, but only the first diagnosis was selected as the object of data analysis [37] . Exclusions included conditions associated with several electrolyte disturbances complicating the provider's decisionmaking process, resulting in a very complex and potentially unreliable analysis [1, 2, 19] . re 2021, 9, x FOR PEER REVIEW 3 of 16 [19] . In this study, patients under 18 years old (n = 19) were excluded. The MIMIC-III database contains a list of diagnoses in order of relevancy for each patient, but only the first diagnosis was selected as the object of data analysis [37] . Exclusions included conditions associated with several electrolyte disturbances complicating the provider's decision-making process, resulting in a very complex and potentially unreliable analysis [1, 2, 19] . Consequently, events involving the transfusion of packed red blood cells (n = 630), rhabdomyolysis (n = 1167), a parathyroid disease of any kind (n = 65), sarcoid disease (n = Consequently, events involving the transfusion of packed red blood cells (n = 630), rhabdomyolysis (n = 1167), a parathyroid disease of any kind (n = 65), sarcoid disease (n = 62), chronic kidney disease (n = 530), acute kidney disease (n = 33,945), and end-stage renal disease (n = 5300) were eliminated. Additionally, encounters where the patient had coronary artery disease (n = 1255), congestive heart failure (n = 3399), atrial fibrillation (n = 4104), nutritional deficiency (n = 9), and paralysis (n = 171) were excluded from further analysis. Additionally, events where a dialysis procedure was performed were also excluded (n = 187). From the remaining 5275 ICU stays, potassium pre-repletion values below 2 or above 7 (n = 74), magnesium pre-repletion values greater than 5 or less than 1 (n = 68), and phosphate pre-repletion values below 0.5 and above 5 (n = 39) were filtered out as errors.

The nominal reference values in the MIMICS-III database are 3.7-5.7 for potassium, 2.5-4.5 for magnesium, and 2.5-4.5 for phosphate, with values outside of these flagged as abnormal.

In this process, the timeframe for analysis followed a 24-h window from the time when the lab value was drawn. This approach resulted in three different scenarios. In the first instance, a lab value was drawn, but there was no instance of electrolyte replacement after 24 h (L only/No repletion (Figure 2A ). If a singular lab value was followed by a repletion, it was assigned to the second scenario (1L → 1R) ( Figure 2B ). These cases constituted 51.1% of all cases where repletion took place. Finally, in cases where multiple lab values preceded electrolyte replacement in 24-h windows, the closest value was used, and the others were ignored ( Figure 2C ). This was called a multiple lab scenario (ML → 1R), and the frequency of these events was 48.9%. Subsequent electrolyte lab values were searched for within a 24-h window, and the lab value closest to the order placement in all three scenarios were considered. This workflow is a reasonable reflection of routine hospital performance [8, [40] [41] [42] . In the first case, only electrolyte lab draw was obtained without preceding electrolyte replacement, and subsequent supplementation was given within 24 hours (A). The lab draw was preceded by a singular electrolyte replacement (B) or multiple electrolyte replacements (C) in two other scenarios.

The Shapiro-Wilk W test and distribution plots were used to test the normality of distribution variables. The parametric variables are expressed as mean ± SD and compared using t-Student. The effect size was estimated as d-Cohen statistics. Regression analysis was conducted by addition, and a subtraction technique with the ordinary least squares' method was used for all linear regressions. Correlations were reported as Pearson r values. A double-sided p-value less than 0.05 was considered statistically significant In the first case, only electrolyte lab draw was obtained without preceding electrolyte replacement, and subsequent supplementation was given within 24 h (A). The lab draw was preceded by a singular electrolyte replacement (B) or multiple electrolyte replacements (C) in two other scenarios.

The Shapiro-Wilk W test and distribution plots were used to test the normality of distribution variables. The parametric variables are expressed as mean ± SD and compared using t-Student. The effect size was estimated as d-Cohen statistics. Regression analysis was conducted by addition, and a subtraction technique with the ordinary least squares' method was used for all linear regressions. Correlations were reported as Pearson r values. A double-sided p-value less than 0.05 was considered statistically significant for all tests. Statistical analyses were performed with specified R packages [7, 38, 39, [42] [43] [44] [45] [46] [47] .

The blood collection and lab reporting followed certain time patterns with only 3.24% of potassium, 10.37% magnesium, and 16.62% of phosphate lab orders resulting in repletion ( Figure 3A ). Overwhelmingly, orders for the replacements of potassium and magnesium were placed when serum levels were in the nominal ranges for every scenario analyzed here. In contrast, phosphate was replaced more often when lab values indicated hypophosphatemia ( Figure 3B ; Table 1 ). Irrespective of the clinical scenario discussed in this manuscript, the frequencies of the electrolytes below, within, or over the nominal value were similar ( Figure 3B ). In the case of potassium and magnesium repletion, trigger values were within the normal limit but not in the case of hypophosphatemia, which was a more frequent trigger for replacement than the normal phosphate level (B).

When considering instances where repletion took place, the preceding lab value of the electrolytes was mostly within the normal value for potassium and magnesium but not for phosphate ( Figure 4A ; Table S1 ). When the instances where electrolytes were not replaced (1L → NR) versus when they were replaced (1L → 1R and ML → 1R) were compared, the preceding electrolyte level was different. With respect to effect size, repletion mattered little for both potassium and magnesium (Table S1 ). At the same time, phosphate had many more cases where providers decided not to replace at nominal and above reference levels ( Table S1 ). The top ten diagnoses for each electrolyte repletion falling above In the case of potassium and magnesium repletion, trigger values were within the normal limit but not in the case of hypophosphatemia, which was a more frequent trigger for replacement than the normal phosphate level (B). When considering instances where repletion took place, the preceding lab value of the electrolytes was mostly within the normal value for potassium and magnesium but not for phosphate ( Figure 4A ; Table S1 ). When the instances where electrolytes were not replaced (1L → NR) versus when they were replaced (1L → 1R and ML → 1R) were compared, the preceding electrolyte level was different. With respect to effect size, repletion mattered little for both potassium and magnesium (Table S1 ). At the same time, phosphate had many more cases where providers decided not to replace at nominal and above reference levels ( Table S1 ). The top ten diagnoses for each electrolyte repletion falling above and below thresholds were distinct for each electrolyte (Supplementary Figure S1 ; Supplementary  Table S3 ). 

The analysis revealed instances when preceding electrolyte levels were above the limit. The frequency of these instances was 0.38% for potassium, 1.4% for magnesium, and 0% for phosphate (Table S1 ). The relative scarcity of these cases precluded any meaningful analysis, but their clinical characteristic is heterogeneous (Supplementary Table S2 ; Supplementary Figure S1 ).

Most of the electrolyte lab values were taken around 6 a.m., followed by repletion at 9 a.m. in cases where an electrolyte was replaced ( Figure 5A ). On average, the replacement was ordered within 217.20 min (3 h 36 min) for potassium, 323.29 min (5 h 18 min) for magnesium, and 339.83 min (5 h 36 min) for phosphate for the charting of the initial lab values. A follow-up lab value was ordered on average within 443.45 min (7 h 18 min), 643.10 min (10 h 42 min), and 567.13 min (9 h 24 min) for potassium, magnesium, and phosphate, respectively ( Figure 5B ).

The threshold of electrolyte replacement remained steady for 24 h but varied between 1L->1R and 1L->MR versus 1L->NR ( Figure 5B ). In addition, if multiple electrolyte lab values were requested, a different bimodal distribution of lab orders was seen ( Figure 5C ). values. A follow-up lab value was ordered on average within 443.45 min (7 h 18 min), 643.10 min (10 h 42 min), and 567.13 min (9 h 24 min) for potassium, magnesium, and phosphate, respectively ( Figure 5B ).

The threshold of electrolyte replacement remained steady for 24 hours but varied between 1L->1R and 1L->MR versus 1L->NR ( Figure 5B ). In addition, if multiple electrolyte lab values were requested, a different bimodal distribution of lab orders was seen ( Figure 5C ). Figure 5 . When the electrolyte order was placed, the lab value and replacement seemed to follow hospital routine (A). However, the threshold did not fluctuate considerably over 24 hours, and the most significant difference in the threshold was when multiple lab values preceded the electrolyte replacement order (B). The follow-up lab values were executed in bimodal distribution (C), but the peaks depended on electrolyte type. 

Orders to replace electrolytes resulted in a highly significant difference before and after repletion, but the effect size of repletion varied by electrolyte and, at best, should be judged as moderate as demonstrated by the d-Cohen values. Potassium demonstrated a small effect size of 0.32, while magnesium and phosphate resulted in large effect sizes of 1.08 and 0.87, respectively ( Figure 6A,B) .

Multiple regression analysis determined several factors that correlated with changes in the potassium level ( Table 2 ). The primary predictor of a significant increase in electrolyte levels was pre-repletion levels ( Figure 5C ). However, analysis of several other factors and their contribution had a significantly less impactful effect, as demonstrated by the regression coefficient (Table 2 ). 

Finally, the heterogeneity of the threshold for order replacements was examined in the context of the patients' locations and medications, which had a profound effect on electrolyte levels. The thresholds to replace were identified as being different when the 

Finally, the heterogeneity of the threshold for order replacements was examined in the context of the patients' locations and medications, which had a profound effect on electrolyte levels. The thresholds to replace were identified as being different when the ICU type was considered for potassium (F(4;6691) = 162.056; p < 0.0001) and magnesium (F(4;3472) = 52.41; p < 0.0001) but not phosphate (F(4;282) = 1.96; p = ns) ( Figure 7A ). Being on medications commonly associated with electrolyte abnormalities did not result in a statistically significant effect on the threshold for serum electrolyte replacement ( Figure 7B ). 

This study demonstrated that a significant number of orders measuring the serum level of electrolytes resulted in no subsequent clinical interventions. The practice of electrolyte repletion followed hospital workflow. When healthcare providers placed an order for electrolyte replacement, the pre-existing lab value was, overwhelmingly, within normative values. This behavior is consistent with a prior study where a different data set was used [19] . However, significantly fewer instances where electrolytes were replaced when the pre-existing value was above nominal were identified [19] . This may suggest a 

This study demonstrated that a significant number of orders measuring the serum level of electrolytes resulted in no subsequent clinical interventions. The practice of electrolyte repletion followed hospital workflow. When healthcare providers placed an order for electrolyte replacement, the pre-existing lab value was, overwhelmingly, within normative values. This behavior is consistent with a prior study where a different data set was used [19] . However, significantly fewer instances where electrolytes were replaced when the pre-existing value was above nominal were identified [19] . This may suggest a different safety or practice policy. Irrespective of the reason, this demonstrates a difference in iatrogenic risk between the two institutions. The average serum value-producing repletion remained constant throughout the day, but the average lab value that prompted nonrepletion fluctuated over 24 h. The effects of electrolyte repletion on the post-repletion levels of electrolytes were small. The threshold for electrolyte replacement varied depending on the unit but not the concomitant utilization of medications interfering with potassium homeostasis that also had some diurnal variability. In general, the data showed significant waste due to over-ordering lab values and ineffective repletion patterns.

Investigating electrolyte repletion in the MIMICS ICU database revealed over-repletion patterns by providers from a major hospital system in the Boston area. Similarities in physician behavioral patterns in other institutions support the presence of over-repletion patterns in the ICU. This may allude to a universal issue of over-repletion secondary to the perception that less risk is in inaction vs. initiation of treatment [15, 19, 21] . This bias may be particularly strong in units where electrolyte and cardiac abnormalities are of particular concern (HVICU and CICU), even though evidence has demonstrated the ineffectiveness of pre-emptive repletion of electrolytes in general [1, 2, 4, 12, 18, 29] . However, in this study, cardiac units did not demonstrate a lower threshold for electrolyte replacement. Conversely, the threshold for potassium replacement was the highest, which suggests that a higher goal is set with providers attempting to use even higher levels. Moreover, the diagnoses related to under-repletion were mostly non-cardiac, which is counterintuitive to the standard practice of being more aggressive with potassium replacement in the cardiac population. Furthermore, there was a weak correlation between electrolyte lab values and the propensity of the provider to replace electrolytes. The relation was linear and fuzzy instead of being a discrete value as evidence-based data would suggest. This study demonstrated that healthcare providers decide to replace or not replace electrolytes at a similar threshold, suggesting that other factors play a role. The regression analysis demonstrated that most of the factors listed by the literature as a critical determinant of electrolyte replacement do not contribute significantly to the decision as to whether to replace or not. Consequently, the factors that are likely to drive the electrolyte replacement process is the hospital culture instead of well-established evidence [1, 10, 20, 21, 28, 34] . The fact that shift changes seemed to correlate with repletion is a strong argument for this. Significant common motifs in making these decisions judging from the variables incorporated into the analysis could not be found, including a potassium-wasting or -sparring diuretic. The influence of several other variables was small, even though their potential role was identified in alternative studies. This does not come as a surprise, as the medical decision-making process is often rooted in several variables, and the intricacies are often not accessible to the decision-making individual. Cultural inertia is often linked to the continuation of medical treatment, even if no evidence of support is present [48, 49] .

This study also analyzed the time dynamics of electrolyte replacement and found many electrolyte repletion orders for potassium, phosphate, and magnesium at around 7 a.m. (major peak) and then again at 6 p.m. (minor peak). Interestingly, this "circadian rhythm" pattern was found in a similar study analyzing provider behavior in another hospital system [19] . These major and minor peaks coincide with provider shift changes more than patient factors and may allude to a more informal repletion pattern. While patient outcomes and adverse events were not analyzed in this study, the hypothesis remains that this informal pattern of electrolyte repletion does not decrease the incidence of adverse events. One reason for the low effectiveness of potassium and magnesium replacement may be related to a large gap in care. Replacing potassium takes a significant amount of time. Furthermore, since potassium is an intracellular cation, most potassium will be translocated beyond serum [1, 2] . Consequently, electrolyte replacement should be more frequent and potentially at higher doses to achieve a meaningful change in serum. This would suggest that more aggressive techniques of replacement should be utilized, especially in the case of critically ill patients. One approach to improvement is the development of a protocolized algorithm, but most studies have demonstrated increased compliance and an improved level of serum electrolytes despite a lack of evidence for a change in the clinical outcome [1, [27] [28] [29] . This study demonstrates that electrolyte replacement ineffectiveness results from the providers' behavior and healthcare organization tailored to patients' needs and hospital workflow. That problem was identified before, but only a few studies could quantify the level of routine care [50, 51] . Reduction is achievable using simple education or software solutions, yet the practice persists [52, 53] . A similar problem in other testing modalities pointed toward an overall culture-driven mechanism instead of toward patientcentered care. A reduction in lab tests reduces error, improves quality of life, and accelerates healing [51] .

Several limitations urge caution when analyzing the data. It is difficult to know when the healthcare providers were made aware of the results from the lab draws. Note that the times associated with these results are the time of fluid acquisition, not the time that the values were made available to the clinical staff. This may result in a situation where the lab is delayed beyond 24 h. However, this seems to be unlikely considering the performance of current lab reporting systems. Electrolyte replacement protocols from the hospital providing the data used in this study were unfortunately not available. Reference ranges dictated from the computer system were used as the guide to repletion. Consequently, there was assumed to be a lack of protocol, given that most of the protocolization of healthcare delivery happened in the last ten years. It was also assumed that typical printouts give the result and reference ranges, thus giving an immediate frame for the provider. As a result, it is not possible to assess whether the repletion orders made without specific indication were in line with existing protocols [10, 15, 21, 30, 34] . Deviation from these standard levels may be attributed to the disparity in optimal electrolyte levels across practices. Ascertaining the effect of provider type (physicians, advanced practice providers, etc.) was not possible, which may be a significant factor in electrolyte decision-making processes. Yet, the variability within provider behavior highlights the lack of compliance to a repletion protocol. Similarities to replacement patterns found in another hospital system and from a different time frame raise the question of whether existing protocols are a strong influencing factor in these behaviors. Alternatively, the patterns may result from non-compliance or the lack of standardized guidelines. In magnesium, electrolyte replacement is often driven not by the absolute level but by a clinical indication [2, 13] . This can potentially introduce heterogeneity and is most likely reflected in some incidences of repletion of magnesium when the level was over the normative value.

This study suggests that a more effective process is needed in the field of electrolyte replacement. Currently, there is a significant over-ordering of lab values with little actionable decision. When decision to intervene is made, the decision does not seem to adhere to clinical evidence and is inconsistent. This study most likely confirmed the impression of all practicing physicians in the USA regarding over-ordering and unclear guidance in several "routine" tasks. Implementation of the protocol was suggested to improve the situation, but protocols are rigid, met occasionally with resistance, and cannot account for other variables. More effective changes may be made by focusing on a more targeted system of ordering from the lab, starting with abandoning scheduled lab draws and designing a demand-or clinical question-driven system. Regarding replacement prescription, utilizing mathematical modeling to predict optimal timing and dose is a potential solution.

Analysis of the MIMICS ICU database highlights a pattern of electrolyte repletion not in line with the needs of critical care patients in terms of workflow and effectiveness.

Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/healthcare9101373/s1, Table S1 . List of exclusion ICD-9 diagnoses and their codes. Table S2 . Diagnoses and frequencies of the instances when over-repletion occurred. Table S3 . Diagnoses and frequencies of the instances when under-repletion occurred. Figure S1 . Top ten diagnoses when repletion was ordered when the preceding lab value was above or below the reference for potassium and magnesium. 

Electrolyte disturbances in the intensive care unit

Cations: Potassium, calcium, and magnesium

Atrial fibrillation and changes in serum and urinary electrolyte levels after coronary artery bypass grafting surgery

Changes in serum electrolyte levels and their influence on the incidence of atrial fibrillation after coronary artery bypass grafting surgery

Changes in serum electrolyte levels during marathon running

Hypomagnesemia associated with other electrolyte imbalances

Is routine electrolyte testing necessary for diabetic patients who present to the emergency department with moderate hyperglycemia?

Association between day and time of admission to critical care and acute hospital outcome for unplanned admissions to adult general critical care units: Cohort study exploring the 'weekend effect'

Linking evidence-based medicine therapeutic summary measures to clinical decision analysis

When is rational to order a diagnostic test, or prescribe treatment: The threshold model as an explanation of practice variation

Severe electrolyte disorders following cardiac surgery: A prospective controlled observational study

Potassium and Magnesium Supplementation Do Not Protect Against Atrial Fibrillation After Cardiac Operation: A Time-Matched Analysis

Protective role of magnesium in cardiovascular diseases: A review

Evidence-based practice is not synonymous with delivery of uniform health care

Van den Ende, J. Rationality, practice variation and person-centred health policy: A threshold hypothesis

Healthcare 2021

American Heart Association's Get with The Guidelines-Resuscitation I: The administration of dextrose during in-hospital cardiac arrest is associated with increased mortality and neurologic morbidity

Does the prophylactic administration of magnesium sulphate to patients undergoing thoracotomy prevent postoperative supraventricular arrhythmias? A randomized controlled trial

Pharmacokinetics of Magnesium Bolus Therapy in Cardiothoracic Surgery

Behavioural patterns of electrolyte repletion in intensive care units: Lessons from a large electronic dataset

Decision-making when data and inferences are not conclusive: Risk-benefit and acceptable regret approach

Thinking Styles and Regret in Physicians

Does admission time matter in a paediatric intensive care unit? A prospective cohort study

Hospital mortality is associated with ICU admission time

An examination of cardiovascular intensive care unit mortality based on admission day and time

Protocol-driven vs. physician-driven electrolyte replacement in adult critically ill patients

Development of a massive transfusion protocol

Effects of Electrolyte Replacement Protocol Implementation in a Medical Intensive Care Unit

Evaluation of an electrolyte replacement protocol in an adult intensive care unit: A retrospective before and after analysis

Evaluation of an electrolyte repletion protocol for cardiac surgery intensive care patients

Narrative Review of Decision-Making Processes in Critical Care

Complexity perspectives on clinical decision making in an intensive care unit

Cognitive complexity of the medical record is a risk factor for major adverse events

The Characterization of the Toll of Caring for Coronavirus Disease 2019 on ICU Nursing Staff

Cognitive elements in clinical decisionmaking: Toward a cognitive model for medical education and understanding clinical reasoning

Health professional decision-making in the ICU: A review of the evidence

Implementing and Enhancing Multifaceted Systems Interventions to Improve Goals-of-Care Discussions Among Patients with Serious Illness-An Opportunity for Human-Centered Design

MIMIC-III, a freely accessible critical care database

Pipe-Friendly Framework for Basic Statistical Tests

Elegant Graphics for Data Analysis

Effects of time and day of admission on the outcome of critically ill patients admitted to ICU

Assessing the association between occupancy and outcome in critically Ill hospitalized patients with sepsis

Ggpubr: 'ggplot2' Based Publication Ready Plots

Dbplyr: A 'dplyr' Back End for Databases

Rcpp' Interface to 'PostgreSQL

Knitr: A General-Purpose Package for Dynamic Report Generation in R, version 1.36; CRANE

Data Table: Extension of 'Data Frame

Dates and Times Made Easy with lubridate

Clinical Inertia in a Randomized Trial of Telemedicine-Based Chronic Disease Management: Lessons Learned

Is therapeutic inertia present in hyperglycaemia, hypertension and hypercholesterolaemia management among adults with type 2 diabetes in three health clinics in Malaysia? a retrospective cohort study. BMC Fam

Implementation of Later Morning Specimen Draws to Improve Patient Health and Satisfaction

Evaluation of the Order SMARTT: An Initiative to Reduce Phlebotomy and Improve Sleep-Friendly Labs on

Reducing the number of unnecessary routine laboratory tests through education of internal medicine residents

Reduction in Unnecessary Clinical Laboratory Testing Through Utilization Management at a US Government Veterans Affairs Hospital

We would like to acknowledge Joanne Levy, MBA, from the Leonard Davis Institute for HealthCare economic for support during this project.

The authors declare no conflict of interest.