Development of a handoff continuity score to improve pediatric ICU physician schedule design for enhanced physician and patient continuity
© Smalley et al.; licensee BioMed Central Ltd. 2011
Received: 1 June 2011
Accepted: 21 October 2011
Published: 21 October 2011
Few studies investigate the benefits of familiarity or continuity during physician-to-physician handoff of inpatients. Factors such as how recently physicians (MDs) have worked and successive days caring for patients increase continuity, and thus could lead to enhanced handoff efficiency. Evaluating the efficacy of MD scheduling to enhance continuity is currently subjective.
An MD group consisting of 9 attending physicians and 7 fellows redesigned its pediatric intensive care unit (PICU) coverage schedule with the goal of enhancing continuity of care. The attending PICU MDs were formally surveyed to rate the impact of the schedule change on continuity and efficiency (5 point Likert scale: 1 = worse, 3 = no change, 5 = better). A Handoff Continuity Score (HCS) was developed and used to analyze the 30-bed PICU MD schedule for continuity and handoff efficiency. MD service and call schedules were evaluated for 6-month periods before and after the schedule redesign. The HCS for each schedule was calculated by considering every shift change, or handoff, in the scheduling horizon, and assigning scores to oncoming physicians based on previous days worked. Specifically, for each handoff, each oncoming MD receives a score between 0 and 1, calculated as the summation of a series of 'familiarity factors', one for each recent day worked. The scores for all oncoming MDs are averaged to determine the score for that specific handoff, and the HCS is the average of all handoff scores. The HCS was incorporated into an integer programming (IP) model for scheduling MDs to maximize continuity. A z-test was used to assess the significance of improvement in the HCS.
The HCS before and after redesign was 0.57 and 0.68, respectively (19% increase, p < 0.01). Mean MD rating was 4.22 ± 0.56 for continuity, and 4.00 ± 0.65 for efficiency. With the goal of further improving the HCS and (partly) automating and streamlining the scheduling process, the IP was developed to populate physician service and night-call schedules while conforming to scheduling constraints; IP-generated schedules improved the HCS to 0.79 (39% increase).
The increased HCS was associated with the MD qualitative assessment of enhanced continuity and efficiency after implanting a schedule change. The IP identified the potential for additional scheduling improvements.
Continuity of care improves patient satisfaction , whereas the fragmentation of care may have a negative impact on patient outcomes. Epstein and colleagues  found a statistically significant relationship between fragmentation of patient care and length of stay (LOS) among patients admitted for pneumonia or heart failure. As staffing models for inpatient hospital settings transition to shift-work because of increased duty hour restrictions and life-style choices by physicians (MDs), serious concerns arise regarding the resulting lack of continuity for patients and MDs and potential impacts on quality of care and education of residents .
Fragmented care leads to increased handoffs, or the transfer of patient care from one MD to another. Communication between MDs during handoff of inpatients is a 'vital link in the continuity of patient care'  but may be insufficient to relay all pertinent information, and 'communication failures have become widely recognized as a leading safety hazard in health care' . On the basis of a review of controlled studies of interventions, Arora and colleagues  provide preliminary recommendations for improving handoffs, but further studies are needed to identify best practices [7, 8].
It is intuitive that an MD's familiarity with a patient prior to starting work has the potential to make the handoff process more efficient. Few studies investigate the benefits of familiarity or continuity during MD-to-MD handoff of inpatients, and improving scheduling models to enhance continuity and familiarity could provide a tool to help with handoff efficiency and quality. Evaluating the efficacy of MD scheduling to provide continuity is currently subjective.
We propose, as one method for improving patient and MD continuity, a Handoff Continuity Score (HCS) as a novel approach for objectively assessing continuity and familiarity with various scheduling models. We use the HCS to analyze MD duty schedules before and after the implementation of a schedule structure redesign intended to maximize patient and MD familiarity and continuity within the context of defined scheduling constraints. We incorporate the HCS into a mathematical model developed to automate and improve the process of MD duty schedule generation. The purposes of this study are (a) to develop the HCS as an objective measure of MD and patient continuity, (b) to validate the HCS and its correlation to MD perceptions of continuity and familiarity within a small group of MDs, (c) to use the HCS to measure improvement in continuity following an MD duty schedule redesign, and (d) to assess the application of a mathematical model for constructing MD duty schedules that optimize the HCS.
Materials and methods
An MD group, consisting of nine pediatric critical care medicine (PCCM) attending MDs and seven PCCM fellows, redesigned its pediatric intensive care unit (PICU) MD coverage schedule to enhance continuity. The PICU is a 30-bed multidisciplinary medical-surgical quaternary care unit and provides care for acutely ill patients, including those receiving liver transplants, continuous renal replacement therapies, or extracorporeal life support. The weekly service schedule for the unit consists of two teams (A and B), each including one attending MD and one fellow providing daytime coverage. The teams treat two separate groups of patients, but MDs from team A are somewhat familiar with the patients treated by team B, and vice versa, because of proximity, collegiality, early-morning fellow sign-outs (informal), and formal sign-outs that include both teams and the oncoming night call MDs. Night call shifts are covered by one attending MD and one fellow.
Physician schedule redesign of weekly night-call schedule
Attending on call
MD from team A
MD from team B
Fellow on call
Night float fellow
MD from team A
MD from team B
Night float fellow
Night float fellow
MD from team A
MD from team B
The institutional review board of the Children's Healthcare of Atlanta approved this study and waived the need for informed consent as this was a quality initiative that did not involve individual patients or interventions. We developed a scoring method, based on the following assumption, to effectively capture the continuity at handoff.
The level of familiarity that an MD has with their patients (a) increases with successive (or multiple) on-duty days in which the MD sees the patients and (b) decreases as the number of off-duty days increases (for example, to more than the average LOS of patients). Intuitively, an MD who sees a patient for a number of consecutive days will be more familiar with and informed about the patient's condition. Conversely, an MD who sees a patient after a multi-day break would need some time to become refamiliarized with the patient's progress and current state.
Notation of variables in continuity score
Patient length of stay (5 days in our analysis)
A familiarity factor dependent on the length of time since last shift
1 if the physician worked between i-1 and i days ago, 0 otherwise
Number of days since previous shift
1 to < 2
2 to < 3
3 to < 4
4 to < 5
Each oncoming MD could receive a score of between 0 and 1 (1 = most familiar). The score for the handoff is calculated as the average continuity score over all oncoming MDs. For a complete MD schedule, the HCS is calculated as the average of all handoff scores and ranges from 0 to 1.
Sample 1-week schedule of attending physicians and fellows
8 a.m. to 4 p.m.
4 p.m. to 8 a.m.
Sample calculation of Handoff Continuity Score
Number of days since previous shift
1 to < 2
2 to < 3
3 to < 4
4 to < 5
Sum (continuity score)
The handoff score is the average of the physician continuity scores →
(0.15+1)/2 = 0.575
To compare the HCS before and after the schedule redesign, we evaluated MD service and night call schedules for 6-month periods. A z-test was used to assess significance of improvement in the HCS. MDs were asked to rate the impact of the schedule change on continuity and efficiency (see Additional file 1 for survey questions).
To (partially) automate and improve the process of schedule generation, we developed an integer program (IP), which is a mathematical model (a system of equations and inequalities) consisting of an objective function and constraints . The constraints ensure that the requirements of the Accreditation Council for Graduate Medical Education (ACGME)  and staffing requirements for the unit (for example, two attending MDs and two fellows on duty for each daytime shift and one attending and one fellow on duty for each night call shift) are not violated. The IP also incorporates soft constraints to accommodate MD requests for time off and the preferred service schedule structure. The soft constraints may be violated only if it is necessary to construct a feasible schedule (and not for the purposes of improving the HCS). The IP was used to construct an MD schedule which maximized the HCS while minimizing violations of the soft constraints. This IP-generated schedule HCS was compared to that achieved by the pre- and post-redesign schedules." For additional details on the IP, see Smalley and colleagues .
While the current service schedule structure for the PICU (7-day blocks for attending MDs and 14-day blocks for fellows) was designed with continuity in mind, a better structure with respect to patient and MD continuity may be possible. Using different versions of the IP, we constructed schedules that follow four alternatives to the current service schedule structure and computed the corresponding HCS for each. The five total service schedule structures that we investigated are (1) 7-day blocks for attending MDs and overlapping 14-day blocks for fellows (current schedule structure), (2) overlapping 7-day blocks for attending MDs (in which a new attending MD starts a 7-day service block every Monday and Friday) and overlapping 14-day blocks for fellows, (3) overlapping 4-day blocks for attending MDs (in which a new attending MD starts a 4-day service block every two days) and overlapping 14-day blocks for fellows, (4) overlapping 14-day blocks for fellows but no predefined service block structure for attending MDs, and (5) no predefined service block structure for attending MDs or fellows. Structure 1 is the current service schedule structure in the PICU. Structures 2 and 3 correspond to service schedule structures that MDs in the PICU proposed as potential improvements. We include structures 4 and 5 to determine the best achievable HCS when there is no predefined service block structure (that is, when there is no requirement that MDs be on service for a certain number of multiple consecutive days).
Physician survey results
Physician rating for impact on continuity and handoff efficiency because of schedule redesign
4.22 ± 0.56
4.00 ± 0.65
Handoff Continuity Score resultsa
HCS of all shifts
19% (P = 0.0011)
39% (P < 0.001)
HCS of service shifts only
19% (P = 0.0244)
22% (P = 0.0104)
HCS of night call shifts only
22% (P = 0.008)
60% (P < 0.001)
HCS of attending physicians only
29% (P < 0.001)
62% (P < 0.001)
HCS of fellows only
13% (P < 0.001)
19% (P < 0.001)
Night call shifts without an on-service physician (n = 175)
Handoff Continuity Scores for alternative service schedules generated by the integer program
Service schedule structure and description
HCS of attending physicians only
HCS of fellows only
Night call shifts without an on-service physician
7-day blocks for attending physicians and overlapping 14-day blocks for fellows
Overlapping 7-day blocks for attending physicians (Mon.-Sun. or Fri.-Thurs.) and overlapping 14-day blocks for fellows
Overlapping 4-day blocks for attending physicians and overlapping 14-day blocks for fellows
Overlapping 14-day blocks for fellows but no predefined service block structure for attending physicians
No predefined service block structure for attending physicians or fellows
On the basis of survey results, the attending MDs found that continuity and handoff efficiency were improved by redesigning the on-service and night call shift schedule structure. This was positively correlated with the HCS numerical estimate of familiarity and continuity.
There was a statistically significant increase in the HCS after the schedule redesign. When viewing the HCS by time of day, we see that the HCS for night call shifts is considerably less than the HCS for service day shifts for both pre- and post-redesign schedules, an expected result given the service schedule structures in both. The greatest increase (29%) in the post-redesign HCS is associated with the schedule for attending MDs. The HCS for fellows increased by only 13%. The fellows' schedule is structured to provide concentrated clinical time (a key component for the HCS) in order to provide protected research time during their training. Increasing from 7-day service blocks, which exceeds the PICU LOS (5 days), to 14-day service blocks resulted in only a modest improvement to the HCS. In addition, ACGME duty hour restrictions provide a rigorous hard constraint that may have limited the improvement in the HCS.
While the schedule redesign improved the HCS, the IP identified the potential for additional scheduling improvements. However, the best HCS was achieved by a schedule that required attending MDs who are on service during a given week to work alternating call shifts that week, without any other attending MDs scheduled. Note the resulting much-improved HCS for night call shifts for the IP-generated schedule over the pre- and post-redesign schedules. There is only a modest improvement in the HCS for service day shifts over the post-redesign schedule because this score is determined largely by the service block structure implemented post-redesign.
The post-redesign schedule ensures that at least six night call shifts per week have an on-service MD (attending MD or fellow) scheduled. Table 7 indicates that, post-redesign, only 8 night call shifts in the 6-month schedule had no on-service MD scheduled, a decrease of 85% from the 52 pre-redesign. The overall HCS increased by 19% post-redesign. The IP-generated schedule did not greatly reduce (in comparison with the post-redesign schedule) the number of night call shifts without an on-service MD (6 versus 8), but the overall HCS increased by 16% over the post-redesign schedule. Thus, there is not a linear relationship between the HCS and the number of night call shifts without an on-service MD. Other factors impact the HCS such as daytime continuity and familiarity among off-service MDs assigned to night call shifts.
While the IP-generated schedule maximizes continuity, its implementation may lead to fatigue and negatively impact quality of care. In cases like this, adjustments can be made in two ways. First, additional constraints can be added to the IP to prevent unfavorable scheduling situations and to address MD preferences (for example, to set a limit on the number of night call shifts per week) to generate a schedule that maximizes continuity without relying on manual generation of schedules. The alternative is that the IP-generated schedule can be manually manipulated and the HCS recalculated to determine the impact of changes on continuity. Given that MD schedules often are made well in advance, changes are needed from time to time to adjust for new requests from MDs and other unavoidable situations. The HCS provides some objective assessment of the impact of such changes on continuity of patient care as MDs evaluate alternatives.
The mathematical model we developed is not the first for scheduling MDs in a hospital setting. Sherali and colleagues  addressed the problem of scheduling medical residents to night shifts over the course of 4 to 5 weeks by using a mixed-integer program and heuristics that exploit the network structure of the problem. Topaloglu [13, 14] used goal programming to assign emergency medicine residents to day and night shifts over a 1-month planning horizon and applied sequential and weighted methods for solving an optimization model with multiple objectives to the problem of assigning residents to on-call shifts based on levels of seniority. Ovchinnikov and Milner  developed a user-friendly spreadsheet model for assigning first- to fourth-year residents to on-call and emergency rotation shifts in a radiology department over a 1-year period, and Cohn and colleagues  addressed the problem of assigning 10 to 20 residents to five types of on-call shifts in three distinct hospitals over a 1-year period by solving multiple nested IP formulations to address the different metrics. Our IP has some similarities with previous work but, to the best of our knowledge, is the first model designed to address the impact of MD scheduling on MD and patient continuity.
The HCS gives consideration to various schedule structures to be tested with an objective quantitative assessment of the impact on continuity. In comparing the scores reported for the various service schedule structures applied to the PICU, we see that the greatest HCS can be achieved by having no predefined service block structure for attending MDs or fellows. The version of the IP used for this schedule structure has fewer constraints than others (that is, no predefined service block length for attending MDs or fellows or both), and therefore we would expect this alternative structure to achieve the highest HCS. Removing predefined service block length constraints, the IP can generate an optimal MD duty schedule with respect to continuity. However, in this case, the schedule generated with this structure requires that attending MDs and fellows be on service for several weeks to a month at a time and may not adhere to MD preferences or be feasible. The number of call shifts without an on-service MD is minimized by structure 3, but the corresponding HCS is also the lowest of the service schedule structures that we tested. The HCS and number of night call shifts without an on-service MD are useful assessments that MD groups can use to tailor a service and call schedule to the needs of their unit or MDs or both. The current service schedule structure (structure 1 in Table 8) is preferred among MDs in the study PICU because, among candidate schedules with the 7-day preferred service structure for attending MDs, it has the fewest night call shifts without an on-service MD associated with the highest HCS.
Familiarity among oncoming MDs has the potential to improve the handoff process itself. However, familiarity alone does not reduce the need for careful discussion during the handoff of a patient's care from one MD to another. Owing to the dynamic nature of illnesses, a patient's condition evolves over time. New information may become available and the course of treatment may change from one day to the next. Furthermore, factors such as new admissions, discharges, bed occupancy, and MD fatigue impact handoff efficiency. Future research opportunities include developing a fatigue factor for assessing the impact of scheduling changes on continuity and fatigue as it impacts the handoff process.
We did not take into account Hawthorne effects among the nonblinded participating MDs. As the first application of this approach, the analysis above relies on the perceptions of an MD group within one PICU. Furthermore, only attending MDs were surveyed and therefore we can make no assumptions regarding the perceived improvement (if any) felt by fellows in this MD group following the schedule redesign. A larger sample of MDs is needed to truly validate the HCS. The potential far-reaching implications of the HCS, particularly as it might impact patient outcomes, has not been studied and is beyond the scope of this article.
To maintain good quality of care, it is important that MDs be familiar with their current patients; familiarity contributes to MD and patient continuity. MD familiarity and continuity are tools that can be used to help improve efficiency and reduce miscommunications that can occur during the handoff process. The proposed HCS is a novel approach for objectively assessing continuity and familiarity given our intuitive assumptions. The results of the MD survey reveal that the HCS is associated with perceived improvement in continuity and handoff efficiency from the schedule redesign. With a mathematical model, an MD schedule can be optimized with respect to the HCS with the goal of maximizing continuity. The IP and HCS together facilitate resource optimization, particularly in the face of increased duty hour constraints. While this initial study assessing an objective scoring method for measuring MD and patient continuity is limited to a small group of MDs, implications may be far-reaching, and further study is needed to assess the applicability of such an approach to other units and institutions.
A proposed Handoff Continuity Score correlates with physicians' perceptions of enhanced continuity and patient familiarity.
An objective scoring method can be used to analyze a physician duty schedule for physician and patient continuity.
With a mathematical model, a physician schedule can be optimized with respect to continuity.
Accreditation Council for Graduate Medical Education
Handoff Continuity Score
length of stay
pediatric critical care medicine
pediatric intensive care unit.
The study group collaboration was funded by a seed grant from the Clinical Outcomes Research and Public Health Pilot Grant Program, jointly sponsored by Georgia Tech Health Systems Institute and Children's Healthcare of Atlanta. This research is also supported in part by the Mary Anne and Harold R. Nash Endowment and the Smalley Endowment at Georgia Tech. The authors are grateful to two anonymous reviewers, whose comments and suggestions greatly improved the manuscript.
- Saultz JW, Albedaiwi W: Interpersonal continuity of care and patient satisfaction: a critical review. Ann Fam Med 2004, 2: 445-451. 10.1370/afm.91PubMed CentralView ArticlePubMedGoogle Scholar
- Epstein K, Juarez E, Epstein A, Loya K, Singer A: The impact of fragmentation of hospitalist care on length of stay. J Hosp Med 2010, 5: 335-338. 10.1002/jhm.675View ArticlePubMedGoogle Scholar
- Fortuna RJ, Palfrey JS, Shelov SP, Samuels RC: Paediatric experiences with work-hour limitations. J Eval Clin Pract 2009, 15: 116-120. 10.1111/j.1365-2753.2008.00964.xView ArticlePubMedGoogle Scholar
- Solet DJ, Norvell JM, Rutan GH, Frankel RM: Lost in translation: challenges and opportunities in physician-top-physician communication during patient handoffs. Acad Med 2005, 80: 1094-1099. 10.1097/00001888-200512000-00005View ArticlePubMedGoogle Scholar
- Landrigan CP: The handoff: a critical point of vulnerability. CRICO/RMF Forum 2007, 25: 8-9.Google Scholar
- Arora VM, Manjarrez E, Dressler DD, Basaviah P, Halasyamani L, Kripalani S: Hospitalist handoffs: a systematic review and task force recommendations. J Hosp Med 2009, 4: 433-440. 10.1002/jhm.573PubMed CentralView ArticlePubMedGoogle Scholar
- Cohen MD, Hilligoss PB: The published literature on handoffs in hospitals: deficiencies identified in an extensive review. Qual Saf Health Care 2010, 19: 493-497. 10.1136/qshc.2009.033480PubMedGoogle Scholar
- Riesenberg LA, Leitzsch J, Massucci JL, Jaeger J, Rosenfeld JC, Patow C, Padmore JS, Karpovich KP: Residents' and attending physicians' handoffs: a systematic review of the literature. Acad Med 2009, 84: 1775-1787. 10.1097/ACM.0b013e3181bf51a6View ArticlePubMedGoogle Scholar
- Nemhauser GL, Wolsey LA: Integer and Combinatorial Optimization. New York: John Wiley & Sons, Inc.; 1999.Google Scholar
- Accreditation Council for Graduate Medical Education: Common program requirements.[http://www.acgme.org/acWebsite/home/Common_Program_Requirements_07012011.pdf]
- Smalley HK, Keskinocak P, Vats A: Physician scheduling for continuity: an application in pediatric intensive care.[http://www.isye.gatech.edu/~hkolberg3/MSOMHCS_WorkingPaper.pdf]
- Sherali HD, Ramahi MH, Saifee QJ: Hospital resident scheduling problem. Production Planning and Control 2002, 13: 220-233. 10.1080/09537280110069667View ArticleGoogle Scholar
- Topaloglu S: A multi-objective programming model for scheduling emergency medicine residents. Computers and Industrial Engineering 2006, 51: 375-388. 10.1016/j.cie.2006.08.003View ArticleGoogle Scholar
- Topaloglu S: A shift scheduling model for employees with different seniority levels and an application in healthcare. Eur J Oper Res 2009, 193: 943-957.View ArticleGoogle Scholar
- Ovchinnikov A, Milner J: Spreadsheet model helps to assign medical residents at the University of Vermont's College of Medicine. Interfaces 2008, 38: 311-323. 10.1287/inte.1070.0337View ArticleGoogle Scholar
- Cohn A, Root S, Kymissis C, Esses J, Westmoreland N: Scheduling medical residents at Boston University School of Medicine. Interfaces 2009, 39: 186-195. 10.1287/inte.1080.0369View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.