PCP indicates primary care physician. P < .001 for both Cochran-Armitage trend tests for continuity with any outpatient physician or with a PCP. For all point estimates, the 95% confidence intervals are less than 0.6%.
Gulshan Sharma, Kathlyn E. Fletcher, Dong Zhang, Yong-Fang Kuo, Jean L. Freeman, James S. Goodwin. Continuity of Outpatient and Inpatient Care by Primary Care Physicians for Hospitalized Older Adults. JAMA. 2009;301(16):1671–1680. doi:10.1001/jama.2009.517
Author Affiliations: Department of Internal Medicine (Drs Sharma, Kuo, Freeman, and Goodwin) and Sealy Center of Aging (Drs Sharma, Zhang, Kuo, Freeman, and Goodwin), University of Texas Medical Branch, Galveston; and Clement J. Zablocki Veterans Affairs Medical Center and Department of General Internal Medicine, Medical College of Wisconsin, Milwaukee (Dr Fletcher).
Context Little is known about the extent of continuity of care across the transition from outpatient care to hospitalization.
Objectives To describe continuity of care in older hospitalized patients, change in continuity over time, and factors associated with discontinuity.
Design, Setting, and Participants A retrospective cohort study of 3 020 770 hospital admissions between 1996 and 2006 using enrollment and claims data for a 5% national sample of Medicare beneficiaries older than 66 years. Data files were constructed to include the patients’ demographic and enrollment information (denominator file) and claims for hospital stays (MEDPAR file) and physician services (carrier claims file). Characteristics of the hospitals were included in annual provider of services files. Being seen by a physician was defined as when a physician had submitted a bill for evaluation and management services for that patient.
Main Outcome Measures Percentage of patients who during hospitalization were seen by any outpatient physician they had visited in the year before hospitalization (continuity with any outpatient physician) or by their primary care physician (PCP) (continuity with a PCP).
Results In 1996, 50.5% (95% confidence interval [CI], 50.3%-50.7%) of hospitalized patients were seen by at least 1 physician that they had visited in an outpatient setting in the prior year, and 44.3% (95% CI, 44.1%-44.6%) of patients with an identifiable PCP were seen by that physician while hospitalized. These percentages decreased to 39.8% (95% CI, 39.6%-40.0%) and 31.9% (95% CI, 31.6%-32.1%), respectively, in 2006. Greater absolute decreases in continuity with any outpatient physician between 1996 and 2006 occurred in patients admitted on weekends (13.9%; 95% CI, 12.9%-14.7%) and those living in large metropolitan areas (11.7%; 95% CI, 11.1%-12.3%) and in New England (16.2%; 95% CI, 14.4%-18.0%). In multivariable multilevel models, increasing involvement of hospitalists was associated with approximately one-third of the decrease in continuity of care between 1996 and 2006.
Conclusion Between 1996 and 2006, physician continuity from outpatient to inpatient settings decreased in the Medicare population.
Continuity of care is a defining attribute of primary care and a core element of the Institute of Medicine definition of primary care.1 Continuity is generally recognized to have 3 dimensions—continuity in information, continuity in management, and continuity in the patient-physician relationship.2,3 Relationship continuity is the ongoing interaction of a patient with one physician, which results in increased knowledge of patient preferences, better communication, and improved trust.4,5 Guthrie et al6 have argued that relationship continuity is especially important for older patients, because of the greater likelihood of multiple chronic conditions that benefit from informed management and shared decision making.
Almost all current research on continuity of care focuses on relationship continuity in outpatient settings. Such outpatient continuity has shown to be associated with improved patient satisfaction,3- 5,7- 10 increased use of appropriate preventive health services,11- 15 greater medication adherence,16 lower hospitalization rates,17- 19 more appropriate end of life care,20- 22 and lower cost.23
One aspect that has received relatively less attention is continuity of care across transitions, such as from community to hospital or from hospital to nursing home. Such transitions threaten all aspects of continuity—information, management, and relationships.24- 26
We examined relationship continuity of care across the transition from community to hospitalization. We hypothesized that outpatient to inpatient continuity with hospitalized older adults decreased between 1996 and 2006, and that the decrease was greater in academic hospitals. We used 2 measures of continuity. A broad measure defined continuity as a hospitalized patient seeing any physician that he or she had seen in an outpatient setting at least once in the prior year. A narrow measure was restricted to patients with identified primary care physicians (PCPs) and whether those PCPs saw the patient in the hospital.
We conducted a retrospective cohort study using enrollment and claims data for a 5% national sample of Medicare beneficiaries.27 The Centers for Medicare & Medicaid Services selected these beneficiaries based on the eighth and ninth digits (05, 20, 45, 70, 95) of their health insurance claim number. Data files were constructed to include their demographic and enrollment information (denominator file) and claims for hospital stays (MEDPAR file) and physician services (carrier claims file).28 Characteristics of the hospitals were included in annual provider of services (POS) files.29
The study cohort consisted of all 5 487 771 hospital admissions between 1996 and 2006 for the 5% sample of Medicare beneficiaries aged older than 66 years. Excluded were admissions involving patients enrolled in health maintenance organizations or those who did not have Medicare part A and B for the entire year before the admission (n = 408 175). Patients with more than 20 admissions during the study period (n = 173 072 admissions) were considered outliers (exceeding the 99th percentile) and were excluded. Also excluded were 11 671 admissions that could not be linked to POS files, resulting in 4 894 853 admissions. For patients with more than 1 admission in a given year, 1 admission per patient per year was randomly selected, resulting in 3 020 770 admissions for the analyses in this study.
Patient Characteristics. Medicare enrollment files were used to categorize patients by age, sex, and race/ethnicity. Race/ethnicity was self-reported during the initial enrollment with the Social Security Administration30 and it was included because continuity may vary by race. A Medicaid indicator (state buy-in coverage) in the denominator file was used as a proxy of low socioeconomic status. This variable indicates whether the beneficiary applied for eligibility and met the low income requirement for Medicaid.31 Distance between the hospital and the patient's residence was calculated using zip codes.32 Comorbidity was measured with a score developed by Klabunde et al.33
Hospital Admission Characteristics. MEDPAR files were used to obtain characteristics of the admission, including origin of the admission (via the emergency department vs other), day admitted (weekend vs weekday), and discharge diagnosis related group (DRG). The DRGs were classified as medical or surgical. Medical DRGs included medical cardiology (121-145), medical neurology (9-35), medical pulmonary (78-102), medical gastroenterology (172-184), and medical other (43-48, 64-74, 185-190, 235-256, 271-284, 294-301, 316-333, 346-352, 366-369, 376, 378-380, 382-391, 395-399, 403-405, 409-414, 416-423, 425-433, 444-455, 462-467, 475, 487, 489-490, 492, 505, 508-511, 521-524). Surgical DRGs included surgical orthopedics (113-114, 209-213, 216-220, 223-230, 232-234, 285, 491, 496-503, 519-520, 531-532, 537-538, 544-546), surgical cardiology (103-111, 115-120, 515-518, 525-527, 535-536, 547-558), surgical gastroenterology (146-171, 191-201, 288, 392-393, 493-494), and surgical other (1-8, 36-42, 49-63, 75-77, 257-270, 285-293, 302-315, 334-345, 353-365, 370-371, 374-375, 377, 381, 394, 401-402, 406-408, 415, 424, 439-443, 461, 471-472, 476-477, 479-482, 484-486, 488, 495, 504, 506-507, 512-513, 528-530, 533-534, 539-543).
Physician and Hospital Characteristics. The specialty code for physicians was based on the specialty designation in the carrier claims files.34 The 2004 POS files were used to obtain hospital information including zip code, metropolitan area size, state, total number of beds, type of hospital, and medical school affiliation. Metropolitan area size and total number of hospital beds were categorized by quartiles; states were grouped by Census regions35; type of hospital was categorized as nonprofit, for profit, or public; and medical school affiliation was categorized as none, minor, or major.36
Outpatient Physicians. Outpatient physicians were identified as any physician who had billed the hospitalized patient for at least 1 outpatient evaluation and management (E&M) code in the prior year. Of the 3 020 770 admissions studied, 2 759 795 (91.4%) had at least 1 outpatient physician. The American Medical Association common procedural terminology (CPT) E&M codes 99201 to 99205 (new patient evaluation) and 99211 to 99215 (established patient follow-up) were used to establish outpatient visits.
Primary Care Physician. There is no agreed-upon algorithm to identify PCPs using administrative data.37- 43 We defined PCPs as a general practitioner, family physician, general internist, or a geriatrician who had billed an outpatient E&M code for the patient on 3 or more occasions in the year before the hospitalization. This definition has 76% concordance with the patient's self-identified PCP.43
Outpatient to Inpatient Continuity of Care. For most of the analyses, we assessed outpatient to inpatient continuity in 2 ways. First, we identified all hospitalizations in which the patient had seen (been billed by) a physician in an outpatient setting at least once in the year before admission. Among these hospitalizations (n = 2 759 795), we determined what percentage were associated with an E&M charge during the hospital stay by any physician who had billed for outpatient services in the prior year. This was termed continuity with any outpatient physician.
Second, we identified all hospitalizations in which the patient had an identifiable PCP (n = 1 547 513) and determined in what percentage of those hospitalizations the patient was seen by their PCP. This was termed continuity with a PCP. Inpatient claims were identified using American Medical Association CPT E&M codes 99221 to 99223 (initial hospital visit), 99251 to 99255 (inpatient consultation), and 99231 to 99233 (subsequent hospital visit).
We also explored different definitions of continuity. Instead of using an outpatient charge in the 1 year before the admission to establish the relationship, we expanded to 2 years and included physicians who had seen the patient either in an outpatient or inpatient setting during that time. For this analysis, our study cohort was limited to patients aged 67 years or older (2 years from enrollment age of 65 years) and hospitalized in 2006. In all cases where we use the term seen by a physician, we refer to a situation where the physician has submitted a bill for evaluation and management services for that patient.
Hospitalist. As previously described,44 we defined hospitalists as general internal medicine physicians who derive 90% or more of their Medicare claims for E&M services from care provided to hospitalized patients. If a patient had an inpatient E&M claim by a hospitalist physician during the hospitalization, he or she was considered to have been cared for by a hospitalist physician.
The study was approved by the University of Texas Medical Branch Institutional Review Board.
The proportions of admissions experiencing continuity of care with any outpatient physician or with their PCP were calculated and then stratified by selected patient and hospital characteristics. Linear trends in continuity were assessed by the Cochran-Armitage trend test. Three hierarchical generalized linear models with a logistic link, adjusting for clustering of admissions (level 1) within hospitals (level 2), were built for the outcome of continuity with any physician. In model 1, only admission year was included. Model 2 added patient characteristics (including age, sex, race/ethnicity, socioeconomic status, emergency admission, weekend admission, DRG groups, distance from hospitals, and comorbidity) and hospital characteristics (including region, metropolitan size, medical school affiliation, type of hospital, and hospital size). Model 3 added whether the patient was cared for by a hospitalist during the hospitalization. The mediating effect of hospitalist care on the association between admission year and continuity of care was evaluated by the method of Sobel.45 Finally, 2 prespecified 2-way interactions (admission year × medical school affiliation, and hospitalist care × medical school affiliation) were added to model 2.
All analyses were performed with SAS version 9.1 (SAS Institute Inc, Cary, North Carolina). The SAS GLIMMIX procedure was used to conduct multilevel analyses. All reported P values are 2-sided and P < .05 was considered statistically significant. The actual sample of 3 020 770 is overpowered for the analyses and likely to detect very small differences as statistically significant. Thus, the interpretation of analyses depends more on the magnitude of differences than on their statistical significance.
Between 1996 and 2006, 45.2% (95% confidence interval [CI], 41.1%-45.2%) of hospitalized patients received care during hospitalization by a physician who had seen them at least once as an outpatient in the prior year, and 38.0% (95% CI, 37.9%-38.1%) of hospitalized patients with an identified PCP received care from that PCP during hospitalization.
Table 1 and Table 2 show the percentage of the study sample from 1996-2006 seen during hospitalization by any physician the patients had visited as an outpatient in the prior year, or by their PCP, stratified by patient and hospital characteristics. Patients more likely to have continuity with any outpatient physician or with their PCP were older, had multiple comorbidities, were admitted through the emergency department, had medical (vs surgical) diagnoses, lived closer to the hospital or in smaller metropolitan areas, or were hospitalized in community hospitals. There was also geographic variation in continuity of care, with the Mountain region having the lowest continuity.
The Figure shows the percentage of patients experiencing outpatient to inpatient continuity of care with any outpatient physician, or with their PCP, by year of hospitalization. Outpatient to inpatient continuity with any outpatient physician decreased from 50.5% (95% CI, 50.3%-50.7%) in 1996 to 39.8% (95% CI, 39.6%-40.0%) in 2006 (P < .001). Similarly, outpatient to inpatient continuity with a PCP decreased from 44.3% (95% CI, 44.1%-44.6%) in 1996 to 31.9% (95% CI, 31.6%-32.1%) in 2006 (P < .001).
Table 3 and Table 4 show the absolute decreases in continuity of care from 1996 to 2006 by patient and hospital characteristics. Decreases in outpatient to inpatient continuity with any outpatient physician were observed in all strata, with greater absolute declines observed in those patients admitted on the weekend, those with medical diagnoses, and those living in larger metropolitan areas. Decreases varied markedly by geographic area, with the largest absolute decreases observed in New England. Similar decreases were observed in continuity with a PCP.
The results of multilevel multivariable analyses of factors associated with outpatient to inpatient continuity with any outpatient physician who saw the patient at least once in the year before hospitalization are shown in Table 5. In model 1, the odds of hospitalized patients receiving care from any outpatient physician decreased by 3.9% (95% CI, 3.8%-4.0%) per year between 1996 and 2006. This was not substantially changed (4.4%; 95% CI, 4.3%-4.4%) after adjusting for other relevant factors in model 2. Other factors associated with continuity were higher socioeconomic status; higher comorbidity; medical (vs surgical) DRG; admission on a weekday; smaller metropolitan area; and care in a nonteaching, nonprofit, or smaller-sized hospital. There were also substantial variations by geographic region.
Model 3 included all the variables in model 2 and introduced whether the patient was cared for by a hospitalist during hospitalization. Adding this variable did not substantially alter the odds of continuity associated with the other variables, with the exception of admission year. The decrease was 4.4% per year in model 2 and reduced to 2.8% per year (95% CI, 2.7%-2.9%) when care by a hospitalist was added in model 3.
We assessed whether changes in continuity of care varied between academic vs community hospitals by testing two 2-way interactions in model 3. The interaction of admission year and hospital teaching status (major medical school affiliation vs none) and the interaction of hospitalist and hospital teaching status were both significant. The odds of experiencing continuity of care decreased more rapidly at major teaching hospitals, (4.4% per year; 95% CI, 4.2%-4.5%) than at community hospitals (2.3% per year; 95% CI, 2.2%-2.4%; P < .001). Care by a hospitalist was associated with a 65.4% (95% CI, 65.0%-65.9%) reduction in odds of continuity of care at community hospitals vs a 53.4% (95% CI, 52.5%-54.4%) reduction in major teaching hospitals (P < .001).
When performing the analyses in Table 5 using the outcome of continuity with a PCP, the results were similar to those presented for continuity with any outpatient physician. Finally, we assessed the percentage of patients hospitalized in 2006 who were cared for in the hospital by any physician who had provided care for them in the prior 2 years, whether in an inpatient or outpatient setting. Using this definition, 47.7% of patients hospitalized in 2006 experienced continuity of care.
The proportion of patients experiencing continuity between outpatient and inpatient settings decreased substantially between 1996 and 2006. By 2006, only 39.8% received care from any physician they had seen at least once in an outpatient setting in the prior year, and 31.9% of hospitalized patients with an identifiable PCP received care from that PCP. Decreases in continuity of care occurred in all areas of the country, in all types of hospitals, and for all diagnoses.
Outpatient to inpatient continuity was considerably less common in academic than community hospitals. Also, the decrease in continuity was steeper over time at academic hospitals. By 2006, only 21.2% of Medicare patients hospitalized at academic hospitals saw their PCPs compared with 36.8% of patients at community hospitals.
Approximately one-third of the decrease in continuity between 1996 and 2006 was associated with growth in hospitalist activity, and there is a rough correspondence of regions of the country with the biggest decreases in continuity and those with the greatest increases in hospitalist activity.44
Maintenance of continuity of care is well recognized as a challenge for hospitalist care. Medicare allows reimbursement for only 1 generalist physician for concurrent care in the hospital,46 so there is a financial disincentive for outpatient physicians to follow their patients in the hospital if they are receiving hospitalist care. Problems with information transfer at admission and unclear responsibility after discharge have been identified,26 with direct communication between the hospitalist and PCP occurring in less than 20% of cases.47
Conversely, there are clear benefits to the hospitalist system, including greater efficiency in both hospitalized48 and outpatient care.49 It is also important that the increase in hospitalist care was related to only approximately a third of the decrease in continuity over the 1996-2006 period. None of the other patient or hospital attributes included in Table 5 were strongly associated with the decrease in continuity. We assume many organizational changes, of which the hospitalist movement is only the most visible, contributed to the decrease, including increasing participation in group practices. Shared call schedules as well as improvements in communication, such as electronic medical records, may reduce the need for a patient's outpatient physician to actively participate in their inpatient care.
Patients with comorbid illnesses and the oldest patients were more likely to have continuity with their outpatient physicians and with their PCP during hospitalization. This would seem an appropriate adjustment in a situation in which continuity may be more important.6
The limitations of our study are those inherent in analyses of administrative databases. For example, our method of assessing continuity was by determining whether a patient's outpatient physician submitted a charge to Medicare for E&M services while the patient was hospitalized. It is likely that some physicians visited their patients without billing. However, our estimate of continuity with a PCP was close to that obtained in a recent survey of 556 PCPs in California, in which 34% stated that they usually or always visited their hospitalized patients.50 A number of instruments using administrative data17,51- 56 or patient questionnaires11- 13,57- 59 have been developed to assess relationship continuity. Rodriguez et al60 assessed the concordance of patient-reported and administratively derived measures of continuity in a sample of 15 126 patients and concluded that administratively derived measures are less susceptible to bias.
A second limitation is that our study only examined changes in relationship continuity and did not have information on other domains of continuity, such as information or management continuity. A hospitalized patient may experience continuity in information without relationship continuity. For example, the patient might be followed up in the hospital by a group practice associate of his or her PCP. The growing use of the electronic health record should promote informational continuity.61- 63 In the era of electronic communication, physicians and patients are more likely to communicate via cell phones or emails to relay information. This is not captured in the administrative data sets, which may lead to an overestimation of the decrease in continuity with a PCP.
A third limitation is that our study used Medicare data; therefore, the findings may not generalize to non-Medicare patients. In addition, the findings may not be generalizable to Medicare beneficiaries enrolled in managed care plans or those without Medicare part B coverage. A change in health maintenance organization enrollments during the study period might have affected the analysis of the time trends in outpatient to inpatient continuity. Another limitation is the algorithm used to identify PCPs, which required 3 outpatient visits in the prior year and thus may have missed many PCPs. A recent study43 showed a 76% concordance between this algorithm and self-report by patients. Moreover, the absolute decreases in outpatient to inpatient continuity over time were similar whether we looked at continuity with a PCP or with any outpatient physician who saw the patient at least once in the prior year.
This study suggests that a majority of older hospitalized patients are not seen by any physician that they had visited previously. A recent study of patients hospitalized at an academic hospital found that 75% were unable to name any physician participating in their care.64 Hospitalization is a stressful time and is also the occasion for serious and sometimes difficult decisions regarding different treatments. The ongoing participation of a physician well known to the patient might help lead to decision making that reflects the patient's values. For example, patients hospitalized with advanced lung cancer in the last 6 months of life are less likely to spend time in an intensive care unit if they are cared for by their outpatient physician during that hospitalization.22
In conclusion, continuity of care from the outpatient to hospital setting is low and is decreasing. Only a part of this decrease appears to be related to the growth of the hospitalist model. Future research should explore whether the lack of continuity contributes to suboptimal care and whether interventions might ameliorate any detrimental effects of discontinuities in care.
Corresponding Author: Gulshan Sharma, MD, MPH, Department of Internal Medicine, University of Texas Medical Branch, 301 University Blvd, JSA-5.112, Galveston, TX 77555-0561 (email@example.com).
Author Contributions: Dr Zhang had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Sharma, Freeman, Goodwin.
Acquisition of data: Zhang, Goodwin.
Analysis and interpretation of data: Sharma, Fletcher, Zhang, Kuo, Goodwin.
Drafting of the manuscript: Sharma, Fletcher, Freeman, Goodwin.
Critical revision of the manuscript for important intellectual content: Sharma, Zhang, Kuo, Goodwin.
Statistical analysis: Zhang, Kuo.
Obtained funding: Sharma, Goodwin.
Administrative, technical, or material support: Freeman.
Study supervision: Goodwin.
Financial Disclosures: None reported.
Funding/Support: This work was supported by grant P50 CA 105631 from the National Cancer Institute and grants K-08 AG 031583 and P60 AG 17231 from the National Institute on Aging.
Role of the Sponsors: The funding agencies had no role in the design and conduct of the study, in the collection, management, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.