Multisite Study to Examine the Amount of Inpatient Physician Continuity Experienced by Hospitalized Patients

Continuity of care is important in ambulatory medicine,^1–3  and for hospitalized patients, continuity has gained notice because of the resident physician duty hour limits. When the Accreditation Council for Graduate Medical Education instituted duty hour limits for residents⁴  in 2003, these rules limited weekly hours to 80 hours. Furthermore, the 2011 revision to the standards placed additional limits on the number of consecutive hours.⁵ 

Residency programs' strategies to adjust to the limits have frequently resulted in reduced continuity.^6,7  Residents, faculty, and program directors have expressed concern about the effect of reduced inpatient continuity of care on the quality of care and resident learning.^8–12  Despite these concerns, the concept of inpatient continuity has remained vague, and in contrast to continuity in the outpatient setting,¹³  measures of continuity of inpatient care have remained elusive. In this study, we aimed to measure the amount of continuity that is experienced by hospitalized patients and to identify factors associated with continuity.

We conducted a multisite prospective study of continuity of care for internal medicine residents using retrospective patient chart review. The study was conducted at 3 sites. The first site is a tertiary care Veterans Affairs hospital. Site 2 is a private teaching hospital and the primary affiliate of the Medical College of Wisconsin's (MCW) internal medicine residency program, and site 3 is a community-based hospital that has its own transitional year interns. The MCW interns and residents also rotated on the general medicine services at site 3 (table 1). At sites 2 and 3, some patients were assigned to hospitalist teams and some were assigned to resident teams. Our report includes only patients admitted to resident teams.

Residents and patients were subjects in this study. We recruited residents rotating on general internal medicine teams between March 2009 and March 2010. Collecting data for 1 year eliminated the possibility of seasonal variation in case mix or continuity. All residents on a team had to agree to participate for the team to be included. Physician participants gave written informed consent. A random number table was used to select a random sample of patients from participating teams. We chose 7 to 8 patients per week at each site, and we obtained a waiver of informed consent for patient chart review. Exclusion criteria for patients were (1) having a fourth-year medical student (subintern) as the primary hospital clinician; (2) being admitted directly to an intensive care unit; (3) being admitted for observation at Froedtert Memorial Lutheran Hospital (this service was organized differently); and (4) being hospitalized for less than 48 hours. Data were recorded in REDCap.¹⁴ 

A research assistant at each site met with each resident to ask them what time they left the day before, what time they arrived that morning, and whether they had left at any other time (eg, for the clinic). This information was recorded in 15-minute increments, and the dates that residents rotated off the participating teams (“switch days”) were abstracted from the available schedule data.

We conducted chart reviews after each discharge. Trained nurses abstracted baseline data by using a standardized guide.^15–17  Nurses collected information on age, sex, race, comorbidities (using the Charlson comorbidity index¹⁸), admitting diagnosis, date/time of admission, date/time of discharge or transfer to another facility, and transfers of care to another service.

Written sign-out documents were downloaded from electronic folders daily in the early morning, assuming that they had not been changed since the night before (the convention for covering teams was not to change the sign-out documents). The sign-out templates varied by site and sometimes by a team within a site.

A physician reviewer or a trained research assistant abstracted the admitting and discharging physicians from the health record. The admitting physician was determined based on who wrote the admitting history and physical document. When more than 1 such document was present, the most junior physician who wrote a history and physical was deemed the admitting physician. The discharging physician was identified as the physician who wrote the final discharge order. The primary inpatient intern was identified from the sign-out document. When this information was missing, the primary intern was determined by who wrote the preponderance of daily progress notes. For patients admitted over team switch days, the new primary intern was determined in the same manner.

This study was approved by the Institutional Review Boards at all 3 hospitals.

We combined the physician schedule data and the patient admission and discharge data into 3 primary variables.

The percentage of hospital time covered by the primary intern is the percentage of the patient's hospitalization time that the intern from the inpatient team provided care, calculated as a ratio of the hours the primary in-hospital intern was in the hospital and the hours of the patient's stay. In the event that a patient's hospitalization extended across a switch day, the newly assigned intern was considered a primary intern, and his or her information was incorporated into the numerator.

The amount of time between the patient's admission and the first handoff of care was determined by calculating the time that elapsed between each patient's admission to the hospital ward and the time that the admitting physician first handed off that patient's care to another physician. Because we did not collect schedule data from the residents on night float or hospitalists who were covering nights, we had some patients admitted by providers whose schedule data were not accessible. These patients were excluded from this analysis.

For admission-discharge continuity, patients who experienced admission-discharge continuity were admitted and discharged by the same physician. In addition to the 3 main continuity variables, we also calculated the number of patients who were hospitalized over a “switch day,” defined as either of their resident team members rotating off the team.

We conducted univariate analyses to look at factors associated with continuity. We looked at patient factors (age, race, sex, comorbidities), length of stay, and site. All analyses were 2-tailed, with P < .05 considered to be significant. Comorbidities were quantified using the Charlson score.¹⁹  This score was converted into a dichotomous variable with a score of 0 versus 1 or more, and age was converted to a dichotomous variable: younger than 65 years versus 65 years and older.

We constructed multivariable models to predict continuity and included the factors that were significant at P < .10 in the univariate analyses.

We included a total of 869 patients in the study (44 patients were excluded because their primary intern had incomplete schedule data, and 78 patients were excluded because they did not have an identified admitting physician). The mean age was 62.6 years (SD = 17.2), and 34% (299 of 869) were women (table 2). Ninety-three percent (204 of 220) of the eligible residents participated in the study. Three residents withdrew early. For the site that contributed the most patients (n = 387), a total of 1905 charts were reviewed. We excluded 294 charts (15%) for length of stay shorter than 48 hours.

The mean percentage of a patient's hospital time that was covered by their primary intern was 39% (SD = 16%). The first handoff of care occurred at a mean of 13 (SD = 7) hours after admission, 350 (40%) patients experienced admission-discharge continuity, and 237 (27%) were hospitalized over a switch day.

In univariate and multivariable analyses, the site was statistically significantly associated with percentage of hospital time covered by the intern, time to first handoff, and admission-discharge continuity (table 3). Some patient demographic factors were associated with the continuity measures in univariate analyses, but most of these associations were nonsignificant in the multivariable models. The multivariable models explained little of the variance (r² = 0.027 for hospital time covered by primary intern; r² = 0.21 for time to first handoff; and Wald score = 61.4 for admission-discharge continuity).

We found that the primary intern was available in the hospital during 39% of a patient's hospital stay, and that fewer than half of patients were admitted and discharged by the same physicians. We also found a large range in the amount of time between admission and the first handoff of care. There was significant variation among the sites for all 3 measures of continuity. Schedules varied considerably across sites, and the association between site and continuity offers some validity evidence that schedules affect continuity.

Many studies²⁰  have used proxy measures for continuity that are not generalizable. For example, 1 study²¹  compared patients admitted by a “short call” team (ie, a team that admits during the day, but goes home in the evening) to those admitted by a “long call” team (ie, a team that admits patients and stays overnight). However, short call as a measure of continuity (or discontinuity) offers limited information. In our study, measuring the time between admission and the first handoff of care allowed us to assess the admitting resident's presence during the intense, early hours of the hospitalization across 3 sites.

Percentage of hospital time covered by the primary intern is a concept that is frequently discussed, but to our knowledge has not previously been measured. This was the most complicated measure to construct, but it provides an important point of reference: primary interns were available for less than 40% of patient stays. This raises the question whether the primary intern is the only one with enough knowledge to provide high-quality care. In a survey study, interns reported that they do not feel the same degree of responsibility for their co-interns' patients as they do for their own, and that they know less about them.²² 

Even after adjustment for other factors, the sites remained significantly associated with all 3 continuity measures. Each site had advantages and disadvantages with respect to maximizing continuity, and we hypothesize that scheduling plays a major role in these measures. For example, while site 2 had a significantly higher percentage of hospital time covered by the primary interns, it had the shortest time between admission and the first handoff of care. No site maximized all measures of continuity, making it important to study the impact of these types of continuity on patients and residents. Programs may need to choose what aspect of continuity is the most important and design approaches to maximize it.

This study has limitations. While it is a multi-institution study, it used residents in a single internal medicine program, with most rotating at all 3 sites. This limits generalizability. We also used self-reports to determine resident hours and did not further validate this measure. Furthermore, collecting data daily from residents was challenging, and we experienced missing data because of that. Third, because the study was conducted after the 2003 duty hour rules went into effect, but before the 2011 rules were implemented, the new 16-hour limit for interns may result in differences in continuity for current patients and residents. Fourth, our definition of admission-discharge continuity is limited because we defined the discharging physician by who wrote the final discharge order. Finally, we excluded patients with stays shorter than 48 hours, and these patients may have experienced more continuity than our sample did.

Future research needs to be done. Programs should study continuity, adapting our approach and using definitions that make sense for their workflow to gain a more robust understanding of continuity of care experienced by hospitalized patients. Future research also should explore the relationship between measures of continuity and outcomes of interest. For example, resident learning and experiential outcomes and patient satisfaction and safety outcomes may be influenced by inpatient continuity of care.

In this multisite study, we present a new approach for measuring continuity of inpatient care and provide baseline data for 3 aspects of continuity for patients hospitalized for more than 48 hours. Our findings showed significant variations by site, which appeared to result from scheduling. While sites may be able to maximize 1 type of continuity, it is unlikely that they could maximize all types, suggesting programs and institutions should focus on the type of continuity deemed most important by them.