The week Winter Storm Uri struck Texas, Medicare mortality rose more than it did over the comparable period the following year. A new analysis estimates the size of that difference. Hospital admissions showed a different timing pattern, with a clearer increase across the four weeks from storm onset.

THE SHORT READ
  • The model estimated 2.94 excess deaths per 10,000 Texas Medicare beneficiaries during the storm week (95% CI 2.12 to 3.76).
  • The estimate for the four-week window was 3.08 excess deaths per 10,000 per week, averaged over a period that includes the storm week.
  • Power outages were measured at county level. Individual household outage histories were unavailable.
THE STUDY AT A GLANCEWinter Storm Uri and Medicare · Berlin et al.
Paper date
20 November 2026
Event studied
February 2021
Main study sample
3,955,190 Texas Medicare beneficiaries in 2021

Original publication: 20 Nov 2026 · The date above refers to this brief.

THE NUMBERS, IN CONTEXT

Estimated extra deaths per 10,000 Medicare beneficiaries

deaths per 10,000 beneficiaries per week

Storm week2.94
Weekly average over four weeks3.08
05
Point estimates from Figure 2. The four-week window includes the first week. The comparison table gives 95% confidence intervals.
THE EXPERIMENT, EXPLAINED

How the comparison estimates excess outcomes

  1. Set the window

    Choose the four pre-storm weeks and either the storm week or the four weeks from onset.

  2. Measure the change

    Compare weekly county rates within the 2021 storm year.

  3. Use the same dates in 2022

    Account for how the same counties changed over the comparison period.

Press-News diagram of the reported study design.

The comparison starts before the storm

The researchers compared weekly county rates for Medicare beneficiaries in Texas. In 2021, they measured the change from the four weeks before February 14 to the storm week. They compared that change with the change across corresponding dates in the same counties in 2022. The main samples contained about 3.96 million and 4.05 million Medicare beneficiaries, respectively. The annual samples can include some of the same people.

Source 1 ↗

Which number belongs to which window

The storm-week estimate was 2.94 extra deaths per 10,000 beneficiaries. The four-week estimate was 3.08 per 10,000 per week, averaged across February 14 to March 13. Changing the scale to 100,000 gives 30.8 per week for the longer window, using the authors’ rounded estimate. The four-week figure already includes the storm week. Both are estimates for the study population after accounting for the comparison year.

Source 1 ↗

Hospital admissions took a different course

The estimate for admissions during the storm week was −1.26 per 10,000 beneficiaries (95% CI −2.90 to 0.39), an interval that includes zero. Over four weeks from onset, the estimated increase was 2.04 per 10,000 per week (95% CI 1.30 to 2.78). The authors suggest that disrupted travel may have contributed to the timing. The study measured admissions. It did not measure how many people needed care but could not reach it.

Source 1 ↗

What the outage analysis can tell us

In the quarter of counties with the fewest outages, the storm-week mortality estimate was 0.66 per 10,000, with an interval from −2.16 to 3.48. The wide interval leaves room for several plausible effects. The researchers could compare counties with different outage patterns. Linking actual household exposure with outcomes would answer a more specific question about the role of electricity loss.

Source 1 ↗

The comparison year matters

Mortality was already declining before the selected weeks in both years. Differences in the amount or timing of seasonal illnesses, local conditions and recovery could affect the modeled result. The authors tested alternative specifications and identify choice of trend adjustment as a source of uncertainty. The findings describe this event and this Medicare population.

Source 1 ↗
TRACE THE EVIDENCE

Match the estimate to the time window

Primary journal article. Checked the county-level design, Results, Figures 2 and 3 captions and limitations. Publisher metadata and abstract also checked. Supplements, source-data files, code and clinical records were not independently audited.

01What time period does 3.08 describe?

What was observed
It is the mean weekly estimate across the four weeks from February 14.
Where the conclusion stops
The four weeks include the storm week. The rate uses 10,000 Medicare beneficiaries as its denominator.

Source 1 · Journal Results and Figure 2 caption

02Can the investigators link a death to one household outage?

What was observed
Exposure was categorized at county level.
Where the conclusion stops
The study estimates associations for groups of beneficiaries living in counties with different outage levels.

Source 1 · Journal Methods and Limitations

Numbers you can inspect

MeasureValue & unitOrigin & method
Storm-year sample3955190 beneficiariesReported Main Texas Medicare sample in 2021.
Source 1 · Journal Results and Table
Comparison-year sample4053788 beneficiariesReported Medicare beneficiaries in the same Texas counties in 2022.
Source 1 · Journal Results and Table
Storm-week excess mortality2.94 deaths per 10,000 per weekReported Adjusted estimate. 95% CI 2.12 to 3.76.
Source 1 · Journal Figure 2A
Four-week average excess mortality3.08 deaths per 10,000 per weekReported Adjusted weekly average from storm onset. 95% CI 2.72 to 3.44.
Source 1 · Journal Figure 2B
Four-week average on a 100,000-person scale30.8 deaths per 100,000 per weekCalculated Multiply the reported 3.08 per 10,000 by 10. Same population and time window.
Source 1 · Journal Figure 2B
Storm-week change in admissions-1.26 admissions per 10,000 per weekReported Adjusted estimate. 95% CI −2.90 to 0.39.
Source 1 · Journal Figure 3A
Four-week average change in admissions2.04 admissions per 10,000 per weekReported Adjusted weekly average. 95% CI 1.30 to 2.78.
Source 1 · Journal Figure 3B
Storm-week excess deaths in lowest outage quartile0.66 deaths per 10,000 per weekReported Estimate for one county group. 95% CI −2.16 to 3.48.
Source 1 · Journal Figure 2A

Compare the actual experiments

These studies answer different questions. Read the unit and endpoint before comparing results.

StudyUnit & settingReadoutInterpretation boundary
County-level mortality

Source 1 · Journal Methods and Figure 2

Texas Medicare populations in the same counties in two yearsChanges in weekly all-cause deathsThe estimate depends on observed trends and the selected control year.
County-level hospital admissions

Source 1 · Journal Figure 3 and Discussion

Texas Medicare beneficiaries over the same two windowsChanges in weekly admissionsAccess to hospital care can affect observed admission rates.
Download evidence table (CSV)

The export includes claims, available numbers, methods and source locations. It contains our reading notes and published summaries; it is not raw participant data or an independent reanalysis.

Evidence update · 21 Nov 2026
First publication. Reported observations, arithmetic and proposed follow-up questions are separated. No raw-data reanalysis. AI source check; no human editorial or clinical review.

CONNECT THE EVIDENCE

The same units, different outcomes and windows

Outcome and windowDifference per 10,000 per week (95% CI)How to read it
Deaths, storm week2.94 (2.12 to 3.76)One week, compared with the change during the corresponding dates in 2022.
Deaths, four weeks3.08 (2.72 to 3.44)Mean weekly difference across a period that includes the storm week.
Admissions, storm week−1.26 (−2.90 to 0.39)The interval includes zero.
Admissions, four weeks2.04 (1.30 to 2.78)Mean weekly difference from the start of the storm.

Estimates from journal Figures 2 and 3. All rows refer to Texas Medicare beneficiaries.

READER QUESTIONS

Your questions, answered

What do excess deaths mean in this study?

The term describes the estimated change in deaths during the event period compared with the change in the same counties a year later, after the authors’ adjustments.

Does the four-week figure give a four-week total?

The reported 3.08 is a weekly average over four weeks. Its unit is deaths per 10,000 beneficiaries per week.

What was known about power loss at individual homes?

The authors compared outage levels by county. Whether a particular beneficiary’s home lost power was unavailable.

Why is the admission result different for the first week?

The estimates cover different windows. The authors discuss access to hospitals and travel as possible explanations for the timing.

LIMITATIONS

Limits of this interpretation

  • County outage levels are an indirect measure of a beneficiary’s own electricity supply.
  • The comparison relies on corresponding weeks in one control year and adjustments for trends and measured county characteristics.
  • Results describe Medicare beneficiaries in Texas. Other residents and other storms can have different outcomes.
SOURCE NOTES

Sources & transparency

  1. Berlin, Phelan, Calianos et al. (2026): Mortality From Winter Storm Uri and Texas Power Outages Among Medicare Patients

    Primary journal article. Checked the county-level design, Results, Figures 2 and 3 captions and limitations. Publisher metadata and abstract also checked. Supplements, source-data files, code and clinical records were not independently audited. · Accessed 21 Nov 2026

    DOI: 10.1001/jamanetworkopen.2026.45376

Prepared and source-checked with AI. Press-news Team is the collective publication byline, not a medical reviewer. No human editorial or clinical review has taken place. This educational article examines published findings and how they were measured. Claims and calculations are based on the sources listed below. Photographs provide context for the topic.

Source check: AI source check — selected primary results, captions and study boundaries

Clinical review: Not applicable to this educational guide

Suggest a correction