The caller is rarely the patient

In a home care service the phone rings all day. Whoever answers sits alone in the office while the rounds are out.

The obvious assumption is that a client is on the other end. It is almost always wrong.

The callers are: the daughter who wants to know whether her mother has eaten today. The GP about a prescription. The pharmacy about a repeat. The medical supply shop about an appointment. The hospital about a discharge.

Why this decides a caller recognition

A search covering only the client file is technically flawless and recognises the minority of calls in daily work. It finds half — and the rarer half at that.

And it does not report this. The screen says “Unknown”, exactly as for a genuine stranger. The difference between “we do not know them” and “we searched in the wrong place” is invisible from outside.

What a standard achieves here

In healthcare there is good news. FHIR — HL7’s data standard — defines the phone search itself:

The search parameter `phone` is defined there as `telecom.where(system='phone')`. So we would invent nothing and a vendor would document nothing.

And the standard knows the relatives. Besides `Patient` there is `RelatedPerson` for next of kin and `Practitioner` for clinicians — each with their own contact details.

The case reaches far beyond care

Once you look for it, you find it everywhere. In many businesses the caller is systematically not the person in the record:

• Therapy practice: the referring doctor calls, or a relative about an appointment.
• Property management: the tenant calls, the tradesman, the owners’ committee — four roles to one flat.
• Freight: the driver calls, the consignee, the client — three people to one shipment.
• Law firm: the other side calls, a court, another firm.

The same structure everywhere: the software holds the case, but the call comes from a person beside it.

What we built from that

So we do not search only the CRM. Alongside the specialist system the portal queries the company’s Microsoft 365 address book and a phone book of its own that grows from previous calls.

63 %

The gap between 63 and 33 per cent is the whole explanation: frequent callers are more often known. Measuring the recognition rate per number weighs the one-off caller as heavily as the daily one — and thereby fails to describe what happens on the phone.

The point we raise ourselves

A care client’s phone number is health data and thus a special category under Art. 9 GDPR. Technically the process is the same as for a business contact; the weight is different.

So our enquiries to care software houses carry a paragraph on it — not because anyone asked, but because otherwise the question comes from their data protection officer in the second meeting. And then it looks as though we had not thought about it.

What we do: no permanent storage, only a cache (hits 14 days, unsuccessful queries 24 hours), operation and data in Germany, processing on instruction. For every single number a service can give information and delete in the portal — that is built, not planned.