Patient comorbidities increase postoperative resource utilization after laparoscopic and open cholecystectomy

An understanding of the relationship between patient factors and healthcare resource utilization represents a major point of interest for optimizing clinical care and overall net savings, yet maintaining financial margins for provider revenues. This study aims to review resource utilization after cholecystectomy in order to characterize patient factors associated with increased postoperative ED visits and 30-day readmissions. A total of 53,632 open and laparoscopic cholecystectomies were reviewed from July 2009 to December 2010 in a large private payer claims database. ICD-9 and CPT codes were available for each event, as well as basic demographics. Data regarding 30-day postoperative resource utilization metrics (emergency department visits and inpatient hospitalizations) were analyzed and stratified by key patient comorbidities. Differences between subgroups were evaluated with univariate and multivariable methods. Of the 53,632 patients studied, 71.2 % (38,171) were female and 28.8 % (15,461) male. Resource utilization within 30 days of surgery included: 6.6 % (3538) of patients with an ED visit and 7.7 % (4103) with an inpatient hospitalization. The most common comorbidities in the study population were: hypertension, hyperlipidemia, GERD/hiatal hernia, and diabetes mellitus. Patients with heart failure, cirrhosis, and a history of MI or acute ischemic heart disease all had a significant association with postoperative ED visit and the highest likelihood of inpatient hospitalization. Angina, diabetes, and hypertension similarly increased both ED utilization and inpatient readmissions to a lesser but still significant extent. Although patients with GERD/hiatal hernia and sleep apnea had a significant association with ED use, they did not have an increased likelihood of readmission. Patient comorbidity indexing plays a major role in clinical risk stratification and resource utilization for cholecystectomy. These factors should be considered in bundled reimbursement packages and in the creation of preventive postoperative ambulatory strategies given their role in determining potential resource utilization in the postoperative setting.

Were health accessibility, affordability, and outcomes similar between the US and the other countries listed above, perhaps this number would not be so troubling; existing data, however, show that these exorbitant costs did not and still do not translate to better care.

Accessibility
Before the implementation of the Patient Protection and Affordable Care Act (ACA), 16% of the entire US population -consisting of approximately 50-million people -were completely uninsured. Moreover, millions of Americans remained underinsured, meaning they lost a significant portion of their income to health care costs not covered by insurance 6,7 . In some states such as Nevada and Texas, the percentage of uninsured soared at an incredible 27% of the adult population 8 . This remains a stark contrast to the universal health coverage provided by our western counterparts in Europe, where the percentage of uninsured is virtually 0%. Figure 2 from the Organisation for Economic Co-operation and Development (OECD)'s health indicators 9 depicts the percentages of the population that were either insured (via public or private insurance) or uninsured across the various OECD participating countries in 2007.

Affordability
In the Schoen et al study 6 previously mentioned that included ten of the world's major developed nations in addition to the US, adults in the US were more likely than adults in any other nation to have healthcare access issues related to cost. For example, they were the most likely to have not seen a doctor when ill or not get recommended care due to cost. They also did not fill their prescription medications and skipped medication doses 2.6 times as often as adults in the country second most likely to have done so, Canada.
Finally, adults in the United States were more likely than adults in all other countries to have had $1,000 or more out-of-pocket medical expenditures.

Health Outcomes
Life expectancy and child mortality: In 2007, life expectancy at birth in the US (78.1 years) was one year less than the OECD average (79.1 years) and approximately two years less than that of any western European nation. Although life expectancy in the US (78.8 years) increased slightly in 2013, it staggered and lingered almost one year behind the OECD average (80.5 years) and 2-4 years behind the life expectancy in western European nations 9,10 .
Low birth weight: The US had one of the incidence levels of low birth weight infants in 2007, with 8.3% of infants born weighing less than 2,500g; the OECD average at the time was 6.8% of infants. Both incidence levels decreased in 2013, with the US at 8.0% of infants weighing less than 2,500g at birth and the OECD average at 6.6% of infants 9,10 .
Prevalence and incidence of various illnesses: The estimated prevalence of diabetes in the US (10.3%) ranked only second to Mexico (10.8%) in 2010, and led the OECD average In order to begin to address these overwhelming issues of poor access to health, inadequate insurance coverage, unsustainable costs, and the inferior value of health care in the US, the ACA was implemented in 2010 13 14 . Through these approaches, implementation of the ACA has slowly initiated a transformation in economic incentives that increase provider participation in order to achieve its goals of creating high value, affordable health care available to most 15 . Thus, the ACA triggered a closer look at healthcare spending with the aim to decrease inefficiencies, increase the quality of care that patients receive, and lower overall healthcare costs.

RISK STRATIFICATION OF BUNDLED PAYMENTS
In order to account for fluctuations in costs per patient, the ACA called for a risk adjustment strategy. Accordingly, bundled payments in the ACA and BPCI initiative were founded in risk stratification based on the existing Medicare Severity Diagnosis Related Group (MS-DRG). The MS-DRG attaches significance to acute illnesses or acute decompensations of chronic illnesses 16 . Thence, bundled payments to ACOs were to include increased payments based on MS-DRG risk stratification to to compensate for the higher costs of treating more severely ill patients.
The higher cost of treating sicker patients does not solely correlate with the acute disease burden represented by the MS-DRG, however; chronic disease burden significantly impacts a patient's total cost of care and may in fact play a greater role in resource utilization than acute disease burden does. Vertrees et al found that a patient's chronic disease burden may in fact be more predictive of a patient's resource use than the acute disease burden 16 . Hughes et al showed the predictive power of integrating the chronic disease burden of patient comorbidities into a system of risk categories in order to stratify patients by their expected use of healthcare resources in a future year 17 . Therefore, although existing risk adjustment under the ACA was founded in the known effects of acute illness, chronic disease burden may also prove a significant target.

CASE: THE CHOLECYSTECTOMY AND CONTRIBUTORS TO RESOURCE USE
Gallstone disease is the fourth most costly gastrointestinal disease in the United States; direct and indirect costs for care related to cholelithiasis totaled $6.2-billion in 2004 18 . Not surprisingly, the cholecystectomy is the most common elective abdominal surgical procedure performed in the United States with more than 750,000 cases occurring each year 19 . Considering the factors that contribute to the estimated $15,651 median cost per case 20 , complications and comorbidities may play a significant role in the total cost of care. Injury to the common bile duct is well known as a major contributor to cost and resource utilization, with immediate associated costs ranging from $13,612 to $30,000 and lifetime costs reported as high as $300,000 [21][22][23][24] . In addition, various patient, hospital, and surgeon factors -such as age, sex, presence of comorbidities, urgency of admission and surgery, length of stay (LOS), treatment in a regional or district hospital, hospital volume, and surgeon experience 25-32 -have been shown to affect cholecystectomy outcomes and associated overall resource use and costs.
As previously mentioned, chronic disease burden has been shown to be more predictive of patient cost and resource use than acute illness 16,17 . Prior studies have also evinced the influence of patient comorbidities on the utilization of resources such as office visits, emergency department (ED) visits, length of stay, drug costs, procedures, and overall costs in other study populations [33][34][35][36][37] .
Given the significance of chronic illness burden and the knowledge of how patient factors influence resource utilization in other conditions, this study serves to elucidate the impact of specific patient comorbidities on resource utilization after cholecystectomy. In the context of our current healthcare system's movement towards innovative reimbursement models, quantification of these impacts may be useful to payers for guiding the development of risk-stratified, case-mix appropriate bundled reimbursement packages to providers for cholecystectomies. Furthermore, this information may be useful to providers in order to create strategies to prevent excess resource utilization, thereby decreasing inefficiencies, increasing the value of care for patients, and simultaneously creating profit margins under the ACA's Shared Savings Program.

!
This was a retrospective observational cohort study of privately insured patients undergoing cholecystectomy in which predictor variables were defined by the presence or absence of various chronic comorbid conditions. Approval from the Cambridge Health Alliance institutional review board (IRB) was obtained.

STUDY DESIGN
This study was a retrospective cohort study utilizing a private payer claims database The endpoints in this study were 30-day ED visits, 30-day readmissions, and length of stay for the 30-day readmissions. A readmission was defined as utilization classified as inpatient in the source data. Length of stay was calculated as discharge date minus admission date.

STATISTICAL ANALYSES
Descriptive statistics were used to summarize baseline patient characteristics, including demographics and comorbid conditions. Outcomes were summarized for the overall population and by subgroups with and without each comorbid condition. Pearson's chi-squared p-values were calculated to evaluate the statistical significance of all differences in outcomes between cohorts. No adjustments were made to these p-values to account for multiple comparisons.
Logistic regression was utilized to determine the impact of baseline comorbid conditions on 30-day readmission rates, while adjusting for demographic differences. Stepwise selection was used to choose the statistically significant predictor variables from among all studied comorbid conditions, age, gender, and region. Sensitivity regression models were also run for the subsets of readmissions including and not including an overnight stay. All statistical analyses were conducted using SAS Version 9.2 (SAS Institute, Cary, North Carolina).

!
A total of 53,632 patients was included in this study; 71.2% (38,171) were female, 28.8% (15,461) male. Baseline patient characteristics are shown in Table 1. Of all cholecystectomies studied, 94.9% (50,900) were laparoscopic, 63.9% (34,285) were outpatient procedures, and 51.2% (27,439) occurred in the South (US Census Region 3) of the United States. The prevalence of chronic comorbidities seen in the study population are reported in Table 2.  Table 4).
ED visits were not significantly different across sexes but inpatient hospitalizations were, with males comprising a greater proportion of readmitted patients (9.9% males versus 6.8% females, p<0.0001) ( Table 5). Resource utilization within 30-days of cholecystectomy, stratified by comorbid conditions, is shown in Table 6. All of the comorbidities studied, except for hyperlipidemia and hypertensive heart disease, had a significantly increased postoperative ED utilization.
Odds ratios and 95% Wald confidence limits for the effect of baseline comorbidities on 30-day readmission are shown in Table 7. All studied comorbid conditions except GERD and Lengths of stay were calculated by subtracting discharge date from admission date. LOS of 0-days informs that the admission and discharge date were the same, suggesting discharge from the hospital before midnight on the same day as admission. In this study, patients with heart failure were the most likely of all to visit to the ED and to be readmitted; in fact, these patients are 2.1 times more likely than healthy patients to be readmitted within 30-days of their cholecystectomy. Patients with either cirrhosis or a history of acute IHD also had significant association with postoperative ED visit and were 1.5-1.6 times as likely to have an inpatient readmission. Those with chronic IHD, diabetes mellitus, and hypertension had a significant association with ED visit and also had a modestly increased likelihood of inpatient readmission that was 1.2-1.3 times that of their healthy counterparts.
Forty-six percent of cholecystectomy costs are attributable to room and board 20 .
Therefore, it is prudent to note the influence of the comorbidities that increase readmission rates and inpatient LOS after cholecystectomy in addition to those that increase ED utilization.
Hyperlipidemia did not increase ED utilization and was a negative predictor of inpatient readmissions in this study population. This may reflect a population of otherwise healthy patients presenting with a 'laboratory comorbidity' as defined by strict numbers yet without any type of actual visceral organ disease. These data demonstrate how two patients with similar costs but different chronic conditions lead to different utilization of resources after surgery: a patient with GERD who visits the ED repeatedly for reflux symptoms and pain in the postoperative setting versus a patient with cirrhosis and heart failure requiring a prolonged inpatient hospitalization postoperatively due to fluid shifts and heart failure exacerbation after intraoperative fluid resuscitation. Although these patients may be very similar in a strictly monetary sense, their use of hospital resources is vastly different. These insights will prove extremely valuable for applying paradigms for care coordination and cost savings under the ACA.

RELEVANCE TO PAYERS
An understanding of the specifics of resource utilization by patient comorbidity will first allow for appropriate departmental reimbursements by insurance companies. In this analysis, not every comorbidity that increased utilization of ED resources led to increased readmission rates.
The two comorbidities that were associated with an increased likelihood of ED visit that did not lead to inpatient readmissions were GERD / hiatal hernia and sleep apnea. Similarly, the impact of resource use will vary from organization to organization based upon the characteristics of their patient population.
Second, an understanding of the impact of each comorbidity from this paper can form the foundation for the risk adjustment of bundled payments for cholecystectomies based on a defines the 'break-even point' for laparoscopic cholecystectomy at 454 cases -this presents a major financial limitation to lower-volume centers performing this operation and therefore may negatively impact access to care. Analysis of the comorbid conditions in a particular healthcare group's coverage will aid in the projection of resource utilization and subsequent administrative planning for cholecystectomy. Such preparation will enable various processes: Opportunities for the creation of preventive strategies, optimization of healthcare delivery, and appropriate distribution of reimbursement profiles to enable increased profit margins for providers, in the context of the risk-adjusted reimbursements from payers mentioned above.
From the clinician perspective, the knowledge gained from this study will be used to guide strategies to optimize patients in the perioperative period and consequently, decrease postoperative resource utilization. Strategically preventing exacerbation of existing patient comorbidities in the perioperative setting represents one of the major goals of the BPCI and is currently a hot topic in the surgical care realm. To undertake this daunting task, administrators, physicians, and ancillary staff have come together in the few years since the implementation of the ACA to create the perioperative surgical home (PSH) model of care 41 .
A mélange of evidence-based practices from lean six sigma theory and its predecessor,

Preoperative Period
Physicians could coordinate close follow-up in the 1-2 weeks prior to surgery for patients most likely to be readmitted. Such care could include weekly in-office visits or frequent phone calls with ancillary medical staff for careful titration of medications aimed at optimizing the patient's physiologic balance before surgery.

Intraoperative Period
During the intraoperative period, anesthesiologists must maintain equilibrium despite often opposing physiologic and surgical needs, with the added burden of chronic illness.
If a patient is afflicted with a comorbid condition that significantly increases their likelihood of readmission such as heart failure in this study's dataset, the anesthesiologist could consider immediate preoperative evaluation in the holding area with a bedside pulmonary and/or transthoracic ultrasound. This cheap and quick bedside examination could provide valuable information regarding the patient's cardiac function and overall volume status via the presence of pathology such as pulmonary edema, pleural effusions, and inferior vena cava variability 44 . The anesthesiologist could then use this information to guide a restrictive or liberal volume resuscitation protocol intraoperatively, depending on other surgical factors as well.

Postoperative Period
In addition to the existing foundations for surgical recovery laid out by ERAS, clinicians could coordinate close follow-up care targeted at preventing exacerbations of chronic conditions for the patients statistically most likely to return to the hospital based on their disease states. Perioperative care physicians could coordinate follow-up telephone calls, home ambulatory services such as Visiting Nurse Association (VNA) visits, and outpatient follow-up with primary care providers. Without the added assurance of inhouse hospital monitoring, the postoperative period provides a unique environment in which the use of innovative technology could have great potential. Mobile health technologies could be provided to the highest-risk patients for real-time monitoring by third-party agencies and reporting to ancillary staff for health optimization. This novel application of technology could prevent chronic health exacerbations at the first sign of pathology, before symptoms ever developed and well before patients visited the ED or were readmitted. Devices could include smart watches for heart rate monitoring, at-home scales and sphygmomanometers with real-time data transmission, and more.
These services could not only lead to less ED visits and inpatient hospitalizations, but also result in better coordination of care, create greater value of services received for patients, and yield lower overall costs to the healthcare system. Although a study using aggregates of comorbidity information could have led to similar findings in terms of identifying trends in resource utilization with increasing chronic disease burden, aggregate patient data would not have allowed for the specific information provided by this study's data to create targeted prevention strategies for individual comorbidities.
The knowledge gained from this study can be used to: (1) Influence payers to create evidence-based payment models with global reimbursements focused on clearly defined packages of patient-centered and risk-stratified post-discharge resources in the postoperative setting after cholecystectomy; and (2) Guide providers to create targeted cost saving interventions in the pre-, intra-, and post-operative periods aimed at decreasing resource utilization for patients in the highest risk groups.

SUGGESTIONS FOR FUTURE WORK
Although the prevention strategies mentioned are certainly possible for reducing postoperative resource utilization, an understanding of the cause of these readmissions must first be established. Suggestions for future work include root cause analyses to determine the factors contributing to ED visits and readmissions in the postoperative setting for the comorbidities mentioned in this study. Additional work thereafter could include application of tactics targeted at these causes and an analysis of the impact of such interventions.

LIMITATIONS
Despite the advantages of using a large private payer database for studying postoperative resource utilization after cholecystectomy, there are several limitations to this approach. The major limitations relate to the use of claims data, namely that we must use ICD-9 diagnosis codes to determine patient comorbid conditions and that the study population is not nationally representative. The data source used in this study only included subscribers to a single private insurance provider and did not include the uninsured population. Using ICD-9 codes to determine comorbidities can lead to under-representation of the conditions in the study population, though coding of chronic conditions tends to be much more complete at the time of hospitalization and/or surgical intervention. Although the comprehensive nature of patient utilization captured in claims data is helpful for elucidating 30-day postoperative resource utilization, it likely overestimates the use of resources associated with cholecystectomy, as the etiology of the visit or hospitalization could be another, unrelated source.
Obesity is known to both correlate with and contribute to the pathophysiology of many of the conditions associated with increased utilization of resources in this study, such as hypertension, hyperlipidemia, GERD and hiatal hernia, diabetes mellitus, heart disease, sleep apnea, and heart failure 45 . The multivariate analyses we conducted allowed for the evaluation of each risk factor independently of the other but did not allow for the evaluation of the effects of obesity, a variable excluded from the study due to severe under-reporting of obesity codes.
Martin et al's Canadian study 46 revealed that a diagnosis of obesity had a sensitivity of a mere 7.75% in their database of 17,380 patients; similarly Januel et al 47 show the scanty sensitivity of obesity coding in Switzerland (ranging from 29.4% in 1999 to 51.5% in 2003). These studies convey the extent to which obesity is under-reported in administrative databases. Therefore, it was not possible to determine the resultant effect of obesity on postoperative resource utilization in this study. In accordance with the ACA mandate that healthcare claims data be used extensively to assess resource utilization and quality of care starting in 2015 13 , it is inevitable that large payer and other administrative databases will be increasingly used to evaluate healthcare systems in the near future. In order for authorities involved in all aspects of healthcare to draw accurate conclusions from these databases, more precise coding information regarding obesity needs to be included.  Figure 1: Profiles of health spending and coverage in eleven countries 6