I'll say it up front: when the next big outage hits, our hospitals will mostly be fine. Fine is relative, and I'll get to their bad days, but hospitals are among the best-prepared parts of the health system, and when disaster hits they tend to be first in line for help. The main federal program for health care readiness is the Hospital Preparedness Program, which ASPR calls the primary source of federal funding for health care preparedness and response. That money flows through states and healthcare coalitions, and hospitals are one of the four core members of every coalition, alongside public health, emergency management and EMS. During COVID, MACPAC found that as of late November 2020, virtually all hospitals had received Provider Relief Fund money, while only 18.4% of potentially eligible Medicaid and CHIP providers outside Medicare had, a group MACPAC said included many pediatric, behavioral health and home- and community-based providers. The places I worry about are the doctor's office down the street, the independent pharmacy, the free clinic that only opens on Saturdays, and the daughter who manages her dad's prescriptions from her phone between shifts.

I sat down this afternoon to write about last week's Microsoft outage, and my phone told me Apple was down. iCloud sign-ins and the App Store were failing, and user reports added Apple Pay, Apple Music and even device activation. By early afternoon Apple's status page had iCloud sign-in listed as slow or unavailable, and plenty of people had been signed out of their accounts and couldn't get back in. As of this afternoon, Apple hasn't said why.

Heck of a coincidence. Here's how the last six days went:

  • Thursday, Nov. 12. Microsoft 365 went down around midday Eastern and wasn't fully restored until Saturday. At US health systems it took out email, voicemail and pagers. ChristianaCare's faxes and pagers went down with its email, so it told patients with urgent needs to call their practices directly. University of Iowa Health Care pointed people to MyChart. Atlantic Health said the impact reached patient care and supply chain. In the UK, the National Pharmacy Association said pharmacies faced delays in getting medications to people.
  • Friday, Nov. 13. At about 3:30 p.m. Eastern the FAA had what it calls a technical problem and halted departures nationwide until 8:15. Roughly 12,500 flights were canceled that day, though the FAA notes that not all of them were caused by the stop, and it says life-saving flights kept flying. Around the same time, outdated content on some National Weather Service webpages fed rumors of a nationwide radar outage, which NWS had to knock down. That evening Workday, the HR and payroll platform used by more than 11,500 organizations, locked people out, including some who were trying to make their open enrollment elections.
  • Sunday, Nov. 15. An incident in AWS's US-EAST-1 region, more or less in my backyard here in Northern Virginia, disrupted 11 services for close to four hours, mostly sign-in, sign-up and network connectivity.
  • Monday, Nov. 16. United had a system outage that took about an hour and a half to resolve and delayed more than 400 flights. The same afternoon Chase said a software bug was slowing transactions, some involving other banks.
  • Today. Apple. And Yahoo Finance, citing CNN, reported that a US national security communications network went down, with no sign so far of a cyberattack.

For anyone tempted to connect the dots: so far, nothing ties these together. Microsoft says there's no indication of malicious activity. The FAA says it has no indication of a cyberattack. United said there was no indication its problem was cybersecurity-related, and Chase blamed a software bug. AWS's updates didn't attribute its incident to anything malicious, the reporting on the security network found no evidence of an attack, and Apple hasn't named a cause. The one confirmed attack I've seen reported this week was a different animal: someone exploited a vulnerability in Optimizely's Experimentation tool and injected malicious code into online tools that nearly 200 organizations use. Mostly, though, this looks like an ordinary week now, which is the unsettling part.

Two systems on the same wires

When people in our field say "health infrastructure," we usually mean the formal system: hospitals, clinics, pharmacies, EMS and 911, health departments, and the EHRs, clearinghouses and payment rails underneath. It has regulators, plans and people whose job is continuity.

A lot of health happens outside it. Family caregivers, community health workers, food pantries, faith communities, mutual aid groups, and the phone in someone's pocket that holds a patient portal, a telehealth app, pharmacy refills and a benefits balance. I'd call that the informal system. Most of it has never been asked to write a continuity plan. Some of it, like a free clinic that still keeps paper charts, may actually ride out an IT outage better than the rest of us.

Both systems now sit on the same handful of cloud platforms, payment networks and phones. Last week showed us the hospital side. For everyone else, we have to look back to 2024.

The hospitals held (mostly)

Hospitals, along with health centers, rural health clinics, nursing homes and a long list of other Medicare- and Medicaid-certified providers, fall under CMS's emergency preparedness rule. It's an all-hazards rule, not a vendor-outage rule, but it does require an emergency plan, policies and procedures, a communication plan with primary and alternate means of communication (the parallel rule covers health centers and rural health clinics), and training and exercises. Federal guidance on EHR contingency planning goes further and recommends paper forms for downtime and staff trained and tested on downtime procedures. Add accreditation and in-house emergency managers, and hospitals start from a stronger position than anyone else in health care.

The early read on the Microsoft outage, and it is early, is that clinical systems mostly stayed up. Tom Walsh of tw-Security told BankInfoSecurity he knew of no clinical system outages or interruptions to patient care. That isn't the whole picture. Atlantic Health said the outage reached patient care. During Change Healthcare, 74% of hospitals reported a direct hit to patient care. When ransomware hit Ascension in May 2024, its EHR, MyChart and some phone systems went down and several hospitals diverted ambulances. A UC San Diego study found network disruptions at 759 of the 2,232 US hospitals it studied after the CrowdStrike failure, though CrowdStrike disputes it. And the front door to emergency care has its own weak spots: a botched AT&T network change in February 2024 blocked more than 25,000 calls to 911.

So "fine" is relative. But hospitals have the plans, the people and the practice to work through a bad day, and that's more than most of the system can say.

Primary care is another story

Go one level down and it thins out fast. I mean physician practices, independent pharmacies and free and charitable clinics, which is where a lot of people get routine care, refills and help managing chronic disease. Health centers and rural health clinics are covered by the CMS rule. Independent physician offices, community pharmacies and free clinics aren't on the list. HIPAA does require covered entities, which includes most practices and pharmacies, to keep a contingency plan with data backup, disaster recovery and emergency mode operations. But testing that plan is "addressable," which means a practice can document a reason not to, and nobody surveys for it the way CMS surveys hospitals. HHS proposed tightening the Security Rule in January 2025, and it still isn't final. And a free clinic that doesn't bill insurance electronically may not be a covered entity at all.

We've already seen what happens when this layer loses a shared system. The February 2024 Change Healthcare cyberattack hit a company that touches one in every three patient records. The AMA ran an informal survey in late March and early April 2024, a convenience sample of more than 1,400 respondents, about 78% of them from practices with 10 or fewer physicians. Eighty percent reported lost revenue from unpaid claims, 55% used personal funds to cover practice expenses, 44% couldn't buy supplies and 31% couldn't make payroll. The AMA said small practices looked especially hard hit, and that isn't a niche group: 47.4% of physicians worked in practices that size in 2024.

Pharmacies had it rough too. The National Community Pharmacists Association, which represents 19,400 independent pharmacies, told the Senate Finance Committee that pharmacies relying on Change's claims switch were cut off from all payers for weeks and that fee-for-service Medicaid was down in six states. Pharmacists who fronted emergency supplies without being able to bill were putting payroll, wholesaler payments and rent at risk.

Health centers sit in between. They're covered by the rule, but they carry a huge load: HRSA-funded health centers saw 32.4 million patients in 2024, and about 90% of those with known income were at or below 200% of the federal poverty guideline. Further out are the free and charitable clinics. According to the association's latest data report, as reported by Mississippi Today, its roughly 1,400 affiliated clinics provided 8.2 million patient visits in 2025, up from 6 million the year before. Project HOPE, where I used to work, reported last year that most run on annual budgets under $250,000 and that about 83% of the people they serve are uninsured. Some, like the Jackson Free Clinic in Mississippi, open one day a week with volunteer staff.

All of these places run on the same commercial email, identity, e-prescribing and claims systems as everyone else, usually with a fraction of the IT support. My bet is that a hospital rides out a bad IT day far better than a three-doctor practice or a free clinic, which can't easily make up lost visits and whose patients have the least room to come back later.

Outside the clinic walls

Sixty-three million Americans are family caregivers, close to 1 in 4 adults, according to AARP and the National Alliance for Caregiving. More than half handle complex medical or nursing tasks (injections, wound care, managing medications), only about one in five have had formal training for it, and six in 10 also hold jobs. When a patient portal, a pharmacy app or a payroll system goes down, the caregiver makes the extra calls and trips and takes the time off.

Sixteen percent of US adults are smartphone dependent, meaning they have a smartphone but no home broadband, according to Pew Research Center's 2025 survey. That's 34% in households making under $30,000 and 4% above $100,000, and 28% for Hispanic adults and 19% for Black adults. For those folks, "email's down, use MyChart" assumes a lot. Today, plenty of people who sign in with an Apple account got a small taste of what losing that front door feels like.

Friday's Workday outage left employees unable to clock in or out, see pay information, request time off or file benefits claims in the middle of open enrollment. On Monday, Chase said a software bug in on-premise hardware systems was slowing transactions, though customers kept access to their accounts and funds. If you're salaried, that's an irritation. If you're hourly and can't clock in, it can be your rent. The Federal Reserve's latest household survey found 4 in 10 adults earning under $50,000 couldn't cover even a $100 emergency expense with only their savings. SNAP benefits ride on EBT systems run by a small number of vendors, each covering several states, and when one went down in August 2022, the Food Research & Action Center warned that some families would have no way to buy essential groceries.

Friday's NWS episode showed how fast outdated content from isolated webpages can turn into a rumor of a nationwide radar outage. If one app is how you get your warnings, your clinic updates and your benefits notices, you're stuck when it goes dark, or when it's wrong.

Why this keeps happening

Where there's any explanation so far, it points to a change or maintenance work. Microsoft blames an earlier network capacity upgrade. A source told Fox News the Chase bug came with a software release that morning. Workday reportedly hit a problem during testing. And the FAA's early findings point to legacy equipment that became unstable when data flooded its network, at two sites, one of them where a contractor had been installing a new router. None of those investigations is finished, and AWS, United and Apple haven't explained theirs. The common thread is concentration. Looking at nine years of publicly reported outages, the Uptime Institute finds that third-party providers like cloud, telecom and colocation companies account for about two-thirds. Microsoft alone says more than 430 million people use its 365 apps.

What I'd actually do

Leaving the cloud isn't realistic or necessary. Five things I'd push for:

Put your vendors in your hazard vulnerability analysis. Email platform, identity provider, EHR host, clearinghouse, paging vendor, payment processor. Ask each one which region you're running in, what fails over where, and how fast they'll tell you when something breaks.

Build communications that fail for different reasons than your daily stack. During the Microsoft outage, Health-ISAC members used text messages, Signal, WhatsApp and mass notification tools, and Errol Weiss of Health-ISAC stressed that those channels need to be approved ahead of time, with privacy and security safeguards. A backup that shares an identity provider, cloud platform or network path with your primary isn't a backup.

Exercise degraded operations and rumor control together. Run it with the charge nurses and the practice managers, not just IT. Test how you reach staff, partners and patients with email and chat gone, and how fast you can correct bad information.

Bring primary care and the informal system into coalition planning, and help pay for it. Healthcare coalitions and health departments should have free and charitable clinics, independent pharmacies, community health workers, food banks, faith groups and caregiver organizations at the table for technology outages, not only hurricanes. A three-doctor practice can't afford two internet providers and a second clearinghouse on its own. Coalitions, health departments and funders can broker shared backup contracts and group purchasing, and support pharmacy status tracking like Healthcare Ready's Rx Open, which maps open and closed pharmacies during disasters. (Full disclosure: I used to lead Healthcare Ready.) Whether the money comes from preparedness grants or philanthropy, someone has to put it in a budget. HHS's voluntary cybersecurity performance goals start with basic incident planning, with drills at the next tier, and they're a reasonable starting point for small providers.

Write it into contracts and grants, and measure who got shut out. Uptime notes that the shift toward third-party providers may require changes to contracts and finances, so put notification timelines and continuity expectations in procurement language. And after-action reports should count declined EBT transactions, missed telehealth visits, unfilled prescriptions and 911 calls that didn't connect, along with where those people live.

A redundancy checklist

A few places to start in the formal system:

  • Connectivity. Two internet providers with physically separate entry points, plus cellular or satellite failover at critical sites.
  • Sign-in. Break-glass admin accounts that don't depend on your single sign-on, kept offline and tested.
  • Clinical data. Read-only downtime copies of what you need to keep seeing patients (the census and medication administration records for a hospital; the day's schedule, problem lists, allergies and active prescriptions for a practice), printable without the network.
  • Cloud. Spread critical applications across availability zones and, for anything that truly can't go down, across regions. AWS's advice during Sunday's incident, to rebuild load balancers across multiple Availability Zones, was aimed at customers stuck in one zone. A regional sign-in failure needs a regional fallback.
  • Claims. A second clearinghouse you've actually tested. After Change, 67% of hospitals said switching was difficult or very difficult. Picture a three-doctor practice trying it mid-crisis.
  • Payroll. A manual timekeeping fallback and a plan to pay staff off the last good cycle.

And in the informal system:

  • A printed medication list, allergy list, pharmacy and clinic numbers and insurance information for every patient and caregiver. Health centers, free clinics and community health workers can hand these out and update them at each visit.
  • Where coverage and state law allow, a few extra days of critical medications.
  • SNAP retailers can take manual vouchers when the system is down, but only with the forms on hand and a processor that can clear them. Food banks and health departments can check that the corner stores their clients use are set up.
  • Wireless Emergency Alerts and NOAA Weather Radio, which the NWS calls two of the most effective ways to get warnings directly, paired with trusted local messengers.
  • Libraries, health centers, houses of worship and community centers with backup power and connectivity, identified and equipped now, as places to charge a phone, check a portal or call a pharmacy.

Same lesson, new hazard

We learned this with natural disasters. Following nearly 3,500 families from 1999 to 2013, Junia Howell and James Elliott found that disaster damage widened wealth gaps: in counties with more damage, white households and homeowners gained wealth compared with similar households in counties with little damage, while Black and Latino households and renters lost ground. Outages are starting to work the same way. Hospitals have downtime binders. The caregiver juggling three prescriptions on a phone doesn't.

There will be another big one, probably soon. The hospitals will be fine, relatively speaking. What I want to know is whether the rest of us have planned for everyone else, especially the people who rely on the informal system because the formal one was never really built for them.