Administrators at multi-department hospitals tend to describe the same slow-building problem in different words. Nobody can say exactly who still has administrator rights on the scheduling platform. Contractors who finished a project last spring can still reach the file server. Approvals are given in a hurry and forgotten. The symptoms look unrelated, but one underlying cause explains most of them: elevated access is handed out freely and almost never taken back. A reference catalogue gives the operational side of this problem a name, AccessOps, and what follows treats it as a diagnosis exercise, covering symptoms, measurements, likely causes and fixes, with the attention kept on administration and systems rather than anything clinical.
Symptoms that point to an AccessOps problem
Begin with what staff actually say. “I think she still has the master login, but I would have to check.” “We gave the vendor a password for the migration.” “Nobody remembers who approved that.” Each remark quietly admits that five basic facts about an elevated permission are missing: the holder, the scope, the purpose, the expiry and the approver.
The catalogue behind this article, the Ops Atlas run by xOps School, calls those five items the minimum record behind any grant. A site that cannot complete them for its most sensitive systems has found its first measurable gap.
Measure before you change anything
An evidence-led approach starts with a few counts, none of which needs special software. The Atlas sketches a small set of indicators, and the ones that suit a hospital office are these:
- Standing privileged accounts. How many people hold powerful roles around the clock on the most sensitive systems?
- Coverage of time-limited access. What share of elevated work happens through a request with an end date, as opposed to a permanent role?
- Time to get access. The median wait and the slower ninetieth-percentile wait, because a process that feels slow gets bypassed.
- Utilisation. Of the access that was approved, how much was ever used?
- Emergency-account review lag. After an emergency login, how quickly does someone look at what happened? The Atlas suggests within one business day.
Its illustrative month is useful as a model of what the numbers can look like. Standing powerful accounts on the top two tiers fell from thirty-four to nine. Two of those nine were emergency accounts, leaving seven that needed an explanation. Of one hundred twenty approvals, thirty were never used, a quarter of the total. None of these figures describes your organisation, but they show the shape of a before-and-after record that management can read at a glance.
Collecting the numbers is a modest job. A spreadsheet fed by the identity directory export and the ticket queue is enough for a first pass, and the exercise itself often exposes the problem. Expect awkward moments: a department head who insists a role is vital for a colleague who changed jobs two years ago, or an approval trail that simply does not exist. Record these without blame. The point of the baseline is to compare next month against this month, so consistency of method matters more than precision of any single figure.
Likely causes, ranked
Once the counts exist, look for the cause behind them. Four turn up repeatedly.
Approvals that approve everything
When every request goes to a line manager, and managers click through dozens a week, the reasons typed into the form become generic. “Needed for work” tells a reviewer nothing. The Atlas names this as an anti-pattern: blanket manager approval encourages blanket justifications. The remedy is to route requests by sensitivity, so that routine ones are approved by rule or by the on-call coordinator and only the rare, powerful ones need a human who reads the reason.
Reviews carrying too much weight
Many organisations run an annual access review and assume the problem is handled. A review is a useful safety net, and the CIS controls call for one at least yearly, but a permanent permission checked once a year is still a permanent permission. The CISA maturity model, as the Atlas summarises it, actually places “standing access plus periodic review” at its most traditional stage. So if reviews are your only control, you have found a cause.
Tools without rules
A purchase order for a vault or session broker feels like progress. The Atlas is blunt that having such a product does not mean access is managed. If nobody has decided who may request what, how long it lasts and who approves, the product merely stores the same sprawl in a nicer cabinet.
Emergency accounts used for convenience
An emergency login that gets used because it is quicker than asking is a backdoor with a good reputation. The page quotes a cloud provider making the same observation. The fix is to alarm on every use and to review each one the next working day.
Remedies, in a sensible order
- Inventory the top tiers. List the systems whose misuse would hurt operations most: scheduling administration, billing configuration, identity directory, backups, the mail domain.
- Put an end date on every grant. The Atlas’s default for ordinary requests is ninety to one hundred eighty days. Anything sensitive gets hours rather than months.
- Move to “eligible” instead of “active”. People who occasionally need a powerful role request it with a reason and receive it for the length of the task.
- Keep two emergency accounts and drill them. Verify that they work, that they are protected with strong sign-in, and that using them sets off a notification.
- Add a second approver for the very top tier. The Atlas pairs that with session recording and a daily reconciliation in its ninety-day plan.
Notice what is absent: nobody is asked to stop doing their job. The aim is to shrink the amount of idle power, and the evidence from the Atlas’s worked month is that most of it is idle.
Limits of the fix
Honesty matters in a diagnosis. Time-limited access does not stop someone who has already stolen a live session, and an automated approver is itself a high-value target that needs the same protection as the systems it guards. The Atlas also points out that an ended grant and an ended session are different things: in its lab, a remote session continued for twenty seconds after its certificate lapsed, and cloud permission changes may take minutes to propagate. A remedy plan that ignores this will look finished on paper and leak in practice.
Finally, remember the name is an editorial label. As the Atlas page describing AccessOps itself acknowledges, no formal standard uses the term, and several unrelated products carry it. Judge the discipline by its component practices, which are long established.
Where this meets the rest of hospital operations
Better access hygiene is rarely the headline of any hospital story, yet it sits under everything patients and families read about. Our article on how families decide when persistent back pain calls for a specialist centre describes a journey that passes through referrals, records and appointments, each handled in systems somebody must administer carefully. A similar thread runs through our overview of the usual reasons behind back pain, where scheduling and information flow matter as much as the topic itself.
If your own audit finds more questions than answers, that is a normal result for a first pass. Fix the top tier first, measure again in a month and repeat. The measurements, not the tooling, are what tell you whether the diagnosis was right.
