Postdischarge needs identified by an automated text messaging program: A mixed‐methods study

Abstract

Background

Text messaging has emerged as a popular strategy to engage patients after hospital discharge. Little is known about how patients use these programs and what types of needs are addressed through this approach.

Objective

The goal of this study was to describe the types and timing of postdischarge needs identified during a 30‐day automated texting program.

Methods

The program ran from January to August 2021 at a primary care practice in Philadelphia. In this mixed‐methods study, two reviewers conducted a directed content analysis of patient needs expressed during the program, categorizing them along a well‐known transitional care framework. We describe the frequency of need categories and their timing relative to discharge.

Results

A total of 405 individuals were enrolled; the mean (SD) age was 62.7 (16.2); 64.2% were female; 47.4% were Black; and 49.9% had Medicare insurance. Of this population, 178 (44.0%) expressed at least one need during the 30‐day program. The most frequent needs addressed were related to symptoms (26.8%), coordinating follow‐up care (20.4%), and medication issues (15.7%). The mean (SD) number of days from discharge to need was 10.8 (7.9); there were no significant differences in timing based on need category.

Conclusions

The needs identified via an automated texting program were concentrated in three areas relevant to primary care practice and within nursing scope of practice. This program can serve as a model for health systems looking to support transitions through an operationally efficient approach, and the findings of this analysis can inform future iterations of this type of program.

INTRODUCTION

The transition from hospital to home is fraught with challenges for patients, as they recover strength, manage medications, and coordinate follow‐up care. ¹ , ² A variety of postdischarge outreach programs have been developed and tested, with the overarching goal of proactively identifying and addressing patient needs. ³ , ⁴ These programs have been shown to reduce hospital revisits in a variety of settings, ⁵ , ⁶ although not consistently so. ⁷ In general, higher intensity interventions have been more effective. ⁸

Nurse‐led phone calls—whether from the primary care practice or a hospital‐based team—are the predominant mode of outreach after discharge, but these are time intensive and therefore difficult to scale beyond a single, postdischarge contact. ⁹ , ¹⁰ To overcome this limitation and to lower the barrier to two‐way communication between the patient and their primary care practice, we designed and tested a 30‐day automated text messaging program to support patients from the primary care setting after hospital discharge. This intervention was associated with a reduction in acute care revisits in a single site pilot, ¹¹ but not in a larger RCT. ¹² In both studies, however, patient engagement and satisfaction with the program was high.

Relatively little is known about how patients use these types of transitional care programs and the postdischarge needs that arise. This is important for two reasons. One is that limited frameworks exist, generally, for considering the types of needs that arise postdischarge, and this program offers a rich data source for observing these needs in real time. In addition, as digital health technologies proliferate, transitional care programs that leverage these approaches to expand and simplify communication will similarly grow. ¹³ The growth of electronic patient portals has shown us that this is a double‐edged sword, ¹⁴ , ¹⁵ raising concern for over‐utilization and low‐value uses of clinician time. ¹⁶ It is important to understand whether increasing access to clinical care via texting leads to high‐value interactions.

To this end, we conducted a secondary, mixed‐methods analysis of a previously published pilot study. ¹¹ We reviewed the needs that were identified during the course of this pilot study and organized them thematically along the Ideal Transition in Care framework. ¹⁷ Our objective was to describe the types and timing of postdischarge needs identified during this 30‐day program. This, in turn, could (1) help us better understand what mediated the positive clinical effect observed in this pilot study, and (2) inform the design of future interventions and the types of needs that might be targeted therein.

METHODS

Overview

We conducted a mixed‐methods analysis of the timing and types of needs patients expressed during a pilot of a 30‐day postdischarge, automated texting program. We used a directed content analysis approach. The pilot ran from January 27 to August 27, 2021 at a single primary care practice within Penn Medicine. This study was determined by the University of Pennsylvania Institutional Review Board to meet criteria for quality improvement and therefore did not require review or informed consent, and follows the Standards for Reporting Qualitative Research (SRQR).

Participants

Patients who were 18 years and older, discharged from an acute care hospitalization, and followed at the specific Penn Medicine primary care practice were eligible. Patients were excluded if they had been admitted to the hospital for (1) planned chemotherapy admissions; (2) scheduled surgical procedures, such as gastric bypass, joint replacements, transurethral resection of the prostate, transplants, and spinal surgery; and (3) obstetrics admissions. Patients were also excluded if they did not have a phone capable of texting or were not able to text in English. All eligible patients were enrolled in the program at the time of the Transitional Care Management (TCM) outreach phone call, whether or not they answered the call. Patients could opt‐out at any time via text message. Further details about patient identification for inclusion can be found in the primary analysis of the pilot program. ¹¹

Program and escalations

Patients enrolled in the program received a standard‐of‐care telephone call from a nurse within 2 business days of being discharged, designed to identify any needs soon after discharge. They were also informed about the automated texting program at the time of this call. Upon enrollment in the program (Day 0) patients received a message introducing the program and providing instructions on how to reach out or opt out. They were asked whether they had an appointment scheduled with their primary care clinician or a specialist within the next 2 weeks, and provided with assistance in doing so if they did not.

The next day (Day 1), patients began receiving check‐in messages on a tapering schedule (see Appendix for full schedule and script of messages). These messages asked whether they needed any help. If a patient responded “No,” no further action was taken.

An escalation was an event in which a patient (a) responded “Yes,” that they needed help, or (b) reached out requesting help outside a check‐in window, which they were told that they could do at any time. In response to any escalation, patients would receive a follow‐up text message asking them to categorize their need (e.g., “I need help with medications”). These responses were then routed to an electronic medical record (EMR) in‐basket, and the practice care manager (a registered nurse) would follow up with a telephone call. Patients who did not respond to the follow‐up message asking them to categorize their needs would receive a reminder prompt; if they did not respond, this was deemed an “incomplete” escalation, and no follow‐up call occurred.

Escalation messages were read during business hours and were responded to within one business day of receipt. Although patients received the text prompt according to the pre‐determined schedule, patients were not limited in the number of escalations during the course of their participation in the pilot study. They were provided clear instructions on how to reach out via text at any time during the course of the program (by texting in HELP or CALL). After a need was addressed via phone, a summary encounter note of the need/s and action/s taken was entered in the EMR by the care manager who handled the escalation.

The automated texting program was developed and managed by Way to Health, a platform developed with support from the National Institutes of Health. ¹⁸ This platform was designed to offer automated technological infrastructure to aid in clinical care and research related to innovative care delivery.

Review of escalations

We conducted a review of the EMR encounter notes documented by the care manager which described the reasons for escalation. Two reviewers (Aiden Ahn and Katherine Honig) conducted a thorough chart review and summarized the content of these notes. An escalation, as described above, was an event in which a patient requested help from the practice; one escalation event could contain multiple needs.

The reason/s for escalations were categorized according to a modified version of the Ideal Transition in Care framework. ¹⁷ We used a directed content analysis strategy ¹⁹ —a deductive approach—because this existing framework is grounded in theory related to the design and role of transitional care programs, and is well suited to provide insight into where existing and future programs should focus their effort. The original framework outlines 10 domains that comprised a successful transition from the hospital to the community setting. Because these domains were originally developed from a slightly different perspective (the health system coordinating the transition, rather than the patient experiencing it), we modified the categories to reflect the needs that patients communicate after discharge. In particular, we combined “Availability, Timeliness, Clarity, & Organization of Information” and “Complete Communication of Information” into a single category; changed “Medication Safety” to “Medication Needs and Reconciliation”; and removed “Discharge Planning.” In addition, we included a “Miscellaneous” category to describe needs that did not fit neatly within the framework, and an “Unknown” category for escalations that could not be categorized at all. The modified categories and definitions (what the category is and what it is not), as developed by the study team (including one of the original authors of the Ideal Transitions framework [R. E. B.]), are outlined in Table 1.

Table 1.

Postdischarge need categories.

Categorization	Category explanation
Monitoring and managing symptoms after discharge	Is: Questions pertaining to symptoms and their treatment. Includes questions about symptoms not related to their original diagnosis (e.g., COVID symptoms).
	Isn't: Questions regarding diagnostic tests, medication/equipment used for self‐management, or symptoms that were experienced during the admission.
Outpatient follow‐up	Is: Scheduling an appointment with the primary care team and questions about what's going to happen during the appointment.
	Isn't: Questions about appointments with members outside of the primary care team.
Medication needs and reconciliation	Is: Questions regarding medication use, refills, prescriptions, and side effects.
	Isn't: Questions regarding usage of medications for patient symptoms or self‐management via dose changes or use of other equipment.
Coordinating care among team members	Is: Scheduling an appointment with providers outside of the primary care team. Questions about whether certain information had been sent to/by any members outside of the primary care team, about what's going to happen during the appointment, and about home care needs.
	Isn't: Questions about appointments or documents related to the primary care team.
Educating patients to promote self‐management	Is: Questions regarding medication dose changes, help with getting and using durable medical equipment (DME), as well as other techniques and precautions that allow for self‐management. Includes preoperative questions.
	Isn't: Questions regarding medication use, refills, prescriptions, or side effects.
Availability, timeliness, clarity, organization, and complete communication of information	Is: Did the patient have the information needed to take care of themselves after discharge, including: questions regarding diagnosis, tests, treatment, and use of medical equipment. This information includes records and information related to the hospitalization.
	Isn't: Questions related to medication changes postdischarge, appointments, or requests for information to be communicated to other members of the care team.
Enlisting help of social and community supports	Is: Letters, notes, and paperwork that are requested outside of the hospitalization, as well as referrals that address social challenges (transportation, meals, qualifying for insurance/coverage).
	Isn't: Requests for notes/other paperwork to be sent to other members of the care team.
Miscellaneous	Is: Documented escalations that involved requests, questions, and information not categorized under categories 1–8.
	Isn't: Unclear notes, unknown reasons for escalations, accidental escalations, and requests for phone number changes or opting out.
Advance care planning	Is: Questions/directions regarding patient's care if they become unable to make decisions on their own, such as end‐of‐life care decisions (DNR, living will, organ/tissue donation) and health care proxies.
	Isn't: Questions about the current treatment plan.
Unknown	Is: Unknown reasons for escalations, accidental escalations, unclear encounter notes, incorrect phone numbers, and patients not picking up.
	Isn't: Miscellaneous reasons for escalations that were clearly documented but not under categories 1–8.

Each need was assigned to one of these categories. When multiple needs were addressed within the same escalation event, each need was assigned its own category. Escalations were reviewed once, with the two reviewers each conducting half of the reviews. During the review process, the reviewers (Aiden Ahn and Katherine Honig) and PI (Eric Bressman) met regularly to discuss any challenges in assigning categories. Where there was uncertainty as to which category a need belonged, this was adjudicated jointly by the entire study team (Aiden Ahn, Katherine Honig, Eric Bressman, and Anna U. Morgan) until there was consensus.

Data collection and analysis

Timing and frequency of escalations were collected from the Way to Health platform. Information related to the needs addressed in escalation phone calls was collected using Penn Medicine's EMRs system (Epic Systems Corporation). Escalation phone calls—when documented—were described in encounter notes by the practice care manager. The time of discharge was also captured in the EMR.

We provide descriptive statistics on the characteristics of all patients in the program, stratified by those who had an escalation versus those who did not. Characteristics include age, sex, race, ethnicity, payer, and risk score. The University of Pennsylvania Health System risk score is an Epic Systems Corporation‐developed and validated point score used to estimate a patient's risk of adverse health events in the next year based on clinical information presented in prior literature. ²⁰ , ²¹

For comparison testing of the characteristics of these two groups we used two sample t‐tests (continuous variables) and Pearson's chi‐squared testing (binary variables). We calculated the frequency and proportion of each need category. In addition, we calculated the time from discharge to each escalation overall and by need category.

All statistical analyses used Stata software, version 16.1 (StataCorp LLC).

RESULTS

Characteristics of patients

A total of 405 individuals were enrolled in the program; the mean (SD) age was 62.7 (16.2); 64.2% were female; 47.4% were Black and 46.4% White; 49.9% had Medicare insurance and 42.7% had commercial insurance (Table 2). Of this population, 178 (44.0%) had at least one escalation. There were no statistically significant differences in any measured characteristics between those who had an escalation and those who did not.

Table 2.

Patient characteristics, stratified by whether they escalated.

		Total	Did not escalate	Escalated	p‐Value
Observations		405	227	178
Age, mean (SD)		62.7 (16.2)	61.6 (16.8)	64.1 (15.3)	.12
Sex, %	Female	260 (64.2)	139 (61.2)	121 (68.0)	.24
	Male	145 (35.8)	88 (38.8)	57 (32.0)
Race, %	Black	192 (47.4)	100 (44.1)	92 (51.7)	.14
	White	188 (46.4)	115 (50.7)	73 (41.0)
	Other/unknowna	25 (6.2)	12 (5.3)	13 (7.3)
Ethnicity, %	Non‐Hispanic or Latino	391 (96.5)	217 (95.6)	174 (97.8)	.24
	Hispanic or Latino	14 (3.5)	10 (4.4)	4 (2.2)
Insurance, %	Commercial	173 (42.7)	107 (47.1)	66 (37.1)	.22
	Medicare	202 (49.9)	106 (46.7)	96 (53.9)
	Medicaid	24 (5.9)	11 (4.8)	13 (7.3)
	Otherb	6 (1.5)	3 (1.3)	3 (1.7)
UPHS risk scorec		4.5 (2.6)	4.3 (2.5)	4.8 (2.6)	.08
Escalations, mean (SD)		0.8 (1.2)	0	1.8 (1.2)	N/A

Abbreviations: SD, standard deviation; UPHS, University of Pennsylvania Health System.

^aIncludes American Indian or Alaska Native, Asian, East Indian, Pacific Islander, and those who self‐described as “Other.”

^bIncludes self‐pay and Department of Defense coverage.

^cSee the Methods section for description of the UPHS risk score.

Escalations

There were a total of 319 escalation events. Across the entire program population, individuals had a mean (SD) of 0.8 (1.2) escalations over the course of the 30‐day period; among those who had at least one event, there was a mean (SD) of 1.8 (1.2) escalations. There were 120 escalations that could not be categorized due to a variety of factors; these included (frequency in parentheses): missing documentation (46), incomplete escalations (36), patients not answering the follow‐up phone calls (20), issue resolved by time of call (8), accidental escalation (8), and patient wanting to unenroll (2).

Of the remaining 199 escalation events, 32 had multiple needs addressed, for a total of 235 needs categorized. The most frequent needs addressed were: (1) Monitoring and managing symptoms after discharge (26.8%), (2) Outpatient follow‐up (20.4%); and (3) Medication needs and reconciliation (15.7%) (Table 3). Example quotes for each category, taken from encounter notes, are presented as well (potential patient identifiers masked or removed).

Table 3.

Frequency of need categories, timing from discharge, and example quotes.

Categorization	Frequency no. (%)	Days since discharge, mean (SD)	Example quotes
Monitoring and managing symptoms after discharge	63 (26.8%)	11.8 (8.3)	“Woke up this morning and glands under neck are very tender… Patient educated to call the office if he experiences fever, throat swelling, difficulty swallowing, or controlling secretions.”
			“Patient wanted to know how long COVID symptoms will last, reports she is still having a slight headache, no worsening SOB.”
			“[Patient] finished Bactrim course this morning. Incisional area still a little red with tender areas…and drainage from area below navel. Patient wanted to know whether this is okay.”
Outpatient follow‐up	48 (20.4%)	10.3 (8.2)	“Call returned. Patient would like to make an appointment. Routed to PSR.”
			“Pt reached and scheduled…for virtual visit w/pcp.”
			“[Patient] wanted to know if all blood work would be drawn [after visit]…Patient reassured lab will be able to see all orders and draw all blood work.”
Medication needs and reconciliation	37 (15.7%)	10.0 (7.3)	“Requested new clopidogrel script to pharmacy.”
			“Pt called with question regarding taking Motrin while on enoxaparin injections. Advised she should avoid NSAIDs until anticoag therapy complete.”
			“Message from patient husband with update on recent hospitalization and question about Keppra side effects.”
Coordinating care among team members	33 (14.0%)	10.9 (6.7)	“Responded to text. Patient needs assistance with scheduling rheumatology. Scheduled in care management encounter.”
			“Patient's son…wanted to know if records from [outside hospital] were received.”
			“Patient would like a second opinion. Asked for assistance to transfer follow‐up care to Penn Medicine.”
Educating patients to promote self‐management	21 (8.9%)	10.1 (8.3)	“She asked for assistance with obtaining new CPAP machine.”
			“[Patient] is still on the Humulin 70/30 pen, but now takes 11 units before breakfast, and 6 units before dinner… Advised daughter to administer patient's morning dose of 11 units, and that patient does not have to eat anything if she does not want to.”
			“Patient asking where he can go for BPPV therapy…Advised on doing home Epley maneuver, demonstrated in clinic.”
Availability, timeliness, clarity, organization, and complete communication of information	15 (6.3%)	11.7 (7.0)	“Patient wanted to know if blood cultures from [outside hospital] had shown any growth. Per Epic blood cultures show no growth after 5 days.”
			“Patient is concerned because she has been on oxygen in the hospital and at home since discharge…and is not sure if she still needs oxygen…Patient educated to certainly wear oxygen when she is short of breath.”
			“Patient's daughter asked about lab results. Labs were reviewed by [provider].”
Enlisting help of social and community supports	12 (5.1%)	11.4 (6.4)	“Patient needs a letter from doctor for court.”
			“Responded to text to receive assistance with FMLA since she did not hear back from surgeon.”
Miscellaneous	6 (2.6%)	12.2 (8.9)	“Patient wanted to make sure shingrix and DTAP vaccines were entered into her medical record.”
			“Patient reports she is currently in the hospital at [outside hospital] and will be transferred to [this hospital]. Will f/u at discharge.”
Advance care planning	0.0%	N/A	N/A

Abbreviations: CPAP, continuous positive airway pressure; PSR, patient services representative; SOB, shortness of breath.

Timing from discharge

Escalations were most frequent within 0–5 days of discharge (30.1%); the frequency of escalations generally tapered off over the 30 days after discharge (Figure 1). Overall, 77.0% of escalations occurred within the first 15 days after discharge. The mean (SD) days from discharge overall was 10.8 (7.9). There were no significant differences in timing from discharge based on the category of need being addressed.

Figure 1.

Frequency of days since discharge. The histogram shows the frequency of when escalations occurred in windows of time after discharge (bins of 5 days). *Escalations may have occurred more than 30 days after discharge if enrollment in the program was delayed relative to discharge.

DISCUSSION

This study is, to our knowledge, the first to assess the types of needs addressed via a postdischarge text‐messaging program. Automated outreach has grown over the last several years as a means to enhance support and promote access after hospital discharge, and this work can help our understanding of how these programs can best be designed and implemented. ²² , ²³ , ²⁴

It is important to understand whether these programs are being used as expected and whether—when embedded within the primary care setting—the needs being addressed are within the clinic's usual purview. A concern with lowering barriers to access to the clinic is the appropriateness and volume of patient communication, an issue that has been more commonly examined in the context of patient portals. ²⁵ , ²⁶ Portal burden has contributed to burnout for some providers and even led some health systems to bill for these interactions. ¹⁶ , ²⁷ In this analysis, we found that 44% of patients had an escalation during the 30‐day program. Across the entire patient population, the average number of escalations over the course of the 30‐day program was 0.8. This was a manageable additional work burden for the practice and, given the design of the program, was seamlessly integrated into the workflow of the clinic's care manager.

In terms of the content of the escalations, the most common needs addressed related to symptoms (26.8%), coordinating follow‐up care (20.4%), and medication issues (15.7%). As the sample quotes demonstrate, there is variation within categories and overlap between categories, and the types of needs range from directly related to the recent admission to more longitudinal issues. On the whole, the needs being addressed appear to be well within the scope of what transitional care programs are designed to address: confusion about medications and therapies (e.g., insulin regimens and newly prescribed home oxygen), continued or recurrent symptoms (e.g., ongoing signs of infection after finishing antibiotics), and challenges with care coordination (e.g., accessing records and getting appointments). There are also a smaller number of interactions that would not normally be triaged by clinical staff (e.g., certain paperwork), which future iterations of this type of program could address.

Our findings are consistent with the types of needs that have been addressed in nontext message‐based transitional care programs, which have typically relied on phone calls. ²⁸ , ²⁹ Most importantly, these align with the challenges that patients are known to face after discharge, the types of problems that the primary care practice is well‐equipped to address, and the intended purpose of transitional care support programs. ¹ , ³⁰ , ³¹ Taken together, these findings support the use of an automated approach to replace the more labor intensive, manual calls that are common practice. In addition, the concentration of needs within three main categories (symptoms, medications, and follow‐up care coordination, which accounted for 63% of the total escalations) suggests the possibility of (a) a narrower, more focused approach and/or (b) inclusion of added resources or content to address these specific domains (e.g., auto‐routing medication questions to pharmacist staff).

The clustering of escalations closer to discharge is not surprising. Some of this had to do with the design of the program, given that the frequency of check‐in messages tapered over time, though patients were able to request help at any time, independent of check‐ins. There was notably no major difference in the types of needs being addressed closer to or further from discharge. Still, this is instructive for the design of future programs which could probably be shortened and still retain most of the benefit. Indeed, studies of postdischarge follow‐up visits with the primary care practice—which have been associated with fewer readmissions—seem to demonstrate a waning effect further out from discharge, suggesting that the critical period for support may be in the first 2 weeks. ³²

Limitations of this study include that it was conducted at a single, urban primary care practice and the needs in this population may not be generalizable to all settings. Patients who did not have a texting‐capable phone or the ability to text in English were excluded from the original study, which again may limit the generalizability of our findings. The categorization of needs did entail some subjectivity, although we attempted to mitigate this through a consensus‐based review process. A substantial proportion of needs could not be categorized due to missing or inadequate documentation, and it is unclear whether this impacted the relative proportions of need categories. In addition, while there were potential differences in the characteristics of patients who escalated versus those who did not, the study was underpowered to detect statistically significant differences.

The strengths of this study include the use of a novel, rich data source from a high‐touch transitional care program designed to proactively identify postdischarge needs. This enabled us to clearly identify escalations, which were time‐stamped in the texting platform; under usual care it is often more difficult to identify postdischarge needs from chart review, unless the notes are labeled in a particular way. We also used an established framework to categorize needs, and ensured there was agreement through a team‐based consensus approach.

CONCLUSION

This analysis demonstrates that patients express similar needs in a text‐messaging postdischarge system as reported in comparable call‐based programs. Three quarters of needs arose within the first 2 weeks of the 30‐day program. The needs identified were concentrated in three areas (symptoms, medications, and follow‐up care coordination) relevant to primary care practice and generally fell within nursing scope of practice. This program can serve as a model for health systems looking to support safer transitions through an operationally efficient approach, and the findings of this analysis can inform future iterations of this type of program.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

See Table A1.

Table A1.

Schedule of outreach.

Practice	Within 2 business days of discharge	Week 1	Week 2	Week 3	Week 4	Day 30
Control	Standard Transitional Care Management (TCM) phone call from the practice	X	X	X	X	X
Intervention	Standard TCM phone call from the practice	Three check‐in messages	Two check‐in messages	One check‐in message	One check‐in message	Closing message
	Enrollment in automated texting platform
	Appointment question

Script of messages:

• −
  Enrollment messages

  • ○
    Hello {name}. The care manager from {PCP's name}'s office enrolled you in this Penn Medicine texting program. This program is for people who have been recently discharged from the hospital.
  • ○
    We will check in regularly to see how you are doing and make sure you get the care you need. We will respond to you within 1 business day. Please note that texting is not 100% secure. Message & data rates may apply.
  • ○
    If you want to speak with us in between check‐ins, text “CALL.” If you need immediate medical help, call 911. If you do not want to receive these messages, text “BYE” at any time.
• −
  Appointment question

  • ○
    Do you have an appointment with you Primary Care doctor or specialist within the next 1–2 weeks?
• −
  Regular check‐in question

  • ○
    Is there anything we can help you with today?

    • ▪
      If patient answers “yes”: Thanks for reaching out. Before we call we'd like to get a little more information. What can we help you with?

      • (A) I don't feel well
      • (B) I need help with my medicines
      • (C) I need help with my appointments
      • (D) I need help at home
      • (E) More than one or something else
• −
  Closing message

  • ○
    You have reached the end of the 30‐day discharge follow‐up program. You will no longer receive messages from us. If you have a new need or a nonurgent medical issue, call your doctor. If you need immediate medical help, call 911.

Ahn A, Morgan AU, Burke RE, et al. Postdischarge needs identified by an automated text messaging program: A mixed‐methods study. J Hosp Med. 2024;19:1138‐1146. 10.1002/jhm.13466