5.1 Networks and Communications

5.1 Networks and Communications

Quick Answer: Domain I B.2 treats the network as a clinical safety system. CPHIMS items expect you to separate clinical, biomedical, and guest traffic; reserve bandwidth for imaging and voice with quality of service (QoS); and design downtime communications so care continues when the EHR or WAN is down.

Why networks are a Domain I competency

A CPHIMS professional is not asked to configure a switch port. You are asked to decide whether radiology can share a guest SSID, whether a tele-stroke cart is allowed to compete with staff video, and how the hospital talks when the EHR is dark. Network failures show up as delayed images, dropped nurse-call, silent ADT, and clinicians inventing workarounds. Task B.2 tests whether you can design for safety, segmentation, and continuity, not whether you can recite OSI layers from a vendor slide.

/practice/cphimsPractice questions with detailed explanations

LAN, WAN, and WLAN — use the terms as operators use them

A local area network (LAN) is the campus fabric: access switches, distribution, core, and the wired drops that still carry most life-safety and imaging traffic. A wide area network (WAN) is the set of circuits and tunnels that stitch campuses, clinics, the data center, and cloud regions. A wireless LAN (WLAN) is RF access that must be designed, not hoped for: channel plans, controller capacity, and SSID-to-VLAN mapping.

Do not treat “we have Wi-Fi” as a design. A 300-bed hospital can have excellent public wireless and still fail a code cart because the clinical SSID is oversubscribed on the same airtime as visitor streaming. Wired remains the default for PACS diagnostic stations, nurse-call controllers, and many biomedical devices that cannot tolerate RF jitter. Wireless is how mobile clinicians move; it is not a license to put every modality on the same radio as a waiting-room tablet.

Campus versus clinic. A hospital LAN is dense, multi-VLAN, and QoS-heavy. A five-provider clinic may be a single switch stack plus a managed WLAN, but the same rules apply: guest traffic stays off clinical VLANs; EHR and payment terminals do not share an open SSID; the WAN circuit is sized for images and visits, not only email. Rural clinics on a single commodity circuit need an explicit statement of what happens when that circuit dies—because it will.

Segmentation: clinical, biomedical, and guest are not one network

Segmentation is the practice of putting different trust and traffic classes on different VLANs, SSIDs, firewalls, and routing domains so a compromise or a flood in one class cannot starve or observe another.

Minimum healthcare pattern:

1. Clinical / EHR VLAN — staff workstations, thin clients, medication cabinets, barcode scanners that talk to the EHR.
2. Biomedical / IoMT VLAN — infusion pumps, monitors, imaging modalities, smart beds. These devices are often unpatchable and should not browse the internet.
3. Voice / real-time VLAN — VoIP, nurse call, RTLS, some video.
4. Guest / BYOD VLAN — visitors, patients, and personal phones. Internet only, no route to EHR, PACS, or biomedical subnets.
5. Management / jump VLAN — infrastructure admin, out-of-band, privileged jump hosts.

Exam trap: putting guest Wi-Fi on the clinical VLAN is a design failure. Convenience is not a compensating control. A visitor device that can ARP-spoof, scan, or saturate the same broadcast domain as the EHR is an availability and confidentiality incident waiting for an OCR letter. A captive portal authenticates a human; it does not isolate Layer-2 attacks or guest floods.

Micro-segmentation (host-level policy) is a refinement, not a substitute for these coarse VLANs. CPHIMS stems usually test the coarse failure: guest and clinical mixed, or pumps sitting on the same flat network as laptops that browse the web. Firewalls between these zones default-deny. Clinical systems initiate only the flows they need (EHR to interface engine, PACS to archive, pump gateway to a vendor cloud only if a BAA and risk analysis support it). Guest traffic never originates a session into clinical space.

QoS: imaging and voice lose if everything is “best effort”

Quality of Service (QoS) marks and queues packets so delay-sensitive and clinically time-critical traffic is not crushed by bulk transfers. Healthcare has two classes that fail visibly:

• Voice and nurse-call need low latency and low jitter. Dropped packets become unintelligible speech or a silent code.
• Diagnostic imaging (PACS, ultrasound, CT reconstructions, cine loops) needs throughput and predictability. A radiologist waiting on a 400 MB study is a length-of-stay problem, not an IT inconvenience.

QoS is not “make the EHR faster.” It is protect the classes that break care when the pipe is full. Typical markings: voice in an expedited-forwarding queue, interactive EHR in a low-latency data class, imaging in a dedicated assured-forwarding class, guest and software updates in scavenger or best-effort. WAN circuits need the same markings honored by the carrier or SD-WAN policy; campus QoS that dies at the MPLS edge is theater.

Telehealth bandwidth is a QoS and capacity problem, not a “we bought a video app” problem. A tele-ICU or tele-stroke cart needs reserved uplink, wired fallback, and a tested codec. Shared clinic Wi-Fi that also carries guest streaming will freeze a neurologist’s exam. Size for concurrent video rooms, not average daily users. Store-and-forward dermatology is delay-tolerant; real-time stroke is not. Document the minimum megabits and jitter the clinical program accepted, then monitor them. If the program cannot state a number, the infrastructure team cannot defend a circuit.

Downtime communications are part of the network design

When the EHR, DNS, or WAN fails, the network team’s job is not only restore. Care continues on downtime procedures: paper MAR, downtime EHR viewers, analog or cellular backup for codes, overhead page, and a pre-agreed command channel.

Design elements CPHIMS expects you to name:

• Out-of-band path — cellular, radio, or a separate ISP so the incident channel does not ride the failed WAN.
• Read-only downtime viewers — locally cached ADT, meds, allergies, and recent results that do not depend on the primary data center.
• Role-based comms tree — who pages the house supervisor, who tells the ED to stop electronic triage, who notifies affiliated clinics.
• Reconciliation path after restore — how paper orders re-enter the legal record.

A dual WAN with no tested failover is not redundancy. Quarterly tabletops plus a live fail-over of a clinic or a unit are how you prove availability. Tie this to Chapter 12’s business-continuity design; here the point is that communications architecture is a continuity control.

Scenarios and traps

• Trap: “Wi-Fi is Wi-Fi.” Clinical, biomedical, and guest SSIDs must map to different VLANs and firewall policies.
• Trap: “Guest on clinical VLAN is fine if we use a captive portal.” A portal authenticates humans; it does not isolate Layer-2 attacks or guest floods.
• Trap: “QoS is only for VoIP.” Imaging and real-time telehealth fail without reserved capacity.
• Trap: “The WAN is the vendor’s problem.” You still own latency SLAs, encryption, and the downtime script when the circuit dies.
• Trap: “Telehealth works if the app launches.” Measure concurrent video bandwidth and wired fallback for stroke and ICU programs.

A 12-site ambulatory network that backhauls every image over a single 50 Mbps circuit will look “up” in a dashboard and still cancel afternoon MRI reads. Capacity and class, not just uptime, are the CPHIMS measures.