External scoring rules now flag patients for review

随访与健康通知，
按计划自动完成

Digital employees call patients on the follow-up schedule, collecting recovery, medication and scale information item by item and writing results with the patient's own words into the follow-up sheet. Incomplete or abnormal cases go to clinicians for review.

Free trialFree trial5-minute quickstart →5-minute quickstart →
mRS follow-up assessmentDay 90 after discharge
Patient
Wang**
Phone
138****0412
Respondent
Self
  • DeceasedNo
  • Still has symptomsYes
  • Can't do pre-stroke activitiesYes
  • Needs help with own affairsNot stated
  • Needs someone to help walkingYes
  • Needs help dressing or eatingNot stated
  • Needs constant careNot stated
mRS
4
Moderately severe disability
Exact score
Verbatim evidence
「I have to hold on to someone to walk.」
「When I walk on my own my knees hurt.」
「Stairs… can't do them.」
Solution overview

From discharge to long-term care, every milestone reached on time

For hospitals' discharge follow-ups, chronic care and health notices. Once the list is imported, calls are scheduled by milestone and every result is written into the follow-up sheet as structured data.

Post-discharge follow-upScale assessmentsChronic disease managementFollow-up and check-up remindersSatisfaction follow-up
Follow-up cycle · what the digital employee handles
  1. Discharge day
    List import

    Bulk import by discharge date and diagnosis, scheduled automatically

  2. Early
    Symptom and medication check-in

    Checks recovery, medication and symptoms

  3. Mid-term
    Functional scale assessment

    Asks scales such as mRS item by item, outputting scores with verbatim evidence

  4. When due
    Follow-up and check-up reminders

    Reminds patients of follow-up visits and checks, confirming attendance

  5. Ongoing
    Chronic care and satisfaction follow-ups

    Regular follow-ups track changes and run satisfaction surveys

What's included
  • AI Outbound Calling →Calls follow the plan and time slots, retrying unanswered calls at other times
  • Data extraction and external scoringConfigured fields extracted on hang-up, with external scoring rules supported
  • Voice and pacing →A 'gentle, patient' setting suited to elderly patients
  • Data analytics →Completion rate, result distribution and review list
Runs on schedule automatically

Calls go out automatically when due, no scheduling needed

Structured results in the sheet

Scores, extracted fields and verbatim evidence go straight into the follow-up sheet

Clinicians focus on review

Staff only handle incomplete or abnormal cases

Real project
Example: neurology stroke follow-up

Measured follow-up data from a neurology department

Stroke patients are followed up at set times after discharge, using one standard set of functional questions.

1,347calls
follow-up calls
633
patients or family members reached
79s
Avg call duration
7items
Functional questions covered per call
Neurology dept., a tertiary hospital in Beijing · Post-stroke discharge follow-up · Mar–Oct 2026From platform call records, de-identified
Modified Rankin Scale
Example: mRS scale assessment

mRS 0–6, graded on what was actually said

The modified Rankin Scale (mRS) is the most widely used measure of functional outcome after stroke. The digital employee asks seven functional questions one by one, then grades according to the scale.

0
1
2
3
4
5
6
  • 0 No symptoms at all · Same as before onset
  • 1 No significant disability · Has symptoms but can carry out daily activities
  • 2 Mild disability · Some activities limited, can live independently
  • 3 Moderate disability · Needs some help, can walk independently
  • 4 Moderately severe disability · Unable to walk independently, needs help with daily life
  • 5 Severe disability · Bedridden, needs constant care
  • 6 Dead
Where four real calls landed on the scale
「Same as before the stroke. Dressing, eating, stairs — all fine.」
—— The patient
0
1
2
3
4
5
6
「He's a bit hard of hearing… goes out for a walk every morning.」
—— Family member
0
123
4
5
6
「Have to hold on to someone to walk, can't do stairs.」
—— The patient
0
1
2
3
4
5
6
「Can't move the left side, can't even sit in a wheelchair, we can't manage at home.」
—— Family member
0
1
2
3
4
5
6
Sufficient information: exact score Insufficient information: score range, sent for human review
Three outcomes

Score when possible; say so when not

Not every call yields complete information. When it falls short, a range is given and sent for human review, with no guesswork scoring.

Exact score
5

The family says the patient's left limbs can't move and the patient can't sit in a wheelchair. They can't provide care at home, and the patient has been moved to a nursing facility.

Bedridden and needing constant care: meets the criteria for a score of 5.

Written directly to the follow-up sheet
Score range · Needs review
1–3

The family says the patient's hearing has declined and responses are slow, but the patient goes for a walk every morning. Key functional items weren't clarified, so a 1, 2 or 3 can't be distinguished.

Information still needed
  • · Whether dressing, eating and toileting need assistance
  • · Whether shopping, cooking and managing medication need assistance
  • · How much slowed responses affect daily independence
Sent to a follow-up nurse to ask again
Not assessable
—

The call only got as far as identity confirmation, with no functional information obtained.

Missing items noted in the result: whether the patient is alive, residual symptoms, walking and self-care.

No forced score; reason noted
Privacy

De-identified before scoring

Below is an excerpt from a real follow-up call, with personal information replaced by placeholders.

Transcript · de-identifiedExcerpt
AIHello, this is the Hospital patient follow-up team. Am I speaking with Name?
PatientYes, it's me.
AIIs anything different now compared with before the stroke?
PatientIt's the same as before the stroke.
AIAnd dressing, eating, stairs — can you manage those yourself?
PatientAll fine.
ScoreNo symptoms at all0
  1. 01De-identified before scoringNames, forms of address, phone numbers and hospital names are replaced with placeholders, so the scoring model only sees the conversation and cannot identify the patient.
  2. 02Encrypted at restPhone numbers and transcripts are stored encrypted and masked in the UI according to role.
  3. 03Traceable scoresEach score carries the model version, rule version and an input digest, so the same conversation reproduces the same result.
  4. 04Quotes as evidenceEvery score includes the patient's or family's own words, so nurses don't need to replay the full recording when reviewing.
Customer story

The follow-up center's setup

What the follow-up center says
「90-day functional scores used to take several nurses a week of calls. Now we import the list and the scores and review list are ready the next day.」
—— Neurology follow-up center, a tertiary hospital in Beijing
Lists
Discharged patient lists imported in batches, with diagnosis and discharge date
Timing
Separate projects for days 7, 30 and 90 after discharge, dialed automatically when due
Script
Asks through mRS item by item: symptoms, usual activities, walking, self-care, care needs
Score
External scoring rules connected; call records show the score, review flag and evidence
Speed
Interruption and recognition set to "Gentle · Wait longer" so elderly patients can finish before the reply
Redial
If unanswered, redialed the next day in the morning or afternoon

Learn more

The product capabilities behind this solution, how to configure them, and more industry use cases.

Each solution combines several product capabilities. Explore what each one does and how well it works.

Ready to get started?