Referrals from an Emergency Room to Primary Care Practices at an Urban Hospital

Referrals from an Emergency Room to Primary Care Practices at an Urban Hospital JOHN H. STRAUS, MD, SUEZANNE TANGEROSE ORR, PHD, AND EVAN CHARNEY, MD Abstract: Three hundred ninety-eight patients treated in the emergency room for non-urgent complaints who stated that they had no regular source of primary medical care were referred to one of several medical care centers in the area. Overall, 34 per cent of such patients complied with the referral. Correlates of compliance were: age (very young and very old), patient-perceived health status, medically determined need for follow-up care, and having an appointment made by the emergency room provider. Another group of 500 successfully referred patients became excellent users of the primary care office, complying with requested health maintenance and follow-up visits. However, emergency room utilization by the successfully referred patients did not decrease more than among referred patients who did not enroll in the primary care source. (Am J Public Health 1983; 73:57-61.) During the 20-year period from 1954 through 1974, there was a nearly 600 per cent increase in visits to hospital emergency rooms. A significant proportion of the more than 77 million visits to emergency rooms annually were made by persons who need primary, not emergency, medical care.' This phenomenon has been noted in various sized metropolitan, rural, and suburban communities.2 In 1975, close to 5 per cent of all visits to physicians occurred in hospital emergency rooms, and for low income persons this proportion was larger.3 It has been suggested that one reason for this non-urgent use of the emergency room is the lack of an available alternative. In fact, the majority of patients who visit emergency rooms-particularly in urban areas-have no regular primary care provider.4,5 Several investigators have demonstrated a decrease in emergency room use among recent enrollees in Neighborhood Health Centers or other primary care programs. Most of the enrolled patients had previously utilized hospital-based sources for primary care, such as outpatient departments or emergency rooms.6-'0 The emergency service of Sinai Hospital of Baltimore, Maryland has attempted over the past five years to reduce non-urgent use, first by establishing several hospital-related primary care programs to complement existing community resources, and second by referring those patients without a regular source of primary care to one of the primary care providers in the community.* The present study was under- Address reprint requests to Dr. John H. Straus, Senior Pediatrician, Primary Care Center, Department of Pediatrics, Sinai Hospital, 5101 Lanier Avenue, Baltimore, MD 21215. Dr. Orr is Director, Division of Pediatric Research; Dr. Charney is Pediatrician-in- Chief, Department of Pediatrics, also at Sinai Hospital. This paper, submitted to the Journal October 14, 1981, was revised and accepted for publication April 5, 1982. *Selected because of nearness to hospital, acceptance of all classes of patients and payment mechanisms, and easy access to appointment and other records. 1982 American Journal of Public Health taken to determine what proportion of unaffiliated patients are successfully referred from an emergency room to primary care provider, and to identify characteristics which are associated with compliance with such a referral. A secondary purpose of the study was to determine use patterns of primary care and emergency services after referral. This would allow us to determine if successful referral was associated with changes in emergency room use and appropriate primary care use. Materials and Methods The study sample was selected from among patients seen for non-urgent** conditions in the Emergency Department of Sinai Hospital of Baltimore, a 516-bed community hospital located on the northwest urban-suburban border of the city. There are 60,000 visits annually to this service. Patients were enrolled into the study in two cohorts: a concurrent cohort consisting of 398 patients enrolled prospectively from October to December 1979, and a nonconcurrent cohort of 500 patients enrolled by reviewing all emergency room records from November 1977 to December 1978. The concurrent cohort was included to obtain detailed data about health-related attitudes which might be associated with successful referral, while the non-concurrent cohort was included to ascertain the appropriateness of health services use over the two years following referral. All study subjects had the following characteristics: * They received care in the medical or pediatric non-urgent area of the emergency room;** * They identified no regular primary care physician; **Entering patients were triaged to "screening" (non-urgent) or "acute/surgical" (urgent, emergent) areas of the emergency room by a registered nurse. Approximately 70 per cent of all pediatric and 50 per cent of all adult patients were triaged to the non-urgent area. AJPH January 1983, Vol. 73, No. 1 57

STRAUS, ET AL. * They resided in the Sinai Hospital catchment area; * They were not admitted to the hospital for their presenting problem. Following the acute care visit, patients eligible for inclusion in the study were given the name and telephone number of one of three primary care programs and encouraged to call for an appointment for a follow-up or health maintenance visit. The three sites were a Sinai Hospital based Primary Care Center, a Sinai Hospital resident continuity program, and a nearby Community Health Center. All three offer comprehensive services, a 24-hour answering service, and continuity of provider. The referral was made by the emergency physician and/or staff ntrse. The effort made by the emergency room provider to accomplish the referral ranged from only giving a brochure, to giving a discussion about the importance of primary care, to actually making an appointment for the patient. For the concurrent cohort, this level of effort (a three-point scale from "minimum" to "maximum") was recorded by the provider making the referral. The techniques used to accomplish the referrals were those that had been in use over the preceding several years; they had been initially selected because they were simple and involved no extra personnel or expense. For each cohort of patients, data were collected for each of the independent variables listed in the Appendix. The rating of "medical need" was determined by a review by two primary care providers of the emergency roomn visit. Raters, working independently, Utilized information from the history, physical examination, laboratory data, and diagnosis to place the visit in one of eight categories. Using this method, inter-rater reliability was 80 per cent. Disagreements were resolved by consensus with little difficulty; most of the disagreements were not conceptual ones but occurred because one reviewer misread the record. The eight categories of the scale of medical need are shown in the Appendix. The self-administered questionnaire was given to the concurrent cohort while they were waiting to be seen; completed questionnaires were obtained from 86 per cent of the subjects. The study had four outcome variables: * Enrollment and completion of an initial visit at the primary care site (both cohorts); * Completion of requested health maintenance visits at the primary care site (non-concurrent cohort only); * Completion of requested follow-up visits at the primary care site (non-concurrent cohort only); and * Change in the rate of emergency room visits for non-urgent problems before and after the referral (non-concurrent cohort only). The primary dependent variable of interest was enrollment in the referral site. Patients were defined as "compliers" with the referral if they accomplished a visit at the primary care source within six months of the referral date. In reality, 80 per cent of patients who complied with a referral had their first visit within three months following the referral. Data about initial visits were collected from computerized records and appointment logs at the three sites. For the non-concurrent cohort only, emergency room use was determined for the 24 months following and 18 58 months prior to the referral.*** ER use by patients who did and those who did not comply with the referral were compared. To identify factors associated with compliance, both contingency tables (with chi square tests) and discriminant analysis were utilized. Contingency table analysis was conducted to evaluate the associations between each independent variable and compliance. Data were organized into strata to evaluate these associations at various levels of confounding variables when appropriate. Chi square tests were used to assess the statistical significance of observed associations. Two-group discriminant analysis" was utilized to identify the best combination of variables to predict group membership (complier, non-complier). This multivariate technique involved a categorical dependent variable and a mixture of categorical and continuous independent variables. The analysis focused upon the development of the "best" mathematical model to classify members of the sample as "compliers" or "non-compliers". Variables were allowed to enter the model in a forward stepwise fashion. Model selection was based upon minimization of Wilks' lambda, the accuracy of classification (per cent classified correctly), and the multivariate F-statistic. Results The demographic characteristics of both study cohorts were essentially the same. For the concurrent cohort, 95 per cent of the patients were Black, 49 per cent were Medicaid recipients, 61 per cent were female, 57 per cent had a high school education or more, and the sample was equally divided between children and adults. Overall, 34 per cent of the referred emergency room patients enrolled and made an initial visit to the primary care site (Table 1). Several independent variables were found to be significantly related to successful referral.t Seventy per cent of all patients who had a primary care appointment made at the time of the emergency service visit enrolled in the primary care source, compared to only 28 per cent of patients without this appointment (Table 1). This difference remained even when other independent variables were controlled. However, the patients for whom an appointment was made were not selected randomly by the providers; they had a greater need for health care (more chronic disease) and poorer self-perceived health status. Additionally, the provider in the emergency room may have sensed some attitudinal factor not measured in this study which suggested that an appointment for the primary care site would be kept. Be- ***Actual exposure time before referral ranged from 13-26 months (mean = 20.1 months), and after referral ranged from 17-30 months (mean = 22.9 months). 4These results are given only for the 398 subjects in the concurrent cohort; the data for independent variables obtained by the qtiestionnaire were only available for this cohort and analysis of data for the non-concurrent cohort produced no significantly different findings. AJPH January 1983, Vol. 73, No. 1

EMERGENCY ROOM REFERRALS TABLE 1-Appointment Status and Enrollment at Primary Care Site, Concurrent Cohort # Enrolled % Enrolled N Appointment not made in ER 99 28.4 348 Appointment made in ER 35 70.0 50 TOTAL 134 33.7 398 x2 = 31.0 (p <.0001) cause of the strong association of having an appointment made with the dependent variable and several independent variables, it was viewed as a confounder, and data were analyzed separately for those with and without such appointments. For the concurrent cohort, among those patients who were not given an appointment in the emergency room, four factors were significantly associated with enrollment: age of patient, patient-perceived health status, medical need for follow-up care, and prior experience in a primary care system (Table 2). Discriminant analysis was used to identify a set of variables (model) which would discriminate between enrollees and non-enrollees. The final model contained nine independent variables, shown in Table 3 in order of inclusion in the model. This model correctly classified 70 per cent of the study population (75 per cent of the non-compliers and 60 per cent of the compliers). The discriminant analysis was conducted utilizing only those patients for whom there were no missing values for any of the variables in the model (N = 288). It was of interest to determine not only enrollment in a primary care system, but also how well patients subsequently utilize that care. Using data from the non-concurrent cohort, use of the primary care sites for periods of one to two years following the referral was examined, focusing upon TABLE 2-Independent Variables Associated with Enrollment In Primary Care Site, Concurrent Cohort* Per Cent Variable Category Enrolled Patient Age* 0-5 yrs 48 5-34 yrs 24 >34 yrs 45 Patient Perceived Health Less healthy 41 Statust About the same 28 More healthy 15 Medical Need for Follow-up needed 39 Follow-up Caret Follow-up not needed 25 Previous use of a Primary Yes 44 Care Systemf No 26 *N = 348 (excludes patients for whom appointment was made in emergency room). **0-5 yrs vs. 5-34 yrs, p <.05 >34 yrs vs. 5-34 yrs, p <.05 tp <.05 completion of requested health maintenance and follow-up visits. Among the 233 patients in the non-concurrent cohort who made an initial visit to their primary care provider, 85 per cent made a subsequent health maintenance visit. Over the follow-up period, 90 per cent of 380 requests for health maintenance were kept and 87 per cent of 598 requests for follow-up visits were kept. These high percentages suggest that the successfully referred patients became excellent users of the primary care office. The last dependent variable examined was the utilization of the emergency room during the 24 months following referral as compared to the 18 months preceding referral. Figure 1 shows that although both groups decreased their utilization rate after referral, the decreased utilization of enrolled patients was no greater than that of the non-enrolled (p >.7). This remained true even when utilization was analyzed controlling for medical need, financial status, and age. Furthermore, when the proportions in each group that returned to the emergency room were compared, no significant differences were found. Discussion The findings suggest several directions for further research. The striking effect of making an appointment should be tested as an intervention using a randomized clinical trial. Such a trial should include patients of different ages, levels of medical need, and health status. The study should involve a careful examination of appointment breaking and longterm utilization patterns at the referral site, since it is inefficient to burden these practices with patients who will increase their administrative problems with inappropriate use. Another direction for future research concerns the use of medical need ratings and/or patient's self-perceived health status to identify patients who are likely to comply with a referral. Intervention strategies developed to refer patients to primary care systems could then be more specifically designed. Patients who are likely to comply because of their health status require a different type of referral strategy than those who are not as likely to comply. If our goal is to refer all patients to primary care systems, then individualized interventions are needed. A further use of the categorization of medical need is as a tool for comparing emergency rooms. Comparisons can be made along two dimensions. First, one can measure the proportion of non-urgent emergency room users who have a problem which requires follow-up, such as a chronic disease. In an emergency room, this proportion should ideally be low. Second, one can use an assessment of medical need to measure the proportion of patients in need of follow-up who then receive a referral to a primary care provider while in the emergency room. This percentage should ideally be high. The finding that emergency room utilization would be unchanged by successful referral to a primary care setting was unexpected. This finding has persisted for another sample of patients analyzed in a manner similar to this study. This result is consistent with the observation at Sinai Hospi- AJPH January 1983, Vol. 73, No. 1 59

STRAUS, ET AL. TABLE 3-Variables in Discriminant Model, Concurrent Cohort* Variables Coding/Interpretation Dependent Variable 0 = not enrolled in primary care site 1 = enrolled in primary care site Independent Variables Age **0 = <5, >35 years 1 = 6-34 years Appointment Status 0 = appointment not made **1 = appointment made by emergency room provider for primary care site Health Status **0 = worse self-perceived health status 1 = better self-perceived health status Insurance **0 = Blue Cross or Commercial 1 = Medicaid Last Visit to Regular Doctor **0 = within last year 1 = prior to last year Education 0 = <12 years **1 = >12 years Medical Need **0 = need follow-up care 1 = do not need follow-up care Regular Doctor **0 = past history of regular doctor 1 = no history of regular doctor Travel **0 = lives close to primary care site 1 = lives far from primary care site Multivariate F = 6.55 (p <.001) Wilks' lambda =.8218 (p <.001) * N = 288 **Associated with enrollment in primary care site. tal that while during the past four years, the percentage of children treated in the emergency room who have a regular physician has increased from 50 per cent to 80 per cent, the number of visits to the emergency room by children has remained constant. To decrease non-urgent emergency room utilization may require not only the availability of a primary care system, but also an increase in the barriers to the receipt of non-urgent care in the emergency room. From this study, we are encouraged that many of the people using emergency rooms who have no alternative Visits per * =Erolled Patient per Year Non-emlled 1.50 1.25 _-*(1.22) 1.00 A (1.01).75 Before Referral A (.79) After Referral +"Before Referral": 18 nonths prior to referral "After Referral": 24 nonths follcwing referral FIGURE 1-Emergency Room Utilization Before and After Referral, Non-Current Cohort + source of care can be successfully referred to a more appropriate primary medical care provider. REFERENCES 1. Piore N: Ambulatory care issues in the United States today. In: Bryant JH, Ginsberg AS, Goldsmith SB, Olendski MC, Piore N (eds): Community Hospitals and Primary Care. Cambridge MA: Ballinger, 1976, pp 3-26. 2. Orr ST: Changing Patterns of Emergency Room Utilization in Winston-Salem, 1971-1973. Unpublished thesis, Wake Forest University, Winston-Salem, NC 1974. 3. Health, United States, 1976-1977. US Dept of Health, Education, and Welfare, Public Health Service, Health Resources Administration, NCHS and MCHSR, Hyattsville, MD. Publ. No. (HRA) 77-1232. 4. Haynes MA, McGarvey MR: Physicians, hospitals and patients in the inner city. In: Norman JC (ed): Medicine in the Ghetto. New York: Appleton-Century-Crofts, 1969, pp 117-124. 5. Roth JA: Utilization of the hospital emergency department. J Health Social Behav 1971; 12:312. 6. Skinner TJ, Price BS, Gorry, GA: Directing the hospital outpatient to the neighborhood health clinic. Am J Public Health 1977; 67:1191. 7. Hochheiser LI, Woodward K, Charney E: Effect of the neighborhood health center on the use of pediatric emergency departments in Rochester, New York. N Engl J Med 1971; 285:148. 8. Gold M, Rosenberg RG: Use of emergency room services by the population of a neighborhood health center. Health Services Reports 1974; 89:65. 9. Ullman R, Block JA, Boatright NC, Stratmann WC: Impact of a primary care group practice on emergency room utilization at a community hospital. Med Care 1978; 16:723. 10. Hillman B, Charney E: A neighborhood health center: what the patients know and think of its operation. Med Care 1972; 10:336. 11. Nie NH, et al (eds): Statistical Package for the Social Sciences. New York: McGraw-Hill, 1975. 60 AJPH January 1983, Vol. 73, No. 1

EMERGENCY ROOM REFERRALS ACKNOWLEDGMENTS This research was supported by grant No. HS 04005 from the National Center for Health Services Research, OASH. An earlier version of this paper was presented at the 1980 Annual Meeting of the American Public Health Association in Detroit, MI. APPENDIX Collection and Coding of Independent Variables Collected by Chart Review (both cohorts) * Age, sex, race * Medical insurance status * Primary and secondary diagnoses * Medical need for follow-up care for identified problems Categories of Medical Need 1. Chronic disease present or suspected-follow-up needed within two months (e.g., hypertension, asthma) 2. Chronic disease present or suspected-follow-up needed occasionally (e.g., eczema) 3. Chronic disease present or suspected and follow-up required for acute condition (e.g., urinary tract infection in a diabetic patient) 4. Follow-up needed for acute condition only (e.g., bronchopneumonia) 5. No follow-up necessary-return only if condition not resolving (e.g., gastroenteritis) 6. Psychosocial problem which would interfere with compliance with referral (e.g., alcoholism, drug abuse, major psychopathyology) 7. Need for follow-up is controversial (e.g., otitis in children, urinary tract infection for adults) 8. Referral to subspecialist in addition to primary care provider is appropriate (e.g., fracture follow-up care by orthopedist) Collected by Self-administered Questionnaire (concurrent cohort only) * Educational level of patients (or parent) * Previous involvement in a primary care system * Patient's perception of their own health status * Location and frequency of medical visits over past year * Ease of transportation to the referral site during office hours * Satisfaction with present health services I American College of Physicians 64th Annual Session The American College of Physicians (ACP) will hold its Annual Session April 11-14, 1983 at the Moscone Center, San Francisco, California. About 6,000 doctors of internal medicine will attend the four-day event, at which more than 230 physicians will deliver about 340 sessions on the current ideas, procedures, and developments in internal medicine. Keeping up with medical advances is essential to practicing physicians, medical researchers, instructors, and physicians-in-training, and the ACP convention seeks to provide them with the information necessary to apply discoveries, breakthroughs, and changes to everyday professional experiences. For further information, contact Jane Ayers, American College of Physicians, 4200 Pine Street, Philadelphia, PA 19104. Telephone 215/243-1200 or 800/523-1546. I AJPH January 1983, Vol. 73, No. 1 61