IT'S COMMON for nurses to struggle to accurately assess the acuity of a patient's clinical status. While some simple-to-use patient acuity assessment tools are available to healthcare providers, powerful new predictive approaches such as the Military Medical Acuity Model (MAM) are always welcome. This article describes MAM, its parameters, and its evolution toward clinical use.
Background and applications
MAM is a patient-centered acuity system based on objective physiologic measures mined from the electronic health record (EHR). MAM translates existing data into recognizable patterns that can be displayed to the clinician. The MAM is designed for use throughout a facility, rather than on a specific unit or specialty area, and allows comparisons among units and facilities regardless of location. It's intended to provide a common lens through which clinicians can view individual patients across multiple specialties. It can also be applied to specific requirements for resource planning and utilization within existing workflow patterns.
MAM is an assessment tool developed by the author from existing literature. MAM mines existing EHR reporting tools and is easily incorporated into daily workflow. The model can also enhance interprofessional communication by generating a common operating picture of patient acuity.
MAM framework
The framework for the MAM includes four concepts from the American Association of Critical Care Nurses (AACN) Synergy Model for Patient Care:1
* complexity (the entanglement of two or more systems)
* stability (the ability to maintain equilibrium)
* vulnerability (susceptibility to stressors)
* resiliency (the capacity to return to higher function after insult).
The AACN Synergy Model recognizes that all patients have similar needs across the continuum from health to illness. In this model, complexity and stability identify the degree of physiologic dysfunction. Vulnerability and resiliency measure the patient's ability to adapt to physiologic dysfunction and identifies the risk of future decline.
In the MAM tool, these four concepts are measured by 17 individual objective parameters. The MAM values are based on a simple rubric: A normal score is zero, with any deviation from 1 to 3 for each measure depending on the variable and degree of dysfunction. All scores are positive, meaning that a higher score equals a higher acuity. In the MAM scoring methodology, the absence of a value results in a score of zero. The intent is to treat missing data the same whether the omission is a conscious decision or an error (such as a data query error). Multiple data sampling points over time help normalize the patient's MAM score.
The MAM is a multidisciplinary tool derived from evidence-based studies and various objective measures. The objective measures were selected because of their presence in more than one tool, or more than one evidence-based study in several different disciplines. Collection of these objective measures and physiologic parameters is nearly universal in inpatient facilities.
Complexity parameters
Comorbidity as a measure of complexity is found in multiple scales and impacts readmissions as well as mortality. In a prospective study in a large outpatient practice setting, the most effective measure for predicting mortality and healthcare cost was a simple count of the number of medications prescribed for the patient.2 Medication count is also used in the Groningen Frailty Indicator; more than three medications is one predictor of frailty.3 In a retrospective cohort study involving two large hospitals and over 5,000 patients, the number of medications at discharge was a strong predictor of readmission in medical patients, as were low hemoglobin and low serum sodium levels.4
Vulnerability parameters
In identifying indicators for vulnerability, readmission rates surfaced in multiple studies. In a prospective cohort study of 142 patients over a 6-month period, three factors were found to be of significance for readmission: chronic disease, depressive symptoms, and underweight. Most (72%) underweight patients and 50% of obese patients were readmitted, compared with 27% of the normal weight patients and 37% of the overweight patients.5 In a separate retrospective study of 10,359 admissions, the readmission rate was 17% and weight loss was one of nine significant factors.6
Other vulnerability factors include age, which is associated with increased risk for morbidity and mortality, and falls, which are the most common cause of nonfatal injuries for people over 65.7 Fall-related injuries are also the most common cause of accidental death in those over 65.8
Stability parameters
Critical illness is preceded by well-defined physiologic signs. The most sensitive sign is respiratory rate.6 Unfortunately, very few reliable methods have been developed for trending these physiologic signs. The Modified Early Warning Score (MEWS), used in the United Kingdom since the early 2000s and extensively studied, applies this concept of trending physiologic changes.9-11 The MEWS consists of six readily available measures. A score of 5 or more on the MEWS was associated with an increased risk of death and ICU admission, and patients over age 70 were at higher risk than younger patients.9 In a retrospective study involving 2,974 patients over a 3-year period, researchers found a strong relationship between the probability of death and the MEWS score.12
Intervention lead time and time to recognition and action are also important to consider with early warning scores. In a prospective study of 551 patients admitted to the ICU, patients from general units had a higher mortality (47.6%) than those admitted either from the ED (31.5%) or OR (19.3%).13 Patients from general units also had more serious antecedents before ICU admission, such as hypotension, tachycardia, tachypnea, and sudden change in the level of consciousness.
A similar study showed that among patients in general units, nearly 80% had physiologic parameters outside normal ranges in the 24 hours preceding ICU admission, and 75% had a life-threatening antecedent factor within 8 hours of ICU admission.14 Respiratory rates of >27 breaths/minute occurred at least once during the 72 hours before cardiac arrest. Typically, Spo2 is used as a surrogate for respiratory function. However, in this same study, the increase in respiratory rate preceded a decrease in Spo2 by between 5 and 126 minutes.14
Similar vital sign parameters were mentioned in an obstetric (OB) study.15 Although OB patients have different risk factors, this study illustrates the commonalities in the significance of these objective measures across multiple patient types.
Resiliency parameters
Identifying patients who have the capacity to recover from psychological and physiologic stressors is important. In a study of psychological resiliency, Bartone found an apparent biophysical connection between hardiness and resiliency to stress and high-density lipoproteins (HDL): Hardy or resilient subjects had higher levels of HDL.16
In a retrospective study of 439 patients, Fraser et al. found that the lowest HDL levels in both males and females were associated with the highest cortisol excretion rates and body mass index. The researchers believed that cortisol affected HDL levels, and that the long-term effect of excess cortisol could explain the increase in cardiovascular risk associated with low HDL.17
In other studies, low HDL was associated with higher mortality and decreased ability to respond to high BP and cardiovascular disease.18,19
Testing the new model
The MAM was tested in a retrospective study at a military hospital. The population for the study was a convenience sample of all hospital inpatients from March 1 to September 30, 2012. Because the MAM was developed for use on adults, all patients under age 18 were eliminated from study, as were all OB and inpatient psychiatric patients. The sample population was 2,461 patients. All patients were eligible for care in the Military Health System under one of several benefit programs.
Deidentified retrospective raw data for each parameter were pulled from an EHR used by the Department of Defense into a spreadsheet to protect patient privacy. These raw data were then put in the MAM tool. Total scores were tallied for each patient every 12 hours. Some scores were measured only once in the patient's stay (age, number of diagnoses on admission, and serum HDL level). Because some data points are rarely collected more than once every 24 hours (lab values, for example), at least one value for the other data points needed to be present in the record in the previous 24-hour collection period to achieve a score greater than zero. The total MAM score was then analyzed as described in the next section.
Results
Because this is a newly developed tool, an average MAM score was calculated using the sample population of 2,461 as a starting point for discovering variation. The average MAM score for this population was 11.41, with a standard deviation of 4.27. Using 11.41 as the mean and adding one standard deviation above the mean (4.27) as a break point, a score of 15 or greater was determined to be high acuity. This was chosen as it represented a score higher than scores for approximately 85% of the population (assuming a normal population distribution curve).
Data were analyzed to see if length of stay (LOS) was predicted by changes in MAM scores for several diagnostic-related groups (DRGs) such as major chest procedures, pneumonia, and limb reattachment. Mean and standard deviations for each DRG were calculated and scored. The total DRG subsample size was 209.
The individual DRG results displayed high degree of variance between standard deviations (range of 1.5 to 5.16 days LOS). The individual DRG results were combined to determine sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for the sample. The result was an overall sensitivity of 58.2%, specificity of 80.3%, PPV of 74.7%, and NPV of 85.9%.
In the sample population of 2,461 patients, HDL levels were available for 923 patients within the hospital stay. The average LOS was 3.6 days for the sample population. High HDL was associated with shorter LOS independent of diagnosis, age, or gender. Patients with an HDL of >45 (n =311) had an average LOS of 2.95, and patients with an HDL of <35 (n = 168) had an average LOS of 7.48 days. The relationship between HDL and LOS was linear with the LOS increasing as the HDL decreased. Although HDL doesn't change quickly, even with intervention, this could support HDL as a reasonable marker for resiliency based on chronic stress load.
From one month's data, eight patients were identified as having MAM score changes significant enough to cause an ICU transfer. Of the eight patients identified, four were actually transferred as predicted by the MAM score, resulting in a 50% accuracy rate in this limited sample. In these four ICU transfer cases, the increase in MAM score preceded the transfer by 11.2 hours on average.
Study limitations and MAM improvements
The data were pulled retrospectively from the EHR database. Data were retrieved twice per day, 12 hours apart for each patient. This collection interval caused limitations because patient data can change rapidly, and it reduced the number of available data points for analysis. One LOS might have only 8 measures (4 days 2 data pulls/day), which makes identifying patterns in these patients difficult.
Due to the sometimes unpredictable nature of the data in the EHR, all data points display missing data for the sample. The grouping of these data is largely due to the program that pulled the data from the source and doesn't reflect any increase or decrease in importance of the data. Cases were identified as missing data if one point was missing from the parameter in the entire stay. Only one parameter (age) had 100% of the data present for the entire LOS.
As a result of this study, six additional data points were added to the MAM version 2.0: diastolic BP (to calculate the mean arterial pressure for the vulnerability score), serum albumin level (to add to the resiliency score), oxygen saturation (Spo2) (for the stability score), serum glucose level, immature white blood cells (bands), and platelets (for the complexity score). All of these elements have been noted in the literature as measures indicating risk for negative outcomes. These lab studies also fit the data profile of using simple data sources such as complete blood cell count and basic metabolic panels.
More specialized labs, such as serum lactate and C-reactive protein, could also be added, but these occur too rarely in the EHR to be of practical use at this time. The alert/verbal/painful/unresponsive scale was removed from the measure as this is rarely documented in the EHR and didn't add any significant refinement to the results.
An easy path to individualized care
In an increasingly busy healthcare system, using existing data and tools that can be built into current workflow is critical for accuracy and efficiency. Using standard lab and clinical data, the MAM can be used to determine an individual patient's acuity, rather than an aggregate acuity for a group of similar patients. The tool is designed to assist the clinician in monitoring changes in patient status, but relies on the clinician to make the appropriate decisions on how to manage the changes. In its current form, the MAM tool is in a spreadsheet, but it will work with nearly any data organization format because it doesn't depend on a specific EHR or system.
Application of knowledge management techniques and manipulating big data is an important part of healthcare today. The ability to extract knowledge from existing systems and present it in an explicit way to others is a hallmark of the MAM. It can provide a common operating picture of the patient to all members of the healthcare team and supports refined application of scarce resources.
REFERENCES