The purpose of this study was to examine the factors associated with discharge lag time and how lag time relates to the discharge process.

Delays in patient discharge have major implications for patient bed allocation and hospital throughput. Discharge lag time is a metric that can be used to monitor discharge efficiency.

This was a retrospective review of all patients undergoing colorectal surgery who were discharged from a tertiary care facility between January 1, 2007, and May 21, 2014.&ensp;The effects of various factors were examined using χ

 tests for categorical data and 2-sample 

 tests for continuous data.&ensp;Any factor that had a 

 value <.20 on univariate analysis was incorporated into the multivariate ordinal logistic model.

During the study period, 1707 patients were discharged. Factors that were found to correlate with a longer discharge time using univariate analysis (

 <.05) were gender, primary diagnosis of cancer, type of surgical procedure, time the order was written, use of a discharge navigator, and discharge destination. Day of the week did not meet statistical significance on univariate analysis, but was included in the multi&shy;variate analysis. Upon multivariate analysis, the only factors that correlated with a longer lag time were gender (

 = .012), day of the week that the discharge occurred (

Discharge orders placed earlier in the morning cause an increase in lag time and a backlog of patient discharges. It is important for hospitals to have adequate personnel to handle these early discharges.

The American Journal of Accountable Care. 2018;6(1):20-24

With healthcare spending continuing to increase while reimbursements are becoming more regulated and fixed, many healthcare systems have begun looking at ways to reduce costs.

 Efficient hospital discharge is one component of a strategy to decrease healthcare spending.

Delays in discharge not only increase the cost of the hospitalization for the patient and the facility, but they also decrease hospital revenue by occupying beds that could go to other patients.

 Discharge delays also affect patient satisfaction, because other patients remain in the emergency department or postanesthesia care unit for extended times, and can hinder patient care by preventing prompt transfers to tertiary care facilities. Reasons for delays in discharge and increased length of stay for seemingly unnecessary reasons have been well studied. The problem is multifactorial and usually due to poor coordination of care and lack of communication,

 which are issues that could be remedied with personnel dedicated to the discharge process.

The use of discharge navigators, coordinators, and/or planners has been increasing in an effort to help streamline the transition to posthospital care.

 Discharge navigators have been shown to be effective at increasing patient satisfaction with the discharge process and are more likely to have discharge documentation, follow-up, and prescriptions prescribed and filled earlier, while alleviating some of the burden carried by the patient’s medical team.

 These discharge navigators act as a major component of the continuum of care between the inpatient and outpatient settings.

Discharge lag time, the time from discharge order entry to actual patient exit from the hospital, is one metric used to quantify discharge efficiency. The ability to better understand this variable will allow hospitals to plan for bed allocation. The goal of this study was to examine discharge lag times for patients undergoing colorectal operations and examine factors that increase or decrease these times.

This study was a retrospective review of all patients 18 years or older undergoing major colorectal surgery who were discharged from Geisinger Medical Center (GMC) in Danville, Pennsylvania, between January 1, 2007, and May 21, 2014. This study was reviewed and approved by the GMC institutional review board.

The primary outcome of the study was discharge lag time, which was defined as the interval between when the discharge order was written and the patient’s departure from the hospital. Discharge lag time was categorized in 3 groups: less than 2 hours, 2 to 4 hours, and more than 4 hours.

The discharge navigator is a physician extender who meets with the surgical team early in the morning to get the plan for patients about their possible discharge and what needs to be done in order to get the patient out of the hospital. These individuals work closely with the surgical team, patient, and care manager to facilitate each patient’s discharge. They place the order for discharge, schedule follow-up appointments, handle wound and drain care education, and provide prescriptions for medications upon discharge. They are only employed Monday through Friday; therefore, all discharges occurring during the weekend are the responsibility of the physician.

Patient, perioperative, and peridischarge variables were used to test associations among the variables and discharge lag times. Patient characteristics such as age, gender, body mass index (BMI), Charlson Comorbidity Index (CCI) score, and distance from hospital (using great-circle distance calculation from the patient’s home zip code to the hospital) were captured. Surgical characteristics such as primary diagnosis, procedure, approach, and duration of surgery also were collected. The discharge characteristics collected were day of discharge (eg, day of the week), time the order was written, presence of a discharge navigator, number of days since surgery (postoperative day), and discharge destination. Data were retrospectively collected from the electronic health record and the institution’s comprehensive enterprise-level data warehouse.

The characteristics of the study population were described using mean ± standard deviation for continuous data, median (interquartile range) for nonparametric data, and frequency (percentage) for categorical data. Ordinal logistic regression was used to test the association of each variable with discharge lag time, and data were represented as odds ratios (ORs) and 95% CIs. Any variable with a 

value <.20 on univariate analysis was incorporated into a multivariate ordinal logistic model. All significance testing was 2-sided, with α set at 0.05. Analyses were performed using SAS 9.4 (SAS Institute; Cary, North Carolina).

A total of 1707 patients met inclusion criteria; a breakdown of the demographics of the population can be seen in 

. Of these, 746 patients had a discharge lag time of less than 2 hours, 612 had a discharge lag time between 2 and 4 hours, and 349 patients had a discharge lag time longer than 4 hours.

Patient, perioperative, and peridischarge variables used to test associations among the variables and discharge lag times are shown in 

. The mean patient age was 61.4 years, which was not associated with discharge lag time (OR, 1.004; 95% CI, 0.998-1.009). The mean BMI was 29.6, which was not associated with delays in discharge (OR, 1.006; 95% CI, 0.994-1.018). Female patients (51.9%) had significantly longer lag times (OR, 1.224; 95% CI, 1.025-1.461) than male patients. The median CCI score was 2 and was not associated with discharge lag time (OR, 0.999; 95% CI, 0.961-1.040). The median distance patients lived from the hospital was 21.5 miles and was not associated with discharge efficiency (OR, 1.003; 95% CI, 0.996-1.010).

Patients with colon or rectal cancer (39.2%) had significantly longer lag times (OR, 1.294; 95% CI, 1.079-1.551) compared with patients who were undergoing surgery for a noncancer-related reason. Patients who had undergone colon resection (67.1%) were found to have shorter lag times compared with those who had undergone rectal resection (OR, 0.826; 95% CI, 0.658-0.997). Surgical duration and approach were not associated with discharge efficiency.

The median postoperative day discharge was 5 days and was not associated with lag time (OR, 0.997; 95% CI, 0.985-1.009). The majority of patients were discharged to their home (76.2%), followed by a skilled nursing facility (SNF) (19.2%) and other (eg, prison, long-term acute care hospital, nursing home, death) (4.6%), which showed a difference in lag time among the 3 groups in that patients discharged to their home had a shorter discharge lag time.

The majority of discharge orders were written before 11 am (51.9%), which was found to be associated with a significant increase in lag time (OR, 9.789; 95% CI, 7.313-13.103) compared with

orders placed after 2 pm. Orders written by a discharge navigator (24%) also increased lag times (OR, 1.798; 95% CI, 1.462-2.212) compared with discharge orders written by a physician. The day of the week (including weekends) was not associated with a delay in discharge.

Next, a multivariate analysis was performed; results are shown in 

. Female patients were again found to have longer discharge lag times (OR, 1.272; 95% CI, 1.054-1.535) than males. Primary diagnosis of cancer and type of resection were no longer associated with lag time; however, day of the week was significantly associated with lag time (

= .03). Discharge orders written before 11 am had significantly longer lag times compared with orders written after 2 pm&#8203; (OR, 9.845; 95% CI, 7.283-13.309). Patients discharged to their home had shorter lag times than those who were discharged to SNFs (OR, 0.738; 95% CI, 0.580-0.939).

Similar to the findings of prior research regarding the discharge process, the results of this study also indicate that delays in discharge are multifactorial. Women had longer discharge lag times, as did patients discharged to SNFs. The hour at which the discharge order was written had a significant effect on discharge efficiency, with earlier orders associated with longer lag times.

Findings of a study by Bozorghadad et al had demonstrated that the use of discharge navigators allowed for earlier discharge from the hospital.

 However, that study did not specifically look at lag time. Interestingly, our study revealed that discharge navigators increase lag time. One can theorize that the effect of the discharge navigator and the time of the order entry are interrelated. Increased efficiency of the discharge navigator in order entry may be creating a backlog of patients and thereby creating a backlog of work for nursing to get these patients discharged.

The most important finding of this study is that the lag time was highest when discharge orders were written before 11 am. This is most likely due to the fact that the highest number of discharges were taking place in that time period. This phenomenon can be remedied by adding personnel during these hours. With this additional support, hospitals can achieve the goals of early discharge and shorter lag time.

Discharge destination was found to be statistically different in terms of lag time

on both univariate and multivariate analysis. Discharge to home had the shortest lag time, whereas being discharged to a SNF increased this time. This could be due to poor communication and coordination with the receiving facility and/or issues with coordination of transportation services. In addition, bed availability at the receiving institution can affect lag time. Transportation is a factor that, in many cases, is beyond the hospital’s control. However, better control of transportation in the future could improve discharge efficiency.

Hospital discharge efficiency is multifactorial and often influenced by factors beyond the hospital’s control. Elements of the discharge process, such as discharge navigators, which improve certain aspects of the discharge process (ie, time of day that orders are written; earlier-in-the-day discharge), can disrupt other aspects (discharge lag time). To achieve a more efficient process, it is necessary for hospitals to have an appropriate number of staff to deal with these earlier discharges to allow beds to be free for incoming patients.

Department of General Surgery, Geisinger Medical Center (CW, SB, LS, C-YS, JD, MH, JW, JB, MS), Danville, PA.

 The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Concept and design (CW, SB, LS, C-YS, JW, JB, MS); acquisition of data (CW, SB, LS, MS); analysis and interpretation of data (CW, LS, JD, JB, MS); drafting of the manuscript (CW, SB, C-YS, JD, MH, JW, JB); critical revision of the manuscript for important intellectual content (CW, SB, LS, C-YS, MH, JW, MS); statistical analysis (JD, JB); provision of study materials or patients (CW, SB, LS, CY-S, JD); obtaining funding (CW, MH, JB, MS); administrative, technical, or logistic support (MH, MS); and supervision (JW, MS).

Charles Walker, MD, Geisinger Medical Center, 100 N Academy Ave, MC 21-69, Danville, PA 17822. Email: cwwalker@geisinger.edu.

1. McClellan M. Reforming payments to healthcare providers: the key to slowing healthcare cost growth while improving quality? 

2. Richardson DB. The access-block effect: relationship between delay to reaching an inpatient bed and inpatient length of stay. 

3. Taheri PA, Butz DA, Greenfield LJ. Length of stay has minimal impact on the cost of hospital admission. 

. 2000;191(2):123-130. doi: 10.1016/S1072-7515(00)00352-5.

4. Bonnema J, van Wersch AM, van Geel AN, et al. Cost of care in a randomised trial of early hospital discharge after surgery for breast cancer. 

. 1998;34(13):2015-2020. doi: 10.1016/S0959-8049(98)00258-5.

5. Chen LM, Freitag MH, Franco M, Sullivan CD, Dickson C, Brancati FL. Natural history of late discharge from a general medical ward. 

. 2009;4(4):226-233. doi: 10.1002/jhm.413.

6. Kim CS, Hart AL, Paretti RF, et al. Excess hospitalization days in an academic medical center: perceptions of hospitalists and discharge planners. 

7. Majeed MU, Williams DT, Pollock R, et al. Delay in discharge and its impact on unnecessary hospital bed occupancy. 

. 2012;12:410. doi: 10.1186/1472-6963-12-410.

8. Ngui NK, Hitos K, Ctercteko G. Preoperative factors prolonging the length of stay in elective colorectal surgery. 

9. Ou L, Young L, Chen J, Santiano N, Baramy LS, Hillman K. Discharge delay in acute care: reasons and determinants of delay in general ward patients. 

. 2009;33(3):513-521. doi: 10.1071/AH090513.

10. Day MR, McCarthy G, Coffey A. Discharge planning: the role of the discharge co-ordinator. 

. 2009;21(1):26-31. doi: 10.7748/nop2009.02.21.1.26.c6871.

11. Finn KM, Heffner R, Chang Y, et al. Improving the discharge process by embedding a discharge facilitator in a resident team. 

. 2011;6(9):494-500. doi: 10.1002/jhm.924.

12. Houghton BA. Discharge planners and cost containment. 

13. Punch L. Cost containment: discharge planners take more active role. 

14. Bozorghadad S, Dove J, Scholtis L, et al. A hospital discharge navigation program: the positive impact of facilitating the discharge navigation process. 

Articles

The American Journal of Accountable Care

Examination of factors associated with discharge lag time and how this metric plays an important role in managing hospital throughput.

Predictive Factors of Discharge Navigation Lag Time

Rates and patterns of referrals by family physicians to specialists have significant effects on numerous aspects of medical care. Understanding the referral patterns of residents is important, as doing so may guide residency program curriculum development to ensure appropriate broad-scope training. This study examined and compared the outpatient referral rates of family medicine residents, residency faculty, and clinical faculty. 

We conducted a retrospective electronic chart review of referrals associated with clinic visits of patients who received primary care services from residents, residency faculty, and clinical faculty.

The primary outcome measured was the proportion of patient visits per year that resulted in a referral for services or a specialist visit.&ensp;We used random effects meta-analysis methodology for comparisons among physicians.

The overall referral rate was less than 20% for all types of referrals, but did increase from fiscal year (FY) 2014 to FY 2015. For medical referrals to other specialists, the residents referred significantly more than the residency faculty in FYs 2014 and 2015.

The referral rates among residents, clinical faculty, and residency faculty caring for similar patient populations can be significantly different. Further multicenter analysis to determine the reasons for this is needed to ensure that residents are receiving the full scope of training necessary to care for a diverse patient population.

The American Journal of Accountable Care. 2018;6(1):25-28

Rates and patterns of referrals by generalists, such as family physicians, to specialists have significant effects on numerous aspects of medical care, including the quality and cost of care provided. The appropriate use of referrals can decrease diagnostic uncertainty and assist with the implementation of proper treatment that improves the quality of care provided. Study findings have suggested that the knowledge base and quality of care provided by specialists for specific conditions may exceed that of generalists; however, specialist care can be lacking in preventive care, disease management, and health maintenance common to generalists.

 The inappropriate use of referrals can be inconvenient, costly, and inefficient for patients, as additional and unnecessary tests and procedures may be ordered.

As reimbursement structures shift toward value-based care, more scrutiny is being placed on the amount and quality of care provided by primary care physicians. One area that warrants evaluation is referral rates, as they have increased over the past decades, leading to higher costs of care. According to the National Ambulatory Medical Care Surveys, between 1999 and 2009, the probability of receiving a referral during an ambulatory patient visit increased from 4.8% to 9.3%, a 92% increase.

 Another study found a wide range of referrals from primary care physicians, from less than 5% of patients per year to greater than 60%.

Limited research has examined the rates of academic ambulatory clinic referrals, including resident referrals. Between 1977 and 1985, the referral rate to specialist physicians was 1.4% for 1 family medicine residency, which was actually less than the 2.7% rate published by the National Ambulatory Medical Care Survey from 1971.

 By the 1990s, the referral rates of internal medicine residents were 19.8%,

 whereas rates of family medicine residents were reported at 4.3%.

 When considering academic generalists’ primary reasons for referrals, most reported referring patients to get diagnostic or therapeutic advice or to have a specialist perform a diagnostic or therapeutic test. For nonmedical referrals, generalists reported that the primary reasons were following perceived standards of care, patient requests, and self-education.

Understanding the referral pattern of academic physicians and residents is important as we strive to produce well-equipped primary care physicians capable of caring for a growing population with diverse needs and to create a more efficient healthcare system. Examining referral patterns may highlight additional services or training needed to address those specific patient care needs and to guide residency program curriculum development. Therefore, the aim of this study was to examine the outpatient referral rate of family medicine residents to begin exploring services or training needed for residents to ensure they maintain a broad scope of training. Additionally, the referral rate of family medicine physicians who teach, supervise, and practice in the same clinic as the residents was examined, as well as referrals from physicians not directly associated with the residency program who were practicing at a separate, yet similar, clinic in the same community.

We conducted a retrospective electronic chart review of referrals associated with clinic visits of patients who received medical care from our Community Health and Family Medicine Clinics associated with the University of Florida over 2 academic years, July 1, 2013, to June 30, 2015, in 3 distinct practices: residency, residency faculty, and clinical faculty. The resident and residency faculty practices are housed in the same clinic building, whereas the clinical faculty practice is in another building located approximately 3 miles away and treats the same underserved patient population. The clinical faculty practice includes a pediatrician and a pediatric nurse practitioner. All other providers in the 3 practices are family physicians. The residency and residency faculty practices include embedded procedures, women’s health, and sports medicine clinics at their site, which are staffed by family medicine faculty at that site but were not available to the clinical faculty practice. The Institutional Review Board and Privacy Office at the University of Florida approved this study.

Data were obtained from the University of Florida Physicians Decision Support (UFPDS) team and the reporting software Cognos Impromptu (IBM; Armonk, New York). The information obtained from UFPDS on patients and associated referrals included patients’ demographic information (medical record number, age, gender, race, ethnicity, and primary health insurance coverage); clinic visit information that resulted in the referral (type of visit, reason for visit, visit location, and provider); 

International Classification of Diseases, Ninth Revision

 diagnoses related to the referral; specialty being referred to; and referral type (internal or external). For referral type, referrals were considered internal if they were to our own family medicine clinic and external if they were to any other provider or service, whether in our academic health system or outside of our system. Cognos Impromptu provided data on all clinic visits, regardless of referrals, occurring in that time period, which were used to calculate the denominator. The Cognos Impromptu report included clinic visit dates, patient information, and primary insurance coverage. The 2 datasets were combined, linking all referrals by the clinic visit encounter from which the referrals were placed.

The primary outcome measured was the proportion of patient visits per year that resulted in a referral for services or a specialist visit outside of the family medicine department (all referrals), termed the referral rate. These excluded the internal referrals to the procedure, women’s health, or sports medicine clinics within our practice that were part of the workflow process. The referrals were broken down into “all referrals,” which encompassed a referral outside of our family medicine clinics for any service or visit, and “medical referrals,” which were referrals placed to physician specialists as opposed to referrals for ancillary services (ie, physical therapy, imaging, etc). Two academic years were used to allow for comparison of resident referral patterns between consecutive years of training to determine if residents were more or less likely to refer patients based on their year of training. Analyses were stratified by physician, and encounters were included only if the physician had encounters in both academic time periods. Proportion estimates and comparisons among them used random effects meta-analysis methodology, using Comprehensive Meta-Analysis version 3.0 (Biostat, Inc; Englewood, New Jersey), using physicians as “studies.” This enabled us to account for repeated measures within each physician and, when years were compared, to pair within physician by year.

To be eligible for analysis, physicians had to have the potential to place referrals in both years. The experimental unit in such an analysis was the physician, not the patient. A patient-level analysis, such as logistic regression, was not an acceptable approach, as it would either treat the physician as a fixed effect or ignore the variance component between physicians, thereby underestimating the true sampling errors of the estimates. Random-effects meta-analysis methodology,

 in which each individual physician takes the role of “study,” controls for between-physician confounding factors when comparing years. Because we have repeated measures by year, we can evaluate between-year differences in a causal manner, but we cannot attribute cause and effect when contrasting centers. 

Referral rates per physician were calculated as number of referrals divided by the number of patient visits per year. Internal referrals within the department to the procedure, women’s, or sports clinics, which were staffed by our family medicine physicians, were excluded. Demographic characteristics of patients were compared by analysis of variance (age) and Pearson’s χ

The referral rates (all referrals and medical referrals) for the 3 practices over the 2 academic years are listed in 

. The overall referral rate significantly increased from 2014 to 2015 for all clinical practices, but the medical referral rates did not significantly change between years. For medical referrals (

), the residents referred significantly more than the residency faculty in 2014 and 2015, but not significantly more than the clinical faculty. There was no significant difference in medical referrals between residency faculty and clinical faculty, nor in referral rates of individual physicians between the 2 years analyzed.

The overall referral rate was less than 20% (Table 1), which is not vastly different than that reported by internal medicine residents in 1992,

 although the medical referral rate range of 8.5% to 13.4% in the current study is substantially higher than the 1.4% average from family medicine residents in the 1980s.

 Study findings have suggested that physicians who are less comfortable with uncertainty or reluctant to disclose uncertainty to patients or other physicians refer more patients to specialists and/or order more testing.

 Perhaps residents, as newer physicians, are less comfortable with uncertainty compared with more experienced faculty. Residents may also feel that referring patients to specialists is educational and complementary to their training, as they follow what the specialist does for further workup and treatment of the patient.

Another possibility for the increased referral rate might be the knowledge gap between residents and residency faculty. However, residents are required to precept all patients with faculty to receive assistance with formulating a plan. Because there is generally less continuity of care in resident practices, perhaps residents and/or preceptors are more likely to refer patients to ensure appropriate follow-up is guaranteed. Patient requests and pressure may also contribute to higher resident referral rates.

 Other studies have shown that younger physicians may experience more pressure from patients for a referral.

A large number of referrals were to ancillary services (Table 1), such as physical therapy, occupational therapy, and nutritional counseling, and this number significantly increased for all groups from 2014 to 2015, whereas the number of medical referrals was not significantly different. Residents actually referred less than both residency and clinical faculty to those services. One reason may be that residency and clinical faculty more often utilize these ancillary services, whereas residents depend more on specialists. Further investigation of this trend is warranted.

There are several limitations to this study. The findings, based on a single academic practice with clinic sites in an underserved geographic location in the Southeast, may not be applicable to all clinic sites or practices. The data analyzed were extracted from electronic health records and billing data, and they may have unintentional errors. Due to the volume of data analyzed, we grouped some types of referrals together, which may not accurately reflect the reasons for referrals. Duplicate referrals were unable to be excluded. These may occur when a physician feels that the previous referral did not get to the intended service or department due to administrative errors or electronic health record errors in transmission. However, this would be an issue for all 3 groups and does not fully explain the increased medical referral rate of the residents. This study does not provide any information on the appropriateness of referrals.

The referral rates among residents, clinical faculty, and residency faculty caring for similar patient populations were found to be significantly different. Further multicenter analysis to determine the reasons for this is needed to ensure that residents are receiving the full scope of training necessary to care for diverse patient populations. As family physicians work to maintain their broad scope of practice, it will be important to ensure that residents are getting the best educational experience possible to allow them to care for medically complex patients in a variety of practice settings.

Department of Community Health and Family Medicine (MP, JM, CM, PB, PJC), and Department of Health Outcomes and Policy (JJS), University of Florida, Gainesville, FL.

Research reported in this publication was partly supported by the National Center for Advancing Translational Sciences of the National Institutes of Health (NIH) under Award Number UL1TR001427. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Concept and design (MP, JM, CM, JJS, PJC); acquisition of data (MP, PB); analysis and interpretation of data (MP, JM, PB, JJS, PJC); drafting of the manuscript (MP, JM, PB, JJS, PJC); critical revision of the manuscript for important intellectual content (MP, JM, CM, JJS); statistical analysis (MP, JJS); provision of study materials or patients (CM); obtaining funding (JJS); administrative, technical, or logistic support (PB); and supervision (PJC).

Maribeth Porter, MD, MS, University of Florida, Department of Community Health and Family Medicine, PO Box 100237, Gainesville, FL 32610. Email: maribethporter@ufl.edu.

1. Donohoe MT. Comparing generalist and specialty care: discrepancies, deficiencies, and excesses. 

. 1998;158(15):1596-1608. doi: 10.1001/archinte.158.15.1596.

2. Nutting PA, Franks P, Clancy CM. Referral and consultation in primary care: do we understand what we’re doing? 

3. Barnett ML, Song Z, Landon BE. Trends in physician referrals in the United States, 1999-2009. 

. 2012;172(2):163-170. doi: 10.1001/archinternmed.2011.722.

4. Franks P, Zwanziger J, Mooney C, Sorbero M. Variations in primary care physician referral rates. 

5. Crump WJ, Massengill P. Outpatient consultations from a family practice residency program: nine years’ experience. 

6. Borowsky SJ, Rubenstein LV, Skootsky SA, Shapiro MF. Referrals by general internists and internal medicine trainees in an academic medicine practice. 

7. Chao J, Galazka S, Stange K, Fedirko T. A prospective review system of nonurgent consultation requests in a family medicine residency practice. 

8. Donohoe MT, Kravitz RL, Wheeler DB, Chandra R, Chen A, Humphries N. Reasons for outpatient referrals from generalists to specialists. 

9. Forrest CB, Nutting PA, Starfield B, von Schrader S. Family physicians’ referral decisions: results from the ASPN referral study. 

10. DerSimonian R, Laird N. Meta-analysis in clinical trials. 

11. O’Donnell CA. Variation in GP referral rates: what can we learn from the literature? 

12. Forrest CB, Nutting PA, von Schrader S, Rohde C, Starfield B. Primary care physician specialty referral decision making: patient, physician, and health care system determinants. 

. 2006;26(1):76-85. doi: 10.1177/0272989X05284110.

13. Armstrong D, Fry J, Armstrong P. Doctors’ perceptions of pressure from patients for referral. 

Referral patterns by family physicians affect numerous aspects of medical care. This study compares the outpatient referral rates of residents, residency faculty, and clinical faculty.

Outpatient Referral Rates in Family Medicine

Given the substantial interest in understanding the drivers of readmissions among patients with chronic illness, our study aimed to identify patient and healthcare system factors associated with 30-day readmissions in a hospital with a Pioneer accountable care organization.

Rates of readmission from 2013 to 2015 were analyzed by attending type, diagnosis, payer, and demographic and clinical factors. Interviews of a matched subset of readmitted and nonreadmitted patients examined patient-reported reasons for readmission.

The readmission rate for the penalty diagnoses of congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), and pneumonia for those managed by geriatricians was lower than the overall readmission rate (11.0% vs 14.1%, respectively). Being divorced was associated with a significantly higher readmission rate (15.7%), whereas Asian/Pacific Islander or “other” race (7.1% and 6.8%, respectively; 

 <.00001) and having private insurance (8.3%; 

 <.00001) were associated with significantly lower readmission rates. Interviews revealed that readmitted patients were more likely than those who were not readmitted to lack an outpatient connection to a primary care provider (10.7% vs 0.0%), to be unable to complete activities of daily living without assistance (32.1% vs 11.1%), and to lack the support of family members during their inpatient stay.

This study found a lower readmission rate for those with CHF, COPD, and pneumonia who were admitted by geriatricians, supporting the notion that enhanced care coordination lowers readmissions. Patient with lower social connectedness were more likely to be readmitted. These findings support the value of rigorous identification of individual risk factors for readmission and of tailoring discharge planning.

The American Journal of Accountable Care. 2018;6(1):e23-e28

In a 2009 analysis of Medicare claims data from all hospital discharges in 2003 and 2004, nearly 1 in 5 patients was readmitted within 30 days of discharge. In monetary terms, this amounts to approximately $17 billion in costs per year.

In an attempt to curb 30-day readmissions and their associated costs, the Hospital Readmissions Reduction Program was implemented in 2012 as part of the Affordable Care Act.

 As a result, CMS began reducing Medicare payments to hospitals whose readmission rates for patients treated for myocardial infarction, congestive heart failure (CHF), and pneumonia were higher than expected for a hospital with a similar patient mix.

 Over the next few years, several admitting diagnoses were added to the list of Medicare “penalty diagnoses”: chronic obstructive pulmonary disease (COPD), total knee replacement, and total hip replacement. This program led healthcare institutions to evaluate and work to better understand the drivers of readmission for patients with these chronic illnesses and acute problems associated with higher rates of readmission.

In this retrospective study of 2013-2015 patient-level data, we sought to contextualize and characterize the 30-day readmission experience of an academic community hospital in Massachusetts. We evaluated factors that were correlated with 30-day readmissions to the medicine service, including the penalty diagnoses of COPD, CHF, and pneumonia; attending physician type; and a number of demographic and patient-specific factors, such as age, gender, race/ethnicity, and marital status.

 As functional status has also been identified as a specific factor that could predict acute care readmission,

 we also collected data on activities of daily living (ADLs). Because previous studies have demonstrated the importance of patient-specific factors on readmission rates

 and the contribution of patients’ experience and postdischarge follow-up to the understanding of quality of care and readmission rates,

It was our goal to examine all patients who were admitted to the medicine service at Mount Auburn Hospital (Cambridge, Massachusetts) and readmitted within 30 days between January 1, 2013, and December 31, 2015. This study was reviewed and approved by the Institutional Review Board of Mount Auburn Hospital. We obtained lists of all patients hospitalized on the medicine service and readmitted within 30 days during this time frame. We analyzed patient outcomes (readmitted or not) by 3 types of variables: demographic characteristics, which included age, sex, race/ethnicity, marital status, and payer type; care characteristics, which included attending type (hospitalist, primary care, or specialist) and discharge disposition (home, home with services, or an acute care facility); and clinical characteristics, which included primary diagnosis and length of stay (LOS).

We performed a univariate analysis, examining each demographic characteristic among readmitted patients, and used the χ

 test to determine which of the factors were significant.

We examined readmitted patients versus nonreadmitted patients by admitting attending type and determined the readmission rate by each type of admitting attending: hospitalist, geriatrician, cardiologist, intensivist, or primary care physician (PCP). We analyzed readmission rates for each category of attending type, stratified by diagnosis. We chose to look at 3 of the Medicare penalty diagnoses (CHF, COPD, and pneumonia) for those years and compared them with all other diagnoses. We then examined the readmission rates for those selected penalty diagnoses in 2013, 2014, and 2015 by 2 attending types, geriatricians and hospitalists, and compared their rates with the overall readmission rates of all attending types. These 2 types of attendings were compared because of the difference in the amount of care coordination and continuity provided by their services; we hypothesized that these differences might be associated with differences in readmissions. Finally, we sought to investigate the readmission rates to other hospitals among this group of patients.

We developed an interview instrument containing both structured and semi-structured qualitative questions and items. The interview substudy focused primarily on several categories of variables that the investigator team believed might be associated with readmission. These included social support, functional status, connection to primary care, connection to relevant specialty care, and experiences with care during and after the index admission. We identified patients from the administrative list who were readmitted within 30 days of an index admission in 2014 with a primary diagnosis of CHF, COPD, or pneumonia. Each patient was contacted by telephone by 1 of 2 interviewers on our study staff (SL or OV) to obtain consent and conduct the interview among those who agreed to participate. Only 2 attempts were made to reach each patient. The interview was conducted only if the patient or a caregiver (designated by the patient) spoke English and was able to understand the consent process. For each patient who agreed to be interviewed, we selected a control patient of the same age and gender who was admitted in 2014 with the same index diagnosis but was not readmitted. If there were multiple potential matches, we chose the first on the list alphabetically who agreed to participate. If there was not a match by age, we chose a patient who was matched in all other ways but the next one chronologically by age (ie, older).

We performed a thematic analysis on the semi-structured items. For the structured items, we performed standard means and compared the means of several variables of the admitted and nonreadmitted patients, including English as a first language, the presence of a PCP in the patient’s care, if the discharge instructions were reviewed, if a family member was involved during the admission, if the patient recalled receiving instructions about who to call post discharge if problems developed, if the patient managed their own medications, if the patient was independent with ADLs, and if the patient felt ready to be discharged at the time of the index admission discharge day.

There were 17,099 admissions to the medicine service from January 1, 2013, through December 31, 2015. As shown in 

, the median age of the patients was 75 years and women accounted for 53.6% of the admissions. With respect to race, we found that 88.3% of those admitted were white, 6.2% were black/African American, 2.4% were Asian or Pacific Islander, and 3.2% identified as “other” race. In terms of marital status, 37.8% were married, 26.1% were widowed, 25.8% were single, and 9.9% were divorced or legally separated. With respect to insurance type, 69.2% of the total admission population had Medicare, including 26.3% who had Medicare fee-for-service and 25.5% who were in our Pioneer accountable care organization (ACO). A subset of Medicare patients were enrolled in Medicare Advantage programs, of which the largest subset was Tufts Medicare Preferred, which included 17.0% of the total admissions. We found that 24.3% of the total admissions had private payers and 6.4% had Medicaid.

Overall, 2226 readmissions occurred within 30 days following an index admission to the medicine service in 2013, 2014, and 2015, for an overall readmission rate of 13.0%. When we examined the demographics of readmitted patients compared with the overall group of admitted patients (

), we found no significant difference in gender between the 2 groups. However, when we examined marital status, we found that being divorced or legally separated (15.7%; 

<.00001) were associated with significantly higher readmission rates compared with being married (11.6%). Those identifying as Asian or “other race” had a significantly lower readmission rate (7.1%; 

 <.000001, respectively) compared with those of other races. Looking at insurance, we found that the readmission rate for patients with private insurance (8.3%) was significantly lower (

<.00001) than that for Medicare patients (14.7%), but this was likely confounded by the lower average age and primarily employer-based insurance of those with private payers. Mean LOS was 3 days for all admissions and 4 days for the readmissions.

The patients were admitted by a diverse group of attending types. Whereas the majority (n = 12,719; 74.3%) of the patients were admitted by hospitalists, 2300 (13.4%) were admitted by PCPs, 1141 (6.6%) were admitted by geriatricians, 707 (4.1%) were admitted by cardiologists, and 232 (1.3%) were admitted by intensivists.

We endeavored to examine readmission rates for each of these attending types for the penalty diagnoses of CHF, COPD, and pneumonia. Therefore, we stratified readmission numbers and rates by attending type by these 3 diagnoses grouped together. We found that the geriatricians had significantly lower readmission rates for the penalty diagnoses of CHF, COPD, and pneumonia (11.0%) compared with other attending types: hospitalists (13.7%), cardiologists (18.5%), intensivists (17.4%), and PCPs (16.7%) (

When we analyzed all readmissions over the 3 years of the study, we saw an increase in readmission rates from 11.5% in 2013 to 13.65% in 2014 and 13.66% in 2015. However, we were surprised to see that the geriatricians’ readmission rate did not follow the general trend. Instead, readmission rates for geriatricians decreased from 16.7% in 2013 to 13.7% in 2014 and to 12.7% in 2015. For hospitalists, the readmission rates increased from 10.8% in 2013 to 13.6% in 2014 and to 13.7% in 2015 (

To examine whether readmissions of our patients to other hospitals might change these findings, we obtained data on readmissions to other hospitals. These data were only available for our Medicare risk-based contracts: Pioneer ACO and a Medicare Advantage program, Tufts Medicare Preferred. We found that although readmissions to other hospitals represented a very small proportion of all readmissions (1.72% for all attending types), geriatricians had lower rates of readmission (0.59%), as shown in 

We interviewed 58 patients with COPD, CHF, or pneumonia, 29 readmitted and 29 nonreadmitted, matched by age, gender, and primary diagnosis of the index admission. The average age was 77 years for readmitted patients and 78 for nonreadmitted, and there were equal numbers of men and women in the groups. We collected data on a diverse array of variables, which we postulated would be related to risk of readmission or explain the cause of readmission from the patient’s perspective. We found that, overall, the readmitted patients were quite ill and had a poor functional status (

). Specifically, they were almost 3 times as likely as those who were not readmitted to indicate that they were not able to complete 1 or more ADLs without assistance (32.1% vs 11.1%, respectively). This was 1 of several ways that readmitted patients differed from nonreadmitted patients; for example, readmitted patients were more than twice as likely to have a first language other than English (17.9% vs 7.1%). They were more likely to lack an outpatient connection to a PCP (10.7% vs 0.0%) and to say they did not recall having their discharge instructions reviewed with them prior to discharge (14.3% vs 0.0%). They were 1.6 times more likely to report that they lacked the support of family members, a significant other, or close friends during their inpatient stay (17.9% vs 10.7%). The readmitted patients were more likely to report that they did not know whom to call with questions or concerns about their health following discharge (10.7% vs 0.0%); they were also more than 3 times as likely as the nonreadmitted patients to report that they did not manage their own medications at home (35.7% vs 10.7%).

However, we did not find substantial differences between the admitted and nonreadmitted patients when it came to educational status or problems with their medications. These medication-related problems included reporting changes to their medications in the hospital, whether they agreed with the mediation changes (or not), reporting difficulty obtaining prescriptions after discharge, or reporting experiencing adverse effects as a result of taking new medications.

This mixed-methods study adds to our understanding of the individual and healthcare system drivers of 30-day readmissions. We found that readmission rates, particularly for the penalty diagnoses of CHF, COPD, and pneumonia, varied significantly by admitting physician attending type. Specifically, we found that geriatricians had lower 30-day readmission rates for the key penalty diagnoses of COPD, CHF, and pneumonia than did other attending types. Small sample sizes did not allow us to achieve statistical significance in this comparison. However, when examining the geriatric service more closely, we noted a continued decrease in readmission rates over time compared with an increase in readmission rates among other services.

To our knowledge, this is the first study to examine the association of attending type with 30-day readmissions. Several elements may have contributed to the lower readmission rates of geriatricians. First, they provide continuity of care to their own patients across the inpatient and outpatient settings. This supports the notion that a comprehensive approach to care during hospital admission and follow-up after discharge reduce the risk of readmissions.

 In addition, they have developed a highly coordinated care model that includes the use of postdischarge home visits, utilizing a multidisciplinary care team familiar with the patient, enhanced use of palliative care, and attention to goals of care. As was previously shown, continuity of care

 may contribute to their lower readmission rates, but the robust infrastructure of the geriatric service probably plays a crucial role as well. All of these special aspects of the geriatrics care model, at least in our hospital, are likely key contributors to their lower readmission rates. Adoption of similar models by other attending types for high-risk patients may be an effective strategy for lowering readmission rates. Furthermore, over the time period of this study, the geriatricians in our hospital focused on strategies for optimizing the discharge process and transitions of care and they hired a nurse practitioner to assist with care of chronically ill patients. They also instituted postdischarge home visits for their most complex patients.

On the individual level, the patient’s social context was also found to be associated with readmission. As in prior studies showing a correlation between community factors, such as access to care, and readmission rates,

 we found that some key social factors increased the risk of readmissions, including lack of social support of a spouse/partner during and after the inpatient stay, lack of a PCP or provider to call with questions, difficulty communicating in English, not remembering discharge instructions, and dependence on others for assistance with medications or ADLs. These findings suggest the need for rigorous identification of individual risk factors for readmission and for tailoring of discharge planning to mitigate the likelihood of readmission for those found to be at high risk.

This study had several limitations. This was a single-center study with a limited number of patients. Therefore, future research is warranted in a larger cohort that involves multiple hospitals. Furthermore, we had data for readmissions to other hospitals only for 2 types of insurers: Medicare Pioneer ACO and Medicare Advantage (Tufts Medicare Preferred). These data included the majority of patients older than 65 years, but we did not have data for other age groups or insurance plans. Finally, the number of successful telephone interviews was lower than ideal, which emphasizes the need to do this type of research either in real time immediately following discharge or on a larger scale with multiple healthcare centers. Finally, it would have been helpful if we had data on LOS by attending types, as this information might explain the lower readmission rates for the geriatricians (ie, if their LOS was significantly higher).

This study adds to the literature regarding effective strategies to decrease 30-day readmissions. However, like many such studies, it consists of a bundled set of interventions studied at a single institution.

 Nonetheless, we believe that we have described another promising strategy for decreasing readmissions, which warrants further study in a larger number of patients and institutions. Taken together, these data suggest that efforts to reduce 30-day readmissions should include comprehensive care teams that focus on the needs of elderly patients with chronic illness and provide continuity in the care team during and following hospitalizations. Furthermore, we show that more attention should be paid to the context of patients’ lives as discharge plans are developed, which may enable identification of those at highest risk of readmission and the provision of enhanced services to meet their needs and mitigate this risk.

Department of Medicine, Mount Auburn Hospital (CCT, TM, VES), Cambridge, MA; Harvard Medical School (CCT, VES), Boston, MA; Sheba Medical Center (NB), Tel Hashomer, Israel; Massachusetts Institute of Technology (KC), Cambridge, MA; FaithCare, Inc (SL), Hartford, CT; University of Rochester School of Medicine (OV), Rochester, NY; CareGroup Parmenter Home Care and Hospice (JK), Watertown, MA.

Presented in part at the Society of General Internal Medicine Annual Meeting (May 2016) and AcademyHealth Annual Meeting (June 2016).

 Dr Kerwin is employed as Director of Quality and Performance Improvement at CareGroup Parmenter Home Care and Hospice. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Concept and design (CCT, NB, TM, SL, OV, VES); acquisition of data (SL, NB, TM, OV, VES); analysis and interpretation of data (CCT, NB, TM, KC, OV, JK, VES); drafting of the manuscript (CCT, NB, KC, JK, VES); critical revision of the manuscript for important intellectual content (CCT, NB, KC,  JK, VES); statistical analysis (NB, TM, KC); provision of study materials or patients (SL); administrative, technical, or logistic support (SL, OV, VES); and supervision (VES).

Valerie E. Stone, MD, MPH, Department of Medicine, South 2, Mount Auburn Hospital, 330 Mount Auburn St, Cambridge, MA 02138. Email: vstone@mah.harvard.edu.

1. Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program [erratum in 

. 2011;364(16):1582. doi: 10.1056/NEJMx110014]. 

. 2009;360(14):1418-1428. doi: 10.1056/NEJMsa0803563.

2. Readmissions Reduction Program (HRRP). CMS website. cms.gov/Medicare/medicare-fee-for-service-payment/acuteinpatientPPS/readmissions-reduction-program.html. Published 2016. Accessed January 21, 2016.

3. Wasfy JH, Zigler CM, Choirat C, Wang Y, Dominici F, Yeh RW. Readmission rates after passage of the Hospital Read&shy;missions Reduction Program: a pre-post analysis. 

. 2017;166(5):324-331. doi: 10.7326/M16-0185.

4. Allaudeen N, Vidyarthi A, Maselli J, Auerbach A. Redefining readmission risk factors for general medicine patients. 

5. Hansen LO, Young RS, Hinami K, Leung A, Williams MV. Interventions to reduce 30-day rehospitalization: a systematic review. 

. 2011;155(8):520-528. doi: 10.7326/0003-4819-155-8-201110180-00008.

6. Goodman DC, Fisher ES, Chang C-H; Dartmouth Atlas Project. The revolving door: a report on U.S. hospital readmissions. The Dartmouth Atlas of Health Care website. dartmouthatlas.org/pages/readmissions2013. Published February 11, 2013. Accessed July 18, 2015.

7. Silow-Carroll S, Edwards JN, Lashbrook A. Reducing hospital readmissions: lessons from top-performing hospitals. The Commonwealth Fund website. commonwealthfund.org/publications/case-studies/2011/apr/reducing-hospital-readmissions. Published April 6, 2011. Accessed July 20, 2015.

8. Hu J, Gonsahn M, Nerenz DR. Socioeconomic status and readmissions: evidence from an urban teaching hospital. 

. 2014;33(5):778-785. doi: 10.1377/hlthaff.2013.0816.

9. Retrum JH, Boggs J, Hersh A, et al. Patient-identified factors related to heart failure readmissions. 

. 2013;6(2):171-177. doi: 10.1161/CIRCOUTCOMES.112.967356.

10. Boulding W, Glickman SW, Manary MP, Schulman KA, Staelin R. Relationship between patient satisfaction with inpatient care and hospital readmission within 30 days. 

This study examined patient clinical and demographic characteristics, healthcare system factors, and patients’ experiences of care associated with 30-day readmissions in a hospital with a Pioneer Accountable Care Organization. 

Thirty-Day Readmissions: Relationship to Physician Attending Type and Social Connectedness

We describe the skills of a physician who practices as an ambula&shy;tory intensivist caring for medically complex patients.

 Provide a detailed referenced review of those skills.

 Succinct summary of the skills and guidance on how one acquires them.

 Crucial elements in the ambulatory intensivist model include the ability to engage patients in order to promote meaningful improvements in health-related behaviors, provide effective leadership for an interdisciplinary team, and interact effectively with a select network of specialists. The ambulatory intensivist develops the ability to collaborate with insurers to mobilize needed social and clinical resources while using predictive modeling to assess concerns, yet still makes use of the more intuitive elements of clinical judgment to minimize clinical risk.&ensp;Advanced clinical skills include in-center diuresis of patients with congestive heart failure, intensive home care that includes extended in-home nursing, and remote monitoring through telehealth devices on an as-needed basis.

The ambulatory intensivist model makes achieving the Triple Aim a reality through improved physician interpersonal, analytic, intuitive, and advanced clinical skills, including the use of telemedicine.

March 2018

Predictive Factors of Discharge Navigation Lag Time

Outpatient Referral Rates in Family Medicine

Thirty-Day Readmissions: Relationship to Physician Attending Type and Social Connectedness

The Skills of the Ambulatory Intensivist: A Review

A Managed Care Organization's Call Center–Based Social Support Role

Chronic Pain as a Driver of Cost in ACO Arrangements

ACO Quality Over Time: The MSSP Experience and Opportunities for System-Wide Improvement

Medicare Accountable Care Spending Patterns: Shifting Expenditures Associated With Savings