Ohio Summit on Children Working Together for Our Future

Length-of-Stay Survival Analysis


  1. Be sure you have Adobe Reader 9 and the latest version of Adobe Flash Player. The files won't work without these programs.

    Download Adobe Reader 9
    Download Adobe Flash Player

    Both programs are free.

  2. Download to your computer the following files:

    2005 Length of Stay Survival Analysis
    2006 Length of Stay Survival Analysis
    2007 Length of Stay Survival Analysis
    2008 Length of Stay Survival Analysis

  3. Open a file.

    You will see a graph representing the length of time children, who have entered care for the first time during the identified time periods, have been in care. The horizontal axis is the number of days.  The vertical axis is the percent of children remaining in care. The gauge in the lower left corner shows the number of children represented in the data.

  4. Using the sample figure below, here's how to interpret the graph. This graph represents children who entered care for the first time in Franklin County, initially placed in a foster home, and were younger  than 3 years old when they were placed.

    2008 length of stay survival analysis

    • Point A states that 100 percent of children will stay longer than one day. Sure, all children will stay longer than one day! This is just a reference point.

    • Point B states that about 82 percent of children will stay longer than 30 days. By subtracting 82 percent from 100 percent, we can say that 18% of children left care within 30 days.

    • Point C states that 50 percent of the children will stay longer than 410 days.  So, 50 percent of the children stayed less than 410 days and 50% stayed longer than 410 days.

    • Point D is a little tricky. When the line ends at 0 percent, it means not enough time has elapsed to know how much longer the children who were still in care after 420 days will be in care. Although we don't know how long 40 percent of the children will be in care, we know that 40 percent will be in care longer than 420 days.

    • You will probably want to know how many children in this profile entered care. This number is shown on a gauge on the lower left of the graph. The gauge says 291 children.

    • Sometimes graphs are a little hard to read. If you have difficulty, drag your mouse over the graph's line, and a message will appear showing the percent of children remaining in care in the time period.You can easily change the profile. By clicking on the arrows above the graph, you can change the filters and see how long each group will stay in care. You can change the county, the initial placement type, and/or the child's age at placement.



This type of analysis is a survival analysis. It is widely used in many fields. For example, drug companies use it to predict how long it takes medicine to be effective. Tire companies, like Goodyear, use it to predict how long your car tires will last. Sony uses it to predict when your new flat screen television will malfunction.  You can bet they use it to determine your warranty! And insurance companies use survival analysis to predict how long people will live. We don't use this procedure to predict how long children will live, but we use survival analysis to predict how long children will remain in custody. Survival analysis is the leading method for measuring how long something will take when we do not know how long it will take for all people or things. Nearly all published research in our field uses this survival analysis to measure duration. It is very well respected by researchers and very powerful.



The population of children we used consists of children who have entered care for the first time. These children have never been in placement prior to this episode.  We could have included children who have entered care for the second, third, or even the fourth time, but since children returning to care have different issues than children entering for the first time, we must compare apples to apples, and analyze these multi-entry kids separately. So, if a child has had more than one episode, we are only using the first episode's data. There will be sequent analyses on multi-entry kids in the months ahead. But to make good decisions and monitor programs, we must assure we are using first entry children.



These PDFs are interactive. You make these files interactive by clicking the drop down boxes at the top of the graph. By modifying the filters (initial placement type and age group), counties should look for patterns within their county. By changing the county, you can compare how any county compares to your county. Nonetheless, we must look for patterns in the data.

Typically, patterns take two forms. The first pattern is a sharp decline in the survival line. This means children are leaving care quickly. The sharper the decline, the faster kids exit -- like a ski slope. If there is such a pattern, we suggest county leaders note when it is occurring. If it is occurring in the first 30 or 60 days of placement, we suggest counties ask themselves if they are taking kids into care that perhaps could be served with less restrictive approaches. In other words, "Why are we taking kids in care and releasing them within 30 days? Couldn't we do something different with a child in this situation?" If there are sharp declines later on, say after 180 in care, it is another opportunity to ask why. For many counties, these decreases are aligned with standard court review dates. When this occurs, we suggest these questions: "Do children only leave care based upon established court review dates; does the court discourage the agency from filing motions to terminate custody between review hearings; or does the agency not file motions to terminate custody between hearings based on their internal procedures? What are the local practices/policies and how can these be modified?" You can see how powerful this kind of analysis is.

So, the first pattern to look for is the sharp decline (ski slope, level 8 or 9). The second pattern is a flat or nearly flat line. This indicates children are not leaving care or are leaving care very slowing (ski slope, level 1 or 2). When we see this pattern, we ask "Why aren't children leaving care? What can be done?  Is this a systemic problem we have?"

While examining the ski slope levels, leaders can also examine the number of children entering each specific cohort. Clearly, the larger the number of children entering and experiencing the level 1 and 9 ski slope patterns, the greater concern. 



This is not a research question, although research can help determine the answer. The answer is dependent on a variety of issues. For instance, how much control the PCSA has over the number of children coming into care. Who controls the decision to take children into care? What is the relationship between the court and the PCSA? Is the county urban or rural? How many services are available to children? Is there a children services levy used to provide critical services? What is the belief system (data- or anecdote-based) of the decision makers? In addition, the child population, local economic conditions, and community values strongly influence the number of children entering care and how long they will stay in care.