Case Study · Hurricane · 2017
September 20, 2017. Category 4. Maria made landfall in Puerto Rico and destroyed 80% of the island's power grid. 2,975 people died — most not from wind or flooding, but from the cascading collapse of medical services during the longest power outage in U.S. history. The storm was catastrophic. But the grid had been failing for years before Maria arrived. The hurricane was the test. The infrastructure had already been failing it.
Puerto Rico · September 20, 2017
Three months before Hurricane Maria made landfall, the Puerto Rico Electric Power Authority — PREPA — filed for bankruptcy. The utility that powered an island of 3.4 million people had accumulated $9 billion in debt. The debt crisis had produced years of deferred maintenance: wooden transmission poles that should have been replaced, distribution lines left vulnerable to tree encroachment, generating plants that hadn't received financing for upgrades, a grid that engineers had described as fragile and deteriorated long before any storm was in the forecast. Three months later, on September 20, 2017, Maria came ashore near Yabucoa as a high-end Category 4 hurricane and tested that grid. The Watchers' analysis documents what happened: wooden poles splintered, transmission towers toppled, substations flooded. Approximately 80% of the island's transmission and distribution infrastructure was destroyed. All 3.4 million residents lost power simultaneously.
Maria was the second major hurricane to hit Puerto Rico in two weeks — Irma had passed close to the island on September 7, leaving widespread outages before power was restored. Maria was significantly more direct and more destructive. For weeks after landfall, NASA satellite imagery showed Puerto Rico as nearly entirely dark from space. Rural municipalities remained blacked out for over six months. The complete restoration of power to all customers took 328 days — the longest blackout in U.S. history. And during those 328 days, people died. Not primarily from wind or flooding. From the absence of electricity.
Sep 20, 2017
Landfall
2,975
Excess Deaths
328 days
To Full Restoration
$90B
Damage
80%
Grid Destroyed
The initial official death toll from Puerto Rican authorities was 64. An independent study commissioned by the Puerto Rican government and conducted by George Washington University found 2,975 excess deaths attributable to Maria and its aftermath. The gap between 64 and 2,975 is not primarily a counting controversy — it reflects the nature of what killed people. Most deaths during Maria were not caused by the storm directly. They were caused by the weeks and months without electricity that followed: hospital generators running continuously and breaking down; dialysis clinics forced to ration treatment; refrigerated medications spoiling; nursing homes without air conditioning growing dangerously hot. The Watchers' analysis is direct: these failures turned a natural hazard into a mass-casualty event. PREPA's $9 billion debt and years of deferred maintenance built the conditions for 2,975 deaths long before the hurricane appeared on a forecast track.
The Science
Think of aging electrical infrastructure like a suspension bridge that's been poorly maintained: each individual wire and beam still holds under normal conditions, so the problem is invisible until a test arrives. Puerto Rico's grid operated on that principle for years. The PREPA debt crisis had reduced maintenance budgets for an already aging system. Power lines were not cleared of vegetation. Wooden poles were not replaced on their required cycle. Generating plants were not upgraded. The system continued to deliver electricity in normal conditions — the kind of conditions that never expose what deferred maintenance has done. Maria was the test. The NPR analysis documents PREPA's assessment: "The system is old and it's undermaintained. Maria brought it to light."
The RAND assessment of Puerto Rico's post-Maria recovery documents that "damage to critical infrastructure resulted in cascading failures of the lifeline systems of energy, transportation, communications, water supply, and wastewater treatment." Electricity is the system on which every other system depends. Water treatment plants require electricity to pump and treat water. Hospitals require it for operating rooms, life-support equipment, and medication refrigeration. Fuel pumps require it to dispense gasoline and diesel. Communications infrastructure requires it. When a grid fails for 11 months rather than a few days, the cascading effects reach every other lifeline system in sequence. The Watchers documents the specific mechanisms: dialysis treatment rationed, refrigerated medicines spoiled, nursing homes without air conditioning overheating, hospitals relying on generators that eventually broke down.
The gap between the official death count of 64 and the GWU-estimated 2,975 excess deaths reflects a methodological distinction: direct deaths (those caused by the storm's immediate physical effects) vs. excess mortality (the statistical difference between expected deaths in a given period and actual deaths, attributed to the disaster's conditions). For a short-duration, acute disaster, direct deaths capture most of the toll. For a long-duration infrastructure collapse — months without power, healthcare, clean water — the excess mortality measure is more accurate. The NBC News analysis of Maria is specific: most of the 2,975 deaths have been attributed to the lack of electricity and its effects on medical and other services. An additional 514 Puerto Ricans who left the island are estimated to have died on the mainland due to displacement effects.
Timeline
01
2006–June 2017: Puerto Rico enters a decade of economic decline. Population drops as working-age residents emigrate. Tax revenues fall. Infrastructure maintenance budgets shrink. PREPA defers maintenance on an aging grid. June 2017: PREPA files for bankruptcy, citing $9 billion in debt. The grid is described by engineers as fragile and significantly deteriorated. Three months later, the test arrives.
02
September 20, 2017, 6:15 AM local: Maria makes landfall near Yabucoa as a high-end Category 4. Wooden poles splinter. Transmission towers topple. Substations flood. 80% of the transmission and distribution grid is destroyed. All 3.4 million residents lose power. The island goes dark. NASA satellites tracking nighttime light show Puerto Rico as nearly entirely black from space. The longest blackout in U.S. history begins.
03
September 20 – March 2018: Hospitals run generators for weeks. Generators break down. Dialysis clinics ration treatment. Refrigerated medications spoil. Nursing homes without air conditioning overheat in September's heat. Two months after landfall, the majority of 3.4 million people are still without power. Police escort fuel convoys to prevent theft. Rural municipalities remain blacked out for six months. 2,975 excess deaths accumulate during this period.
04
August 2018: Day 328. Power restored to all customers — 11 months after Maria. GWU study: 2,975 excess deaths. Congress allocates $71B in recovery funding; 72% hasn't reached communities by 2022. FEMA makes $9.5B available to rebuild the grid — largest ever public infrastructure project. NBC News 5-year anniversary: grid still fragile, outages still frequent. The deferred maintenance problem that pre-dated Maria has not been fully resolved.
Human Decisions
What eventually worked
The NPR account of the immediate post-storm period documents that the New York Power Authority sent a team under Governor Cuomo to assess conditions and help reconstruct the grid — one of the more substantive early external assistance efforts. Multiple states eventually contributed personnel and equipment to grid reconstruction. These state-level responses, organized laterally rather than through federal channels, reflected the scale of the crisis exceeding normal federal response capacity and timelines for U.S. territories.
The independent study commissioned by Puerto Rico's government and conducted by George Washington University produced the 2,975 excess mortality figure that revised the official death count from 64. This kind of independent academic assessment — establishing excess mortality by comparing actual deaths to expected deaths in a baseline period — is the methodologically correct approach for long-duration disasters where death is caused by systemic failures rather than direct storm impacts. The study's existence and findings were essential to a proper accounting of what Maria cost in human lives.
What failed
The RAND assessment of Puerto Rico's pre-storm conditions is comprehensive: a decade of economic decline and a debt crisis produced "a lack of investment in its physical and natural infrastructure" across power, transportation, water, and communications. PREPA's $9 billion debt meant maintenance deferred across the system year after year. The ABC News analysis documents that this produced a grid that engineers knew was old and undermaintained — a condition visible to anyone who looked, but that produced no change in maintenance investment because the economic capacity to address it didn't exist.
The NBC News five-year retrospective documents that Congress allocated at least $71 billion in recovery and reconstruction funds for Puerto Rico, of which $62 billion were made available. The IRP recovery analysis found that approximately 72% of those available funds had not yet reached local communities by 2022 — five years after the storm. Permanent reconstruction work didn't begin until late 2020, three years after landfall. The structural reasons for this — bureaucratic processes, the complexity of managing recovery in a territory rather than a state, pre-existing governance challenges — document a system that was effective at appropriating funds but slow at delivering them to the people and infrastructure they were meant for.
The Watchers analysis documents that hospitals "struggled to keep operating as backup generators ran continuously for weeks, often breaking down." This isn't a simple equipment failure — it's the consequence of chronic underfunding of the backup power infrastructure that hospitals are required to maintain but that budget pressure leads administrators to minimize. The nursing homes without adequate air conditioning, the dialysis clinics rationing treatment, the refrigerated medications that spoiled — each represents a failure of institutional preparedness in facilities that serve the most medically vulnerable populations, where the consequences of power loss are not inconvenience but death.
The cascade lesson
Maria's 2,975 deaths were overwhelmingly caused not by the storm itself but by the eleven months without reliable electricity that followed, which cascaded into medical failures, medicine spoilage, water system failures, and the systematic inability to sustain life-critical services for a population with high rates of chronic illness and medical dependency. The infrastructure that failed had been known to be failing before September 20, 2017. PREPA's bankruptcy, the documented deferred maintenance, the engineers' descriptions of an aging grid — all of this was known. The hurricane didn't create a crisis. It turned a known, documented, pre-existing infrastructure vulnerability into a mass-casualty event. For individual households, the lesson is about medical and power preparedness for long-duration outages, not just 72-hour ones: backup medications, power for medical devices, and the community connections that allow vulnerable members to get help when formal systems are overwhelmed.
What Changed
In 2022, FEMA made $9.5 billion available for Puerto Rico to rebuild its power grid — described in congressional testimony as the largest ever public infrastructure project. The NBC News five-year retrospective documents that as of 2022, only 40 grid reconstruction projects had been approved under this funding. The scale of the project — rebuilding a severely deteriorated grid across an island with complex terrain — has made progress slower than the urgency of the need. The NBC News assessment found that despite the commitment, Puerto Rico's grid remained fragile five years after Maria, with residents losing electricity constantly, before Hurricane Fiona hit the island in 2022 and knocked out power islandwide again.
The gap between Puerto Rico's initial official death count (64) and the GWU study's finding (2,975) became a nationally visible case study in how official death counts from long-duration disasters systematically undercount fatalities from indirect causes. The GWU methodology — comparing actual deaths in a post-disaster period against a baseline of expected deaths — has since been applied to other disasters and is increasingly recognized as the appropriate measure for events where death results from cascading systems failures rather than direct physical causes.
What You Can Do Now
Maria's casualties were primarily medical. The five actions below address the specific vulnerability her death count documents: what happens to people who depend on electricity for medical care when the power stays off for months.
Insulin, certain biologics, and other medications require refrigeration to remain effective. During a power outage, an uninsulated refrigerator keeps food safely cold for approximately 4 hours; a full freezer, 48 hours. For medications, the window may be shorter. Know which medications in your household require refrigeration, know their temperature tolerance, and have a plan: a cooler with ice, a neighbor with a generator, a pharmacy with backup power. Maria's 2,975 deaths include people who ran out of insulin and other medications that spoiled when the grid failed. This is preventable with advance planning.
Medical resilience guideCPAP machines, oxygen concentrators, home dialysis equipment, and other powered medical devices require electricity continuously or at scheduled intervals. Contact your utility and register as a medical baseline customer — most utilities maintain priority restoration lists for medically dependent customers. Have a generator or battery backup sized for your device. Have a written plan for where you will go if a multi-day outage threatens your ability to use your device. Do this before hurricane season in hurricane-prone areas.
Medical device preparednessMaria's blackout lasted 328 days. Even setting aside the extreme duration, the first-month lesson is that supply chains for medications — dependent on functional roads, refrigeration, and distribution infrastructure — can be disrupted for months, not days, after a major hurricane makes landfall in a remote or island location. Most insurance plans allow 90-day fills for maintenance medications. Request 90-day fills for all long-term medications. Keep them filled, not let down to a 5-day supply before refilling. The 90-day buffer is the difference between a supply disruption and a medical emergency.
Long-term medical resilienceThe IRP recovery analysis of Maria documents that "older adults, children, people living with disabilities and chronic illnesses, and women were disproportionately affected." These are the populations for whom power loss is most quickly life-threatening — the elderly in heat, the medically dependent without device power, the chronically ill without medication. Knowing who in your neighborhood or building is medically dependent or elderly — and having a check-in plan for extended outages — is the community action that Maria's death count makes urgent and specific.
Community resilience guideMaria lasted 11 months for the hardest-hit communities. The standard 72-hour preparedness supply is the minimum — it addresses the most common outage scenarios. In hurricane-prone areas, particularly those with long restoration histories or infrastructure fragility, a 30-day supply of food, water, medications, and cash is more appropriate as the planning baseline. This is not extreme preparedness — it is the level that Maria's documented recovery timeline suggests is rational for people who live in areas where a major hurricane could make landfall and leave the grid down for weeks or months.
Two-week preparedness guideHurricane case study series
Katrina (2005) covers levee engineering failure. Andrew (1992) covers building code enforcement. Harvey (2017) covers outdated floodplain maps. Sandy (2012) covers storm surge and evacuation compliance. Together, they document every major failure mode in the hurricane preparedness record. Read the full series or build your hurricane preparedness plan.
Full hurricane case study seriesSources