For patients, staying in touch with a healthcare provider is rarely convenient. Whether it’s a new diagnosis or a routine dental check-up, a test result they’re anxiously waiting on or a prescription that needs renewing, it happens in between work, family and everything else, and often at a moment when they’re already worried.

That’s what customer experience in healthcare really comes down to. Nobody expects medical admin to be enjoyable. But healthcare organizations can make it much easier for patients to hear from them, reply to them and get what they need without having to chase anyone.

Many of the most frustrating moments happen around the care itself. A reminder goes to a phone number the patient stopped using years ago, and the missed appointment is recorded as a no-show, even though the message never arrived. Most guides to healthcare CX cover empathy, wait times and feedback loops, and skip a more basic question: did the message actually reach the patient? This guide covers what healthcare CX means, how to measure it, and the communication layer underneath it, where many of the quiet failures (and the easiest wins) are found.

Customer experience in healthcare is every impression a person forms across their entire relationship with a healthcare organization. Clinical and administrative. Digital and in person. Before the visit, during it, and for months afterward, usually while they’re on hold with billing.

That last part is a joke. Mostly. But it points at something real. When people inside healthcare hear “CX,” they tend to picture the exam room. Patients are living the whole thing: the search, the booking form, the parking garage, the results portal password they’ve reset four times.

These three terms get used interchangeably, but each one describes something different.

Patient satisfaction is a rating of a single encounter. How was your visit on Tuesday? Four stars, lovely nurse, the magazines were from 2019.

Patient experience covers the care journey: how well the care team communicated, whether pain was managed, whether the discharge instructions made any sense.

Customer experience is the whole relationship. It includes everything above, plus scheduling, insurance, reminders, billing, follow-up and the dozens of small administrative moments in between. Most of that is non-clinical, and most of the friction lives there.

The distinction matters because a hospital can have a brilliant cardiology team and a billing process that makes people cry in their cars. Those two things are measured by different instruments, owned by different teams, and fixed (or not fixed) on very different timelines. Look at only one of them and you’ll think you’re doing fine.

Not every touchpoint is equal. These are the ones that shape how people actually feel about a provider, roughly in the order they happen:

Finding and choosing a provider
Booking and rescheduling
Pre-visit instructions and intake forms
Appointment reminders
The wait
The clinical encounter
Results delivery
Discharge and follow-up
Billing and insurance
Ongoing engagement, like recalls, refills and preventive care nudges

Now count the clinical ones. You’ll get two, maybe three if you’re feeling generous. Administrative touchpoints outnumber clinical ones roughly three to one, and almost all of them are communication. Most of what a patient remembers about you happened in a text, an email, a phone queue or a portal. Which is exactly where things tend to go sideways.

Most healthcare CX problems start in the gaps between systems, quietly, where nobody is looking. Here are the four failures that come up again and again, and not one of them shows up on a satisfaction survey.

This is the failure from the start of this guide, and it’s the most common one. Reminders, results notifications and recall messages fail silently. No bounce alert lands on the patient access team’s desk. No survey catches it, because the patient who never got the message isn’t around to fill one in. The missed appointment gets logged as patient behavior, and everyone moves on.

The causes are mundane, which is exactly why they get ignored:

Phone numbers nobody has verified in years
Texts sent to landlines (it happens more often than anyone likes to admit)
Carriers filtering messages from unregistered senders
The wrong channel for the market, like SMS-only outreach in a country where everyone lives on WhatsApp

The takeaway: reach rate and delivery rate are CX metrics. They belong on the same dashboard as NPS.

The good news is that this is a solvable engineering problem. Dialog Health, which runs patient engagement for health systems and health plans on Infobip infrastructure, reaches more than 96% of patients, and 99% on emergency broadcasts.

A patient explains their situation to a chatbot. Then to an agent. Then to a nurse line. Each time from the top, like the world’s most stressful game of telephone.

The usual cause is architecture. The bot, the contact center, the campaign tool and the patient record are often four systems from four vendors, and nobody ever introduced them. Context can’t travel when there’s nowhere for it to go.

When more gets resolved inside the conversation instead of bounced around, the numbers move. Apollo 24|7, one of India’s largest digital health platforms, cut calls to its customer care center by 20% by resolving more inside the conversation itself.

This one is unglamorous, so naturally it’s where things get expensive.

Patient communication is regulated outbound messaging. That means opt-in requirements, preference management and, in the US, carrier registration through 10DLC and short codes, plus HIPAA obligations around what a message can contain. (We’re describing the operational side here, not giving legal advice. Your compliance team gets the final word, as they should.)

When consent lives in one system and sending happens in another, one of two things happens. Either you over-message and patients stop trusting you, or you get nervous, under-message, and patients miss care they needed. Neither is a great look. Dialog Health runs 10DLC and short code registration and number lookup on Infobip infrastructure for Fortune 500 health plans. It’s boring plumbing. It’s also the plumbing that makes everything else work.

Annual and quarterly surveys report on patients who have already left. You know this. Everyone knows this.

What gets talked about less is that a better signal is already sitting right there. Every conversation a patient has with you, on any channel, carries intent and sentiment in real time, and nobody has to fill in a form. The patient who types “this is the third time I’ve asked” is giving you feedback. You just need a system that’s listening.

Now for the part that gets budgets approved.

Each of these instruments sees something different, and each one has a blind spot.

CustomerGauge’s NPS benchmark research puts the average healthcare NPS at 65, the highest of any sector it studied. It’s a useful reference point, with one caveat: the figure reflects self-reported B2B programs rather than published patient scores. CSAT is quick and useful too, but it only hears from the people who answer. And CES, in a sector where so much is hard, is wildly underrated.

Notice the pattern? Every one of these measures the experience of people you successfully reached. None of them can see the communication layer failing.

Here’s your answer to “why should we fund this?”

In the US, patient experience isn’t only a reputation issue. HCAHPS results feed into CMS quality programs for hospitals, including value-based purchasing, so experience scores can influence what a hospital gets paid. On the payer side, CAHPS member experience measures feed into Medicare Advantage Star Ratings, and Star Ratings affect quality bonus payments and how plans compete for members.

The exact weightings shift by program and by year, so we won’t pretend to hand you a number. The mechanism is what matters. When patients have a worse experience, the organization can feel it in revenue as well as in online reviews. That moves the communication layer out of the “patient engagement, nice to have” folder and onto the balance sheet.

These almost never make it onto a CX dashboard. They should:

Message delivery rate
Patient reach rate
Response rate to appointment confirmations
No-show rate
Containment rate on automated conversations
First contact resolution
Average speed to answer

For a sense of what good looks like: Dialog Health sees a 55% response rate on appointment confirmations and around 40% average response across campaigns. Apollo 24|7 delivers 95% of its omnichannel messages, and 67% of its chatbot sessions are actively engaged. You can only report numbers like these if you can see the delivery path end to end, which is why most CX reporting quietly leaves them out.

Okay. Enough diagnosis. Here’s what to actually do, in the order that pays off fastest.

Start here. It has the biggest measurable return of anything on this list, and the sequence is simple even if the execution takes some care:

Step 1: Confirm the number is reachable before the first send. A number lookup catches landlines and dead numbers before they turn into no-shows.
Step 2: Capture channel preference at booking. Just ask. People will tell you.
Step 3: Send the reminder on the preferred channel with a two-way reply path, so “Reply C to confirm, R to reschedule” actually does something.
Step 4: Escalate to a second channel only when the first goes unanswered.

Automated reminders on a patient’s preferred channel can reduce no-shows by up to 50%. Dialog Health sees 55% of patients respond to appointment confirmations, and one recall campaign drove a 15% lift in mammogram appointments. Behind that 15% are real people who finally booked the screening they’d been putting off, because the nudge reached them at a moment when they could act on it.

Channel choice comes down to two questions. What do people in this market actually use? And what does this particular message need to do?

Verified sender identity matters more in healthcare than almost anywhere else. People are rightly suspicious of any text claiming to be from their doctor, so the verified brand name and logo that RCS provides do a lot of heavy lifting. And where WhatsApp is the app people open fifty times a day, that’s where the reminder should be.

Then write down the fallback, and make it automatic. Plenty of organizations have a channel list. Far fewer have a fallback plan, and the fallback plan is the bit that saves the appointment.

Example fallback: Send the reminder on WhatsApp. If it isn’t delivered within a set window, send an SMS. If the SMS goes unanswered, trigger a voice call.

Apollo 24|7 runs WhatsApp, RCS, voice, Instagram, Messenger and Live Chat side by side, and delivers 95% of its messages across all of them.

Consent works best when it’s designed in from the start, so nobody has to run a cleanup project six months later.

Capture opt-in at booking or registration, not after the fact.
Store consent and channel preferences against one unified patient profile, so every system reads the same record.
Make opting out as simple as a single reply, with no phone call or portal login needed.
Keep an auditable trail.

In the US, this also means registering senders through 10DLC or short codes before you send at volume. That registration has a lead time, and teams routinely discover it the week before the flu shot campaign is supposed to go out. Consider this your friendly heads-up. (Again, operational guidance, not legal advice.)

This is where it gets genuinely interesting.

A scripted bot greets everyone the same way. An AI agent working from a unified profile already knows this patient has an appointment on Thursday, prefers Spanish, called last week about a billing question and hasn’t confirmed yet. So it opens with that, instead of making them start over.

Inside AgentOS, Infobip’s agentic AI platform, that’s the job of the AI chatbot builder and AI Agents, both drawing on the Conversational CDP for the patient profile. An agent is only as good as what it knows walking in. Megi Health Platform, which supports people living with chronic conditions, reached 86% CSAT and cut data collection time by 65% with conversational support that gathers structured information ahead of clinical review.

If you want the full tour of what chatbots can do in this space, we’ve covered it in our roundup of healthcare AI chatbot examples.

Automation gets judged by its worst handoff. When a conversation moves from an AI agent to a human, this needs to travel with it:

The full conversation history
Verified identity
Sentiment (is this person calm, frustrated, scared?)
The reason for the escalation
Any open task

In AgentOS, the Cloud Contact Center is the module that receives all of that context, so the human agent picks up mid-sentence instead of from scratch.

Healthcare adds one rule that isn’t optional: anything symptom-related needs an explicit escalation path. Automation should route those conversations to a clinician and leave the advice to them. A chatbot can book the appointment, check the insurance details and send directions to the clinic. It should never be the one deciding whether chest pain is serious.

Apollo 24|7 shows what it looks like to get both halves right: 20% fewer care center calls and 67% engaged chatbot sessions. Better containment and better escalation, at the same time.

Quarterly surveys tell you what went wrong long after the patient has moved on. A faster loop looks like this:

Send a short feedback request right after the interaction, on the channel where it happened.
Route unhappy responses to a named owner with a set response time.
Feed the result back into the patient profile so the next interaction reflects it.

And remember, the conversation is already a signal. The Conversational CDP within AgentOS reads intent and sentiment from conversations as they happen, so you’re not waiting three months for a survey to tell you something a patient already told you on Tuesday.

The last section was about what to do. This one is about who owns it, and how it keeps running after the project team moves on to the next shiny thing.

Healthcare CX is split across patient access, marketing, clinical operations, revenue cycle and IT. The communication layer, the thing connecting all of them to the patient, usually belongs to nobody. It’s the group project where everyone assumed someone else was doing the slides.

A working operating model needs:

One accountable owner
A shared measurement set that includes experience scores and operational metrics like reach and delivery
A regular review cadence that ends in actual decisions
One system of record for patient contact preferences

The hard truth is that fragmented tools enforce fragmented ownership. If five teams run five platforms, you’ll have five versions of the patient, and none of them will be quite right.

Payers and providers often get lumped together, even though their CX looks quite different.

Providers manage access, care delivery and billing, and they’re measured on HCAHPS and patient satisfaction. Payers manage member onboarding, annual enrollment, claims status, prior authorization and plan changes, and they’re measured on Star Ratings and CAHPS.

Their communication looks different too. Payer messaging is higher volume, more seasonal (hello, open enrollment) and more tightly regulated. A health plan might send more messages in October than in the entire summer combined.

Still, plenty of health systems run their own plans, and plenty of plans run their own clinics. Whatever platform you choose has to serve both sides without making anyone pick.

Trends are only useful if they change what you do on Monday morning. These five do.

AI agents are starting to complete tasks for patients, such as rescheduling an appointment directly. To do that, the agent needs write access to scheduling systems and a clear escalation policy, on top of a knowledge base.

Care gap closure, recall campaigns and adherence prompts get triggered straight from the patient profile. Outbound volume climbs fast, which makes consent and deliverability the real constraint.

RCS brings verified sender identity and branded, interactive confirmations to a category where trust is the whole problem. Infobip’s Messaging Trends 2026 data shows RCS growing from 1% to 13% of conversational channel traffic on the platform between 2023 and 2025.

Sentiment and intent get read continuously from conversations, rather than sampled once a quarter from whoever bothered to reply.

As patient information becomes easier to share across care settings, patients reasonably expect the reminder to know what the clinical record knows. “Why are you asking me this, I told my doctor yesterday” is going to become a very common complaint.

Enough theory. Here’s what this looks like when real healthcare organizations actually do it.

Apollo 24|7 runs patient engagement across WhatsApp, RCS, voice, Instagram, Messenger and Live Chat, with booking journeys that happen right inside the messaging channel. No app download, no “please call during business hours.” The results: a 95% omnichannel message delivery rate, 49% more bookings, 20% fewer customer care center calls, 67% engaged chatbot sessions and a 72% increase in average revenue per order.

Dialog Health powers two-way SMS for Fortune 500 health plans, large health systems and ambulatory surgery centers, including carrier registration and number validation. It reaches more than 96% of patients, gets a 55% response rate on appointment confirmations, lifted mammogram appointments by 15% with a single recall campaign, and hits 99% reach on emergency broadcasts.

One more number worth pausing on: 90% reach on a multi-language campaign. Language access is often a legal obligation, and it’s also just basic respect. Sending people messages in a language they can read is CX at its most fundamental.

Megi uses conversational patient support to collect structured health data before clinical review, so care teams spend their time on care instead of chasing forms. The result: 86% CSAT and a 65% reduction in data collection time.

Diagnostics provider Biolab automated patient communication and booking, and saw 13.2% more patients alongside a 17% reduction in operating costs. Growth and savings in the same sentence. It does happen.

Everything in this guide comes back to one idea: healthcare CX breaks at the seams between systems. So the fix is fewer seams.

AgentOS is Infobip’s agentic AI platform, and the modules that matter most for healthcare all live inside it:

The Customer Engagement Platform and Journey Orchestration for patient journeys, reminders and campaigns
The Conversational CDP for a single, unified patient profile
The AI chatbot builder and AI Agents for automated conversations
The Cloud Contact Center for human escalation with full context

So the bot-to-agent handoff and the campaign-to-clinic handoff happen inside one system instead of across four contracts.

And here’s the part that’s hard for anyone else to say. Those modules run on Infobip’s own carrier infrastructure, with direct carrier connections across 190+ countries.

The same platform that decides what to send also delivers it, and reports back on whether it arrived.

So when a reminder fails, you know about it. And far more often, it doesn’t fail at all.

Healthcare is always going to show up at inconvenient moments. Nobody gets to schedule their worries. But the experience a patient remembers rarely starts in the exam room. It starts with a message, and whether that message reached them.

Get that part right and the hard moments get a little lighter. The result arrives in a format people can read. The reminder lands on the app they actually check. The question gets answered without three transfers and a fresh retelling. For a patient who is already worried, having one less thing to chase matters more than any waiting room redesign.

If you’d like to see how this works for your organization, explore AgentOS or talk to our team about patient engagement.