The association between the use of intensivists and the efficiency of care provided by the hospital’s ICU

Introduction

The value of health care services can be conceptualized as the health outcomes achieved per dollar spent on those services (1,2). Many have suggested that the value of healthcare services in the United States is lacking. The concern regarding the value received from healthcare services stems from both sides of the value equation: the rising cost of care (3-5) and the corresponding lackluster (or even harmful) clinical results received from that care (6,7). In other words, Americans are paying more for substandard clinical results. For example, Papanicolas and colleagues found that while the United States spent nearly twice as much as other high-income countries on medical care, it ranked at or near the bottom for health measures such as obesity, life expectancy and infant mortality (4). Because of these deficiencies, improving the value of health care services in the United States is a high priority for both policy makers and providers (8-10).

In response to these concerns, policymakers, payers and providers are all under extreme pressure to develop and implement strategies that improve the quality of the care provided and reduce the cost of providing that care. Recent policy, payer and provider initiatives such as the implementation of electronic medical records (11-14), pay-for-performance (15), bundled payments (16,17) and provisions of the 2010 Affordable Care Act (ACA) (18-20) are designed to either improve quality or reduce the cost of care.

One strategy that has been purported to improve quality while simultaneously increasing efficiency and reducing cost in the intensive care unit (ICU) is the use of intensivists. Intensivists are often employed or contracted by the hospital and are “physicians who specialize in critical care with the experience and skill required to detect and address changes in patients’ clinical conditions, often before complications occur. Intensivists can be surgeons, anesthesiologists, internists, or pediatricians with additional training and board certification in critical care” (21). Intensivists are often the ICU director and are charged with managing and coordinating the costly care provided in the ICU (21-24).

The ICU is a significant component of hospital operations. The ICU can represent as much as 30% of a hospital’s budget (25) and at a national level, the care provided in these units accounts for more than 20% of the total acute care hospital costs (26), while only 15% of hospital beds in the United States are allocated to critical care (27). ICUs care for some of the sickest patients in the health care system (28) and these patients utilize more resources and incur more cost than other patients (29,30). The number of patients being cared for by hospital ICUs has increased, particularly during the coronavirus disease 2019 (COVID-19) pandemic (31-33), and based on demographic trends, the number of patients requiring care in the ICU will continue to increase for the foreseeable future (26). Given these factors, it is easy to see the potential that improvements made in the operations of the ICU by staffing with intensivists could have a considerable impact on the quality, efficiency and costs of the care provided within the hospital.

In this study, we focus on the potential effect of intensivists on hospitals’ efficiency and costs as proxies for value. Several studies have examined the association between the use of intensivists and hospital efficiency and costs. For example, Wortel and colleagues (34) found an association between the ratio of intensivists per ICU bed and ICU efficiency and Dimick and colleagues (35) found that for patients who underwent an esophageal resection, daily rounds with an ICU physician were associated with shorter lengths of stay and reduced hospital expenses (35). Similarly, Breslow and colleagues (36) found that length of stay (LOS) was reduced from 4.35 to 3.63 days after the implementation of an e-ICU program where intensivists provided services remotely using electronic monitoring and communication technologies and Parikh and colleagues (37) found that the average ICU LOS decreased from 3.5 to 2.7 days when staffed with intensivists. Finally, in a structured literature review, Pronovost and colleagues (38) found that in 14 of the 18 studies included in their analysis, LOS was reduced with the staffing of ICUs with intensivists.

While these studies certainly lend credence to the positive effects on cost and efficiency associated with the use of intensivists, it is important to point out that the results have sometimes been mixed, that most of these studies included a very limited number of ICUs, often focused on a single ICU or extremely small number related ICUs, and/or the results related to a single type of patient, diagnosis or procedure treated. This paper seeks to build on the existing literature by expanding the focus to a much larger number of facilities with ICUs across multiple entities and health systems and firmly grounding this study on a conceptual framework based on agency theory. In addition, this study seeks to provide policymakers and administrators a better understanding of the association between the use of intensivists and efficiency and costs within the hospital. For this study, efficiency and costs are operationalized as average LOS and average cost per patient day. This more generalizable understanding of the potential benefits of utilizing intensivists will help leaders in healthcare make decisions that generate the greatest value to patients, hospitals and the overall healthcare system.

Conceptual framework

Agency theory guided this study. In agency theory, there are two key roles within an organization: principals, who own or control the entity, and agents, who are contracted by the principals to perform necessary tasks within the entity on their behalf. Agency theory explains this relationship, conceptualized as a contract, where decision-making authority is delegated from the principal to the agent (39,40). Because the perspectives and interests of the principal and the agent will inherently differ, these contracts are not perfectly efficient. Inefficiencies in contracts are due to agency problems such as differing goals (40), asymmetric information (41,42), and differences in risk preferences (41). Agency theory focuses on determining the most efficient contract between the principal and the agent (43).

To address agency problems, principals must align their goals with the goals of the agent. According to Eisenhardt (43), “organizations are viewed as collectives of self-interested people with partially conflicting goals”, which are “resolved through alignment of goals through the use of incentives”. To align the goals of the principal and the agent, the principal usually relies on various compensation and incentive programs for the agent (44). The incentives can be designed to reward the agent’s actions (behavior-based) or the outcomes of the agent’s actions (outcome-based). The principal must decide whether to base incentives on behaviors or outcomes (45).

A hospital’s goal is to provide high-quality patient care in an efficient, cost-effective manner through various structures and processes. Through a host of buildings, equipment, technology, departments, professionals and processes, hospitals strive to deliver care that results in better clinical outcomes for their patients while utilizing the fewest resources. To achieve this goal, a hospital must align its goals (as the principal) with the goals of each of the staff members and providers (as the agents). Many of these agents, such as nurses, lab techs, radiologists, etc., are employed by the hospital and therefore the hospital can rely on traditional employment rewards and incentives—behavior- or outcomes-based—to help align its goals with the agents’ goals. In addition, the hospital can use information systems and other control processes to reduce information asymmetries that exist between the two parties. However, these mechanisms are less effective with independent referring physicians, who are not directly employed by the hospital.

Traditionally, hospitals have relied on independent physicians to admit patients and manage their care, including in ICUs. These referring physicians are largely autonomous from the hospital and have little incentive to align their personal goals with the goals of the hospital. This traditional model has been labeled as the “physician’s workshop”, where the independent physician has significant control over the resources of the hospital, even though the physician has no direct financial connection with or obligation to the hospital (46). This model provides the hospital (principal) little control over the activities of the referring physician (agent) within the hospital, and little opportunity to align its goals of providing high-quality, efficient, cost-effective care with the goals of the physician. This is especially true regarding the efficiency achieved and the cost incurred by hospitals since physicians can control up to eighty percent of the cost in the hospital (41,47).

This lack of control of referring physicians is especially troubling in the ICU, given the complexity and cost of care. ICUs have the highest case-mix indexes of all hospital inpatient services and include many expensive and unexpected costs driven by the decisions made by referring physicians (46,48,49). Esposto (46) states, “the greater the hospital’s exposure to the risk of physician opportunistic behavior, the greater the probability that the hospital will seek alternative institutional arrangements to reduce the risk.” One of the mechanisms hospitals can use to reduce risk, increase leverage, and more effectively align its goals with those of the physician is through the use of intensivists.

Intensivists are often employed by the hospital, while some are under contract. Under this staffing model, coordination and control of care in the ICU is transferred from the referring physicians to the employed or contracted intensivists. This model allows hospitals to align the physicians’ goals with those of the hospital through traditional incentives, rewards, and information systems. Kohli and Kettinger (41) state, “professional agents are more likely to be committed to the control of management when they are highly dependent on them for career advancement and when management has the legitimacy to distribute rewards.” As such, by employing or contracting intensivists, hospitals can mitigate agency problems and more effectively align the physicians’ goals with those of the hospital.

Methods

As stated above, the goals of the hospital include providing high-quality care in an efficient, cost-effective manner. Efficiency relates to the ability to achieve desired results with the minimum amount of resources. One measure of efficiency within a hospital is the number of patients that are cared for with a given set of resources over a given period of time, which can be operationalized as the average LOS (26). Average LOS depicts the average number of days each patient remains in the hospital per admission. Greater efficiency reduces average LOS and increases throughput for the fixed number of beds within the facility. Given that the intensivist staffing model has the potential to increase the efficiency of the contract between the principal (the hospital) and the agent (the physician) and improve the alignment of the hospital’s goals with those of the physicians serving in the ICU, the following hypothesis is proposed:

• H₁—an increase in the use of intensivists is associated with lower LOS among ICU patients.

Another measure of efficiency for a hospital is the average cost of care. As efficiency improves, the average cost per patient day should decline. Prior research suggests that the use of intensivists is associated with decreased ancillary costs (26,50). This study postulates that the use of intensivists allows a hospital to better align its goals with the employed or contracted intensivists. This alignment increases intensivists’ incentives to coordinate care across providers and adhere to various hospital protocols. Such coordination and adherence are expected to result in more efficient and more cost-effective care. For this study, the average cost per patient day by principal diagnosis was utilized to measure the cost of care provided to ICU patients. The following hypothesis is proposed in regard to the cost of care provided to ICU patients:

• H₂—an increase in the use of intensivists is associated with lower costs among ICU patients.

Data

The unit of analysis for this study was the hospital. Longitudinal hospital-level data from the American Hospital Association’s (AHA’s) Annual Survey for the years 2007–2010 were merged with data from the Area Health Resources Files (AHRF) and summarized patient-level data from the Healthcare Cost and Utilization Project’s (HCUP) State Inpatient Databases (SID). The AHA Annual Survey provides important data points for over 6,200 hospitals across the country. Data points incorporated in the survey include environmental and organizational details, physician and staffing metrics, service offerings, utilization statistics and other applicable details about each hospital. The AHRF provides Medicare Advantage (MA) Health Maintenance Organization (HMO) penetration data. The SID is a patient-level data source that is part of the HCUP catalog of databases that were created through a Federal-State-Industry partnership that is sponsored and coordinated by the Agency for Healthcare Research and Quality (AHRQ). The SID includes inpatient discharge records for all patients, regardless of payer, and contains a robust offering of clinical and non-clinical data for each patient. The SID includes data regarding diagnoses, procedures performed, admissions and discharge statuses, patient demographics, payment sources, total charges, LOS and other pertinent patient-level information.

Sample

The hospitals included in the analysis were acute, short-term, general hospitals in New York and Washington State. While there are SIDs available for approximately 27 states for the years included in this study, not all states nor all years for each state included all variables needed for the analyses performed for this study. The availability of these required variables was the principal driver of the selection of states for this study. By analyzing the variables offered for each available state for the years to be included in the study, it was determined that the SIDs for New York and Washington State contained the required data elements and provided the desired heterogeneity of the geographic, demographic, environmental, ethnic and cultural attributes of the included hospitals and the patients served by those hospitals.

Furthermore, the sample was limited to hospitals that were categorized by the American Hospital Association as non-federal governmental, nongovernment not-for-profit, or investor-owned for-profit facilities. Because of their differing nature, the analysis excluded specialty hospitals and federal governmental hospitals such as facilities operated by Veteran’s Affairs and the armed services. In addition, only facilities that reported ICU beds to the AHA were included in the analysis.

The analytic sample consisted of between 169 and 174 hospitals per year from New York and Washington, resulting in between 614 and 625 hospital-year observations, depending on the principal diagnoses being analyzed, over the 4-year period.

Variables

Table 1 lists the variables utilized in this study and notes the definition and source of each. The dependent variables in this study included: (I) the average total LOS for ICU patients by hospital; and (II) the average total cost per patient day for ICU patients by hospital. A separate measure was created for both of these dependent variables (average total LOS and average total cost per patient day) for each of the four principal diagnoses of interest, resulting in a total of eight hospital-level dependent variables. The four principal diagnoses of interest included acute myocardial infarction (AMI), congestive heart failure (CHF), stroke and pneumonia. Because the average LOS and the average cost per patient can differ significantly by diagnosis, this study chose to segment patients by these four specific principal diagnoses and analyze the results for each of these patient segments independently. These four diagnoses were selected due to the availability of data and relatively high number of patients that had one of these diagnoses and utilized the ICU during their stay.

In order to create each hospital-level dependent variable, patients were first segmented by whether they utilized the ICU during their hospital stay. This was accomplished using the ICU utilization flag included in the SID. HCUP includes up to thirty utilization flags in each SID that indicate whether a patient utilized various services during their visit. These utilization flags were developed by HCUP using International Classification of Diseases, Ninth Revision (ICD-9) procedure codes and Uniform Billing (UB-92) revenue codes. This variable was coded as 1 if ICU services were utilized or 0 if not. This study only included patient stays where the ICU utilization flag was coded to 1, indicating that the patient utilized ICU services during their visit.

Once patients who utilized the ICU were identified, those patients were categorized by principal diagnosis using ICD-9 codes. Table 2 lists each ICD-9 code included for each of the four principal diagnoses included in this study. The list of ICD-9 codes for each principal diagnosis used for this study was sourced from the Centers for Medicare and Medicaid Services’ (CMS) measure methodology reports (51,52) and was validated by reviewing several previous studies that also focused on these principal diagnoses (53-57).

Finally, to convert the patient-level data contained in the SIDs to hospital-level measures for each hospital included in the analysis, the patient-level data were aggregated into hospital-level measures. The average total cost per patient day by each principal diagnosis for all patients who utilized the ICU was calculated for each hospital. The standard SID tables only include the total charges for each patient, but a cost-to-charge conversion file is provided by HCUP that was used to convert total charges to total costs. In addition to the average cost per patient day, the average total LOS for all patients who utilized the ICU was calculated for each principal diagnosis for each hospital.

The independent variable in the study was a dichotomous variable indicating either the use or non-use of intensivists at the hospital. This binary variable was coded to 1 if the facility reported intensivist full-time equivalents (FTEs) or 0 if not. For a post hoc analysis, another set of independent variables was created based on the number of reported intensivist FTEs per patient day for all ICU patients. This set of independent variables included five dummy variables that were created based on the intensity of intensivist staffing. For these dummy variables, hospitals were placed into one of four quartiles based on the number of intensivist FTEs reported per the total number of patient days for all patients that utilized the ICU during their stay. The fifth category was for hospitals that reported no FTEs. This fifth category served as the reference category.

In addition to the dependent and independent variables, several control variables were included in the analysis. Control variables for environmental factors and hospital-level characteristics were included. To control for environmental factors, the level of competition, operationalized as the Hirschman-Herfindahl Index (HHI), was included. The HHI has a range of 0 to 1, with higher values signifying lower competition (higher market concentration). The HHI was calculated based on hospital-reported adjusted patient days using Health Services Areas (HSA) to define the geographic market to which a hospital belonged. HSAs have been used to define geographic markets in numerous hospital studies (58-60).

To control for hospital factors associated with hospital operational performance, such as efficiency and effectiveness, several hospital-level control variables were included in the analysis. Nurse staffing has been found to affect hospital performance (61-63) and was therefore included as a control variable for this study. Occupancy rate and payer mix variables have been included in other studies exploring hospital performance (64,65), therefore hospital occupancy rate, the percentage of patients that were Medicare and the percentage of patients that were Medicaid were included as control variables for this study. Hospital size was controlled for with two variables: total number of beds set up and staffed at the hospital and the total number of medical/surgical ICU beds reported by the hospital. Finally, a dichotomous variable was also included, indicating whether or not the hospital operated a cardiac intensive care unit alongside a medical/surgical ICU unit.

In order to control for average patient acuity at the hospital level, average patient demographics and comorbidities were included. Demographic variables were calculated separately for each principal diagnosis and included the percentage of female patients who utilized the ICU, the percentage of non-white patients who utilized the ICU, and the average age of patients who utilized the ICU. In addition, variables were created to represent the average number of comorbidities per ICU patient for each principal diagnosis. HCUP includes binary variables for up to 29 comorbidities that could be present for each patient. These binary variables indicate the presence of additional preexisting medical conditions that are not directly related to the principal diagnosis or the main reason for the patient's stay in the hospital. These variables are created by HCUP using AHRQ software that determines the existence of each comorbidity based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnoses and the discharge Diagnosis-Related Group (DRG). For the purposes of this study, these binary variables were used to calculate an average number of comorbidities per ICU patient for each principal diagnosis to be included as a control variable in the analyses performed.

Analysis

Once the data from the various sources were merged and tested for missing values, outliers and other underlying multivariate assumptions, univariate analyses were performed to provide basic descriptive statistics regarding the sample hospitals and the patients served by those hospitals. In this analysis, characteristics and attributes regarding the hospitals and patients were described and compared between the base year of 2007 and the final year of 2010.

In addition to the univariate analyses, bivariate analyses were also performed to further describe the data included in the study. In these analyses, hospitals that used intensivists were compared to hospitals that did not use intensivists using t-test and chi-square procedures. These procedures tested whether there was a statistical difference between the two groups of hospitals.

In order to test the two proposed hypotheses, the dependent variables included in the panel data were regressed against the independent and control variables. Separate fixed effects models were run for each of the two dependent variables for each of the four principal diagnoses for a total of eight separate models. Hospital fixed effects models were used in order to control for unobserved, time-invariant factors that may have affected the dependent variable. This would include any unobserved factors at the hospital, market or state level. Although the SID provides more granular data at the patient level, these data sets do not support tracking patients across years due to the absence of a consistent and unique patient identifier. Thus, given the relationships of interest and the potential for endogeneity, we opted for a panel fixed-effect regression model at the hospital-level. For each of these models, year fixed effects were also included to control for any time-based factors that might influence hospital operations. In addition, the models were clustered at the hospital level. SAS Version 9.3 was used for data management processes and STATA Version 13 was used for all of the statistical analyses included in the study. A P value of 0.10 or less was used as the statistical significance threshold.

Results

The results from the univariate analyses are shown in Table 3. These results describe the dependent, independent and control variables for the sample hospitals included in the first and last years of the study period. For the dependent variables, the average costs per patient day for each of the principal diagnoses included in the study were relatively similar in the base and final years. The compound average growth rate of the average cost per patient day from 2007 to 2010 ranged from 0.71% for CHF to 3.67% for patients with a principal diagnosis of stroke.

Average LOS for each of the principal diagnoses was also steady from the base year to the final year. There was less than a day difference in the average LOS for all four of the principal diagnoses. Three of the four principal diagnoses exhibited a decline in the average LOS from 2007 to 2010, while CHF showed a slight increase of 0.13 days from 9.49 to 9.62. The largest decline was seen in stroke patients, declining 0.72 from an average LOS of 9.73 in 2007 to 9.01 in 2010. Pneumonia patients experienced the longest average LOSs of 11.29 and 10.64 in 2007 and 2010 respectively, while AMI patients experienced the shortest stays of 6.97 in 2007 and 6.72 in 2010.

There was a notable increase in the use of intensivists in hospitals from the base year of 2007 to the final year of 2010. In 2007, only 26.44% of the sample hospitals used intensivists, while in 2010, 40.24% used them. In addition to the use or non-use of intensivists, the intensity of intensivist staffing also increased from 2007 to 2010. In 2007, the average intensivist FTE per ICU patient day was 0.111, while in 2010, the average had increased to 0.174 intensivist FTE per ICU patient day, representing a 56.76% increase.

The univariate analysis results show the HHI increased slightly from 0.592 in 2007 to 0.610 in 2010. This slight increase in the HHI indicates that there was slightly less competition and slightly more market power among the sample hospitals in 2010 than in 2007. Hospital-level attributes such as occupancy rate, the percentage of Medicare patients and the presence of cardiac ICU beds were fairly static from 2007 and 2010. There was, however, a 14.5% increase in nurse staffing intensity from 2007 to 2010. There were also increases in the percentage of Medicaid patients and the Medicare Advantage HMO penetration. The percentage of Medicaid patients was 26.59% in 2007 and 28.33% in 2010 while Medicare HMO penetration grew slightly from 24.43% to 26.65%. The average size of hospitals in 2010 was slightly larger than in 2007.

The various patient characteristics included in the study varied slightly from 2007 to 2010. The difference in the average age of patients with each principal diagnosis of interest was less than 1 year, with the exception of pneumonia, where the average age of patients declined 1.7 years. Similarly, the percentage of female patients in 2007 and 2010 was within 200 basis points from each other, with the exception of CHF patients, where the percentage of female patients increased 2.83 percentage points. The percentage of non-white patients for each diagnosis increased between 1.45 and 3.59 percentage points between 2007 and 2010, with pneumonia exhibiting the largest increase. Finally, the average number of comorbidities for patients with all four principal diagnoses increased by an average of 0.39 from 2007 and 2010.

In addition to the univariate descriptive analyses, bivariate analyses were also performed to compare hospitals that used intensivists to those that did not use intensivists. The results from these analyses are shown in Table 4. For the dependent variables in the study, both the average cost per patient day and the average LOS for patients with all four principal diagnoses were found to be higher at hospitals that used intensivists than those that did not use intensivists. The difference in the average cost per patient day ranged from $214 for pneumonia patients to $595 for AMI patients. The difference in the average lengths of stay ranged from 2.74 days for pneumonia patients to 4.17 days for CHF patients. All eight of these differences were statistically significant at the 99% confidence level.

For the control variables included in the study, hospitals that used intensivists were found to be in more competitive markets than those that did not use intensivists. In addition, hospitals that used intensivists tended to be larger, have more medical/surgical ICU beds, be more likely to have a cardiac ICU (and experience higher overall occupancy rates). Hospitals that used intensivists also tended to have a lower percentage of Medicare and Medicaid patients, but were in markets with a higher percentage of Medicare Advantage HMO penetration. Interestingly, there was not a statistically significant difference in nurse staffing levels (0.51 vs. 0.49).

For patient characteristics, hospitals that used intensivists had a higher percentage of ICU patients who were non-white for all four principal diagnoses. In addition, hospitals that used intensivists experienced statistically significant lower percentages of patients who were female for all the diagnoses except pneumonia. The analyses performed showed that hospitals that used intensivists had statistically significantly younger ICU patients. Finally, with the exception of pneumonia, there was not a statistically significant difference in the average number of comorbidities for patients with each of the principal diagnoses of interest.

The results from the multivariate analyses performed to test the average LOS hypothesis (H₁) are shown in Table 5. The multivariate analyses performed found no support for H₁. No statistically significant relationships were found between the use of intensivists and the average LOS for patients with any of the four principal diagnoses included in the study.

While the average LOS hypothesis (H₁) was not supported, the analyses performed did find a relationship between the average number of comorbidities and the average LOS. There was a positive relationship ranging from 0.58 to 1.23 days between the number of comorbidities and the average LOS for three of the four principal diagnoses of interest. In addition, the analyses found a positive relationship between the hospital occupancy rate and the average LOS. Finally, lower competition as measured by the HHI was associated with increased average LOS for patients with a principal diagnosis of pneumonia. All of these results were at the 95% confidence level.

The results from the multivariate analyses performed to test the average cost of care hypothesis (H₂) are shown in Table 6. Limited support was found for H₂. With the exception of ICU patients with a principal diagnosis of AMI, no statistically significant relationships were found between the use of intensivists and the average cost per patient day. For ICU patients with a principal diagnosis of AMI, however, the use of intensivists in hospitals was associated with an average cost per patient day $202.12 less than hospitals that did not use intensivists (P=0.09).

For the control variables used in the average cost per patient day analysis, a statistically significant positive relationship was found between the use of intensivists and Medicare Advantage HMO penetration for three of the four principal diagnoses of interest. For each percentage point increase in MA, the average cost per patient day was found to be $78.63, $61.21 and $81.17 higher for CHF, pneumonia and stroke diagnoses, respectively. Statistically significant negative relationships were also found between the use of intensivists and the percentage of female CHF patients, the average age for AMI patients and the presence of cardiac beds for pneumonia patients. No statistically significant relationships were found for the other control variables included in the analysis.

Discussion

Using agency theory as a theoretical framework, this study posited that the use of intensivists in the ICU would allow hospital administrators to better align the goals of the hospital with the goals of the physicians practicing in the ICU, resulting in more efficient, cost-effective care being provided to the patients. The study focused on patients who utilized the ICU during their hospital stay and had a principal diagnosis of CHF, AMI, stroke or pneumonia. The study included all patients meeting these criteria who were cared for by hospitals in both New York and Washington State for the years 2007 through 2010.

The average LOS was one measure used to operationalize the efficiency and cost of care provided. This study proposed that by utilizing intensivists in the ICU, the goal of the hospital to provide efficient care and the goals of the physicians providing that care would be more closely aligned and would therefore result in reduced average lengths of stay. The analyses performed, however, showed no statistically significant relationships between the use of intensivists and the average LOS experienced by patients with any of the four principal diagnoses of interest. While there were limitations to this study discussed below, based on the analysis performed, the data does not appear to support the notion that the use of intensivists will help hospital management achieve the goal of improved efficiency as measured by average LOS. Given this result, perhaps intensivists working in the ICU focus more on obtaining outstanding clinical results for patients than on moving the patients through the care process in the most efficient manner. Several studies have, in fact, found a relationship between the use of intensivists and improved outcomes such as patient mortality (36-38).

Multiple studies have also found a relationship between the use of intensivists and ICU efficiency (34-36). In our study, the average cost per patient day was used to test the efficiency and cost of care provided to patients. Again, based on agency theory, it was proposed that the use of intensivists would help the hospital better align its goals of providing cost-effective care with the goals of the physicians providing that care. The analysis performed provided limited support for this notion, with the finding that the use of intensivists is associated with a reduced average cost per day for AMI patients. These mixed results spurred additional post hoc analysis concerning the relationship between intensivist staffing intensity and the cost of care provided by the hospital.

Based on the mixed finding regarding the average cost per patient day, additional post hoc analyses were performed to test the relationship between the staffing intensity of intensivists and the average cost per patient day. The results from these post hoc analyses can be found in Table 7. These analyses show that the association between the staffing of intensivists and the average cost per patient day is nonlinear. For AMI patients, only the lowest (β=−245.17; P=0.07) and the highest levels of intensivist staffing intensity (β=−425.37; P=0.06) were associated with a lower cost per patient day, while no statistically significant relationships were noted for the middle two quartiles. For CHF patients, only the highest level of intensivist staffing intensity was associated with a lower cost per patient day β=−339.69; P=0.01). No statistically significant association was found between the intensity of intensivist staffing and a reduced cost per patient day for pneumonia or stroke.

While a nonlinear relationship can often be complex and require further exploration to fully understand, these post hoc analyses seem to indicate that the benefits derived from the use of intensivists can vary based on the type of patient being treated and the intensity of the intensivist staffing being utilized. The results appear to support the notion that it is possible for hospitals to use intensivists to help align the hospital goals of cost-effective care with the goals of the physician, at least for the care of certain types of patients and at certain levels of staffing intensity. The results suggest that hospital administrators should take a close look at the benefits gained or not gained in each specific scenario and should pay close attention to the intensivists’ staffing levels used to staff the ICU. The results of this study indicate that there are some cost benefits in using intensivists to staff the ICU, but those benefits do not extend across all types of patients and all staffing levels.

Policy and research implications

The use of intensivists represents a significant investment for both hospitals and the overall healthcare system. The question is whether this added cost is offset by the ability to provide more efficient, cost-effective care. This study and others before it have had mixed results and seem to indicate that benefits from utilizing intensivists are possible, but not necessarily in all circumstances. As researchers continue to explore the use of intensivists, special attention should be given to the specific scenarios in which the use of intensivists provides the most advantages. Why, for example, does the use of intensivists produce cost savings for AMI and CHF patients at certain levels of intensivist staffing intensity, but does not appear to produce cost savings for other patient types or levels of staffing intensity? Future studies should focus on determining the specific environmental and organizational factors that enable improved efficiencies and reduced costs when utilizing intensivists.

In addition to determining the specific factors that enable improved operations from the use of intensivists, more attention should be given to the appropriate staffing levels of intensivists in the ICU. This study found that the benefits could differ based on the staffing levels. Additional research should continue to explore the relationship between the specific staffing intensity of intensivists and the benefits derived from those staffing levels.

Finally, the use or non-use of intensivists in ICUs has noteworthy policy implications. As stated before, the use of intensivists represents a material investment and adds to the overall cost structure of the U.S. healthcare system. As policymakers grapple with the growing costs of healthcare and the need to improve the quality of the care provided, understanding the benefits of utilizing intensivists becomes imperative. Policymakers need to know whether the investments made in utilizing intensivists pay the dividends needed to offset the cost to the system. Policymakers also need to know whether the investments made in utilizing intensivists help improve the quality of care provided by the system. In order to answer these questions more definitively, policymakers should encourage further studies that explore the relationship between cost and quality and the use of an intensivist.

Limitations

While this study attempted to broaden the knowledge regarding the use of intensivists, there were some limitations encountered during the course of the analyses performed. While this was one of the first studies to include a relatively large number of ICUs, the study was limited to the hospitals with ICUs in New York and Washington State. This limitation was due to many states limiting the availability of data that was needed for the analysis. While this limitation was encountered, the study still included a varied group of hospitals from two diverse states. If the availability of data from various states improves, future research could utilize more recent data from a greater number of states to determine whether these updated and broadened data lead to findings consistent with the ones in this study.

In addition to the geographic limitation, the data utilized from the SID were from 2007–2010. Similar to the geographic limitation, this was driven by the general availability of the data needed for the analysis. The timeframe of the data utilized in the analyses may affect the current applicability of the findings in the study. Much like the geographic limitation, future research could build on this study by utilizing more recent data as it becomes available and determining if the findings are still applicable as processes, practices and policies have evolved over time.

Another limitation is that analysis was performed at the hospital level rather than the patient level. This was due to the limited nature of the data available. The data available regarding the use of intensivists was at the hospital level and only indicated the number of intensivists FTEs that were serving at the hospital at the point the AHA annual survey data was requested. The SID patient-level data used in the analyses did not indicate whether intensivists cared for the specific patients included in the data set, nor did it provide visibility into the specific protocols utilized within the ICU. Given this limitation, patient-level data were aggregated to the hospital level and merged with the hospital-level data regarding the use of intensivists and other hospital characteristics.

Similar to the hospital-level intensivists’ measure, the LOS measure used in the analysis was for the total hospital LOS for each ICU patient. This measure included the time spent in both the ICU and non-ICU areas of the hospital as opposed to the LOS attributed to just the ICU. Much like the other limitations, the availability of data was the driving force behind the limitation. ICU-specific LOS was not available at the patient-level detail and therefore, the average total LOS was calculated for all patients who utilized the ICU during their stay for each of the principal diagnoses of interest. A similar limitation exists for several control variables, where we utilized hospital-level measures due to the unavailability of ICU-specific data.

Finally, the cost measures used in the analyses were based on the total patient charges multiplied by the hospital-level cost-to-charge ratio provided by HCUP. These hospital-level cost-to-charge ratios are updated annually and provide a means to convert the patient charges included in the SID to hospital costs. The ratios, however, are based on all-payer inpatient costs and therefore, changes in payer mix by principal diagnosis were not taken into consideration. While diagnosis-specific cost-to-charge ratios would have been ideal, that data were not available and the annual hospital-specific ratios were deemed to be an adequate proxy across the four diagnoses of interest.

Conclusions

Hospital administrators and policy makers alike continue to search for strategies that improve the value of the care provided to patients. One strategy that has been proposed to reduce costs, improve efficiency and/or improve the quality of care provided is the use of intensivists in the hospital ICU. Several studies have explored the use of intensivists and its relationship with cost, efficiency and quality, yet the results of those studies have been mixed and have not been rooted in a theoretical framework. While the results have been mixed, there has been some support for the purported benefits of utilizing intensivists in the ICU.

This study attempted to build upon the existing knowledge regarding the benefits of intensivists and explored the relationship between the use of intensivists and the cost of care provided and the efficiency of that care. While no statistically significant relationships were uncovered between the use of intensivists and the average lengths of stays for patients, there was a relationship found between the use of intensivists and the average cost per patient day for certain patients and certain intensivist staffing intensities. While this study helped improve the body of knowledge regarding the use of intensivists, much work is still needed to more completely understand the benefits derived from the use of intensivists.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Journal of Hospital Management and Health Policy for the series “Healthcare Finance: Drivers and Strategies to Improve Performance”. The article has undergone external peer review.

Data Sharing Statement: Available at https://jhmhp.amegroups.com/article/view/10.21037/jhmhp-24-62/dss

Peer Review File: Available at https://jhmhp.amegroups.com/article/view/10.21037/jhmhp-24-62/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jhmhp.amegroups.com/article/view/10.21037/jhmhp-24-62/coif). The series “Healthcare Finance: Drivers and Strategies to Improve Performance” was commissioned by the editorial office without any funding or sponsorship. R.W.M. served as the unpaid Guest Editor of the series and serves as an unpaid editorial board member of Journal of Hospital Management and Health Policy from March 2025 to December 2027. L.R. reports receiving research grants from the Agency for Healthcare Research and Quality, the National Institutes of Health, and the Centers for Disease Control and Prevention. All of these grants are awarded to his affiliated institution. He also serves as a member of the Methodology Committee of the Patient-Centered Outcomes Research Institute. The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Porter ME. What is value in health care? N Engl J Med 2010;363:2477-81. [Crossref] [PubMed]
2. Teisberg E, Wallace S, O'Hara S. Defining and Implementing Value-Based Health Care: A Strategic Framework. Acad Med 2020;95:682-5. [Crossref] [PubMed]
3. Bodenheimer T. High and rising health care costs. Part 1: seeking an explanation. Ann Intern Med 2005;142:847-54. [Crossref] [PubMed]
4. Papanicolas I, Woskie LR, Jha AK. Health Care Spending in the United States and Other High-Income Countries. JAMA 2018;319:1024-39. [Crossref] [PubMed]
5. Martin AB, Hartman M, Lassman D, et al. National Health Care Spending In 2019: Steady Growth For The Fourth Consecutive Year. Health Aff (Millwood) 2021;40:14-24. [Crossref] [PubMed]
6. Chassin MR, Galvin RW. The urgent need to improve health care quality. Institute of Medicine National Roundtable on Health Care Quality. JAMA 1998;280:1000-5. [Crossref] [PubMed]
7. Dzau VJ, McClellan MB, McGinnis JM, et al. Vital Directions for Health and Health Care: Priorities From a National Academy of Medicine Initiative. JAMA 2017;317:1461-70. [Crossref] [PubMed]
8. Pronovost P, Thompson DA, Holzmueller CG, et al. Impact of the Leapfrog Group's intensive care unit physician staffing standard. J Crit Care 2007;22:89-96. [Crossref] [PubMed]
9. Nesse RE, Kutcher G, Wood D, et al. Framing change for high-value healthcare systems. J Healthc Qual 2010;32:23-8. [Crossref] [PubMed]
10. Orszag PR, Ellis P. The challenge of rising health care costs--a view from the Congressional Budget Office. N Engl J Med 2007;357:1793-5. [Crossref] [PubMed]
11. Bardhan IR, Thouin MF. Health information technology and its impact on the quality and cost of healthcare delivery. Decision Support Systems 2013;55:438-49.
12. Hillestad R, Bigelow J, Bower A, et al. Can electronic medical record systems transform health care? Potential health benefits, savings, and costs. Health Aff (Millwood) 2005;24:1103-17. [Crossref] [PubMed]
13. Campanella P, Lovato E, Marone C, et al. The impact of electronic health records on healthcare quality: a systematic review and meta-analysis. Eur J Public Health 2016;26:60-4. [Crossref] [PubMed]
14. Rudin RS, Friedberg MW, Shekelle P, et al. Getting Value From Electronic Health Records: Research Needed to Improve Practice. Ann Intern Med 2020;172:S130-6. [Crossref] [PubMed]
15. Petersen LA, Woodard LD, Urech T, et al. Does pay-for-performance improve the quality of health care? Ann Intern Med 2006;145:265-72. [Crossref] [PubMed]
16. Orszag PR, Emanuel EJ. Health care reform and cost control. N Engl J Med 2010;363:601-3. [Crossref] [PubMed]
17. Agarwal R, Liao JM, Gupta A, et al. The Impact Of Bundled Payment On Health Care Spending, Utilization, And Quality: A Systematic Review. Health Aff (Millwood) 2020;39:50-7. [Crossref] [PubMed]
18. Koh HK, Sebelius KG. Promoting prevention through the Affordable Care Act. N Engl J Med 2010;363:1296-9. [Crossref] [PubMed]
19. Kocher RP, Adashi EY. Hospital readmissions and the Affordable Care Act: paying for coordinated quality care. JAMA 2011;306:1794-5. [Crossref] [PubMed]
20. Zhao J, Mao Z, Fedewa SA, et al. The Affordable Care Act and access to care across the cancer control continuum: A review at 10 years. CA Cancer J Clin 2020;70:165-81. [Crossref] [PubMed]
21. Belden H. How pharmacists can work with intensivists on ICUs. Drug Topics 2002;
22. Gershengorn HB, Garland A, Costa DK, et al. ICU Staffing in the United States. Chest 2024;166:743-53. [Crossref] [PubMed]
23. Admon AJ, Kerlin MP. ICU Staffing and Organization: Everything Has Changed. Chest 2024;166:661-2. [Crossref] [PubMed]
24. Nurok M, Flynn BC, Pineton de Chambrun M, et al. A Review and Discussion of Full-Time Equivalency and Appropriate Compensation Models for an Adult Intensivist in the United States Across Various Base Specialties. Crit Care Explor 2024;6:e1064. [Crossref] [PubMed]
25. Dorman T, Pauldine R. Economic stress and misaligned incentives in critical care medicine in the United States. Crit Care Med 2007;35:S36-43. [Crossref] [PubMed]
26. Pronovost PJ, Needham DM, Waters H, et al. Intensive care unit physician staffing: financial modeling of the Leapfrog standard. Crit Care Med 2006;34:S18-24. [Crossref] [PubMed]
27. Pastores SM, Halpern NA. Insights into intensive care unit bed expansion in the United States. National and regional analyses. Am J Respir Crit Care Med 2015;191:365-6. [Crossref] [PubMed]
28. Krizner K. Intensivists bring experience to critical care medicine. Managed Healthcare Executive 2008;18:27-8.
29. Scurlock C, Raikhelkar J, Nierman DM. Targeting value in health care: how intensivists can use business principles to make strategic decisions. Physician Exec 2011;37:18-20, 22, 24.
30. Ohsfeldt RL, Choong CK, Mc Collam PL, et al. Inpatient Hospital Costs for COVID-19 Patients in the United States. Adv Ther 2021;38:5557-95. [Crossref] [PubMed]
31. Rosenfeld BA, Dorman T, Breslow MJ, et al. Intensive care unit telemedicine: alternate paradigm for providing continuous intensivist care. Crit Care Med 2000;28:3925-31. [Crossref] [PubMed]
32. Abate SM, Ahmed Ali S, Mantfardo B, et al. Rate of Intensive Care Unit admission and outcomes among patients with coronavirus: A systematic review and Meta-analysis. PLoS One 2020;15:e0235653. [Crossref] [PubMed]
33. Arabi YM, Azoulay E, Al-Dorzi HM, et al. How the COVID-19 pandemic will change the future of critical care. Intensive Care Med 2021;47:282-91. [Crossref] [PubMed]
34. Wortel SA, de Keizer NF, Abu-Hanna A, et al. Number of intensivists per bed is associated with efficiency of Dutch intensive care units. J Crit Care 2021;62:223-9. [Crossref] [PubMed]
35. Dimick JB, Pronovost PJ, Heitmiller RF, et al. Intensive care unit physician staffing is associated with decreased length of stay, hospital cost, and complications after esophageal resection. Crit Care Med 2001;29:753-8. [Crossref] [PubMed]
36. Breslow MJ, Rosenfeld BA, Doerfler M, et al. Effect of a multiple-site intensive care unit telemedicine program on clinical and economic outcomes: an alternative paradigm for intensivist staffing. Crit Care Med 2004;32:31-8. [Crossref] [PubMed]
37. Parikh A, Huang SA, Murthy P, et al. Quality improvement and cost savings after implementation of the Leapfrog intensive care unit physician staffing standard at a community teaching hospital. Crit Care Med 2012;40:2754-9. [Crossref] [PubMed]
38. Pronovost PJ, Angus DC, Dorman T, et al. Physician staffing patterns and clinical outcomes in critically ill patients: a systematic review. JAMA 2002;288:2151-62. [Crossref] [PubMed]
39. Jenson MC, Meckling WH. Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 1976;3:305-60.
40. Eisenhardt KM. Agency theory: An assessment and review. Academy of Management Review 1989;14:57-74.
41. Kohli R, Kettinger WJ. Informating the clan: Controlling physicians' costs and outcomes. Mis Quarterly 2004;28:363-94.
42. Ludwig M, Van Merode F, Groot W. Principal agent relationships and the efficiency of hospitals. Eur J Health Econ 2010;11:291-304. [Crossref] [PubMed]
43. Eisenhardt KM. Agency-and institutional-theory explanations: The case of retail sales compensation. Academy of Management Journal 1988;31:488-511.
44. Lee WI, Wang YH, Cheng JM, et al. Balancing managerial control systems and service quality: a case study of the national health insurance programme in Taiwan. International Journal of Management 2006;23:576.
45. Mott DA, Schommer JC, Doucette WR, et al. Agency theory, drug formularies, and drug product selection: implications for public policy. Journal of Public Policy & Marketing 1998;17:287-95.
46. Esposto AG. Contractual integration of physician and hospital services in the US. Journal of Management and Governance 2004;8:49-69.
47. Landon BE, O'Malley AJ, McKellar MR, et al. Physician compensation strategies and quality of care for Medicare beneficiaries. Am J Manag Care 2014;20:804-11.
48. Bai G, Coronado F, Krishnan R. The role of performance measure noise in mediating the relation between task complexity and outsourcing. Journal of Management Accounting Research 2010;22:75-102.
49. Valley TS, Schutz A, Miller J, et al. Hospital factors that influence ICU admission decision-making: a qualitative study of eight hospitals. Intensive Care Med 2023;49:505-16. [Crossref] [PubMed]
50. Cooney E. Intensivist team; Hospital's approach improves care for critically ill patients. Worcester, Mass: Telegram & Gazette; 2002:B.1-B1.
51. Yale New Haven Health Services Corporation/Center for Outcomes Research and Evaluation. 2013 Measures Updates and Specifications: Acute Myocardial Infarction, Heart Failure, and Pneumonia 30-Day Risk-Standardized Mortality Measure (Version 7.0). Available online: https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Mortality_AMI-HF-PN_Measures_Updates_Report_FINAL_06-13-2013.pdf
52. Hospital 30-Day Mortality Following Acute Ischemic Stroke Hospitalization Measure Methodology Report. Available online: http://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Measure-Methodology.html. September 2010.
53. Adamczyk P, Attenello F, Wen G, et al. Mechanical thrombectomy in acute stroke: utilization variances and impact of procedural volume on inpatient mortality. J Stroke Cerebrovasc Dis 2013;22:1263-9. [Crossref] [PubMed]
54. Brinjikji W, Rabinstein AA, Kallmes DF, et al. Patient outcomes with endovascular embolectomy therapy for acute ischemic stroke: a study of the national inpatient sample: 2006 to 2008. Stroke 2011;42:1648-52. [Crossref] [PubMed]
55. Foster ED, Cavanaugh JE, Haynes WG, et al. Acute myocardial infarctions, strokes and influenza: seasonal and pandemic effects. Epidemiol Infect 2013;141:735-44. [Crossref] [PubMed]
56. Griffin MR, Zhu Y, Moore MR, et al. U.S. hospitalizations for pneumonia after a decade of pneumococcal vaccination. N Engl J Med 2013;369:155-63. [Crossref] [PubMed]
57. Rodriguez F, Wang Y, Naderi S, et al. Community-level cardiovascular risk factors impact geographic variation in cardiovascular disease hospitalizations for women. J Community Health 2013;38:451-7. [Crossref] [PubMed]
58. Connor RA, Feldman RD, Dowd BE, et al. Which types of hospital mergers save consumers money? Health Aff (Millwood) 1997;16:62-74. [Crossref] [PubMed]
59. Ho V, Hamilton BH. Hospital mergers and acquisitions: does market consolidation harm patients? J Health Econ 2000;19:767-91. [Crossref] [PubMed]
60. Seago JA, Ash M, Spetz J, et al. Hospital registered nurse shortages: environmental, patient, and institutional predictors. Health Serv Res 2001;36:831-52.
61. McCue M, Mark BA, Harless DW. Nurse staffing, quality, and financial performance. J Health Care Finance 2003;29:54-76.
62. Flood SD, Diers D. Nurse staffing, patient outcome and cost. Nurs Manage 1988;19:34-5, 38-9, 42-3.
63. Glandon G, Colbert KW, Thomasma M. Nursing delivery models and RN mix: cost implications. Nurs Manage 1989;20:30-3.
64. Bazzoli GJ, Fareed N, Waters TM. Hospital financial performance in the recent recession and implications for institutions that remain financially weak. Health Aff (Millwood) 2014;33:739-45. [Crossref] [PubMed]
65. Clement JP, McCue MJ, Luke RD, et al. Strategic hospital alliances: impact on financial performance. Health Aff (Millwood) 1997;16:193-203. [Crossref] [PubMed]