Information exchange among physicians caring for the same patient in the community

Study design

We conducted a multicentre prospective cohort study including patients discharged to the community from medical or surgical services of 11 Ontario hospitals in 5 cities after an elective or emergent admission. The 11 hospitals consisted of 6 university-affiliated institutions and 5 community hospitals. We did not include patients who had been discharged to nursing homes. Patients were eligible for enrolment if their cognitive skills were unimpaired, if they had a telephone and if they provided written informed consent.

The study was approved by the Ottawa Hospital Research Ethics Board, with subsequent approval by each participating site.

For the purposes of this study, we called the initial physician visit after the patient's discharge from hospital the “index visit.” In the analysis, visits for which we determined information availability were called “current visits,” and the physician for any such visit was the “current physician.” Visits that occurred before the current visit were called “previous visits,” and the physician for any such visit was a “previous physician.” All nonspecialist physicians were classified as family physicians. Physicians that a patient identified as having provided regular, continuing care before the hospital admission were classified as the patient's “regular physicians.”

Data collection and study outcome

Before each eligible patient's discharge from hospital, a member of the study team interviewed the patient to determine baseline functional status and to identify the patient's regular physicians. We examined the patient's chart and discharge summary, if available, to collect demographic information, a list of chronic medical conditions, information about the hospital stay and medications at discharge. We also determined the Charlson comorbidity index,15 which quantifies the severity of chronic illnesses, with values of 5 or more representing the greatest severity. After discharge, a member of the study team contacted each patient by telephone at 1, 3 and 6 months to identify the date and the physician for all visits that occurred after the discharge.

The primary outcome of the study was whether information about the patient from previous physician encounters was available to the current physician. We used 3 methods to gather the data. First, for any visit to a physician in the 6 months after discharge, we asked the patient to give the physician a paper survey. In this survey, the physician was asked to “List all visits this patient had with other doctors for which you have information (e.g., consult notes, phone calls).” We did not ask for any details about the previous information. The physician returned the completed survey in a postage-paid envelope. Study personnel supplied blank surveys to patients after they were recruited and after each follow-up phone call.

Second, if we did not receive a completed physician survey within 2 weeks after a reported visit, we faxed a copy of a different survey to the physician. This survey listed the dates of all visits that the patient had had with other doctors since the time of discharge, and the physician was asked to indicate whether information about each prior visit had been available for the current visit.

Third, if 4 weeks had passed since the survey was sent by fax and the physician had not returned the survey, we administered the survey by telephone, asking the physician or office personnel to review the patient's chart to determine whether information from other physicians was present. For the analyses in this study, we did not include visits to an emergency department and readmissions to hospital.

Statistical analysis

We calculated the overall probability of information availability as the number of times that information from previous visits was available divided by the total number of previous–current visit combinations. An illustration of the analytical data set and how it was created appears in Appendix 1 (available online at www.cmaj.ca/cgi/content/full/179/10/1013/DC2). We calculated the unadjusted association of information availability with categorical patient and physician factors using the average proportion of previous–current visit combinations for which information was available for patients with or without each factor. For continuous variables, we calculated the Pearson correlation coefficient with the proportion of visits for which information was available.

The visits included in our analysis were not independent. Therefore, we used a hierarchical mixed-effects multivariable logistic regression model to determine the independent association between patient, physician and visit factors and information availability.16 Details of our modelling strategy are given in Appendix 2 (available online at www.cmaj.ca/cgi/content/full/179/10/1013/DC2).

Strengths and limitations

Our study had several strengths that increase the generalizability of the results. It included a large number of patients from a broad range of services at several teaching and community hospitals, and the rate of follow-up was good. The analysis simultaneously accounted for many important patient-level and system-level factors that might influence information exchange. Finally, the study measured information exchange among all physicians who treated a defined group of patients. This represents an improvement over previous studies, which examined particular physician practices.

Our study also had several limitations. First, we were unable to determine the outcome variable for 19.0% of visits. However, an 81.0% response rate compares very well with the results of other survey-based studies. Also, we doubt that the validity of our results would be affected by these missing data, since physician response is unlikely to be associated with information availability.29

Second, we did not explicitly ask physicians what type of information they had received for previous visits, nor did we ask them about the quality or necessity of the information, or whether it was helpful. We simply asked physicians if they had “any” information about previous visits with other physicians. We limited the data requested from responding physicians to optimize response rates; however, this simplicity limited our ability to determine the nature and quality of the information available.

Third, our study relied on physician recall. If recall was poor, or if physicians completed the survey long after the original visit, the data they recorded would be less reliable.

Fourth, our study included only patients who had been discharged from hospital. The postdischarge period is frequently characterized by significant shifts in patient care, which could decrease information exchange. Whether our results can be extended to patients who have not received hospital care requires further study.

Fifth, we did not model the administrative or physical links between physicians treating the same patient. For example, we did not specifically model whether different physicians worked in the same practice. Such data could be important in explaining how information flows between physicians.

Sixth, our study was limited to information exchange between physicians. Although visits to other health care professionals are important, we did not capture data about such visits in this study.

Conclusion

Our study has revealed poor exchange of information between physicians who treat the same patient in the community. If continuity of patient care is to be increased, interventions and systemic modifications will be needed to improve information exchange.

@@ See related commentary by Reid and Wagner, page 987