Written by Alex Harb - Lena Health CEO
When I meet with healthcare executives, I often hear the question, "which engagement channel should be the primary engagement strategy?" The tendency is usually towards the digital front door. It's what feels most modern and appealing.
The problem is that a decade of data tells a different story. It turns out that the phone performs best above all channels. Digital apps and portals get abysmal engagement while call centers continue to grow. The reason is that the phone works. But why is that?
We dove into the data to understand why the phone remains so effective, and we outline how to scale it with AI.
Digital Outreach Misses Large Segments of the Population
Apps, portals, and email underperform across key patient populations, especially among older adults, lower-income groups, and individuals without consistent access to high-speed data.
Health apps have seen strong adoption among younger patients, but not among older adults. Only 28% of adults aged 50 to 80 report using a health app, and 78% abandon it within a week if they’re not properly onboarded. Push notifications fare even worse, with an average open rate of just 4% across healthcare apps.
Patient portals perform slightly better but still miss a wide swath of patients. While 78% of older adults report having a portal account, only 45% log in during a given month. In most systems, that means the majority of portal messages go unread.
Email remains the default for many health systems, but it rarely delivers meaningful engagement. In healthcare, the average open rate is just 21.5%, and click-through rates fall below 3%. For non-urgent updates or educational content, email may suffice. For any task that requires timely action, it consistently underperforms.
Texting offers better visibility. Ninety-eight percent of texts are read, often within minutes. But text messages only go so far. They work best for reminders and binary responses. When patients need to resolve a barrier or ask a question, texting stalls.
The phone still delivers the broadest reach. While only 19% of Americans answer calls from unknown numbers, 94% say they would respond if the caller ID clearly showed the clinic name. Phone ownership now spans socioeconomic status and geography. Seventy-nine percent of adults earning under $30,000 own a smartphone, and almost all have a working mobile or landline phone. In rural communities, where broadband access is often unreliable, 71% of adults still own smartphones, reinforcing the phone’s unique position.
The phone remains the only channel with consistent reach across all age groups, income levels, and geographic regions.
Digital Handles Simple Tasks. The Phone Handles What’s Complex.
Even when a message reaches the patient, the next barrier is whether the channel can support the action required. A refill reminder doesn’t carry the same complexity as scheduling a specialist referral. Some tasks need a live conversation. Others need a click. When health systems apply the same outreach method across every workflow, engagement drops, and staff are left doing redundant follow-up.
Appointment scheduling is a clear example of a channel-task mismatch. While online tools are available, many patients still prefer to speak with someone. A plurality of older adults say they would rather schedule by phone. SMS can help with reminders and confirmations, but as a primary outreach tool for scheduling, it often fails to meet expectations. Patients who receive a text to book care may still choose to call, which can create delays and require additional staff intervention.
Referrals introduce more complexity. Forty-six percent of older adults prefer to coordinate referrals by phone. These tasks often require context, clarification, or troubleshooting to ensure effective completion. When messages about referrals are sent only by portal or email, they frequently go unanswered. Without a direct conversation, referrals stall, and the care gaps persist.
Medication refills are better suited to digital automation, but only when patients are already engaged in their care. Fifty-one percent of Medicare-eligible adults prefer to refill their prescriptions via a portal, but the other half don’t access their portal consistently. Another 33% still call to refill medications. Texting offers improvement at the margins. In one study, interactive SMS prompts improved refill rates by just 14 percentage points among Medicare Advantage members.
In reality, most healthcare scenarios have too many edge cases for a typical digital app. From the patient's perspective, it's easier to just "do it over the phone", despite the long call queues and long holds, instead of tapping around in app that may or may not have the option.
How to Automate Phone Outreach Without Losing Quality
Phone outreach works, but it's expensive and a massive operational challenge: salaries keep growing, half of the team churns within the year, staff require three months of training, and teams are chronically overstaffed or understaffed due to volume peaks.
We've found that AI is perfectly placed to automate outbound calls to engage patients successfully. Below are some key components to successfully scaling the phone:
1. Automate the call attempts
For certain workflows, the entire call can be handled by a voice agent. This includes structured tasks with clear decision trees, such as scheduling an appointment, consenting a patient in a care management program, or completing a screening questionnaire.
Voice AI agents can:
Follow a call script
Be flexible to the random turns in a conversation
Flag patient needs or barriers identified on the call
Follow clear protocol in scenarios like the mention of an urgent clinical symptom
We've found that AI often outperforms humans because it can follow a script with high fidelity while being flexible to changes, and it can confidently navigate challenging questions or skeptical patients.
2. Automate the conversation itself
For certain workflows, the entire call can be handled by a voice agent. This includes structured tasks with clear decision trees, such as scheduling an appointment, confirming interest in a program, or completing an intake checklist.
Voice AI agents could:
Ask and respond to patient questions
Navigate multi-turn conversations
Clarify needs and match them to the right next step
Escalate to a human if the issue is unclear or sensitive
These agents extend staff capacity without reducing quality. They also respond instantly, without wait times or limited hours.
The most efficient systems combine both strategies. They use automation to reduce waste, then reserve human time for the highest-complexity and highest-risk cases. It’s not about removing people, but about letting people work to the top of their license.
3. Evaluate Call Performance at Scale
Agent-led calls generate a level of insight that manual outreach typically doesn't. Every call is an opportunity to assess the transcript in order to measure how the call went. Below are some data points we've built into our system to evaluate call performance:
System Health and Delivery Metrics
Call drops: Are calls completing, or ending early due to network issues?
Audio quality: Was the agent clearly understood by the patient?
Caller ID and anti-spam registration: Are calls being flagged, blocked, or labeled generically?
Answer rates by time and cadence: Are patients picking up during certain windows or attempts?
Conversation Flow and Friction Points
Introduction response: Does the patient hang up immediately, continue normally, or ask skeptical questions?
Abandonment: Where in the call are patients dropping off?
Signals of confusion: Do patients repeat questions, hesitate, or shift sentiment?
Emotional tone and engagement: Does the patient sound confused, cooperative, skeptical, or rushed?
Call Goals to Evaluate the Agent
Each call should have a set of goals to achieve. That could be as simple as ensuring each patient is greeted with a warm and friendly demeanor, or as complex as completing a social determinants of health screening.
Each agent should have a list of targets, or to-dos, to complete on the call. In our system, we've implemented an AI evaluator for agent goals that can be customized to each agent.
This then allows us to offer care teams a dashboard so they can see how their AI is doing as they handle thousands of calls per month. Without automated goal evaluation, it would be impossible to review each call for quality.
The KPIs that Matter Most
Tracking outcomes at the goal level is just one part of the picture. To improve performance, teams need visibility into how patients respond throughout the call, from the moment the phone rings to the final action taken. That’s where agent-level KPIs come in. These metrics highlight which parts of the interaction work, where drop-off occurs, and how to fine-tune the experience for each population.
Reach rate: How many patients are we able to successfully reach with the phone? This often depends on combining multi-attempt outreach, leaving great voicemails, and complementing call attempts with SMS-text.
Intro acceptance: After the agent introduces itself and explains the reason for calling, does the patient continue to engage or hang up? The average human call center gets 80% intro acceptance. We've found we can match or exceed this rate with AI, but performance depends on how the introduction is designed. This is the most critical part of the call.
Conversation completion: Did the patient stay on through the full exchange? This metric measures the effectiveness of the AI in navigating the conversation.
Primary goal completion: Each call usually has one singular goal. A solution should be able to measure how often the AI is achieving this primary goal on each call.
Conclusion: The Phone Still Reaches the Most Patients—and Voice Automation Makes It Scalable
The outreach challenge isn’t just about choosing the right channel. It’s about helping patients complete the right action.
Digital tools often fall short when they treat outreach as a one-way broadcast. Manual staff calls can’t scale when teams spend half their day leaving voicemails. And chasing patients with the same message on five platforms doesn’t fix the problem if no one responds.
The phone still reaches the broadest patient population. But its full potential is only realized when it’s paired with automation, not to remove the human touch, but to reserve it for where it matters most.
By treating agent-led outreach as a precision workflow—one that can be measured, improved, and scaled—health systems move beyond messaging and toward resolution.