Abstract
Background. The benefits of continuity of care (COC) have not been firmly established for pediatric patients.
Objective. To assess whether greater COC is associated with lower emergency department (ED) utilization.
Setting. Outpatient teaching clinic at Children's Hospital and Regional Medical Center, Seattle, WA.
Patients. All 785 Medicaid managed care children ages 0 to 19 years followed at Children's Hospital and Regional Medical Center between 1993 to 1997 who had at least four outpatient visits.
Methods. Retrospective claims-based analysis. COC was quantified based on the number of different care providers in relation to the number of clinic visits.
Results. Attending COC was significantly greater than resident COC. In a multiple event survival analysis, compared with those patients in the lowest tertile of attending COC, those in the middle tertile had 30% lower ED utilization (hazard ratio 0.70 [0.53–0.93]) and those in the highest tertile had 35% lower ED use (hazard ratio 0.65 [0.50–0.80]). Resident COC was not significantly associated with ED use.
Conclusion. Greater COC with attending physicians in outpatient teaching clinics is associated with lower ED utilization.
- COC =
- continuity of care •
- ED =
- emergency department •
- CHRMC =
- Children's Hospital and Regional Medical Center •
- HR =
- hazard ratio
What are the measurable benefits of continuity of care (COC)? Several studies of adults suggest that having a regular and consistent source of care increases patient and provider satisfaction, decreases hospitalization, and lowers costs.1–7 For children, however, data on the effect of COC are sparse and contradictory. Although absent or inconsistent Medicaid coverage has been shown to decrease use of outpatient services for low-income children,8 ,9 there are conflicting reports of how or even whether consistent relationships with a pediatrician can affect children's health status and utilization patterns.10–14
Despite the uncertain value of COC for children, managed care organizations nevertheless have emphasized its role for patients and their families. Continuity of primary care physicians, it is argued, will decrease inappropriate use of expensive health services (eg, emergency department [ED] visits), enhance preventive care, and ideally keep patients from becoming so sick as to require hospitalization.15 ,16 The extent to which any of these relationships exist is poorly understood.
To clarify one aspect of this puzzle, we attempted to assess whether greater COC was associated with lower ED use among Medicaid children who were followed at an outpatient teaching clinic.
METHODS
Setting and Providers
The 7H Clinic, named for its location within the Children's Hospital and Regional Medical Center (CHRMC), is a traditional outpatient resident teaching clinic. Twenty-one trainees spend one-half day per week throughout their residency in the 7H Clinic during which time they follow their own panel of patients for both scheduled and acute visits. The appointment system is structured to enable residents to see their own patients whenever possible. In addition, every second-year resident spends a 2-month block on an outpatient medicine rotation during which they see acutely ill outpatients in the clinic.
Attending physicians on faculty in the Division of General Pediatrics at the University of Washington precept residents in the clinic. They too are typically present for a specified afternoon and are assigned to work with the same set of residents throughout the year. Residents see and evaluate their patients independently and then present them to the attending physician of the day. In many cases, be it for teaching purposes or to clarify an uncertain diagnosis, attending physicians examine a patient after the resident has done so. Treatment decisions for patients are made together and the residents then present the plan to their patients. Although there is no formal faculty practice, on rare occasions, patients are seen only by an attending physician.
Patients
The patients included in this study were on Medicaid, covered by Aid to Families with Dependent Children. Our sample was restricted to those patients assigned to CHRMC as their primary source of care between September 1993 and September 1997. Families in Medicaid managed care in the state of Washington are allowed to change sites of care through a formal process as frequently as monthly although few patients who have chosen or have been assigned to CHRMC have historically transferred care elsewhere.
The study protocol was reviewed and approved by the University of Washington's Institutional Review Board.
Outcomes
Our primary outcome was ED visitation. Because of managed care, all ED visits for managed care Medicaid patients require primary care provider approval. Accordingly, if a patient visits an ED, the physician-on-call for the plan is notified, and the visit is either approved or denied. Both approvals and denials are recorded and kept in a database at CHRMC. Because physicians have different thresholds for approving or disallowing visits to an ED, we used all ED visits as our outcome regardless of whether or not the visit was approved. This is appropriate because approved visits would be a nonuniform measure of illness acuity.
Continuity Measure
Our primary predictor variable was COC. Several indices for measuring COC have been developed.17–20 Crude measures simply quantify the number of providers or the proportion of visits to a given provider and are less useful when visits are few or multiple providers are used. We opted to use a more sophisticated measure that models the dispersion in patient-provider contacts and establishes a concentration index. The COC index developed by Bice and Boxerman18 is of the general form:
where n = total number of visits
nj = number of visits to provider j
s = number of providers
The COC takes on values between zero and one. A value of zero signifies maximum dispersion that occurs when a different provider is seen for every visit. A value of one signifies minimum dispersion, which occurs when the same provider is seen at every visit. To demonstrate the behavior of the COC, several hypothetical patterns each involving eight visits are shown in Table 1. Note that as the contacts with providers become more dispersed—from all visits with provider A to every visit with a different provider—the COC moves from one to zero.
Example of the Continuity of Care Index (COC)
In a study of elderly veterans, an increased COC was associated with decreased hospitalization rates and shorter lengths of stay.7 Because the COC index becomes more robust—that is, less subject to significant change as a result of minor perturbations in care dispersion as the number of visits increases—we restricted our analysis to patients with four or more visits. This decision was in keeping with the construct underlying continuity, because meaningful COC cannot exist when visits are exceedingly few.
Because both residents and attendings were involved in the care of patients, COC indices were computed separately for both types of providers.
Covariates
In addition to age and sex, two other covariates were included as possible confounders: presence of underlying asthma and total outpatient (non-ED) visits. Asthma was included because it is the most prevalent chronic disease of childhood and is associated with increased ED use. Children with asthma are identified in our clinic by means of an asthma registry that contains all patients with an inpatient or outpatient diagnosis of asthma based on ICD-9 codes. We included total outpatient visits as an additional covariate because children with greater illness burden would be more likely both to visit their physician and to visit the ED.
Although we did not formally control for socioeconomic status, all the included patients were on Medicaid.
Statistical Analysis
Paired sample t tests were used to compare resident and attending continuity indices. Multiple event survival analysis was used for the statistical modeling of ED visit frequency.21Survival analysis offered two important advantages. First, it controlled explicitly for the amount of time at risk which was necessary because patients had different enrollment periods. Second, it ensured that the COC index used as a predictor in our model preceded an ED visit.
The COC index changes with each contact a patient has with a provider, because their care either becomes more or less dispersed across providers. We therefore modeled COC as a time-dependent variable and updated it with each visit a patient made.22 Robust estimation of variance was used because of nonindependence of the repeated observations on the same patients.21 Attending and resident COC indices for study patients were divided into tertiles based on their distribution in our sample.
All analyses were conducted with Stata 5.0 for the Macintosh computer.21
RESULTS
During the 4-year study period, there were 785 eligible patients with an average of 10 outpatient visits per patient. Demographic data are summarized in Table 2. A slight majority (54%) of our patients were male; 15% had asthma; 30% had at least one visit to the ED. Ten attendings and 97 residents cared for patients in the 7H Clinic during the study period.
Summary of Patients Included in the Analysis
The COC indices for both residents and attendings are displayed in Fig 1 and Fig 2. There was a significant difference in mean resident COC and mean attending continuity (P < .01). Notably, 211 patients had resident COC indices of zero meaning that they saw a different resident at each of their visits. In contrast, 111 patients had attending COC indices of zero. Ninety-eight patients had perfect resident continuity (COC = 1) whereas 117 had perfect attending continuity.
Histogram of resident continuity of care (COC) indices.
Histogram of attending continuity of care (COC) indices.
In the unadjusted survival analysis, compared with low resident continuity, patients with medium resident continuity were no less likely to visit the ED during the study period (hazard ratio [HR] 0.93 [0.65–1.3]); patients with high resident continuity were significantly less likely to use the ED (HR 0.88 [0.51–0.99]) (Table 3). In contrast, when compared with low attending continuity, medium and high attending continuity were both associated with decreased ER use (HR 0.75 [0.51–0.99]) and (HR 0.70 [0.43–1.0]), respectively.
Multiple Event Survival Analysis of ED Use
In the fully-adjusted multiple event survival analysis, asthma was associated with increased ED use (HR 1.30 [1.0–1.72]) as was total outpatient visits (HR 1.04 [1.02–1.06]). Increasing age was associated with decreased ED use (HR 0.94 [0.91–0.97]). Greater resident COC was not significantly associated with decreased ED use whereas medium attending continuity was (HR 0.70 [0.53–0.93]) and high attending continuity was even more so (HR 0.65 [0.50–0.80]) (Table 3).
DISCUSSION
Our study of COC for children in an outpatient teaching clinic revealed three major findings: that attending continuity was greater than resident continuity; that attending continuity mattered more than resident continuity as a predictor of ED use among these children; and that the association between greater attending continuity and less use of the ED seemed to follow a dose-response relationship, with high continuity being more associated with low ED use than medium continuity.
Of these three findings, the first—that attending COC was greater than resident continuity—is not surprising. The fact that residents work only one-half day per week and also are turning over at a rate of one third per year, makes resident COC difficult to achieve and maintain as others have noted.23–25 We do not know how the level of either attending or resident continuity documented in our study compares to the continuity provided for patients in other nonteaching settings, although one would suspect continuity in the teaching setting to be worse.26
The second finding raises the question: why might attending continuity seem to matter more than resident continuity? In part, this result may be a statistical artifact that might seem to devalue the potential benefits of resident continuity. Because 97 residents are involved in the care of the 7H Clinic patients, COC for most patients is likely to be quite low. Detecting a true significant statistical signal amid so much background noise may have been difficult.
Another possible explanation of the relative importance of attending continuity, as well as of the dose-response relationship between increased attending continuity and decreased ED use, arises from the structure of the clinic. Although residents in teaching clinics may not develop continuous relationships with patients, attendings who hear about and often see the patients presenting to the clinic throughout months and years, may come to know certain patients and their families quite well. This knowledge may breed special insights into families—their medical problems, past experiences, social situations—and these insights may in turn affect what therapies are instituted (eg, initiation of chronic inhalation therapy for asthma) or what services are made available (eg, social work, public health nurse). Such insights might be especially important when serving patients with significant medical and social problems, such as those in our study. This increased, long-term knowledge of patients, coupled with the attending's more advanced medical expertise, may improve the quality of the care implemented based on their recommendation, and ultimately reduce the need for ED use.
An alternative explanation of why patients with greater COC might be less likely to use the ED is that these patients have less severe illnesses. One can imagine that healthier children may have fewer drop-in acute visits, more scheduled visits, and thus greater COC with a given provider working on a particular day of the week. Healthiness would then confound the relationship between continuity and ED use. Despite this possibility, however, we intentionally did not control separately for the number of acute outpatient visits. Although we did control for age (thereby accounting for the numerous routine well-child visits of early life), any further adjustment would have—on clinical grounds—been unwarranted, because the actual care provided during well and acute visits frequently blends together. On the one hand, preventive care (eg, immunizations, inhaler medications) and health maintenance recommendations (eg, regarding safety or anticipatory guidance) are often made during minor acute office visits even when they are not related to the reason for the visit. On the other hand, children with chronic illnesses such as asthma or diabetes may be better served by more frequent well-visits with their providers so that treatment of their potentially acute problems is proactive and not purely reactive. In such cases of chronic illness, increasing the number of nonsick visits may be the mechanism by which COC exerts is salubrious effect, thereby making controlling for them inappropriate.
A final explanation to the associations between either resident or attending continuity and ED use arise from yet another question: to what degree might the conscientiousness of patients and their families in working toward the goal of achieving a highly-continuous relationship with a provider effect the findings of this study? Patients who are willing and able to establish high continuity relationships with a provider—in essence, adhere to the concept of COC—may also be more likely to adhere with medication regimens, to make organized use of office hours clinic visits, and to use the ED less. Although we could not control for this possible source of confounding, previous randomized controlled trials that would have equally distributed this variable suggest that increased continuity is in fact beneficial.7 ,27
Given that our study was conducted in a single outpatient teaching clinic, they must be conservatively generalized. Certainly, many other teaching clinics function in the same way and care for similar patients; nevertheless, the applicability of our results to nonteaching settings is unknown. Additionally, our adjustment for illness severity was limited to the presence of asthma and to the total number of outpatient visits, neither of which may be adequate proxies for actual severity.
Despite these limitations, this study does raise further interesting questions. The ability to quantify COC from existing databases invites comparisons among health care delivery systems. The extent to which managed care plans actually succeed in improving COC seems a worthwhile area of inquiry. Moreover, because continuous relationships between providers and patients have been deemed critical to medical training,28 ,29 the COC may provide a useful tool for bench-marking residency program performance as well as individual resident's outpatient experiences. Lastly, the important, underexplored phenomenon of how COC relationships are established and maintained, and the extent to which health care systems can improve them, begs further study.
In conclusion, our study shows that patients with poor attending continuity can be identified from existing administrative databases, and that they are at increased risk of ED utilization. Efforts to enhance COC in resident teaching clinics seems warranted.
ACKNOWLEDGMENTS
We thank Chris Feudtner, MD, PhD; Robert Davis MD, MPH; Richard Deyo, MD, MPH; and Frederick Rivara, MD, MPH for their review of this manuscript.
Dimitri Christakis was a Robert Wood Johnson Clinical Scholar while this research was conducted.
Footnotes
- Received May 11, 1998.
- Accepted August 31, 1998.
Reprint requests to (D.A.C.) Division of General Pediatrics, Child Health Institute, 146 N Canal St, Suite 300, Seattle, WA 98103.
Drs Christakis and Connell are also affiliated with the Child Health Institute, University of Washington, Seattle, Washington.
The opinions expressed here are those of the authors and are not necessarily those of the Robert Wood Johnson Foundation.
REFERENCES
- Copyright © 1999 American Academy of Pediatrics