Case study

Operation Installation

Dallas, Texas

The short version

Ten years of door-to-door smoke alarm installation in 36 high-risk census tracts, with both the outcome and the cost measured and published in the peer-reviewed literature.
Type
Peer-reviewed evaluation — outcomes measured and published by researchers, not reported by the programme
Period
2001 – 2011
Last reviewed
September 4, 2026

The problem

House fires kill and injure people, and they do not do it evenly across a city. The question Operation Installation set out to answer was not whether smoke alarms work — that was settled — but whether a fire department going door to door in the neighbourhoods that carry the most risk actually changes what happens to the people living in them.

Few programmes had ever been evaluated that way. Most could report how many alarms they had handed out.

What was done

Between April 2001 and April 2011, teams of volunteers and firefighters canvassed houses in 36 high-risk target census tracts in Dallas. Where residents were at home and gave permission, a lithium-powered smoke alarm was installed.

Not handed over at a community event. Installed, in the house, by the people who knocked on the door.

What made the measurement possible

The evaluation was possible because of decisions taken at the start, not at the end.

  • The target area was defined as specific census tracts, so there was a boundary to measure inside
  • The addresses where an alarm was installed were recorded — 8,134 houses
  • Houses in the same tracts that did not receive an alarm became the comparison group — 24,346 houses
  • House fire deaths and injuries were followed for both groups, over a mean of 5.2 years

The comparison group is the part most programmes never have. Without houses in the same neighbourhoods to compare against, a fall in injuries proves nothing.

What was measured

After a mean of 5.2 years of follow-up, the rate of house fire-related deaths and injuries was 68% lower among residents of programme houses than non-programme houses — 3.1 against 9.6 per 100,000 population, a rate ratio of 0.32 with a 95% confidence interval of 0.10 to 0.84. Adjusted for several demographic variables, the rate was 63% lower.

The programme was most effective in the first five years after installation, with the difference in rates declining after the sixth year — probably, the authors say, because the alarms were becoming non-functional. Their conclusion states the duration of effectiveness as less than ten years.

A later cost-benefit analysis put the estimated net saving at $3.8 million over the decade, a return of $3.21 for every dollar spent, taking a societal perspective. Counting only direct programme and medical costs, the incremental cost per fire injury averted was $128,800 in 2013 dollars.

What to take from it

  • Installing beats distributing. The programme put alarms in, in person, with permission.
  • The comparison group was designed in from the start. It is the reason there is a number at all.
  • The effect wore off. An installation programme with no plan to return is a programme with a ten-year clock on it, and the paper is explicit about that.
  • The two figures do different jobs. $3.21 back for every dollar wins an argument about value; $128,800 per injury averted is what it costs before productivity is counted, and a finance director will ask for that one.

The honest reading of both papers together is that this worked, it paid for itself, and it needed doing again.

Sources

How these sources are labeled

Research
Academic or laboratory work.

Nothing here is paid placement. Links are chosen because they are the clearest explanation we could find, and labeled so you know who wrote them.