Presented at the Society for Academic Emergency Medicine annual meeting, Boston, MA, June 2011.
Original Research Contribution
Improving Telephone Follow-up for Patients Discharged from the Emergency Department: Results of a Randomized Controlled Trial
Article first published online: 14 MAY 2013
© 2013 by the Society for Academic Emergency Medicine
Academic Emergency Medicine
Volume 20, Issue 5, pages 456–462, May 2013
How to Cite
Academic Emergency Medicine 2013; 20:456–462 © 2013 by the Society for Academic Emergency Medicine
The authors have no relevant financial information or potential conflicts of interest to disclose.
- Issue published online: 14 MAY 2013
- Article first published online: 14 MAY 2013
- Manuscript Accepted: 26 OCT 2012
- Manuscript Revised: 23 OCT 2012
- Manuscript Received: 1 AUG 2012
- Top of page
- Supporting Information
Reliable telephone follow-up offers physicians a timely method to notify patients of unexpected laboratory and imaging results, clarify discharge instructions, evaluate health status changes, and potentially boost patient satisfaction. This study sought to determine if verifying telephone numbers, obtaining best contact times, and informing patients that they will be contacted would increase the proportion of emergency department (ED) patients contacted at 48 to 72 hours postdischarge. Secondary outcomes included estimating successful postdischarge follow-up across demographic categories.
This was a prospective, randomized controlled trial of adult patients in an inner-city, public hospital ED. Patients were excluded for critical illness, no telephone number, non–English- or non–Spanish-speaking, or anyone admitted as an inpatient. All subjects completed a demographic questionnaire. In the intervention arm, research assistants (RAs) verified the telephone number, obtained best contact times, and informed subjects that they would be called. In the control arm, telephone numbers were obtained from registration materials and were not verified, and subjects were not informed that postdischarge telephone calls were planned. RAs made four telephone attempts to contact each subject 48 to 72 hours after discharge.
The intervention did not significantly improve postdischarge contact. Most subjects, 72.8% in the intervention group and 68.2% in the control group, were successfully contacted (difference = 4.6%, 95% confidence interval [CI] = –2.2% to 11.4%). On multivariate analysis, Hispanic ethnicity and owning a mobile phone were associated with increased odds of successful postdischarge follow-up.
Verifying contact information, obtaining best contact times, and notifying patients of impending follow-up calls did not substantially improve postdischarge telephone contact rates.
Mejora del Seguimiento Telefónico para Pacientes Dados de Alta desde el Servicio de Urgencias: Resultados de un Ensayo Clínico Aleatorizado
El seguimiento telefónico fidedigno ofrece a los médicos un método oportuno para notificar a los pacientes resultados inesperados de pruebas de imagen o de laboratorio, aclarar instrucciones al alta, evaluar los cambios en el estado de salud y potencialmente mejorar la satisfacción del paciente. En este estudio se determinó si verificando los números de teléfono, obteniendo los mejores momentos para llamar e informando a los pacientes que se contactará con ellos es posible incrementar el porcentaje de pacientes del servicio de urgencias (SU) contactados entre las 48–72 horas tras el alta. Los resultados secundarios incluyeron estimar el éxito de seguimiento tras el alta en función de las categorías demográficas.
Ensayo clínico con asignación aleatorizada de pacientes adultos atendidos en un SU hospitalario público en la zona centro. Se excluyeron los pacientes con enfermedades críticas, sin número de teléfono, que no hablasen inglés o español, o con ingreso hospitalario. Todos los sujetos completaron un cuestionario con datos demográficos. En el grupo de la intervención, los asistentes de la investigación (AI) verificaron el número de teléfono, obtuvieron los mejores momentos para contactar e informaron a los sujetos que serían llamados. En el grupo control, los números de teléfono se obtuvieron de los materiales de registro y no fueron verificados y no se informó a los sujetos que había programada una llamada tras el alta. Los AI hicieron cuatro intentos de llamada para contactar con cada sujeto 48–72 horas tras el alta.
La intervención no mejoró significativamente el contacto tras el alta. Se contactó de forma exitosa con la mayoría de los sujetos, 72,8% en el grupo de intervención y 68,2% en el grupo control (diferencia 4,6%, IC95% = −2,2% a 11,4%). En el análisis multivariable, la etnia hispana y tener un teléfono móvil propio se asociaron con un incremento de la probabilidad de éxito de seguimiento tras el alta.
Verificar la información de contacto, obtener los mejores momentos para la llamada y notificar a los pacientes de un contacto inminente de seguimiento no mejoró sustancialmente el porcentaje de contactos telefónicos tras el alta.
Postdischarge telephone follow-up is a valuable tool for improving the safety and efficiency of medical care.[1-4] Reliable telephone follow-up affords physicians a timely method to notify patients of unexpected laboratory results or imaging findings, clarify discharge instructions, and evaluate health status changes in response to treatment. Strategies involving postdischarge telephone calls are effective at reducing readmission rates and emergency department (ED) visits for recently hospitalized patients and have become routine in primary care settings.[1-4] However, there is widespread concern that ED patients are at increased risk for being lost to follow-up or are otherwise less reachable than patients in primary care settings.[5-8] In the absence of readily available follow-up strategies, emergency physicians (EPs) may be more likely to order additional tests or use prolonged observation to rule out unlikely conditions that could otherwise be handled with a watch-and-wait approach. Moreover, postdischarge telephone follow-up has been lauded as an effective mechanism to boost patient satisfaction. Patient satisfaction has become a key metric in value-based reimbursement for health care and is increasingly being publicly reported.[8, 9]
The reliability of telephone follow-up of ED patients has not been clearly characterized in the literature, particularly in resource-poor settings such as public safety-net hospitals.[6, 7, 10-15] The literature suggests the proportion of discharged ED patients who can be contacted by telephone ranges from 42% to 82%, and there have been few intervention studies designed to improve these numbers. Further, most reports were drawn from studies that occurred before the wide dissemination of mobile phones. Recent investigations have demonstrated that over 80% of people, including low-income and minority group members, have mobile phones. Given the proliferation of these new technologies, studies of interventions and patient characteristics that improve or are associated with successful post-ED telephone contact should be updated.
The primary objective of our study was to determine if a brief intervention, consisting of verifying contact information, obtaining best contact time, and notifying patients they will be contacted, will increase the proportion of discharged ED patients who are successfully contacted within 48 to 72 hours of discharge. Secondary goals of the investigation were to evaluate associations between characteristics of patients presenting to the ED and successful postdischarge telephone contact. In particular, we hypothesized that having a stable residence, stable telephone number, and owning a mobile phone are associated with successful postdischarge telephone follow-up.
- Top of page
- Supporting Information
This was a prospective, randomized, controlled single-blind trial from July 2010 to January 2011. The local institutional review board approved the study protocol.
Study Setting and Population
The setting was the ED at Los Angeles County + University of Southern California (LAC+USC) hospital. LAC+USC is the largest provider of uncompensated care in Los Angeles County. The ED sees 180,000 low-income, predominantly Hispanic patients annually. All adult patients physically located in ED beds or gurneys were eligible for inclusion. Patients were excluded for 1) age less than 18 years, 2) critical illness precluding the intake interview, 3) not having a telephone number of any type, 4) being admitted to the hospital, and 5) being non–English- and non–Spanish-speaking.
Enrollment occurred primarily during daytime hours subject to research assistant (RA) availability on 52 separate enrollment days. On each enrollment day, RAs approached a sample of potential subjects based on a randomly generated list of ED beds. Each bed in the ED had equal probability of being on the list. RAs sequentially approached the subjects from this list, screened them for eligibility, and obtained written informed consent. The RAs opened sealed, opaque, externally unmarked envelopes containing computer-randomized allocation assignments. RAs had no clinical involvement with the research subjects. RAs used the following script to inform both intervention patients and controls of the purpose of the study prior to consent:
“Hi, I am ________. I am a research assistant working on a project to determine the accuracy of contact information in the emergency department. They want to make sure they can reach patients to give them information if it is needed. If you decide to participate, you will be randomized to one of two groups and asked to fill out a one-page questionnaire. The questionnaire will take about 2 minutes to complete. Are you willing to participate?”
All enrolled subjects completed a questionnaire regarding demographic characteristics including age, sex, ethnicity, income, stability of residence, and stability of telephone number (see Data Supplement S1, available as supporting information in the online version of this paper). We considered the subject's address and/or telephone number “stable” if this variable had been consistent for the 3 months prior to enrollment. Subjects filled out the one-page multiple-choice questionnaire with the RAs present. RAs reviewed the surveys for completeness with the subjects to minimize missing or incorrect data. RA and subject enrollment interactions were scripted and memorized in both Spanish and English for interviewer consistency. Subjects were not compensated for participation.
The intervention consisted of three study procedures. First, each subject randomized to the intervention arm was explicitly told that he or she would be receiving a telephone follow-up call within 48 to 72 hours as part of the research study. Second, each intervention subject was asked to specifically verify up to three potential telephone numbers to be used during the follow-up phase and, last, each subject indicated the best times to attempt telephone follow-up. In the control arm, RAs obtained the follow-up contact number from the administrative materials, which were collected as part of the routine triage and registration process and not as part of the study procedure. RAs did not verify the numbers with the patients, did not obtain optimal follow-up times, and did not inform the patients when they might be called.
Forty-eight to 72 hours following enrollment, RAs reviewed each subject's electronic medical record to ensure that he or she had not been admitted to the hospital. Subjects admitted to the hospital were excluded from the study after enrollment. RAs then made a maximum of four attempts to call back each subject, with at least 4 hours (but typically 12 hours) between each attempt. Each attempt consisted of the RA calling all numbers listed for the patient on the triage registration form or, if the patient was part of the intervention group, from verified numbers provided by the subject to an RA. RAs could not be blinded to group assignment. If the call was not answered, messages were left asking the subjects to call the study telephone number. All follow-up attempts were documented. Successful telephone contact was defined as speaking with the patient, first-degree relative, or spouse. The follow-up phone call and messages were also scripted and delivered in Spanish or English according to patient preference (Data Supplement S2, available as supporting information in the online version of this paper).
All subjects were analyzed according to intention-to-treat principles. The primary outcome of interest was successful telephone contact within 72 hours across treatment arms. Based on previous literature, we estimated successful postdischarge telephone contact would occur in 65% to 70% of control subjects and determined that an absolute improvement in the proportion of subjects successfully reached of 10% would be clinically meaningful. With two-tailed alpha = 0.05 and beta = 0.8, a sample size of 626 to 698 total subjects was needed. Secondary endpoints included estimating successful postdischarge follow-up proportions across demographic categories. Student's t-test and Fisher's exact test of significance were used as appropriate. The literature and a priori hypotheses suggested that stability of address, primary language, ownership of a mobile phone, type of insurance, race/ethnicity, and our intervention were likely to be important predictors.[7, 12] Multivariate logistic regression was used to evaluate the independent contributions of each of these independent variables. Data were analyzed using STATA 11.0 (StataCorp, College Station, TX). Because we conducted seven tests of hypothesis (one primary, six secondary) we calculate that the family-wise error rate for this study or the probability of observing a chance association is 27%.
- Top of page
- Supporting Information
Figure 1 depicts the flow of patient enrollment and randomization into the study. A total of 1,781 were approached for enrollment based on their room assignment. Of those subjects, 759 were excluded for altered mental status, being critically ill, or unable to speak neither English nor Spanish or because they were admitted to the hospital at the time RAs approached them. Only 5.9% were excluded because they had no contact information. A total of 940 subjects were enrolled and randomized. Of these, 251 of these were excluded after enrollment because they were immediately hospitalized, leaving a total study sample of 689 subjects. Only 41 eligible subjects (8%) declined to participate.
Characteristics of Study Subjects
Subjects in this study were predominantly Hispanic (77.5%), spoke Spanish as their primary language (55.2%), had annual household incomes of less than $15,000 (75.8%), and were without medical insurance (53.7%). The vast majority lived in apartments or homes (89.0%), 84.9% had stable phone lines (defined as having the same phone number for more than three months), and 74.7% owned mobile phones (Table 1).
|Characteristic||Control Group (n = 343)||Intervention Group (n = 346)|
|Age (yr), median (IQR)||45 (33–55)||44 (34–55)|
|Female||183 (54.5)||171 (51.2)|
|Hispanic||260 (76.9)||268 (78.1)|
|Non-Hispanic white||26 (7.7)||27 (7.9)|
|Black or African American||25 (7.4)||22 (6.4)|
|Asian||14 (4.1)||15 (4.4)|
|Other race||10 (3.0)||6 (1.8)|
|Spanish language preference||186 (55.0)||193 (57.3)|
|Less than high school||138 (41.1)||143 (42.4)|
|High school graduate||111 (33.0)||113 (33.5)|
|Some college or higher||55 (16.4)||58 (17.2)|
|Completed college||24 (7.1)||18 (5.3)|
|More than college||8 (2.4)||5 (1.5)|
|Less than $15,000||260 (86.1)||261 (87.0)|
|Between $15,000 and $50,000||37 (12.3)||34 (11.3)|
|More than $50,000||5 (1.7)||5 (1.7)|
|Any private||3 (0.9)||8 (2.4)|
|Medicare||9 (2.7)||9 (2.7)|
|Medicaid||55 (16.5)||58 (17.3)|
|County||60 (18.0)||76 (22.7)|
|None||197 (59.0)||172 (51.3)|
|Other||10 (3.0)||12 (3.6)|
|Has primary care provider||82 (24.9)||109 (32.7)|
|Home or apartment||306 (90.0)||305 (89.1)|
|Shelter||5 (1.5)||4 (1.2)|
|Homeless||17 (5.0)||16 (4.7)|
|Has stable phone number for > 3 months||288 (85.5)||295 (87.3)|
|Has mobile phone||268 (79.5)||244 (74.2)|
Our overall contact percentage was 70.5% of patients across intervention groups, 72.8% in the intervention group and 68.2% in the control group (difference = 4.6%, 95% confidence interval [CI] = –2.2% to 11.4%; p = 0.19). RAs frequently left messages. For example, on the first callback attempts, messages were left 33% of the time (n = 229). In total 438 messages were left among all the callback attempts, but these were rarely replied to (29 replies in 438 messages).
In the secondary analysis, we found substantial differences in successful postdischarge follow-up across subgroups consistent with our a priori hypotheses (Table 2). Specifically, we found that owning a mobile phone (73.4% vs. 61.0%, difference = 12.4%, 95% CI = 3.8% to 21.0%), having a stable address (72.9% vs. 59.4%, difference = 13.5%, 95% CI = 4.0% to 23.1%), and having stable telephone number (73.1% vs. 55.4%, difference = 17.6%, 95% CI = 6.9% to 28.4%) were significantly associated with successful postdischarge contact regardless of the intervention allocation. In exploratory analysis, we observed that Hispanic patients, women, patients with Spanish language preference, and those who stated they had primary medical doctors were more likely to have successful telephone follow-up. We were only able to contact 40.4% of African American patients compared with 74.2% of Hispanic patients and 50.9% of non-Hispanic whites. We did not observe important differences in contact proportions associated with income, education, or insurance (Table 2). Multivariate regression showed that Hispanic ethnicity (odds ratio [OR] = 3.55, 95% CI = 1.73 to 7.29) and having a mobile phone (OR = 1.66, 95% CI = 1.09 to 2.54) were associated with higher odds of successful postdischarge telephone contact (Table 3).
|Characteristic||Percent Contacted (n/N)||Difference,% (95% CI)a||p value|
|Female||74.3 (263/354)||7.2 (0.3–14.1)|
|Spanish||74.1 (281/379)||10.4 (3.1–17.6)|
|Primary care provider|
|Has PCP||77.5 (148/191)||9.5 (2.2–16.7)|
|No PCP||68.0 (321/472)|
|Stable phone number||73.1 (426/583)||17.6 (6.9–28.4)|
|No stable number||55.4 (51/92)|
|Has mobile phone||73.4 (376/512)||12.4 (3.8–21.0)|
|No mobile phone||61.0 (94/154)|
|Over 75||62.2 (28/45)||0.750|
|Non-Hispanic white||50.9 (27/53)|
|Black/African American||40.4 (19/47)|
|Other race||81.3 (13/16)||<0.001|
|Less than high school||68.7 (193/281)|
|High school graduate||73.2 (164/224)|
|Some college or higher||67.3 (76/113)|
|Completed college||73.8 (31/42)|
|More than college||76.9 (10/13)||0.683|
|Less than $15,000||70.2 (366/521)|
|Between $15,000 and $50,000||76.1 (54/71)|
|More than $50,000||70 (7/10)||0.559|
|Any private||81.2 (9/11)|
|Home or apartment||73.5 (449/611)|
|Has mobile phone||1.66||1.09–2.54|
|Stable home or apartment||1.59||0.98–2.59|
- Top of page
- Supporting Information
While patients frequently present to the ED for nonspecific symptoms that most likely represent benign, self-limited conditions, these conditions may also represent early, mild, or atypical presentations of serious disease. Routinely, the EP must decide whether to pursue an aggressive diagnostic approach or a watchful waiting strategy. Numerous variables, including the patient's medical follow-up plan, will influence this decision. Obvious ability to make telephone contact in the event of diagnosis-changing laboratory results, radiographic overread, or clinical deterioration may provide physicians with added confidence to expedite ED discharge. Conversely, poor capacity for telephone follow-up may provoke more cautionary testing and longer ED observation periods.
To date, however, the reliability of telephone follow-up of ED patients has not been clearly characterized in the literature, particularly in the digital age.[6, 7, 12, 13, 17-19] Reported postdischarge telephone follow-up rates in studies between 1992 and 2000 varied from 42% to 82% of general ED patients, depending on the setting and population studied.[6, 7, 12, 13, 17-19] In the largest published study evaluating postdischarge ED follow-up, across 77 EDs in the United States and Canada, 71% of patients seen in the ED for respiratory complaints were contacted 2 weeks after discharge from the ED. Other studies have similarly focused on postdischarge follow-up rates for special populations, including sexual assault victims, pediatric patients, and patients with psychiatric complaints.[6, 7, 10-15, 18, 19] Studies based in private hospitals (dating from 1995 to 2000) found relatively high contact percentages (69% to 82%).[6, 10] However, studies in state-supported hospitals report postdischarge telephone contact percentages between 42 and 55%.[7, 11] These dismal contact rates would seemingly preclude relying on such a follow-up method. However, the reports demonstrating such small chances for telephone follow-up predate the mobile phone era. With increasing use of mobile phones, particularly among poorer populations, these results may no longer be valid. In addition, the association of demographics and stability of telephone access with contact percentages has not been well characterized in prior studies. Perhaps even more importantly, few practical interventions that might improve postdischarge telephone follow-up have been reported in the emergency medicine literature.
In a small randomized controlled trial published in 1997 (n = 134), Ferrigno et al. found a significant improvement in successful telephone follow-up (54% success in controls compared with 77% success in the intervention group) using a similar strategy to the one used in the present study. The study by Ferrigno et al. was conducted in 1997 and used landlines and pagers to contact discharged patients. It is unclear whether their findings remain relevant in an era with widespread mobile phone availability and minimal pager use.
In the present study we attempted an intervention to increase postdischarge telephone contact in a large, public ED. Overall, we contacted 70.5% of the study sample, which compares favorably with previous reports. Although the intervention was simple, feasible, and likely generalizable, the results of this study indicate the intervention was not strong enough to increase postdischarge follow-up by the 10% the study was powered to detect. At a 29.5% failure rate, telephone follow-up is an unreliable mechanism to notify patients of very important clinical findings. Should the EP wish to discharge the patient with important clinical examinations pending, alternate mechanisms to ensure communication with the patient must be undertaken (scheduled ED return visit, documented notification to the primary care physician, etc.) However, for more modest targets, such as to clarify discharge instructions, evaluate health status changes, or improve patient satisfaction, telephone follow-up success (70.5%) appears viable.
Our secondary analysis confirmed the a priori hypotheses that having a stable address and stable telephone number and owning a mobile phone are associated with higher contact percentages. In this sample, 76.9% of subjects had a mobile phone and we were able to contact 73.4% of them, compared with 61.0% of subjects who did not have a mobile phone. This high prevalence of mobile phone ownership, along with the significantly higher contact rates for those with mobile phones, may account for the higher overall contact proportion observed in our study compared with other studies conducted in public hospitals in the pre–mobile phone era.
In exploratory analysis, we observed substantial variation in contact proportions across racial and ethnic lines, with Hispanic patients having higher contact success than African American or non-Hispanic whites. The reasons for this curious finding were not explored in our study. We did not observe any effect of self-reported income on postdischarge telephone contact. However, this may reflect a general lack of variation in income, as 86.5% of the study sample reported annual household incomes below $15,000. Similarly, insurance status was not associated with differential contact rates.
- Top of page
- Supporting Information
We performed this study at a single, large, urban county institution with a primarily uninsured, Hispanic population. This population is likely to be one of the most difficult to reach, making these contact percentages conservative estimates of what might be expected in other ED environments. A second limitation is that we sampled primarily during daytime hours depending on RA availability. Therefore, the convenience sample that was studied may not be fully representative of ED patients who present at different times of day. Additionally, patients had to agree to participate and sign informed consent forms, possibly creating a selection bias. We believe this bias was minimal, however, as only 8% of eligible subjects declined to participate.
We considered a positive outcome if the RA successfully contacted the patient or a first-degree relative or spouse. Because this definition of success expanded the potential number of points of contact for a patient, our estimate may be less conservative compared to other studies that required direct patient contact. However, contact of relatives or spouses following discharge is common clinical practice in most EDs, and therefore we believe this outcome should be considered successful.
We were not always able to call the patient at his or her exact preferred callback times (this occurred less than 20% of the time). Although this may have reduced the effect of the intervention, we believe that this limitation mirrors the reality of a post-ED callback process in clinical settings and therefore preserves the general validity of the findings. Perhaps more importantly, our efforts to call the patient back were not motivated by medical need. It is possible that the contact percentages would have been higher had the messages left indicated some medical urgency requiring a return telephone call.
In our study, the RA making the telephone follow-up call was not blinded as to the randomization assignment, introducing the possibility of observer bias. We believe that this risk is mitigated because 1) the outcome of interest was highly objective (i.e., the person either answered the phone or he or she did not); 2) identical message scripts were left when the RA reach a voice mailbox regardless of allocation; 3) all four calls were made on 100% of subjects in each arm before being called a callback failure (100% adherence to protocol); and 4) control and intervention calls were made during the same sessions, making it less likely that the specific time of day was being used to systematically bias in favor of one arm or the other.
Finally, as the 95% CI of the point estimate of the treatment effect includes 10% (4.6%, 95% CI = 2.2% to 11.4%; p = 0.19), there remains a small possibility the negative findings represent a type II error that arises because the study, by design, only had 80% power to detect a 10% absolute difference in telephone follow-up.
- Top of page
- Supporting Information
Although we were able to contact 70.5% of the total study sample, the simple intervention consisting of verifying contact information, obtaining a best contact time, and notifying the subject of an impending follow-up call did not substantially improve postdischarge telephone contact rates. Patients with a stable address and stable telephone number and those who owned a mobile phone were more likely to be successfully contacted than their counterparts. These findings may help practitioners identify a cohort of patients that are substantially more likely to have successful postdischarge telephone follow-up. However, stronger interventions are required to ensure postdischarge telephone contact for all patients.
- Top of page
- Supporting Information
- 2Reduction of 30-day postdischarge hospital readmission or emergency department (ED) visit rates in high-risk elderly medical patients through delivery of a targeted care bundle. J Hosp Med. 2009; 4:211–8., , , et al.
- 4Telephone follow-up, initiated by a hospital-based health professional, for postdischarge problems in patients discharged from hospital to home. Cochrane Database Syst Rev. 2006; 4:CD004510., .
- 8[No authors listed]. With Medicare revenues at stake, ED managers place new importance on elevating the patient experience. ED Manag. 2012;24:1–4.
- 9[No authors listed]. With patient satisfaction under increasing scrutiny, consider patient callbacks. ED Manag. 2011;23:81–3.
- 16Latinos and Digital Technology, 2010. Pew Hispanic Center. Feb 9, 2011. Available at: http://www.pewhispanic.org/files/reports/134.pdf. Accessed Jan 23, 2013.
- Top of page
- Supporting Information
|acem12128-sup-0001-DataSupplementS1.docx||Word document||96K||Data Supplement S1. Control survey.|
|acem12128-sup-0002-DataSupplementS2.docx||Word document||110K||Data Supplement S2. Intervention survey.|
Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.