Evaluating the Association and Predictability of Complex Medication Regimen Scores with Clinical Outcomes Among the Critically Ill

Introduction Medication Regimen Complexity (MRC) refers to the combination of medication classes, dosages, and frequencies. The relationship between MRC and clinical outcomes in the intensive care unit (ICU) has not been examined. The objective of this study is to examine the association between MRC scoring tools and their utility in predicting clinical outcomes. Methods We conducted a retrospective cohort study that includes 322 adult patients admitted and stayed (>24 hours) to the ICU between February 1, 2020, and August 30, 2020 in a community-based, teaching hospital in Rhode Island .Medication complexity was assessed using two MRC scoring tools: MRC Index (MRCI) and the MRC in ICU (MRCICU). We used a multivariable logistic regression model to identify the association between MRC scores and clinical outcomes and to predict the clinical outcomes. RESULTS Among the 317 patients included in the study (55.2% men with a median age of 62 [IQR: 51-75] years). Higher MRC scores (i.e., > 63 MRCI or > 6 MRCICU) were associated with increased mortality (14% and 15%), longer ICU length of stay (LOS )(30% and 34%), and need for MV (24% and 28%). MRCICU scores at 24 hours were found to be a significant risk factor in all clinical outcomes (ICU mortality, LOS, and MV) with Odds Ratio (ORs) of 1.12 (95% CI: 1.06-1.19), 1.17 (1.1- 1.24), and 1.21 (1.14- 1.29), respectively. In the prediction models, both MRCI and MRCICU models performed similarly (AUC: 0.88 [0.75-0.97] and 0.88 [0.76-0.97] in predicting mortality. The Medication Model included 15 medication classes outperformed others (AUC: 0.82 [0.71-0.93] in predicting ICU LOS and the MRCICU model outperformed others (AUC: 0.87 [0.77-0.96]) in predicting the need for MV. CONCLUSION MRC scores are associated with poorer clinical outcomes and improves the prediction of poorer clinical outcomes which will support clinicians to prescribe safer therapies.


63
Medication regimen complexity (MRC) refers to multiple features of a patient's medication drug 64 regimen rather than absolute number of medications consumed per day [1]. MRC incorporates features 65 such as the number of agents, dosages, administration time intervals, and additional instructions (i.e., 66 take on an empty stomach) [2][3][4]. An increase in MRC burden has been associated with poorer 67 medication noncompliance and caregiver quality of life measures, as well as an increase in healthcare 68 resource utilization [5,6]. Critically ill patients are at significant risk of higher MRC due to the severity of 69 illness, management of multiple chronic conditions, and the complex pharmacotherapies prescribed. It 70 has been estimated that the average critically ill adult may receive up to 13 medications per day and the 71 incidence of experiencing an adverse drug event has been estimated to greater than 25% lending to 72 substantial morbidity and mortality [7][8][9][10][11]. Therefore, examining only the quantity of medications 73 administered may not accurately describe the complex and intricate nature of the critical care 74 medication regimens.

75
Numerous methods have been used to quantify the complexity of medication regimens. Yet, the most 76 commonly utilized and validated objective scoring tool is the 65-item, weighted MRC index (MRCI) which 77 has been developed for outpatient use [12][13][14][15][16]. MRCI has been used to evaluate conditions such as 78 neurological impairment in children, hypertension, diabetes, and chronic kidney disease in adults [17-79 21]. In the intensive care unit (ICU) population, the MRC in ICU (MRCICU) scoring tool was developed 80 and revised (modified MRCICU) in 2019 [22,23]. The modified MRCICU is the first validated quantitative 81 weighted scoring tool intended to predict clinical outcomes (i.e., ICU mortality, length of stay (LOS), and 82 need for mechanical ventilation (MV)) [24,25]. In a previous study, we have developed a novel proof-of-83 concept method demonstrating improvement in the prediction of patient outcomes by incorporating 84 the MRCICU score into the previously established Acute Physiology and Chronic Health Evaluation 85 (APACHE II) scoring tool [26]. To date, no study has assessed the association between these two MRC 86 scoring tools and clinical outcomes within the critical care setting.

87
In this study, we developed two custom MRC scoring algorithms and several statistical and prediction 88 models using the MRCI and MRCICU tools to gain insight into how MRC impacts clinical outcomes (i.e., 89 ICU mortality, LOS, and need for MV). We aimed to (1)  logistic regression models were utilized to examine the association between risk factors and clinical . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) prediction models were constructed without any variable selection. A 'no imputation' approach was 152 used when preparing the data for the prediction model. We assessed correlated variables using the 153 Pearson correlation coefficient. The SAPS II severity score was used in the prediction models due to a 154 high correlation with the APACHE II classification system (S1 Fig). The prediction accuracy was evaluated 155 using the area under the receiver operating characteristic (AUC) curve. An AUC of at least 0.7 was 156 regarded as acceptable. We applied a 'leave-one-out' cross-validation method with 10,000 repetitions 157 and AUC was selected as an overall performance measure. Additionally, sensitivity and specificity 158 analyses were included for the three outcomes. Furthermore, all prediction models recorded variable 159 importance rankings for each outcome. characteristics between survivors and non-survivors . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 4, 2022.  Table). The use of vasopressors was found to be a significant risk factor for all 210 outcomes in the models. When evaluating morality, the use of paralytic agents was significant with an 211 (OR: 3.38 (1.09-11.11). The use of anti-infectives, anticoagulants, and cardiovascular agents were 212 significantly associated with prolonged LOS. Lastly, the use of analgesics, sedatives, psychiatric, 213 cardiovascular, and pulmonary agents were significant risk factors for the need of MV.
214 215 216 is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 4, 2022.  (Fig 2). In the MRCI & SAPS II Model, MRCI scores at 24 and 48 231 hours were identified as the top variables of importance when predicting mortality (Fig 3) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 4, 2022. ; . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 4, 2022. interdisciplinary care team members to review and modify the medication regimen to ensure safer, . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

280
Secondly, our results clearly demonstrate that MRC scores at 24 hours were associated with mortality 281 suggesting they should be incorporated into current practice ( MRC has been shown to be a better predictor of mortality compared to polypharmacy alone [32]. MRC 284 scores at 24 hours were significantly associated with ICU LOS and the incorporation of therapeutic drug 285 classes improved the prediction for LOS and need for MV (Table 3). These findings can be confounded 286 however by a) exclusion of the weights for important medications (i.e., vasopressors and anti-infectives) 287 in MRC calculations, b) exclusion of patient's severity, and c) exclusion of influential combinations of 288 medications regimens.

289
Thirdly, vasopressors were a significant predictor in all clinical outcome models (Fig 3) and were found 290 to be statistically significant in all stepwise models ( Table 2). In practice, use of vasopressors is indicated 291 in patients with poorer health conditions, such as decompensated heart failure and shock [45,46]. The 292 frequent diagnosis of ICU-related delirium has been a known contributing factor to ICU LOS among other 293 undesirable outcomes [47][48][49]. Historically, numerous medications have been used to minimize the 294 duration of delirium, yet studies are lacking to identify a safe and effective agent. Our findings of current 295 psychiatric medications potentially contributing to an increase in TOMV suggests the continued need to 296 identify an agent to minimize the incidence and duration of ICU delirium leading to extended TOMV and 297 LOS.

298
Lastly, the association of pulmonary and paralytic agent use with mortality and the need for MV was 299 anticipated in our findings as these therapeutic classes are commonly associated with high acuity 300 diseases such as acute respiratory distress syndrome (ARDS) and acute brain injury [50,51]. Therefore, 301 our study supports the inclusion of medication class usage to predict critically ill outcomes. Most 302 notably, the lack of medication use into existing severity of illness scoring tools ( We propose adopting MRC scores, preexisting comorbidities, and severity of illness into future modeling 316 to improve the accuracy of prediction. 317 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted February 4, 2022. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.03.22270376 doi: medRxiv preprint