Mapping the structural failure in post-discharge psychiatric care coordination, and designing tools to close it.
The highest-risk moment in psychiatric care is the first 48 hours after discharge, when inpatient care has ended and outpatient care hasn't started. Through my research into mental health hospital operations, I kept arriving at the same question: who actually owns that window, and what happens when nobody does?
I came into this project expecting to find a training problem, or a resource problem. What I found instead was a jurisdictional one. The 48-hour window after acute psychiatric discharge has no formal owner. The inpatient team's responsibility ends at the door. The outpatient team's responsibility doesn't begin until the first appointment. The case manager fills the gap — informally, invisibly, and without institutional support.
I mapped the case manager role across five domains: clinical care, admin and documentation, internal coordination, crisis management, and external liaison. The picture that emerged was someone carrying maximum cross-system responsibility with minimum formal authority. Twenty distinct responsibilities. No system designed to hold them together.
I used two complementary tools to trace the problem from symptom to root. The fishbone diagram spread the causes across six categories: information gaps, discharge process failures, workforce limits, policy barriers, patient vulnerability, and system fragmentation. Each branch felt familiar from clinical literature. But the 5 Whys drill pushed further.
The deepest root wasn't clinical. It was institutional. The inpatient and outpatient systems were never designed to complement each other — they were built for different stages of care, optimised for different goals, and funded around different units of activity. When something goes wrong in the gap between them, the harm lands on the patient. But the system records it as a readmission, not a coordination failure. The loop never closes, so no institutional signal fires.
The risk analysis reframes case manager exposure as a systems design question, not a performance one. I separated protocol-triggered decisions (where the system fires a prompt) from person-dependent ones, where the case manager has to remember without scaffolding. The higher the person-dependence, the higher the structural risk.
The decision tree maps every decision point in the post-discharge window in full, and shows exactly where the void opens. The key finding: the system cannot distinguish a stable patient from one who is deteriorating and unreachable. A logged call attempt reads the same either way. The EHR records process compliance, not patient state.
Five personas. Two journey maps. One service blueprint. This phase was about translating the structural analysis into lived experience: not as illustration, but as evidence. Marcus Tran is discharged on a Tuesday. He takes the bus home alone. He has a follow-up appointment in six days that he doesn't know how to get to. He has a phone number on a piece of paper that he's supposed to call if things get bad, but no one told him what "bad enough" looks like.
Nadia Kowalski (case manager, caseload of 34) logs a contact attempt for Marcus at 10 AM. He didn't answer. The EHR records this as a completed task. Nadia knows something is off but can't act on it within the system. The service blueprint places all of this in a single view: how each layer's invisible failure makes the next one worse. At 72 hours, the EHR shows Marcus as compliant. In his room, he is deteriorating. The only clinical-quality observation belongs to his sister, who was never briefed on what to watch for.
The journey maps below follow the same 72-hour window through two entirely different vantage points. I built both to focus on the structural gaps between staff and patient sides.
Nadia's map shows how coordination work becomes invisible under caseload pressure including the calls that get logged as attempts, the decisions made without system support, the moment where she knows something is off but has no mechanism to act.
Marcus's map shows the same window as someone who has just left a ward and has no framework for what recovery is supposed to feel like. The gap between what Nadia intends and what Marcus experiences is where patients are lost.
The service blueprint below places both of the staff and patient experiences in a single view. Alongside the full backstage experience including the institutional rules, the EHR logic, and the funding structures. It's the artifact that shows why individual fixes don't work: every gap below the surface makes the one above it structurally inevitable.
The Transition Void Monitor is a staff-facing coordination dashboard that surfaces the 48-hour window as an explicit, urgency-ranked object. It makes Nadia's caseload legible at a glance: who is inside the void, who has no contact documented, who has missed appointments, who needs escalation. Right now, that picture lives in her head, assembled manually from spreadsheet rows every morning. The dashboard replaces that labour with structure.
I designed three role-differentiated views because role-scoped design is what determines whether a clinical tool gets used. Nadia sees her full aftercare queue. Dr. Reuben Asante (Consultant) sees only the escalated cases requiring clinical authorisation. Sandra Cheung (General Manager) sees team-level metrics and pattern flags, never individual patient names.
Bridge is the frontstage complement to the staff dashboard. It's designed for a phone, in plain language, for a person whose cognitive capacity may be reduced. Not a clinical record system. Not a medication tracker built for compliance. A tool that answers the three questions Marcus actually has the morning after discharge: What am I supposed to do? When is my appointment? Who do I call?
The daily check-in is the critical piece. When Marcus marks that he's struggling, that signal routes to Nadia's dashboard within two hours. He doesn't need to understand the system is watching — he just needs to feel like someone is there. The two tools together create the structural scaffold the case manager was previously trying to hold alone.
Designing the right thing isn't enough. A dashboard that looks good in a demo but doesn't survive week three of a real ward is just a more expensive version of the problem it was supposed to solve. So the last phase of this project was working out the adoption arc: who introduces the tool, how training runs, what resistance looks like, and how you know it's working.
The plan uses a People-Process-Technology framework. The PPT structure isn't arbitrary — it maps directly onto the three failure modes I had already documented. People: trust in clinical tools runs on peer credibility, not IT announcements. Process: you can't rip out a spreadsheet on day one. Technology: friction at the access layer kills adoption before it starts. Each layer of the plan addresses the corresponding failure mode from the research.